w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4

Available online at www.sciencedirect.com

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

Interaction energy evaluation of soluble microbial products

(SMP) on different membrane surfaces: Role of the
reconstructed membrane topology

Lin Chen a,*, Yu Tian a,b,**, Chu-qing Cao c, Jun Zhang a, Zhi-neng Li a
a
School of Municipal and Environmental Engineering, Harbin Institute of Technology, Harbin 150090, China
b
State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology, Harbin 150090, China
c
School of Mechatronics engineering, Harbin Institute of Technology, Harbin 150090, China

article info abstract

Article history:
Received 8 November 2011
Received in revised form
4 January 2012
Accepted 15 February 2012
Available online 23 February 2012
Keywords:
Soluble microbial products
Membrane surface properties
Extended DLVO theory
Surface reconstruction
Interaction energy
Membrane fouling
Soluble microbial products (SMP), a majority of organic matter in effluents, play a key role
in membrane fouling. A series of filtration experiments were conducted, and demonstrated
that the flux decrement rate was in order of cellulose acetate membrane (CA, 65.4%),
polyvinylidene fluoride (PVDF, 47.9%) and polyether sulfones (PES, 29.2%). Results showed
that the fouling behavior of membrane should be predicted from the combined knowledge
of solution chemistry, surface chemical properties and surface morphology. To better
understand the interactions between the SMP and different membranes, a technique for
reconstructing the membrane surface topology was developed on the basis of statistical
parameters obtained from atomic force microscopy. The interaction energy, represented
by extended DerjaguineLandaueVerweyeOverbeek (XDLVO) potential, was calculated by
surface element integration, allowing exploring the interaction energy profiles for different
surfaces and providing considerable insights into the role of such interactions on the
macroscopic fouling behavior. The resulting interaction energy differed considerably from
the corresponding interaction between perfectly smooth surfaces. The great influence of
protrusion on the membrane surface was to reduce the primary energy barrier height, thus
rendering rough surface more favorable for deposition. An attractive energy region was
immediately surrounded by each positive asperity as demonstrated in the roughness-
engendered interaction energy maps. As the SMP approached closer to the membrane,
they had a high probability of getting trapped in the attractive energy region, leading to
a more rapid loss of flux than smooth membrane.
ª 2012 Elsevier Ltd. All rights reserved.

1. Introduction control of membrane fouling. Soluble microbial products

(SMP), a significant component of effluent organic matter,
Membrane separation technology has been widely used as an have been recognized to play an important role in membrane
alternative method in wastewater reclamation; however, the fouling and flux decline (Jarusutthirak and Amy, 2006;
effective application of membrane process requires the Rosenberger et al., 2006). SMP are known to adsorb at the

* Corresponding author. Tel./fax: þ86 451 86283077.

** Corresponding author. State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology,
Harbin 150090, China.
E-mail addresses: hitlinchen@126.com (L. Chen), hittianyu@163.com (Y. Tian).
0043-1354/$ e see front matter ª 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2012.02.030
2694 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4

solid surfaces in water, wherein the interaction between
membrane surface and solute has usually been explained by
pore blocking, ligand exchange reaction, charge interaction or
hydrophobic interaction (Maximous et al., 2009). Numerous
researchers have been concentrating on studying the gov-
erning roles of hydrophobic and charge interactions on
membrane fouling, typically represented in terms of the
DerjaguineLandaueVerweyeOverbeek (DLVO) theory (Kühnl
et al., 2010; Lee et al., 2007). More recently, Xiao et al. (2011)
applied extended DLVO (XDLVO) to describe the combined
effect of membrane and foulant (dextran, bovine serum
albumin and humic acid) hydrophobicity and surface charge
on adsorptive fouling during microfiltration. These previous
studies provided a sound starting point to understand
membrane fouling by solutes that can be treated as colloidal
particle.
However, most previous evaluation of interaction energy
was conducted assuming perfectly smooth surfaces of the
colloidal particle and the membranes. Instead, atomic force
microcopy (AFM) scans provide considerable insight into the
“roughness property” of membrane surface. Recent studies
(Elimelech, 1997; Park et al., 2005) found that the surface
topological properties of membranes had serious implica-
tions on membrane performance and fouling propensity.
Rough surfaces fouled more easily because roughness may
have an equivalent short-range effect on colloidal interac-
tions (Bhattacharjee et al., 1998) and affect surface properties
(Wong et al., 2009), such as the water contact angle, zeta
potential and surface area. Darbha et al. (2010) pointed out
that surface sections dominated by mean asperities (around
30 nm) and high surface coverage were more effective for
colloidal deposition than areas characterized by isolated
large asperities (70 nm). In light of these observations, it
seemed pertinent that simple means for accurate determi-
nation of the interaction energy between SMP and membrane,
including experiments and simulations, are critically
required along with the consideration of membrane surface
topology. In this respect, the surface element integration (SEI)
(Bhattacharjee and Elimelech, 1997) of scaling technique was
suitable to evaluate the interaction energy taking into
account the surface curvature and shape. While the results of
these studies were encouraging, the reconstructed of
membrane surface was the key issue and the role of
membrane surface properties in SMP fouling was still not well
understood.
In this study, three types of membranes were used to
describe the SMP filtration, providing experimental insight
into the fouling phenomena in terms of interaction forces
between macromolecules and the membrane surface. A
technique of reconstructing membrane topology was
proposed using information from AFM scans. Following this,
the SEI technique was used to calculate the interaction
energy between SMP and the reconstructed rough
membrane surfaces in the framework of the XDLVO theory,
shedding some light on the realistic mapping of the inter-
action energy. The critical flux was calculated for the SMP
filtration with the combination of the XDLVO potential and
the membrane topology, elucidating the important physi-
cochemical property that influenced membrane flux
behavior.
2. Experimental section
2.1. Microfiltration membranes
Three types of commercial membranes with cellulose
acetate (CA), polyvinylidene fluoride (PVDF) and polyether
sulfones (PES) were selected for the study. The membrane
characteristics are listed in Table 1 (a). PVDF membrane was
immersed in 75% (v/v) alcohol for ca. 2 h ensuring the
membranes were sufficiently wetted and degassed. Prior to
use, all membranes were soaked in deionized water for 24 h
with several intermediate water changes to remove impuri-
ties or additives.
2.2. Sampling of SMP
Activated sludge samples were obtained from a lab-scale MBR
(Tian et al., 2011). The SMP sample was separated from the
sludge mixed liquor by centrifugation (4000 rpm for 5 min) and
a successive membrane filtration (0.45 mm, CA membrane).

Table 1 e Properties of membranes and SMP.

Membrane Manufacture Zeta potentiala Contact angle ( ) Average pore Permeability
(mV) size (mm) (L/(m2 h kPa))
W/DI W/formamide W/diiodomethane

(a) Properties of membrane

CA TaoYuan 15.9 67.9 10.6 19.0 0.22 100.6
PVDF Millpore 15.5 83.1 52.8 27.0 0.22 89.4
PES Lubitech 17.4 80.9 54.6 36.8 0.22 108.7

Foulant DOC (mg/L) Zeta potential Contact angle ( ) Ionic strength (M) pH
(mV)
W/DI W/formamide W/diiodomethane

(b) Properties of foulant

SMP 6.1 17.8 37.8 23.4 41.5 0.01 6.8

pH ¼ 6.8, temperature ¼ 20 C.
a Ionic strength ¼ 10 mM NaCl.
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4 2695

The obtained SMP was then stored in the refrigerator at 4  C, The interaction energy per unit area for LW, EL and AB is
and the main characteristics of SMP are shown in Table 1 (b). estimated from Eqs. (2e4) (Oss, 2006) as the function of
The size distribution of SMP and the zeta potential trends as separation distance (h):
function of pH are clearly demonstrated in Supporting
A
123 ðhÞ ¼ 
ELW
Information (SI 1). (2)
12ph2

2.3. Fouling experiments
2 

x1 þ x23 1
123 ðhÞ ¼ εr ε0 kx1 x3
EEL ð1  cothkhÞ þ (3)
2x1 x3 sinhkh
The fouling propensity between SMP and different
membranes was conducted using a stirred dead-end cell  
h0  h
(MSC300, Mosu Corp.) operated at room temperature EAB ðhÞ ¼ DGAB
exp (4)
123 h0
l
(20  0.5  C). Before each experiment, deionized (DI) water was
filtered through the membrane prior to the fouling experiment where A ¼ 12ph20 DGLW h0 at the right-hand side of Eq. (2) is the
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
for 1 h to allow for membrane compaction and other unknown Hamaker constant; DGLW h0 ¼ 2ð rLW
2  rLW 1 Þð rLW
3  rLW 2 Þ is

cause of flux decline. After stable flux was achieved, the the free LW energy per unit area between the surface; h0 is the
membrane permeability was determined by measuring pure minimum cut-off distance due to Born repulsion; rLW is the
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
water flux over a range of applied pressure. The permeability Lifshitzevan der Waals component; 1=k ¼ ε0 εr Rg T=ð2F2 IS Þ is
was 100.6, 89.4 and 108.7 L/(m2 h kPa) for CA, PVDF and PES the Debye length; T is the absolute temperature in Kelvin; ε0 is
membrane, respectively. The filtration pressure was operated the permittivity of free space; εr is the dielectric constant of
at 3e4 kPa to produce the same initial flux for each membrane, the bulk fluid; Rg is the gas constant; F is the Faraday’s
and the stirring speed in the cell was set at 100 rpm constant; Is is the ionic strength; x1 and x3 represent
throughout the experiments. The operation was conducted at the surface potentials of membrane and foulant, respectively;
pffiffiffiffiffi pffiffiffiffiffi
initial flux of 326 L/(m2 h), and permeate flux data were l is the decay length of AB interaction; DGAB h0ffi ¼p 2 rþ ð r
pffiffiffiffiffi pffiffiffiffiffi pffiffiffiffiffi pffiffiffiffiffi
þ
pffiffiffiffiffi
þ
pffiffiffiffiffiþ
pffiffiffiffiffiffiffiffiffi
þ 
ffiffiffiffiffiffiffiffiffiþ2ffi 1

continuously logged using a top-loading electronic balance þ r3  r2 Þ þ 2 r2 ð r1 þ r3  r2 Þ  2ð r1 r3  r

(BL-1200S, Setra Systems) connected to a server computer.
The filtration experiments stopped when 300 mL permeate
was processed.
2.4. Analytical items
AFM (Veeco, Santa Barbara) was used to describe membrane
surface topography in terms of membrane surface roughness
(see SI 2). The Nanoscope control software (Version 5.30r3sr3)
was adopted for image acquisition (10 mm  10 mm areas), and
the following parameters associated with membrane
morphology were calculated from at least 7 images for each
membrane sample: average roughness (Ra), root-mean-square
roughness (Rq), maximum roughness (Rm), peak count (PC )
and surface area difference (SAD). The zeta potential (x) of
membrane was measured by the tangential streaming
potential method (Shim et al., 2002) using the electrokinetic
analyzer (SurPASS, Anton Paar). Surface tensions of SMP and
membranes were determined from contact angles (Contact
Angle meter, Tantec) of DI water, formamide and diiodo-
methane by the sessile drop on filtered lawns of SMP or on
a clean membrane.
3. Model development
3.1. Determination of surface tension energies
The XDLVO theory (Eq. (1)) describes the total interaction
energy per unit area (E ) of particle-surface in terms of Lish-
itzevan der Waals force (LW) energy, electrostatic force (EL)
energy and acid-base (AB) interaction energy.
EXDLVO
123 ¼ ELW
123 þ E123 þ E123
EL AB
(1)
where the subscripts 1e3 represent the membrane, water and
foulant, respectively.
  
1 3r Þ is
the acid-base free energy per unit area between solute and
membrane surface at contact; rþ is the electron acceptor, and
r is the electron donor component.
þ 
The parameters of rLW i ; ri ; ri (i ¼ 1, 3) can be calculated
from YoungeDupré equation (Eq. (5)) (Oss, 2006) after
measuring contact angle data (q) for three probe liquids with
þ 
known surface tension parameters ðrLW pl ; rpl ; rpl Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffi
ð1 þ cos qÞrpl ¼ 2 i rpl þ
rLW LW
rþ 
i rpl þ ri rpl
þ
(5)
3.2. Roughness simulation and surface reconstruction
In order to avoid a rigorous computation of the interaction
energy between the complexity surface and SMP, a mathe-
matical methodology for the reconstruction of the membrane
surface was developed using statistical parameters derived
from AFM analyses. For simplicity and conformity to the
interaction energy model, the simulated surface was con-
structed of a flat plate, interrupted by hemispherical protru-
sions and depressions forming the peaks and valleys. Given
the histogram of asperity (see SI 3), the membrane asperities
were assumed on the basis of a normal distribution ( f(x), x ˛
[Rm/2, Rm/2]), which chose the Ra as the mean value and the
Rq as the standard deviation. Considering the same distribu-
tion of asperity number (n) and asperity, the proportion of PC
can be calculated by integrating the area with jxj higher than
the standard deviation and then dividing the total curve area.
The value of n can be obtained by diving the AFM-measured PC
with the PC proportion. The parameters of standard deviation
and PC were adjusted, then the previous steps were repeated
to make the simulated SAD agree with the corresponding
AFM-measured SAD, and subject to the constraint that the
AFM-derived Ra and Rm were also well approximated. Finally,
the complexity of surface morphology was modeled by
random arrangement of simple elementary asperities.
2696 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4

3.3. Calculation of the sphere-rough surface interaction sphere surface; (wx, wy) is the coordinate of the spherical
energy surface under consideration, depending on the “meshgrid”
function relative to the center of sphere.
The basic governing equation of SEI considered the total Substituting Eqs. (1) and (7,8) into Eq. (6) yielded the inter-
interaction energy by integrating the interaction energy per action energy between the spherical particle and the
unit area between two surfaces over the actual surfaces membrane surface, and then shifting the horizontal location
(Bhattacharjee and Elimelech, 1997; Vrijenhoek et al., 2001). of the sphere center with respect to the rough membrane
ZZ surface produced a map of the interaction energy.
UðSÞ ¼ EXDLVO
123 ðhÞdA (6)

where U is the interaction energy between the particle and the

rough surface, S is the distance of closest approach between
4. Results and discussion
the sphere and the mean-plane of the surface and dA is the
4.1. Membrane fouling and flux decline by SMP
projected differential surface area of the particles.
To provide a narrative description of Eq. (6), the following
SMP sample derived from MBR was tested with CA, PVDF and
discussion was conducted on a model of the rough surface
PES membranes in a stirred cell with similar initial flux. The
(see SI 4), which contained a distribution of hemispherical
flux decline trends by different membranes are shown in SI 5,
protrusions and depressions of different radius interacting
and the observed differences in fouling rates are statistically
with a smooth sphere of radius ur located at a veridical
significant at 95% confidence level. The CA membrane
distance S. A Cartesian coordinate was employed to evaluate
exhibited the fastest diminishing rate of flux among the three
the surface integral with the origin located at one angle of
types of membranes. The flux of CA membrane was found to
mean-plane, thus the center of the sphere and ith asperity can
drop by 65.4% when 300 mL of permeate was collected, while
be fixed with location coordinates of (ux, uy) and (vi,x, vi,y),
the PVDF and PES membranes resulted in only ca. 47.9% and
respectively. Noticeably, the surface integral was calculated
29.2% decrement, respectively. Since the same mixed liquor
over the regions of asperities that overlapped with the pro-
was filtered and similar initial flux was applied, the rate of
jected area of sphere in the energy computation. For each
permeability decline could be much dependent on the char-
location of the sphere center, the numbers of asperity and
acteristics of membrane surface.
regions of each asperity that intersected with the projected
The hydrophilic/hydrophobic interaction between the SMP
area of the sphere were confined by the following inequality:
and membrane may be of much importance since hydro-
2
vi;x  ux þ vi;y  uy <
2
vi;r þ ur
2
(7) phobic substances demonstrated stronger interaction with
hydrophobic membrane, which was supported by previous
where jvi,rj and ur represent the radius of the ith asperity and researches (Maximous et al., 2009; Van der Bruggen et al.,
sphere, respectively. 2004). Hydrophobic interaction is the proclivity for aggrega-
With respect to the vertical distance h between a surface tion of apolar of partly apolar chains, molecules or particles
element on the sphere and the region of the membrane when immersed in water; while hydrophilic interaction is the
surface, it can be directly determined by: proclivity of strongly polar chains, molecules or particles for
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi repelling each other (Van Oss, 1995). Table 2 shows the
2
vi;r vi;r v2 calculated surface tension parameters and free energy of
h ¼ S þ ur 
vi;r cohesion for each membrane and SMP (diameter 100 nm). The
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 u2r  u2 ðri > 0 for protrusion; ri < 0 for depressionÞ (8) D50 (the 50 percentile of the particle size distribution) of
100 nm was adopted for the energy calculation, which was
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
somewhat random. Anyway, the influence of diameter on the
where v ¼ ðwx  vi;x Þ2 þ ðwy  vi;y Þ2 is the radial distance of
magnitude of LW, EL and AB interaction energies was
the asperity surface from the center of the asperity;
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi observed in the same relative magnitude, thus the compara-
u ¼ ðwx  ux Þ2 þ ðwy  uy Þ2 is the radial distance of the bility of the different interactions between the different

Table 2 e Surface tension parameters and surface free energies at the separation distance of h0 (0.157 ± 0.009 nm).
gLW gþ g gAB gTOT DGLW
121 DGAB
121 DGEL
121

(a) Surface tension parameters and the free of cohesion (mJ/m2) for each membrane and SMP
CA 48.07 3.48 2.67 6.09 54.16 10.25 43.51 0.058
PVDF 45.40 0.12 2.23 1.03 46.43 8.56 66.90 0.055
PES 41.18 0.19 3.91 1.72 42.90 6.12 56.70 0.069
SMP 38.87 1.52 34.86 14.56 53.43 4.90 13.04 0.072

CA PVDF PES

DGLW
123 DGAB
123 DGEL
123 DGLW
123 DGAB
123 DGEL
123 DGLW
123 DGAB
123 DGEL
123

(b) Surface free energy parameters (mJ/m2) for the adhesion of SMP on the membrane surface
SMP 7.09 20.63 0.065 6.48 19.11 0.063 5.47 15.57 0.071
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4
membranes was not affected by the choice of the diameter.
P
The ionic strength was determined as I ¼ 1=2 Ci Z2i (Ci is the
molar concentration of the ith ion and Zi is its charge) (Sawyer
et al., 2003), resulting in an approximate value of 0.01 M, which
was in accordance with the ionic strength given by Wang et al.
(1998). The hydrophobicity of the membrane determined the
magnitude of hydrophilic repulsion or hydrophobic attraction
that affected the interactions with SMP. All the membranes
exhibited hydrophobic characteristics (PVDF > PES > CA),
while the SMP presented hydrophilic nature. The free energy
of cohesion is the interaction free energy (per unit area) when
two surfaces of the same material are immersed in water and
brought into contact, providing a quantitative insight
regarding the hydrophobicity/hydrophilicity of the membrane
and solute (Brant and Childress, 2002). The free energy of PES
membrane (20.97 mJ/m2) was lower than that of PVDF
(25.53 mJ/m2) and CA (27.66 mJ/m2) membranes with
respect to SMP in an aquatic environment, suggesting PES has
the lowest SMP fouling tendency in terms of adsorption onto
the membrane surface. Significant difference in fouling rate
was detected between CA and PVDF membranes regardless of
the similar free energy. In terms of Zeta potentials analysis,
the membrane with a more negative zeta potential would
exhibit higher electrostatic double layer repulsion according
to the classical DLVO theory (Vrijenhoek et al., 2001), and thus
be more fouling resistant to negatively charged solutes.
However, the experimental fouling data consistently showed
that the CA fouled more severely than PVDF and PES
membranes. It seemed that the characteristics of hydropho-
bicity and Zeta potential may not be sufficient to predict the
effect of membrane property on membrane fouling.
The actual geometry of the membrane morphology was
another important aspect to be considered. Surface roughness
produced tangential colloidal forces which can immobilize
colloidal particles on the membrane surface (Carnie et al.,
2005). Large-scale surface roughness, of the same order of
magnitude as the SMP interacting with the surface, signifi-
cantly increased the rate of SMP attachment through
providing a larger surface area and greater contact opportu-
nities for SMP with the membrane surface. Additionally, as
the permeation rate was proportional to the thickness of the
active (skin) layer, the bottom of a “valley” presented the
“path-of-least resistance” to permeating water (Vrijenhoek
et al., 2001). These effects of surface roughness resulted in
enhanced attachment of SMP onto the membrane surface,
and hence, more severe fouling. As shown in SI 2, both of PVDF
and PES exhibited relatively fewer but larger protuberances,
compared to the CA with abundant but small asperities. Also,
the roughness feature of CA had much smaller peak-to-peak
separation distance than that of PVDF and PES. Hence, for
a rough membrane (CA), SMP were preferentially transported
onto the membrane surface. The surface quickly became
clogged with multiple layers of densely packed foulants,
leading to an increasing flow resistance and a more rapid flux
reduction.
Regardless, it can be concluded that the relative fouling
behavior of the three commercial membranes should be
qualitatively predicted from the combined knowledge of
physical surface morphology, surface chemical properties and
solution chemistry. Given the irregular morphology of the
2697
membrane surface, it was really hard to calculate the total
interaction between SMP and the actual membrane surface. In
the following section, a mathematical methodology for the
reconstruction of membrane surface was developed, and SEI
technique was used to evaluate the extended DLVO interac-
tions between SMP and the reconstructed rough membranes.
4.2. Interaction energy barrier between SMP and
membrane surfaces
Fig. 1 demonstrates the reconstructed topologies of the three
types of membranes, and Table 3 lists the statistical rough-
ness parameters obtained from the AFM roughness analysis
and the model rough surfaces. It clearly showed minor devi-
ations (<1%) from the measured SAD, Ra and Rm values but
noticeable deviations from Rq (<25%) and PC, indicating a good
match between the statistical roughness parameters of the
model surfaces and those of the real membranes. The
subsequent calculations of the XDLVO interaction energies
between a spherical particle and the three types of rough
membranes were performed based on these simulated
topologies.
Fig. 2a illustrates the dependence of the interaction energy
between SMP and different smooth membranes on the sepa-
ration distance. Noticeably, both LW energy and AB energy
were attractive and played important roles at short-range
scale (<5 nm), while EL energy was repulsive and long-ranged
(>5 nm) which can be attributed to the negatively charged of
surfaces and a thick electrical double layer. At even larger
separation, the XDLVO interaction became attractive, and
a secondary minimum existed at a separation distance of
15 nm. Additionally, the interaction energy decreased from
the primary energy barrier to zero over a distance of about
12 nm for all surfaces, which was also true even a positive
asperity prevented SMP from attaching to the mean-plane of
the membrane surface (as shown in Fig. 2b). The combination
of these three energies determined the SMP-membrane
interactions as the SMP approaching to the mean-plane. At
smooth surface, the primary energy barrier was existed at
a separation distance of 3e4 nm from the mean-plane, and
the value of primary energy barrier was in the order of 8.36,
8.75, 12.64 kT for CA, PVDF and PES membranes, respectively.
Clearly, SMP was subject to the greatest repulsive interaction
with PES membrane, which may inhibit the initial deposition
of the foulant. CA and PVDF membranes exhibited a similar
primary energy barrier but a greatly different fouling rate,
which might be explained by the interaction profiles between
SMP and the rough membrane surface as provided in
Fig. 2(bec) and SI 6.
In one extreme, the interaction between SMP and a single,
positive asperity produced an interaction energy which was
much smaller in magnitude than the calculated sphereeplate
interaction energy profile (Fig. 2b). The respective values of
primary energy barrier reduced to 4.57, 3.79 and 7.62 kT for
positive CA, PVDF and PES asperities, compared with
8.32e12.64 kT for the corresponding planar surface. The
existence of asperities counteracted the repulsive interaction
energy due to the zeta potential; concomitantly, the large
positive asperity physically prevented the particle from
attaining a separation distance of less than 81 nm (CA), 52 nm
2698 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4

Fig. 1 e Representative image of simulated roughness membrane surface: (a) CA; (b) PVDF and (c) PES.

(PVDF) and 98 nm (PES) from the mean-plane. In another
extreme, the primary energy barrier reached up to 30%e96%
larger than the sphereeplate interaction energy profile if SMP
penetrated into a single, negative asperity (Fig. 2c). The
occurrence of the enhanced interaction energy might be
attributed to the fact that the effective separation distance
between sphere and negative asperity surface was nearly
uniform and thus much smaller than the separation distance
between sphere and smooth surface. Since the number of
positive asperity was much higher than that of negative
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4 2699

Table 3 e Measured and simulated membrane statistical roughness parameters (Hemispheres).

Morphological parameter unit CA PVDF PES
a b
AFM SIM AFM SIM AFM SIM
2
Membrane area mm 100 100 100 100 100 100
Average roughness nm 104.2 104.2 262.3 262.3 178.6 178.6
RMS roughness nm 160.4 200.4 338.1 417.5 244.1 302.0
Maximum roughness mm 1.081 1.082 2.715 2.716 1.623 1.624
Number of asperities no. N/Ac 347 N/A 85 N/A 48
Peak count no. 140 140 28 39 13 22
Surface area difference % 45.96 44.75 75.92 76.11 17.95 17.86

a Surface area scanned and analyzed by AFM.

b Surface area reconstructed using hemispherical asperities.
c The number of asperities in AFM surface scans can’t be determined.

asperity and the SMP radius was larger than most of the
negative asperity sizes, it was reasonably assumed that the
number of interactions falling into the negative asperity was
small.
Additionally, Fig. 3 evaluates the influence of asperity
radius on the energy barrier as a single asperity is placed right
under the incoming SMP particle. Similar results can be
observed for these three membranes. It clearly showed that,
for the median radius of SMP (50 nm), a critical asperity radius
(around 14 nm) resulted in the highest reduction of the energy
barrier. For the larger asperity radius, the energy barrier
increased since the XDLVO interaction was approximately
proportional to vir $ur =ðvir þ ur Þ. With respect to the smaller
asperity radius, the asperity size was not large enough to
lower the SMP-planar interaction, which accounted for the
largest part of the energy barrier. Therefore, the complete

Fig. 2 e SEI model predictions of interaction energy profiles for the SMP (radius [ 50 nm) and different simulated membrane
surfaces: (a) smooth surface, (b) positive asperity and (c) negative asperity.
2700 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4
Fig. 3 e Primary energy barrier between SMP and rough
surface with varying asperity radius.
understanding of the roughness distribution was required to
determine its impact on the energy barrier.
The simulations described above were then repeated for 50
additional randomly selected (x, y) locations, and the primary
energy barrier for a given membrane was averaged. For all the
surfaces, the presence of roughness features reduced the
interaction energy barrier compared to that of smooth
surface, wherein the averaged primary energy barriers were
4.35, 6.38 and 9.14 kT for CA, PVDF and PES membranes,
respectively. The practical implication for these results was
that the rough property of membrane surface would tend to
promote fouling or, at least, reduce fouling resistance-
regardless of the chemical properties of a membrane mate-
rial. To provide a thorough understanding of the interaction
between SMP and reconstructed membrane surface, interac-
tion energy map was performed at a series of separation
distances and over a range of (x, y) locations.
4.3. Interaction energy map between SMP and different
membrane surfaces
The calculation and drawing of the interaction energy map at
a fixed separation distance were conducted by varying the
horizontal coordinate (x, y) of the sphere center in a rasterized
pattern (the distance between the sphere center ¼ 50 nm). In
this study, five different separation distances, including 1, 3.5,
10, 25 and 75 nm from the mean-plane, were performed for
the three simulated membrane surface, as shown in Fig. 4.
Two different superimposed color maps with “jet” style were
adopted to show the magnitude of the interaction energy, one
for the 25e75 nm separation distance and another for the
1e10 nm separation distance. Areas filled with red, lavender
and blue represented the attractive, neutral and repulsive
interaction energy, respectively. The overlap between the SMP
and membrane asperity can’t be avoided as the SMP
approaching to the mean-plane. Given the physically impos-
sibility for SMP to occupy the membrane asperities, such an
overlapped situation was simply handled by avoiding any
energy computation, and the overlapped area was depicted in
gray. Noticeably, the XDLVO interaction energy between the
SMP and a smooth membrane surface was significantly
altered by membrane surface topology (roughness), and the
most prominent feature was that the overlapped regions were
surrounded by navy-color contours.
Fig. 4 (a1-a5) illustrates the interaction energy maps for
a simulated CA membrane at different separation distances
ranging from 75 nm down to 1 nm. The simulated CA surface
was composed of abundant asperities, and the radius of some
asperities was large compared with the median radius of SMP
(50 nm). At a large separation distance of 75 nm, the entire
membrane surface presented extremely weak attractive
interaction energy (0.16 kT). Simultaneously, some inter-
esting phenomena can be observed from the careful inspec-
tion of figure. The SMP firstly felt an elevated repulsive energy
such as at the location around (1500, 1200) (Fig. 4 (a-1)), and
then these locations were mapped in gray due to some
physical overlap of the SMP with the surface protrusion as
shown in Fig. 4 (a2-a5). Another interesting point was that the
gray regions were surrounded by blue contours, suggesting
the interaction energy in these contours were highly attractive
since the SMP colloid nearly contacted with the positive
asperity. Some repulsive regions were existed around the
attractive contours, which can be explained by the XDLVO
theory that the primary minimum immediately follows the
primary energy barrier. Additionally, it should be mentioned
here that some repulsive regions around the attractive
contours can’t be reflected during simulation due to the
adopted 50 nm of sphere centerecenter distance during
simulation and the short-range (<5 nm) between the primary
minimum and the primary energy barrier.
The membrane surface demonstrated an increased
attractive energy (0.86 kT) as the SMP approached to 25 nm
above the mean-plane, while the surface presented an aver-
aged repulsive energy (0.55 kT) as the SMP was further brought
to 10 nm from the surface. These phenomena clearly sug-
gested that as the SMP shifted from 75 nm to 10 nm from the
membrane surface, the SMP with enough convective energy
pushed through the secondary minimum and then experi-
enced electrostatic repulsion from the surfaces in closest
proximity. There were also some highly attractive contours
surrounded by the physical overlap of the SMP with the
positive asperities, thus the SMP might be easily trapped into
the energy minimum regions.
The CA membrane exhibited strong repulsive interaction
energy (6.73 kT) as the SMP arrived at a separation distance of
3.5 nm (Fig. 4 (a-4)). Regarding smooth membrane, no
membrane fouling could be observed theoretically if the
interaction energy barrier was higher than the hydraulic drag
energy. However, the interaction energy in some isolated
regions was found to be negligible (mapped in lavender) due to
the existence of negative asperity on the surface; simulta-
neously, the attractive regions around the positive asperity
were obviously increased due to the increment of overlapped
surface. The existence of these asperities not only reduced the
interaction energy barrier but also intensified the SMP depo-
sition, leading to a more rapid flux decrement and severer
membrane fouling.
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4 2701

Fig. 4 e SEI model predictions of interaction energy map for the SMP (radius [ 50 nm) and three different simulated
membranes: (a) CA, (b) PVDF and (c) PES at a series of separation distances. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)
2702 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4
Finally, as SMP was brought to 1 nm from the mean-plane
(Fig. 4 (a-5)), the entire membrane surface presented a fairly
uniform attractive energy (381.29 kT) for the SMP, inter-
spersed with an increased amount of overlapped regions and
a few of valleys. The five energy maps for CA membrane
clearly demonstrated the general trends of interaction energy
conforming to the XDLVO theory, and also showed the great
changes of interaction energy due to the existence of asper-
ities. As the SMP closed to the rough surface, it may simply be
drawn toward the regions with lower interaction energy.
Under the condition of cross-flow membrane filtration, SMP
deposited on the pit wall of rough membranes might be pro-
tected from cross-flow shear due to the large positive asper-
ities and the small peak-to-peak separation distance, and thus
initial SMP fouling may be enhanced.
Fig. 4 (b1-b5) illustrates the corresponding interaction
energy maps for the simulated PVDF membrane. As the SMP
was brought from 75 nm to 10 nm from the mean-plane, the
interaction energy exhibited the general trend of neutrally
zero (D ¼ 75 nm, 0.15 kT), slight attraction (D ¼ 25 nm,
0.78 kT), slight repulsion (D ¼ 10 nm, 0.17 kT), elevated
repulsion (D ¼ 3.5 nm, 7.41 kT) and strong attraction
(D ¼ 1 nm, 350.75 kT). There were also some overlaps of the
SMP with the positive asperities on the PVDF surface as
indicated by the gray-mapped areas. However, the most
striking differences from the CA membrane were the 100
lower of peak count and 31% higher of surface area differ-
ence. The number of asperities on the PVDF membrane was
obviously decreased, resulting in a reduced overlapped areas
and local attractive contours. On the other hand, since the
increased surface area difference was companied by signifi-
cant increment of asperity size, the primary energy barrier
was only subtly decreased on the basis of asperity radius-
energy barrier analysis (see Fig. 3). As the SMP approached
closer to the membrane surface, its lateral movement was
limited by the presence of the peaks, and thus, the SMP
would attach to the sidewalls of the asperities. Compared
with the CA membrane, the PVDF membrane exhibited the
larger peak-to-peak separation distance, less energy barrier
reduction and lower attractive contours around the protru-
sions, indicating the decreased fouling rate of PVDF
membrane from another perspective.
The interaction energy maps for the simulated PES
membrane at different separation distances from 75 nm to
1 nm are depicted in Fig. 4 (c1-c5), exhibiting a similar
tendency as the simulated PVDF membrane. However, some
noticeable differences can still be found between PVDF and
PES membranes, attributing to the surface topology and
physicochemical property. Compared with other two
membranes, the PES membrane was much smoother with the
least asperities, leading to relatively smaller attractive regions
around the positive asperities. As the SMP approaching to the
mean-plane, the probability of SMP trapped into the lower-
energy valleys significantly decreased. Additionally, the PES
membrane showed the higher primary energy barrier than
PVDF due to its own surface properties, becoming much
fouling resistant. As clearly be presented in Fig. 4 (c4), the
interaction energy reached up to 12.43 kT for PES membrane
at the separation distance of 3.5 nm, which was 5.70 kT and
5.02 kT higher than that for CA and PVDF membranes,
respectively. Hence, under the same hydraulic condition, the
PES exhibited the lowest fouling potential, with 18.7% lower of
flux reduction compared with PVDF membrane.
As stated above, the energy barrier was significantly
reduced by the presence of asperity. Since the SMP deposi-
tion was highly dependent on the height of the energy
barrier, the SMP fouling would occur mainly near asperities
where the energy barrier was low. Furthermore, the
membrane fouling would be enhanced by surface roughness,
provided that the asperity and peak-to-peak separation
distance were not too large to decrease the primary energy
barrier and increase the contact probability between the SMP
and membrane surface.
4.4. Implication for membrane process
Generally, the issue of modeling colloidal deposition was
addressed as a two-step process (Adomeit and Renz, 1996):
first colloid was transported by the motions of the flow toward
surface, and second, in the immediate vicinity of the surface,
the forces between the incoming colloid and the surface
were determined with the XDLVO theory. Critical flux defined
as a condition that the hydrodynamic drag force trans-
porting colloids from the bulk to the membrane surface was
roughly balanced by repulsive interaction forces (Lee and
Elimelech, 2006). Accurate assessment and characterization
of membrane-SMP interactions may allow for optimization of
repulsive membrane-SMP interactions to operate under
appropriate operating conditions, resulting in reduced
membrane fouling.
Using the measured SMP and membrane properties, it was
possible to determine the maximum critical flux (vc) for
a membrane process through the modified Bowen and Sharif
method (Richard Bowen and Sharif, 1998). A correction factor
(Urough/Usmooth) was introduced to take into account the
surface roughness as given in Eq. (9), serving as a control
strategy for membrane fouling.
FLW þ FEL þ FAB Urough
jvc j ¼ $ (9)
ð6pmur ÞfH Usmooth
where FLW is the van der Waals force, FEL is the electrostatic
repulsive force, FAB is the acid-base force, m is the solution
viscosity, fH is the hydrodynamic correction factor (Wang
et al., 2005), Urough and Usmooth are the total interaction
energy of the simulated rough surface and smooth surface,
respectively. Note that FLW, FEL, FAB, Urough and Usmooth must be
obtained at the same separation distance between the sphere
and the mean-plane.
Fig. 5 demonstrates the relationship of critical flux against
the particle-membrane separation distance for the SMP and
different membranes, wherein Fig. 5(a) and Fig. 5(b) represent
the critical flux for the smooth surface and rough surface,
respectively. There was a maximum in the critical flux curves
as a function of separation distance. If a flux less than this
maximum critical velocity was applied, then the hydrody-
namic force on the SMP would be balanced by the repulsive
electrostatic force before the SMP contacted with the
membrane surface. It can be seen from Fig. 5(a) that the
distance at which this occurred was 3e9 nm deviations from
the mean-plane, and slight difference of the critical flux
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4
Fig. 5 e Critical velocity as a function of separation distance for different membranes: (a) smooth membrane surface and (b)
rough membrane surface.
curves was observed between the CA and PVDF membranes.
However, the magnitude of critical flux was obviously
decreased with the incorporation of roughness effect
(Fig. 5(b)). At a distance of 4 nm, approximate 11.0% and 6.1%
reduction of critical flux were obtained for PVDF and PES
membranes, respectively. With respect to the CA membrane,
the curve shape was significantly changed with more than
70% decrement of critical flux, and the controllable balance
region was existed in a separation of 6e8 nm from the mean-
plane. Great differences of the critical flux were found
between the PVDF and PES membranes from the aspects of
flux magnitude, tendency and curve shape. These observa-
tions could explain the results of fouling experiments to some
extent: the CA membrane operated above critical flux resulted
in sever membrane fouling while slight fouling occurred for
the PES membrane operating below the critical flux.
Regardless of the present results, further detailed investi-
gation should be performed. Firstly, additional experimental
support was required. The experimental data regarding the
deposition of SMP were rare, and it would be much worth
considering a more detailed experimental approach (e.g.,
a roughness controlled surface with protruding spherical
asperities (Chen et al., 2009)). Then, the numerical deposition
was much dependent on the particle size: larger particle had
a higher tendency to attach to the rough surface. The intro-
duction of surface roughness has been shown to reproduce
nonzero deposition rate, which was in agreement with the
experimental observation (Cerovic et al., 2009). The determi-
nation of critical flux might be investigated in future calcula-
tion within an acceptable fouling rate.
5. Conclusions
Compared with PVDF and PES membranes, the CA membrane
exhibited the fastest fouling potential during the SMP filtra-
tion, which can be qualitatively predicted from the combined
knowledge of physical surface morphology, surface chemical
properties and solution chemistry. Simulations were con-
ducted to evaluate the interfacial interactions between
SMP and reconstructed membrane surfaces via an XDLVO
potential. The resulting interaction energy was changed
2703
considerably by membrane surface topology, particularly near
the repulsive interaction energy barrier. For the SMP inter-
acting with a flat plate, the interaction energy was substan-
tially reduced in the presence of positive asperity; whereas the
primary energy barrier was much larger if SMP penetrated
into a single, negative asperity. The magnitude of XDLVO
potential was on average decreased by surface roughness,
since the number of positive asperity was much higher
than that of negative asperity and the SMP radius was larger
than most of the negative asperity sizes. The roughness-
engendered interaction energy map clearly showed that an
attractive energy contour was immediately surrounded by
each positive asperity, suggesting that SMP had a high prob-
ability of getting trapped in the attractive energy regions as
approaching to the membrane surface. Given the surface
morphology, a more rapid loss of flux and sever membrane
fouling would be observed in the CA membrane filtration of
SMP solution. Additionally, the accurate assessment of
membrane-SMP interactions allowed identification of critical
flux for effective membrane process operation, resulting in
more than 70%, 11% and 6% decrement of critical flux for CA,
PVDF and PES membranes with the incorporation of rough-
ness effect.
Acknowledgments
This study was supported by the National High-tech R&D
Program (863 Program) of China (No. 2009AA064704), the
National Natural Science Fund of China (No. 50978071) and the
State Key Laboratory of Urban Water Resource and Environ-
ment, Harbin Institute of Technology (No. 2011DX01).
The authors also appreciate the National Innovation Team
supported by the National Science Foundation of China
(No. 50821002).
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2012.02.030.
2704 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 9 3 e2 7 0 4
references
Adomeit, P., Renz, U., 1996. Deposition of fine particles from
a turbulent liquid flow: experiments and numerical
predictions. Chemical Engineering Science 51, 3491e3503.
Bhattacharjee, S., Elimelech, M., 1997. Surface element
integration: a novel technique for evaluation of DLVO
interaction between a particle and a flat plate. Journal of
Colloid and Interface Science 193, 273e285.
Bhattacharjee, S., Ko, C.H., Elimelech, M., 1998. DLVO interaction
between rough surfaces. Langmuir 14, 3365e3375.
Brant, J.A., Childress, A.E., 2002. Assessing short-range
membrane-colloid interactions using surface energetics.
Journal of Membrane Science 203, 257e273.
Carnie, S.L., Chan, D.Y.C., Lewis, C., Manica, R., Dagastine, R.R.,
2005. Measurement of dynamical forces between deformable
drops using the atomic force microscope. I. Theory. Langmuir
21, 2912e2922.
Cerovic, L., Lefèvre, G., Jaubertie, A., Fédoroff, M., Milonjic, S.,
2009. Deposition of hematite particles on polypropylene walls
in dynamic conditions. Journal of Colloid and Interface
Science 330, 284e291.
Chen, G., Beving, D.E., Bedi, R.S., Yan, Y.S., Walker, S.L., 2009.
Initial bacterial deposition on bare and zeolite-coated
aluminum alloy and stainless steel. Langmuir 25, 1620e1626.
Darbha, G.K., Schäfer, T., Heberling, F., Lüttge, A., Fischer, C.,
2010. Retention of latex colloids on calcite as a function of
surface roughness and topography. Langmuir 26, 4743e4752.
Elimelech, M., 1997. Role of membrane surface morphology in
colloidal fouling of cellulose acetate and composite aromatic
polyamide reverse osmosis membranes. Journal of Membrane
Science 127, 101e109.
Jarusutthirak, C., Amy, G., 2006. Role of soluble microbial
products (SMP) in membrane fouling and flux decline.
Environmental Science and Technology 40, 969e974.
Kühnl, W., Piry, A., Kaufmann, V., Grein, T., Ripperger, S.,
Kulozik, U., 2010. Impact of colloidal interactions on the flux in
cross-flow microfiltration of milk at different pH values:
a surface energy approach. Journal of Membrane Science 352,
107e115.
Lee, S., Elimelech, M., 2006. Relating organic fouling of reverse
osmosis membranes to intermolecular adhesion forces.
Environmental Science and Technology 40, 980e987.
Lee, S., Kim, S., Cho, J., Hoek, E.M.V., 2007. Natural organic matter
fouling due to foulant-membrane physicochemical
interactions. Desalination 202, 377e384.
Maximous, N., Nakhla, G., Wan, W., 2009. Comparative
assessment of hydrophobic and hydrophilic membrane
fouling in wastewater applications. Journal of Membrane
Science 339, 93e99.
Oss, C.J.v., 2006. Interfacial Forces in Aqueous Media, second ed.
Taylor & Francis, Boca Raton.
Park, N., Kwon, B., Kim, I.S., Cho, J., 2005. Biofouling potential of
various NF membranes with respect to bacteria and their
soluble microbial products (SMP): characterizations, flux
decline, and transport parameters. Journal of Membrane
Science 258, 43e54.
Richard Bowen, W., Sharif, A.O., 1998. Hydrodynamic and
colloidal interactions effects on the rejection of
a particle larger than a pore in microfiltration and
ultrafiltration membranes. Chemical Engineering Science
53, 879e890.
Rosenberger, S., Laabs, C., Lesjean, B., Gnirss, R., Amy, G.,
Jekel, M., Schrotter, J.C., 2006. Impact of colloidal and soluble
organic material on membrane performance in membrane
bioreactors for municipal wastewater treatment. Water
Research 40, 710e720.
Sawyer, C.N., McCarty, P.L., Parkin, G.F., 2003. Chemistry for
Environmental Engineering and Science. McGraw-Hill, New
York.
Shim, Y., Lee, H.J., Lee, S., Moon, S.H., Cho, J., 2002. Effects of
natural organic matter and ionic species on membrane
surface charge. Environmental Science and Technology 36,
3864e3871.
Tian, Y., Chen, L., Jiang, T., 2011. Characterization and
modeling of the soluble microbial products in membrane
bioreactor. Separation and Purification Technology 76,
316e324.
Van der Bruggen, B., Kim, J.H., DiGiano, F.A., Geens, J.,
Vandecasteele, C., 2004. Influence of MF pretreatment on NF
performance for aqueous solutions containing particles and
an organic foulant. Separation and Purification Technology 36,
203e213.
Van Oss, C., 1995. Hydrophobic, hydrophilic and other
interactions in epitopeeparatope binding. Molecular
Immunology 32, 199e211.
Vrijenhoek, E.M., Hong, S., Elimelech, M., 2001. Influence of
membrane surface properties on initial rate of colloidal
fouling of reverse osmosis and nanofiltration membranes.
Journal of Membrane Science 188, 115e128.
Wang, J., Huang, C., Allen, H.E., Takiyama, L.R., Poesponegoro, I.,
Poesponegoro, H., Pirestani, D., 1998. Acid characteristics of
dissolved organic matter in wastewater. Water Environment
Research 70, 1041e1048.
Wang, S., Guillen, G., Hoek, E.M.V., 2005. Direct observation of
microbial adhesion to membranes. Environmental Science
and Technology 39, 6461e6469.
Wong, P.C.Y., Kwon, Y.-N., Criddle, C.S., 2009. Use of atomic force
microscopy and fractal geometry to characterize the
roughness of nano-, micro-, and ultrafiltration membranes.
Journal of Membrane Science 340, 117e132.
Xiao, K., Wang, X., Huang, X., Waite, T.D., Wen, V., 2011.
Combined effect of membrane and foulant hydrophobicity
and surface charge on adsorptive fouling during
microfiltration. Journal of Membrane Science 373,
140e151.