Chemical Engineering Science 163 (2017) 155–166

Contents lists available at ScienceDirect

Chemical Engineering Science

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

The effect of particle surface charge density on filter cake properties

during dead-end filtration
Søren Lorenzen ⇑, Kristian Keiding, Morten L. Christensen
Department of Chemistry and Bioscience, Aalborg University, Fredrik Bajers Vej 7H, 9220 Aalborg East, Denmark

h i g h l i g h t s

Model particles with variable surface charge were used in dead-end filtration.

Specific resistance and compressibility increase with particle charge.

Porosity increase with particle charge and decrease with applied pressure.

Existing models fail to predict specific cake resistance for charged particles.

Osmotic pressure in the filter cake proposed as a possible explanation.

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

Article history: During the filtration of suspended organic material, filter cakes with high specific resistances are often
Received 24 October 2016 formed. The properties of these filter cakes differ markedly from the classical filtration properties derived
Received in revised form 4 January 2017 from inorganic suspensions, not only by having a higher specific resistance but also by being highly com-
Accepted 19 January 2017
pressible. These organic substances typically have a high surface charge density, due to the different func-
Available online 24 January 2017
tional groups on their surfaces, so a set of model particles was synthesized in which the surface charge
density can be varied while keeping all other properties constant. The synthesized monodisperse parti-
Keywords:
cles all have a polystyrene core and varying ratios of charged (i.e. polyacrylic acid) and uncharged (i.e.
Particle charge
Filter cake
hydroxypropyl cellulose) stabilizing polymers covalently grafted to the surface. All used particles have
Charge effects proven to be stable against aggregation during storage and are easily redispersed upon agitation.
Specific cake resistance Aqueous suspensions of the particles were filtered in a dead-end setup at constant pressure to determine
Osmotic pressure the filter cake properties. Results indicate that the specific cake resistance, cake compressibility, and cake
porosity increase with the particle surface charge. Comparison with two empirical models of filter cake
resistance demonstrates that particle charge is more important than is porosity, as highly charged filter
cakes have resistances orders of magnitude higher than calculated, which is not the case for filter cakes of
uncharged particles. DLVO theory suggests that the results cannot be explained in terms of classical
double-layer effects, even in the extreme case of very small interparticle distances. Deviations from this
classical behavior, arising from the confinement of charged particles by charged walls or pores, are a pos-
sible explanation, allowing the high resistances to be interpreted as increased osmotic pressure due to
the distribution of counterions in the pores. Calculations indicate that osmotic pressure increases with
particle charge and with decreasing porosity, but remains close to zero for uncharged particles.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction
Dewatering filtration is a common unit operation whenever
suspended solids must be separated from a liquid. In almost all
applications, the properties of the filter cake deposited on the filter
medium govern the process. The liquid flow through the porous fil-
⇑ Corresponding author.
E-mail address: slo@bio.aau.dk (S. Lorenzen).
ter cake will thus be a limiting factor for the filtration, and efforts
to reduce the hydraulic resistance account for a significant part of
the operational costs of a filtration unit. The liquid flux through
porous media (e.g. filter cakes) is often interpreted using Darcy’s
law, in which the flux is expressed as the pressure drop across
the cake divided by the hydraulic resistance. Properly dimension-
ing the equipment used for dewatering calls for knowledge of
the expected properties of the filter cake formed. For this purpose,
semi-empirical models exist for predicting the hydraulic resistance

http://dx.doi.org/10.1016/j.ces.2017.01.044
0009-2509/Ó 2017 Elsevier Ltd. All rights reserved.
156 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
from the cake thickness, cake porosity, liquid viscosity, and solid–
liquid contact area in the cake (which can be calculated from par-
ticle size and shape). One often-used model is the Kozeny–Carman
model (Eq. (1)), giving the specific resistance, a, i.e. hydraulic resis-
tance per cake mass deposited. In it, k is the Kozeny constant (with
a value of 5 ± 0.5 for spherical particles; Carman, 1937), S0 the
volume-specific surface of a particle, qs the solid density, and e
the filter cake porosity. This model also serves as the basis of some
other models (Xu and Yu, 2008).
S20  ð1  eÞ
aKC ¼ k ð1Þ
qs  e3
Although this and some derived models are extensively used,
they have several limitations. Indeed, the Kozeny–Carman model
has been found to give satisfactory predictions of the filtration of
simple inorganic suspensions or suspensions of uncharged latex
particles (Ogden and Davis, 1990). However, the model is very poor
at dealing with suspensions of biological origin, such as microbial
cell suspensions (Foley, 2006) or activated sludge (Mikkelsen and
Keiding, 2002), and application of the model is complicated by
the difficulty of determining S0 and e, especially for activated
sludge.
Previous work on the filtration of wastewater sludge has found
very high specific filter cake resistances, typically several hundred
times higher than would be expected from the Kozeny–Carman
model, i.e. 10–100 Tm kg1 (Bugge et al., 2012; Chu and Li, 2005;
Mikkelsen and Keiding, 2002; Wu et al., 2001). Among the sug-
gested explanations of this difference are (1) blinding effects
within the cake itself, which occur when the filter cake acts as a
secondary membrane, capturing smaller particles within the pores
(Sørensen et al., 1997; Novak et al., 1988), and (2) skin formation,
leading to reduced filter cake porosity, especially in layers near the
membrane where the solid stress is the highest, caused by the high
degree of compressibility typical of wastewater sludge (Tiller and
Kwon, 1998). Both these effects are possible in systems in which
the suspended solids have a broad range of characteristics, further
complicating the prediction and modeling of cake resistance. As
sludge is a very complex matrix (Le-Clech et al., 2006), it is difficult
to isolate and vary only one property of a sludge floc without
affecting several others, all of which can affect filterability.
To overcome this, model particles can be used to simulate real-
world suspensions, with the added benefit of ruling out effects
caused by blinding and skin formation. A key factor in sludge filtra-
tion is thought to be related to the primarily negatively charged
groups on the surfaces of the sludge flocs, single cells, and exopoly-
mers, in line with the findings of Hodgson et al. (1993) and Ohmori
and Iritani (Ohmori and Glatz, 1999; Ohmori and Iritani, 2004). It
has been found that reducing the surface density of these charges
positively affects the filterability (Mikkelsen and Keiding, 2002).
This positive effect is thought to be caused mainly by increased
flocculation (Bugge et al., 2013), as the low surface charge cannot
provide the interparticle repulsion necessary to keep the sludge
flocs apart, while a more direct influence of particle charge also
have been proven to be of importance (Lorenzen et al., 2014). In
contrast, Mukai et al. found that increasing protein surface charge
would decrease cake resistance during ultrafiltration, due to a
higher porosity resulting from increased electrostatic repulsion
(Mukai et al., 1997). Thus, conflicting results from different sys-
tems have been reported, which only serves to increase the impor-
tance of further research in this area.
Some researchers have also discussed a more direct influence of
charge, exerted mainly through increased osmotic pressure in the
filter cake caused by the colligative properties of the counterions
associated with the charged groups. This osmotic pressure coun-
teracts the applied pressure and is therefore perceived as increased
resistance (Keiding and Rasmussen, 2003; Wijmans et al., 1984).
This effect has only recently been quantified in MBR systems
(Chen et al., 2012; Curvers et al., 2009; Zhang et al., 2013, 2014),
while earlier work has found osmotic pressure to be negligible in
cross-flow microfiltration (Davis, 1992). The complexity and diver-
sity of sludge, however, will make it difficult to compare results
and might even lead to contradictory conclusions, as the effects
of charge are impossible to isolate due to the adverse secondary
effects occurring when attempting to increase or decrease the
charge density, such as changes in floc strength and particle size
distribution.
Although several attempts have been made to modify the
Kozeny–Carman equation (Valdes-Parada et al., 2009; Xu and Yu,
2008), none of these models accounts for variations in particle sur-
face charge and, to the authors’ best knowledge, no such model
exists. As shown, charged surfaces have been found to have a sig-
nificant impact on filter cake resistance in applications containing
organic material (Foley, 2006; Hodgson et al., 1993; Ohmori and
Glatz, 1999; Ohmori and Iritani, 2004). The ability to isolate the
effects of charge in such a system is therefore necessary in order
to develop and test new models. A set of novel monodisperse
model particles, with a core/shell structure and in which the sur-
face charge can be varied without compromising stability against
aggregation, was used to mimic suspensions of biological origin
and thus can be used in filtration experiments to gain knowledge
of the influence of charge in such systems.
2. Materials and methods
2.1. Particle synthesis
Micrometer-sized core/shell particles were synthesized by the
dispersion polymerization of styrene (Sigma-Aldrich, CAS: 100-
42-5, 99%) in methanol (Sigma-Aldrich, CAS: 67-56-1, 99.8%), with
the use of benzoyl peroxide (BPO) (Acros Organics, CAS: 94-36-0,
75%, remainder water) as initiator. Poly(acrylic acid) (PAA)
(Sigma-Aldrich, CAS: 9003-01-4, 35 wt.% in water) and hydrox-
ypropyl cellulose (HPC) (Sigma-Aldrich, CAS: 9004-64-2, P95%)
were used as graftable polymeric stabilizers. Before the synthesis,
benzoyl peroxide was recrystallized in methanol to remove any
water and impurities, while all other chemicals were used as
received. The synthesis was performed in 250-mL round-
bottomed two-necked flasks with reflux and magnetic stirring
and under nitrogen atmosphere. The stirring was kept constant
at 300 RPM throughout the synthesis.
To obtain particles with varying surface charge densities, differ-
ent weight ratios of HPC to PAA were used as stabilizers in the syn-
thesis. The batches of particles are named using the letters A–E,
and the amounts of stabilizers (dry weight) can be found in Table 1.
The corresponding amounts of PAA and/or HPC were dissolved
in methanol in the reaction flask, then preheated and degassed
by being bubbled with nitrogen for 10 min. BPO was dissolved in
styrene, degassed in a similar manner as was the polymer solution,
and kept at a temperature under 5 °C for no longer than 1 h until
use. The BPO/styrene solution was then added to the preheated
polymer solution and left under nitrogen at a reaction temperature
of 70 °C for 24 h. The produced particle suspension was filtered
through a filter cloth with an average pore size of 45 lm to remove
any coagulated material, and then cleaned by repeated sedimenta-
tion/resuspension in methanol five times. The cleaned particles
were all stored in methanol and then rinsed and resuspended in
an aqueous 1 mM NaOH solution before use.
The monomer contained a small amount of inhibitor (4-tert-
butylcatachol). It was deliberately not removed as it slows the
onset of polymerization, allowing the temperature of the reaction
 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
Table 1
Amounts of chemicals used in the synthesis of particle types A–E.
Particle type ID Solvent (100 mL) PAA (g)
A Methanol 0.0
B Methanol 0.3
C Methanol 0.6
D Methanol 0.9
E Methanol 1.2
mixture to rise again after addition of the monomer/initiator
solution.
2.2. Particle characterization
Particle size and size distribution were measured in a 1 mM
NaOH solution using an LS 13,320 MW laser diffraction particle
size analyzer (Beckman Coulter, Brea, CA, USA). Using a pipette, a
concentrated suspension of the particles was added to the sample
vessel under constant stirring until the obscuration was 8–12%;
measurements were made immediately thereafter. The size distri-
bution of the particles is given as the grade of polydispersity, d, cal-
culated from Eq. (2), where di is the volume-average particle
diameter below which i% of the particles are found (Tropea et al.,
2007).
d84  d16
d¼
2d50
The electrophoretic mobility was measured using elec-
trophoretic light scattering with a Zetasizer Nano ZS device (Mal-
vern Instruments, Malvern, UK). Before measuring this, the
particles were diluted to the point of almost zero turbidity in
1 mM NaOH to ensure full deprotonation of the acid groups. The
measured motilities were then converted to zeta potentials using
Smoluchowski theory. Although the zeta-potential concept is
meaningless for core/shell particles, as the surface of shear is
poorly defined, it is a common way to interpret particle charge
and has, for this reason, been chosen as the value to present. Recall,
however, that electrophoretic mobility (and therefore also zeta
potential) is still inherently connected to the charge density of
the particle shell through Ohshima’s (1995) equations and will,
as expected, increase with the increasing weight fraction of PAA
in the synthesis.
Finally, particle morphology was investigated using SEM and
optical microscopy.
2.3. Filtration experiments
The filtration equipment, shown in Fig. 1, was a piston–
cylinder-based dead-end filtration cell developed in our laboratory.
The filtration cylinder and piston used for the dead-end filtra-
tions consisted of a 116-mm-high PVC cylinder with a bore diam-
eter of 50 mm fitted with a piston sealed with a pneumatic piston
seal. The liquid pressure was measured at the piston/water inter-
face by a pressure transducer connected through a pipe. The filter
medium, which was placed between the cylinder and its base, also
functioned as a seal between the two parts. A collection chamber
was installed at the base of the cylinder, under the filter medium,
and connected via a pipe to a beaker.
The filter medium was a GPWP14250 polyethersulfone mem-
brane (Merck Millipore, Darmstadt, Germany) with a pore size of
0.22 lm and a membrane resistance of approximately
1.55  109 m1 (measured from clean water filtration experiments).
The external pressure, liquid pressure, and piston position were
logged using a load cell, pressure transducer, and linear encoder,
respectively, in a data file together with the sampling time.
157
HPC (g) Stabilizer MW (kDa) Styrene (mL) BPO (g)
1.2 100 17.5 0.5
0.9 100 17.5 0.5
0.6 100 17.5 0.5
0.3 100 17.5 0.5
0.0 100 17.5 0.5
Filtration solutions were made by diluting the concentrated parti-
cle suspensions with a 1 mM NaOH solution to a final dry matter
content of 8.33 g L1, resulting in a pH of approximately 9.5 and
a conductivity of approximately 115 lS cm1. One hundred mL of
the filtration solution was used for each filtration run, which was
allowed to last to the end of the consolidation phase to ensure
the maximum obtainable compression of the filter cake at each
pressure.
As the settling velocity of these particles is very low compared
with the filtration duration, particle settling is ignored and the sus-
pension above the filter cake is assumed to remain homogeneous
throughout the filtration process. Furthermore, as the piston drives
the suspension against the filter medium, all particles in the sus-
pension are deposited as a filter cake, as the particle size is much
larger than the pore size of the filter medium. For this reason, pore
blocking is negligible and the filter cake resistance increases with
time as the filter cake grows, while the average specific resistance
ð2Þ
should remain constant at constant pressure, assuming negligible
filter resistance and a linear increase in solid pressure as the posi-
tion in the filter cake approaches the medium, as has proven to be
the case for other synthetic particles (Christensen et al., 2006).
At some point during the filtration, the piston face will contact
the filter cake, marking the transition between the filtration and
consolidation phases. This is seen as a sharp decrease in both liquid
pressure and permeate flux, as the solid pressure in any part of the
filter cake approaches the value of the applied pressure while the
cake porosity homogenizes throughout the filter cake, expelling
only the interstitial water. At the end of the consolidation process,
the filter cake cannot compress any more, as the applied force
equals the interparticle forces and thus no more water can be
expelled.
2.4. Data analysis
The resulting data were analyzed using a modified version of
Darcy’s law of flow through a porous medium, shown in Eq. (3)
(Wakeman and Tarleton, 1999):
t la^ s
¼ Vf ; Peff ¼ DP  Dp ð3Þ
V f 2Peff A2
where t is time (s), Vf the filtrate volume (m3), l the filtrate viscosity
(Pa s), A the filter medium area (m2), Peff the effective liquid pres-
sure (Pa) (corresponding to the applied pressure, DP, minus the
osmotic pressure in the filer cake, Dp), s the amount of dry matter
deposited in the filter cake per volume of filtrate (kg m3) (or sim-
ply the dry matter content of the suspension to be filtered, if the
suspension is dilute and all dry matter is retained, as it is in this
case), and a ^ the average specific filter cake resistance (m kg1).
The membrane resistance is negligible compared with the filter
cake resistance and can be ignored.The obtained values of a ^ have
been fitted with two different empirical models. First, Tiller’s
often-used model of local cake resistance and compressibility (Eq.
(4)) is integrated to give the average cake resistance, a ^ , as a function
of the pressure drop across the cake, equal to Pliq when membrane
resistance is ignored, as shown in Eq. (5) (Tiller and Yeh, 1987). Here
a0 is the apparent zero-pressure resistance, while Pa and n are
158 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
Fig. 1. Schematic of the filtration setup showing the cylinder and piston.
empirical parameters related to cake compressibility. Second, a
modified saturation expression is suggested (akin to the one
derived for Michalis–Menten kinetics), based solely on a best-fit
approach (Eq. (6)). Here b and k are fitting parameters and a ^ 0 the
apparent zero-pressure resistance, as equivalent to Eqs. (4) and
(5). The true maximum specific resistance, a ^ max (at high pressure),
is then the sum of a ^ 0 and b. Filter cake compressibility is evaluated
from k and b, with k determining the filter cake compressive
strength (high k = high strength), i.e. the response to changes in
pressure, while b determines the limiting value that the resistance
can increase to, remembering that in the case of very high pressure
a^ = b+a^ 0 ).
The compressibility, in terms of resistance, can thus be defined
as lim ¼ da ^ =dP, making the expression b/k useful in determining
P!0
the compressibility with k – 0.

n
P is uniform throughout the cake and the local porosity will thus be
a ¼ a0 1 þ liq
Pa
Pliq =Pa
a^ ¼ a^ 0 ð1  nÞ
ð1 þ P liq =Pa Þ1n  1
Pliq b
a^ ¼ ^0
þa
Pliq þ k
Porosity data were obtained on the same system through a pres-
sure step experiment, as data on porosity from the previous filtra-
tion experiments proved not to be useful, partly due to a lower dry
matter content (which increases the uncertainty of the cake thick-
ness), but also due to deviations in suspension volume, to which
the porosity calculations are very sensitive. Therefore, particle sus-
pensions with a dry matter content of 10 g L1 were prepared in a
similar manner as were the suspensions used for filtration, except
that for this experiment the amounts of suspension used were
determined by mass, not volume, to minimize the deviations in
total volume. For each particle class, the suspension was first fil-
tered at 0.1 bar to obtain a compressible cake, the filtration being
allowed to run until no permeate flow was recorded for at least
four hours, at which point cake consolidation was considered to
be finished. The pressure was then increased in a single step to
0.25 bar to compress the cake, again maintaining this pressure
until no permeate flow was recorded for at least four hours. During
the consolidation, the cake was in constant contact with water on
the permeate side to prevent cake drying during the long period
with no water flow. This procedure was repeated with incremental
pressures steps of 0.25 bar to a final value of 1 bar. The volume of
the wet filter cake, Vc (m3), at the different pressure equilibria was
calculated by subtracting the filtrate volume, Vf (m3), and the vol-
ume lost in the tubing, Vloss (m3), from the initially added volume,
Vinit (m3). Vc was then used, along with the total mass of the
deposited particles, to determine the filter cake porosity,
e ¼ 1  /, after consolidation based on Eq. (7). The density of the
particles, qp, was set to 1050 kg m3, the same as the density of
atactic polystyrene.
ðV init  V loss Þs
/¼ ; V c ¼ V init  V loss  V f ð7Þ
V c qp
It is assumed that the cake is fully consolidated, i.e. the porosity
ð4Þ
equal to the average porosity. This allows the use of Eq. (8) to relate
solidosity, /, to solid pressure, Ps, as the local and average porosity
are equal in this situation and therefore can be used interchange-
ð5Þ ably. Similar to Eqs. (4) and (5), /0 is the solidosity at the zero-
pressure limit and b and Pa are empirical parameters related to
cake compressibility.
ð6Þ
b
Ps
/ ¼ /0 1 þ ð8Þ
Pa
To use the measured porosities for further discussion of cake
structure and intra-pore/interparticle effects, the interparticle dis-
tance, L, and pore diameter, dp, can be calculated from Eqs. (9) and
(10), with R being the particle radius. This is done by assuming an
HCP-like arrangement of the particles in the filter cake (i.e. none of
the particles forms aggregates during storage, so it is assumed that
particles repel each other and therefore might be arranged in a way
that maximizes the coordination number). The derivation is
described in detail in the Appendix. As L and dp are functions of
porosity, they will represent average values with respect to dis-
tance from the filter medium, which in this specific case (fully con-
solidated cake) will be equal to the local values as well, due to the
assumed uniformity of the cake porosity.


0:904R
L¼2 p ffiffiffiffiffiffiffiffiffiffiffi  R ð9Þ
3
1e
 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166


ð2R þ LÞ  Sinð30Þ
dp ¼ 2 R
Sinð120Þ
3. Results
During styrene polymerization, soluble polymers will be grafted
onto the growing polystyrene chains as they precipitate out of
solution, eventually forming stable core/shell particles with an
insoluble polystyrene core and a ‘‘hairy” gel-like layer of soluble
stabilizing polymers. The surface properties of the particles will
be determined by the nature of the polymers forming the ‘‘hairy”
layer.
3.1. Particle characterization
The produced particles, types A–D, all have mean particle sizes
of 1.36–1.41 lm, while particle type E has a mean particle size of
2.21 lm. Furthermore, all the particles are monodisperse within
each class, with a grade of polydispersity of 0.14–0.17.
The numerical value of the measured zeta potential increases
linearly with the weight percentage of PAA in the synthesis feed,
as shown in Fig. 2, with values ranging from 1.3 to 51.2 mV.
These results indicate that, even though the two polymers used
for stabilization exert their stabilizing effect in two distinctively
different ways (i.e. electrostatic with PAA and steric with HPC),
these effects can complement each other, as all the produced par-
ticles are stable against aggregation when suspended in either
methanol or aqueous 1 mM sodium hydroxide, settling only very
slowly when in concentrated suspension.
Furthermore, scanning electron micrographs showed the parti-
cles to be spherical and confirmed a narrow size distribution.
3.2. Filtrations
From the filtrations, filtrate volume was recorded as a function
of time for each of the five DP used. Based on these data, t Vf1 was
plotted versus Vf for each filtration run, an example of which is
shown in Fig. 3 for particle type C at a DP of 1 bar. In Fig. 3, the lin-
ear part of the t Vf1 vs. Vf plot, the slope of which is used in calcu-
lating a ^ from Eq. (3), has been fitted with a linear regression, which
is indicated by a dashed grey line. Also seen in this figure is the
transition to the consolidation phase, i.e. the point at which the pis-
ton contacts the filter cake, shown by a very sharp increase in the
slope of the plot to almost vertical. This fast transition is a good
indication that creep effects are nonexistent or negligible, which
can also be assumed to be the case during the filtration phase.
It is also worth mentioning that all of the filtrations fulfilled the
criterion of achieving a constant average specific cake resistance
Fig. 2. Zeta potential of particle types A–E with increasing wt% PAA in the synthesis
feed; error bars show standard deviation.
159
ð10Þ
Fig. 3. Example of a t V1 vs. V plot for particle type C (DP = 1 bar); dashed line
indicates the linear fit, the slope of which is used to calculate a
^.
once a filter cake has formed (after filtering approximately 20%
of the total filtrate volume). There is, however, some deviation
from this at the beginning of every filtration run, just as the parti-
cles begin to deposit. This could be caused by particles not yet cov-
ering the entire filter area or changes in cake structure (i.e. cake
compression and blocking of pores in the membrane surface, if a
particle is deposited on top of a pore).
^ , of the filter cake
Using Eq. (3), the average specific resistance, a
can be calculated from data obtained during dead-end filtration.
Fig. 4 shows the resulting resistances of particle types A–E at pres-
sures of 0.1–1 bar. All particles form compressible filter cakes, as
resistance increases with DP, with the general trend that particles
with a higher surface charge form a more compressible filter cake
while yielding a higher specific resistance at all values of DP.
The model proposed in Eq. (5) is shown in Fig. 4, using particle
type D as an example. Even though this model provides a reason-
ably good fit with a coefficient of determination of 0.95, it does
not seem to follow the experimental data at higher DP very well.
Furthermore, the fitting parameters a0 and Pa are highly correlated
(nonlinear), such that a series of equally good fits can be presented
with different values of a0 and Pa, and hence also of n. This would
make comparison of filtration behavior between the particles
impossible, as the choice of fitting parameters becomes rather arbi-
trary. For this reason, this model will not be used further here.
Fig. 4 also shows the fitted model data based on Eq. (6) (dotted
black lines), with the coefficient of determination exceeding 0.98 in
all cases. Although this model does fit the data very well in the
range between 0.1 and 1 bar, it is difficult to say whether it is accu-
Fig. 4. Specific filter cake resistance as a function of applied pressure; dotted lines
show the model from Eq. (6), while the dashed line shows the model from Eq. (5).
160 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
rate below 0.1 bar, as no experimental data are available in this
range. Despite this, data from the model are used in determining
a^ 0 and filter cake compression, as presented in the Section 4, and
these data should be evaluated bearing in mind the high degree
of uncertainty. The use of this model is preferred, as the fitting
parameters are not correlated to the same degree as are those in
Eq. (5), while the fit to experimental data seems better, as evident
in Fig. 4.
Because of the filter cake compressibility, one would expect the
filter cake porosity to change as well. Based on pressure step
experiments, the filter cake porosities for all particle types have
been measured at approximately the same DP values as used in
the filtrations and are shown as a function of DP in Fig. 5. As
expected, porosity decreases with pressure for all particles, while
average porosity (over the whole DP range) increases with particle
charge. Furthermore, porosity apparently decreases faster and
more linearly when the particle charge is low, as compared with
the highly charged particles. By using an indirect measurement
of cake porosity, the result might be subject to a high degree of
uncertainty. For more certainty, the porosity should have been
measured gravimetrically, as mass is lost during drying, but as
the cake swells as soon as the pressure is released, this would
prove difficult. In the few instances in which it was possible to iso-
late a portion of the filter cake that had not swelled, the gravimet-
rically determined water content was within the 10% error margin
of that determined from piston position. Furthermore, porosities
within the same range have been measured during the analytical
centrifugation of similar particles (unpublished results).
Fig. 5 also shows Eq. (8) fitted to the experimental data. This
model proved to be an exceptionally good fit, with the coefficient
of determination exceeding 0.99 in all cases except that of particle
type B (0.96). The fitting parameters can be found in Table 2, where
most notably b is within the range found in compression experi-
ments on biological suspensions (Tiller and Kwon, 1998). It is dif-
ficult to compare e0, as most suspended material of biological
origin is characterized by a high volume fraction of water inside
the actual particles/cells and/or trapped between the aggregated
particles as interstitial water. This would, depending on how
porosity is measured, lead to assumptions of a pore volume larger
than the actual volume available for flow in the cake, which in turn
would lead to a very high porosity. This is not the case in this work,
as the particle core does not swell in water. Pa is also well above
the range normally encountered for biological material, but as Pa
and e0 are correlated, and e0 is high for the reasons mentioned
above, direct comparison with biological material would not be
meaningful. There is, however a clear dependence on all particle
Fig. 5. Filter cake porosity as a function of applied pressure; dashed/dotted lines
show the model from Eq. (8).
Table 2
Fitting parameters obtained from fitting Eq. (8) to experimentally obtained data on
filter cake porosity.
Particle ID e0 Pa (bar) b
A 0.590 1.303 0.643
B 0.602 1.181 0.576
C 0.615 1.060 0.508
D 0.624 0.911 0.436
E 0.642 0.724 0.360
charge parameters, indicating that charge does play a major role
in filter cake compression.
Based on the obtained porosity values, it is possible to calculate
the interparticle distance from Eq. (9) (i.e. the shortest distance
between two neighboring particles) and the minimum pore diam-
eter from Eq. (10). In doing so, it is assumed that the particles are
arranged in a HCP-like lattice, as explained in the Section 2, an
assumption reviewed later in the Section 4. This can supplement
the porosity values, giving another way of interpreting the filter
cake structure. To assess the influence of particle charge on cake
structure, this has been done for data obtained at DP = 1 bar and
is presented in Fig. 6 as a function of particle f potential. The pore
diameter increases from 300 nm to 710 nm, going from low to high
charge, while the interparticle distance increases from 86 nm to
320 nm.
4. Discussion
Based on fitting Eq. (6) to experimental data (acknowledging
the high degree of uncertainty), values of a ^ 0 and cake compress-
ibility have been obtained and are presented in Fig. 7. From these
values it can be seen that the surface charge affects not only the
compressibility of the filter cake but also the general resistance
to filtration when no compression is taking place, i.e. at the zero-
pressure limit. The actual cake compression experienced will of
course depend on DP, due to the curvature of the model. Any
change in DP will have a large impact on a ^ at low pressures but
a significantly smaller impact at higher pressures. Although there
is no obvious explanation for this effect, the notion of a maximum
resistance at very high DP would make sense, as there is a limit to
how compressed the filter cake can become, due to the hard poly-
styrene core of the particles.
The results of comparing cake porosity and resistance, shown in
Fig. 8, make it clear that the effects of particle charge significantly
outweigh the effects of porosity, even at low porosity values. In this
figure, the measured specific resistances have been divided by s20 to
eliminate any effect caused by differences in particle size.
Fig. 6. Interparticle distance, pore diameter, and porosity (Papp = 1 bar) as a function
of particle zeta potential.
 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
Fig. 7. Filter cake compressibility and zero-pressure resistance from fitting the
experimental data using Eq. (6).
Fig. 8. The influence of porosity on specific resistance, a
^ , including contributions of
particle size.
From these porosities, we can also calculate the expected filter
cake resistance according to the Kozeny–Carman model, both the
original and the modified version of Rumpf and Gupte (1975), Eq.
(11). The latter was chosen partly for its simplicity but also because
it has been found to fit better than the original Kozeny–Carman
model, within the range of 0.3 < e < 0.67, comparable to the exper-
imental data presented here.
5:6
aKC mod ¼ 0:55 d2
ð11Þ
e settle in a position that, ideally, maximizes the coordination num-
These methods also effectively remove the contribution of par-
ticle size, allowing all five particle types (i.e. A–E) to be fully com-
pared. In addition, the effects of porosity are reduced to the effects
caused by changes in particle surface charge density. The original
model (Fig. 9, left) greatly underestimates the specific cake resis-
tance for all particles, although the error increases with particle
charge to a measured value almost 400 times higher than that cal-
culated from the model. The modified model (Fig. 9, right) is better
at estimating the actual cake resistance observed, especially for the
uncharged particles (type A), but still underestimates cake resis-
tance when the particles are charged.
As stated previously, the filter cake is assumed to be in an HCP-
like configuration, or at least a random close packed configuration,
containing significant areas of HCP structure. This assumption was
originally based on the lack of formation of aggregates, during both
storage (sedimentation) and centrifugation (for a short time) and
upon particle deposition in the existing filter cake, as observed
during cross-flow filtration by Lorenzen et al. (2016). As conditions
on the surface of the filter cake during dead-end mode can differ
greatly from those during cross-flow, mainly in the lack of tangen-
161
tial flow, calculations were carried out to evaluate the possibility of
particles adhering either upon impact or due to drag forces on the
top of the filter cake. The calculations are based on the use of DLVO
theory, ignoring van der Waals attractions, as these have been
found to be negligible for this type of particle (the Hamaker con-
stant for the particle shell should be very close to that of water
due to the high degree of swelling). Although this theory cannot
explain the high observed porosities or resistances, it is used here,
as an extreme case, as the experimental results indicate that the
particle repulsion is greater than predicted from DLVO. The parti-
cles in the bulk suspension during filtration will have a net veloc-
ity, normal to the filter cake, equal to the flux of water, assuming
the particles have accelerated to match the flow velocity of the
water. In the case of one particle approaching another particle
(in the filter cake) from directly overhead at velocity vp, for the par-
ticles to adhere on impact, the kinetic energy of the moving parti-
cle must be greater than the maximum repulsive energy that can
arise from bringing the two particles closer together. Using Eq.
(12) with particle type E as an example (neglecting steric effects)
and solving for particle separation, x, we find that the kinetic
energy of the moving particle (after 5 min of filtration at a rela-
tively high flux) is only sufficient to bring the two particles to
within approximately 140 nm of each other, avoiding collision. In
Eq. (12), mp is the mass of one particle, vp is the velocity of the
incoming particle (equal to the water flux through the membrane),
1r and 10 are the relative permittivities of water and vacuum,
respectively, rp is the particle radius, f is the particle zeta potential
(the same for both particles), and j the inverse Debye length.
4p1r 10 r 2p f2 ejx
1=2mp v 2p ¼ ð12Þ
2r p x
The case cited here is an ideal case, however, in which the par-
ticles are perfectly aligned. As the two spherical particles approach
each other, the increasing repulsion between them will force them
to deflect if there is even the slightest offset in force or position.
This would cause the incoming particle to change its trajectory,
instead of just stopping, as downward force exerted by drag forces
from the flowing water is still acting on the particle. Assuming a
purely laminar flow and using Stokes’ Law, the absolute maximum
drag a particle can experience (after 5 min of filtration) will only be
enough to bring two directly opposing particles to within 50 nm of
each other, again using DLVO theory and ignoring steric effects and
van der Waals attractions. As this drag is incapable of making the
particles aggregate, it could be assumed that they will eventually
ber, i.e. HCP with the coordination number 12.
On the other hand, one might argue that the sheer number of
settling particles will make the arrangement more chaotic, pre-
venting the particles from settling into an HCP structure; they
would instead become interlocked in more open structures, though
still without actually being in direct contact. This can be seen as
another extreme, in which the filter cake comprises many inter-
locked clusters, each made of several particles held closely
together. In this case, porosity could be high due to the volumes
between the interlocked clusters, though the water would still
have to flow through some very narrow pores (bottlenecks) if the
clusters were connected to each other, as schematized in Fig. 10.
In this scenario, the distance between individual particles could
be very small, and the filter cake might then behave, in terms of
resistance, like one with a very low porosity, due to the aforemen-
tioned bottlenecks within the structure. Assuming the effective
porosity to be as low as possible with the spheres packed as closely
as possible (i.e. the pore size in the bottlenecks is the same as when
e = 0.26), we can calculate the expected a^ from the Kozeny–Car-
man model used before, which would be 3.9 Tm kg1. Compared
162 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
Fig. 9. Ratio between experimental and theoretically determined a
^ , for the classical Kozeny–Carman model and the modified version of Rumpf and Gupte (1975).
Fig. 10. Schematic of particles interlocked in a highly porous structure but with
water flow bottlenecks.
with the experimental results, this would be sufficient to explain
only the resistances found for particle type A. So even in these
extreme situations, the classical interpretation of filter cake resis-
tance cannot explain the high values measured, emphasizing the
previous observation that a ^ depends more on particle charge than
on porosity.
Further investigating what could cause an increase in resistance
when dealing with charged particles leads to the possibility of elec-
troviscous effects in the filter cake. These effects would cause the
apparent viscosity of the water in the pores to increase, due to
the charged surfaces of the particles. Fig. 11 illustrates this effect
as a function of the zeta potential of pores of different sizes, with
the y-axis showing the increase in viscosity as a percentage of
the normal viscosity. From this it can be seen that even at high zeta
potentials, a very small pore size will be needed to have any notice-
able effect, and as the absolute minimum pore size in this case is
200 nm (e = 0.26), for an HCP structure, this cannot explain the
observed values either. Even in the case of irregular pores, as in
Fig. 11. Percent increase in viscosity, due to electroviscous effects, for different pore
sizes and zeta potentials.
the interlocked, open-structured scenario, the minimum particle
separation (based on drag forces) was 50 nm, which is still not
enough to have an appreciable effect. This is because the electro-
viscous effects are caused by the double layer surrounding a
charged particle, which in this case is only around 10 nm in
diameter.
Another proposed cause is increased osmotic pressure in the
pores due to the high number of counterions associated with the
charged surfaces, as mentioned in the Introduction. Although this
could make sense and is clearly observable in terms of concentra-
tion polarization in nanofiltration and reverse osmosis, the pores
are still too big, as classical double-layer theory predicts that most
counterions will be within the double layer of the surface and
would not extend far enough into the pores to have an effect.
Recent experimental studies have found, however, that charged
particles confined between charged walls might behave very dif-
ferently from the predictions of classical theories. This could also
apply in the case of charged particles confined by other charged
particles, such as in the filter cake. The findings of Larsen and Grier
(Grier, 1998; Larsen and Grier, 1997) and of Bowen and Sharif
(1998) indicate the presence of long-range attractive forces at par-
ticle separations of up to several micrometers. Even though there is
still discussion of the theoretical grounds for this phenomenon,
 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
and even of whether the effect is attractive or repulsive (Sader and
Chan, 1999; Trizac and Raimbault, 1999), one proposed explana-
tion suggests that redistribution of the counterions in the double
layer (Bowen and Sharif, 1998) allows for more evenly distributed
ions throughout the pore volume. Larsen and Grier (Grier, 1998;
Larsen and Grier, 1997) demonstrated the effect to be negligible
for a particle–wall separation of 9.5 lm, while having a pro-
nounced impact at a particle–wall separation of 2.5 lm. Fig. 6
shows that the measured pore diameter in the filter cake at a DP
of 1 bar is 0.3–0.7 lm, implying that such a redistribution of coun-
terions might occur in the filter cake pores, which would serve to
increase the osmotic pressure. Studies of flow in microchannels
obtain similarly mixed results, some indicating long-range surface
interactions in the lm range (Li, 2001; Ren et al., 2001), others
indicating an early transition from laminar to turbulent flow
(Bayraktar and Pidugu, 2006). Using the osmotic pressure model,
based on the redistribution of counterions, it is possible to combine
Eqs. (3) and (11) and then isolate Peff. This would give the actual
effective pressure needed for the experimental data to fit those
modeled by Eq. (11). As also stated in Eq. (3), the effective pressure
is the difference between the applied pressure and the filter cake
osmotic pressure, so it is now possible to estimate the filter cake
osmotic pressure for the different particles and hence also the
counterion concentrations in the pores. This is shown in Fig. 12
and 13, where the osmotic pressure increases with both particle
charge and decreasing porosity. Figs. 12–14 show only results for
particle types B–E, as particle type A has no PAA on the surface.
For highly charged particles, the osmotic pressure would then
almost equal the applied pressure with a linear correlation
Fig. 12. Calculated osmotic pressure in the pores as a function of filter cake
porosity.
Fig. 13. Calculated osmotic pressure in the pores as a function of applied pressure.
163
Fig. 14. The ratio between the concentration of available counterions from PAA and
the concentration needed for the calculated osmotic pressure. Fig. A-1 2-D
representation of three spheres packed in an HCP structure, forming a tetrahedral
pore.
between the two. Based on a grafting density of PAA to the particle
surfaces of 5.7  1015 grafts per square meter, it is possible to cal-
culate the maximum concentration of counterions in the pores
(assuming the available counterions are distributed evenly in the
pore volume). The ratio between this value and the concentration
needed to achieve the previously determined osmotic pressure is
presented in Fig. 14.
Except in two instances (particle type B at the two lowest
porosities), all the data points are >1. This could indicate that the
notion of osmotic pressure in filter cakes, based on the theory of
counterion redistribution proposed by Bowen and Sharif (1998),
might be a valid explanation for the high values of a ^ often
observed when filtering wastewater sludge and other suspensions
of biological origin.
5. Conclusion
Dispersion polymerization has been used to synthesize a series
of core/shell model particles with varying surface charges. The zeta
potential of the particles ranges from 1.31 to 51.24 mV, exhibit-
ing a linear correlation with the weight fraction of polyacrylic acid
used in the synthesis. All the produced particles are monodisperse,
stable against aggregation in 1 mM NaOH, and exhibit no change in
surface properties over time (up to several months).
To investigate the influence of particle charge on filter cake
behavior, the produced particles were used in a model suspension
in a series of dead-end filtration experiments at the following val-
ues of DP: 0.1, 0.25, 0.5, 0.75, and 1 bar. Results indicate that speci-
fic cake resistance generally increases with increasing particle
charge, with a nonlinear relationship between applied pressure
and cake resistance. At low DP, a small change in pressure will
result in a large change in resistance, whereas at high DP, a large
change in pressure will result in only a small change in resistance,
with the resistance eventually leveling off at a maximum value.
Measured porosities have also been found to increase with par-
ticle charge, which stands in contrast to the cake resistance, as a
more porous cake would normally yield a lower specific resistance.
Furthermore, the porosity decreases with DP in all instances.
Comparison of the experimental data with model data based on
two empirical models clearly indicates particle charge to be an
important factor, as the ratio between the experimental data and
the models and increase by a factor of 20–400, depending on the
model, as the particle charge increases.
The pore size calculated from the porosities suggests that the
pores are too big for the electric double layer to have any apprecia-
164 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
ble effect on resistance, as in the case of electroviscous effects,
which should also be the case for osmotic pressure from the coun-
terions. In addition, the interparticle distance is greater than would
be expected from a classical understanding of electrostatic repul-
sion, as applied in DLVO theory. Even though this theory cannot
explain the observed filter cake behavior, recent studies have
found that the behavior of charged colloids confined between
charged walls or within charged pores might differ from that based
on a classical understanding, meaning that the particles interact
over much longer distances than predicted. Based on this, the
osmotic pressure of the filter cake was calculated and found to
increase with particle charge. Furthermore, the concentration of
counterions in the pores (equally distributed) was compared with
that needed in order to obtain the calculated osmotic pressure and
was found to be sufficient in almost all instances.
Acknowledgements
This research is part of the EcoDesign-MBR Research Centre
founded by the Danish Council for Strategic Research. Special
thanks are directed towards the Department of Physics and Nan-
otechnology at Aalborg University for assistance with SEM
imagery.
Appendix - Calculation of pore size and interparticle distance
In a hexagonal close packed (HCP) lattice, formed by spheres of
the same size, two types of channels (pores) form, when looking
from the top down.
One type can be approximated as a line of octahedral voids
(although they are actually concave octahedrons) extending
straight through the entire lattice (assuming no defects or random-
ness). The tortuosity of this type of channel is ideally 1.
Fig. A-1. 2-D representation of three spheres packed in an HCP structure, forming a tetrahedral pore.
The other type can be approximated by a series of tetrahedral
voids (again concave), the corners of which are all connected to
the octahedral voids. This increases the tortuosity, although one
can question whether these tetrahedral voids will participate in
direct flow through the lattice, as they provide connections only
to the straight octahedral channels. If not, they should in principle
be omitted from the porosity when dealing with liquid flow.
Either way, the corners of these tetra- and octahedrals are not
completely pointed, but can instead be described as concave trian-
gular faces (imagine cutting off the tips of the tetrahedral voids
described above). These faces provide the entry points to all the
voids and will constitute the narrow points in the pores, regardless
of whether the pores are based on tetrahedrons or octahedrons. If
it is assumed that the sides of this triangle are not concave, and if
these sides are all tangential to the surface of the surrounding
spheres, then the incircle of the triangle will describe the cross-
sectional area, and thus the radius, of the largest sphere that can
pass through the pores without moving the other spheres. This
radius can then be viewed as the absolute minimum pore radius
possible in an HCP lattice, recalling that the pores all widen when
moving into the tetra- and octahedral space.
A.1. Calculation of minimum pore radius
Fig. A-1 is a two-dimensional representation of the pore open-
ing, i.e. a cross section of the point where the pore is narrowest.
The inner triangle (grey) represents the triangular face described
above, while the small circle (grey) is the incircle of that triangle.
The corners of the outer triangle are positioned in the centers of
the big circles, with the angle bisectors drawn in as well.
Based on the knowledge that the big triangle is equilateral, all
the angles must be 60°, the length of the sides must be 2R (R being
the radius of the big circles), and the angle bisectors must be per-
pendicular to the opposite side. It is therefore possible to draw the
 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166 165

right-angled triangle (bottom right-handed side of the outer trian- Bowen, W.R., Sharif, A.O., 1998. Long-range electrostatic attraction between like-
charge spheres in a charged pore. Nature 393, 663–665. http://dx.doi.org/
gle) where the hypotenuse equals R + r (r being the radius of the
10.1038/31418.
small circle) and consequently to calculate r. Bugge, T.V., Larsen, P., Saunders, A.M., Kragelund, C., Wybrandt, L., Keiding, K.,
Christensen, M.L., Nielsen, P.H., 2013. Filtration properties of activated sludge in
Cosð30Þ ¼ Rþr R
! r ¼ Cosð30Þ
R
R municipal MBR wastewater treatment plants are related to microbial
  community structure. Water Res. 47, 6719–6730. http://dx.doi.org/10.1016/j.
r ¼ R p2ffiffi3  1 watres.2013.09.009.
Bugge, T.V., Jørgensen, M.K., Christensen, M.L., Keiding, K., 2012. Modeling cake
buildup under TMP-step filtration in a membrane bioreactor: cake
At e > 0.26, the minimum pore radius can be calculated from the
compressibility is significant. Water Res. 46, 4330–4338. http://dx.doi.org/
interparticle distance according to 10.1016/j.watres.2012.06.015.
Carman, P.C., 1937. Review of literature, confirming validity of Kozeny’s theory.
ð2R þ 2DrÞ  Sinð30Þ Trans ICHemE 15, 150–166.
rp ¼ R
Sinð120Þ Chen, J., Zhang, M., Wang, A., Lin, H., Hong, H., Lu, X., 2012. Osmotic pressure effect
on membrane fouling in a submerged anaerobic membrane bioreactor and its
experimental verification. Bioresour. Technol. 125, 97–101. http://dx.doi.org/
10.1016/j.biortech.2012.08.038.
A.2. Interparticle distance Christensen, M.L., Sedin, P., Theliander, H., Keiding, K., 2006. Pressure and
concentration profiles in filter cake consisting of core/shell latex particle.
The interparticle distance is the shortest distance between par- Colloids Surf. A Physicochem. Eng. Asp. 290, 295–303. http://dx.doi.org/
10.1016/j.colsurfa.2006.05.037.
ticle surfaces. By again assuming an HCP lattice of equally sized Chu, H.P., Li, X.Y., 2005. Membrane fouling in a membrane bioreactor (MBR): sludge
spheres in its densest state (e = 0.26), the interparticle distance is cake formation and fouling characteristics. Biotechnol. Bioeng. 90, 323–331.
per definition zero (i.e. the spheres are touching each other). For http://dx.doi.org/10.1002/bit.20409.
Curvers, D., Usher, S.P., Kilcullen, A.R., Scales, P.J., Saveyn, H., Van der Meeren, P.,
this system, it has been demonstrated that the volume fractions 2009. The influence of ionic strength and osmotic pressure on the dewatering
of the unit cell can be calculated as follows (Granton and Fraser, behaviour of sewage sludge. Chem. Eng. Sci. 64, 2448–2454. http://dx.doi.org/
1935). 10.1016/j.ces.2009.01.043.
Davis, R.H., 1992. Modeling of fouling of crossflow microfiltration membranes. Sep.
V total ¼ 5:65R3 Purif. Rev. 21, 75–126. http://dx.doi.org/10.1080/03602549208021420.
Foley, G., 2006. A review of factors affecting filter cake properties in dead-end
V sphere ¼ 4:18R3 microfiltration of microbial suspensions. J. Memb. Sci. 274, 38–46. http://dx.doi.
org/10.1016/j.memsci.2005.12.008.
V v oid ¼ 1:47R3 Granton, L.C., Fraser, H.J., 1935. Systematic packing of spheres: with particular
relation to porosity and permeability. J. Geol. 43, 785–909.
Now imagine that each sphere is moved away from any other Grier, D.G., 1998. Colloids: a surprisingly attractive couple. Nature 393, 621–623.
sphere by an equal distance, Dr. This gives a structure in which http://dx.doi.org/10.1038/31340.
the spheres are still positioned in an HCP-like lattice, but are no Hodgson, P., Leslie, G., Fane, A., 1993. Cake resistance and solute rejection in
bacterial microfiltration: the role of the extracellular matrix. J. Memb. Sci. 79,
longer as closely packed (e > 0.26). The new distance between 35–53. http://dx.doi.org/10.1016/0376-7388(93)85016-P.
two spheres will thus be 2Dr, as both are moved distance Dr rela- Keiding, K., Rasmussen, M.R., 2003. Osmotic effects in sludge dewatering. Adv.
tive to their former positions. This can be thought of as each sphere Environ. Res. 7, 641–645. http://dx.doi.org/10.1016/S1093-0191(02)00043-6.
Larsen, E.A., Grier, G.D., 1997. Electrodynamical like-charge attractions in
swelling an equal amount, with Dr + R being their new radius, and metastable colloidal crystallites. Nature. http://dx.doi.org/10.1142/
then shrinking again, while keeping their newly obtained posi- S0217984998001013.
tions. Noting that the sphere volume does not change, the new vol- Le-Clech, P., Chen, V., Fane, T.A.G., 2006. Fouling in membrane bioreactors used in
wastewater treatment. J. Memb. Sci. 284, 17–53. http://dx.doi.org/10.1016/j.
ume fractions will then be: memsci.2006.08.019.
Li, D., 2001. Electro-viscous effects on pressure-driven liquid flow in microchannels.
V total ¼ 5:65ðR þ DrÞ3 Colloids Surf. A Physicochem. Eng. Asp. 195, 35–57. http://dx.doi.org/10.1016/
S0927-7757(01)00828-7.
V sphere ¼ 4:18R3
Lorenzen, S., Hinge, M., Christensen, M.L., Keiding, K., 2014. Filtration of core-shell
V v oid ¼ V total  V sphere ¼ 5:65ðR þ DrÞ3  4:18R3 colloids in studying the dewatering properties of water-swollen materials.
Chem. Eng. Sci. 116, 558–566. http://dx.doi.org/10.1016/j.ces.2014.05.029.
The new porosity will then become: Lorenzen, S., Ye, Y., Chen, V., Christensen, M.L., 2016. Direct observation of fouling
phenomena during cross-flow filtration: influence of particle surface charge. J.
Memb. Sci. 510, 546–558. http://dx.doi.org/10.1016/j.memsci.2016.01.046.
V v oid 5:65ðR þ DrÞ3  4:18R3 4:18R3
e¼ ¼ 3
¼1 Mikkelsen, L.H., Keiding, K., 2002. Physico-chemical characteristics of full scale
V total 5:65ðR þ DrÞ 5:65ðR þ DrÞ3 sewage sludges with implications to dewatering. Water Res. 36, 2451–2462.
http://dx.doi.org/10.1016/S0043-1354(01)00477-8.
and from this, Dr can be isolated, recalling that the interparticle dis- Mukai, Y., Iritani, E., Murase, T., 1997. Effect of protein charge on cake properties in
dead-end ultrafiltration of protein solutions. J. Memb. Sci. 137, 271–275. http://
tance is 2Dr, as follows: dx.doi.org/10.1016/S0376-7388(97)00222-6.
3 Novak, J.T., Goodman, G.L., Pariroo, A., Huang, J., 1988. The blinding of sludges
1  e ¼ 5:65ðRþ
4:18R
DrÞ3
during filtration. Water Pollut. Control Fed. 60, 206–214.
p ffiffiffiffiffiffiffiffiffiffiffi q ffiffiffiffiffiffiffi Ogden, G.E., Davis, R.H., 1990. Experimental determination of the permeability and
3
1  e ¼ 3 4:18  R ¼ 0:904  1þ1Dr
5:56 RþDr
relative viscosity for fine latexes and yeast suspensions. Chem. Eng. Commun.
R 91, 11–28. http://dx.doi.org/10.1080/00986449008940698.
1 þ DRr ¼ p3ffiffiffiffiffiffi
0:904 Ohmori, K., Glatz, C.E., 1999. Effects of pH and ionic strength on microfiltration of C.
1e glutamicum. J. Memb. Sci. 153, 23–32. http://dx.doi.org/10.1016/S0376-7388
3ffiffiffiffiffiffi  1
Dr
R
¼p0:904 (98)00239-7.
1e
  Ohmori, K., Iritani, E., 2004. Effect of addition of amino acids on microfiltration
Dr ¼ R p 3ffiffiffiffiffiffi  1
0:904 behavior of a microorganism slurry. J. Chem. Technol. Biotechnol. 79, 1169–
1e 1173. http://dx.doi.org/10.1002/jctb.1047.
p3ffiffiffiffiffiffi  R
Dr ¼ 0:904R Ohshima, H., 1995. Electrophoretic mobility of soft particles. Colloids Surf. A
1e Physicochem. Eng. Asp. 103, 249–255. http://dx.doi.org/10.1016/0927-7757
(95)03293-M.
Ren, L., Qu, W., Li, D., 2001. Interfacial electrokinetic effects on liquid flow in
References microchannels. Int. J. Heat Mass Transf. 44, 3125–3134. http://dx.doi.org/
10.1016/S0017-9310(00)00339-2.
Rumpf, H., Gupte, A.R., 1975. The influence of porosity and grain size distribution on
Bayraktar, T., Pidugu, S.B., 2006. Characterization of liquid flows in microfluidic
the permeability equation of porous flow. Chem. Ing. Techn. 43, 367–375.
systems. Int. J. Heat Mass Transf. 49, 815–824. http://dx.doi.org/10.1016/j.
Sader, J.E., Chan, D.Y.C., 1999. Long-range electrostatic attractions between
ijheatmasstransfer.2005.11.007.
identically charged particles in confined geometries: an unresolved problem.
Sørensen, Birgitte L., Keiding, Kristian, Lauritzen, Steffen L., 1997. A Theoretical
J. Colloid Interf. Sci. 213, 268–269.
Model for Blinding in Cake Filtration. Water Environ. Res. 69, 168–173.
166 S. Lorenzen et al. / Chemical Engineering Science 163 (2017) 155–166
Tiller, F.M., Kwon, J.H., 1998. Role of porosity in filtration: XIII. Behavior of highly
compactible cakes. AIChE J. 44, 2159–2167. http://dx.doi.org/10.1002/
aic.690441005.
Tiller, F.M., Yeh, C.S., 1987. The role of porosity in filtration. Part XI: filtration
followed by expression. AIChE J. 33, 1241–1256. http://dx.doi.org/10.1002/
aic.690330803.
Trizac, E., Raimbault, J.-L.L., 1999. Long-range electrostatic interactions between
like-charged colloids: steric and confinement effects. Phys. Rev. E 60, 6530–
6533. http://dx.doi.org/10.1103/PhysRevE.60.6530.
Tropea, C., Yarin, A.L., Foss, J.F., 2007. Springer Handbook of Experimental Fluid
Mechanics. Springer. http://dx.doi.org/10.1007/978-3-540-30299-5.
Valdes-Parada, F.J., Ochoa-Tapia, J.A., Alvarez-Ramirez, J., 2009. Validity of the
permeability Carman-Kozeny equation: a volume averaging approach. Phys. A
Stat. Mech. its Appl. 388, 789–798. http://dx.doi.org/10.1016/
j.physa.2008.11.024.
Wakeman, R.J., Tarleton, E.S., 1999. Filtration: Equipment Selection, Modelling and
Process Simulation. Elsevier Advanced Technology.
Wijmans, J.G., Nakao, S., Smolders, C.A., 1984. Flux limitation in ultrafiltration:
osmotic pressure model and gel layer model. J. Memb. Sci. 20, 115–124. http://
dx.doi.org/10.1016/S0376-7388(00)81327-7.
Wu, R.M., Lee, D.J., Wang, C.H., Chen, J.P., Tan, R.B.H., 2001. Novel cake
characteristics of waste-activated sludge. Water Res. 35, 1358–1362. http://
dx.doi.org/10.1016/S0043-1354(00)00513-3.
Xu, P., Yu, B., 2008. Developing a new form of permeability and Kozeny-Carman
constant for homogeneous porous media by means of fractal geometry. Adv.
Water Resour. 31, 74–81. http://dx.doi.org/10.1016/j.advwatres.2007.06.003.
Zhang, M., Peng, W., Chen, J., He, Y., Ding, L., Wang, A., Lin, H., Hong, H., Zhang, Y., Yu,
H., 2013. A new insight into membrane fouling mechanism in submerged
membrane bioreactor: osmotic pressure during cake layer filtration. Water Res.
47, 2777–2786. http://dx.doi.org/10.1016/j.watres.2013.02.041.
Zhang, Y., Zhang, M., Wang, F., Hong, H., Wang, A., Wang, J., Weng, X., Lin, H., 2014.
Membrane fouling in a submerged membrane bioreactor: effect of pH and its
implications. Bioresour. Technol. 152, 7–14. http://dx.doi.org/10.1016/j.
biortech.2013.10.096.