You are on page 1of 10

Chemical Engineering Science 58 (2003) 2893 2902

www.elsevier.com/locate/ces

Particle size distributions and viscosity of suspensions undergoing


shear-induced coagulation and fragmentation
G. Barthelmesa , S. E. Pratsinisa; , H. Buggischb
a Particle Technology Laboratory, Department of Mechanical and Process Engineering, Sonneggstrasse 3, ML F25, ETH Z!urich,
CH-8092 Z!urich, Switzerland
b Institut f!
ur Mechanische Verfahrenstechnik und Mechanik, Universit!at Karlsruhe (TH), D-76128 Karlsruhe, Germany
Received 23 September 2002; received in revised form 13 December 2002; accepted 21 February 2003

Abstract

The dynamic behavior of concentrated suspensions (up to a solids volume fraction of 20%) of non-spherical particles is investigated
theoretically by coupling a rheological law to a population balance model accounting for coagulation and fragmentation of the detailed
particle size distribution. In these suspensions, the immobilization of matrix liquid renders the viscosity dependent on the particle
aggregation state. The e5ect of initial solids volume concentration and shear rate on the transient behavior of particle size distribution
and suspension viscosity is examined. Power law correlations for the equilibrium 6ow curves of aggregating suspensions are deduced
and compared to experimental data. Steady-state or equilibrium particle size distributions are found to be self-preserving with respect to
solids volume fraction and shear rate.
? 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Particles; Fractals; Self-preserving; Population balance; Modeling; Rheology

1. Introduction increase the e5ective solids volume fraction in the sus-


pension and its viscosity (Graham, Steele, & Bird, 1984;
Simultaneous coagulation (or aggregation) and fragmen- Tsutsumi, Yoshida, Yui, Kanamori, & Shibata, 1994).
tation by 6uid shear is encountered in processes involving Increasing the viscosity increases the shear stresses that
polymerization (Blatz & Tobolsky, 1945), liquidliquid dis- enhance breakage of aggregates. The structure of the ag-
persion (Coulaloglou & Tavlarides, 1977), emulsi>cation gregates and, thus, the rheology strongly depends on the
(Danov, Ivanov, Gurkov, & Borwankar, 1994), 6occula- applied shear stresses.
tion (Ives, 1978) and even diesel soot formation (Harris & Aggregating suspensions show shear thinning behavior
Maricq, 2002). At low particle concentrations the 6uid even with Newtonian matrix liquids. At low shear rates
and particle dynamics can be decoupled, facilitating thus their viscosityshear rate correlation can be approximated
the description of the suspension dynamics (e.g. Spicer & by power laws (Doi & Chen, 1989; Chen & Doi, 1989). A
Pratsinis, 1996a). variation of the primary particle concentration (Chen & Doi,
At high solids concentrations, however, the state of parti- 1989; Folkersma, van Diemen, Laven, & Stein (1999)) or
cle aggregation can a5ect the rheology as has been encoun- interparticle forces (Kurzbeck, Kaschta, & MBunstedt, 1996)
tered in polymer slurry processing (Horn & Patterson, 1997), leads to a parallel shift of the 6ow curves in the low shear
metallic alloy solidi>cation (Barb@e, Perez, & Papoular, rate region. At higher shear rates, the 6ow curves level o5
2000), food processing (Weipert, Tscheuschner, & Wind- to a Newtonian plateau, which is the state of a deaggre-
hab, 1993) and even blood circulation (Schmid-SchBonbein, gated suspension where the aggregates can no longer break
Gallasch, Gosen, Volger, & Klose, 1976). Aggregates im- (Snabre & Mills, 1996). Rheological studies on aggregating
mobilize part of the matrix 6uid inside them and therefore suspensions mainly consider equilibrium, when the aggre-
gate structure is fully developed and in steady state. de Rooij,
Corresponding author. Tel.: +41-1-632-2510; fax: +41-1-632-1595. Potanin, van den Ende, and Mellema (1994), in particu-
E-mail address: pratsinis@ptl.mavt.ethz.ch (S. E. Pratsinis). lar, studied the transient rheological behavior of aggregating

0009-2509/03/$ - see front matter ? 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0009-2509(03)00133-7
2894 G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902

suspensions of polystyrene latex. They pre-sheared the sus- The >rst two terms of the right-hand side (RHS) describe
pensions at high shear rate for complete deaggregation and the birth of aggregate particles of section i by coagulation
then reduced sharply the shear rate resulting in a pronounced of smaller aggregates, the next two RHS terms account for
viscosity increase and later on in a shear rate-dependent the death of aggregates of section i by coagulation with
equilibrium viscosity. other aggregates, while the last two RHS terms describe
Here, the dynamic behavior of suspensions of polydis- death and birth, respectively, of aggregates in this section
perse particles is examined accounting for shear-driven by breakage.
coagulation and fragmentation by population balances. The i; j is the so-called collision kernel of particles or
Fluid and particle dynamics are coupled by replacing the aggregates. The collisions can be caused by the ever-present
solids concentration in material laws with an e5ective thermal motion (Brownian motion) or by shear deformation
solids volume fraction which includes the immobilized of the 6uid (Smoluchowski, 1917). Here, only particles of
6uid (Wolthers, van den Ende, Duits, & Mellema, 1996). at least micron size are considered where the in6uence
That way it overcomes the limitations of the earlier mod- of thermal motion is negligible compared to shear. This
els of de Rooij et al. (1993, 1994) that were con>ned to collision rate i; j of Sa5man and Turner (1956) for ho-
monodisperse particle suspensions and of Kramer and Clark mogeneous, isotropic turbulence with spherical particles
(1999) that assumed constant viscosity. The suspension smaller than the Kolmogorov microscale (for which par-
dynamics are investigated at various initial solids fractions ticle inertia can be neglected) was extended by Flesch,
and shear rates and the transient behavior of representative Spicer, and Pratsinis (1999) to account for the aggregate
macroscopic quantities (mass mean diameter, e5ective con- structure:
centration, viscosity) is studied. Special focus is placed on
the equilibrium state, particularly on a sensitivity analysis 1=Df 1=Df 3
i; j = 0:31 G vp (xi + xj ) ; (3)
of various model parameters on the 6ow curves, which are
compared to experimental data. where G is the so-called shear rate, i.e. the spatially aver-
aged velocity gradient that is constant and homogeneous.
The symbol G is utilized for both, laminar and turbulent
2. Theory shear. Possible e5ects of viscous retardation, van der Waals
attraction or particle repulsion are neglected, although at-
2.1. Sectional population balance model tractive and repulsive forces are usually not fully compen-
sating and the collision eOciency is depending on shear rate
The evolution of aggregate size distributions undergoing (cf. Vanni & Baldi, 2002). Here, completely destabilized
simultaneous coagulation and fragmentation is described suspensions are assumed where all collisions are successful.
by the population balance equation (Blatz & Tobolsky, This assumption of a collision eOciency of unity appears re-
1945). To solve this equation, Spicer and Pratsinis (1996a) alistic as the streamlines of 6ow only penetrate the outmost
extended the sectional model of Hounslow, Ryall, and layers of compact aggregates (Veerapaneni & Wiesner,
Marshall (1988) to account for fragmentation of aggregates. 1996). Therefore the primary particle density in the outer
They used a sectional spacing factor Vi =Vi1 = 2, where Vi regions of the aggregates must be suOciently large to ob-
is the mass-equivalent volume of an aggregate in section i, tain collisions of primary particles when the aggregates
i.e. the volume of a fully coalesced aggregate that consists start to overlap. The slight penetration of 6ow into the ag-
of xi = 2i1 spherical monodisperse primary particles of gregate reduces, however, hydrodynamic repulsion found
diameter dp : in squeezing 6ow between two approaching solid spheres
  (see e.g. Happel & Brenner, 1973 or Kim & Karrila, 1991),
Vi = d3i = xi vp = xi dp3 : (1)
6 6 which would have decreased the collision eOciency.
With this sectionalization, the population balance for the Aggregates are irregularly shaped with a void fraction or
number concentration Ni of aggregates in section i becomes inversely an aggregate density that depends on aggregate
(cf. Spicer & Pratsinis, 1996a): size. As they exhibit fractal-like properties they can be char-
acterized by a fractal dimension Df (Mandelbrot, 1983). In
i2
dNi  ji+1 1 2 sheared suspensions the fractal dimension is typically in the
= 2 i1; j Ni1 Nj + i1; i1 Ni1 range of 2.12.7. Here, the fractal dimension is a constant
dt 2
j=1
though it can depend on shear rate, shear history (Spicer et
i1 imax 1 al., 1998) and on aggregate size (Oles, 1992). The collision
 
Ni 2ji i; j Nj Ni i; j Nj diameter dc; i of an aggregate of section i is related to the
j=1 j=i
number of primary particles xi of diameter dp in it, by (cf.
Matsoukas & Friedlander, 1991)
imax

1=Df
Si Ni + i; j Sj Nj : (2) dc; i = dp xi : (4)
j=i
G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902 2895

The smaller the Df , the less compact the aggregate is and primary particles is distinguished from the relative aggregate
the larger is its collision diameter. size. Introducing Eqs. (6)(8) in Eq. (5), it becomes
The fragmentation kernel Si is the rate of particle breakage  q  3=Df
and is related to shear rate and aggregate size by (Pandya (tot ) G 1=3 dc; i
S i = kb vp : (9)
& Spielman, 1983)  dp

Si = A G q Vi1=3 ; (5) In dilute suspensions ( 0), where the viscosity is con-


stant, and for spherical particles (Df = 3), Eq. (9) nicely
where Si is proportional to the mass-equivalent diameter reduces to Eq. (5).
di Vi1=3 = (xi vp )1=3 . The fragmentation parameter A and The viscosity of suspensions with fractal-like aggregates
the exponent q are determined experimentally (Spicer & corresponds to that of suspensions of solid spheres with the
Pratsinis, 1996a). Since the aggregates are actually ruptured hydrodynamic diameter of the aggregates (Snabre & Mills,
by the forces acting upon them, the shear rate is replaced by 1996). There is a variety of correlations between particle
the external shear stress in the suspension. This is similar to volume fraction and suspension viscosity (Quemada, 1977).
the approach of Kramer and Clark (1999) who utilized the Here, a one-parameter viscosity function is chosen that de-
product of viscosity and strain rate which is proportional scribes polydisperse suspensions of hard spheres (Quemada,
to the shear stress in simple shear 6ows. The shear stress 1977) or of structural units like clusters or aggregates
is proportional to viscosity that depends on the e5ective (Quemada, 1998):
volume fraction of the aggregates. A characteristic shear 0
stress  is introduced to non-dimensionalize the shear stress (tot ) = : (10)
(1 tot =m )2
term. It is a measure of the aggregate strength: the larger
the  , the less susceptible to breakage are the particles at a In this equation, tot is the e5ective or total volume concen-
certain shear stress: tration of the aggregates:
 q
G imax

Si : (6) tot = Ni vc; i ; (11)

i=1
A comparison of Eqs. (5) and (6) shows that  can be
related to A by where vc; i = (=6) d3c; i is the collision volume of the ag-
 1=q gregates. The m is the maximum volume concentration at
kb which the viscosity diverges to in>nity, ranging from 0.58
 = ; (7)
A (Wolthers et al., 1996) to 0.69 (Lee, So, & Yang, 1999),
while m = 0:6 is used here as an average.
where kb = 1 cm1 s1 is used to match the dimensions of The fragment distribution function i; j in Eq. (2) gives
both sides of the equation. From the data of Oles (1992) the fraction of fragments produced in section i, when an
for an aqueous dilute polystyrene suspension, Spicer and aggregate of section j breaks. Here only binary breakage (or
Pratsinis (1996a) extracted A = 0:0047 cm1 sq1 and q = aggregate splitting) is considered, with
1:6. With the viscosity of water  = 103 Pa s, this results 
in a characteristic shear stress  = 0:0285 Pa. 2; j = i + 1;
As the disrupting shear forces are acting on the collision i; j = (12)
0 else:
diameter of the aggregate, the mass equivalent diameter di
is replaced by the collision diameter dc; i in Eq. (5). How- This splitting into two roughly equally sized fragments is
ever, aggregates with identical collision diameter but lower consistent with photographic observations by van de Ven
fractal dimension are more porous. Thus they contain less (1989) and simulations by Higashitani and Iimura (1998).
particle bonds per aggregate volume and their strength is Splitting into more than two daughter fragments can be con-
lower (Tang, Ma, & Shiu, 2001). So they are ruptured more sidered easily as in Spicer and Pratsinis (1996a).
easily by shear forces. Therefore, the linear dependency on The width of the aggregate size distribution can be de-
the aggregate collision diameter can be replaced by a power scribed by a number-based geometric standard deviation gn
law so the breakage rate increases with decreasing fractal or by a volume based geometric standard deviation gv . The
dimension, i.e. with increasing porosity of the aggregates. gn is given by (Hinds, 1999)
 3=Df  1=2
1=3 dc; i ni (ln di ln dgn )2
Si vp : (8) ln gn =  ; (13)
dp ni
This is consistent with Pandya and Spielman (1982) who where the geometric mean diameter is de>ned by
proposed S V m where the exponent m is determined 
experimentally (for their data: m = 0:33). As porosity is ni ln di
ln dgn =  : (14)
independent of the primary particle size, the volume of the ni
2896 G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902

The volume-based geometric standard deviation gv is 0.25


t*=0.5
obtained likewise by replacing ni by ni Vi in Eqs. (13) t*=0.7

relative mass fraction, (ni Vi)/0


and (14). 0.2
t*=0.8
t*=2.1
The sectional population balance equation (2) for imax =40 t*=21
and the above-de>ned kernels (3) and (9) and the corre-
0.15
lations (10)(12) is solved numerically using the VODPK
solver (Byrne, 1992; Netlib, 2001).
0.1

3. Results and discussion


0.05

3.1. Dynamics of aggregate size distributions


0
1 10 102 103 104
Table 1 shows the parameters employed here. The model
relative aggregate diameter, di/dp
was evaluated >rst for dilute suspensions ( 0) of spher-
ical particles with Df = 3 at the conditions of Oles (1992). Fig. 1. Particle size distribution at increasing dimensionless time
The evolution of the detailed size distribution and the mass t =t G 0 and Df =2:3 in a moderately concentrated suspension (primary
mean particle diameter as well as the attainment of equilib- particle concentration 0 = 0:01), starting with a completely deaggregated
rium size distribution were identical to that of Spicer and suspension.
Pratsinis (1996a).
Initially aggregate-free suspensions of particles of
dp = 2 m and 0 = 0:01 are considered. Fig. 1 shows the Therefore, they are disrupted by the increasing shear forces
evolution of the dimensionless aggregate mass distribution from the increasing viscosity. After about t = 2:1, di5er-
over di . Starting with the initial monodisperse primary ences in distribution curves can no longer be discerned (cf.
particles, >rst small aggregates of Df = 2:3 are formed t = 2:1 and 21), i.e. the suspension has reached steady state
(t = t G 0 = 0:5) with a low collision cross-section and or equilibrium between particle coagulation and fragmenta-
thus a slow growth. Particle growth accelerates when larger tion.
fractal-like aggregates are formed since the newly formed Transiently (e.g. around t = 0:7), the distribution curves
aggregates have a larger collision volume than the sum of undergo signi>cant changes in shape by the presence of a
the collision volumes of their predecessors, resulting in the signi>cant fraction of primary particles and the appearance
large-aggregate mode in the size distribution. This mode of the large aggregate mode. The observation of the very
can be observed after some time (t = 0:7) as a hump in large aggregates or superclusters is in agreement with simu-
the distribution. This hump grows rapidly and moves to lations of Silbert, Melrose, and Ball (1999), who calculated
smaller sizes (t = 0:8) because the large aggregates have a a transient particle gel network that is subsequently rup-
high porosity and thus a low strength (Tang et al., 2001). tured by the ongoing shear forces. Gustafsson, Nordenswan,
and Rosenholm (2001) explained also their experimental
results on the shear history dependence of aggregating sus-
Table 1 pensions using the concept of transient gelation.
Parameters used for simulation of >gures (unless otherwise noted) The shape of the steady-state or equilibrium particle
Parameter Figs. 16 and 9 Figs. 7 and 8 size distributions is independent of the initial solids vol-
ume fraction or shear rate (Fig. 2). When these are plotted
External
as a function of a normalized aggregate size, i.e. of the
dp 2:0 m 5:1 m
0 103 Pa s 24 103 Pa s mass-equivalent aggregate diameter di of the aggregates
divided by the mass mean diameter dmm , all distribution
 
0:01 (Figs: 13) curves collapse to one line. This indicates that steady-state
0 0.17
0:1 (Figs: 4; 9) aggregate size distributions are self-preserving with re-

spect to both, initial solids volume fraction 0 and shear
10 s1 (Figs: 1; 2; 4; 5) rate G. The latter is consistent with results of Spicer and
G Pratsinis (1996a) for dilute suspensions. Self-preservation
1 s1 (Fig: 3)
here means that particles attain a size distribution of which
Model the shape is invariant with respect to 0 and G even though
Df 2.3 2.3 the actual size changes. This holds however only as long
q 4 4 as the fraction of primary particles is negligible. At high
 
0:1 Pa (Figs: 1; 2; 46)
 16 Pa (Fig. 8) shear rates, when fragmentation dominates resulting in
0:01 Pa (Figs: 3; 9) the decomposition of aggregates into primary particles,
the self-preservation is no longer valid as has been shown
G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902 2897

0.2 with 0 =0:01 and G =1 s1 . These normalized steady-state


0=0.001 distributions broaden with increasing Df , so they are not
0=0.01
relative mass fraction, (ni Vi)/0

0=0.1
1
self-preserving with respect to Df . This behavior of suspen-
0.15 G=1s 1
G=10s 1
sions with shear-induced aggregation and fragmentation is
G=100s similar to that of suspensions undergoing Brownian coagu-
lation in the continuum regime (Vemury & Pratsinis, 1995).
0.1
The peak of the distribution curve is sharpened and shifted
to smaller sizes for decreasing Df . The increased fragility
0.05
of large aggregates at low Df may dominate over the en-
hanced collision cross-section narrowing the distributions.
In Fig. 3 the number-based geometric standard deviation
0 gn (squares) and the volume-based geometric standard de-
0.1 1 10 viation gv (triangles) are plotted as a function of the frac-
normalized aggregate diameter, di /dmm
tal dimension. The gn decreases from 2.2 for Df = 3 to
Fig. 2. Normalized particle size distribution for various initial solids
1:64 for Df = 2:1, whereas the dependence of gv on Df is
volume fractions at G = 10 s1 (lines) and various shear rates at 0 = 0:1 less pronounced (only a decrease from 1.8 to 1.62 for the
(symbols): ratio of particle mass in one section to total particle mass, same Df ). At Df = 3, the size distribution and both geo-
as function of the mass-equivalent diameter of the aggregates divided by metric standard deviations closely coincide with the values
the mass mean diameter for Df = 2:3. of Spicer and Pratsinis (1996a) for binary fragmentation
as it is used here. The decrease in gn with lower Df re6ects
2.5 the narrowing of the distributions in the inset of Fig. 3.
Fig. 4 shows the dmm evolution as a function of dimen-
sionless time t = t G 0 for various initial volume frac-
numberbased gn tions 0 (a) and various shear rates (b). With this scaled
volumebased gv time, the initial aggregation kinetics coincide regardless of
0 (Fig. 4a) or G (Fig. 4b). The scaling of the dynamics of
shear-induced 6occulation of suspensions with t = t G 0
geometric standard deviation g

2 has been found in dilute systems also (Oles, 1992; Serra,


Colomer, & Casamitjana, 1997). The transient formation of
large particles (Fig. 1) results in a short overshoot of the
dmm (Fig. 4a and b) at high initial volume fractions (e.g.
0 0:01) and low shear rates (e.g. G 6 10 s1 ). Later
Df=2.3 on the dmm reaches asymptotically an equilibrium diame-
Df=2.5
ter depending on the initial primary particle concentration
relative mass fraction, ni vi/0

1.5 Df=3.0
and on the shear rate as has been shown amply with dilute
suspensions (Oles, 1992; Serra et al., 1997). The fractal ag-
gregates tend to occupy as much space as possible, i.e. to
establish a space->lling particle network (Folkersma, van
Diemen, Laven, van der Plas, & Stein, 1998; Folkersma et
al., 1999; Dickinson, 2000). This can be achieved, however,
1
only when the suspension is at rest. Under such conditions,
0.1 1
normalized aggregate diameter, di/dmm
10
the cluster size scales with (Varadan & Solomon, 2001):

2 2.2 2.4 2.6 2.8 3 dc; i 1=(D 3)


0 f : (15)
fractal dimension Df dp
Fig. 3. Geometric standard deviations (number based: gn , squares; volume
based: gv , triangles) of the steady-state particle size distributions as In sheared concentrated suspensions, the viscosity increases
a function of fractal dimension Df . The inset shows the normalized at higher tot enhancing fragmentation. Therefore, the ag-
steady-state particle size distribution by coagulation and fragmentation of gregates become even smaller than at rest and the depen-
concentrated suspensions of 0 = 0:01 0.2 (here: 0 = 0:01) for Df = 2:3, dence on 0 is more pronounced. This result contrasts that
2.5 and 3.
of dilute suspensions where increasing particle concentra-
tions give larger steady-state size distributions (Spicer &
by Spicer, Pratsinis, Trennepohl, and Meesters (1996) for Pratsinis, 1996b). At high shear rates (e.g. G = 1000 s1 ),
dilute suspensions undergoing shear-induced 6occulation. no overshoot in average aggregate size can be observed as
The inset of Fig. 3 shows the equilibrium or steady-state the high shear stresses cause high fragmentation rates even
particle size distributions normalized by dmm for Df =2:33, of small aggregates.
2898 G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902

3
10 10
0=0.001
rel. mass mean diameter, dmm(t)/dp

rel. mass mean diameter, dmm(t)/d0


G=0.1s1
1
G=10s
2
10
0=0.01

3 1
10 G=10 s

0=0.1

1 1
0.01 0.1 1 10 0.01 0.1 1 10
(a) dimensionless time, t G 0 (b) dimensionless time, t G 0
Fig. 4. Evolution of the mass mean diameter as a function of time for various (a) initial solids volume fractions (for G = 10 s1 ) and (b) shear rates
(for 0 = 0:1).

103 The very rapid dynamics of viscosity evolution predicted by


the model can be attributed to the model assumptions: ide-
ally homogeneous shear >eld, constant fractal dimension,
and completely eOcient aggregate collisions.
relative viscosity, (t)/0

The correlation between viscosity and applied shear is ex-


102 0=0.1
amined and the in6uence of various model parameters on
the viscosity is elucidated. A common presentation of the
0=0.01
rheological properties is the so-called 6ow curve where
the suspension viscosity (or the relative viscosity) is plot-
0=0.001
10 ted against the applied shear rate. To study the in6uence
of the initial volume fraction on the 6ow curves, 0 was
varied over some range typical for moderately concentrated
suspensions (Fig. 6). The simulated 6ow curves are straight
lines in the loglog plot in the range of low shear rates (here
1
0.01 0.1 1 10 G = 0 G= 6 1) and can be approximated by power laws.
dimensionless time, t G 0 At high G the viscosity approaches the value of completely
deaggregated suspensions. A change of the primary particle
Fig. 5. Evolution of the relative viscosity for the same parameters as in concentration causes a parallel shift of the 6ow curves at low
Fig. 4a. shear rates. This parallel shift of the 6ow curves is consistent
with direct numerical simulations of aggregating concen-
3.2. Viscosity and =ow curves trated suspensions (Chen & Doi, 1989; Doi & Chen, 1989)
and with experimental data of Folkersma et al. (1999) for
The appearance of the large aggregates leads to an in- aqueous polystyrene latex suspensions (dp = 2:0 m) with
crease in viscosity (Fig. 5). This quantity does not exhibit 0:5 M NaCl [cf. inset in Fig. 6], and of Zaman, Moudgil,
the overshoot behavior of the mass mean diameter, as only Fricke, and El-Shall (1996) for aqueous silica suspensions
a few of the large particles are formed, but bends smoothly (dp = 1 m) with 0:03 M NaNO3 . However, the extend of
towards an equilibrium value. By coagulation, the fractal ag- the parallel shift is underpredicted by the model: The ex-
gregates are >lling the available space gradually until the si- perimental data show an increase of viscosity by more than
multaneously increasing viscosity is leading to shear stresses one order of magnitude when the solids concentration in-
that counteract growth by fragmentation. Experimental ob- creases from 2% to 10% (de Rooij, Potanin, van den Ende,
servations of de Rooij et al. (1994) on the viscosity evo- & Mellema, 1993) or from 3% to 21% (Folkersma et al.,
lution of aggregating suspensions show moderate slopes of 1999), while the model predicts only an increase by a factor
viscosity evolution after a step from high to low shear rate. of 3 when the solids concentration is increased from 1% to
G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902 2899

104
3
10
Df=2.1

0 = 0.216 Df=2.3
Df=2.5

relative viscosity /0
Df=2.7

0 = 0.031 102 Df=2.9
3
10
relative viscosity /0

Folkersma et al .(1999) 10

2
10
1
102 101 1 10 102
dimensionless shear rate G*=(G 0/*)

Fig. 8. Negligible in6uence of the fractal dimension in the typical range


10 0 =0.2 Df = 2:12.7 on the 6ow curves for the system of Kurzbeck et al. (1996).
Parameters:  = 16 Pa, otherwise as in Fig. 7.
0 =0.1
0 =0.01
0 =0.001 of Kurzbeck et al. (1996) of a suspension of a hydrophobic
1 molten wax at 50 C >lled with a volume fraction 0 =0:17 of
103 102 101 1 10 chromium oxide pigment particles (dp =5:1 m) at di5erent
dimensionless shear rate, G*= (G 0/*)
moisture content (points in Fig. 7). A variation of  in the
simulations leads to a parallel shift of the 6ow curves in the
Fig. 6. Steady-state 6ow curves for various initial solids fractions or low shear rate range (Fig. 7), when the relative viscosity is
primary particle concentrations. Inset: experimental data adapted from plotted over the shear rate. The moisture enhances the aggre-
Folkersma et al. (1999): 6ow curves for primary particle concentrations gation of the particles. So increasing the moisture shifts the
from 0 = 0:031 () to 0.216 (). 6ow curves in the low shear rate range (G 300 s1 ) par-
allel to higher viscosity. The  is determined by matching
103 the model predictions and experimental results. At higher
*=16 Pa
*=8 Pa shear rates, the experimentally measured e5ective viscosity
13% H2O decreases and approaches a value (at G 1000 s1 ) above
6% H2O the viscosity of a completely deaggregated suspension. The
relative viscosity /0

2 dry
10 highest applied shear stresses were apparently not suOcient
to disintegrate the aggregates completely down to primary
particles (Kurzbeck et al., 1996).
In the range of values that is typical for sheared sus-
10
pensions (Df = 2:12.7), the fractal dimension Df has
negligible in6uence on the slope of the 6ow curves of con-
centrated suspensions (Fig. 8). Only at fractal dimensions
1 close to Df = 3 the viscosity is reduced signi>cantly. As
10
2
10
3
10
4
Df 3, the aggregates become more compact and immobi-
1
shear rate G/s lize less matrix liquid, and eventually none at all for Df =3 .
The insensitivity of the 6ow curves to the fractal dimension
Fig. 7. Comparison between model predictions for the e5ect of varied
particle interaction forces on 6ow curves (lines) and experimental results in the typical range for shear-induced aggregation (Df =
(dots) of Kurzbeck et al. (1996) of a pigment-wax suspension at 50 C 2:12.7) indicates that the apparent viscosity is not
at various moisture contents. signi>cantly a5ected by the selection of Df . Possible
shear-induced restructuring, resulting in a compaction of
10%. A possible reason for this di5erence may be in the frag- aggregates and an increase of Df over time, can be ne-
mentation kernel and that the coagulation kernel is derived glected for particles larger than 1 m (Spicer et al., 1998;
for binary collisions without any in6uence of surrounding Selomulya, Bushell, Amal, & Waite, 2002).
particles, which can lead to multiple collisions, enhancing Fig. 9 shows the e5ect of the fragmentation exponent q
coagulation and thus the viscosity. on the e5ective viscosity as a function of G for q = 2
Fig. 7 shows the e5ective suspension viscosity as a func- 6. The q changes the slope of the 6ow curve in contrast to
tion of shear rate for  equal to 8 Pa (dotted line) and the parallel shift of R by the previous parameters (0 , 
16 Pa (solid line) by the present model as well as the data or Df ). This property of q facilitates the estimation from
2900 G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902

104 uid within the aggregates. This model describes the evolu-
q=2
q=3
tion of aggregate size distribution and the associated change
q=4 of the suspension viscosity. The equilibrium aggregate size
103 q=6 distributions are self-preserving with respect to the initial
relative viscosity /0

solids volume fraction and shear rate.


2
Flow curves of suspensions of polydisperse particles (up
10
to 20% by solids volume) are calculated that are consis-
tent with experimental data in the literature. The correla-
10
tion between viscosity and shear rate can be described by a
power law. Two material parameters are in6uencing signif-
icantly the predicted 6ow curves and need to be determined
1 by comparison with experimental results: the fragmentation
4 3 2 1 2
10 10 10 10 1 10 10 exponent q and the characteristic shear stress  . The expo-
dimensionless shear rate G*=(G 0/*)
nent q de>nes the sensitivity of the aggregates on the shear
Fig. 9. In6uence of the fragmentation exponent q on the slope of the
rate and is the only parameter with signi>cant in6uence on
simulated 6ow curves. the slope of the 6ow curves. The fractal dimension has little
in6uence on the 6ow curves.

Table 2
Analysis of 6ow curves from literature

Author Negative
Notation
6ow curve Calculated
exponent q d diameter (cm)
Experimental Df fractal dimension
de Rooij et al. (1993): Fig. 3 0.64 0.84 2.8 6.4 G shear rate (s1 )
Folkersma et al. (1999): Fig. 2 0.69 0.85 3.2 6.5 kb dimension correction factor (1 cm1 s1 )
Kosmulski, Gustafsson, and Rosenholm 0.80 5.0
N aggregate number concentration (cm3 )
(1999): Fig. 3
Kurzbeck et al. (1996): Fig. 6 0.71 0.79 3.4 4.7 q fragmentation exponent
Verduin, de Gans, and Dhont (1996): Table 1 0.43 0.77 1.75 4.3 S fragmentation kernel (s1 )
Zaman et al. (1996): Fig. 1 0.55 0.84 2.2 6.4 t time (s)
v aggregate or particle volume (cm3 )
Simulated
V mass equivalent volume (of fully coalesced sphere)
Chen and Doi (1989): Fig. 14 0.56 2.3
Chen and Doi (1999): Fig. 11 0.67 3 (cm3 )
Silbert, Melrose, and Ball (1997): Fig. 5 0.75 4 x number of primary particles per aggregate
Silbert et al. (1999) (in text) 0.84 6.25
Greek letters

experimental data by regression of  at small G . There the  collision kernel (cm3 s1 )


6ow curves can be approximated by the following power  fragment distribution function
law correlation:  dynamic viscosity (Pa s)
0 dynamic viscosity of matrix liquid (Pa s)
R G (1q)=q : (16) g geometric standard deviation of diameter (cm)
Table 2 summarizes the analysis of 6ow curves in the  characteristic shear stress (Pa)
literature. As the negative 6ow curve exponents scatter  aggregate volume concentration
signi>cantly (0.43 0.85), the extracted q also vary over
some range (q = 1:75 6.5). Similar values (1.9 5.6) are Superscripts and subscripts
reported by Serra and Casamitjana (1998) for dilute sus-
pensions. c collision
i; j section number
m maximum (concentration)
4. Summary and conclusions mm mass mean
p primary particle
A sectional population balance model for shear-induced R relative
coagulation fragmentation of concentrated suspensions is tot total
developed accounting for the immobilization of matrix liq- 0 initial, reference
G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902 2901

Acknowledgements Hinds, W. C. (1999). Aerosol Technology (2nd ed.). New York: Wiley.
Horn, J. A., & Patterson, B. R. (1997). Thermally induced reversible
coagulation in ceramic powder polymer liquid suspensions. Journal of
This research was supported partially by the Deutsche
the American Ceramic Society, 80(7), 17891797.
Forschungsgemeinschaft (DFG, Grant No. Bu485/21-1) and Hounslow, M. J., Ryall, R. L., & Marshall, V. R. (1988). A discretized
the Swiss Commission for Technology and Innovation (KTI, population balance for nucleation, growth, and aggregation. A.I.Ch.E.
TOP Nano21 Project No. 5773.2). Journal, 34(11), 18211832.
Ives, K. J. (1978). The scientiEc basis of =occulation. Alphen aan den
Rijn, Netherlands: Sijtho5 & Noordho5.
References Kim, S., & Karrila, S. J. (1991). Microhydrodynamics. London:
Butterworth-Heinemann.
Barb@e, J. C., Perez, M., & Papoular, M. (2000). Microstructure and Kosmulski, M., Gustafsson, J., & Rosenholm, J. B. (1999). Correlation
viscosity of semi-solid mixtures. Journal of Physics: Condensed between the zeta potential and rheological properties of anatase
Matter, 12(12), 25672577. dispersions. Journal of Colloid and Interface Science, 209(1),
Blatz, P. J., & Tobolsky, A. V. (1945). Note on the kinetics of 200206.
systems manifesting simultaneous polymerisationdepolymerisation Kramer, T. A., & Clark, M. M. (1999). Incorporation of aggregate breakup
phenomena. Journal of Physical Chemistry, 49, 7780. in the simulation of orthokinetic coagulation. Journal of Colloid and
Byrne, G. D. (1992). Pragmatic experiments with Krylov methods Interface Science, 216(1), 116126.
in the sti5 ODE setting. In J. Cash & I. Gladwell (Eds.), Kurzbeck, S., Kaschta, J., & MBunstedt, H. (1996). Rheological behaviour
Computational ordinary di@erential equations (pp. 323356). Oxford: of a >lled wax system. Rheologica Acta, 35, 446457.
Oxford University Press. Lee, J.-D., So, J.-H., & Yang, S.-M. (1999). Rheological behavior and
Chen, D., & Doi, M. (1989). Simulation of aggregating colloids in shear stability of concentrated silica suspensions. Journal of Rheology,
6ow. II. Journal of Chemical Physics, 91(4), 26562663. 43(5), 11171140.
Chen, D., & Doi, M. (1999). Microstructure and viscosity of aggregating Mandelbrot, B. B. (1983). The fractal geometry of nature. New York:
colloids under strong shearing force. Journal of Colloid and Interface Freeman.
Science, 212(2), 286292. Matsoukas, T., & Friedlander, S. K. (1991). Dynamics of aerosol
Coulaloglou, C. A., & Tavlarides, L. L. (1977). Description of interaction agglomerate formation. Journal of Colloid and Interface Science,
processes in agitated liquidliquid dispersions. Chemical Engineering 146(2), 495506.
Science, 32(11), 12891297. Netlib, (2001). Netlib Repository at University of Tennessee and Oak
Danov, K. D., Ivanov, I. B., Gurkov, T. D., & Borwankar, R. P. Ridge National Laboratory. URL: http://www.netlib.org/
(1994). Kinetic-model for the simultaneous processes of 6occulation Oles, V. (1992). Shear-induced aggregation and breakup of polystyrene
and coalescence in emulsion systems. Journal of Colloid and Interface latex particles. Journal of Colloid and Interface Science, 154(2),
Science, 167(1), 817. 351358.
de Rooij, R., Potanin, A. A., van den Ende, D., & Mellema, J. Pandya, J. D., & Spielman, L. A. (1982). Floc breakage in agitated
(1993). Steady shear viscosity of weakly aggregating polystyrene latex suspensions: Theory and data processing strategy. Journal of Colloid
dispersions. Journal of Chemical Physics, 99(11), 92139223. and Interface Science, 90(2), 517531.
de Rooij, R., Potanin, A. A., van den Ende, D., & Mellema, J. (1994). Pandya, J. D., & Spielman, L. A. (1983). Floc breakage in agitated
Transient shear viscosity of weakly aggregating polystyrene latex suspensions: E5ect of agitation rate. Chemical Engineering Science,
dispersions. Journal of Chemical Physics, 100(7), 53535360. 38(12), 19831992.
Dickinson, E. (2000). Structure and rheology of simulated gels formed Quemada, D. (1977). Rheology of concentrated disperse systems. I.
from aggregated colloidal particles. Journal of Colloid and Interface Minimum energy dissipation principle and viscosityconcentration
Science, 225(1), 215. relationship. Rheologica Acta, 16(1), 8294.
Doi, M., & Chen, D. (1989). Simulation of aggregating colloids in shear Quemada, D. (1998). Rheological modelling of complex 6uids. I. The
6ow. Journal of Chemical Physics, 90(10), 52715279. concept of e5ective volume fraction revisited. European Physical
Flesch, R., Spicer, P. T., & Pratsinis, S. E. (1999). Laminar and turbulent JournalApplied Physics, 1(1), 119127.
shear-induced 6occulation of fractal aggregates. A.I.Ch.E. Journal, Sa5man, P., & Turner, J. (1956). On the collision of drops in turbulent
45(5), 11141124. clouds. Journal of Fluid Mechanics, 1(1), 1630.
Folkersma, R., van Diemen, A. J. G., Laven, J., van der Plas, G. A. J., & Schmid-SchBonbein, H., Gallasch, G., Gosen, J. V., Volger, E., & Klose,
Stein, H. N. (1998). Visualization of the breakdown of dilute particle H. J. (1976). Redcell aggregation in blood 6ow. 2. E5ect on apparent
gels during steady-shear. Rheologica Acta, 37, 549555. viscosity of blood. Klinische Wochenschrift, 54(4), 159167.
Folkersma, R., van Diemen, A. J. G., Laven, J., & Stein, H. N. (1999). Selomulya, C., Bushell, G., Amal, R., & Waite, T. D. (2002). Aggregation
Steady shear rheology of dilute polystyrene particle gels. Rheologica mechanisms of latex of di5erent particle sizes in a controlled shear
Acta, 38, 257267. environment. Langmuir, 18, 19741984.
Graham, A. L., Steele, R. D., & Bird, R. B. (1984). Particle clusters Serra, T., & Casamitjana, X. (1998). E5ect of the shear and volume
in concentrated suspensions. 3. Prediction of suspension viscosity. fraction on the aggregation and breakup of particles. A.I.Ch.E. Journal,
Industrial & Engineering Chemistry Fundamentals, 23(4), 420425. 44(8), 17241730.
Gustafsson, J., Nordenswan, E., & Rosenholm, J. B. (2001). Shear-induced Serra, T., Colomer, J., & Casamitjana, X. (1997). Aggregation and
aggregation of anatase dispersions investigated by oscillation and low breakup of particles in a shear 6ow. Journal of Colloid and Interface
shear rate viscometry. Journal of Colloid and Interface Science, Science, 187, 466473.
242(1), 8289. Silbert, L. E., Melrose, J. R., & Ball, R. C. (1997). Colloidal micro-
Happel, J., & Brenner, H. (1973). Low Reynolds number hydrodynamics. dynamics: Pair-drag simulations of model-concentrated aggregated
Leyden, Netherlands: Noordho5. systems. Physical Review E, 56(6), 70677077.
Harris, S. J., & Maricq, M. M. (2002). The role of fragmentation in Silbert, L. E., Melrose, J. R., & Ball, R. C. (1999). The rheology
de>ning the signature size distribution of diesel soot. Journal of and microstructure of concentrated, aggregated colloids. Journal of
Aerosol Science, 33(6), 935942. Rheology, 43(3), 673700.
Higashitani, K., & Iimura, K. (1998). Two-dimensional simulation of the Smoluchowski, M. von (1917). Versuch einer mathematischen Theorie der
breakup process of aggregates in shear and elongational 6ows. Journal Koagulationskinetik kolloider LBosungen. Zeitschrift f!ur Physikalische
of Colloid and Interface Science, 204, 320327. Chemie, 92, 129168.
2902 G. Barthelmes et al. / Chemical Engineering Science 58 (2003) 2893 2902

Snabre, P., & Mills, P. (1996). I. Rheology of weakly 6occulated Vanni, M., & Baldi, G. (2002). Coagulation eOciency of colloidal
suspensions of rigid particles. Journal de Physique III, 6(12), 1811 particles in shear 6ow. Advances in Colloid and Interface Science,
1834. 97(13), 151177.
Spicer, P. T., & Pratsinis, S. E. (1996a). Coagulation and fragmentation: Varadan, P., & Solomon, M. J. (2001). Shear-induced microstructural
Universal steady-state particle size distribution. A.I.Ch.E. Journal, evolution of a thermoreversible colloidal gel. Langmuir, 17(10),
42(6), 16121620. 29182929.
Spicer, P. T., & Pratsinis, S. E. (1996b). Shear-induced 6occulation: The Veerapaneni, S., & Wiesner, M. R. (1996). Hydrodynamics of fractal
evolution of 6oc structure and the shape of the size distribution at aggregates with radially varying permeability. Journal of Colloid and
steady state. Water Research, 30(5), 10491056. Interface Science, 177, 4557.
Spicer, P. T., Pratsinis, S. E., Raper, J., Amal, R., Bushell, G., & Meesters, Vemury, S., & Pratsinis, S. E. (1995). Self-preserving size distributions
G. (1998). E5ect of shear schedule on particle size, density, and of agglomerates. Journal of Aerosol Science, 26(4), 175185.
structure during 6occulation in stirred tanks. Powder Technology, 97, Verduin, H., de Gans, B. J., & Dhont, J. K. G. (1996). Shear induced
2634. structural changes in a gel-forming suspension studied by light
Spicer, P. T., Pratsinis, S. E., Trennepohl, M. E., & Meesters, G. H. M. scattering and rheology. Langmuir, 12(12), 29472955.
(1996). Coagulation and fragmentation: The variation of shear rate and Weipert, D., Tscheuschner, H. D., & Windhab, E. (1993). Rheologie der
the time lag for attainment of steady state. Industrial & Engineering Lebensmittel. Hamburg, Germany: Behrs.
Chemistry Research, 35(9), 30743080. Wolthers, W., van den Ende, D., Duits, M. H. G., & Mellema, J. (1996).
Tang, S., Ma, Y., & Shiu, C. (2001). Modelling the mechanical strength The viscosity and sedimentation of aggregating colloidal dispersions
of fractal aggregates. Colloids and Surfaces A, 180(12), 716. in a Couette 6ow. Journal of Rheology, 40(1), 5567.
Tsutsumi, A., Yoshida, K., Yui, M., Kanamori, S., & Shibata, K. Zaman, A. A., Moudgil, B. M., Fricke, A. L., & El-Shall, H. (1996).
(1994). Shear viscosity behavior of 6occulated suspensions. Powder Rheological behavior of highly concentrated aqueous silica suspensions
Technology, 78(2), 165172. in the presence of sodium nitrate and polyethylene oxide. Journal of
van de Ven, T. G. M. (1989). Colloidal hydrodynamics. London: Rheology, 40(6), 11911210.
Academic Press.

You might also like