Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Floc Structure and Growth Kinetics for Rapid Shear Coagulation

of Polystyrene Colloids

FRANCISCO E. TORRES, WILLIAM B. RUSSEL, 1


AND WILLIAM R. SCHOWALTER*
Department of Chemical Engineering, Princeton University, Princeton, New Jersey 08544-5263 and *Department of
Chemical Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801
Received July 5, 1990; accepted September 13, 1990
Colloid coagulation in a flowing suspension is governed by floc structure and by the colloidal and
hydrodynamic interactions between the aggregates. We have investigated both the internal structure and
growth rate of rigid flocs formed by rapid coagulation in a linear shear flow, examining structure by
static light scattering and measuring sizes by dynamic light scattering. Comparison of the results for
sheared suspensions with Brownian coagulation data reveals a similar structure for the two modes; in
both cases the flocs exhibit characteristics ofa fractal with dimension d = 1.8 _+0.1, and for both modes
the average number of nearest neighbors is approximately the same. In contrast, the growth kinetics are
inherently different, as expected. To model growth in a shear flow, we treat a porous floc as a body with
a hydrodynamic radius which is smaller than the capture radius corresponding to floc-floc contact, and
we also find it necessary to invoke a criterion for the maximum size a floc can obtain at a given shear
rate. Necessary structural details are provided by the static light scattering results. We compare kinetics
predicted by this model with data at several shear rates and find qualitative agreement. © 1991Academic
Press, Inc.

1. INTRODUCTION structure and dynamic light scattering to mea-


At the appropriate conditions, particles in sure floc size.
a colloidal suspension undergoing shear co- Much of the previous fundamental work on
agulate to form flocs containing a large num- colloid coagulation in a flow field has dealt
ber of the individual particles connected into with two solid spheres or, in some studies, two
complex structures. As two such flocs ap- nonspherical particles, but relatively few in-
proach, they experience hydrodynamic and vestigators have ventured beyond the doublet
colloidal interactions, and these interactions formation step. Basic questions about the hy-
both govern the structure of any larger floc drodynamic and colloidal forces experienced
formed and depend on the structure of the by porous flocs remain unanswered, and al-
coagulating flocs. Therefore, there is a basic though some data on floc structure for shear
interdependence between the structure and the coagulation have been reported (1-3), con-
hydrodynamic and colloidal forces. To help clusions about general behavior are not avail-
provide a better understanding of this process, able. However, many factors that influence the
we present an experimental study of the in- coagulation process have been identified (see,
ternal structure and coagulation kinetics for e.g., (2, 4-7 )). For example, the interparticle
forces, particle size, shear rate, and particle
rigid flocs formed by rapid coagulation in a
linear shear flow. This study is limited to dilute surface characteristics can affect whether co-
agulated particles rotate freely relative to one
suspensions, and it relies on static light scat-
another, with rotation observed for bonding
tering to extract information about the internal
within a secondary minimum in a DLVO-type
To whom correspondence should be addressed. potential and rigid bonds (as desired in our

554
0021-9797/91 $3.00
Copyright © 1991 by Academic Press, Inc.
All rights of reproduction in any form reserved. Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 199-1
SHEAR-INDUCED COLLOID COAGULATION 555

experiments) observed for primary minimum coagulation rate within a random suspension
coagulation (8, 9). Thus, rearrangement and of uniform spheres:
breakup can be influenced by these factors, dnl
and studies by Lindsay, et al. (10) and Sonntag = - K i l n 2. [1.1]
dt
and Russel (11 ) provide some experimental
results. The term nl is the number concentration of
Regarding the current understanding of floc singlets, and Kll is the rate constant.
hydrodynamics, Wiltzius and co-worker ( 12, In his pioneering study of colloid coagula-
13) report experimental work on the hydro- tion, Smoluchowski (18) derived equations for
dynamic behavior of fractal flocs, and Chen Kll for both flow- and Brownian motion-in-
et aI. (14) compare these data with computer duced coagulation. In the former case, he as-
simulations reported in a different context sumed particles follow streamlines undis-
earlier ( 15 ). Also, Pusey et al. (16) discuss the turbed by other particles except at contact,
effect of floc size polydispersity on the mea- giving
surements reported by Wiltzius. Taken to- KII = K0 = X ( a l q- a2)3q/ [1.2]
gether, these papers and comments appear to
bring the experiments and simulations into for linear flows with a strain rate ,) and par-
reasonable accord, but the assumptions in- ticles of radii al and a2. The constant x de-
voked need further study. The above authors pends on the type of flow and equals 4 for
all report their results in terms of the ratio of simple shear flows, the type of interest here.
the floc hydrodynamic radius RH to the radius The Smoluchowski result K0 often provides a
of gyration Re, where RH is defined as the useful estimate of the actual coagulation rate
radius of a sphere having the same Stokes drag K, and we adopt the convection of expressing
as the given floc, and they find/3 = R H / R a K in terms of the stability ratio
has a value between 0.72 and approximately
K0
1.0, independent of size for large flocs. In an- W K " [1.3]
other investigation, Rogak and Flagan (17)
calculate approximate values of/3 for various Later investigators included hydrodynamic
geometries, providing some insight into the interactions and colloidal forces in predicting
effect of structure on drag. However, a detailed W for particles and studied a range of Peclet
understanding of the hydrodynamics of two numbers
interacting flocs is lacking, even for flocs
67r#a32/
treated as equivalent spheres. Consequently, Pe- kBT ' [1.4]
the present analysis relies on simple descrip-
tions of the hydrodynamics which capture the providing information on the interaction of
essential features expected for two porous Brownian motion and shear flow. In particu-
bodies. lar, for purely flow-induced coagulation van
At the doublet formation level, coagulation de Ven and Mason (8) and Zeichner and
rates due to Brownian motion, flow, and cer- Schowalter (19) calculated trajectories of two
tain combinations thereof have been calcu- spheres with different interaction potentials to
lated by utilizing known hydrodynamic func- obtain collision cross sections, and Feke and
tions and colloidal force equations, and to- Schowalter (20) extended the infinite Peclet
gether with experimental data these results number results to large but finite values
provide a basis for predicting two sphere be- through a regular perturbation expansion of
havior in detail (of. (4, 5, 7 )). Because these the trajectory equations in Pe-l. Interestingly,
results form a framework for our studies, we the latter investigators found rather large ef-
briefly review relevant points, beginning with fects for small amounts of Brownian motion.
the differential equation governing the initial These analyses centered on the equation for
Journal of Colloid and Inter-face Science, Vol. 142, No. 2, March 15, 1991
556 TORRES, RUSSEL, A N D S C H O W A L T E R

the relative velocity of two spheres in a linear ligible interparticle repulsion), and to obtain
flow field, which can be written as (20) a model system for which flocs undergo little
rearrangement, we study particles that appear
v=ft×r+E-r
to form strong and rigid primary-minimum
bonds. Examining the structure and kinetics
+ A(r,X)7~+B(r,X ) I- .E.r
together, we demonstrate that the kinetics can
be described by a model which combines three
C(r, X) features: the structural information inferred
+--Fc-D(r,X)'Vlnp2. [1.5]
6rqzal from the first set of experiments, a simple de-
The functions A(r, X), B(r, X), and C(r, X) piction of floc hydrodynamics, and an ap-
are hydrodynamic functions whose detailed proximate criterion for the maximum floc size
tabulation made the trajectory approach pos- attainable at a given shear rate. Insofar as our
sible. Other variables are the center-to-center structure results lead to a reasonable kinetics
particle separation r, the ratio of the particle model, we conclude that the proposed de-
radii X, the vorticity vector ft and rate of strain scription of the aggregation process is consis-
tensor E of the bulk flow, the colloidal force tent.
Fo, the tensor D characterizing mutual diffu- The next section describes the experiments
sion of two spheres, and the pair distribution performed to examine structure and coagu-
function P2 whose gradient drives a Brownian lation rates. A discussion of the static light
velocity (21 ). Note that the Smoluchowski re- scattering results follows in Section 3, after
sult given by [ 1.2 ] follows from [ 1.5 ] by setting which we discuss the rate data obtained by
A(r, X), B(r, X), Fc, and Vp2 equal to dynamic light scattering in Section 4. The
zero (19). calculations of the floc growth process are
Thus, the two-particle studies mentioned presented and compared with our data in
above demonstrate the role of hydrodynamic Section 5.
and colloidal forces in coagulation, and to an-
alyze floc coagulation the importance of these 2. E X P E R I M E N T A L DESCRIPTION

interactions for aggregates must likewise be 2.1. Particle Synthesis and Characterization
assessed. Floc interactions, however, depend
on the floc structure, an important factor The particles used in our experiments were
which the two sphere results do not address. monodisperse polystyrene latexes purchased
On the other hand, previous investigators of from Interfacial Dynamics Corporation (batch
rapid Brownian coagulation ( e.g., (22-26)) 7-108-7), who synthesized the particles by
studied floc structure and detected fractal-like surfactant-free emulsion polymerization. The
configurations with a fractal dimension d manufacturer reports a surface charge density
= 1.8 + 0.1 for nonrearranging flocs, in good of 1.16 ~tC/cm2 for these IDC particles, equiv-
agreement with simulations (27, 28 ). alent to 1378 A2 per charge group, with the
To learn more about shear-induced aggre- charge arising predominantly from sulfate
gation of colloidal flocs, we have conducted groups generated by the initiator for the po-
an experimental study of the internal structure lymerization. The reported mean diameter,
arising from both shear- and Brownian mo- determined by transmission electron micros-
tion-induced rapid coagulation, and we have copy, is 0.098 #m, with a coefficient o f vari-
measured the growth rate for the former, ap- ation equal to 10.9%. Our dynamic light scat-
plying new results for dynamic light scattering tering measurements give an intensity aver-
from flocs. These experiments are the subject aged diameter equal to 0.108 gm in a 10-3M
of this paper. The colloidal suspensions con- HC1 glycerol/water solution (36.0 wt% glyc-
tained high NaCI concentrations to realize the erol), in reasonable agreement with the mi-
rapid coagulation conditions (i.e., with neg- croscopy results; the HC1 concentration was
Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
SHEAR-INDUCED COLLOID COAGULATION 557

chosen to minimize diffuse double layer effects TABLEI

without inducing coagulation.


Properties of test suspension

2.2. Flocculation Studies 36.00 _+ 0.05 wt% glycerol


5.20 -+ 0.01 wt% NaC1
Using these particles, both Brownian and 58.80 ± 0.12 wt% H20, deionized
shear coagulation experiments were carried 2.5 X 10-4 M H C I
q~ = 1.00 X 10 -6 _+ 1%, I D C particles
out at ~b = 1.00 X 10 -6 in 36.0 wt% glycerol
a = 0.05 # m
solutions, the glycerol added to raise the vis- # = 3.12 _+ 0.06 cP at 25°C
cosity. These conditions slow Brownian co- = 2.79 _+ 0.05 cS at 2 5 ° C
o = 1.12 g / c m 3 at 22.8°C
agulation to levels that permit shear coagula-
no = 1.3872 at 2 5 ° C (Xo = 538 nm)
tion and sampling in reasonable times. At
these conditions and T = 25°C, the charac- Properties of precursor solutions A and B ~
teristic time for doublet formation by Brown-
Solution A Solution B
ian motion,
36.00 _+ 0.05 wt% glycerol 36.00 -+ 0.05 wt% glycerol
67r~a 3 64.0 _+ 0.1 wt% H20, deion. 10.10 _+ 0.02 wt% NaC1
~h = 2.00 × 10 -6 ± 1% 5.0 X 10-4 M H C I
IBR = chkBT' [2.1]
53.9 +_ 0.1 wt% H20, deion.
p - 1.09 g / c m 3 at 2 5 ° C p = 1.15 g / c m 3 at 2 5 ° C
is 30 rain, allowing sufficient time for mixing, = 2.67 +_0.01 cP at 2 5 ° C
shearing, and sampling. Since tBR ~ R 3 for
fiocs with hydrodynamic radius RH, the a E q u a l v o l u m e s o f s o l u t i o n s A a n d B a r e m i x e d to o b -
tain the test suspension for the coagulation experiments.
Brownian times for the coagulated suspensions
greatly exceed 30 rain, providing sufficient
time for light scattering measurements after experiments the final solution was then slowly
sampling as well. The sheared suspensions poured into a concentric cylinder apparatus,
contained 1.0 M NaC1 to eliminate double the entire preparation taking less than 2 to 4
layer repulsion, putting all experiments in the min. After shearing for a predetermined time,
rapid coagulation regime. Furthermore, the we would remove a sample for analysis. For
suspensions contained 2.5 X 10 - 4 M HC1 to the Brownian coagulation experiments, we
associate weak acid surface groups, but this placed the mixed solution directly in a light
step is not expected to be crucial. The re- scattering cell and followed coagulation in situ.
mainder of the solvent was deionized water. We maintained the temperature at 25.0
Prior to adding particles, we passed all of the _+ 0.1 °C for all light scattering measurements
solvents at least twice through 0.22-#m Mil- and the Brownian coagulation experiments,
lipore GS filters to eliminate dust, a term while shearing was done at 25.0 _+ 1.7°C. Since
which refers to any unwanted particles, the characteristic time for shear coagulation is
fibers, etc. inversely proportional to +q~ and the strain
Because the particles start to coagulate im- rate + is less than 50 s -1 during careful mixing
mediately in a 1.0 M NaC1 solution, we pre- and pouring, the low value of q~ ensures min-
pared two solutions for each experiment, one imal shear coagulation during these steps in
containing the NaC1 and HC1, and the other addition to guaranteeing minimal Brownian
containing the particles. Both contained 36.0 coagulation as stated above. Note, however,
wt% glycerol, with the exact contents listed in that raising 4~to 10 -5 or decreasing the particle
Table I together with the contents and prop- radius by a factor of 2 would cause Brownian
erties of the final test suspension. At the be- coagulation to occur during the few minutes
ginning of an experiment, equal volumes of of mixing and pouring.
each were mixed in a 250-ml Erlenmeyer flask We performed shear experiments at rates
by swirling gently approximately 12 times at between 780 and 1590 s -1 for shearing times
1 rotation per s, and for shear coagulation between 5 and 30 min _+5 s, and at these shear
Journal of Colloid and InterfaceScience, Vol. 142, No, 2, March 15. 1991
558 TORRES, RUSSEL, A N D S C H O W A L T E R

rates the characteristic time for shear coagu- by a Brookhaven Instruments BI-2030AT
lation, digital correlator, and the results were stored
71" on an AT-clone personal computer for later
rs 4+4' [2.21
analysis. We verified alignment of the appa-
ranges from 8.2 to 16.8 min. The shearing de- ratus by examining the light scattered from
vice (29) is a chrome-plated apparatus with the silicone oil used as a bath for the light scat-
the outer cylinder rotating to prevent Taylor tering cell, and over the full range of angles
instabilities, and it has a gap with an outer the results behaved as expected for a Rayleigh
r a d i u s Rcouett e = 6.417 X 10 -2 ___ 4.0 X 10 s scatterer within <2.3%. After alignment, all
m and width h = 1.55 X 10 - 3 ± 1.2 X 10 - 4 experiments took place without further ad-
m. The sample volume is approximately 150 justment to permit accurate comparison. We
ml. Sampling at approximately 0.5 m l / s measured the static light scattering for the co-
through a valve with a diameter of 2.5 X 10-3 agulated suspensions at 12 ° , 15 ° , 20 ° , 30 ° ,
m assured no significant shear coagulation oc- 60 ° , 90 ° , 125 ° , 140 ° , and 160 ° , while auto-
curred during this step, and, prior to each correlation functions were measured and an-
sampling, we collected and discarded 2 to 3 alyzed at the three larger angles. The scattering
ml of solution. Furthermore, we performed volume at 90 ° is estimated to be 18-32 X 106
all experiments well below a conservative es- ~m 3 based on dynamic light scattering mea-
timate of the critical Reynolds n u m b e r given surements discussed elsewhere (31 ).
by For all the measurements, dust proved to
be a~major source of error and often made it
wRcouetteh
Reont- - 1500, [2.3] necessary to repeat experiments. When ~b
2v = 10 - 6 for 0.1-~zm diameter particles in an
which corresponds to the onset of turbulence aqueous suspension, the scattering intensities
for the above geometry (see (30, p. 591 )). In are small, making cleanliness crucial. Conse-
[2.3 ], w is the device angular velocity and u is quently, in addition to passing all solvents
the suspension kinematic viscosity. twice through 0.22-#m filters, the shearing de-
We measured light scattering with a Brook- vice and all glassware were critically cleaned
haven Instruments dynamic light scattering prior to use. The cleaning procedure consisted
goniometer and 514.5 n m light from a Lexel of soaking the items in Micro cleaner and
Argon ion laser, providing data for scattering thoroughly rinsing with deionized water, fol-
angles from 12 ° to 160 ° . This range corre- lowed by a thorough rinsing with filtered
sponds to deionized water and covering the items to pre-
vent contamination. Dust problems still oc-
3.54 # m -1 ~< q = - ~ 0 sin curred despite these steps; hence, we discarded
any experiment for which the intensities were
~< 33.37 # m -1, [2.4] abnormally high. Additionally, in the later ex-
0 being the scattering angle, no being the sol- periments we checked the NaC1 solutions by
vent refractive index, and Xo being the wave- light scattering prior to use, and the NaCl-free
length of the incident light in vacuum, with particle solution precursors were likewise
the incident light polarized perpendicular to checked immediately prior to adding the par-
the scattering plane. Since the scattering vol- ticles.
ume varies with scattering angle as 1 /sin 0 for 3. STATIC L I G H T S C A T T E R I N G
this instrument, all intensities are multiplied The intensity of light scattered from a floc
by sin 0 before analysis, and all intensities re- at a given angle depends on the structure of
ported here have been corrected in this m a n - the floe and is most sensitive to structural de-
ner. The autocorrelation function and average tails on the length scale q-l; therefore, static
intensity of the scattered light were compiled light scattering data covering a range of q pro-
Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
SHEAR-INDUCED COLLOID COAGULATION 559

vide a useful fingerprint of the floe. In addition,


this spectrum can be analyzed in a straight- n 1 o Id <N>w -
forward m a n n e r for weak scattering, making
the technique a useful tool for our purposes.
Before discussing our results, we briefly review
the theory for scattering from a floc based on
the Rayleigh-Gans-Debye approximation and 2
discuss our procedure for examining data.

3.1. Theory
For a floc containing M identical particles,
nlol d - .f,,
the static light scattering in the Rayleigh-
I I I I I I
G a n s - D e b y e approximation is log (air g) 1

I = So ~i=1 M sin(qArij)
j=l qArij [3.1]
Log (a'q)

FIG. i. Static light scattering for a fraetal-likeaggregate


with dimension d. The intensity is divided by the single
where I0 is the scattered intensity correspond- panicle form factor P(aq), where a is the particle radius.
ing to an independent particle, Arij is the cen- The small aq asymptotereflectsthe flocsize, and the linear
portion represents the internal structure, nl0 is the total
ter-to-center distance between particles i and number concentration of individual particles, regardless
j, and the double sum represents a summation of whether they are part of an aggregate,and [d is the zero-
over all particle pairs (32). The above sum is angle limit of the scattering intensity from a singlet mul-
sensitive to the floc structure (i.e., the config- tiplied by an instrument-dependent factor that accounts
uration of particles in the floe) because the for the scattering volume, detector efliciencies, etc.
structure determines which Arij are included
in the sum. Furthermore, the s u m m a n d in high angle limit where aq ,> 1 and therefore
[3.1] changes most for q-1 ~ Arij, resulting qAr o ~> 1 unless i = j, only the i = j terms in
in a strong dependence on the distribution of [3.1] contribute significantly, and the spec-
Arij nearest q 1. t r u m becomes independent of floc structure
To elaborate on the behavior of [3.1], in and size, asymptoting to the value expected
Fig. 1 we present a sketch of a static light scat- for a suspension ofuncoagulated independent
tering plot over a large range o f w a v e n u m b e r s particles. Since we divide the intensity by
q for a hypothetical suspension of flocs con- P(aq), this asymptote is a horizontal line. In
sisting of identical particles. To emphasize the the intermediate region, the spectrum has a
floc configuration dependence in favor of par- shape that reflects the internal structure; for
ticle size effects, we divide the intensity by Fig. 1 the intermediate region is a straight line
P(aq), the single particle form factor. At qRo with slope - d as would be expected for a frac-
1, RG being the aggregate radius of gyration, tal floc with dimension d (22). We will ex-
the length scales probed exceed the aggregate amine the value of a fractal description for
size and, consequently, any Ar 0. As a result, floes formed by rapid coagulation in the fol-
the s u m m a n d is approximately one for all lowing sections.
particle pairs, and in the low q limit the spec- The static light scattering spectra reported
t r u m asymptotes to a value proportional to in this investigation cover the domain 0.177
nlolo<N)w, [3.2]
< aq < 1.67, corresponding to the conditions
cited above. In addition, the data are all pre-
the sum of the scattering from the individual sented as log(I/P(aq)) vs log(aq) for the rea-
particles multiplied by the weight averaged sons cited above. To demonstrate the sensi-
n u m b e r of particles in a floe. Likewise, in the tivity to floc structure, calculated spectra based
Journal of Colloid and lnlerface Science, Vol. 142, No. 2, M a r c h 15, 1991
560 TORRES, RUSSEL, AND SCHOWALTER

on [3.1] are shown in Fig. 2 for various ag- 5.50

gregate structures. The line drawn through the


diamond symbols in Fig. 2 represents a typical 5.00
set of data, and the figure compares this curve
with spectra for a BCC floc with a radius equal
4.50
to t6 particle radii, a linear chain of 10 par-
tides, a doublet, and a singlet. The most im-
portant aspects of this figure are the large, eas- " ~ 4.0O
ily discernible differences between these curves o
over the range o f l o g ( a q ) examined in our ex-
3.50
periments. Also, at aq > 1 the curves should
all approach the singlet line in an oscillatory
m a n n e r (31 ), and this asymptotic behavior 3.00 H H ' ' H I [ ' ' ' ' I IIHllH HH''IIHI HH' IIHH I IH II'IHH H I,,,,,,,H I
-1.00-0.80-0.60-0.40-0.20 0.00 0.20 0.40
begins to appear at the larger aq. The curve Log (aq)
for the BCC floc is based on calculations at
relatively few aq values shown as solid circles; FIG. 3. Static light scattering for a sample sheared 17.5
min at ~/= 1170 s -~. The data represent internal structure
nevertheless, the curve drawn through these since the small aq asymptote depicted in Fig. 1 does not
points shows more oscillations than the other affect these data.
curves, a consequence of the long-range reg-
ularity. For log(aq) < 0.2, the BCC curve has
an average slope of ~ - 3 . The linear chain
the beginning o f an oscillation; the horizontal
spectrum is roughly linear with a slope of ~ - 1
portion indicates the light scattering probes
and starts to turn up at the highest aq. The
length scales larger than the doublet size for
spectrum for the doublet is horizontal over
these lower values of aq. Just as the structures
m u c h of the low aq values and exhibits a min-
of these various types of aggregates differ, so
i m u m at a point above aq = 1, presumably
do the static light scattering spectra in the range
of aq studied, justifying the assertion that our
measured spectra are sensitive to structure in
%00
a practical sense.
6.00
3.2. Results and Discussion
doto
5.00
linear ~ _ Experiments at shear rates of 780, 1170,
t~ 1350, and 1590 S - 1 w e r e performed for shear
4.00 o:22 ,
times varying between 5 and 30 min.
~a 3.00 Data for 17.5 min of shearing at + = 1170
O
s ~ are displayed in Fig. 3, with the scatter
2.00 resulting from the small scattering volume and
the corresponding low scattering intensities,
1.00
the low diffusivity of the large floes, and small
amounts of dust contamination floating in and
0"00--1.2'0.... '--'0
~8'0.... '---b~4~..... 0.dd ..... o.4{) '
out of the scattering volume. The points at
Log (aq)
log(aq) -- 0.18, 0.20, 0.22 are for measure-
FIG. 2. Calculated staticlight scattering for various hy- ments averaged over 300 to 400 s, whereas the
pothetical aggregate structures compared with typicaldata measurements at lower aq are averaged over
from our expcrimcnts (~). BCC: Particleson a body-cen-
60 to 100 s. Also, the data are roughly linear
tcrcd-cubic latticeextending to R = 16a, with the actual
calculated points (@) connected by a spline fit.Linear: for log(aq) < 0 and go through a m i n i m u m
Chain of I0 particlesin a straight line. at larger aq. The slope of the linear portion is
Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
SHEAR-INDUCED COLLOID COAGULATION 561

a 5.50 : b 5.50
shear time: shear time:
~. (-10 min
I5 rain 5.00 ~ . ~ 22.5 r a i n
5.00
5.00 n ~ 26.2r5nin
5 mi
" •4.50 4.50

4.00 O~O4 . 0 0
,--1
3
-- 20 r a i n
S.60 ~25 rain 3.50 ~--.~--
_J30 rain

3.00 i i . . . . . . i i i 11 i l l i i i i i l l i i l l i l l l l l l i 1 1 1 1 I I I I I I I I I I I l l I I S l l l l l I I I I I I I I I l
-~.00-0.50-0.50-0.40-0.~0 0.00 0 . ~ 0 0.40 --1.00 -0.80 -0.60 -0.40 --0.20 0.00 0.20 0.40
Log (~q) Log (aq)

F~O. 4. ( a ) Static light scattering for shearing at ~/ = 1170 s -~ for various shear times. (b) Static light
scattering for shearing at + = 780 s -~ for various shear times.

- 1.8 _+ 0.1, indicating a similarity to a frac- negligible for smaller shear times. Analogous
tal of dimension d = 1.8. data for -~ = 780 s -1 shown in Fig. 4b resemble
Figure 4a displays results for all shear times that in Fig. 4a.
at -~ = 1170 s -1, the curves being smooth fits Additionally, comparing the results for four
to the data with the actual points excluded to shear rates at constant +t = 1.05 × 106 in Fig.
minimize clutter. (The scatter for the different 5 shows that the data for at least the three
shear times is comparable to that in Figs. 3 higher shear rates are independent of shear
and 5.) At shear times less than 17.5 min, the rate. The data for 780 s -~ appear to be system-
spectra change with shear time at lower aq be- atically lower than the other sets, but the de-
cause the flocs for these shear times are small viation is certainly within the scatter observed
enough for the low aq asymptote to be affecting
the data. As the shear time increases, the
5.50
curves approach the 17.5-min curve and the
intensity becomes less dependent on the floc A

size, the spectra for these longer shear times 5.00


being representative of the internal structure.
Floc size data discussed in Section 4 support
4.50
this assertion. Above t = 17.5 rain, the curves
retain their shape while assuming lower po-
sitions in intensity, but we do not believe this 4.00
change reflects a further change in structure 0
,-.1
because the type of structural evolution which 3.50
would cause such a shift is rather unlikely. In-
stead, this drop probably results from a com-
bination of effects including possible multiple 3.00

scattering and loss of particles due to floc sed- Log (aq)


imentation or coagulation on the walls of the
FIG. 5. Static light scattering for shearing at various
shear device. Conversely, the occurrence of shear rates for ~,t = 1.05 × 106. The data correspond to
this shift only for shear times greater than 17.5 + = 1590 s -1 (C3), -~ = 1350 s - ' ( ~ ) , -~ = 1170 s -~ ( . ) ,
rain indicates that these effects are probably a n d "~ = 780 s -~ ( © ) .

Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
562 TORRES, RUSSEL, A N D S C H O W A L T E R

for these experiments. Thus, the internal may be significant. The strong similarity of
structure of the flocs formed by shear coagu- the internal structure for Brownian and shear
lation apparently does not depend strongly on flocs is quite interesting, as well as being un-
shear rate over the range of rates examined. expected. The shear coagulation process may
The data in Figs. 3 through 5 imply the in- feasibly approach scale invariance with respect
ternal floc structure is independent of shear to structural evolution as the flocs grow to sizes
rate, at least at the shear rates examined, but much larger than the particle radius, thereby
complementary data in Fig. 6 supports a more eliminating the natural length scale; thus, a
general conclusion. This figure compares scat- fractal structure for these flocs would not be
tering intensities for shear coagulation with unreasonable. However, the static light scat-
those for Brownian coagulation, and as with tering results indicate that both the dimension
the different spectra for shear coagulation, (he of the scale invariant structure and the density
two spectra in ibis figure are indistinguishable are approximately the same for the two modes
with respect to both shape and magnitude. of coagulation, which means the flocs pro-
Since previous experimental studies of flocs duced by the two modes not only exhibit a
formed by rapid Brownian coagulation dem- fractal-like structure but also exhibit nearly the
onstrate the structure to be approximately same apparent scale invariance.
fractal with d = 1.8 + 0.1, it follows that the
flocs formed by shear coagulation may also 3.3. Data Fitting
have fractal characteristics. (Also, this knowl-
To further investigate the static light scat-
edge about the behavior of Brownian flocs
tering spectra, we fit the data for t = 17.5 min
gives us confidence that the small deviations
at ~, = 1170 s -1 using
at the lowest aq in Fig. 6 result from experi-
mental scatter of the Brownian data.) In both N(r)= l+2t~(r - 2 a ) . B ( r ) d, [3.3]
cases, we report measurements at coagulation
times long enough that the data become in-
where N ( r ) equals the number of particle cen-
dependent of floc size, assuring the spectra
ters within a radius r of a given particle in a
represent the internal structure, but we also
floc, and Y ( r - 2a), the Heaviside step func-
avoid floc sizes for which multiple scattering
tion, accounts for the excluded volume. From
[3.3] one can derive the corresponding spher-
5.50 ically averaged pair distribution func.t.io~
0 namely,

02(r) = ~¢~(r - 2 a ) - - -
47rr 3
~4.50 2dB
+ 6(r- 2a) 16~ra2. [3.41

4.00 The first term in [ 3.4 ] is the power-law decay


with an excluded volume contribution, and
¢
the second term is the nearest neighbor con-
3.50
tribution. The function 6 (x) is the Dirac delta
function. We choose a power-law dependence
3.00 , ,,, .... , ....
-1.00 2518'0'-0.60 -0.40 -0.20 0.00
, .... , ....
0.20
~ .... ,
0.40 characteristic of fractals and fit this particular
Log (aq) set of data because it represents internal struc-
ture.
FIG. 6. Comparison of the static light scattering spectra
for shear-induced ( • ) and Brownian motion-induced ((?)
Using the above pair distribution function
coagulation. for d = 1.8--chosen to match the slope of the
Journal of Colloid and Interface Science, Vol. 142,No. 2, March 15, 1991
S H E A R - I N D U C E D COLLOID C O A G U L A T I O N 563

6.00
B a higher value of log (aq) and a lower intensity.
0.80 Consequently, we conclude that the simple
0.68
5.50 0.62 power-law fit outlined above reasonably rep-
0.57
0.46
0.34
resents the internal structure at larger length
,~5.00 0.23 scales but not near contact.
With the fitted values for B and d, one can
~ 4.80 estimate some characteristic floc parameters.
First, these values correspond to an average of
~ 4.00 2.2 nearest neighbors for a particle, which im-
plies the floc does not have a tightly packed
3.50 structure. Consistent with the number of
nearest neighbors, a plausible description of
8.00 ..,,.,,i.,.,,..i.,...,,i.,..,,,i.. floc-floc bonding for rapid coagulation might
-1.20 --0.80 --0.40 0.00 0.40
Log (aq) incorporate interpenetration of two flocs until
a particle in one contacts a particle in the other,
FIG. 7. Calculated static light scattering corresponding
at which point a bond forms. For our exper-
to [3.4] for d = 1.8 and various values of B.
imental conditions, the attractive forces be-
tween particles are expected to dominate only
low aq d a t a - - a n d various values of the coef- at small separations and, as mentioned earlier,
ficient B, we calculated the family of static the bonds are believed to be rigid; hence, for
light scattering curves shown in Fig. 7. These two flocs to form multiple bonds these contacts
curves have all been normalized using the would have to occur at almost the same time,
measured scattering intensity for uncoagulated an unlikely possibility. Thus, it is reasonable
IDC particles, thereby accounting for the scat- to expect two flocs to form a new floc by
tering from a single particle, the scattering bonding at only one point in most coagulation
volume, and the optics of the detection system. events.
Specifically, we measured the scattering inten- In addition to the number of nearest neigh-
sity as a function of aq for particles suspended bors, one can estimate porosities by approxi-
in a salt-free solution and corrected the inten-
sities using Rayleigh-Gans-Debye theory to
5.80
account for the actual particle concentration
and refractive index of the sheared suspension.
As a reference, Figs. 7 and 8 show the single 8.00
particle scattering determined in this manner
as horizontal lines at log(I/P(aq)) = 3.67 "• 4.00
+0.03. For small log(aq), the calculated
curves have values significantly higher than
this line and asymptote to lines with slope ¢~ 4.00

-1.8, each curve displaced an amount depen-


dent on the value of B. However, the curves 3.50
all cross the single particle line at the same
point, as required by the form of[ 3.3 ]. Figure
8 shows a fit to the data corresponding to B
= 0.62 (chosen "by eye"), and this curve is Log (a,~
seen to cross the single particle line near where
FIG. 8. Fit based on [3.4] of the data in Fig. 3. The
the data approach this line. The calculated fit chosen values of B and d are B = 0.62 and d = 1.8. The
does not capture the upturn in the data at horizontal line represents the scattering for uncoagulated
log(aq) > 0, but it does have a minimum at particles.
Journal of Colloid and Interface Science, VoL I42, No. 2, March 15~ 1991
564 TORRES, RUSSEL, AND SCHOWALTER

mating the number of particles in a floc with cared. Even if one assumes the floc to be
outer radius Rci a s spherical, rotational motion affects the auto-
correlation function. Lindsay and co-workers
i=B +1, (34-36) addressed this problem for a fractal
structure by numerically evaluating the au-
withB=0.62andd= 1.80, [3.5] tocorrelation function of simulated flocs at aq
< 1 and all values of qRc, and they later pre-
B and d being the values obtained from the sented an approximate method; however, we
fit in Fig. 8 and R~i defined as the radius of will apply an alternate methodology that ad-
the smallest sphere containing all particle cen- dresses the behavior at aq >~ 1 by examining
ters of an /-particle floc. The power-law de- the differential equations governing dynamic
pendence o f [3.5] is chosen to be the same as light scattering.
that for N ( r ) without the excluded volume
term, and the factor 1 is added to give reason- 4.1. Theory Accounting for Rotations
able behavior for i = 2 and i = 1; nonetheless,
the choice of [3.5] is ad hoc, as is [3.3]. For Consider a collection of a large number of
R~i = 10a, [3.5 ] yields a solids volume fraction flocs c~ = 1, 2 . . . . . n~, with centers at R,.
4~o~ of 0.04. Each floc contains Ms particles located at b~
To summarize, the flocs formed by rapid (m = 1. . . . . M,) relative to its center. The
shear coagulation of particles that form strong, electric field scattered from a floc at a given
rigid bonds appear to be porous, tenuous ag- scattering vector q follows
gregates with ~ 2.2 nearest neighbors for each M~
particle on average. The structure is remark- E~(t) "~ E 0 . e iq'x%~t~ ~ e iq'b~"~t~, [4.1]
m=l
ably similar for Brownian motion- and shear-
induced coagulation at these conditions and where E0 is the incident electric field assumed
appears to be independent of shear rate. The to be polarized perpendicular to the scattering
2.2 nearest neighbors, the porous structure, plane. Therefore, the intensity autocorrelation
and the similarity between shear and Brown- function defined by
tan results are indications that there is little
rearrangement, as desired for our studies. G2(q, t) = (I(q, t)I(q, 0 ) ) , [4.2]
an ensemble average over all floc positions and
4. DYNAMIC LIGHT SCATTERING orientations, can be written as
Unlike the structure, the rates correspond-
ing to shear- and Brownian motion-induced G2(q,t)= C( ~ exp{iq.[R~(0)
cq/~,3,,6= 1
coagulation differ substantially, and measure-
ments on the former mode at the conditions - l i e ( 0 ) + R~(t) - li~(t)])
of interest are lacking. In order to obtain data
for a characterized model system, we measured × ~] exp{ iq. Ibm(0) - b ~ ( 0 )
average floc sizes by dynamic light scattering l,m,n,p=l
for the shear coagulated suspensions described
+ b.~(t)- b~t)])5. [4.3]
above. Specifically, we calculated floc hydro-
dynamic radii from the initial decay of the The constant C multiplying the right-hand side
intensity, or second order, autocorrelation of [4.3 ] is not important insofar as data anal-
functions. This calculation requires a theory ysis is concerned. If one assumes the flocs dif-
describing the initial decay in terms of the size, fuse independently and rotational diffusion is
and for single spheres it is known that the de- independent of translational diffusion, the en-
cay follows a simple exponential in t/a (33). semble averaging eliminates most of the terms
For a floc, though, the decay is more compli- in the summations of [4:3], and terms cor-
Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
SHEAR-INDUCED COLLOID COAGULATION 565

responding to translation factor from those for To derive an equation for F(q, t, bl, bp,
rotation (see similar arguments in (34, 33 pp. 01p), we examine P(b, t; b(0)), the probability
62-64)). Assuming for simplicity that the that a particle in a floc starting at b(0) at t = 0
translational diffusion coefficients of all flocs resides at b at time t. This probability is the
are identical, we then obtain solution of the initial value problem (37)
G2(q, t) = G0exp { -2q2Dt} OP
- - = Ob2V~P [4.7.1]
nT Me Ot
× E [~ (exp{iq.(b;(0)-bp(t))})
a,B4~o~ Lp ~ 1 t.C.:P=6(b-b(0)) att=0. [4.7.2]
M~ The operator V2~--the angular part of the
× ~ (exp{-iq'(b~m(O)-h~(t))})] + G~ divergence operator--is defined as V~
m,n ~ l
= Vb~" Vb~, where Vb~ = (I - bb/b2) • V, and
[4.4]
the rotational diffusion coefficient O governs
when the scattering volume contains a large how quickly the rigid body changes its orien-
number of flocs, a condition met in our ex- tation due to Brownian motion. The ensemble
periments. Equation [4.4] has a baseline G~ average in [4.5 ], the defining equation for F(q,
and a time dependent part which depends on t, bl, bp, Olp), is similar to a Fourier transform
the diffusion coefficients of interest. Thus, the of P; hence, an initial value problem describ-
problem of determining the dynamic light ing F(q, t, bt, bp, 0lp) can be obtained from
scattering from a floc with a given structure Eqs. [4.7] by applying methods similar to
reduces to solving for those invoked to Fourier transform the three-
F( q, t, bz, bp, Otp) dimensional diffusion equation. Working
through the details yields
=(exp{iq.[b,(0)-bp(t)]}) [4.5]
for two points at bl and bp in a rigid body which - q2b2F- b2ff-~ [4.8.1]
undergoes rotational diffusion independent of
its translational diffusion. The independent sin(qs)
variable Ozprepresents the angle between bt and I.C.:F--- att=0. [4.8.2]
qs
bp, the vectors whose magnitudes are bl and
In Eqs. [4.8], b can equal the magnitude of
bp, respectively.
either bl or bp, and s = [bz(t) - bp(t)[. One
The function F(q, t, bl, bp, Olp) describes
method of solving Eqs. [4.8] involves applying
the correlation between the electric field scat-
separation of variables to obtain
tered from the point bt at time t = 0 and that
scattered from the point bp at time t, with the
F = ~ (2n + 1 ) e x p { - n ( n + 1)Or}
brackets denoting an ensemble average over n-0
all orientations at time t = 0. As a result of
the averaging over all orientations, F depends × jn(qbt)jn(qbp)Ln(cos Olp); [4.9]
on bl and bp only in terms of their magnitudes
then this series can be summed over all particle
and the angle Otp,as reflected in the functional pairs, each pair (l, p) corresponding to specific
form given above. Once F( q, t, bt, bp, Otp) is
values for bl, bp, and 0tp, to obtainf(q, t). (In
obtained, one can determine
[4.9],jn(x) is the nth spherical Bessel function
M . M~
of the first kind and Ln(x) is the nth Legendre
f(q, t) = ( Z Z F(q, t, bl, bp, Ozp)> [4.6] polynomial.) However, this procedure can be
/=I p=l
quite cumbersome for all but the simplest
for a given floc structure. The brackets in [4.6] structures. In addition, it requires knowledge
represent an average over an ensemble offlocs, of the structure in terms of the number of par-
replacing the outer summation in [4.4]. ticle pairs in a floc corresponding to a partic-
Journal o f Colloid and lnterface Science, Vol. 142, No. 2, March 15, 1991
566 TORRES, RUSSEL, A N D S C H O W A L T E R

ular set of values for bl, bp, and 01p.Thus, we Therefore, the contribution to Of(q, t)/Ot at
utilize an approach which circumvents the t = 0 from the self-correlation term in [4.13 ]
need for a priori knowledge of the structure at becomes
this point in the derivation.
To proceed, we convert G2(q, t) to the first- -O(N)qZRg. [4.16]
order autocorrelation function Gl(q, t)by In a manner analogous to the derivation of
subtracting the baseline G~ from G2 ( q, t) and [ 4.16 ], we substitute the initial condition for
taking the square root, yielding F(q, t, bl, bp, 0lp) into the governing differ-
G~(q, t) = A e x p { - q 2 D t ) f ( q , t). [4.10] ential equation [4.8.1] to provide an equation
for F at t = 0, and we sum the result over all
The time derivative of Gl(q, t) at t = 0 is pairs of distinct particles to give the pair cor-
Gl(q, o) relation term in [ 4.13 ]. Equivalently,

of
=A{-q2D.f(q,O)+-~t=o }. [4.11] f dblpf(b,) f ds p2(S; bl)F(t = 0) [4.17]

Furthermore, at time t = 0, represents the summation over all pairs of dis-


tinct particles as a double integral, where pf(bl)
f(q, O) = ( ~ pZ sin(qS);qs/ , [4.12] is the probability density for a particle in a floc
1 to be at bl relative to the floc center and p2(s;
b/) is the pair probability density for a particle
which is just the static light scattering for the
to be at s relative to position b/given that there
floc divided by the single particle scattering.
is a different particle at bl. Assuming spherical
To analyze [4.11], we separate f ( q, t) into
symmetry of p2 (s; bl) with respect to s makes
self-correlation and pair correlation terms and
it possible to integrate [ 4.17 ] over all orien-
differentiate to obtain
tations of s to yield
df Mo
- ~ = ( E Fo( q, t, bi) )
i=1 87rq23 o [ f dblpf(bl)b~
M~ M~
+ ( ~ ~ F(q, t, bt, bp, Olp)}. [4.13]
l=1 p=l × f ds o2(s; b,)szSin(---qs)l. [4.18]
p#l qs j
The second term on the right-hand side of
[4.13 ] is a double sum over all distinct particle In making this assumption, we are focusing
pairs, distinguishing it from the sum in [4.6 ]. our attention on the behavior at qRG ~> 1,
The function F0 represents self-correlation and whereas at smaller qRG the outer boundary of
satisfies the initial value problem given by Eqs. the flocs would cause a dependence on the di-
[4.8] for s = 0, rection o f s . (Note that one must account for
the dependence o f / " on the orientation of s
10Fo _ q2bZFo _ ~b ( b2 OFoI relative to bl when performing this integra-
00t Ob ] [4.14.1] tion.) If one further assumes that 02(s; hi) is
independent of bl, [ 4.18 ] simplifies to
I.C.: Fo = 1 at t = 0. [4.14.2]
Upon substituting the initial condition for F0 87rq2 OR 2
into [4.14.1 ], one finds 3
OFo
Ot~-t=O = - O q Z b 2 " [4.15] - - s
qs )
2 . [4.19]

Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
SHEAR-INDUCED COLLOID COAGULATION 567

Our motivation for making this assumption becomes smaller as q becomes larger.
comes from its validity with fractal objects, Because the pair correlation term in [ 4.13 ]
but the form of p2(s) is not restricted to be a is small relative to the self-correlation term at
power of s, giving added generality. Again, in large q, the assumptions regarding 02 are not
making this assumption we are considering expected to be important in this limit. Indeed,
wavenumbers satisfying qRG >> 1, rendering [ 4.18 ] furnishes a basis for determining the
the behavior ofp2(s; bl) at length scales com- order of magnitude of the pair correlation term
parable to Rc unimportant. Finally, recogniz- even when the assumptions about 02 are not
ing that f ( q, 0) can be written as strictly valid. In our experiments the ratio
given by [4.23] is not always negligible, but it
f ( q , O) = ( N } + 47r is small enough to provide confidence that the
assumptions about 02 do not lead to unac-
sin(qs)
× f dbtpAbl) f ds p2(s)s2 qs ceptable errors in the data analysis.
The only other analysis of dynamic light
scattering from flocs similar to the type seen
= ( N ) { 1 + 47r f ds p2(s)s 2 sin(qs)}
qs in this investigation is work reported by Lind-
say et al. (34, 35 ) and Lin et aL (36) for fractal
[4.201 flocs with aq < 1. Comparing their results with
[4.21], one finds the term
using the same assumptions invoked to arrive
at [ 4.19 ], we obtain the result 1 (N}
[4.24]
2 f ( q , 0)
rl __ (~1 (t = 0) { 2
G l ( t = 0 ) - q2 D + ~ 0 R 2 absent from their solution, but their results do
address the behavior for all qRG when aq ~ 1.

× ( 1+
2 f ( q , O)
)}
'
[4.21]
Equation [4.21], on the other hand, was de-
rived using assumptions for p2 which only
make sense ifqR~ >> 1. Although of theoretical
where F1 is the first cumulant of the first-order interest, the additional contribution from the
autocorrelation function (33). Eq. [4.21] is above term is comparable to the uncertainty
the desired equation relating the diffusion in our experimental data and therefore is not
coefficients to the autocorrelation function. crucial here. Nonetheless, our development
Note that the term explicitly distinguishes the contributions of the
self and pair correlation terms.
(N}
[4.22] 4.2. Determining the Hydrodynamic
f ( q , O)
Radius of a Floc
equals the ratio of the single particle baseline
to the scattered intensity and therefore can be In order to determine a floc radius using
measured by static light scattering, and recall [4.21], one must relate the diffusion coeffi-
that [ 4.21 ] was derived without explicitly in- cients to the radius. Unfortunately, exact
voking a power-law pair distribution function. equations are not available. Consequently, we
Also, note that the ratio of the pair correlation follow the lead of previous investigators (e.g.,
term to the self-correlation term in [ 4.13 ], (12, 34, 36)) and define an effective hydro-
dynamic radius RH according to
2 ( N } - f ( q , O)
[4.23] D - - -
kBT [4.25]
3 (N} ' 67r~zRH
Journal of Colloid and Interface Science. Vol. 142, No. 2, March 15, 1991
568 TORRES, RUSSEL, AND SCHOWALTER

2.50
Also, Lindsay et al. (34) assumed

O- kBT
87r~R3n [4.26] 2.00

and Chen et al. (14) reported /3 = RH/RG

J
= 0.875; hence, to be consistent with their re- ~-~ 1.50

suits we adopt

2 0R 2 1.00 - -
3
- - = 0.65. [4.27]
D

Thus, our final working equation is 0.50


0.1
, , , , , , , i

1
, , , , , , , ,

1'o
qRQ
['1 = q 2 D 1 +0.65 1 +2f(q, 0) " FIG. 9. FI/q2D as a function ofq/~ at Rda = 11.3.
The points (*) depict calculationsbased upon [4.28] for
[4.28] the data in Fig. 3, and the solidcurverepresentsthe results
from Lindsayet al. (35) for fractal flocs.The valueRG/a
Compared with the large qRG asymptote = 11.3 followsfrom/3 = 0.875 and measurementsOfRn
correspondingto the experimentdepicted in Fig. 3.
I'1 1
q2----~= 1 + 2 7 [4.29]
4.3. Discussion of Results
reported by Lin et al. (36), our analysis differs Figure 10 shows a typical plot of G~(q, t)
quantitatively only in the extra term given by at longer shear times for correlation times cor-
[4.24], a term which is important when aq responding to the first 64 correlator channels.
>~1. To calculate G1 (q, t) from the measurements
To illustrate the expected behavior of r l/ of Gz(q, t), one needs a value for the baseline
q2D, Fig. 9 shows points calculated from G~, and the Brookhaven Instruments corre-
[4.28] based on our data (the necessary scat- lator/computer determines this quantity by
tering intensities were obtained from Fig. 3) two methods: from the total scattered intensity
as well as a curve representing results from and from six delay channels. The difference
L~ndsay et al. (35). As illustrated, [ 4.28 ] pre- between the two yields an uncertainty which
dicts behavior at aq >~ 1 that is not reported we include in the calculation of error bars. For
by Lindsay et al. (34, 35) or Lin et al. (36), our purposes, we determined F1 by plotting
while their results address behavior at qRG <~ 1 ln{ G1 (q, t) } vs t and graphically measuring
that our approach misses. the initial slope; due to noise in the data, un-
For our purposes, [ 4.28 ] together with scat- certainty of the baseline, and a lack of a simple
tering intensity data provide an adequate equation describing the autocorrelation func-
means for determining average floc hydrody- tion at t > 0, more elegant methods would not
namic radii. Alternatively, one could measure significantly increase the accuracy.
I'1 and the static light scattering over a range Figure 11 presents data for floc size scaled
of q and plot I'1/q2 against the ratio given by by the particle radius as a function of time
[4.22]; then O . R~ and D could be deter- scaled by ~-s, the characteristic shear time at t
mined directly from the slope and intercept = 0 (see [ 2.1 ]). Each point corresponds to
by invoking [4.21]. Unfortunately, this pro- one shear experiment, and the error bars rep-
cedure requires data at aq >~ 1 that are more resent uncertainties of the baseline and differ-
accurate than our measurements provide. ences between data collected at the three angles
Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
SHEAR-INDUCED COLLOID COAGULATION 569

-0.60
of interest. The Brownian calculations are

-0.80

-1.00
<. based on the equation ( 11 )

RH
a - Y(2 -
(t/tBr) ~/a
1/d) "
In addition to Fig. 11, we present data for
[4.31]

floc size as a function of shear rate at constant


~ -1.20 "i,l = t.05 × 106 in Fig. 12. If there were no
breakup or rearrangement and no significant
-1.40 hydrodynamic or colloidal interactions, then
the data would all lie on a horizontal line be-
cause the shear rate' would only affect the time
-1.60 ......... f ......... ~ ......... = .........
O.OE+O00 1.0E+O03 2.0E+003 3.0E+003 4.0E+003 scale; however, we see that floc size does vary
time, /.t,s e c with the shear rate, being smallest at the high-
FIG. 10. First-order autocorrelation function G~(t) cal- est values. We discuss the possibility of this
culated from G2(t) measured at a scatteringangle of 140° effect arising from floc breakup in the next
for 15 min of shearing at ~/= 1170 s 1. section.

5. KINETICS CALCULATIONS
125 ° , 140 ° , and 160 ° . This figure shows results
for shear rates of 1170 and 780 s -1, and the To calculate shear coagulation kinetics we
scaling given by rs partially collapses the data use the equations (4)
for the two rates. As discussed in the following dNi_ 1 i~ 1
section, this time scaling is expected to be less dz 8 j=l ~ ( R j Jv Ri_j)3NjNi_j
effective at long shear times because breakup
of large flocs occurs. Note that the radii at t~ 1 o~ 1
rs < 1.5 for + = 1170 s -I are comparable to 4 j~ -~o (Ri + Rj)3NiNj [5.1.1]
the values o f ( a q ) -I at which the curves in Fig.
4a deviate from the 17.5-min curve, providing I.e. 1: N~ = 1 at r = 0 [5.1.2]
further evidence that the 17.5-rain curve rep-
resents the internal structure. A similar com-
l.C. 2 : N i = O , i > O , a t r = 0. [5.1.3]
parison can be made with Fig. 4b, but the The quantities N~ represent the n u m b e r con-
scatter makes the effect less obvious. centrations of flocs containing i particles di-
Another important aspect of these experi- vided by the initial n u m b e r concentration of
ments is the contribution of Brownian motion. singlets, r is the time scaled by rs, Ri = Rci/a
At the initial stages of the shearing when most is the capture radius of an/'-particle floc scaled
particles exist as singlets, the Peclet n u m b e r by a (a quantity discussed below), and Wa is
a t e / = 1170s - l i s the stability ratio for coagulation of an i-par-
ticle floc with a j-particle floc.
Pe = 2.15, [4.30]
The stability ratios W/j differ from unity due
and Brownian motion obviously contributes to hydrodynamic and colloidal interactions,
significantly. Note, however, that Pe ~ R 3 but these effects are expected to be small for
for a floc; hence, when RH = 6a, Pe = 465. the system investigated. As stated above, dou-
To compare the rates of shear and Brownian ble layer repulsions are negligible because the
coagulation at longer times, Fig. 11 shows the solvent has a high salt concentration, and the
calculated growth for Brownian coagulation, hydrodynamic and attractive dispersion forces
demonstrating that this mode is m u c h slower between the porous flocs are much weaker
than shear coagulation for most of the times than those for two solid bodies of comparable
Journal of Colloidand InterfaceScieuce, Vol. 142, No. 2, March 15, 1991
570 TORRES, RUSSEL, AND SCHOWALTER

25.00
and colloidal floc behavior at a level that is
appropriate to the current understanding.
20.00 To proceed, the hydrodynamic and capture
radii must be related to the number of particles
15.00
in a floc. Equation [ 3.5 ] suffices for Rci, and
an equation for RHi results from setting /3
= 0.875 as in Section 4 (14). In order to relate
10.00
Rri to i, we assume the distribution of particles
in a floc follows
5.00

0.00 ' ' ' ' l ' ' ' ' l ' ' ' ' ' ' ' ' l l l ' ' ' ' ' ' ' ' l ' ' ' ' ' ' ' ' ' t ' ' ' ' ' ' ' ' ' l
0.00 0.50 1,00 1.50 2.00 2.50
t/-r,
with B = 0.62, d = 1.8, and b equal to the
F~G. 11. Floe hydrodynamic radius vs r = t/zs for
distance from the center of the floc. This
shearing at + = 780 s -l ([])'and + = 1170 s -l (0). The
solid line ( ) represents kinetics calculations for + equation follows upon eliminating the nearest
= 1170 s -l with a maximum-floe-sizecriterion, and the neighbor term and excluded volume step
dashed line ( ..... ) represents calculations for Brownian function from [3.4 ], retaining the power-law
coagulation ( 11). dependence. (The former terms are necessary
in [3.41 but not in [5.2] .) Also, the 6(b) term
in [5.2] is added to maintain at least one par-
size. In fact, the latter two forces are only sig- ticle in the floc. F r o m [5.2] and the chosen
nificant very near contact. Taking these factors value of 13
into account, we develop a two-radius floc
model and use it to examine by what extent (i-1)d] °'5
W o. exceeds unity. RHi = 0.875 ~-~--_~2) ] Rci
The proposed two-radius model describes
the floc as a spherical body with both a capture for i >~ 2, [5.3]
radius Rc and a smaller hydrodynamic radius
RH. The latter describes the drag in a flow field
and is used to estimate the hydrodynamic 20.00
functions, whereas the former radius repre-
sents the distance within which another floc
must approach for coagulation to occur. Be- 15.00
cause two flocs bond according to our model
when they are separated by a distance (Rcl
+ Re2) that is greater than ( R m + RH2), lu-
brication forces will not exhibit a singularity
at contact, and as a consequence the form of
10.00
t
the equation invoked to describe attraction is
5.00
not as important as for two solid bodies. In-
deed, trajectory calculations suggest the at-
traction can be replaced with adhesion at con-
0.00
tact without significantly affecting the results. 50~ ~o . . . . . . io'od.~6 . . . . . 1'5~d.6d ..... ~o~d.oo
Our model also accommodates the expected •.f, sec -I

fluid flow within the outer portions of a porous FIG. 12. Floc hydrodynamic radius vs ~/for shearing at
floc. Thus, our simple description encom- +t = 1.05 × 106. The points ( e ) represent kinetics cal-
passes the basic features of the hydrodynamic culations incorporating a maximum-floc-size criterion.
Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
SHEAR-INDUCED COLLOID COAGULATION 571

1.60
but this equation gives RHi < a for i = 2. This
discrepancy is not too important because our
investigation considers flocs containing many 1.40
particles, but to maintain a realistic value of
RH2 the factor 0.445a is added as an ad hoc
1.20
correction, yielding our working equation
((i- 1)all°5
RHi = 0.875 ~ + 2) ] Rci + 0.445a. 1.00

[5.41
0.80
Single particles are assigned the values RHi
= R c i = a.
0.60
Comparison of the results of the kinetics , , , , ,,,,i

10
, , , , ,,1,1

100
, , , , ,,,11

1000
calculations with the dynamic light scattering i

data also requires determining the intensity FIG. 13. Estimate of the stability ratio W. as a function
averaged hydrodynamic radius ( R H ) , and as- of i based on the two-radius floc model.
suming Ii is proportional to i as suggested by
the data for the larger scattering angles, this
average can be written as already limits our ability to calculate W~j ac-
curately. Thus, in the following calculations
we set Wo = 1.
When all the elements of the above model
(RH) = Z ~ I Nil . [5.5]
are put together and Eqs. [ 5.1 ] solved, the re-
Combined with known hard-sphere hydro- sults predict floc growth that increases at a
dynamic functions (21, 38-41), our model continuously increasing rate, contrary to the
can be used with [1.5] to calculate stability data in Fig. 11. Evidently the flocs cannot grow
ratios for two interacting flocs. The details are to arbitrarily large sizes in the shear flows
presented elsewhere (31 ), but the result for studied, indicating breakup should be included
two equal-size flocs in the absence of Brownian in our calculations. For the case of interest
motion is two flocs likely coagulate at one particle-par-
ticle contact as discussed in Section 3, and
{ 1 -- 87]---
3~ f 27/i~ ( B1( r.~) ~'-?, ) ) r breakup occurs when the hydrodynamic force
Wff I:
pulling the flocs apart exceeds the attractive
force at the contact point. However, the hy-
× exp[-2 ~ig (B(F)
-1---A-~
- A(?) IF
ll drodynamic force changes as the flocs rotate
together in the fow; hence, to develop a
breakup criterion we assume that two flocs
×exp[3~n~(B(-lr)-A(r)-_~-~ ,I ldr]
r [5.6] which come into contact will separate if the
maximum hydrodynamic force acting to sep-
with rh = Rci/RHi. Figure 13 illustrates that arate them exceeds the attractive force holding
the stability ratio W~ differs only moderately two particles together. For stability between i-
from unity and does not depend strongly on and j-particle flocs, this criterion is
i. Further computations would provide all the 3
necessary W~j, but Wij = 1 is a reasonable ap- 7 r # ( R H i -t- R H j ) 2 / y <. FAu [5.7]
proximation in view of Fig. 13, the uncertainty
in our data, and our intent to identify trends. for simple shear and small separation limits
In addition, the neglect of Brownian motion of the appropriate hydrodynamic functions,
Journal of Colloid and InterfaceScience, Vol. 142,No. 2, March 15, 1991
572 TORRES, RUSSEL, AND SCHOWALTER

with the extrema in the radial component of


the viscous force occurring at an angle of 45 °
1.0 E-1
relative to the bulk velocity (4, p. 301 ). Pri-
mary m i n i m u m attraction at a separation 6
1.0 E-2
with a Hamaker coefficient An yields

aAH 11/2 z" 1.o ~.-s


(RHi + R , j ) <~ 18--~-52 / [5.81
1.0 E - 4

To evaluate [5.8 ] we simply choose 5 = 1 nm,


l.O E - 5 .
a value that corresponds to length scales at
which molecular structure, roughness, etc.,
1.0 E-6. ~-~
become important, and we calculate A H for
polystyrene latexes in the solvent of interest i
according to an approximation reported by FIo. 14. Dimensionless number concentration N~vs i
Russel et al. (4, Eq. [5.9.3]), obtaining AH calculatedfor r = 1.2 at -y = 1170 s-1. The peak at i = 390
= 5.5 × l0 -21J. It is also likely that rearrange- correspondsto the maximum floc size at this shear rate.
ment begins to play a role when the shear
forces are comparable to the bonding forces,
and a detailed analysis of the structure at the dependence on shear rate agrees nicely
longer times should consider such effects. For with the data. Recognizing that the calcula-
our purposes, however, the simple criterion tions would give a horizontal line in the ab-
above captures the essential physics that limit sence of a breakup criterion, these results sup-
,further growth of aggregates. port the hypothesis that the force holding two
Figure 11 shows the solution of Eqs. [ 5.1 ] floes together is comparable to one particle-
for ~ = 1170 s -1 using [3.5] and [5.4] for the particle bond. Furthermore, it is interesting to
floc capture and hydrodynamic radii, [5.5] for note that the calculated curve would lie closer
the intensity averaged hydrodynamic radius, to the data if stability ratios closer to those
[5.8] for floc breakup, and Wij = 1. We per- given in Fig. 13 were used instead of Wij = 1.
formed the integration using the routine Evidently, the two-radius floc model, struc-
ODESSA (42-44) on a CONVEX vector pro- tural information from our fit of the static light
cessor. Although the calculated curve and the scattering spectra, and the proposed floc
data differ somewhat, the agreement is good breakup criterion provide the necessary ele-
considering the uncertainties in the data and ments for describing the coagulation process
the approximations in the theory. In addition at a level of detail appropriate for our data.
to limiting the sizes at large times to values Reconsidering the effects of shear on the floc
consistent with measurements, the breakup structure, recall that the two possible conse-
criterion causes the change in curvature of the quences of large shear stresses are rearrange-
calculated growth curve at intermediate times. ment and breakup. As examples, L i n e t al.
To examine the size distribution at longer (45) found rearrangement occurs without sig-
shearing times, we show a representative plot nificant breakup when fractal gold clusters
in Fig. 14, and this figure illustrates the de- formed by Brownian aggregation experience
velopment of a peak at the maximum floc size, large shear stresses, whereas we propose the
as well as the broadening peak expected at stresses in our system restrict the flocs to a
smaller sizes. maximum size. In the absence of breakup,
Figure 12 shows the prediction for shearing rearrangement of our floes would yield more
at different shear rates with ~t = 1.05 × l06, compact aggregates and slow the kinetics ac-
and although the predicted values are high, cordingly, but Figs. 11 and 14 suggest the ab-
Journal of Colloid and Interface Science, Vol. 142, No, 2, March 15, 199 !
SHEAR-INDUCED COLLOID COAGULATION 573

sence of an upper limit on the n u m b e r of par- neighbors inferred from the static light scat-
ticles in a floc would still lead to m u c h larger tering spectra. An approximate breakup cri-
radii than observed. On the other hand, Eq. terion then follows by equating the force hold-
[ 5.8 ] with AH(gold) = 3 × 10 -19 J (4, p. 148), ing two particles together to the m a x i m u m
6 = a / l O , and a = 7.5 n m implies the Rr~ hydrodynamic force acting to separate two
= 500 n m flocs studied by L i n e t al. (45) re- flocs near contact. Rearrangement without
main intact as long as ~+ < 280 N / m 2, and breakup m a y also occur when these forces be-
the largest shear stress they report equals 90 come comparable, but we neglect this more
N / m 2 ; therefore, [ 5.8 ] is consistent with their complicated possibility because it does not
conclusion that breakup does not occur. The appear to be capable of explaining our data.
reason the gold aggregates restructure at length A further paper will describe simulations
scales between 100 and 500 n m while ours that provide complementary information on
apparently do not remains an interesting topic the structure of flocs formed by shear coagu-
for study. lation.

6. SUMMARY ACKNOWLEDGMENTS
In a suspension undergoing rapid shear co- The authors gratefully acknowledge NSF and the Ford
agulation, polystyrene particles with a low Foundation for graduate fellowship support (FET), IFPRI
surface charge can form strong, rigid bonds for financial support, and Prof. Y. B. Kevrekidis and
H. S. Brown for assisting with the kinetics calculations
and aggregate into large, tenuous flocs. Static and providing access to the CONVEX computer.
light scattering detects an internal structure
which is independent of shear rate and indis-
REFERENCES
tinguishable in our experiments from that as-
sociated with Brownian coagulation; namely, 1. Ali, S. I., and Zollars, R. L., J. Colloid Interface Sci.
117, 425 (1987).
both modes give a floc structure that can be
2. Hunter, R. J., Adv. Colloid Interface Sci. 17, 197
approximated as a fractal with d ~ 1.8 _+ 0.1. (1982).
Also, a fit to the static light scattering data sug- 3. Ekdawi, N., and Hunter, R. J., J. Colloid Interface
gests an average of 2.2 nearest neighbors per Sci. 94, 355 (1983).
particle. 4. Russel, W. B., Saville, D. A., and Schowalter, W. R.,
"Colloidal Dispersions." Cambridge Univ. Press,
With results for the dependence of the first
Cambridge, 1989.
cumulant on the rotational and translational 5. Schowalter, W. R., Ann. Rev. Fluid Mech. 16, 245
diffusion coefficients, we also examine floc size (1984).
as a function of shear rate and time by dy- 6. Sonntag, R. C., Ph.D. dissertation. Princeton Uni-
namic light scattering. To interpret these data, versity, Princeton, NJ, 1985.
7. Van de Ven, T. G. M., Adv. Colloid Interface Sci. 17,
we develop a two-radius floc model that de-
105 (1982).
scribes the hydrodynamic behavior of our 8. Van de Ven, T. G. M., and Mason, S. G., J. Colloid
tenuous aggregates. Analysis of the model Interface Sci. 57, 505 (1976).
suggests hydrodynamic interactions can be 9. Van de Ven, T. G. M., and Mason, S. G., J. Colloid
neglected in kinetics calculations and, further, Interface Sci. 57, 535 (1976).
that a sticky-floc attraction suffices for our 10. Lindsay, H. M., Lin, M. Y., Weitz, D. A., Sheng, P.,
Chen, Z., Klein, R., and Meakin, P., Faraday Dis-
purposes. Calculations based on this model cuss. Chem. Soc. 83, 153 (1987).
predict behavior consistent with our measure- 11. Sonntag, R. C., and Russel, W. B., Z Colloid Interface
ments, but only when a m a x i m u m size crite- Sei. 113, 399 (1986).
rion is invoked. 12. Wiltzius, P., Phys. Rev. Lett. 58, 710 (1987).
To develop a size criterion, we assume two 13. Wiltzius, P., and van Saartoos, W., Phys. Rev. Lett.
59, 2123 (1987).
flocs join at one particle-particle bond, a hy- 14. Chen, Z. Y., Meakin, P., and Deutch, J. M., Phys.
pothesis consistent with the number of nearest Rev. Lett. 59, 2121 (1987).

Journal of Colloid and Interface Science, Vol. 142, No. 2, March 15, 1991
574 TORRES, RUSSEL, AND SCHOWALTER

15. Meakin, P., Chen, Z. Y., and Deutch, J. M., J. Chem. 31. Torres, F. E., Ph.D. dissertation. Princeton University,
Phys. 82, 3786 (1985). Princeton, NJ, 1991.
16. Pusey, P. N., Rarity, J. G., Klein, R,, and Weitz, 32. Van de Hulst, H. C., "Light Scattering by Small Par-
D. A., Phys. Rev. Lett. 59, 2122 (1987). ticles." Dover, New York, 1981.
17. Rogak, S. N., and Flagan, R. C., J. Colloid Interface 33. Berne, B. J., and Pecora, R., "Dynamic Light Scat-
Sci. 134, 206 (1990). tering." Wiley, New York, 1976.
18. Smoluchowski, M. yon, Z. Phys. Chem. 92, 129 34. Lindsay, H. M., Klein, R., Weitz, D. A., Lin, M. Y.,
(1917). and Meakin, P., Phys. Rev. A Gen. Phys. 38, 2614
19. Zeichner, G. R., and Schowalter, W. R., AIChE J. (1988).
23, 243 (1977). 35. Lindsay, H. M., Lin, M. Y., Wetiz, D. A., Ball, R. C.,
20. Feke, D. L., and Schowalter, W. R., J. Fluid Mech. Klein, R., and Meakin, P., in "OSA Proceedings
133, 17 (1983). on Photon Correlation Techniques and Applica-
21. Batchelor, G. K., J. FluidMech. 74, 1 (1976). tions," Vol. 1, p. 122. Washington, DC, May 31-
22. Schaefer, D. W., Martin, J. E., Wiltzius, P., and Can- June 2, 1988.
nell, D. S., Phys. Rev. Lett. 52, 2371 (1984). 36. Lin, M. Y., Lindsay, H. M., Weitz, D. A., Ball, R. C.,
23. Weitz, D. A., and Oliveria, M., Phys. Rev. Lett. 52, Klein, R., and Meakin, P., Proc. R. Soc. A 423, 71
1433 (1984). (1989).
24. Aubert, C., and Cannell, D. S,, Phys. Rev. Lett. 56, 37. McQuarrie, D. A., "Statistical Mechanics." Harper &
738 (1986). Row, New York, 1976.
25. Dimon, P., Sinha, S. K., Weitz, D. A., Safinya, C. R., 38. Batchelor, G. K., and Green, J. T., J. Fluid Mech.
Smith, G. S., Varady, W. A., and Lindsay, H. M., 56, 375 (1972).
Phys. Rev. Lett. 57, 595 (1986). 39. Adler, P. M., J. Colloid Interface Sci. 84, 461 ( 1981 ).
40. Feke, D. L., Ph.D. dissertation. Princeton University,
26. Matsushita, M., Hayakawa, Y., Sumida, K., and Sa-
Princeton, NJ, 1981.
wada, Y., in "Science on Form: Proceedings of the
41. Jeffrey, D. J., and Onishi, Y., J. FluidMech. 139, 261
First International Conference for Science on
(1984).
Form" (Y. Kato, R. Takaki, and J. Toriwaki, Eds.).
42. Leis, J. R., and Kramer, M. A., A C M Trans. Math.
KTK Scientific Pub., Tokyo, 1986.
Software 14, 45 (1988).
27. Meakin, P., J. ColloidlnterfaceSci. 102, 491 (1984). 43. Leis, J. R., and Kramer, M. A., A C M Trans. Math.
28. Jullien, R., Kolb, M., and Botet, R., J. Phys. Lett. 45, Software 14, 61 (1988).
L211 (1984). 44. Hindmarsh, A. C., ACM-SIGNUM Newslett. 15(4),
29. Zeichner, G. R., Ph.D. dissertation. Princeton Uni- 10 (1980).
versity, Princeton, N J, 1978. 45. Lin, M. Y., Klein, R., Lindsay, H. M., Weitz, D. A.,
30. Schlichting, H., "Boundary-Layer Theory." McGraw- Ball, R. C., and Meakin, P., J. Colloid Interface
Hill, New York, 1979. Sci. 137, 263 (1990).

JournalofColloidandInterfaceScience.Vol,142,No. 2, March15, 1991 /

You might also like