You are on page 1of 11

JOURNAL OF COLLOID AND INTERFACE SCIENCE 180, 360370 (1996)

ARTICLE NO. 0314


Low-Shear Viscosities of Dilute Dispersions of Colloidal
Rodlike Silica Particles in Cyclohexane
ANIEKE M. WIERENGA AND ALBERT P. PHILIPSE
1
Vant Hoff Laboratory for Physical and Colloid Chemistry, Utrecht University, Padualaan 8, 3584 CH Utrecht, The Netherlands
Received June 26, 1995; accepted November 13, 1995
The choice for a model system of rodlike particles is an
Low-shear viscosities have been studied for isotropic dispersions
obvious one. The simple-shear viscosity of dilute samples
of uncharged colloidal silica rods in cyclohexane as a function of
of axisymmetric particles has been studied theoretically (6,
volume fraction F and aspect ratio r
p
. Intrinsic viscosities in the
7). For more concentrated systems of long, thin rods (either
limit F r 0 agree with theoretical results [Brenner, H., I nt. J .
semidilute, concentrated isotropic, or nematic dispersions)
Multiphase Flow 1, 195 ( 1974) ] for colloidal cylinders. The F
2
qualitative theories for concentration-dependent diffusion
term in the low-shear viscosity, accounting for two-particle inter-
have been developed (810).
actions, is considerably larger than predicted by D. H. Berry and
Many viscosity studies on systems with organic, rodlike
W. B. Russel [J . FluidMech. 180, 475 ( 1987) ] for dilute hard-rod
macromolecules have been performed. Most of them have dispersions. The discrepancy is very likely due to weak interparti-
cle attractions, which also account for the novel observation of an been carried out on solutions of helix-forming polymers,
isotropicisotropic phase separation in the silica rod dispersions.
such as poly-g-benzyl-L-glutamate (11), and biopolymers,
Our results demonstrate that, in comparison to spheres, high as-
such as tobacco mosaic virus (TMV) (12), FD virus (13),
pect ratios markedly increase both the shear viscosity and its sensi-
or triple-helical polysaccharide Schizophyllan (14). These
tivity to even minor attractions. 1996 Academic Press, Inc.
polymers can be obtained in monodisperse samples, but,
Key Words: low-shear viscosity; colloidal silica rods; isotropic
apart from TMV particles, they all exhibit some degree of
isotropic phase separation; rheology rodlike colloids.
exibility. In addition, the biopolymers are charged, have
xed particle dimensions, and can be obtained only in small
quantities. With respect to inorganic rodlike particles, studies
1. INTRODUCTION
have been reported on bers (15) and whiskers (16). Here
particle dimensions are generally above the colloidal range.
The rheological behavior of dispersions of colloidal
The rheology of suspensions of these non-Brownian rods
spheres with either hard-core interactions, weak attractions,
differs from that of colloidal dispersions, especially at low
or double-layer repulsions has been studied extensively (1
volume fractions and low shear rates, where Brownian rota-
3). The dependence of the low-shear viscosity on the con-
tions are expected to increase the shear viscosity consider-
centration of hard spheres up to volume fractions of about
ably (7, 17, 18).
0.1 is understood quantitatively (4). For higher concentra-
Rodlike boehmite particles, synthesized following Buin-
tions, the dependence can be described qualitatively (5).
ing et al. (19), form stable dispersions of charged inorganic
Since many (inorganic) colloidal particles in nature and
needles in water. We performed preliminary rheological ex-
technology are not spherical, it is important to know the
periments on these systems which were difcult to interpret
inuence of particle anisotropy on rheological properties of
because the rheological behavior was found to be strongly
the dispersions, such as the low-shear viscosity. However,
time dependent. This complicated behavior is possibly
the rheological behavior of such practical suspensions,
caused by particle surfaces carrying either opposite charges
like dispersions of clay particles, is generally difcult to
or different charge densities. Recently, Philipse et al. (20)
study, due to the presence of electrical double layers, surface
developed a method to disperse these needles in nonaqueous
charge inhomogeneities, contaminants, and a large spread in
solvents by rst precipitating silica on the boehmite surface
both particle shapes and dimensions. Therefore, to study the
and subsequently grafting polymers on the silica shell. These
effect of particle anisotropy separately, a model dispersion,
grafted inorganic rodlike particles, dispersed in cyclohexane,
containing particles of fairly monodisperse size and shape,
have no signicant surface charge and are rigid. These prop-
is required.
erties make the silica rods promising for investigating
the inuence of particle shape.
1
To whom correspondence should be addressed. We compare the low-shear intrinsic viscosity of silica rods
360
0021-9797/96 $18.00
Copyright 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.
AID JCIS 4137 / 6g0f$$$321 05-19-96 20:36:49 coida AP: Colloid
361 SILICA ROD VISCOSITY
in cyclohexane to theoretical predictions of Brenner (6) for tion to the dispersion viscosity can be expressed in material
tensors, which depend on the exact particle shape. The mate- cylinders and spheroids. Furthermore, we investigate the appli-
cability of the theory of Berry and Russel (17) on the effect rial tensors can be calculated by solving Stokes equations
with appropriate boundary conditions for the exact particle of two-particle interactions on the low-shear viscosity in dilute
dispersions of hard rods. To clearly establish the role of particle shape equations [ for cylinders and spheroids see (24) and
(21)]. In this paper, the theoretical values for the intrinsic shape we have studied two dispersions of silica rods with aver-
age particle aspect ratios of, respectively, r
p
5.6 and r
p
22.4. viscosities, given in Table 2, are calculated from expressions
for material tensors as given in the extensive survey on The time-consuming synthesis and purication procedures (20)
obstruct a systematic variation of r
p
. We therefore choose the rheology of dilute dispersions of axisymmetric particles by
Brenner (6). two extremes in aspect ratio that can be achieved with current
synthesis procedures (19, 20). The rotary Peclet number Pe
rot
( g
h
/ D
rot
) compares the
Apart from the low-shear viscosity as a function of rod shear rate to the rotational diffusion coefcient D
rot
. In the
volume fraction for the separate dispersions, mixtures of the limit of dominant Brownian motion, i.e., when Pe
rot
0,
dispersions were also investigated. This was done to amplify evaluation of the particle stresslets for spheroidal particles
the particle polydispersity in a controlled way, and to see if with very large aspect ratios yields (6)
the inuence of a distribution of aspect ratios can be under-
stood. The dispersions were also studied with light scattering
[h]
0
to detect qualitatively the presence of any rod clusters. Fur-
thermore, the colloidal stability of dispersions was checked

r
2
p
5

1
3(ln 2r
p
0 1.5)
/
1
ln 2r
p
0 0.5

/ 1.6. [2.2]
over extensive periods, before and after rheological experi-
ments. Any indications for other than excluded volume inter-
This expression was already derived in 1945 by Kuhn and actions, of course, must be taken into account when interpre-
Kuhn (7), who calculated the energy dissipation of a disper- ting rheological data.
sion of spheroids using a two-dimensional description for Before discussing experiments (Section 3) and results
the ow eld around the rods. (Section 4), we briey explain the background of equations
with which we compare our measurements.
2.2. Effect of Two-Rod Interactions
on the Shear Viscosity
2. SHEAR VISCOSITY OF A DILUTE DISPERSION
OF HARD RODS
Hydrodynamic interactions will change the hydrodynamic
particle stress since they modify the ow eld around the
2.1. Intrinsic Viscosity of Noninteracting Rods
dispersed particles. The orientation distribution function
P(e) is affected by the excluded volume of interacting rods.
The average contribution of one dispersed particle to the
The effect of two-particle interactions on the hydrodynamic
dispersion viscosity is called the intrinsic viscosity, [h],
stress has been evaluated by Berry and Russel (17). In their
which is dened as
calculation they used a far-eld approximation to describe
both the altered ow eld and the excluded volume.
[h]
h
r
0 1
F

h
s
S
g
h
. [2.1]
In the far-eld approximation, which is assumed to be
valid if the shortest distance between two rods is much larger
than the rod diameter, only leading terms in the two-particle
Here, h
r
is the dispersion viscosity, relative to the viscosity
interaction are taken into account. The effect of the presence
h
s
of the solvent, F is the particle volume fraction, S is
of a second rod j on the angular velocity of rod i can be
the average particle stresslet, and g
h
is the shear rate. In a
written as
dilute dispersion of noninteracting axisymmetric particles,
which are not subjected to external forces or couples, the
V
i
V
i0
/
3
2

1
01
ds
i
s
i
(u
i
0 (u
j
re
i
)e
i
). [2.3] contribution of the particles to the dispersions shear viscos-
ity can be written as the sum of a shear term and a Brownian
term. The particle shear stresslet is related to the motions of
Here, V
i 0
is the angular velocity of particle i in the absence
the particle as imposed by the shear eld, following the
of particle j , and the integral term denotes the disturbance
equations of motion as derived by Jeffery (21). The
of this velocity as caused by the ow eld of particle j ,
Brownian stresslet is due to the Brownian rotational motion
with u
j
the disturbance velocity:
of the particles which restores the initial isotropic orienta-
tional distribution (22, 23).
The relation between both the shear rate, g
h
, and the
u
j

1
3 ln r
p
O

V
j
dS f (r )rT(r 0 r ) [2.4]
Brownian rotational velocity, V
Br
, and the particle contribu-
AID JCIS 4137 / 6g0f$$$322 05-19-96 20:36:49 coida AP: Colloid
362 WIERENGA AND PHILIPSE
with f (r ) the force per unit area exerted by a surface ele- ow of demineralized water to remove soluble silicates. The
silica surface of the particles was grafted with stearyl alcohol ment of rod i on the uid, and T(r 0 r ) the Oseen tensor.
Berry and Russel (17) argued that in the case of far-eld and nally dispersed in cyclohexane (20).
interactions, the excluded volume of a single rod can be
approximated by a preaveraged form, obtained by averaging 3.1.2. Characterization
the excluded volume of a rod with arbitrary orientation over
3.1.2a. Transmission electron microscopy. From trans-
all orientations. For Pe
rot
0, the shape of this average is
mission electron micrographs (see Fig. 1), the particle
a sphere. If a weak ow eld is present, such that Pe
rot

widths and lengths were measured using interactive image
1, the sphere is slightly deformed to an ellipsoid. Using these
analysis. The dimensions of both boehmite cores, BT2 and
approximations Berry and Russel (17) found an expression
BU1, were obtained from 100 particle counts. Using trans-
for the viscosity of dilute dispersions of interacting rods:
mission electron microscopy (TEM) and X-ray powder dif-
fraction, Buining et al. (19) were also able to determine the
approximate thickness of the boehmite crystals (8 nm). For
h
dispersion
h
0
1 / (1 0 0.020 Pe
2
rot
)[h]
0
F
the grafted rods, BT2S2C18 and BU1S3C18, 500 particle
counts were made for both length and width. Assuming that
/ k
H
(1 0 0.0142 Pe
2
rot
)[h]
2
0
F
2
. [2.5]
the average increase of this thickness, caused by the precipi-
tation of sodium silicate, is the same as that of the width of
Inspection of [2.5] reveals that the theory of Berry and
the particles, we have made an estimate for the thickness of
Russel predicts a shear-thinning behavior for dilute disper-
our particles. Since the stearyl alcohol layer cannot be ob-
sions of axisymmetric particles. Their calculations show that
served by TEM, the particle dimensions we obtained are
the value of the Huggins parameter is k
H
0.4. In the far-
those of the bare silica rods. The thickness of the grafted
eld approximation of Berry and Russel the pair interaction
layer is approximately 1.5 nm.
term for rods is dominated by Brownian torques. This is
In Table 1 the so-obtained particle dimensions are shown.
why we prefer to compare our results for colloidal rods with
The average aspect ratio r
p
is the quotient of the average
their calculations, instead of the more detailed numerical
length and width of the particles, both corrected for the
results of Claeys and Brady (18), which do not include
presence of the grafted layer. These aspect ratios are used for
Brownian motions.
calculations of the viscosities. Although for the BU1S3C18
particles, the thickness of the particles seems to be somewhat
3. EXPERIMENTAL
smaller than the width, the difference is not signicant.
Therefore, calculation of the average aspect ratio as de-
3.1. Rodlike Silica Particles
scribed above is a reasonable approximation.
3.1.1. Preparation
3.1.2b. Static light scattering. Static light scattering
measurements of dilute samples of BU1S3C18 were per-
The preparation of the colloids, described extensively in
formed on a Fica-50 light scattering photometer using verti-
Refs. (19, 20), can be summarized as follows. Aqueous
cally polarized incident and detected light with a wavelength
dispersions of crystalline boehmite (AlOOH) needles, coded
of 546 nm. Light scattering curves were calculated from the
BU1, with an average length of approximately 100 nm, were
particle dimensions of the boehmite cores as obtained from
synthesized from a mixture of aluminum tri-sec-butoxide
TEM, following Buitenhuis (25), using the coupled dipole
(ASB, Fluka) and HCl in demineralized water, which was
method for isolated nite cylinders. Figure 2 shows the mea-
stirred for 1 week and then autoclaved for 22 h at 150C.
sured Rayleigh ratios and the calculated curves, which are
Crystalline boehmite needles with an average length of 290
scaled to coincide with the experimental values at high wave-
nm, coded BT2, were prepared by stirring a mixture of ASB,
number k. At low k, the measured intensities are much larger
aluminum triisopropoxide (Janssen Chimica), and HCl in
than predicted by Buitenhuis (25). This discrepancy is very
demineralized water for 1 week. Here, 1.5 h before autoclav-
likely due to the presence of some clusters, which can be
ing (22 h, 150C), a small amount of tetraethoxysilane
either permanent clusters, formed during the synthesis and
(Merck) was added to the alkoxide mixture (20), which
grafting of the silica rods, or temporary aggregates of rods
promotes the formation of particles with high aspect ratios.
resulting from attractions between the grafted silica particles.
To remove alcohols and electrolyte, both dispersions were
dialyzed for 2 weeks in cellophane tubes against demineral-
3.2. Rheological Measurements
ized water. A silica layer a few nanometers thick was precipi-
tated on the boehmite cores in water by slowly adding a
3.1.1. The Rheometer
3% sodium silicate solution (pH 11.5) to the dialyzed
dispersion (20). The silicaboehmite dispersions were The viscosity measurements are performed on a Contraves
LS40 rheometer, a constant-shear apparatus, with a Couette again dialyzed for 2 weeks in cellophane tubes against a
AID JCIS 4137 / 6g0f$$$322 05-19-96 20:36:49 coida AP: Colloid
363 SILICA ROD VISCOSITY
provided with a vapor lock. The rheometer was calibrated
using viscosimetric reference liquids (Nederlands Meet In-
stituut, NMI, Delft, The Netherlands). During the measure-
ments, the temperature within the rheometer was kept 25C
({0.1C) by a thermostatic bath that circulated water around
the measuring geometry.
3.2.2. Measuring Procedure
For each series of measurements, the vapor lock was lled
with 0.5 ml of a viscosimetric reference liquid with h
0.699 mPars. Then 1 ml of the sample was inserted in the
measuring geometry, using a Finn pipet. In every run, the
shear stress was measured for 50 different shear rates, evenly
spaced between 0.5 and 50 s
01
. For dilute dispersions, this
shear rate corresponds to rotational Peclet numbers of
9.310
05
Pe
rot
9.310
03
for BU1S3C18 and 1.210
03
Pe
rot
1.210
01
for BT2S2C18. The slope of a linear
t of the data was dened as the measured viscosity h
m,d
for
the dispersion and h
m,chx
for cyclohexane, which was the
solvent in all cases. This measured viscosity was corrected
for the additional friction caused by the presence of the vapor
lock, h
vl
. This friction was determined by performing two
runs without lling the geometry, one run using a geometry
without vapor lock, and a second run in a geometry provided
with a vapor lock, lled in the way described above. The
slopes of the ow curves were interpreted as the viscosities
of air, measured either with (h
air,vl
) or without (h
air
) a vapor
lock. The relative viscosity was nally obtained from
h
r

h
m,d
0 h
vl
h
m,chx
0 h
vl

h
m,d
0 (h
air,vl
0 h
air
)
h
m,chx
0 (h
air,vl
0 h
air
)
. [3.1]
3.2.3. Sample Preparation
Samples were prepared by dilution of weighted amounts
of concentrated stock dispersions with weighted amounts of
cyclohexane. Both the volume fractions of the stock disper-
sions and the mass densities of the particles were obtained
from weighing and drying a known volume of dispersion.
The mass densities of BU1S3C18 and BT2S2C18 particles
were found to be 1.23 and 1.64 g/ml, respectively. Philipse
et al. (20) also determined the mass density of BT2S2C18
from elemental analysis, to nd 1.6 g/ml.
FIG. 1. (a) Transmission electron micrograph of BU1S3C18, silica
rods with boehmite core, average length L 90.9 nm, average width
D 13.7 nm. (b) Transmission electron micrograph of BT2S2C18, silica
4. RESULTS
rods with boehmite core, average length L 292 nm, average width D
10.2 nm.
4.1. Colloidal Stability of the Dispersions
The initial stock dispersions of both types of grafted rods
geometry (MS 41S/ 1S). This equipment is particularly suit- in cyclohexane did not show any sign of occulation for a
able for low-shear viscosity measurements. Using the geom- period of at least 18 months. (The stock concentrations are
etry depicted in Fig. 3, the rheometer is appropriate for the about F 5%.) Diluted samples of BT2S2C18 remained
stable for at least half a year. The diluted samples of shear rate range 210
04
g
h
110
2
s
01
and for the shear
stress range 910
04
s 5.50 Pa. The geometry was BU1S3C18, however, only remained stable for approxi-
AID JCIS 4137 / 6g0f$$$323 05-19-96 20:36:49 coida AP: Colloid
364 WIERENGA AND PHILIPSE
TABLE 1
Particle Dimensions of the Boehmite Cores and Grafted Silica Rods as Obtained from Transmission Electron Micrographs
Length s
length
a
WIdth s
width
a
Thickness
b
Length/width
System (nm) (%) (nm) (%) (nm) (TEM) r
p
c
BU1
d
100 48 10.1 29 8 9.9
BT2
d
288.8 38 7.7 26 8 37.5
BU1S3C18
e
90.9 25 13.7 24 11.6 6.6 5.6
BT2S2C18
e
292 25 10.2 22 10.5 28.6 22.4
a
s
(x)
denotes standard deviation in x.
b
Estimate, based on observations of Buining et al. (19).
c
Estimated aspect ratio, including grafted organic layer.
d
Boehmite cores.
e
Grafted silica rods.
mately 6 weeks, after which they separated into a clear upper dispersions with higher volume fractions from the drawn
line. [This behavior is also predicted by Eq. [2.5].] layer, which hardly scattered any light, and a turbid, very
viscous bottom layer. This concentrated bottom layer of the The Huggins coefcient k
H
can be deduced from C
2
using
(compare Eq. [2.5]) dispersions did not exhibit optical birefringence; thus rods
are randomly oriented. The bottom layer could be redis-
persed by gently shaking the sample tube. After that, how-
C
2
k
H
[h]
2
0
. [4.1]
ever, the separation reappeared within a few days. The vol-
ume of the bottom layer depended on the particle concentra-
In Table 2 the so-obtained values for the extrapolated intrin-
tion of the sample: it was the smallest for the lowest volume
sic viscosities, [h]
0
, are compared with theoretical values as
fraction and the largest for the highest volume fraction of
calculated from the expressions evaluated by Brenner (6)
BU1S3C18. Samples of BT2S2C18 with volume fractions
for spheroids (exact ) and cylinders (only exact for r
p

below 0.05% showed a similar separation behavior.
1) from the number-average dimensions determined from
The rheological experiments also appeared to inuence
transmission electron micrographs. The experimental Hug-
the stability of both types of rod dispersions. After a shear
gins coefcients are in Table 2 compared with the value of
experiment, no immediate change in the appearance of the
0.4, calculated by Berry and Russel (17).
dispersions was observed. After a few weeks though, the
sheared samples (of both types of grafted rods) became less
4.3. Effects of Polydispersity on the Intrinsic Viscosity
translucent than the nonsheared samples. After a few months,
at Innite Dilution
most of these samples also contained small pieces of gelated
The effect of the particle size distribution on the intrinsic
dispersion adhering to the walls of the test tubes in which
viscosity at innite dilution was studied by measuring ow
they were stored. By shaking intensively, these dispersions
curves of mixtures of BT2S2C18 and BU1S3C18. From
could be homogenized for a short while, after which the
these curves, the intrinsic viscosity was calculated as de-
small ocs reappeared.
scribed above. The results, shown in Fig. 5, indicate that
[h]
0
can be simply calculated by proportional addition of
4.2. Intrinsic Viscosity as a Function of Volume Fraction
the contributions of both systems:
and Particle Size
In the applied shear rate range, the samples of BT2S2C18
[h]
0
with volume fractions F 0.006 and BU1S3C18 with F
0.018 showed Newtonian behavior. The slope of the ow

F
BU1
F
BU1
/ F
BT2
[h]
0,BU1
/
F
BT2
F
BU1
/ F
BT2
[h]
0,BT2
. [4.2]
curves corresponds to the dispersion viscosity. For both sys-
tems, the intrinsic viscosity is plotted against the particle
volume fraction in Fig. 4. The intercept of the linear regres- Note that for the separate components the polydispersity
is small in comparison with the mixtures. This allows an sion t through the experimental data is interpreted as the
intrinsic viscosity at innite dilution in the low-shear limit. estimation (see the Appendix) of [h]
0,BU1
and [h]
0,BT2
on
the basis of number averages, since the correlation between The slope, C
2
, of that curve is a measure of the inuence
of two-particle interactions in the system. Shear-thinning particle length and diameter is weak. [Buitenhuis (25) found
a correlation coefcient of about 0.5.] rheological behavior seems to be a plausible explanation for
the deviation of the measured intrinsic viscosities of the Values for the Huggins coefcients were again estimated
AID JCIS 4137 / 6g0f$$$323 05-19-96 20:36:49 coida AP: Colloid
365 SILICA ROD VISCOSITY
smooth, monodisperse cylinders as input parameters. Never-
theless, both types of grafted rods have a polydispersity of
about 25%. In addition, transmission electron micrographs
of the particles show a considerable surface roughness (Fig.
1) and the static light scattering measurements (Fig. 2) indi-
cate the presence of some particle clusters in both types of
dispersions. These complicating factors apparently hardly
affect [h]
0
for reasons that can be understood as follows.
The experiments with the bidisperse samples indicate that
the zero-shear intrinsic viscosity of an innitely dilute sam-
ple of polydisperse particles is simply the volume average
of the intrinsic viscosities of the individual particles present
(Fig. 5), as one would expect from the denition of [h] in
Eq. [2.1]. For the separate components, BU1S3C18 and
BT2S2C18, we estimate that this volume average is only
about 10% larger than the number-average viscosity (Appen-
dix). Therefore, for our dispersions, extrapolation to F 0
of the zero-shear intrinsic viscosity yields a value that is
comparable with the theoretical intrinsic viscosity (6) as
calculated on the basis of number-average particle dimen-
sions as observed by TEM.
Furthermore, the surface roughness of the particles does
not seem to complicate the ow eld around the particles.
One can imagine that because of stagnant solvent in the
surface irregularities, the particles have a fairly smooth hy-
drodynamic shape.
Finally, any permanently clustered particles will have a
more pronounced effect on static light scattering than on
viscosity measurements. In light scattering, at small scatter-
ing angles, the contribution of a certain class of particles to
FIG. 2. Static light scattering curves (l) of BU1S3C18 ( L 90.9
nm), at a volume fraction F 0.02% (a) and BT2S2C18 ( L 292 nm),
at F 0.003% (b), compared with theoretical curves ( ) for free cylin-
ders according to Buitenhuis (25). Here, R(k) is the Rayleigh ratio and k
the wavenumber.
from the slope of the intrinsic viscosityvolume fraction
curves. Table 3 shows that, as for pure dispersions, the exper-
imental Huggins coefcients differ considerably from the
theoretical value 0.4 of Berry and Russel (17).
5. DISCUSSION
5.1. Intrinsic Viscosities
For both types of grafted silica rods, the experimental
FIG. 3. LS40 Rheometer with MS 41S/ 1S geometry () and vaporlock
intrinsic viscosity at innite dilution [h]
0
is in rather good
(). The radius of the bob is 5.5 mm, the inner radius of the cup is 6.5
agreement with the value calculated from the expressions as
mm, the height of the geometry is 8 mm, and the volume of the space
given by Brenner (6). In these calculations we use number-
between the cup and the bob is 0.3 ml. During the measurements, the vapor
lock is lled with 0.5 ml of a reference liquid (h 0.699 mPars). averaged TEM dimensions and the material tensors for
AID JCIS 4137 / 6g0f$$$323 05-19-96 20:36:49 coida AP: Colloid
366 WIERENGA AND PHILIPSE
FIG. 4. Intrinsic viscosity versus volume fraction in low-shear Newtonian region for both BU1S3C18 () (with average length L 90.9 nm and
average aspect ratio r
p
5.6) and BT2S2C18 (l) ( L 292 nm, r
p
22.4), measured in the shear rate range 0.550 s
01
. The results illustrate the
strong inuence of the aspect ratio, as explained in the text.
the scattered intensity is proportional to the number density We note here that the contributions of permanent and
temporary clusters to the light scattering curves cannot be and the squared particle volume. So, a small volume fraction
of large clusters will inuence the light scattering curve at separated. Therefore, neither the number of permanent clus-
ters nor the strength of attractions between the grafted silica low k values signicantly. (At higher angles, the interparticle
interference reduces the intensity scattered by large objects.) rods can be extracted from the static light scattering data, a
point to which we will return to in the next section. The contribution of permanent clusters to the viscosity of a
dispersion can be written as (Eq. [4.2])
5.2. Interaction Effects
The experimental Huggins coefcients (see Table 2) are Dh(due to clusters) [h]
clusters
F
clusters
. [5.1]
approximately 10 times larger than the value 0.4, predicted
by the theory of Berry and Russel (17), which takes both If the volume fraction of clusters is small, and the intrinsic
viscosity of the clusters does not differ very much from the hydrodynamic interactions and Brownian motion into ac-
count. Furthermore, this theoretical value is derived for hard intrinsic viscosity of the single particles (comparable aspect
ratios), the inuence of the presence of permanent clusters rods without attractions or additional repulsions. A compari-
son of these assumptions and properties of our dispersion on the measured intrinsic viscosity will indeed be small.
TABLE 2
Experimental Values of the Intrinsic Visocisity [h]
0
, the Pair Interaction Term C
2
,
and the Huggins Coefcient k
H
for Grafted Silica Rods in Cyclohexane
[h]
0
[h]
0
b
[h]
0
b
K
H
c
K
H
d
Sample (exp.) s
([h]
0
)
a
(cylinder) (spheroid) C
2
s
(C2)
a
(exp.) (B/R)
BU1S3C18 12.79 0.86 10.0 3.46 692.4 85.5 4.23 0.4
BT2S2C18 50.17 2.35 47.3 46.2 8697 1004 3.49 0.4
a
s
(x)
denotes standard deviation in x.
b
Calculated from the average TEM dimensions, using the expressions as given by Brenner (6).
c
Calculated from the two-particle interaction term C
2
, using Eq. [4.1].
d
As predicted by the theory of Berry and Russel (17).
AID JCIS 4137 / 6g0f$$$324 05-19-96 20:36:49 coida AP: Colloid
367 SILICA ROD VISCOSITY
FIG. 5. Intrinsic viscosity (), measured in the shear rate range 0.550 s
01
, of mixtures of both types of grafted rods (BU1S3C18 and BT2S2C18)
versus the contribution of BT2S2C18 to the total particle volume fraction. The straight line represents the expected values based on Eq. [4.2], i.e.,
proportional addition of the experimental intrinsic viscosities of the separate components ( l), as obtained from Fig. 4.
suggests that the origin of the discrepancy is the presence separation in layers on a time scale of months. The upper
of other than hard-rod interactions, i.e., either weak short- layer is a very dilute dispersion, a colloidal rod gas. The
range attractions or long-range repulsions between the silica bottom layer, however, is a concentrated uid that can be
rods.
easily redispersed and in which no directional ordering of
Double-layer repulsions are absent for these grafted rods
the silica rods occurs (no birefringence). Probably, the sepa-
in cyclohexane. The grafted layer (thickness 1.5 nm)
ration in layers is a gasliquid phase separation for iso-
contains octadecyl chains for which cyclohexane is a good
tropic rods, which should be clearly distinguished from the
solvent, so this layer causes at most a slight increase in the
more familiar isotropicnematic phase separation (26). In-
effective rod diameter. Therefore, the large Huggins coef-
dications for a phase separation are the increase in volume
cients very likely reect the inuence of attractions between
of the concentrated phase with the volume fraction of the
the rods.
starting suspension, the reproducibility of the phase volume,
The presence of weak short-range attractions is indicated
and the reversible nature of the phenomenon. Such an iso-
by the change in appearance of the diluted dispersions, as
tropic phase behavior for rods seems not to have been recog-
described in Section 4.1. The attractions are not strong
nized earlier. Probably this behavior will occur only within
enough to induce immediate occulation, but do cause a
a narrow range of attractions. Above this range (space-
lling) gel formation can be expected. The weakness of the
attraction, and thus the subtlety of its inuence on the Hug-
TABLE 3
gins coefcient, is also supported by the following ndings.
Experimental Intrinsic Viscosity at Low Shear, [h]
0
, Two-Parti-
cle Interaction Term C
2
, and the Huggins Coefcient k
H
for Mix- We observe a time-independent viscosity for all samples
tures of BU1S3C18 and BT2S2C18
and a Newtonian behavior for most samples at the applied
shear rates. Furthermore, we nd a linear volume fraction
dependence of the intrinsic viscosity at low volume fractions.
F(BT2S2C18)/F
tot
[h]
0
s
([h]
0
)
a
C
2
s
(c
2
)
k
H
This rheological behavior suggests that the particles do not
0.109 21.61 4.60 755.7 941.3 1.62 form a permanent network, but that they are only slightly
0.377 26.37 7.19 3478 1244 5.03
sticky. When rods collide with another, they occasionally
0.724 41.30 3.53 6596 1206 3.87
stick together for a short while, before separating again. The
0.288 26.95 3.05 2996 786 4.12
orientation of the particles during this sticking period is not
a
s
(x)
denotes standard deviation in x. likely to be parallel, because the dispersions are isotropic
AID JCIS 4137 / 6g0f$$$324 05-19-96 20:36:49 coida AP: Colloid
368 WIERENGA AND PHILIPSE
in the surface which are accessible for any other particle,
decreasing the distance between the boehmite cores. While
the Hamaker constant of boehmite in cyclohexane (29) is
much larger than the Hamaker constant of silica in cyclohex-
ane (20), at such pits the attractive interaction between the
particles would thus locally be increased.
Since the number of possible contact points per particle
is limited, the rods will form an open network structure,
which will ll up the whole dispersion as long as the particle
concentration is high enough (as in the stock dispersions),
but is not strong enough to cover the whole dispersion vol-
ume at lower concentrations. In dilute dispersions this effect
leads to a separation in layers. In the BT2S2C18 systems,
the separation in layers has been observed only at extremely
low volume fractions. Those particles are longer than the
BU1S3C18 particles and their surface is somewhat smoother
(Fig. 1) and is thus providing a smaller number of possible
contact points than that of the BU1S3C18 particles. These
features will promote the development of an even more open
FIG. 6. Schematic drawing of two silica rods, showing how surface
network structure than that being formed in the
irregularities of rodlike particles can act as contact points for short-time
BU1S3C18 systems, which will possibly be able to ll up
sticking or long-time aggregation. Any small attraction resulting from such
larger volumes.
irregularities may increase the pair interaction term in the viscosity consider-
As far as the authors are aware, a reversible separation ably.
into two isotropic layers, as observed in the described disper-
sions, has not been predicted for rods with aspect ratios r
p
at these volume fractions and low shear rates. Moreover,
3.5. The reason for the occurrence of an isotropiciso-
irregularities on the surface will obstruct the approach of two
tropic phase separation in our dispersions may be the nature
rodlike particles in a parallel way (Fig. 6). The weakness of
of the attractions. As mentioned earlier, the silica rods proba-
the attraction also seems to be in accordance with a nonparal-
bly possess local, weakly attractive sites on the particle sur-
lel orientation of aggregated particles, since in a parallel
face rather than a homogeneous attractive well around the
orientation, the van der Waals attractive force between two
whole particle surface, as in the theory of Khokhlov and
rods would be much larger than in a nonparallel orientation.
Semenov (30).
The van der Waals energy between two identical, crossed
These authors have studied the inuence of attractive in-
cylinders is given by (27)
teractions on phase behavior of dispersions of rodlike parti-
cles. In their theory, which is comparable with van der Waals
theory for attractive spheres, the attractions are included in
W

0
AD
12d
, [5.2a]
the free energy by using a concentration- and orientational-
order dependent mean eld potential:
whereas for parallel orientation,
U
att
0u
0
c 0 u
a
c P
2
(cos u) P
2
(cos u). [5.3]
W

0
AL

D
24

2d
3/ 2
. [5.2b]
Here, c denotes the rod concentration, P
2
(cos u) the order
parameter, and u
0
and u
a
are unspecied constants. They
calculated that for rods with an aspect ratio of r
p
20, an In these expressions, A is the Hamaker constant, L and D
the length and diameter (width) of the particles, respectively, attraction force of e 0.25 kT may already induce an iso-
tropicnematic phase separation. For the initiation of an and d the shortest distance between the surfaces of two parti-
cles. For two BT2S2C18 particles at d 3 nm, W

00.28 isotropicnematic phase separation in dispersions of rods


with lower aspect ratios (3.5 r
p
15) they expect at- A, while W

05.3 A, where the Hamaker constant for


silica in cyclohexane at 298 K is A 0.15 kT (28). The traction forces of a few kT (e 2.5 kT) to be sufcient.
This gives some support to the notion that for an isotropic bochmit cores, however, may increase the attractions and
promote clustering. isotropic phase separation in our silica dispersions, indeed,
only weak attractions are required. A nonparallel clustering can be attained in the following
way. The irregularities on the surface may lead to small pits In the absence of theoretical models, it is not possible to
AID JCIS 4137 / 6g0f$$$324 05-19-96 20:36:49 coida AP: Colloid
369 SILICA ROD VISCOSITY
extract the exact strength of the rodrod attraction forces Viscosity measurements on samples of bidisperse grafted
rods show that the intrinsic low-shear viscosity of a mixture from the observed phase behavior. This state of affairs is
typical of the complex nature of the calculation of phase of rods corresponds to the volume-averaged dimensions of
the rods. However, for the separate dispersions the polydis- diagrams for dispersions of attractive rodlike particles. Van
der Schoot and Odijk (31) calculated that attractions that persity is sufciently small to estimate [h]
0
from number-
averaged rod dimensions as obtained from electron micros- are strong enough to cause a phase separation will not only
affect the second virial coefcient, but also have a major copy (Appendix). The fact that the experimental values for
[h]
0
also compare quite well with predictions (Table 2) that contribution to the third and fourth coefcients in the virial
expansion. This is one reason why light scattering proles are based on number-averaged particle dimensions justies
our use of average aspect ratios as input parameters. as shown in Fig. 2 are used by us as a mere qualitative
indication for particle clustering. The concentration dependence of the intrinsic viscosity
for (mixtures of) rod dispersions is a factor of 10 larger The reversible isotropicisotropic separation is no longer
observed after dispersions have been subjected to shear ex- than expected from the calculations of Berry and Russel
(17) for hard rods. Most likely the discrepancy is caused by periments. Instead, long-time aggregation occurs in both sys-
tems, which is rather irreversible and does not lead to a clear the presence of very weak attractions between the rods,
which also give rise to an isotropic phase separation at low separation in layers (Section 4.1.) as does the aggregation
of the unmeasured samples. Imposing a shear eld on the volume fractions.
The rodlike silica colloids in this study are clearly dispersions thus leads to stronger attractions between the
particles. Brownian particles, and their low-shear intrinsic viscosity
(both at innite dilution and in less dilute dispersions) is If the shear forces would induce a signicant acceleration
of the aggregation mechanism that leads to the phase separa- much higher than has been found for non-Brownian bers
[e.g., by Milliken et al. (33)]. Comparing our results with tion as described above, the buildup of the network would
cause a time-dependent rheological behavior, which has not sphere rheology (1), we nally conclude that high aspect
ratios considerably increase the magnitude and concentration been observed. Possibly, the shearing affects the grafted
stearyl alcohol layer which provides a barrier against irre- dependence of the viscosity and, at the same time, amplify
the inuence of weak attractions. The practical implication versible aggregation in the unsheared dispersions. Shear deg-
radation of the stearyl alcohol layer has also been observed is that uncharged Brownian bers will be very effective as
thickeners. The theoretical challenge is to incorporate in sphere dispersions (32).
For dispersions of adhesive spheres, Woutersen and de small attractions into the Huggins coefcient.
Kruif (3) found a trend in the inuence of attractions on the
F
2
term, which is similar to our observations for rods. The
APPENDIX: INFLUENCE OF POLYDISPERSITY ON
volume fraction dependence of their dilute dispersions could
THE ZERO-SHEAR INTRINSIC VISCOSITY AT
be described by Einsteins formula for hard spheres: h
r
1
INFINITE DILUTION FOR A DISPERSION
/ 2.5F. At higher volume fractions (F 0.025), however,
OF RODLIKE PARTICLES
the attractive interactions increased the F
2
coefcient in the
expansion of the relative viscosity with a factor 1.042.9,
Equivalent with Eq. [4.2], one may write the zero-shear
depending on the temperature. Due to the larger excluded
intrinsic viscosity of an innitely dilute dispersion as
volume of rodlike particles, two-particle interactions in our
dispersions occur at much lower volume fractions and, fur-
[h]
0

i
n
i
V
p,i
[h]
0,i
/
i
n
i
V
p,i
. [A.1]
thermore, their effect on the viscosity is much more pro-
nounced than in Woutersens sticky sphere dispersions.
Here, n
i
denotes the number of particles with volume V
p,i
6. CONCLUSIONS
p/ 4L
i
D
2
i
and intrinsic viscosity [h]
0,i
. In absence of a
signicant correlation between particle length and width, Eq.
The rheological behavior of the particles at innite dilu-
[A.1] can be written as follows for large aspect ratios r
p
tion compares very well with the behavior that is expected
(compare [2.2]):
for colloidal cylindrical particles (6). Surface roughness and
some rod clusters do not seem to affect the viscosity. The
agreement with calculations for spheroidal particles is less
[h]
0

n L
3

n L D
2

1
5

1
3(ln r
p
0 1.5)
good, in particular for the BU1S3C18 particles with the
lowest aspect ratios. This is in accordance with the transmis-
sion electron micrographs, which also show that the variation
/
1
ln 2r
p
0 0.5

/ 1.6. [A.2]
of the diameter with the position along the main axis is not
related to a spheroidal shape.
AID JCIS 4137 / 6g0f$$$324 05-19-96 20:36:49 coida AP: Colloid
370 WIERENGA AND PHILIPSE
5. Krieger, I. M., and Dougherty, T. J., Trans. Soc. Rheol. 3, 137 (1959).
For a narrow distribution of X, the normalized moments M
n
6. Brenner, H., Int. J. Multiphase Flow 1, 195 (1974).
X
a
/ X
a
can be approximated by (34)
7. Kuhn, W., and Kuhn, H., Helv. Chim. Acta 37, 97 (1945).
8. Doi, M., and Edwards, S. F., J. Chem. Soc. Faraday Trans. 1, 74, 918
(1978); ibid. 74, 560 (1978).
M
n
1 /
a(a 0 1)
2
s
2
/ rrr, s 1, [A.3]
9. Jain, S., and Cohen, C., Macromolecules 14, 759 (1981).
10. Fixman, M., Phys. Rev. Lett. 54, 337 (1985); ibid., 55, 2429 (1985).
11. Yang, J. T., J. Am. Chem. Soc. 80, 1783 (1958); Hermans, J., Jr., J.
Colloid Sci. 17, 638 (1962); Ookubo, N., Komatsubara, M., Nakajima, where s M
2
0 1. Applying Eq. [A.3] to Eq. [A.2] yields
H., and Wada, Y., Biopolymers 15, 929 (1976).
12. Nemoto, N., Schrag, J. L., Ferry, J. D., and Fulton, R. W., Biopolymers
14, 409 (1975); Reinhardt, U. T., Meyer de Groot, E. L., Fuller,
[h]
0
(1 / 2s
2
)
L
2
D
2
1
5

1
3(ln r
p
0 1.5) G. G., and Kulicke, W. M., Macromol. Chem. Phys. 196, 75 (1995).
13. Graf, C., Kramer, H., Deggelman, M., Hagenbuchle, M., Johner, C.,
Martin, C., and Weber, R., J. Chem. Phys. 98, 4929 (1993).
14. Enomoto, H., Einaga, Y., and Teramoto, A., Macromolecules 17, 1573
/
1
ln 2r
p
0 0.5

/ 1.6. [A.4]
(1984).
15. Powell, R. L., J. Stat. Phys. 62, 1073 (1991).
16. Tsao, I., and Danforth, S. C., J. Am. Ceram. Soc. 76, 1977 (1993).
So the proper volume-averaged intrinsic viscosity is a factor
17. Berry, D. H., and Russel, W. B., J. Fluid Mech. 180, 475 (1987).
18. Claeys, I. L., and Brady, J. F., J. Fluid Mech. 251, 443 (1993).
of order 1 / 2s
2
1.13 larger than the value calculated in
19. Buining, P. A., Pathmamanoharan, C., Jansen, G. B. H., and Lekkerk-
Table 2 from number averages L and D .
erker, H. N. W., J. Am. Ceram. Soc. 74, 1303 (1991).
20. Philipse, A. P., Nechifor, A. M., and Pathmamanoharan, C., Langmuir
ACKNOWLEDGMENTS
10, 4451 (1994).
21. Jeffery, G. B., Proc. R. Soc. London Ser. A 102, 163 (1923).
22. Giesekus, H., Rheol. Acta 2, 1, 50 (1962). We thank A. M. Nechifor and C. Pathmamanoharan for providing us
23. Batchelor, G. K., J. Fluid Mech. 41, 3, 545 (1970). with the BT2S2C18 and BU1 rod systems and TEM results. Dr. J. K. G.
24. Cox, R. G., J. Fluid Mech. 44, 4, 791 (1970). Dhont is thanked for his helpful comments on Section 2, and Dr. G. J.
25. Buitenhuis, J., Thesis, Utrecht University, Ch. 4, 1994.
Vroege and Professor H. N. W. Lekkerkerker are thanked for their remarks
26. Onsager, L., Ann. NY Acad. Sci. 51, 627 (1949).
on the effects of attractive interactions on the phase behavior of rod disper-
27. Israelachvili, J., Intermolecular and Surface Forces, 2nd ed. Aca-
sions. This work was supported by The Netherlands Foundation for Chemi-
demic Press, London, 1991.
cal Research (SON) with nancial aid from The Netherlands Organization
28. Jansen, J. W., de Kruif, C. G., and Vrij, A., J. Colloid Interface Sci.
for Scientic Research (NWO).
114, 471 (1986).
29. Buining, P., thesis, Utrecht University, Ch. 5, 1992.
REFERENCES
30. Khokhlov, A. R., and Semenov, A. N., J. Stat. Phys. 38, 161 (1985).
31. Van der Schoot, P. P. A. M., and Odijk, T. J., Chem. Phys. 97, 515
1. Russel, W. B., J. Fluid Mech. 85, 209 (1978). (1992).
2. van der Werff, J. C., and de Kruif, C. G., J. Rheol. 33, 3, 421 (1989). 32. De Gans, B. J., and Verduin, H., unpublished results.
3. Woutersen, A. T. J. M., and de Kruif, C. G., J. Chem. Phys. 94, 5793 33. Milliken, W. J., Gottlieb, M., Graham, A. L., Mondy, L. A., and Powell,
(1991). R. L., J. Fluid Mech. 202, 217 (1989).
34. Pusey, P. N., Fijnaut, H., and Vrij, A., J. Chem. Phys. 77, 4270 (1982). 4. Batchelor, G. K., and Green, J. T., J. Fluid Mech. 56, 401 (1972).
AID JCIS 4137 / 6g0f$$$325 05-19-96 20:36:49 coida AP: Colloid

You might also like