Langmuir 2005, 21, 5761-5769 5761

Aggregation and Charging of Colloidal Silica Particles:

Effect of Particle Size
Motoyoshi Kobayashi,† Frédéric Juillerat,‡ Paolo Galletto,† Paul Bowen,‡ and
Michal Borkovec*,†
Laboratory of Colloid and Surface Chemistry, Department of Inorganic, Analytical and
Applied Chemistry, University of Geneva, Sciences II, Quai Ernest Ansermet 30,
CH-1211 Geneva 4, Switzerland, and Laboratory of Powder Technology, Institute of Materials,
Ecole Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland
Received December 21, 2004. In Final Form: April 25, 2005

We studied systematically aqueous suspensions of amorphous well-characterized silica particles by

potentiometric titration, electrophoretic mobility, and time-resolved light scattering. Their charging behavior
and aggregation rate constants were measured as a function of pH and ionic strength in KCl electrolytes
for three types of particles of approximately 30, 50, and 80 nm in diameter. The charging behavior was
consistent with the basic Stern model; the silica particles carry a negative charge, and its magnitude
gradually increases with increasing pH and ionic strength. On the other hand, their early-stage aggregation
(or coagulation) behavior is complex. The aggregation of the largest particles shows features resembling
predictions of the Derjaguin, Landau, Verwey, and Overbeek (DLVO) theory. On one hand, the rate constant
decreases sharply with increasing pH at low ionic strengths and attains fast aggregation conditions at high
ionic strengths. On the other hand, we observe a characteristic slowing down of the aggregation at low
pH and high ionic strengths. This feature becomes very pronounced for the medium and the small particles,
leading to a complete stabilization at low pH for the latter. Stabilization is also observed at higher pH for
the medium and the small particles. From these aggregation measurements we infer the existence of an
additional repulsive force. Its origin is tentatively explained by postulating hairy layers of consisting of
poly(silicilic acid) chains on the particle surface.

1. Introduction and the mutual diffusion coefficient enter.7,9-12 The DLVO

Colloidal silica particles are widely used in many
industrial applications, for example, as polishing slurries,
catalysts, composite coatings, and adsorbents.1,2 Silicate
particles (i.e., quartz, feldspars, clay minerals) further
play an important role in determining the fate of nutrients
and contaminants in subsurface environments, be it as
adsorbents or as mobile carriers.3,4 In all these processes,
colloidal stability of silica particles is of major importance,
and for this reason it is essential to properly understand
the aggregation properties of these systems in aqueous
suspensions.
The famous theory of Derjaguin, Landau, Verwey, and
Overbeek (DLVO) represents the classical framework to
discuss the stability of colloidal suspensions.5-9 The theory
treats the particle interactions as a superposition of
repulsive double-layer overlap forces and an attractive
dispersion (van der Waals) force. The aggregation rate
constant is in turn evaluated from the steady-state solution
of the diffusion equation, where the interaction potential
†
University of Geneva.
‡ Ecole Polytechnique Fédérale de Lausanne.
* Corresponding author. E-mail: Michal.Borkovec@unige.ch.
(1) Iler, R. K. The Chemistry of Silica; John Wiley: New York, 1979.
(2) Castelvetro, V.; De Vita, C. Adv. Colloid Interface Sci. 2004, 108-
109, 167.
(3) Ryan, J. N.; Elimelech, M. Colloids Surf. A 1996, 107, 1.
(4) Kretzschmar, R.; Borkovec, M.; Grolimund, D.; Elimelech, M.
Adv. Agronomy 1999, 66, 121.
(5) Derjaguin, B. W.; Landau, L. Acta Physicochim. URRS 1941, 14,
633.
(6) Verwey, E. J.; Overbeek, J. T. G. Theory of the Stability of
Lyophobic Colloids; Elsevier: Amsterdam, 1948.
(7) Russel, W. B.; Saville, D. A.; Schowalter, W. R. Colloidal
Dispersions; Cambridge University Press: Cambridge, 1989.
(8) Israelachvili, J. N. Intermolecular and Surface Forces; Academic
Press: London, 1991.
(9) Hiemenz, P. C.; Rajagopalan, R. Principles of Colloid and Surface
Chemistry; Marcel Dekker: New York, 1997.
theory predicts a well-defined critical coagulation con-
centration (CCC), which separates the regimes of slow
and fast aggregation. Furthermore, the theory suggests
an increase of the suspension stability with increasing
surface charge and with decreasing salt levelstrends
which are observed in most suspensions of charged
particles. In the fast regime, the aggregation rate constants
are predicted reasonably well, typically within a factor of
2 or better.12-17 As shown recently, the theory can also
quantify the aggregation rate constants in the slow regime
for some systems, for example, carboxylated latex,18,19
weakly charged sulfate latex,20 and negatively charged
hematite.19,21 In the slow regime, however, the validity of
the DLVO theory is restricted to systems, where the
potential barrier is located at distances exceeding a few
nanometers. This condition limits the validity of DLVO
theory to weakly charged systems, which have the CCC
at low salt levels. For highly charged systems, the CCC
is located at higher salt concentrations, and the barriers
(10) von Smoluchowski, M. Z. Phys. Chem. (Munich) 1917, 92, 129.
(11) Fuchs, N. Z. Z. Phys. 1934, 89, 736.
(12) Adachi, Y. Adv. Colloid Interface Sci. 1995, 56, 1.
(13) Lips, A.; Willis, E. J. Chem. Soc., Faraday Trans. 1 1973, 69, 1226.
(14) Van Zanten, J. H.; Elimelech, M. J. Colloid Interface Sci. 1992,
154, 1.
(15) Higashitani, K.; Kondo, M.; Hatade, S. J. Colloid Interface Sci.
1991, 142, 204.
(16) Holthoff, H.; Egelhaaf, S. U.; Borkovec, M.; Schurtenberger, P.;
Sticher, H. Langmuir 1996, 12, 5541.
(17) Holthoff, H.; Schmitt, A.; Fernandez-Barbero, A.; Borkovec, M.;
CabrerIzo-Vilchez, M. A.; Schurtenberger, P.; Hidalgo-Alvarez, R. J.
Colloid Interface Sci. 1997, 192, 463.
(18) Behrens, S. H.; Christl, D. I.; Emmerzael, R.; Schurtenberger,
P.; Borkovec, M. Langmuir 2000, 16, 2566.
(19) Behrens, S. H.; Borkovec, M.; Schurtenberger, P. Langmuir 1998,
14, 1951.
(20) Behrens, S. H.; Semmler, M.; Borkovec, M. Prog. Colloid Polym.
Sci. 1998, 110, 66.
(21) Schudel, M.; Behrens, S. H.; Holthoff, H.; Kretzschmar, R.;
Borkovec, M. J. Colloid Interface Sci. 1997, 196, 241.

10.1021/la046829z CCC: $30.25 © 2005 American Chemical Society

Published on Web 05/24/2005
5762 Langmuir, Vol. 21, No. 13, 2005
move to sub-nanometer distances. In this case, one
observed deviations from DLVO theory, which are prob-
ably related to surface roughness and/or discreteness of
charge.18,21-25
The stability of silica particles has been repeatedly
reported to disagree with DLVO theory.1,15,26-31 Silica
particles are claimed to be stable at low pH, where silica
has basically no charge.26,28,32-35 Under such conditions,
the DLVO theory incorrectly predicts that silica particles
should be unstable. Other silica systems show a minimum
in the CCC as a function of pH.26,28,30 Again, this minimum
cannot be reconciled with DLVO theory, which predicts
a continuous increase of the CCC due to the increase of
the magnitude of surface charge with increasing pH. Such
anomalies have been mainly reported for nanosized silica
particles,1,15,26-30 while sub-micrometer silica particles
seem to behave DLVO-like.15,24,29,36-38
We find many of these experimental findings difficult
to judge. In many studies, the suspension stability was
assessed only in a qualitative fashion by visual observa-
tion, sedimentation tests, or turbidity measurements.26,28-30
While such assays are easy to perform, the results are
difficult to compare with the DLVO theory quantitatively.
Stability studies were sometimes not carried out on well-
characterized samples, and the particle size, pH, and ionic
strength were not always varied systematically.15,24,36-38
For these reasons, we have systematically studied well-
characterized spherical silica particles by means of
potentiometric titration, electrophoretic mobility, and light
scattering techniques. We present results concerning the
particle charge and aggregation rate constants as a
function of ionic strength and pH for particles of different
particle size. We find that submicron-sized silica particles
aggregate in a fashion, which is similar to predictions of
the DLVO theory, while for nanometer-sized particles
substantial deviations are observed.
2. Experimental Section
Materials. This study used three different batches of com-
mercial colloidal Klebosol silica particles supplied by the
manufacturer (Clariant), namely large (150H50), medium (1508-
35), and small (20H12) particles. The large particles were washed
three times with 0.1 M HCl and rinsed with deionized water by
repeated centrifugation and decantation until the electrical
conductivity of supernatant dropped below 2 µS/cm.39 Deionized
water was prepared with the Milli-Q A10 UV/UF (Millipore)
system and used throughout. The small and medium particles
were suspended in 0.1 M HCl solution, dialyzed against three
times exchanged 1 mM HCl solution, and finally extensively
dialyzed against pure water.
(22) Cooper, W. D. Kolloid Z. Z. Polym. 1972, 250, 38.
(23) Prieve, D. C.; Lin, M. M. J. J. Colloid Interface Sci. 1982, 86, 17.
(24) Kihira, H.; Ryde, N.; Matijevic, E. J. Chem. Soc., Faraday Trans.
1992, 88, 2379.
(25) Hiemstra, T.; van Riemsdijk, W. H. Langmuir 1999, 15, 8045.
(26) Allen, L. H.; Matijevic, E. J. Colloid Interface Sci. 1969, 31, 287.
(27) Allen, L. H.; Matijevic, E. J. Colloid Interface Sci. 1970, 33, 420.
(28) Depasse, J.; Watillon, A. J. Colloid Interface Sci. 1970, 33, 430.
(29) Harding, R. D. J. Colloid Interface Sci. 1971, 35, 172.
(30) Depasse, J. J. Colloid Interface Sci. 1999, 220, 174.
(31) Milonjic, S. K. Colloids Surf. 1992, 63, 113.
(32) Bolt, G. H. J. Phys. Chem. 1957, 61, 1168.
(33) Yates, D. E.; Healy, T. W. J. Colloid Interface Sci. 1976, 55, 9.
(34) Hiemstra, T.; van Riemsdijk, W. H.; Bolt, G. H. J. Colloid Interface
Sci. 1989, 133, 91.
(35) Borkovec, M.; Jonsson, B.; Koper, G. J. M. Colloids Surf. Sci.
2001, 16, 99.
(36) Chang, S. Y.; Ring, T. A.; Trujillo, E. M. Colloid Polym. Sci.
1991, 269, 843.
(37) Killmann, E.; Adolph, H. Colloid Polym. Sci. 1995, 273, 1071.
(38) Barany, S.; Cohen Stuart, M. A.; Fleer, G. J. Colloids Surf. A
1996, 106, 213.
(39) de Keizer, A.; van der Ent, E. M.; Koopal, L. K. Colloids Surf.
A 1998, 142, 303.
Kobayashi et al.
Particle Size Measurements. Particle morphology and the
size distribution were assessed by electron microscopy. For
transmission electron microscopy (TEM) normal-resolution and
high-resolution microscopes (Philips CM20 and CM430) were
used. Scanning electron microscopy (SEM) was carried out with
a high-resolution instrument (Philips XL-30SFEG). A few
microliters of dilute particle suspensions at particle concentra-
tions of around 0.2-0.4 g/L was allowed to air-dry slowly on
carbon-coated copper grids for TEM and on glass slides for SEM.
The samples were imaged without any metal coatings. For the
medium and large particles, the number-averaged radius 〈r〉 and
the coefficient of variance CV ) x〈r2〉/〈r〉2-1 were evaluated by
measuring 500-1000 particles on the micrographs. For the small
particles only 100 were counted due to difficulties in obtaining
sharp focus.
Particle size was further measured on two light scattering
instruments. First, a multiangle goniometer (ALV CGS-8,
Langen) with ALV-5000 correlators and a 532 nm solid-state
laser as light source was used. Second, the ZetaPALS instrument
(Brookhaven) equipped with a BI-9000AT digital correlator and
a 661 nm solid-state laser was utilized. Dynamic light scattering
(DLS) was performed in stable dilute suspensions and 0.1 mM
KCl at pH around 8 at 25 °C and a scattering angle of 90°. From
dynamic light scattering data the number-averaged radius 〈r〉
and CV were obtained with regularized least-squares deconvo-
lution (CONTIN),40 while from second-order cumulants the
hydrodynamic radius, rh, was evaluated. The radius remained
constant within 2 nm, and we could not detect any systematic
dependence on pH or the ionic strength. The latter was compared
with the TEM data using the relationship rh ) 〈r6〉/〈r5〉. The higher
moments were estimated by assuming a Schultz distribution.
Angle-dependent static light scattering was interpreted with the
Rayleigh-Gans-Debye (RGD) theory,41 from which an apparent
radius rS was extracted, which can be compared with the TEM
data by rS ) x〈r8〉/〈r6〉. Centrifugal sedimentation was per-
formed with an X-ray disk centrifuge (XDC, Brookhaven). For
such measurements, the suspensions were initially diluted in
deionized water down to a concentration of 20 g/L and dispersed
with a 150 W ultrasonic horn for 15 min. From the measured
sedimentation distribution, the particle densities and CV were
obtained on the basis of the particle radii known from TEM.
Density and Surface Area Measurements. Density mea-
surements of the particles were performed in suspensions by
standard water pycnometry and after drying by helium pyc-
nometry (Microline 380, Micromeritics). The specific surface area
of the silica particles, a, was determined with nitrogen adsorption
BET measurements9 performed with a Gemini 2375 (Micromer-
itics) on air-dried and freeze-dried samples (Alpha 1-4, Christ).
The specific surface area was compared with the TEM measure-
ments with the relation a ) (3/F)〈r2〉/〈r3〉, where F is the particle
density.
Potentiometric Acid-Base Titration. The surface charge
density of silica particles was measured with acid-base poten-
tiometric titrations. They were carried out on a home-built
computer-controlled instrument with four automatic burets
containing 0.25 M HCl, 0.25 M CO2-free KOH (Baker Dilut-It),
3.0 M KCl, and pure water. All solutions were freed of CO2 by
boiling. The potential between a glass electrode (6.0123.100,
Metrohm) and a separate Ag/AgCl reference electrode (6.0733.100,
Metrohm) was measured by a high-impedance voltmeter. The
titrations were carried out at 25 °C, and the vessel was
continuously flushed with moist CO2-free nitrogen gas. Acidified
silica suspensions with particle concentrations around 5.5, 32,
and 54 g/L for small, medium, and large particles were titrated
at constant ionic strength constant from pH 3.5 to 9 and back.
The pH range was chosen in order to minimize effects of silica
dissolution and to ensure an experimental error below 5%. After
such a run, the overall ionic strength was increased, and the
procedure was repeated. Electrodes were calibrated with blank
titrations, which also served to determine the exact base
concentration and the activity coefficients. The surface charge
was obtained from the difference between the titration curves
(40) Provencher, S. W. Comput. Phys. Commun. 1982, 27, 229.
(41) Kerker, M. The Scattering of Light and Other Electromagnetic
Radiation; Academic Press: New York, 1969.
Colloidal Silica Particles
Figure 1. Determination of aggregation rates by time-resolved
light scattering. (a) Temporal increase of the apparent DLS
hydrodynamic radius for the small silica particles in 1 M KCl
solution at different pH values. Linear extrapolations to short
times in the early stage of the aggregation are shown as solid
lines. The intercept is the hydrodynamic radius of isolated
particles of 18 nm. (b) Scatter plot of the apparent static and
dynamic aggregation rates obtained from an SSDLS experi-
ment. The data are fitted to a straight line, whose intercept
gives the absolute aggregation rate constant.
with and without the sample and normalized to the BET surface
area. Details on the technique are given elsewhere.21,34,35
Electrophoretic Mobility. A laser Doppler velocimetry in-
strument (Zeta Sizer 2000, Malvern) was used to carry out
electrophoretic mobility measurements at 25 °C as a function of
pH and at different ionic strength. Samples were prepared from
silica suspensions diluted with water and by adding appropriate
volumes of stock solutions of KCl, HCl, or KOH to adjust the
ionic strength and pH. A combination electrode (6.0234.110, Met-
rohm) was used to measure the suspension pH. Particle con-
centrations were around 0.1, 1, and 2 g/L for large, medium, and
small particles, respectively. Mobility values were found to be in-
sensitive to changes in particle concentrations within a factor of 4.
Aggregation Rate Constants. Time-resolved static and
dynamic light scattering was used to determine aggregation rates
with the ALV multiangle goniometer (see above). The hydro-
dynamic radius rh(q,t) was measured as a function of time t and
given magnitudes of the wave vector q from second-order
cumulant analysis. The DLS measurement were carried out at
90° with a time resolution between 1 and 20 s. Simultaneous
static and dynamic light scattering (SSDLS) measurements16,17
were performed for the largest particles only, and in addition to
rh(q,t), the scattering intensity I(q,t) was measured as well. The
angular resolution improved by repositioning the detectors during
the run, typically leading to an angular resolution of 8°. The
sample temperature was kept at 25 °C.
Quartz or borosilicate glass cuvettes used were cleaned by
soaking in a hot 3:1 mixture of concentrated sulfuric acid and
30% hydrogen peroxide and rinsed extensively with water. Prior
to experiments, the necessary amounts of stock solutions of KCl,
HCl, or KOH were mixed with water in the cuvette. In the case
of large silica particles, the stable silica suspension was added,
shaken with a vortex mixer, and monitored by DLS or SSDLS
light scattering during up to 60 min with a time resolution of 20
s. The particle concentration was chosen around 0.5 mg/L (1015
m-3), which leads to an increase of the hydrodynamic radius of
about 2-50% during the measurement. The initial radius did
Langmuir, Vol. 21, No. 13, 2005 5763
not show any systematic variations with the solution composition.
The suspension pH was determined with a combination micro-
electrode (6.0234.110, Metrohm) in the cuvette right after the
measurement. In the case of the small- and medium-sized
particles, the cuvette was inserted in the vat, and the suspension
was injected with a pipet, mixed, and monitored by DLS. The
particle concentrations were in the range 8-60 mg/L (7 × 1016
and 7 × 1018 m-3). In the case of fast aggregation, the smallest
time resolution chosen was 1 s, and the experiment was carried
out for a few minutes. This procedure permits us to keep the rate
increase of the hydrodynamic radius with time sufficiently small.
By extrapolating to the initial time, the radius of the particles
in a stable suspension could be recovered (see Figure 1). In this
way it was assured that the early stages of the aggregation were
monitored even for the small particles.
The absolute aggregation rate constant k was extracted from
the apparent initial DLS rate based on the RGD approximation
and the relative hydrodynamic radius of the doublet R ) 1.39
with the relation16,17
1 drh(q,t)
| ) 1+
rh(q,0) dt tf0 2rq[
sin (2rq) 1
1 - kN0
R ]( ) (1)
where r is the (number-averaged) particle radius and N0 is the
particle number concentration. This equation has been shown
to be reliable for small particles.16,17,42,43 For the largest particles,
the values of the absolute rate constants were verified with
SSDLS. This method requires besides the apparent DLS rate
also the apparent static light scattering rate as a function of the
angle. The quantities are plotted in a scatter plot and fitted to
the linear equation16,17
1 dI(q,t) 1 drh(q,t) R
I(q,0) dt
|tf0 ) |
rh(q,0) dt tf0 R - 1 (
- kN0 (2) )
This method makes no assumptions about the optical and
hydrodynamic properties of the particle dimers. Representative
data and this fit are shown in Figure 1. In the pH range 7-11
and in 1 M KCl, we have performed five independent measure-
ments with DLS and SSDLS. The average rate constant of (1.5
( 0.2) × 10 - 18 m3 s-1 obtained from DLS is in very good agreement
with the SSDLS value of (1.6 ( 0.2) × 10-18 m3 s-1. The latter
technique also gives a relative hydrodynamic ratio of R ) 1.38
( 0.03, which is close to the theoretical value stated above.
3. Modeling
Surface Charge. The classical 1 - pK basic Stern
model is used to evaluate the surface charge density as
a function of pH and ionic strength. This model has been
successfully applied to silica and other oxides previ-
ously.34,35,44 Surface charge of silica is assumed to originate
from the deprotonation of silanol groups according to the
reaction
SiOH T SiO- + H+ (3)
The site densities of both surface groups ΓSiOH and ΓSiO-
obey the mass action law
ΓSiO-aH
) K exp(βeψ0) (4)
ΓSiOH
where aH is the activity of the protons (with pH ) - log
aH), K the intrinsic equilibrium constant (with pK ) -log
K), ψ0 the surface potential, e the elementary charge, and
β the inverse thermal energy. The total number density
of these sites is given by
(42) Yu, W. L.; Borkovec, M. J. Phys. Chem. B 2002, 106, 13106.
(43) Lin, W.; Galletto, P.; Borkovec, M. Langmuir 2004, 20, 7465.
(44) Hiemstra, T.; de Wit, J. C. M.; van Riemsdijk, W. H. J. Colloid
Interface Sci. 1989, 133, 105.
5764 Langmuir, Vol. 21, No. 13, 2005 Kobayashi et al.

Γ0 ) ΓSiOH + ΓSiO- (5) assumed to be

while the surface charge density is V(h) ) VvdW(h) + Vdl(h) (13)

σ ) -eΓSiO- (6) where VvdW(h) and Vdl(h) are the interaction potentials
due to van der Waals and double-layer overlap forces. By
The surface charge density is related to the potential drop invoking the Derjaguin approximation, we can write
over the Stern layer
Ar
σ ) CS(ψ0 - ψd) (7) VvdW(h) ) - (14)
12h
where CS is the Stern layer capacitance and ψd the diffuse where A is the Hamaker constant across the liquid. The
layer potential. The charge potential relationship of the double-layer overlap potential Vdl(h) is obtained within
diffuse layer (so-called Grahame equation) the Derjaguin approximation from a numerical solution
of the Poisson-Boltzmann equation including full regula-
σ)
20κ
βe
sinh
2 ( )
βeψd
(8)
tion.48 The solution in use is equivalent to the results based
on elliptic functions.18,49
4. Results and Discussion
closes the system of equations, whereby 0 is the total
Morphology and Particle Size. The SEM and TEM
permittivity of water, and the Debye length is given by
micrographs for the three silica batches are shown in
Figures 2 and 3. The photographs illustrate the spherical

x
0
κ-1 )
2βe2NAI
where NA is Avogadro’s number and I the ionic strength.
Equation 8 is valid when rκ . 1, which is a good
approximation in most situations discussed here. Cur-
vature corrections are considered by using the corre-
sponding charge-potential relationship for a sphere
obtained from the numerical solution of the Poisson-
Boltzmann equation.
Electrophoretic Mobility. The standard electroki-
netic model due to O’Brien and White is used to evaluate
the electrophoretic mobility.45 The model takes into
account the double-layer relaxation and retardation effects
and uses the potential at the shear plane (so-called ζ
potential) as input. We calculate this potential in an
approximate fashion from the potential profile of the
diffuse layer18
shape and the narrow size distributions of the larger
(9)
particles. The average number-weighted radii 〈r〉 and the
coefficients of variation (CV) are summarized in Table 1.
The most reliable results are obtained from TEM, giving
particle radii of 10, 23, and 38 nm for the small, medium,
and large particles, respectively. The large- and medium-
sized particles are spherical and monodisperse, with a
CV in the range 0.10-0.15. The small particles are
sometimes ellipsoidal and much more polydisperse (CV
≈ 0.4). The SEM and light scattering (DLS and SLS) lead
to consistent results, indicating only marginal swelling of
the particles in water.
The particle densities are summarized in Table 2. Air
pycnometry gives a value near 2.2 g/cm3, as expected for
an amorphous precipitated silica.50 From water pycnom-
etry and XDC,51 one infers densities in the range 1.7-2.2
g/cm3, which indicate porosity or swelling of the surface
layer.
The BET specific surface areas are listed in Table 2 as

ζ)
4
βe [
arctanh exp(-κd) tanh
βeψd
4 ( )] (10)
well. They are consistently higher than the geometrical
surface areas calculated from the TEM data, irrespective
of the drying method. This difference represents an addi-
where d is the distance of the shear plane from the surface, tional indication of particle porosity or surface roughness.
and the diffuse layer potential is calculated from the 1 - Surface Charge Density. Figure 4 shows the surface
pK model discussed above. charge density as a function of pH at different ionic
Aggregation Rate Constants. The steady-state solu- strengths in KCl electrolytes measured with potentio-
tion of the diffusion equation in a force field leads to the metric titrations. The silica particles have a negative
classical expression of the aggregation rate constant7,9,18 charge, the magnitude of which increases with increasing
pH and increasing ionic strength. The solid lines are

{ }
B(h) -1 results of calculations based on the 1 - pK basic Stern
k)
8
3βη
2r ∫0∞ (2r + h)2 exp[βV(h)] dh (11) model with a total site density Γ0 ) 8 nm-2, a Stern
capacitance CS ) 2.9 F/m2, and an ionization constant pK
) 7.6. The model predicts a saturation of the surface charge
where η is the viscosity of solution, h the distance between at very high pH, but this saturation cannot be observed
particle surfaces, V(h) interaction potential, and B(h) the within the accessible pH range.
dimensionless hydrodynamic resistance function. The The parameters used give adequate fits to the experi-
latter is approximated by46,47 mental data for the data at lower ionic strengths but give
inferior fits at 1 M, probably due to neglect of specific
6(h/r)2 + 13(h/r) + 2 binding of potassium ions. The same values for the site
B(h) ) (12)
6(h/r)2 + 4(h/r) density and the Stern capacitance were proposed previ-
ously for silica, while the intrinsic constant was estimated
Within the DLVO theory, the interaction potential is to be pK ) 7.5 on the basis of the multisite surface

(45) O’Brien, R. W.; White, L. R. J. Chem. Soc., Faraday Trans. 2
1978, 74, 1607.
(46) Honig, E. P.; Roeberse, G. J.; Wiersema, P. H. J. Colloid Interface
Sci. 1971, 36, 97.
(47) Elimelech, M.; Gregory, J.; Jia, X.; Williams, R. Particle
Deposition and Aggregation; Butterworth: Oxford, 1995.
(48) Pericet-Camara, R.; Papastavrou, G.; Behrens, S. H.; Borkovec,
M. J. Phys. Chem. B 2004, 108, 19467.
(49) Behrens, S. H.; Borkovec, M. J. Phys. Chem. B 1999, 103, 2918.
(50) Jelinek, L.; Kovats, E. S. Langmuir 1994, 10, 4225.
(51) Staiger, M.; Bowen, P.; Ketterer, J.; Bohonek, J. J. Dispersion
Sci. Technol. 2002, 23, 619.
Colloidal Silica Particles
Figure 2. Scanning electron micrographs (SEM) of the silica
particles: (a) large (bar 200 nm), (b) medium (bar 200 nm), and
(c) small (bar 100 nm) particles. The results of their particle
size analysis are given in Table 1.
complexation (MUSIC) model.34,35,44 With the latter value,
a slightly inferior fit of the experimental data is obtained.
Electrophoretic Mobility. Figure 5 shows the ex-
perimental results for the electrophoretic mobilities as a
function of pH at different ionic strengths. In line with
their negative charge, the particles exhibit negative
mobility, and its magnitude increases with increasing pH.
The solid lines are calculations based on the standard
electrokinetic model using the diffuse layer potential from
the 1 - pK basic Stern model as input. In the calculation,
we assume that the position of the shear plane is located
at 0.5 nm from the surface. This shift reduces the
magnitude of the mobility at higher pH at 100 mM but
does not much affect the results otherwise. The model
calculations agree reasonably well with the experimental
results, particularly when one realizes that the model
contains only one adjustable parameter. Note that al-
though the mobility approximately saturates to a common
value at pH > 9, the surface potential keeps increasing
with increasing pH. This phenomenon is related to the
existence of a mobility maximum, which is predicted by
Langmuir, Vol. 21, No. 13, 2005 5765
the standard electrokinetic model.45 Moreover, the model
correctly captures the intersections of the mobility curves
for 3 and 10 mM, occurring for the large- and medium-
sized particles. The shear plane is located twice as far
from the surface than the distance of 0.25 nm proposed
previously for carboxylated latex particles.18
Aggregation Rate Constants. Before we discuss the
experimental results, let us briefly review the predictions
of DLVO theory. With the known charging behavior, the
DLVO theory can be used to calculate the aggregation
rate constants. The only additional parameter is the
Hamaker constant of silica across water. Its value is known
fairly accurately, and we use A ) 8.3 × 10-21 J reported
earlier.8,52 Figure 6 shows the calculated aggregation rate
constants for the large particles as a function of pH for
different ionic strengths. At low pH, the rate constant is
close to diffusion-controlled value, and with increasing
pH the aggregation rate decreases sharply beyond a critical
pH, due to the buildup of charge on the silica surface. The
rate constant also increases with increasing ionic strength,
due to screening of the electrostatic interactions. Because
of this effect, the critical pH moves toward higher pH
values, until the aggregation rate constant remains
diffusion-controlled through the whole accessible pH. This
transition is predicted to happen at high salt concentra-
tions near 4 M. The predictions of DLVO theory are very
similar for the smaller particles, with the exception that
the decrease of the rate constant with pH is less pro-
nounced. Note that similar pH dependencies for the
aggregation rates were experimentally confirmed for
carboxylated latex and negatively charged hematite
particles.18,19,21
Let us now discuss the measured absolute aggregation
rate constants as a function of pH at different ionic
strengths (see Figure 7). The rate constants were obtained
with time-resolved light scattering, mostly with DLS and
verified with SSDLS for the largest particles. We are thus
confident that the rate constants correspond to the
absolute rate constants of dimer formation and that they
are measured in the early stages of aggregation even for
the smallest particles (see section 2).
Focus now on the data for the large silica particles shown
in Figure 7a. The observed aggregation rate constants
show a more complex behavior than what is predicted by
the DLVO theory (see Figure 6). Nevertheless, several
important features are reproduced, the least in a semi-
quantitative fashion. At high salt levels and high pH, the
aggregation rates are fast and close to the diffusion-
controlled value. At lower salt levels, one observes a rapid
decrease of the rate constants with increasing pH, which
is also in line with the DLVO theory.
However, there are several important differences to the
DLVO predictions. The first difference is the lower value
of the fast aggregation rate constant, the second is the
marked decrease of the rate constant at high ionic
strengths with decreasing pH, and the third reflects an
increase in the rate constant at pH > 11. Let us now discuss
these three differences. However, we do not attempt to
model these differences quantitatively.
The first difference to DLVO theory is a lower fast
aggregation rate constant around 2 × 10-18 m3/s, while
the theory predicts 8 × 10-18 m3/s. While the use of the
Derjaguin approximation in the calculation leads to a
slightly (say 10-20%) larger rate constant, this difference
of a factor of 4 is hard to reconcile. Comparable values for
the fast aggregation rate constant have been reported for
submicron-sized silica particles in the range (2-10) × 10-18
(52) Hough, D. B.; White, L. R. Adv. Colloid Interface Sci. 1980, 14, 3.
5766 Langmuir, Vol. 21, No. 13, 2005 Kobayashi et al.

Figure 3. Transmission electron micrographs (TEM) of silica particles for small (left) and large (right) magnification: (a) large,
(b) medium, and (c) small particles.

Table 1. Particle Size and Polydispersity

〈r〉 (nm)a CVb rh (nm)c rs (nm)d
silica particles TEM SEM CON TEM SEM CON XDC CUM TEM CON SLS TEM CON
large (150H50) 38 36 39 0.11 0.11 0.06 0.18 47 40 39 49 40 40
medium (1508-35) 23 24 25 0.15 0.13 0.16 0.20 28 26 28 28 26 29
small (20H12) 10 10 13 0.37 0.15 0.23 0.27 18 17 16 17 19 17
a Number-weighted particle diameter 〈r〉 determined by transmission electron microscopy (TEM), scanning electron microscopy (SEM),

and CONTIN deconvolution of dynamic light scattering data (CON). b Coefficient of variation (CV) obtained by the same methods and by
X-ray disk centrifugation (XDC). c Apparent hydrodynamic radius rh obtained by cumulant (CUM) analysis of dynamic light scattering data
and calculated from TEM and CONTIN data. d Intensity-weighed radius rs obtained from the static light scattering data and calculated
from TEM and CONTIN data. Experimental errors are (1 nm in the radii and (0.02 in the CVs.

m3/s for other silica systems.15,36-38 For other systems,
fast aggregation rate constants of about factor 2-3 below
the DLVO predictions have been noted,12-15,18,19,21 while
other studies have reported good agreement between
experimentally measured rate constants and DLVO theory
in the fast regime.7,47,53 In any case, this difference is not
enormous but points toward an additional repulsion or
drag force.
The second difference to DLVO theory is a marked
decrease of the rate constant with decreasing pH at high
ionic strengths. The theory, on the other hand, predicts
at high ionic strengths a constant rate which is determined
by diffusion control and independent of pH. Necessarily,
an additional repulsive force must cause this decrease.
This force sets in rather abruptly below pH < 6 and
remains roughly constant at lower pH.
Colloidal Silica Particles Langmuir, Vol. 21, No. 13, 2005 5767

Table 2. Particle Densities and Specific Surface Areas

particle density F (g/cm3)a specific surface area a (m2/g)b
silica particles helium water XDC air-dry freeze-dry TEM CON
large (150H50) 2.20 2.14 1.7 51 52 36 35
medium (1508-35) 2.20 2.27 1.7 77 80 57 53
small (20H12) 2.19 1.7 228 227 104 95
a Particle density measured by helium and water pycnometry (errors (0.01 g/cm3) and X-ray disk centrifugation (XDC, errors (0.1

g/cm3). b Measured specific surface area by nitrogen adsorption on air-dried and freeze-dried samples and compared with calculated values
based on TEM and CON data with a particle density of F ) 2.2 g/cm3 (see Table 1, errors are (1 m2/g).

Figure 4. Surface charge of silica particles as a function of pH

different ionic strengths adjusted with KCl. The data points
are obtained by potentiometric titrations, while the solid lines
are the results of the basic Stern model with a total site density
of 8 nm-2, a Stern capacitance of 2.9 F/m2, and an ionization
constant pK ) 7.6. (a) Large, (b) medium, and (c) small particles.
The third difference to DLVO theory reflects an increase
in the rate constant at pH > 11. Note that the observed
aggregation rate constants are in closer agreement to
values in the fast aggregation regime in other sys-
tems.18,19,21 We suspect that this acceleration is related to
silica dissolution (see below).
Turning toward the results for the medium size particles
shown in Figure 7b, we find substantial differences with
respect to the larger particles. The diffusion-controlled
plateau at high pH can be still identified, but only for high
ionic strengths. The decrease around pH 6 is much more
(53) Lichtenfeld, H.; Knapschinsky, L.; Sonntag, H.; Shilov, V.
Colloids Surf. A 1995, 313.
Figure 5. Electrophoretic mobility of silica particles as a
function of pH at different ionic strengths adjusted with KCl.
The data points are obtained by electrophoresis, and the solid
lines are the results of the basic Stern model together with the
standard electrokinetic model. The plane of shear is located at
0.5 nm from the surface. The other parameters are the same
as in Figure 4. (a) Large, (b) medium, and (c) small silica
particles.
pronounced than for the larger particles, and one cannot
identify a constant plateau at lower pH any longer. A
decrease of the aggregation rate with increasing pH is
only visible at 0.6 M. Furthermore, a minimum in the
rate constant at an ionic strength of 0.8 M appears around
pH 10. The rate constant increases at pH > 11 similar to
the larger particles.
For the small particles, as shown in Figure 7c, no
aggregation could be detected below an ionic strength of
0.5 M and for pH < 6. The latter feature is analogous to
the slowdown of the aggregation at pH < 6, as observed
5768 Langmuir, Vol. 21, No. 13, 2005
Figure 6. Theoretical predictions of DLVO theory of aggrega-
tion rate constants of large silica particles plotted as a function
of pH for different ionic strengths adjusted with KCl. The
Hamaker constant was chosen as 8.3 × 10-21 J.
Figure 7. Experimental results of the aggregation rate
constants for silica particles as a function of pH at different
ionic strength: (a) large, (b) medium, and (c) small silica
particles. The lines serve to guide the eye.
for the medium and large particles. The rate at very high
ionic strength of 2 M remains about constant for pH > 6
but is substantially slower than the diffusion-controlled
rate observed in the previous two cases. The rate constants
show a minimum near pH 10 at ionic strengths of 1.2 M
and below. This minimum is similar, but more pronounced,
to the one observed for the medium-sized particles.
Silica Dissolution. At sufficiently high pH, silica
dissolves. Dissolution rates of silica have been measured
Kobayashi et al.
Figure 8. Hydrodynamic radius measured by the dynamic
light scattering as a function of time for the medium-sized silica
particles for different pH values in a 0.1 mM KCl solution. The
lines correspond to constant dissolution rates of 10-7, 10-8, and
10-9 mol m-2 s-1.
by various authors and were found to vary about 1-2
orders of magnitude depending on the silica type (e.g.,
quartz, tridymite, cristobalite, amorphous silica).1,54,55 The
dissolution rate strongly increases with increasing pH.
At pH 12, dissolution rates in the range of 10-11-10-9 mol
m-2 s-1 have been reported.33,54-57 The dissolution rate of
amorphous silica is close to the fastest rates observed.
We have assessed silica dissolution by time-resolved
DLS performed on the medium-sized particles at different
pH in 0.1 mM KCl. The results are shown in Figure 8. At
pH < 11 we see basically no decrease of the particle size
as a function of time, while at pH 12.1 a clear decrease
is observed. The rate of decrease of the particle radius can
be expressed as
dr M
) - kdis (15)
dt F
where kdis is the dissolution rate, M ) 60.1 g/mol the
molecular mass of silica, and F ) 2.2 g/cm3. When one
assumes a constant dissolution rate, the observed decrease
agrees reasonably well with eq 15 for the dissolution rate
of kdis ) 10-9 mol m-2 s-1 at pH 12.1. However, the observed
decrease is not consistent with a constant dissolution rate,
but one observes faster dissolution initially, which then
slows down for longer times. In general, the effect of
dissolution is small, and a substantial decrease in the
particle size can be only observed at pH > 12. We thus
conclude that for most conditions the effect of silica
dissolution is small.
Hairy-Layer Hypothesis. The observed anomalies of
silica aggregation can be possibly reconciled with the
existence of a hairy (or gel) layer. Such a layer has been
proposed by various authors,58-63 and it might be pictured
as consisting of short flexible polymer-like segments of
poly(silicilic acid) anchored to the surface. The clearest
(54) Hiemstra, T.; van Riemsdijk, W. H. J. Colloid Interface Sci. 1990,
136, 132.
(55) House, W. A.; Orr, D. R. J. Chem. Soc., Faraday Trans. 1992,
88, 233.
(56) Brady, P. V.; Walther, J. V. Chem. Geol. 1990, 82, 253.
(57) van Lier, J. A.; De Bruyn, P. L.; Overbeek, J. T. G. J. Phys.
Chem. 1960, 64, 1675.
(58) Kitchener, J. A. Faraday Discuss. 1971, 59, 379.
(59) Vigil, G.; Xu, Z.; Steinberg, S.; Israelachivili, J. J. Colloid Interface
Sci. 1994, 165, 367.
(60) Israelachvili, J.; Wennerstrom, H. Nature (London) 1996, 379,
219.
(61) Tadros, T. F.; Lyklema, J. J. Electroanal. Chem. 1968, 17, 267.
(62) Frens, G.; Overbeek, J. T. G. J. Colloid Interface Sci. 1972, 38,
376.
(63) Churaev, N. V.; Sergeeva, I. P.; Sobolev, V. D.; Derjaguin, B. V.
J. Colloid Interface Sci. 1981, 84, 451.
Colloidal Silica Particles
evidence for the existence this layer is based on observa-
tions of short-range repulsive forces between silica
surfaces, suggesting a thickness below a few nanom-
eters.59,64,65 More indirect indications of the existence of
this layer are the relatively large value of the Stern layer
capacitance with respect to other metal oxides66 and the
large distance of the shear plane of 0.5 nm needed to fit
the present electrophoresis data.
The hairy-layer hypothesis explains the additional
repulsive forces, which are clearly established on the basis
of direct force measurements59,64,65 and the present stabil-
ity data. This hypothesis can be reconciled with the
observed particle size dependence, since its relative
influence decreases with increasing particle size. There-
fore, we expect smaller deviations from the DLVO behavior
for larger particles than for the smaller ones. The
acceleration of the rate around pH 12 might be related to
the dissolution of the hairy layer, similarly as the fast
initial decrease in the dissolution experiment (see Figure
8). A similar mechanism was suggested on the basis of
earlier reflectometry measurements.67 The fact that this
layer induces a stronger repulsion at lower pH is probably
related to the deprotonation of the silanol groups around
pH 7, and differences in the interaction forces could be
(64) Zhmud, B. V.; Meruk, A.; Bergstrom, L. J. Colloid Interface Sci.
1998, 207, 332.
(65) Adler, J. J.; Rabinovich, Y. I.; Mouldgil, B. M. J. Colloid Interface
Sci. 2001, 237, 249.
(66) Hiemstra, T.; van Riemsdijk, W. H. Colloids Surf. 1991, 59, 7.
(67) van Duijvenbode, R. C.; Koper, G. J. M. J. Phys. Chem. B 2001,
105, 11729.
Langmuir, Vol. 21, No. 13, 2005 5769
related to surface charge heterogeneities,22-25 hydration
forces,60 or effects of hydrogen bonding.1,15,26,27
5. Conclusions
We have presented extensive set of experimental data
on charging and aggregation rate constants as a function
of pH and ionic strength of well-characterized silica
particles of approximately 30, 50, and 80 nm in diameter.
We found the charging behavior to be consistent with the
basic Stern model, where the particles are neutral at low
pH and their surface charge decreases with increasing
pH and increasing ionic strength. Their aggregation
behavior, on the other hand, is complex. Only the
aggregation of the largest particles exhibits features
similar to predictions of the DLVO theory. The smaller
particles, on the other hand, aggregate much more slowly
at higher pH and are completely stable at low pH.
Additional non-DLVO repulsive forces must be present.
We propose that the repulsive forces originate from the
overlap of hairy layers on the silica surface.
Acknowledgment. This research was supported by
the University of Geneva, by grants from the Swiss
National Science Foundation, and by the Commission for
Technology and Innovation, Switzerland, Research Pro-
gram TOP NANO 21. M.K. is grateful for the Postdoctoral
Fellowship for Research Abroad (2003-2005) from the
Japan Society for the Promotion of Science.
LA046829Z