Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

1296 J. Phys. Chem.

B 2007, 111, 1296-1303

Long-Range Interactions between Soft Colloidal Particles in Slit-Pore Geometries

Sabine H. L. Klapp,*,†,§ D. Qu,‡ and Regine v. Klitzing‡


Stranski-Laboratorium für Physikalische und Theoretische Chemie, Sekretariat C7, Technische UniVersität
Berlin, Strasse des 17. Juni 115, D-10623 Berlin, Germany, Institut für Theoretische Physik,
Sekretariat PN 7-1, Fakultät II für Mathematik und Naturwissenschaften, Technische UniVersität Berlin,
Hardenbergstrasse 36, D-10623 Berlin, Germany, and Stranski-Laboratorium für Physikalische und
Theoretische Chemie, Sekretariat TC9, Technische UniVersität Berlin, Strasse des 17. Juni 124,
D-10623 Berlin, Germany
ReceiVed: September 13, 2006; In Final Form: December 1, 2006

Combining theoretical and experimental techniques, we investigate the structure formation of charged colloidal
suspensions of silica particles in bulk and in spatial confinement (slit-pore geometry). Our focus is to identify
characteristic length scales determining typical quantities, such as the position of the main peak of the bulk
Downloaded via NANYANG TECHNOLOGICAL UNIV on September 25, 2019 at 09:40:55 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

structure factor and the period of the oscillatory force profile in the slitpore. We obtain these quantities from
integral equations/SANS experiments (bulk) and Monte Carlo simulations/colloidal probe-AFM measurements
(confinement), in which the theoretical calculations are based on the Derjaguin-Landau-Verwey-Overbeck
(DLVO) potential. Both in bulk and in the slitpore, we find excellent qualitative and quantitative agreement
between theory and experiment as long as the ionic strength chosen in the DLVO potential is sufficiently low
(implying a relatively long-ranged interaction). In particular, the bulk properties of these systems obey the
widely accepted density scaling of ξ ∝ φ-1/3. On the other hand, systems with larger ionic strengths and,
consequently, more short-ranged interactions do not obey such power law behavior and rather resemble an
uncharged hard-sphere fluid, in which the relevant length scale is the particle diameter.

Introduction A change of the interaction parameters is reflected in the


Colloidal dispersions are omnipresent in daily life. Therefore, characteristics of the force oscillations. For instance, atomic
the understanding and control of the interaction between force microscope (AFM) measurements on a solution of silica,
colloidal particles is of great interest not only in scientific sulfonated polystyrene particles, or both entrapped between a
contexts but also for technical applications. Due to ongoing silica microsphere and a flat silica plate show that with
miniaturization, in addition to their volume properties, the increasing diameter, increasing surface potential, and increasing
structure and dynamics of colloidal dispersions in confined concentration of the particles, the force oscillation becomes more
geometries become more and more important.1 In the present pronounced (i.e., the minima become deeper and the maxima
paper, we are particularly interested in the characteristic length higher).10
scales determining structure formation of charged colloids in Theoretically, structural forces have been studied using
slit-pore geometries and in bulk. We investigate these questions various statistical-mechanical methods, such as density func-
using various theoretical and experimental tools. Our investiga- tional theory (see, e.g., refs 11 and 12) and computer simula-
tions are based on silica suspensions that present a suitable tions.13 The studies have shown that theoretical tools can
model system with tunable interactions. describe oscillatory forces in a variety of model systems, such
Confining colloidal dispersions in slit-pore geometries leads as hard spheres,12,14 polar fluids,15 liquid crystals,16 and poly-
to the so-called structural forces caused by ordering of mol- electrolytes.17
ecules, particles, or aggregates. These forces were shown The present paper deals with structure formation of charged
experimentally for small molecules such as octamethylcyclotet- particles in aqueous bulk solution and in thin liquid films. In
rasiloxane2 and water,3 which were entrapped between two mica bulk solution, the decisive quantity for structure formation is
plates in a surface force apparatus. The approach of the plates the structure peak, which is a measure for the mean particle
toward each other leads to damped oscillatory forces that are distance. We determine this quantity experimentally by small-
explained by a layering of the molecules parallel to the mica angle neutron-scattering (SANS) and theoretically by hyper-
surfaces.4 With increasing external force, the molecules are netted chain (HNC) integral equations, in which the calculations
squeezed out of the slitpore layer by layer. The oscillation period are based on the well-known Derjaguin-Landau-Verwey-
is connected to the distance of the molecular layers. Oscillatory Overbeck (DLVO) potential.18 In slit-pore geometry the
forces also occur in thin films of molten salt,5 liquid crystals,6 relevant quantity is the period of the force oscillation, which
or colloidal particles.7-9 we measure by AFM. The corresponding theoretical calculations
are carried out using Monte Carlo (MC) simulations. The period
* To whom correspondence should be addressed. E-mail: sabine.klapp@
fluids.tu-berlin.de. of the force oscillation gives a measure for the layer thickness,
† Fakultät II für Mathematik und Naturwissenschaften. that is, the particle distance just before the particles are pressed
‡ Stranski-Laboratorium für Physikalische und Theoretische Chemie,
out.
Sekretariat TC9.
§ Stranski-Laboratorium für Physikalische und Theoretische Chemie, One central question of our study is the effect of the range
Sekretariat C7. of the interaction, which we can control by the silica density
10.1021/jp065982u CCC: $37.00 © 2007 American Chemical Society
Published on Web 01/24/2007
Soft Colloidal Particles in Slit-Pore Geometries J. Phys. Chem. B, Vol. 111, No. 6, 2007 1297

and the ionic strength, on the bulk structure peak, focusing on the actual calculations, the parameter ff has been set such that
the appearance of power law behavior. Another point we wish the SS interactions are essentially neglible against the DLVO
to explore is the density dependence of the period of force repulsion (see below).
oscillations in the slit pore, which has been discussed in the To model the slit-pore confinement, we consider the silica
literature,19,20 and has generated some controversy. Connected solution to be squeezed between two plane, parallel, smooth,
to these questions is the mathematical description of the colloidal surfaces separated by a distance h along the z axis of the
interactions, which we achieve by comparing theory and coordinate system and of infinite extent in the x-y plane. In
experiment. the present work, we confine ourselves to the investigation of
uncharged surfaces only. Specifically, we employ the fluid-
Model System and Parameters wall potential21
In the present work, we adopt an effective one-component
4π σ 9 σ 9
description of the real colloidal solution. That is, we focus on
the structure and ordering of the macroions (silica particles),
ufw(z) )
45
fw[(
h/2 + z
+) (
h/2 - z )] (7)
whereas the “solvent particles” (counterions plus additional salt
ions) are treated only implicitly. For a system with low macroion From the preceding equations, it follows that the physical
density, F (and moderate macroion charges Z), the corresponding properties of the colloidal model system depend on temperature
effective interaction between two macroions with separation r (T), density (F), valency (Z), and diameter (σ) of the macroions,
can be derived on a rigorous statistical-mechanical basis,21 the ionic strength (I) of the additional salt, and the numerical
leading to the well-known DLVO potential,18 parameters ff and fw characterizing the fluid-fluid and fluid-
wall repulsion, respectively. To mimic the experimental condi-
exp(-κ(r - 2R)) tions (see Experimental Methods), the theoretical results pre-
uDLVO(r) ) (Z̃e0)2 (1)
4π0r sented in this work have been obtained with the parameters T
) 298 K and σ ) 26 nm. The valency has been set to Z ) 35,
where e0 is the elementary charge, 0 is the permeability of as estimated from the Grahame equation for the surface charge
vacuum,  is the dielectric constant of the solvent, and R is the density, σ22
particle radius. In addition, Z̃ is the effective valency which
(assuming that the colloidal concentration is low compared to
that of the added salt) is given by σ ) x80kBT sinh ( )e 0ζ
2kBT
(8)

Z
Z̃ ) (2) where the ζ potential is given by (about) -80 mV at an ionic
1 + κR
strength of (about) I ) 10-4mol/L (see Particles and Suspen-
Finally, κ is the inverse Debye screening length, defined as sions). However, since the precise value of I is not known from
the experiments, and to clarify the role of I for the system’s

x
1 K properties, we have performed calculations also at several other
κ)
0kBT
(Fc(zce0)2 + ∑ Fk(zke0)2) (3) values of the ionic strengths in the range I ) 10-1-10-5 mol/
k)1 L. From a more physical point of view, variation of I allows us
to explore the influence of the range of the repulsive interactions,
where kB is Boltzmann’s constant, T is the temperature, and Fc which depend on the (inverse) Debye length and, therefore, also
and zc are the number density and valency of the counterions, on I (see eq 4).
respectively. The remaining sum refers to the additional salt The density of macroions has been characterized through the
ions. Assuming univalent counterions (|zc| ) 1), the condition volume fraction φ ) Fσ3, and we have considered systems in
of charge neutrality between counter- and macroions requires the range φ ) 0.2-30 %. Finally, the repulsion parameters have
Fc ) |Z|F. Equation 3 can then be rewritten as been fixed such that ff/kBT ) fw/kBT ) 1.0.

x
e02 Integral Equations for the Bulk System
κ) (ZF + 2INA) (4)
0kBT In bulk experiments, the characteristic length describing the
1/ ∑K F z2
local structure in the fluid was deduced from the location of
where we have introduced the ionic strength I ) of
2 k)1 k k the first (“structural”) peak in the scattering intensity I(q) ∝
the additional salt, and NA is Avogadro’s constant. We note P(q)S(q), where P(q) and S(q) are the form factor and the
that due to the appearance of the Debye length, the DLVO structure factor, respectively, and q ) |q| is the length of the
potential depends on both the density of the macroions and on scattering vector. The relevant structural information is contained
the temperature. in S(q), which is related to the pair correlation g(r) via21
The total fluid-fluid interaction potential used in the present
calculations is given by
S(q) ) 1 + F ∫ dr exp[iq ‚ r](g(r) - 1) (9)
uff(r) ) uSS(r) + uDLVO(r) (5)
In the present work, we calculate S(q) or, equivalently, the
where the first term represents an additional soft-sphere repul- pair correlation function g(r) ≡ h(r) + 1 (with h(r) being the
sion. The latter is defined as total correlation function) by integral equations theory. This
involves simultaneous (numerical) solution of the exact Orn-
uSS(r) ) 4SS(σ/r)12 (6) stein-Zernike equation23
with the particle diameter σ ) 2R. The soft-sphere interactions
have been introduced purely for technical reasons. In fact, in h(r12) ) c(r12) + F ∫ dr3 h(r13)c(r32) (10)
1298 J. Phys. Chem. B, Vol. 111, No. 6, 2007 Klapp et al.

TABLE 1: HNC Results for the Dimensionless ensemble, P⊥ can be expressed as a statistical average involving
Compressibility, Internal Energy, (Virial) Pressure, and the instantaneous normal component of the (virial) pressure
Excess Chemical Potential of the Colloidal Solution at tensor,13 that is
Volume Fraction O ) 5.0 % (Gσ3 ) 0.095) and Various Ionic

〈∑ ∑ 〉
Strengths Ia

|
I [mol/L] κσ βχ/F βU/N βP/F βµex
〈N〉µ,T 1 N N z2ij ∂uff(r)
P ⊥ ) kB T - r)rij -
10-5 1.11 0.033 10.6 (10.6) 1.46 (1.46) 26.04 (25.16) Ah Ah rij ∂r

〈∑ 〉
i)1 j*i µ,T
10-4 1.37 0.052 5.98 (5.95) 0.93 (0.93) 15.76 (15.22)
10-3
|
2.91 0.224 0.76 (0.75) 0.25 (0.25) 2.90 (2.84) 1 ∂ufw(z)
N

a
The corresponding values for the dimensionless inverse Debye
zi z)zi (12)
Ah i)1 ∂z µ,T
lengths κσ are given in the second column. MC data are shown in
parentheses.
where 〈...〉µ,T denotes an average in the grand canonical
where c(r) is the direct correlation function, combined with an ensemble.
approximate closure relation relating the correlation functions A second quantity of interest is the local density profile. Due
to the pair potential. We choose the hypernetted chain ap- to the planar, homogeneous character of the confining walls,
proximation,21 defined as the fluid is translationally invariant in the x and y directions
such that the local density depends only on z. A statistical
g(r) ) exp[-βu(r) + h(r) - c(r)] (11) expression is then given by

where β ) 1/kBT. Using the HNC correlations to calculate 〈N(z)〉µ,T


various thermodynamic quantities,21 such as compressibility,
F(z) ) (13)
A∆z
internal energy, chemical potential, and pressure, we have found
a very good internal consistency. For example, the pressure data where N(z) is the average number of particles inside slices of
resulting from virial and free energy route agreed with each thickness ∆z ) 0.05h.
other within 0.5%, indicating high accuracy of the HNC results
under the conditions considered in the present work. An Experimental Methods
alternative choice would be the Rogers-Young (RY) closure,24 Particles and Suspensions. Silica beads (Ludox TMA-34,
which interpolates between the HNC and the Percus-Yevick deionized) were purchased from Aldrich (Taufkirchen, Ger-
approximation and has been widely used in the theory of many). To remove the salt, the samples were dialyzed against
macroion solutions (see, e.g., ref 25). Within our calculations, Milli-Q water (Millipore, Billerica, MA) for 10 days. The tubes
however, HNC and RY gave identical results. with a MWCO of 1000 were from Roth (Karlsruhe, Germany).
Moreover, the HNC thermodynamical data are in very good The particle size was determined by scanning electron
agreement with those from (canonical) MC simulations (see microscopy (SEM S-4000, Hitachi, Tokyo, Japan) at a primary
Monte Carlo Simulations). Some exemplary data are given in electron energy of 20 keV and atomic force microscopy with
Table 1, indicating that an increase in the ionic strength, I (i.e., tapping mode with a multimode nanoscope (Digital Instruments,
a decrease of the range of the repulsive interactions), at fixed Santa Barbara, CA). Both methods gave a particle diameter of
volume fraction yields an increase in the (dimensionless) ∼26 nm. The ζ potential was determined by electrokinetic
compressibility. At the same time, the total internal energy, the measurements (Zeta Sizer, Malven). At an ionic strength of
pressure, and the (excess) chemical potential decrease, reflecting ∼10-4 mol/L, the value of the potential is about -80 mV.
the reduction of overall repulsion. For most of the experiments, Milli-Q water was used as
solvent. Only for small-angle, neutron-scattering experiments
Monte Carlo Simulations
was D2O (99.9% D, euriso-top, Saarbrücken) the solvent. The
To investigate the spatially confined silica solutions, we weight percentage and the density of the solutions were deter-
performed MC simulations in the grand canonical (GC) mined by weighing the sample before and after drying (24 h at
ensemble, that is, at fixed temperature, T; fixed volume, V ) 400°). The density of the silica particles is 2 g/cm3 (determined
Ah of the simulation box (with A being the box area parallel to by thermogravimetric measurements), corresponding to a con-
the surfaces); and fixed chemical potential, µ. The GCMC version factor of 0.5 from weight percentage to volume
method assumes that the confined fluid is in contact with a bulk percentage.
reservoir at the same chemical potential,26 whereas the particle SANS Measurements. The neutron experiments were carried
number N (in the pore) can fluctuate. This situation corresponds out at the SANS-I-Spektrometers at the Paul-Scherrer Institut
precisely to the situation under which the present experiments (Switzerland). To get enough contrast, D2O was used as the
are performed. solvent for the Ludox particles.
To identify the chemical potentials of interest, we have first Colloidal-Probe AFM. The colloidal probe (CP) technique
conducted MC simulations of the bulk systems in the canonical was first developed by Ducker et al.27 In our experiments, a
(NVT) ensemble, from which the chemical potential has been silica particle is glued with epoxy to a tipless cantilever
extracted using Widom’s test particle method.26 The resulting (Ultrasharp Contact Silicon Cantilevers, CSC12) produced by
values have then been used to perform GCMC simulations at µMasch. The silica particles are produced by Bangs Laborato-
various surface separations h in the range 1.6σ e h e 4.5σ ries, Inc., and all have a radius, R, of ∼3.35 µm. The tip is
(≈ 40-120 nm). In all of these calculations, the inverse Debye cleaned with plasma cleaning for 10 min right before each
length has been fixed at the value corresponding to the bulk measurement cycle to remove all the organic components on
system. its surface. The substrate used is a silicon wafer with a native
For reasons explained in the Colloidal Films section, a central SiO2 top layer, cleaned using the RCA method,28 and stored in
quantity in the present study is the normal pressure, P⊥, exerted Millipore water before usage. Just before each experiment, the
by the fluid on the two confining surfaces. In the grand canonical substrate is taken out of the water and dried in a nitrogen stream.
Soft Colloidal Particles in Slit-Pore Geometries J. Phys. Chem. B, Vol. 111, No. 6, 2007 1299

Figure 2. Structure factor at φ ) 5.0% and various ionic strengths, I.


Figure 1. Structure factor at ionic strength I ) 10-5 mol/L and various
volume fractions, φ. behavior, on one hand, to the increasing importance of packing
effects when the system becomes denser and denser (an effect
Then a drop of the Ludox solution is put onto the substrate, that would also be present without any charges). On the other
and the probing head is immersed in the solution. Force-vs- hand, an increase in the silica volume fraction also implies an
distance F(h) curves were measured with a commercial atomic increase in charges in the system (in other words, an increase
force microscope MFP (molecular force probe) produced by of κ; see eq 4), which induces stronger repulsion (on the average)
Asylum Research, Inc. and distributed by Atomic Force and a more pronounced local ordering of the particles.
(Mannheim, Germany). Because both the silica particles in the We now consider the high-density limit of the structure factor
solutions and the interfaces are negatively charged, no adsorption in Figure 1. According to the so-called Hansen-Verlet crite-
occurs on the surfaces. For each solution, altogether 10-20 rion,29 a system freezes at S(qmax) ≈ 2.85. The present
force-distance curves were measured at different lateral posi- calculations (at I ) 10-5 mol/L) indicate that this is the case at
tions on the same substrate as well as on different substrates to φfreezing ≈ 44 %. Interestingly, this value is essentially indepen-
ensure reproducibility and to get good statistics. Oscillatory force dent of the ionic strength of the additional salt. This seems to
curves occur, and the period can be determined by the distance be in contrast to experimental observations30 on the solidification
between two adjacent minima. The final result of the force of charged latex spheres in which a strong dependence of the
period is the average period of 10-20 curves, and the error ionic strength has been observed. We note, however, that the
bars are from the standard deviation. total charge (and size) of the latex spheres was much higher
than that of the silica spheres considered in the present work,
Results and Discussion making the Coulomb interactions (of the latex spheres) more
dominant.30 Therefore, a direct comparison of the systems is
Bulk Structure. To understand the local structure in the bulk quite difficult. Within the present experiments, the validity of
silica solution, we have performed several HNC calculations the theoretical predictions could not be checked, since the silica
of the structure factor, S(q), as well as SANS measurements of particles could not be stabilized in this high concentration
the scattering intensity, I(q). These two quantities differ by the regime. Nevertheless, at the relatively low concentrations
form factor, P(q), such that their full q-dependence is not directly considered, we did not find any evidence for crystallization.
comparable. However, since P(q) is almost constant in the q So far, results were presented for a very low ionic strength
range of interest, we can safely assume that I(q) and S(q) exhibit (I ) 10-5 mol/L) of the additional salt. In the following, we
their main (“structural”) peak at the same wavenumber, qmax. consider in more detail screening effects on S(q) induced by
The corresponding length, ξ ≡ 2π/qmax, roughly coincides with variation of the ionic strength. Indeed, the ionic strength does
the position of the main peak of the pair correlation, g(r), of have a strong influence on S(q) at small volume fractions. This
the silica particles as a consequence of the fact that S(q) and is demonstrated in Figure 2 where we have plotted S(q) for
g(r) are directly related by a Fourier transform (see eq 9).21 various values of I at an exemplary value of φ typical for the
Due to this close connection, the length ξ ) 2π/qmax, indeed, dilute regime. Starting from low salt concentrations (I ) 10-5
characterizes the local structure of the system and can be used mol/L), it is seen that an increase in I has several effects. First,
as a measure of the average spacing between two silica particles. the height of the main peak decreases, reflecting that the local
A further interesting feature of the structure factor is the height structure around a given particles becomes less pronounced.
of the main peak, S(qmax), characterizing the degree of the local Second, the magnitude of S(q) in the limit q f 0, that is, the
(shell) structure around a given particle, that is, roughly compressibility, increases, indicating a weakening of the
speaking, the number of oscillations in g(r). Finally, the long repulsive interactions between the silica particles. This is
wavelength limit S(q f 0) is proportional to the system’s consistent with the thermodynamic state data presented in Table
compressibility.21 1, reflecting that, apart from the compressibility, pressure and
Theoretical results for S(q) at ionic strength I ) 10-5 mol/L average internal energy also decrease with increasing ionic
and several volume fractions are plotted in Figure 1. At the strength. A further conclusion from the theoretical data in Figure
small values of φ characterizing the experimental range explored 2 is that an increase in I shifts the position of qmax toward larger
in the present work, the structure factor has only one (main) values, that is, smaller lengths, ξ, in real space. Compared to
peak, whose position shifts to higher q values with increasing the effect of particle concentration, however, the influence of
φ. This is expected from the preceding remarks and reflects the ionic strength on qmax is much lower. A change of ∼1 order
that the average spacing between two silica particles becomes of magnitude in φ yields a shift in qmax of 0.1 nm-1, whereas
smaller. Upon further increase in φ, one also observes the even a change of 4 orders in I is not sufficient to get the same
development of additional oscillations in S(q) and a strong change in qmax. This weak sensitivity of the structure peak with
increase in the height of the main peak. We attribute this respect to the ionic strength has also been observed in scattering
1300 J. Phys. Chem. B, Vol. 111, No. 6, 2007 Klapp et al.

geometric arguments33 for isotropic fluids in three spatial


dimensions, one would expect that ξ scales with the packing
fraction as ξ ∝ φ-1/3. As seen from the double-logarithmic
representation in Figure 3b, power law behavior of the form

ξ ) aφ-b (14)
is, indeed, shown by the data for I ) 10-5 mol/L and for the
SANS data, whereas the data at I ) 10-4 mol/L are less
conclusive. Using the above power law as a fit formula, one
obtains for the theoretical curves (including that for I ) 10-4)
the values a ≈ 80, b ≈ 0.30 (I ) 10-5) and a ≈ 70, b ≈ 0.21
(I ) 10-4), respectively. The corresponding SANS parameters
lie between these values, that is, aSANS ≈ 80, bSANS ≈ 0.25.
Strictly speaking, the geometric scaling (b ) 1/3) should be
restricted to weakly interacting (ideal gaslike) systems, such as
the ones considered in the above-mentioned SANS experiments.
Interestingly, however, the present theoretical data at I ) 10-5
mol/L can be fitted to the above power law up to volume
fractions φ ≈ 20 % (see Figure 3b). Only at even higher
concentrations do the functions ξ(φ) approach a limiting value
given roughly by the particle diameter, σ ≈ 26 nm.
Turning now to the theoretical model systems with high salt
concentrations, one sees from Figure 3a that the density
dependence of ξ(φ) is much less pronounced. Indeed, a fit to
eq 14 turned out to be impossible for these systems; instead, ξ
Figure 3. (a) Characteristic length as function of the volume fraction
for various ionic strengths, I. Included are the experimental (SANS) is close to the particle diameter for all volume fractions
data. (b) Double-logarithmic representation of the data corresponding considered in Figure 3a. This is reminiscent of pure hard-sphere
to low I, together with the SANS data. (HS) systems, where ξ (as obtained from the Verlet-Weiss fit34
of the exact hard-sphere structure factor) nearly coincides with
spectra of polyelectrolyte solutions.31 In all these systems, the the particle diameter throughout the fluid density range.
addition of salt has two effects. One is the screening, that is, It seems worth noting that in a very recent study of silica
the increase in κ (see eq 4), which leads to a stronger damping solutions, Waltz and co-workers19 presented and analyzed
of the oscillations. The other effect is ion condensation at the another measure (different from ξ) for a mean particle distance
macroion, which reduces the effective charge (see eq 2). that is not based on the structure factor, but on first-order
Having seen the effect of the additional salt, it is worthwhile perturbation theory. According to this measure, even HS systems
to briefly come back to the density dependence of S(q) shown obey power law behavior up to high concentrations, indicating
in Figure 1. Due to charge neutrality, an increase in macroion that the precise definition of the mean distance is crucial for its
concentration implies a simultaneous increase in the counterion density dependence.
concentration and, consequently, in the screening effect (that Colloidal Films. Experimentally, the primary quantity from
is, an increase in κ, see eq 4). Therefore it is interesting to ask which we extract relevant length scales in the confined silica
for the role of the counterion screening for the structure factor. solutions is the force-distance profile, that is, the force, F,
For the studied system, a particle concentration of 0.5 vol % between a solid surface and that of a spherical nanoparticle
corresponds to 10-6 mol/L and 30 vol %, to 6 × 10-5 mol/L. (radius R) as function of the thickness, h, of the liquid film
Under the assumption that each particle carries 35 charges, the confined between them. When the diameter of the nanoparticle
contribution by the counterions of the particles to ionic strength is sufficiently large as compared to the distance between the
increases from 3.5 × 10-5 mol/L to 2.1 × 10-3 mol/L in the interfaces, which is the case in the present experiments (see
considered particle concentration regime. Hence, the screening the Colloidal-Probe AFM section), the two surfaces can be
length would be dominated by the counterions, and its value assumed to be locally planar (Derjaguin approximation22,35). As
varies between 53 nm (for 0.5 vol %) and 6.9 nm (for 30 vol a consequence, the normalized force, F/R, becomes a function
%). But obviously, the screening effect of the counterions is of the film thickness alone, that is, F/R ) F(h)/R. In typical
much weaker than the influence of the increase in charges that colloidal suspensions, this function displays oscillatory character,
leads to a pronounced structure peak in Figure 1 for high particle and the period of these oscillations, dslit, characterizes the
concentrations. An analog experimental result was found for structure formation in the film.
other charged macroions, such as polyelectrolytes, in which an Within the GCMC simulation technique, we do not have
increase in concentration or degree of charge leads always to a direct access to the force profile, but rather, to a closely related
higher amplitude of the structure peak.32 quantity, the pressure, P⊥, exerted by the fluid onto two plane-
The theoretical results for the length ξ ) 2π/qmax as function parallel confining surfaces separated by a distance h (see eq
of both the volume fraction and the ionic strength are sum- 12). Connection between P⊥ and F/R involves again the
marized in the two parts of Figure 3, where we have included Derjaguin approximation,22,35 which allows one to derive the
the experimental (SANS) data. From Figure 3a, one clearly relation16,12
observes that the two theoretical curves obtained at low salt

concentrations (I ) 10-5-10-4mol/L) are closest to the F(h)
experimental data, both in magnitude of the characteristic length,
2πR
) ∫ dh′ (P⊥(h′) - Pbulk) (15)
ξ, and in its dependence on the volume fraction, φ. From simple h
Soft Colloidal Particles in Slit-Pore Geometries J. Phys. Chem. B, Vol. 111, No. 6, 2007 1301

TABLE 2: Bulk Volume Fractions and Corresponding


Dimensionless Chemical Potentials at Two Values of the
Ionic Strength
φbulk [%]
5 8 10 15 20 30
βµ (I ) 10-4) 12.0 17.1 22.7 34.9
βµ (I ) 10-5) 15.2 17.0 18.7 22.0 26.8 40.1

where Pbulk is the pressure of a bulk fluid characterized by the


same temperature and chemical potential. From eq 15, it follows
that

(2π)-1
dh R ( )
d F(h)
) -(P⊥(h) - Pbulk) ≡ -f(h) (16)

reflecting that the normalized normal pressure, f(h), is essentially


the deriVatiVe of the force profile. As a consequence, the
maxima and minima of the force profile correspond to zeros in
the function f(h), and the period of oscillations in the force
profile, dslit, can be estimated numerically as the distance
between two next-nearest zeros of the normalized normal
pressure. We prefer this method to estimate dslit over a numerical
integration of f(h), since we are primarily interested in the period
of force oscillations (rather than in other characteristics of these
functions). We also note that the functions f(h) obtained for the Figure 4. (a) Simulation results for the normalized normal pressure
as a function of film thickness at ionic strength I ) 10-4 mol/L and
present systems could not be easily fitted to the standard shapes various (bulk) volume fractions, φbulk. (b) Corresponding density profiles
known for simple (e.g., Lennard-Jones-like) fluids.16 In the at film thickness h ) 2.8σ (73 nm).
following, we focus on results obtained at ionic strengths I )
10-5 mol/L and I ) 10-4 mol/L, since the corresponding model
systems seem to be closest to the real colloidal suspension (see
Figure 3). The corresponding chemical potentials used in the
GCMC simulations are given in Table 2.
Numerical results for the function f(h) are shown in Figure
4a, where we focus on the influence of the volume fraction φbulk
of the bulk fluid with which the confined fluid is in thermo-
dynamic equilibrium (I ) 10-4 mol/L). For all values of φbulk
shown in Figure 4a, the functions f(h) have damped oscillatory
character similar to what is known for simple fluids16 (for
volume fractions φbulk < 10%, the oscillations vanish at this
particular ionic strength). Furthermore, a comparison of the data
indicates two main effects of an increase in φbulk on the function Figure 5. Force curve of the silica particle suspension confined
f(h): First, the amplitude of the oscillations increases, reflecting between two silica surfaces as measured by colloidal-probe AFM. The
that the stratification of the film (i.e., the formation of fluid concentration of the respective volume phase is given in volume
percentage.
layers parallel to the surfaces) becomes more and more
pronounced. This can also be seen more directly from Figure of the force oscillation decreases from 50 nm (4.1%) to 40 nm
4b, where we have plotted the local volume fraction φ(z) ) (12.3%). At much higher silica concentrations (30 vol %), the
(π/6)σ3F(z) (see eq 13) of silica particles at a fixed thickness oscillation period is on the order of the particle diameter36 or
and various values of φbulk. Clearly, the density profiles become the effective diameter (geometric diameter plus twice the Debye
sharper when the corresponding bulk volume fraction increases. length).20 This could be a hint for a starting crystallization at a
Turning back to Figure 4a, the second feature displayed by lower particle concentration than predicted for the volume phase
the pressure profiles is that an increase in φbulk yields a decrease (see discussion in the Bulk Structure section).
in the period of the corresponding force profile, dslit. A similar Beyond the influence of (bulk) volume fraction, it is also
observation has been made in an earlier MC study by Jönsson interesting to consider the role of the ionic strength (of added
et al.17 salt), I, on the oscillatory force and pressure profiles. To explore
Furthermore, both features, the increase in the amplitude and this point from the theoretical side, we have performed
the decrease of the period of the oscillations of the normal additional GCMC simulations at I ) 10-5 mol/L. A comparison
pressure, resemble the behavior observed in experimental of two of the resulting functions, f(h), with the corresponding
measurements of the force-distance curves of confined silica ones at the larger ionic strength I ) 10-4 mol/L is shown in
solutions. This may be verified from Figure 5, where we have Figure 6a. The data indicate that addition of salt somewhat
plotted various results from CP-AFM measurements at various reduces the amplitude of the oscillations, which is consistent
volume fractions. While at φ ) 1.25%, an oscillation can be with the softening of the corresponding density profiles plotted
hardly detected, and the oscillatory forces are well-pronounced in Figure 6b. On the other hand, the resulting period of the force
at a volume fraction of 1 order of magnitude higher. The oscillations, dslit, is more or less unaffected by a variation of I,
comparison between the volume fractions of 4.1% and 12.3% at least at this particular volume fraction. Indeed, at lower
shows that in addition to an increase in amplitude, the period volume fractions characterizing the experimental systems (φ
1302 J. Phys. Chem. B, Vol. 111, No. 6, 2007 Klapp et al.

such as the position of the main peak of the structure factor


(volume) and the period of the force oscillations (slitpore), we
find excellent agreement between the experimental results from
SANS (volume) and CP-AFM measurements (slit-pore), on
one hand, and the theoretical results from HNC integral
equations (bulk) and MC simulations, on the other hand. From
a theoretical point of view, the present results indicate that the
DLVO potential used to describe the silica interaction yields a
very good description of the systems’s properties not only under
bulk conditions (which was known before; see, e.g., refs 25,
37, 38) but also in the presence of spatial confinement (which
is not at all obvious38). On the other hand, from an experimental
point of view, the present agreement shows the importance of
theoretical calculations to determine the ingredients of the
interaction potential (such as the ionic strength) of the real
system.
Concerning bulk properties, a particularly interesting finding
is that the commonly accepted power law ξ ∝ φ-1/3 for the
average particle distance ξ ) 2π/qmax in three spatial dimensions
(and isotropic particle distributions) describes the present
theoretical (HNC) results only for low ionic strength, that is,
for relatively long-ranged potentials characterized by small
values of κ. These particular ionic strengths coincide with those
estimated for the real system, where ξ, according to SANS
Figure 6. (a) Normalized normal pressure as a function of film measurements, does indeed respect the above power law.
thickness at two volume fractions and two typical values of I. (b) Increasing the ionic strength in the model system yields a
Density profiles at film thickness h ) 2.8σ (73 nm). decrease in the exponent of φ, and at high ionic strengths, ξ is
close to the particle diameter throughout the fluid density range,
similar to what is predicted for a (completely screened) hard-
sphere fluid (κ f ∞). A second effect of an increasing ionic
strength is the reduction of oscillations (and peak heights) of
the structure factor due to screening effects and a related partial
disordering. On the other hand, an increase in the particle
concentration (at fixed ionic strength) yields more structure in
S(q) (reflecting a more pronounced local ordering in real space)
and a reduction of ξ.
The force oscillation period dslit characterizing the structure
formation in slitpore geometry displays on a qualitative level
the same dependence on the particle concentration and the ionic
strength as S(q). We note that theoretical (MC) results in the
Figure 7. Period of force oscillations for a silica particle suspension experimentally investigated concentration range could only be
confined between two silica surfaces as obtained by MC simulations obtained at low ionic strength; increasing the latter yields a
and by CP-AFM. reduction and eventually a vanishing of the oscillations. In the
concentration range accessible for both methods, theory (MC)
e12%, see Figure 5), the MC results become quite sensitive to and experiment (CP-AFM) yield again results in excellent
the ionic strength in the sense that the oscillations actually vanish agreement. We also note that the resulting lengths characterizing
when there is too much salt in the system. the system’s properties in bulk and confined geometry have
Finally, we compare in Figure 7 theoretical (I ) 10-5/10-4 similar magnitudes (for a given concentration).
mol/L) and experimental (CP-AFM) data for the period of the On the basis of the results of the present work, it seems
force oscillations, dslit. From the theoretical side, only the
worthwhile to investigate in more detail the relation and
calculations with I ) 10-5 mol/L gave reliable results at volume
interchange between bulk properties, on one hand, and properties
fractions below 10%, and even at this very low ionic strength,
in spatial confinement, on the other hand. Furthermore, one may
it was impossible to determine dslit in the range φ e 5 %.
expect that the properties of the confining surfaces have a
Nevertheless, in the concentration range accessible by both
significant effect on the structure formation. Simulations and
theory and experiment, the data in Figure 7 indicate very good
experiments in this direction are underway.
agreement. This concerns not only the concentration dependence
of the force period but also the absolute values of these
quantities. Acknowledgment. We thank Madeleine Kittner for fruitful
discussions regarding the GCMC simulations of the colloidal
Conclusion films. Financial support from the Deutsche Forschungsgemein-
schaft (DFG) via Sonderforschungsbereich 448 “Mesoskopisch
This paper deals with the structure formation of suspensions strukturierte Verbundsysteme” (projects B6 and B10) is grate-
of silica particles in the volume phase and under confinement fully acknowledged. S.H.L.K. thanks for funding via the Emmy-
in a slit-pore geometry. Focusing on length-sensitive quantities, Noether-Programm (DFG). D.Q. thanks the Max Planck Society
Soft Colloidal Particles in Slit-Pore Geometries J. Phys. Chem. B, Vol. 111, No. 6, 2007 1303

and R.v.K. the Fonds der Chemischen Industrie and the (18) Verwey, E. J. W; Overbeek, J. T. G. Theory of Stability of
European Community (RTN EU-project, SOCON) for financial Lyophobic Colloids; Elsevier: Amsterdam, 1948.
(19) Tulpar, A.; van Tassel, P. R.; Walz, J. Y. Langmuir 2006, 22, 2876.
support.
(20) McNamee, C. E.; Tsujii, Y.; Ohshima, H.; Matsumoto, M. Langmuir
2004, 20, 1953.
References and Notes (21) Hansen, J. P.; McDonald, I. R. Theory of Simple Liquids, 3rd ed.;
Academic Press: Amsterdam, 2006.
(1) Stubenrauch, C.; v. Klitzing, R. J. Phys.: Condens. Matter 2003,
15, R1197. (22) Israelachvili, J. Intermolecular and Surface Forces, 2nd ed., 7th
(2) Horn, R. G.; Israelachvili, J. N. J. Chem. Phys. 1981, 75, 1400. printing; Academic Press: San Diego, 1998.
(3) Israelachvili, J. N.; Pashley, R. M. Nature 1983, 306, 249. (23) Ornstein, L. S.; Zernike, F. Proc. Acad. Sci. (Amsterdam) 1914,
(4) Mitchell, D. J.; Ninham, B. W.; Pailthorpe, B. A. J. Chem. Soc., 17, 793.
Faraday Trans. II 1978, 74, 1116. (24) Rogers, F. R.; Young, D. A. Phys. ReV. A 1984, 30, 999.
(5) Horn, R. G.; Evans, D. F.; Ninham, B. W. J. Phys. Chem. 1988, (25) Gisler, T.; Schulz, F. S.; Sticher, H.; Schurtenberger, P.; D’Aguanno,
92, 3531. B.; Klein, R. J. Chem. Phys. 1994, 101, 9924.
(6) Ivanov, I. B.; Dimitrov, D. S. In Thin Liquid Films; Ivanov, I. B., (26) Frenkel, D.; Smit, B. Understanding Molecular Simulations;
Ed.; Marcel Dekker: New York, 1988; p 379. Academic Press: San Diego, 2002.
(7) Perez, E.; Proust, J. E.; Ter-Minassian-Saraga, L. In Thin Liquid (27) Ducker, W. A.; Senden, T. J.; Pashley, R. M. Nature 1991, 353,
Films; Ivanov, I. B., Ed.; Marcel Dekker: New York, 1988; p 891 239.
(8) Basheva, E. S.; Nikolov, A. D.; Kralchevsy, P. A.; Ivanov, I. B.;
Wasan, D. T. Surfactants in Solution; Mittal, K. L., Shah D. O., Eds.; Plenum (28) Riegler, H.; Engel, M. Ber. Bunsen-Ges. Phys. Chem. 1991, 95,
Press: New York, 1991; p 467. 1424.
(9) Nikolov, A. D.; Wasan, D. T. Langmuir 1992, 8, 2985. (29) Hansen, J. P.; Verlet, L. Phys. ReV. 1969, 184, 151.
(10) Piech, M.; Walz, J. Y. J. Colloid Interface Sci. 2002, 253, 117. (30) Russel, W. B.; Saville, D. A.; Schowalter, W. R. Colloidal
(11) Evans, R.; Henderson, J. R.; Hoyle, D. C.; Parry, A. O.; Sabeur, Dispersion; Cambridge University Press: Cambridge (UK),1999; p 347.
Z. A. Mol. Phys. 1993, 80, 755. (31) v. Klitzing, R.; Kolarić, B.; Jaeger, W.; Brandt, A. Phys. Chem.
(12) Götzelmann, B.; Evans, R.; Dietrich, S. Phys. ReV. E 1998, 57, Chem. Phys. 2002, 4, 1907.
6785. (32) Qu, D.; Pedersen, J. S.; Garnier, S.; Laschewsky, A.; Möhwald,
(13) Schoen, M.; Klapp, S. H. L. Nanoconfined Fluids. Soft Matter H.; v. Klitzing, R. Macromolecules 2006, in press.
between Two and Three Dimensions; John Wiley & Sons: New York, 2007; (33) Raman, C. V. Philos. Mag. 1924, 47, 671.
in press. (34) Verlet, L.; Weis J. J. Phys. ReV. A 1972, 5, 939.
(14) Trokhymchuk, A.; Henderson, D.; Nikolov, A. D.; Wasan, D. T.
Langmuir 2001, 17, 4940. (35) Derjaguin, B. V. Kolloid Zeits. 1934, 69, 155.
(15) Klapp, S. H. L.; Schoen, M. J. Chem. Phys. 2002, 117, 8050. (36) Qu, D.; v. Klitzing, R. Unpublished results.
(16) Schoen, M.; Gruhn, T.; Diestler, D. J. J. Chem. Phys. 1998, 301, (37) Härtl, W.; Versmold, H.; Wittig, U. Langmuir 1992, 8, 2885.
301. (38) Goulding, D.; Hansen, J. P. In New Approaches to Problems in
(17) Jönsson, B.; Broukhno, A.; Forsman, J.; Akesson, T. Langmuir Liquid State Theory; Caccamo, C., Hansen, J.-P., Stell G., Eds.; Kluwer
2003, 19, 9914. Academic Publishers: Dordrecht, Boston, 1999; p 321.

You might also like