Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

13052 J. Phys. Chem.

B 2010, 114, 13052–13058

Brownian Dynamics Study of Gel-Forming Colloidal Particles

P. H. S. Santos,*,† O. H. Campanella,† and M. A. Carignano‡


Department of Agricultural and Biological Engineering, Purdue UniVersity, 225 South UniVersity Street, West
Lafayette, Indiana 47907, United States, and Department of Biomedical Engineering and Chemistry of Life
Processes Institute, Northwestern UniVersity, 2145 Sheridan Road, EVanston, Illinois 60208, United States
ReceiVed: June 21, 2010; ReVised Manuscript ReceiVed: September 9, 2010

Brownian dynamics simulations are used to study the formation process of colloidal gels and the effect of
particle concentration on their rheological properties. To model the interaction between particles, we adopted
an R-shifted 12-6 Lennard-Jones (LJ) potential, which allows independent control of the particle size and
attractive range. For a short-ranged potential, the percolated network characteristic of a gel exhibited a
viscoelastic behavior of weak gels. The storage modulus (G′) increased with the particle concentration increase.
Moreover, the dependency of storage modulus (G′) on the particle concentration followed a power law function,
which is commonly reported in the literature for experimental data. Simulating frequency sweep tests showed
that the system behaved as a liquid-like or solid-like material, depending on the frequency applied. The crossover
frequency, i.e., the frequency at which G′ and G′′ are equal, appears to shift slightly to lower values when the
particle concentration increases, suggesting a more solid-like behavior for systems with higher particle
concentration. During gelation, the storage modulus increases as a stretched exponential and reaches a constant
value at long times.

Introduction order to get a better understanding about irreversible colloid


Gels formed by colloidal particles are soft materials rich in aggregation.15-18 Reversible gelation leads to the formation of
solvent with distinctive viscoelastic properties.1 Colloidal transient gels. In this article we will only consider reversible
suspensions can be induced to aggregate by adding destabilizing gels: gels formed by particles that do not make permanent bonds.
agents.2 If the particle concentration is high enough, a desta- Namely, colloidal gels form under temperature and concentration
bilizing colloidal suspension will form a space-filling network conditions that will eventually lead to a phase separation of the
rather than disconnected clusters. The resulting percolated cluster system. Therefore, the gel is a transient state undergoing a very
constitutes the skeleton of the gel. This skeleton has a slow process that tends to evolve toward a heterogeneous system
characteristic internal dimension around 10 to 100 times the with a particle-rich phase in contact with a solvent-rich phase.
particle size, leaving large open regions that are occupied by Under certain conditions, this slow phase separation process
the low molecular weight solvent. There are many examples of may be completely arrested,1,19 and the system remains trapped
colloidal gels in many different areas. Yogurt and many dairy in the gel state. In the cases where the phase separation
products are colloidal gels where the particles are large proteins continues, the time scale of this process is usually very long
or protein aggregates.2-5 In the pharmaceutical industry, col- (days to years) compared with the time scale of the application
loidal topical formulations formed with liposomes are used as of the gel. It is known that particle gels exhibit local ordering;
drug carriers.6 In more recent years, there has been a surge of however, they have no long-range correlations.20,21
interest in nanotechnology and the use of nanocomposite gels
Computer simulations of transient colloidal gels have been
in the design of biocompatible complex materials for biomedical
carried out for systems with short- and long-range attractive
devices and drug delivery systems.7 Biodegradable and noncy-
totoxic three-dimensional (3-D) gels arise as an emergent potentials.22-24 In this context, we refer to short- and long-range
alternative in tissue engineering and in cell culture.8-10 For other interactions as determined by the ratio between the effective
applications, the flow properties of some colloidal gels may also attractive range and the size of the particle. A ratio that is larger
convey high performance and improved safety for rocket or equal to one characterizes a long-range potential, as is the
propellants, combining properties and advantages of solid and case for 12-6 Lennard-Jones. A ratio that is much smaller
liquid fuels.11,12 than 1 corresponds to a short-range potential, as is the case for
The destabilized and aggregated colloidal network may be 72-36 Lennard-Jones. Lodge and Heyes20,25-29 extensively
either permanent or transient, depending on the attractive forces studied systems interacting with Lennard-Jones 2n-n poten-
between particles in the system.13 Very strong attractions or the tials. By taking different values for n, the ratio of range to
formation of chemical bonds at relatively low volume fractions particle size changes, affecting the gelation properties of the
may result in irreversible gels by permanent fractal aggregation, model. It is clear that the magnitude of the range of the
resulting in a large interconnected network.14 Several experi- interparticle attraction relative to their size has a profound effect
ments and computational simulations have been carried out in on the system phase diagram.30 For example, Lennard-Jones
particles have a well known phase diagram with the approximate
* To whom correspondence should be addressed. E-mail: psantos@ shape of an inverted parabola that displays a gas/liquid
purdue.edu.

Purdue University. coexistence curve characterized by a approximate critical point

Northwestern University. E-mail: cari@northwestern.edu. T* ∼ 1.31 and F ∼ 0.316.30 When the attractive range becomes
10.1021/jp105711y  2010 American Chemical Society
Published on Web 09/28/2010
Brownian Dynamics Study J. Phys. Chem. B, Vol. 114, No. 41, 2010 13053

extremely short, the liquid phase becomes metastable and there which relates the friction constant to the radius of the sphere,
is a gas/solid phase transition for those systems.30-32 a, and the viscosity of the pure solvent, η. Finally, the
Weak gels are thermodynamically unstable systems in which Stokes-Einstein relation establishes the connection between the
the active gellant agent forms a fragile and space-filling network. diffusion process and the thermal energy
Spherical particles may form weak gels if the range of the
attraction between the particles is considerable smaller than the k BT kBT
characteristic size of the particles. At sufficiently low temper- D0 ) ) (4)
ature, the short-ranged potentials lead to the formation of gels, γ 6πηa
while longer range potentials (for example Lennard-Jones
12-6) tend to result in a faster phase separation. However, Considering that our model particles interact by a smooth
depending on the time scale studied, a transient gel structure continuous interaction potential, there is no clear definition of
can also be observed for long potential range.28 Performing the particle radius and the range of attraction between the
Brownian dynamics simulations for a short-ranged potential, particles. We will define an effective diameter following the
we show the formation of transient gels at low packing fractions Weeks-Chandler-Andersen theory:33,34
for relatively long simulation times. Most of the efforts on
colloidal systems focus on gelation routes of transient gels and
on the development of a general theoretical framework that
σeff ) ∫0∞ [1 - e-U rep(r)/kBT
]dr (5)
covers all concentration ranges, including high packing fractions.
It is still a challenge to link macroscopic properties of materials where Urep(r) is the repulsive part of the interparticle potential.
with their microstructural properties and conformation. In Therefore, our time unit is expressed in terms of σeff instead of
practice, macroscopic properties of gels provide important and the radius of particle a. Then the time unit can be defined as
useful inputs to the design of new materials, and that is the 2
σeff /4D0.
reason why most fundamental research on gels describes their To model the interaction between colloidal particles, we
properties in terms of viscoelasticity. In this paper we use a adopted an R-shifted 12-6 Lennard-Jones potential:
rheological approach to study short-ranged transient gels and

{ [(
the effect of particle concentration on the gel formation and on
∞, r e R0

) ( )]
their mechanical properties using computer simulations. The
main objective is to get a better understanding about relations U(r) ) 4ε σ 12
-
σ 6
, r > R0 (6)
between microstructural and macroscopic (viscoelastic) proper- r - R0 r - R0
ties of colloidal gels.
where r is the distance between the centers of the particles and
Model and Simulation Details R0 is the shifting distance. In all the simulations, a spherical
We performed Brownian Dynamics (BD) simulations in the cutoff was imposed at rc ) R0 + 2.5σ. This interaction potential
overdamped limit. Namely, we numerically integrate the position allows for an independent control of the size of the particles
Langevin equation: through the shifting distance R0 and the range of attraction
through Lennard-Jones parameter σ. Taking ε as the unit of
energy, and σeff as the unit of length, we can rewrite the
dri 1 interaction potential in the reduced form:
) fi(t) + θi(t) (1)
dt γ
U(r)
)

{ [(
where γ is the friction constant, the same for all the particles, ε
fi(t) is the force acting on particle i due to the interaction with ∞, r' e R0/σeff
the other colloidal particles, and θi(t) is a random noise satisfying
〈θi(t)θi(t)〉 ) 2δijδ(t)kBT/γ. Here kB is the Boltzmann constant
and T the absolute temperature. The hydrodynamic interaction
4
1
r' - R0/σeff ) (
12
-
1
r' - R0/σeff )]6
, r' > R0/σeff (7)

among the particles is beyond the scope of this work and is not
considered. It is also convenient to introduce the reduced temperature
Within the Einstein-Smoluchowski theory of Brownian T* ) kBT/ε. All the simulations presented here are with N )
motion, the mean square displacement of a particle is given 2000 particles and constant volume V. In order to prevent
by: particle overlap, a sufficiently small time step was used (8 ×
10-6σ2eff/4D0) and no special provision was implemented to avoid
the region of infinite potential energy. We use σeff as a nominal
〈r2〉 ) 6D0t (2) size (diameter) of the particle and define the volume fraction
by

where D0 is the diffusion coefficient of the particles at infinite Nπ 3


dilution. From here we can define the unit of time, as is usually φ) σ (8)
V 6 eff
done in BD simulations, as a2/D0, where a is the radius of the
particles. This time unit is proportional to the time needed by
an isolated particle to diffuse a distance proportional to its size. For particles interacting through a potential with R0 ) 9,
At this point and to complete the computational framework, σ ) 1, and T* ) 0.2, the nominal diameter or σeff corresponds
we introduce Stokes law: to 10.075σ.
The initial configuration for all simulations corresponds to a
random configuration obtained by running a short molecular
γ ) 6πηa (3) dynamics simulation at high temperature. The initial potential
13054 J. Phys. Chem. B, Vol. 114, No. 41, 2010 Santos et al.

Figure 1. Potential energy as a function of time for a system interacting Figure 3. Potential energy of a system with 2000 particles, interacting
via standard Lennard-Jones 12-6 potential. The insert shows the initial with the R-shifted potential with R0 ) 9 and σ ) 1. Different colors
100 time units in which the particles form a percolated cluster. The correspond to different volume fractions: 0.10 (black), 0.13 (red), 0.16
drop in U between t ) 200 and t ) 300 corresponds to the breaking in (green), and 0.20 (blue). For the intermediate volume fraction, we run
the connectivity of the percolated cluster in one of the Cartesian two independent simulations starting from a different initial condition
directions, and the subsequent reorganization of the system’s structure. and with different time steps ((8 × 10-6)σeff
2
/4D0 and (8 × 10-7)σeff 2
/
4D0). Longer simulation was performed for the system with volume
fraction 0.13.
energy is zero or a very small negative number for all the cases,
since there is practically no contact between the particles. indication of the advance of the separation process. Figure 2D
shows the final structure, at t ) 10 000, that has a cylindrical
Results
shape as the connectivity across the box is maintained only in
We started our work considering the pure Lennard-Jones one of three directions. Figure 2D also reveals the crystalline
system. This is a particular case of an interaction potential with order in parts of the system. Lodge and Heyes20,25-29 previously
R0 ) 0. Even though this system is not in our main interest, it reported the formation of transient gels using a Lennard-Jones
represents a good test case for the simulation methodology due 2n-n potential, volume fractions of 0.10, 0.16, and 0.20, and
to the existence of previous works in similar systems. In Figure T* ) 0.3. They used 864 particles and a total simulation time
1 we show the time evolution of the potential energy of the of 81 time units. In their conclusion, they discussed the
system, for a simulation corresponding to a volume fraction of surprisingly gel-like behavior of the longest range potential
φ ) 0.13 and a temperature T* ) 0.2, and in Figure 2 we show system (n ) 6). Moreover, they commented that there were no
a series of snapshots characterizing the whole process. The signs of complete phase separation for the long-range system,
potential energy shows a monotonic decreasing behavior as a which could be due to the time scale of their simulations, since
function of time. At first the decrease is very sharp as the those systems tend to phase separate more rapidly. We also
particles cluster themselves to form a gel-like structure with observed visually a gel-like structure using Lennard-Jones
several branches. Figure 2A shows a snapshot of this transient potential (n ) 6), volume fraction of 0.13, and T* ) 0.2 at
gel that remains as such up to a time of t ≈ 200 (in reduced short simulation times (t ) 100). However, if the simulation
units). Then there is a major reaccommodation of the branches goes on for longer times, the system no longer exhibits a
as they cluster together to form a single bulky structure, still percolated network characteristic of gel, which shows the
percolated, but reducing its surface area as shown in Figure 2B important role that the potential range and time scale play in
which corresponds to t ) 500. The formation of the central transient gels.
bulky structure is reflected by a change in the slope of the We now consider particles interacting through a short-range
potential energy curve between t ) 200 and t ) 300. The system potential, as is the case with R0 ) 9, and σ ) 1. In Figure 3 we
keeps evolving and the connectivity across the box in one of show the time evolution of the potential energy for six
the directions is lost, as shown in Figure 2C corresponding to trajectories corresponding to four different volume fractions.
t ) 1000. The breaking of the space-filling network is an All the simulations were done at the same temperature T* )

Figure 2. Snapshots of the standard Lennard-Jones system at (from left to right) t ) 100, t ) 500, t ) 1000, and t ) 10 000. In the first
two snapshots, the system forms a percolated cluster interconnected through all the faces of the simulation box. At t ) 1000 the connectivity
in the vertical direction has been lost, and at t ) 10 000 the connectivity in a second direction is also broken and the cluster adopts a
cylindrical shape.
Brownian Dynamics Study J. Phys. Chem. B, Vol. 114, No. 41, 2010 13055

Figure 4. Snapshots of a system with volume fraction of 0.13 interacting with the R-shifted potential with R0 ) 9 and σ ) 1 at (from left to right)
t ) 100, t ) 500, t ) 1000, and t ) 10 000. The typical network of a gel remains up to 10 000σeff 2
/4D0.

0.2, and two trajectories starting from different initial conditions


and using a different time step ((8 × 10-6)σeff 2
/4D0 and (8 ×
10-7)σeff
2
/4D0) have been obtained for the intermediate volume
fractions. Judging by the potential energy and also visually
inspecting the simulated trajectories, our results show no
important differences between these two time steps. The volume
fraction has some effect on the kinetics of gelation and on the
final potential energy of the system. All four different volume
fractions show qualitatively the same behavior. There is an initial
fast (t < 10) clustering, followed a process of gel formation (10
< t < 100). For t > 100, the gel keeps evolving at a very low
rate, and no kinetically arrested final state is ever reached. Note
that Figure 3 shows the time evolution for one of the simulations
corresponding to a volume fraction of 0.13 up to t ) 10 000,
confirming that the system keeps decreasing its potential energy,
indicating an accommodation of the particles in more energeti-
cally favorable conformations, and may be a result of aging
process.35 Comparing the structures corresponding to t ) 1000
and t ) 10 000, we see a minor global reaccommodation, but
a larger degree of local ordering is achieved at long times. As
shown in Figure 4, the characteristic network of a gel remains
for the longest simulation time (t ) 10 000). The times in which
the snapshots were taken are equivalent to those presented in
Figure 2 for the pure Lennard-Jones particles.
In Figure 5 we present (A) the time evolution of the size of
the largest cluster in the system, and (B) the time evolution of
the average cluster size for the same systems considered in Figure 5. (A) Size of the maximum cluster as a function of time for
Figure 3. A particle is considered to be part of the cluster if the the system characterized by R0 ) 9 and σ ) 1, and different volume
distance to any of the particles already in the cluster is smaller fractions. (B) Average cluster size as a function of the packing fraction.
than a cutoff distance that corresponds to the first minimum in Different colors correspond to different volume fractions: 0.10 (black),
0.13 (red), 0.16 (green), and 0.20 (blue).
the particle-particle radial distribution function. The four
systems display the same behavior: except for short time
fluctuations, there is a monotonic increase in both the average respectively. These peaks correspond to an fcc lattice. The
and the largest cluster sizes. The time scale to reach total reason why only eight first neighbors are found and not twelve,
percolation is longer for the most diluted systems. This result as is the optimum for hexagonal close packing, is that the system
is expected, since the average initial interparticle distance is has a large exposed surface, as can be seen in Figure 4. The
smaller in the more concentrated system favoring the cluster insert of Figure 6 shows a detail of the local crystalline structure
development. However, it is interesting to note that the time to corresponding to φ ) 0.13, showing sections of two crystalline
achieve total percolation for φ ) 0.10 is approximately 2 orders grains in different relative orientations.
of magnitude longer than the corresponding time for φ ) 0.20. From an experimental point of view, gels can be defined as
This strong dependency could be interpreted in terms of the a system with at least two components: (i) a solvent and (ii) a
slow diffusion of the intermediate clusters that need to travel a gellant that forms a macroscopic space-filling structure that
longer distance in the systems with low concentrations in order provides mechanical rigidity.36 In terms of deformation and flow
to aggregate into a single percolated structure. properties, complex materials such as gels do not follow
The overall structure of the systems at t ) 1000 corresponds Newton’s law of viscosity or even Hooke’s law of elasticity.
to a collection of small clusters fused together. The local Unsteady shear material functions are commonly used to
structure of the system is displayed in Figure 6 for the four characterize gel properties and address their applications. Small
studied systems. The number of first and second neighbors is amplitude oscillatory shear (SAOS) strain is applied to the fluid,
8.0 and 11.6, respectively, and the positions of the first and producing a shear stress wave of the same frequency but usually
second peaks in the g(r) are approximately 1.0σeff and 1.4σeff, not in phase. Knowing the resulting shear stress and phase
13056 J. Phys. Chem. B, Vol. 114, No. 41, 2010 Santos et al.

Figure 6. Cumulative g(r) for the system characterized by R0 ) 9


and σ ) 1 at different volume fractions: 0.10 (black), 0.13 (red), 0.16
(green), and 0.20 (blue). The insert shows the crystalline structure of
two flocs in different relative orientation. The g(r) and the snapshot
correspond to t ) 1000 σeff2
/4D0.

difference between stress and strain waves, we can calculate


two properties: storage modulus (G′) and loss modulus (G′′).37
The storage modulus (G′) describes the solid-like (elastic)
characteristic of the system whereas the loss modulus (G′′)
describes its liquid-like (viscous) behavior. These properties are
often determined experimentally38-42 to characterize the rheo-
logical properties of gels since most of them are viscoelastic
materials, i.e., they exhibit both viscous and elastic properties.
In practice, they can describe the material in terms of strength
and weakness, which are useful properties in designing materials
Figure 7. Simulated small amplitude oscillatory shear (SAOS) for the
and products. In addition, previous simulations also calculated system R0 ) 9, σ ) 1, and φ ) 0.13, taken at t ) 1000σeff 2
/4D0: (A)
the rheological properties of colloidal gels for Lennard-Jones Simulated oscillatory strain (black), stress response (red), and phase
potential and for systems with irreversible bonds.15,18,28,29,43,44 delay (δ) at ω ) 100. (B) Stress response fitting for G′ and G′′
However, it is still a challenge in different areas such as material calculations (green). Lxy and Pxy are the xy components of stress and
and food sciences to optimize the mechanical properties of strain tensors.
systems and to obtain a better understanding of the relation
between microscopy properties and rheology.2 the gel (t ) 1000), the shear stress response, and the difference
In order to calculate the viscoelastic properties of the in phase between them. Figure 7B shows the fitting of the shear
simulated gels, i.e., the storage modulus (G′) and loss modulus stress response, and constants were used to compute the storage
(G′′), we followed an approach similar to the experimental modulus (G′) and loss modulus (G′′):
setup.15 Namely, we performed Brownian dynamics simulation
on systems undergoing a small amplitude oscillatory shear. The
simulation box was deformed by displacing the top wall normal σ0
to the z-axis following an oscillatory pattern with a fixed G'(ω) ) cos δ (11)
γ0
frequency ω. A general triclinic simulation box can be described
by a 3 × 3 matrix. A cubic box is a particular case of a triclinic σ0
box, with all the diagonal elements equal to the length of the G'(ω) ) sin δ (12)
γ0
box’s edge, and the off diagonal elements equal to zero. The
oscillatory shear strain that we apply to the simulation box
corresponds to set the xy component of the matrix equal to For a purely elastic material, the stress response is in phase
with the strain whereas for a purely viscous material the angle
γxy(t) ) γ0 sin(ωt) (9) phase between them corresponds to 90°. The phase angle δ for
viscoelastic materials ranges from 0 to 90°.
The first goal in this work was to determine the regime in
This oscillatory shear was applied for a very short time, as which the stress is linearly related to strain, a standard test
compared to the time required for the system to have a performed in experimental rheology to find the range of linear
significant change in structure. To calculate the rheological viscoelasticity of the sample. Strains of different magnitudes
properties of the system, we used the xy-component of the stress were applied to the gel, and the storage modulus (G′) was
tensor, which follows an oscillatory function with the same calculated at a constant frequency (ω ) 10 000 4D0/σ2eff). Figure
frequency of the applied strain: 8 shows the strain dependence of G′ for gels of different packing
fractions. For all particle concentrations, G′ does not vary for a
σxy(t) ) σ0 sin(ωt + δ) (10) range of small strains and exhibits significant reduction when a
certain strain is achieved. This range characterizes the gel’s
linear viscoelastic region, and the strain in which this relation-
The amplitude σ0 and the phase shift δ are obtained from a ship is no longer linear is defined as critical deformation value
fit. Figure 7A shows the simulated sinusoidal strain applied to (γc). In addition, Figure 8 shows that the higher the particle
Brownian Dynamics Study J. Phys. Chem. B, Vol. 114, No. 41, 2010 13057

Figure 8. Storage modulus (G′) dependency on maximum strain for


the system R0 ) 9, σ ) 1, and different volume fractions: 0.10 (black), Figure 9. Storage (G′) and loss (G′′) moduli dependency on frequency
0.13 (red), 0.16 (green), and 0.20 (blue). All calculations were for the system R0 ) 9, σ ) 1, and different volume fractions: 0.10
performed at a frequency of ω ) 10 000. The calculation was performed (black), 0.13 (red), 0.16 (green), and 0.20 (blue). All calculations were
at t ) 1000σeff
2
/4D0. performed in the linear viscoelastic region (γ0 ) 2.5%) and at t )
2
1000σeff /4D0. The insert shows G′ and G′′ as a function of volume
fraction at fixed frequency ω ) 1000.
concentration, the higher the storage modulus (G′), for that
particular frequency. Shih et al.45 developed the scaling fractal
theory which considers that the gel structure is formed by material. As the frequency increases, both of them (G′ and G′′)
clusters that aggregate with each other. This theory allows the increase. When a certain frequency is achieved, the storage
collection of structural information on the gels from experi- modulus (G′) exhibits higher values than those of the loss
mental rheological data. From experimental datan they showed modulus (G′′). The frequency at which G′ and G′′ are the same
a power law dependency of the storage modulus (G′ ∝ Øn) and is defined as the crossover frequency. At higher frequencies,
of the critical deformation (γc ∝ Øn) on the particle concentra- the solid-like characteristic of the material is predominant.
tion. Two different behaviors were found to describe the gels, The effect of particle concentration on this solid/liquid like
and they were defined in terms of strong and weak links. In a “behavior transition” can also be observed in Figure 9. As the
weak-link regime, both G′ and γc increase with increasing particle concentration is increased, the crossover frequency
particle concentration. In a strong-link regime, G′ also increases shifted to lower frequency values, suggesting that higher
as concentration increases (higher dependence than in a weak- concentration systems exhibit a more solid-like behavior. For
link gel); however, γc decreases as concentration increases. In the range studied, the higher the concentration, the higher the
addition, the interactions between the colloidal flocs (aggregates) frequency range in which the material exhibits solid-like
are stronger than the interactions within flocs in the strong-link behavior. The variation of the crossover frequency is not large,
systems. In contrast, in a weak-link gel the interaction within but the change in behavior at ω ) 1000 4D0/σeff 2
confirm this
flocs is stronger than that between aggregates. trend. In the insert of Figure 9, we display G′ and G′′ at ω )
We also observe a power law dependency of G′ on particle 1000 4D0/σeff2
as a function of the volume fraction φ, showing
concentration for our transient gels. The storage modulus (G′) a larger slope for G′ than for G′′. This dependency implies that
satisfies a power law function with an index of 1.04. A weak the transition to solid-like behavior happens at lower frequencies
dependence of storage modulus on the particle concentration is
for higher particle concentration. All the macro properties
observed, which means that G′ increases slowly as packing
presented so far were calculated for the gel network formed
fractions increases. The critical deformation (γc), which limits
after a time of 1000σ2eff/4D0. In order to obtain some information
the linear dependency of G′ on strain, showed a weak
on the rheology of the system during the gel formation,
dependence on particle concentration for the studied range.
conformations of the system at different times were used to
These results suggest that our system behaves as a weak-link
determine G′ during the aggregation process. In Figure 10, the
colloidal gel.
The characteristic behavior of a weak gel shows up also when calculated G′ for φ ) 0.10 and φ ) 0.20 packing fractions is
simulated frequency sweep tests are performed in the linear plotted as function of time. This calculation was performed at
region (γ ) 2.5%) for different volume fractions. This test is a frequency in which the gels exhibited predominantly solid-
commonly used in experimental studies to investigate the like properties (ω ) 10 000). As the aggregation process goes
frequency dependence of the storage modulus (G′) and loss on, the storage modulus increases, approaching an asymptotic
modulus (G′′). From rheological point of view the flow value. The higher the particle concentration, the higher the
properties of weak gels depend on the angular frequency (ω) asymptotic value, implying that solid-like behavior is enhanced
of the applied stress.36 The storage (G′) and loss (G′′) modulus by concentration. In addition, the G′ values during gel formation
are frequency dependent, exhibiting solid-like or liquid-like increase exponentially according to a stretched exponential of
the form G′ ) G∞ exp-(B/t) . Fitting the data to this function the
k
properties depending on the frequency applied. Solid-like gels,
i.e., strong gels, exhibit an elastic modulus much larger than constants G∞ ) 2.28, B ) 7.29, k ) 0.36, and G∞ ) 4.81, B )
the viscous modulus (G′ . G′′). Figure 9 shows the simulated 2.37, k ) 0.34, are obtained for φ ) 0.10 and φ ) 0.20,
frequency sweep tests for our transient gels. For all particle respectively. The stretching factor k is approximately the same
concentrations, the colloidal gels displayed a typical viscoelastic for the two cases, but the time constant B is significant larger
behavior of a weak gel. At low frequencies, the loss modulus for the lower concentration. A stretched exponential time
(G′′) is the dominant response of the system, which means that dependency has been observed in the other properties of the
for that frequency range, the system behaves as a liquid-like gels, such as the relaxation of incoherent scattering function.46
13058 J. Phys. Chem. B, Vol. 114, No. 41, 2010 Santos et al.

References and Notes


(1) Zaccarelli, E. J. Phys.: Condens. Matter 2007, 19, 323101.
(2) Mezzenga, R.; Schurtenberger, P.; Burbidge, A.; Michel, M. Nat.
Mater. 2005, 4, 729.
(3) Donald, A. M. Soft Matter 2008, 4, 1147.
(4) Aichinger, P. A.; Michel, M.; Servais, C.; Dillmann, M. L.; Rouvet,
M.; D’Amico, N.; Zink, R.; Klostermeyer, H.; Horne, D. S. Colloid Surf.,
B 2003, 31, 243.
(5) Schurtenberger, P.; Stradner, A.; Romer, S.; Urban, C.; Scheffold,
F. Chimia 2001, 55, 155.
(6) Esposito, E.; Drechsler, M.; Mariani, P.; Sivieri, E.; Bozzini, R.;
Montesi, L.; Menegatti, E.; Cortesi, R. Int. J. Cosmet. Sci. 2007, 29, 39.
(7) Schexnailder, P.; Schmidt, G. Colloid Polym. Sci. 2009, 287, 1.
(8) Wang, Q.; Wang, J.; Lu, Q.; Detamore, M. S.; Berkland, C.
Biomaterials 2010, 31, 4980.
(9) Xie, B. J.; Parkhill, R. L.; Warren, W. L.; Smay, J. E. AdV. Funct.
Mater. 2006, 16, 1685.
Figure 10. Storage modulus calculation for the system R0 ) 9, σ ) (10) Pek, Y. S.; Wan, A. C. A.; Shekaran, A.; Zhuo, L.; Ying, J. Y.
1, and volume fractions of 0.10 (squares) and 0.20 (circles) during gel Nat. Nanotechnol. 2008, 3, 671.
formation. The calculations were performed in the linear viscoelastic (11) Natan, B.; Rahimi, S. Combust. Energ. Mater. 2002, 172. DOI:
region (γ0 ) 2.5%), at frequency ω ) 10 000 and t ) 1000σeff 2
/4D0. 10.1615/IntJEnergeticMaterialsChemProp.v5.i1-6.200.
(12) Santos, P. H. S.; Arnold, R.; Anderson, W. E.; Carignano, M. A.;
Conclusions Campanella, O. H. Eng. Lett. 2010, 18, 41.
(13) Dickinson, E. J. Colloid Interface Sci. 2000, 225, 2.
Brownian dynamics simulations were performed to study the (14) Dietsch, H.; Malik, V.; Reufer, M.; Dagallier, C.; Shalkevich, A.;
relationship between microstructural and macroscopic properties Saric, M.; Gibaud, T.; Cardinaux, F.; Scheffold, F.; Stradner, A.; Schurten-
of transient colloidal gels. We initially considered a standard berger, P. Chimia 2008, 62, 805.
(15) Whittle, M.; Dickinson, E. Mol. Phys. 1997, 90, 739.
Lennard-Jones 12-6 system corresponding to a volume (16) Lin, M. Y.; Lindsay, H. M.; Weitz, D. A.; Ball, R. C.; Klein, R.;
fraction of φ ) 0.13 and a temperature T* ) 0.1. A transient Meakin, P. Nature 1989, 339, 360.
gel is formed at relatively short times (t ≈ 100), but due to (17) Del Gado, E.; Fierro, A.; de Arcangelis, L.; Coniglio, A. Europhys.
major reaccommodation of the branches, the system collapses Lett. 2003, 63, 1.
(18) Wijmans, C. M.; Dickinson, E. J. Chem. Soc., Faraday Trans. 1998,
to form a single bulky structure with some crystalline local order. 94, 129.
To model particles interacting through a short-range potential, (19) Foffi, G.; De Michele, C.; Sciortino, F.; Tartaglia, P. J. Chem. Phys.
we adopted a R-shifted 12-6 Lennard-Jones model. Since the 2005, 122.
potential range plays an important role in the gel formation,24 (20) Lodge, J. F. M.; Heyes, D. M. J. Chem. Phys. 1998, 109, 7567.
(21) Toledano, J. C. F.; Sciortino, F.; Zaccarelli, E. Soft Matter 2009,
the advantage of using this interaction potential is the indepen- 5, 2390.
dent control of the size of the particles and the range of (22) Bijsterbosch, B. H.; Bos, M. T. A.; Dickinson, E.; vanOpheusden,
attraction. In addition, we observed that for the case corre- J. H. J.; Walstra, P. Faraday Discuss. 1995, 51.
(23) Bos, M. T. A.; vanOpheusden, J. H. J. Phys. ReV. E 1996, 53, 5044.
sponding to R0 ) 9 and σ ) 1, a percolated network (24) Lu, P. J.; Zaccarelli, E.; Ciulla, F.; Schofield, A. B.; Sciortino, F.;
characteristic of a gel is formed for packing fractions varying Weitz, D. A. Nature 2008, 453, 499.
from 0.10 to 0.20 and it remains for relatively long simulation (25) Lodge, J. F. M.; Heyes, D. M. Mol. Simul. 1996, 18, 155.
2
times (up to 10 000σeff /4D0). Using a rheological approach, we (26) Lodge, J. F. M.; Heyes, D. M. Phys. Chem. Liq. 1996, 31, 209.
(27) Lodge, J. F. M.; Heyes, D. M. J. Chem. Soc., Faraday Trans. 1997,
characterized those systems and determined their viscoelastic 93, 437.
behavior to investigate the effect of particle concentration on (28) Lodge, J. F. M.; Heyes, D. M. Phys. Chem. Chem. Phys. 1999, 1,
their mechanical properties. Our simulation results using the 2119.
R-shifted Lennard-Jones potential show similar trends and (29) Lodge, J. F. M.; Heyes, D. M. J. Rheol. 1999, 43, 219.
(30) Vliegenthart, G. A.; Lodge, J. F. M.; Lekkerkerker, H. N. W. Phys.
behavior to those observed experimentally for weak gels. First A (Amsterdam, Neth.) 1999, 263, 378.
of all, the regime in which the stress is linearly related to strain (31) Hagen, M. H. J.; Frenkel, D. J. Chem. Phys. 1994, 101, 4093.
was determined for all the considered concentrations. As particle (32) Hasegawa, M.; Ohno, K. J. Phys.: Condens. Matter 1997, 9, 3361.
concentration increased, the storage modulus (G′) in the linear (33) Andersen, H. C.; Weeks, J. D.; Chandler, D. Phys. ReV. A: Gen.
Phys. 1971, 4, 1597.
viscoelastic region increased. Moreover, it satisfies a power law (34) Barker, J. A.; Henderson, D. ReV. Mod. Phys. 1976, 48, 587.
function with index of 1.04. The simulated frequency sweep (35) d’Arjuzon, R. J. M.; Frith, W.; Melrose, J. R. Phys. ReV. E: Stat.
tests showed that our colloidal gels behave as a weak gel. For Nonlinear Soft Matter Phys. 2003, 67, 061404.
(36) Terech, P.; Weiss, R. G. Chem. ReV. 1997, 97, 3133.
all particle concentrations, the gels displayed a typical vis- (37) Understanding rheology; Morrison, F. A., Ed.; Oxford University
coelastic behavior. As the applied frequency is increased, the Press: New York, 2001.
system undergoes a transition from a liquid-like to a solid-like (38) Kamp, S. W.; Kilfoil, M. L. Soft Matter 2009, 5, 2438.
material. The frequency at which G′ and G′′ are the same, which (39) Tan, L. J.; Liu, S. P.; Pan, D.; Pan, N. Soft Matter 2009, 5, 4297.
(40) van Riemsdijk, L. E.; Sprakel, J. H. B.; van der Goot, A. J.; Hamer,
is defined as the crossover frequency, shifts slightly to lower R. J. Food Biophys. 2010, 5, 41.
values as the particle concentration increases, suggesting that (41) Asai, H.; Masuda, A.; Kawaguchi, M. J. Colloid Interface Sci. 2008,
higher concentration systems exhibit a more solid-like behavior. 328, 180.
During the gelation process, the storage modulus (G′), which (42) Oppong, F. K.; Coussot, P.; de Bruyn, J. R. Phys. ReV. E: Stat.
Nonlinear Soft Matter Phys. 2008, 78, 021405.
in practice characterizes the gel strength, increases as a stretched (43) Whittle, M.; Dickinson, E. J. Chem. Soc., Faraday Trans. 1998,
exponential as the aggregation goes on, reaching a constant value 94, 2453.
for very long times. (44) Whittle, M.; Dickinson, E. J. Chem. Phys. 1997, 107, 10191.
(45) Shih, W. H.; Shih, W. Y.; Kim, S. I.; Liu, J.; Aksay, I. A. Phys.
Acknowledgment. This work was supported by the U.S. ReV. A 1990, 42, 4772.
(46) Del Gado, E.; Kob, W. Soft Matter 2010, 6, 1547.
Army Research Office under the Multi-University Research
Initiative (MURI) grant number W911NF-08-1-0171. JP105711Y

You might also like