You are on page 1of 104

ADVANCES IN

COLLOID AND
INTERFACE
Advances in Colloid and Interface Science
ELSEVIER 68 (1996) 97-200
SCIENCE

Correlation of viscoelastic properties of stable and


flocculated suspensions with their interparticle
interactions
Th.F. Tadros
ZENECA Agrochemicals (formerly part of the ICI Group), Jealott’s Hill Research Station,
Bracknell, Berkshire RG12 6EY, UK

This review starts with a general introduction on the properties of concentrated


suspensions. The distinction between “dilute “, “solid” and “concentrated” suspensions
is given in terms of the balance between Brownian, hydrodynamic and interparticle
interactions. A section is given on inter-particle interactions and their combinations. The
four different types of interactions, namely hard-sphere, electrostatic, steric and van der
Waals are described. The flocculation of both electrostatically and sterically stabilized
suspensions is also discussed. The next section covers the principles of rheological
measurements. Transient (static), dynamic (oscillatory), shear wave propagation and
steady state measurements are described. The last part of the review deals with the
viscoelastic properties of concentrated suspensions. Four different systems were de-
scribed and examples were given: (al Suspensions with hard-sphere interactions; (b)
Stable systems with SOR(electrostatic) interaction; (c) Sterically stabilised systems; (d)
Flocculated and coagulated systems. Both weakly and strongly flocculated systems were
discussed.
In the above review, particular emphasis was given to the relationship between the
viscoelastic properties of concentrated suspensions and their inter-particle interactions.
As far as was possible, the results obtained from rheological measurements were
quantitatively analysed in terms of such interparticle forces. The review demonstrated
that such correlation is generally followed and this illustrated the powerful use of
rheology for studying interparticle interactions.

0001~8686/96/$32.00 0 1996 - Elsevier Science B.V. All rights reserved.


PZZzSOOOl-8686(96)00305-3
98 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

Contents

Summary.. ..................................... ..Q 7


1. Introduction. ................................... .98
2. Distinction between dilute and concentrated suspensions ........... 99
3. Interparticle interactions and their combinations ............... 105
3.1. Interparticle interactions .......................... 105
3.2. Combination of interparticle interactions ................. 113
3.3. Flocculation of suspensions ......................... 117
4. Principles of rheological measurements ..................... 128
4.1. Transient (static) measurements ...................... 129
4.2. Dynamic (oscillatory) measurements ................... 133
4.3. Shear wave propagation measurements .................. 136
4.4. Steady-state measurements ........................ 137
5. Viscoelastic properties of concentrated suspensions .............. 139
5.1. Suspensions with hard-sphere interactions ................ 140
5.2. Stable systems with soft (electrostatic) interactions ........... 143
5.3. Sterically stabilised suspensions ...................... 153
5.4. Flocculated and coagulated systems .................... 167
5.4.1. Weakly flocculated systems ..................... 170
5.4.2. Strongly flocculated systems .................... 181
References ....................................... 197

1. Introduction

Solid/liquid dispersions (suspensions), both of the aqueous and non-


aqueous type, find applications in many industrial preparations, of
which the following may be worth mentioning: paints, dyestuffs, pig-
ments, paper coatings, printing inks, cosmetics, household products,
ceramics, pharmaceuticals and agrochemicals. More recently, non-
aqueous suspensions of magnetic oxides have attracted considerable
attention because of their application in the electronic industry. There
has been also some interest in non-aqueous suspensions of silica and
many other polymer particles for application in electro-rheological flu-
ids. The control of the properties of these suspensions is critical in the
stages involved in their preparation, their long-term stability and in
their subsequent application. Some of the parameters which control
such properties are: particle size and shape distribution, interparticle
interaction forces and the volume fraction of the dispersed phase. These
parameters control the flow characteristics (rheology) of these systems.
Control of rheology, or viscoelastic properties of the suspension is crucial
during preparation of the system, its behaviour on standing (e.g., its
settling characteristics) and its application. For example, if one aims at
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 99

preparation of a suspension with high volume fraction, such as for


example in ceramics, it is essential to maintain reasonable flow during
moulding and the solid-packed system should “set” at the required
shape. With many paint systems, thixotropy is essential for the appli-
cation of the paint. Thus, the “gel-like” paint must flow during applica-
tion (by the shear action of the paint brush or roller) for adequate
coverage of the substrate. Once the shear force is removed, the system
must recover its “gel” structure within a well defined period. If the
recovery is too fast, the paint film will leave brush marks, whereas if
the recovery is slow, some dripping may occur.
The present review is aimed at correlating the rheological properties
of suspensions with interparticle interactions. As we will see later
suspensions may show viscous, elastic or viscoelastic (mixed viscous and
elastic) response depending on the ratio between the relaxation time of
the system (which depends on its volume fraction, particle size, hydro-
dynamic and interparticle interaction) and the experimental time. Be-
fore describing how viscoelasticity of a concentrated dispersion may be
investigated, it is essential to describe two main topics. Firstly, one
should be able to distinguish between “dilute”, “solid” and “concen-
trated” suspensions. These systems show predominantly viscous, pre-
dominantly elastic and a mixed viscoelastic response respectively. These
responses are determined by the volume fraction, hydrodynamic and
interparticle interactions. The second part of this review will summarise
the inter-particle interactions and their combinations. Particular atten-
tion will be paid to the formation of three dimensional structural units
which are produced at high volume fractions. A brief description of
viscoelasticity and its measurement will be given in the following
section. The last part will give examples and a comprehensive discussion
of the viscoelastic properties of concentrated suspensions, both stable
and flocculated. The correlation between these viscoelastic results and
the interparticle interactions will be stressed.

2. Distinction between dilute and concentrated suspensions

A possible distinction may be obtained if one considers the role of


hydrodynamic and interparticle interactions as the number concentra-
tion in a suspension is increased. At one extreme, a suspension may be
considered dilute if the thermal (Brownian) motion of the particles
predominates over the effect of hydrodynamic and interparticle forces
[ 1,2]. In these dilute systems, the distance between the particle surfaces
100 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

is large compared to the range of interaction forces, whether hydrody-


namic or surface. In this case, the particle translational motion is large
and in such dispersions only occasional contact occurs between the
particles, i.e. the particles do not “see” each other until a collision occurs
131. In other words, the particle interactions can be represented by
two-body collisions. Provided the gravitational force can be neglected
(i.e. no settling occurs), the properties of such dilute systems are essen-
tially time-independent. Any time-average quantity, such as light scat-
tering, osmotic pressure and viscosity, may be extrapolated to infinite
dilution to obtain fundamental properties of the system such as its
particle or hydrodynamic radius. In these dilute systems, the particles
move independently and their Brownian motion has been examined
experimentally in detail by Perrin 141and theoretically by Einstein 151.
Such dilute systems are seldom encountered in industrial preparations
and their investigations provide only fundamental information.
As the particle number concentration in a suspension is increased,
the volume of space occupied by the particles relative to the total volume
increases and thus a proportion of space is excluded in terms of its
occupancy by any single particle 111.The probability of particle-particle
interaction, both hydrodynamic and surface increases. In this case, the
hydrodynamic and surface forces play dominant roles in determining
the properties of the system, such as its structure in space and its flow
characteristics. With gradual increase in particle number concentration,
a situation is reached whereby the interparticle distances become rela-
tively small compared to the particle radius. In this case, any particle in
the system interacts with many neighbours and the repulsive interactions
produce specific order between the particles to the extent that a highly
developed structure is reached. When the spacing between the particles
is very small, they can only undergo vibration with amplitudes smaller
compared to the particle size. The system will behave like a solid
showing elastic response and it shows no time dependence of its prop-
erties. Such systems are conveniently referred to as “solid suspensions”,
as for example encountered with ceramics (the so-called green body).
In between the above two extremes of dilute and solid suspensions
one may define concentrated suspensions. In this case, the volume
fraction is sufficiently high for many-body interaction to occur. Both
hydrodynamic and surface interactions play a role in determining the
properties of the system. However, the interparticle distance in this case
is comparable to the particle size. This allows the particles to diffuse,
albeit slowly, and the system shows time dependence of its properties.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 101

The particle arrangements show spatial correlations, i.e. the position of


one particle in space becomes determined by the position of the other
particles in the neighbourhood. In addition, the diffusion coefficient of
the particles become a strong function of the volume fraction, i.e. the
system shows temporal correlations.
Thus, for any suspension the first step towards understanding its
properties is to consider the arrangements of particles in the system.
Three cases may be considered: random arrangement with free diffusion
(dilute or “vapour-like”), loosely ordered (concentrated or “liquid-like”)
and highly ordered (solid or “crystal-like”). This microstructure of the
suspensions may be investigated using small-angle X-ray or neutron
scattering techniques [6,7]. Once the system is understood at the micro-
structure level, one must investigate how the particl+particle interactions
influence the macroscopic properties of the suspension, such as its osmotic
pressure, its elasticity and flow characteristics. Clearly the techniques that
need to be used for studying concentrated suspensions should address
themselves to such spatial and temporal correlations [6,7]. Such meth-
ods should enable the bulk properties to be related to the microstructure.
A convenient method of describing the structure of suspensions is
through the particle radial distribution function, as used to describe the
structure of liquids [1,21. Consider a system containing NP particles in
a volume V, then the average macroscopic number density, po, is ex-
pressed by,

32
PO= V

However, if the container is examined more closely on a microscopic


scale, one can obtain the distribution of particles around any reference
particle. This is illustrated in Fig. 1, in which the centres of each particle
surrounding an arbitrary chosen “central” particle (the black one) are
placed in concentric shells of thickness dr at a distance r from the central
particle. It is clear that, in the immediate vicinity of the central particle,
there is a space in which the particle density is zero. With an increasing
distance r from the centre of the chosen particle, the number of particles
with their centres in any shell of thickness dr varies. Hence, one may
define a number density p(r) which varies with r and which describes
the distribution of the particles. This density function will have two
limiting values: p(r) + 0 as r + 2R (where R is the particle radius) and
p(r) + p. as r * 00.
102 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

Fig. 1. Microscopic view (schematic) of the distribution of particles around a central one.

The pair distribution function, g(r), can be defined as,

r
g(r) =y (2)
0

which has the properties g(r) + 0 as r + 2R and g(r) + 1 as r + 00. The


radial distribution function is simply 4nr2p(r). The quantity g(r) allows
one to describe the distribution of particles in space around a central
particle. An additional feature ofg(r) is that it is directly related to the
potential Q(r) of mean force acting between the particles, i.e.,

1
--Nr>
g(r) =exp kT (3)
t

and

4(r) = V(r) + y(r) (4)

where V(r) is the simple pair potential and ur<r>is a perturbation term
which takes into account the effect of many-body interactions. For very
dilute systems, with particles undergoing Brownian motion, the distri-
bution will be random, and only occasional contacts will occur between
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 103

.
FIXED SITE
"CRYSTAL"
g(r) -

CONCENTRATX

'MBILE"

"LIQUID LIKE"

t r
g(r) a exp(-@/1(T)

r DILUTE

"VAPOUR-LIKE"

Fig. 2. Radial distribution function for (a) “dilute”; (b) “concentrated”; (c) “solid”
suspensions [ 1,2].

the particles, i.e. there will in general be pairwise interaction. Under


these circumstances G(r) can be replaced by V(r). In this case, g(r)
increases rapidly from the value of zero at r = 2R to its maximum value
of unity beyond the first shell. This is illustrated in Fig. 2a. Thus, with
such dilute systems (‘tapour-like”) no structure develops. In contrast,
“solid” suspensions (Fig. 2c) show distinct, sharp peaks similar to those
observed with atomic and molecular crystals. In between these two
extremes, is the case for concentrated dispersions (“liquid-like”) where
g(r) shows the form represented in Fig. 2b. This consists of a pronounced
first peak followed by a number of oscillatory peaks damping to unity
beyond four or five particle diameters. Moreover, as one proceeds out-
wards from the first shell, the peaks become broader.
104 Th.F. TadmslAdu. Colloid Interface Sci. 68 (1996) 97-200

2.0
.** (a)
.. -9
. ’
:. ...

Fig. 3. g(r) versus F for polystyrene latex dispersions in lo4 mol dm3 NaCl at various
volume fractions: solid line, $ = 0.01; dashed line, (I = 0.04; dotted line, $ = 0.13.

,2-

l- . .. . . . . . . . . . . . . .
m

L I I I ??

400 800 1200 1600

Fig. 4. g(r) versus F for sterically stabilized polytmethyl methacrylate) dispersions in


dodecane at various volume fractions: solid line, 9 = 0.23; dashed line, $ = 0.28,
dashed/dotted line, (I = 0.36; dotted line, $ = 0.42.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 105

The dependence of g(r) on r for colloidal dispersions can be deter-


mined by scattering techniques. This is illustrated in Fig. 3 for polysty-
rene latex particles (R = 16 nm) at constant electrolyte (NaCl) concen-
tration of lOA mol dm3 and at various volume fractions [6,8].
The most dilute dispersion (Q = 0.01) shows a curve of g(r) versus r
similar to that shown in Fig. 2a. At $ = 0.04, the initial portion of the
curve has become much steeper as the particles are packed close to each
other, and a clear maximum occurs followed by an oscillatory curve. At
$ = 0.13, the amplitude of the first maximum has increased, indicating
more structure. Also the particles have moved even closer and are
interacting more strongly.
Similar results were obtained for sterically stabilized dispersions as
illustrated in Fig. 4. In this figuregcr) is plotted versus r for dispersions
of poly(methy1 methacrylate) in dodecane stabilized with poly-12-hy-
droxystearic acid 171.Again by increasing the volume fraction of the
dispersion, the amplitude of the first peak increases and it becomes
shifted to lower r values, indicating closer approach of the particles and
some compression of the stabilizing layer.

3. Interparticle interactions and their combinations

3.1. Interparticle interactions

Four different types of interparticle interactions may be identified


and these are schematically illustrated in Fig. 5. The first type of
interaction (Fig. 5a) is the so-called “hard-sphere”, which is also some-
times referred to as neutral stability systems, i.e. whereby both repul-
sive and attractive forces are screened. In this case, the particles are
considered to consist of hard spheres with a radius Rhs that is only
slightly larger than the actual radius R. When the particles approach
each other to centre-to-centre separation distances that are slightly
smaller than the hard sphere diameter (i.e. 2&J, strong repulsion
occurs and the energy of interaction rises very steeply to plus infinity.
In this case the dispersion changes from “fluid-like” to “solid-like”
behaviour as soon as the volume fraction reaches the maximum hard-
sphere volume fraction, which for random packing is in the region of
0.64. As we will see later, the viscosity of the suspension rises very
sharply when one approaches the maximum packing.
The second type of interaction is that between electrostatically stabi-
lized suspensions with extended double layr;*s 191,i.e. at low electrolyte
106 Th.F. TadroslAdv. Colloid Interface Sci. 66 (1996) 97-200

;p :
;:
‘I

d
:
,

1
(
‘n, , I
- RetI

b)

c
CAPTURE
OISTINCE

(4

Fig. 5. Types of interparticle interactions in suspensions: (a) %ard-sphere”; (b) ((soft”


(electrostatic); (c) steric; (d) van der Waals.

concentration. The energy-distance curve for such systems is schemati-


cally represented in Fig. 5b. When two particles each bearing a double
layer of the same sign approach each other to a surface-to-surface
Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 107

distance h where the double layers begin to interfere, neither of these


layers can develop completely. Therefore, work must be done on the
system, i.e. the interaction of double layers gives rise to a repulsion
between the particles. In this case, the energy-distance curve shows an
exponential decay of the form shown in Fig. 5b. The extent of the decay
depends on the double layer thickness, (UK), which depends on electro-
lyte concentration and valency as given by the following expression,

(5)

where E, is the relative permittivity of the dispersion medium, E, is the


permittivity of free space, K is the Boltzmann constant, T is the absolute
temperature, 2 is the valency of the ions, e is the electronic charge, NA
is the Avogadro’s constant and C is the electrolyte concentration in mol
dmM3.
For 1:l electrolytes, cl/~) is equal to 100, 10 and 1 nm in lo”, 10”
and 10-l mol dmm3respectively. This shows that the repulsive interac-
tion decays more slowly with increase of interparticle distance as the
electrolyte concentration is decreased. In other words, the interaction is
“softer” the lower the electrolyte concentration.
For two spherical particles of radius R and surface potential vo, the
double layer repulsion V, is given by the following expression (lo), when
lcR < 3,

v =4~%E,~2tie~p(--1Ch) (6)
R
2R+h

From Eq. (6), it is clear that VR depends on the exponential term


which depends strongly on K and hence on electrolyte concentration C.
At low C, VR operates in a long-range leading to a gradual decrease of
VR with increase of h. This means that VR persists to separation
distances comparable to (UK), which for 1:l electrolytes is of the order
of 100 nm. As h decreases VR increases and a point is reached at which
VR shows a rapid increase with further reduction of h. This is illustrated
in Fig. 5b by the vertical line. The distance h at which this rapid rise
occurs denotes the effective radius of the particle, R,, (i.e. R + the
contribution from the double layer). The value of R,, can be orders of
magnitude larger than R, depending on particle radius and electrolyte
108 Th.F. Tadros /Adv. Colloid Interface Sci. 68 (1996) 97-200

concentration. For example, for 10 nm particles in 10” 1:l electrolyte,


R,, is - 110 nm. In this case the effective volume fraction can be several
orders of magnitude higher than the actualvolume fraction. In the above
example &.&j is of the order of 103. In this case the maximum effective
packing fraction of the particles is reached at relatively low actual
volume fractions. Assuming this packing fraction is 0.74 (hexagonal
maximum packing), the actual volume fraction under these conditions
is 10e3 lower, i.e. of the order of 7.4 x 10”. This illustrates the low
particle number concentration at which a “structure” is produced, as
evidenced by the iridescence colours produced by many latex suspen-
sions at low electrolyte concentration. These colours are manifestation
of the long-range interaction between the particles. As we will see later,
these systems show viscoelastic behaviour as a result of the strong
double layer interaction.
The above electrostatically stabilized suspensions undergo signifi-
cant changes on increasing the electrolyte concentration. This results
from double layer compression which leads to the change from long-
range to short-range interaction. For example, if the electrolyte concen-
tration in the above example is increased to 10” mol dma, the double
layer thickness is reduced from 100 to 10 nm. This will result in the shift
of the point at which rapid rise in repulsion occurs to lower h values. In
other words the suspension changes from “soft” to a “hard” sphere
system, This behaviour is reflected in the viscoelasticity of the system
as will be discussed later.
The third type of interaction is that when the particles possess
adsorbed or grafted layers of surfactants and/or macromolecules. In this
case, particle interaction occurs as soon as the adsorbed or grafted layers
begin to interact 1111.This is illustrated in Fig. 5c. It is convenient to
divide the interaction into two contributions. The first is the mixing or
osmotic interaction, which arises from the unfavourable mixing of the
adsorbed or grafted layers, when these are in good solvent conditions.
This is schematically illustrated in Fig. 6 for a very simple case of two
spheres surrounded by an adsorbed or grafted layer of thickness 6. The
assumption is also made in this case that the layer has uniform segment
density and the volume fraction of the surfactant or polymer in the layer
is $z. The chemical potential of the solvent in the layer before overlap is
pi. If the two particles are forced to approach to a surface-to-surface
distance h that is less than 26, then some overlap of the layers will occur
in a volume element dV, whereby the chemical potential of the solvent
is now l.~zthat is lower than 11-i.This reduction in the chemical potential
Th.F, TadroslAdu. Colloid Interface Sci. 66 (1996) 97-200 109

Fig. 6. Schematic representation of overlap of adsorbed or grafted layers.

of the solvent in the overlap region is equivalent to an increase in the


osmotic pressure (as a result of the increase in segment density) in the
overlap region. As a result solvent will diffuse from the bulk to the
overlap region thus separating the particles, i.e. leading to repulsion.
The free energy of interaction as a result of this unfavourable mixing,
Gtix, can be calculated using the Flory-Krigbaum theory [12] for the
free energy of mixing of two polymer solutions. This free energy, 6(Gti,,),
is given by the following expression,

(7)
where Q2 is the volume fraction of surfactant or polymer in the layer,
6nr and 6n2 are the number of solvent and surfactant or polymer
molecules in the layer and x is the chain-solvent interaction parameter,
usually referred to as the Flory-Huggins interaction parameter [13]. x
is related to the second virial coefficient B, which is a measure of the
non-ideality of mixing of polymer solutions.
The total change in free energy of mixing for the whole interaction
zone is obtained by summing over all the volume elements in dV. For
flat plates this leads to the following expression,
110 Th.F. TadroslAdu. Colloid Interface Sci. 68 (1996) 97-200

2kTc
%,ix= v (8)
1

where V, and V, are the molar volumes of the polymer (or surfactant)
and solvent respectively, v2 is the number of chains per unit area and
R,,(h) is a geometric function which depends on the form of the segment
density distribution of the chain normal to the surface, p(z).
For two spherical particles, the following expression can be derived,
assuming a uniform segment density in the adsorbed or grafted layer
114,151,

Clearly Gti,, can be positive or negative (i.e. repulsing or attractive)


depending on the magnitude of x. When the chains are in good solvent
conditions, i.e. x < 0.5, then GA,, is positive, i.e. repulsion occurs as a
result of polymer or surfactant overlap. In other words mixing of the
chains in this case is unfavourable. In contrast, when x > 0.5, i.e. the
chains are in poor solvent conditions (worse than e-solvent), Gti,, is
negative and mixing of the chains is favourable, i.e. the mixing interac-
tion in this case is attractive. The condition 2 = 0.5 is referred to as the
theta condition, which determines the onset of attraction [ll]. Many
authors have found good correlation between the onset of attraction
(flocculation) and the O-point. For example many suspensions showed
flocculation by addition of a non-solvent to the chains at the composition
whereby the chains reached their e-condition. With many aqueous
suspensions stabilized with chains based on poly(ethylene oxide), floc-
culation occurred at a particular temperature close to the e-temperature
of the chains.
The second term in the steric interaction is the volume restriction or
elastic contribution which arises from the loss in configurational entropy
of the chains on the approach of a second particle. This is schematically
illustrated in Fig. 7, following Mackor and van der Waals [16,17].
Considering the chain to be represented by a rigid rod connected to
the surface by one point, then at infinite separation distance between
the particles (considered as flat plates) the rod sweeps a volume repre-
sented by the hemisphere shown in Fig. 7. When a second surface
approaches to a distance h, the hemisphere loses part of its volume and
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 111

Volume avaiable for the


1////////u chain is reduced (restricted).
4
??
8’‘-------% h =co
: ‘; - mh
///////////

At h=ao No. of At h No. of


configurations = G configurations = n

Fig, 7. Schematic representation of the loss in configurational entropy as a result of


volume restriction.

this results in the chain sweeping a smaller volume than that before
close approach. This volume restriction leads to a reduction in the
configurational entropy of the chain. The effect is referred to as volume
restriction or elastic interaction, G,,. For two flat plates the elastic
interaction is given by the following expression [l 11,

G,, = 2 K!“v2 In RO = 2 kT v2 R,,(h) (10)


W-)

Q(h) and Q(m) are the number of configurations available to the chain
at h = h and h = = respectively; R,,(h) is a geometric function whose form
depends on p(z). G,,, is always a repulsive contribution.
It is clear from the above discussion that steric interaction leads to
steep repulsion as soon as the surface-to-surface distance becomes
smaller than twice the adsorbed or grafted layers. This is particularly
the case when the chains are in good solvent conditions, as illustrated
in Fig. 5c. If the chains are short and densely packed, the interaction
may be represented by a hard-sphere type with R,, = R + 6. This is the
case, for example, with nonaqueous suspensions stabilized by surfactant
molecule [ 161.In this case the interaction is felt at relatively high volume
fractions since the effective volume fraction is close to the actual (core)
volume fractions. However, with many sterically stabilized dispersions
the adsorbed or grafted layer extends to a considerable distance from
the particle surface (tens of nms). This implies that the effective radius
is significantly higher than the core radius of the particles. The effective
volume fraction, &,fl,becomes considerably larger than the core volume
fraction, +, since
112 Th.F. TadroslAdv. Colloid Interface Sci. 66 (1996) 97-200

(11)

In this case, the interaction becomes of long-range nature occurring at


large interparticle distances as soon as the long dangling tails of the
chains begin to overlap. Under these conditions, steric interaction
becomes significant at relatively low core volume fractions. For example,
if&-0.5R,the&- 3 4. Thus, a suspension with a core volume fraction
of 0.2 has an effective volume fraction of - 0.6, which is close to the
maximum packing fraction. As we will see later, this leads to significant
viscoelasticity of the system 1171.
The fourth type of interaction is the van der Waals, which is universal
with all disperse systems. Generally speaking, the van der Waals
attraction between atoms or molecules consists of three main types,
namely dipol+dipole (Keesom), dipole-induced dipole (Debye) and dis-
persion (London) interaction [91.The most important contribution is the
dispersion interaction which occurs between polar and non-polar atoms
or molecules. It arises from the charge fluctuation within an atom or
molecule associated with the motion of its electrons. Although the
London force between two atoms or molecules is short-range (being
inversely proportional to the seventh power of the distance between
atoms or molecules), the net force between two particles is of much
longer range and its magnitude depends on the particle radius, its
nature as determined by the Hamaker constant and the interparticle
distance [ 181.For example, for two spherical particles, the van der Waals
interaction, V,, is given by [Ml,

1
x2+ 2x
&=_A
L-.2-
12 x2+2X+X2+2.%+1
1
+ In
x2+2x+1
(12)

where x = (r - 2R)/2R (r is the centre-to-centre distance of separation


between the particles) and A is the composite Hamaker constant that is
given by the expression,

(13)

where A,, is the Hamaker constant of the particles and A,, that of the
medium and Ai2 is the Hamaker constant between particle and medium.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 113

The Hamaker constant of any material i is related to the London


dispersion constant pii and the number of atoms or molecules per unit
volume, 4i, by the expression,

(14)

pii is related to the polarizability of the atoms or molecules and it has a


value in the range 10-78-10-76 J m6.
At short distances between the particles, i.e. when r - 2R << 2R, Eq.
(13) reduces to,

AR AR
VA-- (15)
12(r-R) =-- 12h

Equation (15) shows that V, changes inversely with h and it increases


rapidly at small values of h. The interaction, therefore, shows a capture
distance significantly larger than the particle radius; this is illustrated
in Fig. 5d. The contribution of the van der Waals interaction to the total
energy of interaction depends to a large extent on the nature of the
particles and medium, as determined by their Hamaker constants, the
separation distance between the particles and the magnitude of the
repulsive force at this separation. This will be shown below when the
total interaction between the particles is plotted versus distance of
separation between the particles, in which the various contributions are
taken into account depending on the stabilization mechanism. These
energy-distance curves form the basis of the theories of colloid stability
and they also can be used to account for the viscoelastic properties of
the suspensions.

3.2. Combination of interparticle interactions

The universal van der Waals attraction may be combined with the
various repulsive interactions leading to three general energy-distance
curves. These are illustrated in Fig. 8.
Figure 8a represents the case for combination of electrostatic repul-
sion and van der Waals attraction. This forms the basis of the well
known theory of stability of lyophobic colloids derived independently by
Derjaguin and Landau and Verwey and Overbeek [lo], abbreviated as
Electrostatic Steric Electrostatic + sterlc
(Ionic Surfactants) (Polymers) (Polyelectrolytes)

Totar po~cmal energy

V V \

Pnmar-v
+ maawnum + +

0 0 0 !L
I
_

!a ib 1 iC

Fig. 8. Total energydistance curves for three stabilization mechanisms.


Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 115

the DLVO theory. The form of the energy-distance curve shown in Fig.
8a results from the way in which both repulsion and attraction change
with distance of separation. The repulsion has the feature of an expo-
nential function with a range of the order of the thickness of the double
layer. The attraction, however, decreases as the inverse power of the
distance. For very small distances, it goes to very large negative values
and at very close contact, VA changes sign due to the Born repulsion,
giving rise to a deep primary minimum in the potential energy-distance
curve (Fig. 8a). At intermediate distances, the repulsion normally pre-
dominates (at low electrolyte concentrations), giving rise to an energy
maximum or energy barrier. At large distances of separation, VA domi-
nates V, since the latter decays much faster then V, with h at large
distances. This is due to the fact that an exponential decreases much
faster than a power law at large distances. The result is the appearance
of a secondary minimum in the energy-distance curve as shown in Fig.
8a. The depth of the secondary minimum depends on the size (and
shape) of the particles and the Hamaker constant. With large and
asymmetric particles, V,,, may reach several tens of KT units (12is the
Boltzmann constant and T is the absolute temperature; KT is simply the
thermal energy of the particle). Thus, the shape and magnitude of the
energy-distance curves of charge stabilized dispersions depend on four
main parameters, namely the surface (or zeta) potential, electrolyte
concentrations, particle size and Hamaker constant. The electrolyte
concentration, in particular, has a pronounced effect on the height of
the energy maximum and the depth of the secondary minimum. With
increasing electrolyte concentration, V,.,.,,, decreases and V,,, increases.
A critical electrolyte concentration is reached where the energy maxi-
mum is reduced to zero and this indicates the onset of rapid flocculation.
It should be mentioned that the DLVO theory has been developed for
dilute systems, whereby the total interaction is simply considered to be
represented by a pair potential, namely the sum of V, and V,. Applica-
tion of the DLVO theory to concentrated dispersions requires some
modification to take into account the pair-wise and multibody interac-
tions. To a first approximation, the total energy of any one particle may
be taken as the sum of the pair-wise interactions over all nearest
neighbours , i.e.,

(16)
116 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

where iV is the number of nearest neighbours which interact with a


potential energy (VTIi.
The combination of steric repulsion and van der Waals attraction is
represented in Fig. 8b. The total interaction energy is given by the sum
of G, (which is formed from the two contributions Gtix and G,,) and GA.
At large distances of separation, GA predominates, but at separation
distances just smaller than twice the adsorbed layer thickness (i.e.
where the layers begin to overlap), a very steep repulsion occurs with
any further reduction in h. This prevents any close approach between
the particles to distances smaller than 26. At such distances (usually of
the order of tens nms, depending on the adsorbed layer thickness), the
van der Waals attraction is usually small. Thus, unlike the DLVO type
of interaction, the energy-distance curve in Fig. 7b shows only one
minimum, whose depth Gti,, is only a few M’ units, depending on the
adsorbed layer thickness, the particle radius and the van der Waals
attraction. The depth of this minimum, everything else is equal, becomes
smaller the larger the value of 6. This is illustrated in Fig. 9 whereby
the energy-distance curves for flat plates calculated using the Hesselink
et al. theory [19] is shown as a function of adsorbed layer thickness [201.
These calculations were made using poly(viny1 alcohol) (PVA) polymers

Fig. 9. Energy of interaction-distance curves for polystyrene latex with adsorbed layers
of PVA with various molecular weights.
Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 117

with various molecular weights and polystyrene latex as the substrate.


With the highest molecular weight polymer (M = 43000) with an ad-
sorbed layer thickness of 19.7 nm [21,221 Gti, is very small and it does
not appear in Fig. 9. As the molecular weight of the polymer is decreased
(M = 28000 and 17000) the adsorbed layer thickness decreases (to 14.0
and 9.8 nm respectively) and Gti, increases. At sufficiently low molecu-
lar weight (M = 8000) with a small 6 of 3.3 nm, Gti,, becomes significant.
Under these conditions, weak flocculation can occur. This was confirmed
using freeze fracture and scanning electron microscopy.
The free energy of interaction of particles having an adsorbed or
grafted polymer (or surfactant) layer depends on the following parame-
ters: the number of adsorbed or grafted chains per unit area v, the
average number of segments per loop or tail, which determines the
adsorbed layer thickness 6, the quality of the solvent which determines
the chain-solvent (Flory-Huggins) interaction parameter x, the mode of
attachment of the chain, the particle size and Hamaker constant [19].
The third type of energy distance curve is shown in Fig. 8c. This
represents the case where a nonionic polymer (or surfactant) is adsorbed
on a charged surface (i.e. where there is a contribution from double layer
repulsion) or where the stabilization is produced by adsorbed or grafted
polyelectrolytes. In this case, the total interaction is the sum of G,, G,
and GA. The energy-distance curve shows a shallow minimum at large
separation distances, an ill-defined maximum at intermediate distances
(characteristic of the DLVO type interaction) and a steep rise at shorter
distances (characteristic of the steric-type interaction). In other words,
the energy-distance curve combines the features of Fig. 8a and 8b.

3.3. Flocculation of suspensions

The flocculation of suspensions is the process leading to aggregation


of particles to give three dimensional clusters without coalescence
occurring, i.e. the particles retain their identities in the clusters (or
floes). For electrostatically stabilized suspensions, this occurs when the
energy maximum (Fig. 8a) is small (say < 5 K7’)or absent. This situation
arises on the addition of electrolytes which leads to the compression of
the double layer, thus reducing V, to a point where VA exceeds V, at all
distances of separation. In the absence of any energy barrier, fast
flocculation occurs and the process is diffusion controlled. This situation,
which is sometimes referred to as “sticky collisions”, may be represented
by a deep energy well, whereby any two particles diffusing in a suspension
118 Th.F. TadroslAdv. Colloid InteTface Ski. 68 (1996) 97-200

stick together as a result of this strong energy of attraction. This process


of fast flocculation was considered by von Smoluchowski [23] who
assumed that no interaction between two colliding particles occurs until
they come into contact, whereupon they adhere irreversibly. For such a
system initially containing n, particles, the number concentration, n, at
time t may be obtained if one assumes the process of flocculation to be
represented by a second order kinetics. Thus, n and no may be related
by the simple expression,

n0
(17)
n=l+k,n,t

where k, is the rate constant (fast flocculation rate) that is related to


the diffusion coefficient of the particles, D, by the expression,
k,=hDR (18)
The diffusion coefficient is given by the Stokes-Einstein equation, i.e.,

(19)
D=6i;R

where q is the viscosity of the medium (- 1 mPas for water at 25°C).


Thus, for particles dispersed in water at 25”C,

k
0
-_8kT
- =5.5 x lo-l8 m3 s-l (20)
67-1
In the present of an energy barrier, with height G,,, slow flocculation
occurs with a rate k depending on the height of this barrier. This
situation was analysed by Fuchs 1241,who related k to k, by the following
equation,

k+ Cm

where W is defined as the stability ratio that is related to G,, by the


expression (241,

G
W=2RJexp 5 h-2 dh (22)
2R ( 1
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 119

An approximate form of Eq. (22) has been given by Reerink and Over-
beek [25],

(23)

These authors also derived the following theoretical relationship be-


tween W and electrolyte concentration C. For aqueous suspensions at
25°C

log W = constant - 2.06 x 10’ -RYz log c (24


22

where,

[exp (Zev, /2kT) - 11


‘= [exp (Zev,, /2kZ’) + 11
cm

Equation (24) predicts that experimental plots of log W versus log C


should be linear in the slow flocculation regime. However, log W = 0 (W
= 1) in the fast flocculation regime. This is illustrated in Fig. 10 for 1:l
and 2:2 electrolytes. The plots show two linear portions intersecting at
a critical electrolyte concentration at which W = 1. This is defined as the
critical flocculation concentration (CFC) of the particular electrolyte.
Note that in Fig. 10, W is less than 1 in the fast flocculation regime. This
is the consequence of the contribution of the van der Waals attraction
which should be added to Eq. (22). It can be seen from Fig. 10 that the
CFC for 1:l electrolyte is - 10-l mol dm3, whereas that for a 2:2
electrolyte is - 5 x lo3 mol dm3. This illustrates the dependence of the
CFC on electrolyte valency as described by the Schulze-Hardy rule [9].
It can be seen from Fig. 8a that flocculation of suspensions may occur
as a result of attraction at longer separation distances, i.e. resulting
from the presence of the secondary minimum in the energy-distance
curve. This flocculation is, however, weak and reversible. In this case
one must take into account the backward rate of deflocculation. In this
case the rate of decrease of particle number with time is given by the
following expression,

&E-k n2+k n
dt f b
120 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

log c
Fig. 10. Log W versus log C for 1:l and 2:2 electrolytes.

where 12,and k, are the rate constants for the forward (flocculation) and
backward (deflocculation) processes. The rate constant of deflocculation
may well depend on floe size and the exact way in which the floes break
down, i.e. how many “contacts” are broken. This would mean that the
second term on the right hand side of Eq. (26) would need to be replaced
by a summation over all possible modes of breakdown, making the
analysis of the kinetics complicated.
Another complication in the analysis of the kinetics of reversible
flocculation is that the process is a critical phenomenon rather than a
chain (or sequential) process. Thus, a critical particle number concen-
tration, nwit, has to be exceeded before flocculation occurs. Above ncrit
flocculation becomes a thermodynamically favoured process. Thus, the
kinetics of weak, reversible flocculation has more in common with
nucleation kinetics than with chemical (e.g. polymerization) kinetics.
This does not mean that doublets, triplets, etc. will not form transiently
below ncrit.These units, if produced, are thermodynamically unstable,
but their effective concentrations may be calculated from a suitable
kinetic analysis.
The above diffusion controlled flocculation is referred to as periki-
netic in which the assumption is made that particle collisions arise solely
from Brownian diffusion of the particles, i.e. without the application of
any external field. The diffusion coefficient of the particles is given by
the Stokes-Einstein equation, i.e. D = kT/f, where f is the friction
coefficient. If an external field is applied on the system, such as shear,
ultrasonic or centrifugal, or the system is not at thermal equilibrium
(resulting in convection currents), the rate of particle collisions is
Th.F. Tadros lAdv. Colloid Interface Sci. 68 (1996) 97400 121

modified (usually increased) and the flocculation in this case is referred


to as orthokinetic. For example, in a shear field the rate of flocculation
is related to the shear rate, ‘y,by the expression,

_J!!L=.E&R~
(27)
dt 3
where a is the collision frequency, i.e. the fraction of collisions which
result in permanent aggregates. For an irreversibly coagulating system,
i.e. with a deep primary minimum, orthokinetic conditions lead to an
increased rate of flocculation at any given time interval. This is due to
the fact that application of shear enhances the collisions between the
particles, thus increasing the rate of flocculation. In contrast, with many
weakly flocculated systems, application of shear results in a decrease in
the rate of flocculation. In this case, the shear forces tend to break the
weakly aggregated structure and hence the second term on the right
hand side of equation [27] increases. In some cases, the shear force may
be sufficient to increase the deflocculation rate and a stable system may
result under shear.
The flocculation of sterically stabilized suspensions (produced by
adsorption of nonionic surfactants or polymers) may occur as a result of
a number of factors. Firstly, if the chains are not strongly adsorbed
(anchored) to the surface, then with time and during particle collisions,
some desorption may take place. The creation of bare patches on the
surface of particles can cause flocculation as a result of van der Waals
attraction between the bare particles, or bridging. In the latter case, the
chains from one particle may adsorb on the bare patches of another
particle. Secondly, flocculation may occur as a result of reduction of
solvency of the medium for the chains. As discussed above, if the solvent
becomes worse than &solvent for the chains, i.e. x > 0.5, the mixing
interaction, Gtii,, becomes negative, i.e. it becomes attractive. Unless
there is sufficient repulsion from the elastic contribution, the dispersion
flocculates when x > 0.5. This is illustrated in Fig. 11, which shows the
interaction free energy-separation curves for a sterically stabilized
suspension in good (x c 0.5, better than O-solvent) and poor (x > 0.5,
worse than Cl-solvent) solvents. When x c 0.5, both G,, and GvR are
positive and when these are added to the van der Waals attraction, the
energydistance curve shows a shallow minimum. Under these conditions,
flocculation is thermodynamically unfavourable. On the other hand, when
x > 0.5, Ghi, becomes negative and the total energy-distance curve
shows a deep minimum (Fig. 11). Under these conditions, flocculation
122 Th.F. Tadros IAdu. Colloid Interface Sci. 68 (1996) 97-200

Gel

reduce
solvency \

ai

Fig. 11. Interaction free energy-separation curves for sterically stabilized suspensions
in good (x < 0.5) and poor (x > 0.5) solvent conditions.

becomes thermodynamically favourable. The flocculation of sterically


stabilized suspensions, induced by reduction of solvency is sometimes
referred to as incipient flocculation. It usually occurs just beyond the
(j-condition for the chains. Napper 1111demonstrated close correlation
between the critical flocculation point and the O-point for the chains. For
example, correlations were obtained between the critical volume frac-
tion of an added non-solvent that produces Cl-conditionsfor the chains and
the volume of the non-solvent at which flocculation of the suspension takes
place. Napper 1111also showed good correlation between the critical
flocculation temperature of a suspension and the e-temperature for the
chains. However, such correlations are not always produced, particu-
larly with chains having multipoint anchoring to the surface [ll] .
As discussed above, weak flocculation of sterically stabilized suspen-
sions may occur, even if the chains are in good solvent conditions. This
arises when the depth of the minimum, Gti in the energydistance curve
(Fig. 8b) reaches a critical value. The value of Gti,, at which flocculation
occurs depends on the volume fraction of the suspension. This could be
easily understood from a consideration of the total free energy of
flocculation which consists of an energy and entropy terms, i.e.,
Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 123

The value of @ is negative and it is determined by Gti,. However,


on flocculation the suspension loses entropy as a result of aggregation
and hence ASflocis negative. This means that the second term on the
right hand side of Eq. (28) is positive on flocculation. For a dilute
suspension, the reduction in entropy is significant and hence a large
value of Gti,, is required for flocculation to occur. In the limit of infinite
dilution, the value of Gti,, may not be sufficient for any flocculation to
occur. For a concentrated suspension, on the other hand, the reduction
in entropy on flocculation is not significant, since the particles are close
to each other before flocculation. In this case, a small value for Gti,, is
sufficient for flocculation to occur. Thus, for any given volume fraction
of a suspension there is a critical value for Gti, above which flocculation
occurs. The higher the volume fraction of the suspension, the smaller
the critical value of Gti,, required for weak flocculation to occur.
Two other flocculation mechanisms may be described, namely bridg-
ing and/or charge neutralization by polymers and polyelectrolytes and
flocculation induced by “free” (non-adsorbing) polymer referred to as
depletion flocculation. The process of flocculation by polymer bridging
and charge neutralization by polyelectrolytes was discussed by Gregory
[26]. Essentially, bridging flocculation occurs because segments of a
polymer chain adsorb on different particles, thus linking the particles
together. Adsorption is an essential step, and this requires some favour-
able interaction between polymer segments and particle surface.
Ruehrwein and Ward [27] pointed out that typical polymeric flocculants
could have molecular dimensions comparable to the size of colloidal
particles (0.1-l pm> and that attachment of a polymer chain to several
particles could explain the effectiveness of these materials as floccu-
lants. The floes produced by polymer bridging are considerably stronger
than aggregates formed by coagulation of suspensions by electrolytes.
With polyelectrolytes, the picture is more complicated since the dimen-
sions of polyelectrolyte chains depend on ionic strength. At low ionic
strength screening of the charges on the chain is limited and the
polyelectrolyte adopts an extended configuration, which makes bridging
more likely. At high ionic strength, on the other hand, screening of the
charge produces a more compact configuration, thus reducing the
chance of bridging. However, increasing the electrolyte concentration
also results in reduction in the range of interparticle repulsion and
hence the compact chains could still bridge the particles. In addition, at
124 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

high electrolyte concentrations, adsorption of polyelectrolytes with the


same charge sign on the particle surface is enhanced and this also
increases the likelihood of bridging.
In many practical applications, it has been found that polyelectro-
lytes with a charge opposite to that of the particles are very effective in
flocculation of suspensions. With oppositely charged polyelectrolytes, it
is likely that adsorption gives a rather flat configuration of the adsorbed
chain, as a result of the strong electrostatic attraction between the ionic
groups on the polyelectrolyte chain and the charge sites on the surface.
This would probably reduce the possibility of bridging. However, the
adsorption of say a cationic polyelectrolyte on a negatively charged
surface would reduce the surface charge of the particles, and this charge
neutralization could be an important factor in the cause of flocculation.
However, this simple mechanism of flocculation cannot explain other
effects of the molecular weight and ionic strength. An alternative
mechanism was proposed by Gregory 1281,namely the “electrostatic-
patch” model. This applies to cases where the particles have a fairly low
density of immobile charges and the adsorbing polyelectrolyte has a
fairly high charge density. Under these conditions, it is not physically
possible for each surface to be individually neutralized by a charged
segment on the polymer chain, even though the particle may have
sufficient polyelectrolyte to achieve overall neutrality. There are then
“patches” of excess positive charge, corresponding to adsorbed polyelec-
trolyte chains (probably in a rather flat configuration), surrounded by
areas of negative charge, representing the original particle surface.
Particles having this “patchy or “mosaic” type of surface charge distri-
bution may interact in such a way that positive and negative “patches”
come into contact, giving quite strong attachment (although not as
strong as polymer bridging). A schematic representation of this type of
interaction was given by Gregory [26] as shown in Fig. 12.
The flocculation of suspensions by the addition of “free” (non-adsorb-
ing) polymers has been the subject of many investigations in recent
years. Above a critical concentration or volume fraction of free polymer,
4); 7 weak flocculation of the suspension occurs. The mechanism of
flocculation is as follows. If the added polymer does not adsorb on the
particle surface, the polymer coil cannot approach the surface closely
since the loss in entropy of the chain as it becomes very close to the surface
is not compensated by an adsorption energy term. As a result the polymer
coil can only approach the surface to a distance A, beyond which there
will be no polymer segments. This region is defined as the depletion zone,
_-_-
Th.F. TadroslAdv.

m ,+

+
Colloid Interface Sci. 68 (1996) 97-200

+
---__
---Be

m-m-_
e--m

--

----
-_
125

w___ . --__I

Fig. 12. “Electrostatic patch” model for the interaction of negatively charged particles
with adsorbed cationic polyelectrolyte.

i.e. the distance that is void of any polymer segments. At relatively low
free polymer concentrations, the depletion thickness is equal to the
radius of gyration of the free polymer coil, R,. At higher polymer
concentrations, A decreases with increase in free polymer concentration
(in which case A is better equated to the correlation length of the polymer
chain). When two particles with their depletion zones of thickness A
approach to a distance h c 2A, the two depletion layers overlap and the
free polymer coils are “squeezed out” from in between the particles. This
is illustrated in Fig. 13 for two sterically stabilized particles to which a
free polymer is added. As a result, the osmotic pressure of the solution
outside the particles will be larger than that in between the particles
whereby a polymer-free zone (pure solvent) is formed. This results in
attraction between the particles. This phenomenon is usually described
as depletion flocculation. The magnitude of the depletion interaction is
of the order of the osmotic pressure of the polymer solution, whereas its
range is of the order of A or R,. As we will see later, the critical volume
fraction, @z , at which depletion flocculation starts depends on the
molecular weight of the free polymer. The higher the molecular weight
the lower the value of (I; . Other parameters that may affect +G are the
126 Th. F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

T
be’ow oc

of cx
ot=oc
/..... ---*-.....*_* oc

ococ
+
above 9
P

w
w
osmotic
pressure

Fig. 13. Schematic representation of depletion flocculation.

volume fraction of the suspension and the particle size. As we will see
later, the flocculation produced by the addition of free polymer is usually
weak and reversible.
Several theories have been developed to quantify the flocculation
produced by the addition of free (non-adsorbing) polymer. Asakura and
Oosawa [29,30] introduced a simple model whereby the particles and
polymer coil were considered to be represented by hard-spheres. The
exclusion of the polymer coils from the particle interstices produces an
osmotic pressure which gives a measure of the depletion free energy of
attraction, Gdep. In other words, Gdepwas simply equated with -poS,
where p. is the osmotic pressure of the polymer solution and S is the
area of the depletion zone. Asakura and Oosawa derived the following
expression for Gdep(which is valid for the case where the particle radius
is much larger than the polymer coil radius),
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 127

O<XCl

where 42 is the volume concentration of the free polymer that is equal


to (4xA3Nz/3u) with A being the depletion layer thickness (that is equal
to the radius of gyration of free polymer, R,), N2 is the total number of
polymer molecules in a volume u of solution. p = u/A, where a is the
particle radius and z = [A- (h/2)1/A,where h is the distance of separation
between the outer surfaces of the particles. Clearly, when h = 0, i.e. at
the point where the polymer coils are “squeezed out” from the region
between the particles, x = 1.
The above theory has been reconsidered by Vrij [31] and by Sperry
[32] who derived similar expressions. Later, Fleer, Scheutjens and Fleer
(FSV model) [33] developed a general approach of interaction of hard-
spheres in the presence of a free polymer. The model was based on the
original approach of Scheutjens and Fleer [34,35] for polymer adsorption
at the solid/liquid interface and subsequent interaction between sur-
faces containing polymer layers. The FSV model takes into account the
dependence of range of interaction on free polymer concentration and
any contribution from the non-ideal mixing of polymer solutions. This
theory gives the following expression for Gdep,

(30)

where p1 is the chemical potential at bulk polymer concentration 3,


py is the corresponding value in the absence of free polymer and uy 1s
the molecular volume of the solvent. The difference in chemical potential
(cl1- ~7) can be calculated from the volume fraction of the free polymer
and the polymer solvent interaction (Flory-Huggins) parameter [13],

(31)

where n2 is the number of polymer segments per chain.


A useful model, that takes into account the softness of the interaction
between grafted polymer chains on the particles and the free polymer
coil, was introduced by Vincent, Edwards, Emmett and Jones [36]
(VEEJ model). The distancep the solution polymer coil interpenetrates
the adsorbed layer of thickness, 6, was calculated and this led to the
128 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

following expression for Gdep,for the case p < 6,

(32)

The key problem is to calculate p, which can be estimated from the


adsorbed polymer profiles on the particle surfaces. Vincent et al. [36]
used four different profiles for the grafted polymer layer, namely uni-
form, linear, pseudo-homopolymer, and pseudo-tails. Only the latter
profile fitted the segment volume fraction profiles and this was used for
calculation of Gdep,giving the following expression at h = 26,

Gdep= 27cu

It should be mentioned that the above theories on depletion floccula-


tion were developed for the case of dilute dispersions. As we will see
later, the volume fraction of the suspension plays a role in determining
the onset of flocculation. Indeed, Gast et al. [371 developed a statistical
mechanical approach to predict the phase separation of colloidal disper-
sions by the addition of a non-adsorbing polymer. These theories take
into account both the volume fraction of the suspension and that of the
free polymer. They also predict the trend obtained when the ratio of
particle radius to polymer coil radius is changed. As we will see later
such ratio determines the dependence of the critical free polymer volume
fraction at the onset of depletion on the volume fraction of the suspension.

4. Principles of rheological measurements

For full evaluation of the rheology of concentrated dispersions one


needs to carry out experiments within well defined time scales. Two
main types of experiments may be carried out. In the first type, referred
to as transient measurements, a constant stress or strain is applied
within a well defined short period and the relaxation of strain or stress
respectively is followed as a function of time. In the second type, referred
to as dynamic measurements, a stress or strain is applied within a well
characterized frequency regime (usually sinusoidal) and then resulting
strain (or stress) is compared with the stress (or strain) respectively.
Although the two types of measurements are equivalent, they provide
different information.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 129

The above-mentioned experiments are usually referred to as low


deformation since measurements can be made before the “structure” of
the suspension is broken down. Another low deformation experiment is
that of measuring the velocity of a generated wave in the suspension
(shear wave measurements). These low deformation experiments pro-
vide valuable information of the “structure” of the suspension at any
interparticle interaction. However, experiments can also be carried
under conditions whereby the “structure”of the suspension is deformed or
broken down during the measurement. This is the case, for example,
whereby the suspension is subjected to continuous shear while measuring
the stress in the sample. Such measurements are sometimes referred to
as steady state (high deformation) measurements. Below a summary of
the basic principles of the above-mentioned measurements is given.

4.1. Transient (static) measurements

As mentioned above one can either measure the relaxation of stress


after sudden application of strain or the strain relaxation after sudden
application of stress (creep measurements). In stress relaxation a sud-
den shear strain y is applied on the system within a very short shear
period, i.e. keeping the rate of strain y constant [38]. The stress decays
exponentially with time as illustrated in Fig. 13 for a system with a
single relaxation time that follows a simple Maxwell model.
As shown in Fig. 14, a single Maxwell model is represented by a
spring-dash pot element in series. The ratio of the stress at any time t
to the constant strain applied is the stress relaxation modulus G(t). As
the stress in the sample relaxes by viscous flow, the modulus decreases.
The instantaneous value of the stress at the moment when the strain is
applied is z. and the corresponding modulus is Go which is simply the
spring constant of the elastic element. The stress relaxes exponentially
with time and reaches a zero value at t = 00. The system is referred to as
a viscoelastic liquid. In this case, the rate of decrease of stress with time
follows an equation similar to first order kinetics [39], i.e.,

(34)

where K is the characteristic rate constant for the relaxation process,


i.e. lz = l/t,, where t, is the relaxation time that is given by the ratio of
the viscosity to the modulus. Equation (34) can be written in the form,
130 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

hl t
Fig. 14. Schematic representation of stress relaxation for a single Maxwell model.

z(t) = z, exp (-t/t,) (35)

G(t) = Go exp (-t/Q (36)

The above equations assume a system with one relaxation time, such as
the case with aqueous monodisperse suspensions with no other addi-
tives. In most practical systems, the particles are not monodisperse and
other ingredients such as micelles or polymers may be present in the
continuous medium. This implies that the various particles and the
components will have different relaxation times. In this case, a gener-
alized Maxwell model with elements having different relaxation times
must be used. In other words, the system is represented by a group of
Maxwell elements in parallel. This is equivalent to a discrete spectrum
of relaxation times, each time t, being associated with a spectral
strength Gi. A summation over all elements may be used, i.e.,
n

G(t) = C Gi exp (-t/t,) (37)


i=l

where n is the number of elements.


As mentioned above, for a viscoelastic liquid the modulus becomes
zero at infinite time. However, for a viscoelastic solid, the modulus
reaches a finite equilibrium value at t = m and in this case one of the
relaxation times is infinite and the equilibrium modulus is G,. For a
continuous spectrum, the summation in Eq. (37) may be replaced by an
integral, i.e.,
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 131

Fig. 15. Schematic representation for creep compliance for a system with a single Kelvin
element.

G(t) = G, + [H exp (-t/t,) d In t,

where H is defined as the continuous relaxation spectrum, i.e. the


product of the modulus of the part of the model with relaxation In t, with
the fractional weighting of that part.
The second type of transient measurements is the constant stress or
creep. In this case, a small stress z is applied on the system and the
strain or compliance J (deformation per applied stress applied, i.e. ylz)
is followed as a function of time. This is illustrated in Fig. 15 for a system
represented by a single Kelvin element, i.e. a spring and dash pot in
parallel. At time t, the stress is suddenly removed and the deformation,
which now reverses sign, is measured over a longer period of time. At t
= 0, there will be a rapid elastic deformation (i.e. no energy is dissipated
within such very short time period), characterized by an instantaneous
compliance Jo that is inversely proportional to the instantaneous modu-
lus Go (Go = z/J,). At t > 0, J increases less rapidly with time and the
system exhibits a retarded elastic response. This can be understood from
a consideration of the Kelvin element. The dashpot simply retards the
motion of the spring and the Kelvin model is characterized by a retar-
dation time tk that is equal to n/G. For a Kelvin model, where the stresses
of the two elements must be added, the creep compliance is the compli-
ance of the spring corrected for by the retarded strain due to the dashpot,
132 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

11 - exp (-t/t,>]
J(t) =
G

The form of the creep curve depends on whether the material behaves
as a viscoelastic liquid or solid. This is particularly important after
sudden removal of stress. The rate of deformation will change sign and
the system will return more or less towards its normal state. This
reversing of deformation is called creep recovery. With a viscoelastic
solid, the system reaches an equilibrium deformation in creep, charac-
terised by an equilibrium compliance J,. Attainment of equilibrium in
this case corresponds to a Maxwell model in which the dashpot is
missing. For a viscoelastic liquid, on the other hand, such as a dispersion
where the particles are not permanently attached to each other, the
system does not reach an equilibrium compliance. Under constant
stress, the strain rate approaches a limiting value and a situation of
steady flow is eventually reached, governed by a Newtonian viscosity q0
which may be visualised as the sum of all viscosities in the generalised
Maxwell model. In their steady state, the springs of the Maxwell
elements are stretched to equilibrium extensions representing elastic
energy storage which is recoverable after the stress is removed.
Similar to the analysis shown above, a group of Voigt elements in
series represents a discrete spectrum of retardation times, each time t,
being associated with a spectral compliance with magnitude Ji. Thus,
for n elements,

J(t) = i Ji [I- exp (-t / tk)] (40)


i=l

For a viscoelastic liquid, one spring has zero rigidity and therefore a
term t/q, must be added to Eq. (40). If the number of elements in the
Voigt model is made infinite, a continuous spectrum of retardation
times, L, may be used in a similar fashion to the relaxation spectrum,
i.e.,

(41)

In Eq. (411, the instantaneous compliance jz (= l/G,) is added to allow


the possibility of a discrete contribution with tk = 0. As mentioned above
the term t/q, is added to allow the possibility of an element with zero
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 133

rigidity. In other words, the generalized Kelvin model corresponds to a


system with one element with a dashpot of zero viscosity and another
element with a zero-modulus spring.
4.2. Dynamic (oscillatory) measurements

In these experiments, a small amplitude sinusoidal strain (or stress)


with frequency v (Hz) or w (rad sM1),where o = 27cv,is applied to the
system and the stress and strain compared simultaneously [391. This is
schematically illustrated for a viscoelastic system in Fig. 16, which
shows the response of stress to a sinusoidal strain of amplitude ‘yo.The
amplitude of the stress is z, and it oscillates with the same frequency,
but out of phase. The phase angle shift 6 is given by the product of the
time shift At between the strain and stress sine waves and the frequency
o. For a perfectly elastic system, the maximum stress occurs when the
strain is a maximum. In this case 6 = 0. For a perfectly viscous liquid,
the maximum stress occurs at the maximum strain rate, and the two
waves are 90” out of phase.
Dynamic measurements are usually expressed in terms of a complex
modulus, 1G* 1 that is given by,

IG*I $ (42)
0

The complex modulus is vectorially resolved into two components, the


elastic or storage modulus G’(the real part of the complex modulus) and
the viscous or loss modulus G” (the imaginary part of the complex
modulus), i.e.,

1G* 1 = (G2 + G”2)1’2

G’and G” can be separately calculated from 1G* I and the phase angle
shift, i.e.,

Fig. 16. Stress-strain response for a viscoelastic system.


134 Th.Z? TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

G’= IG*I cos6 (44)

G”= IG*I sin6 (45)

and

IGI =G’+iG” (46)

where i is equal to (-1>1’2.


G’ is a measure of the energy stored elastically during a cycle,
whereas G” is a measure of the energy dissipated as viscous flow. It is
convenient to define the ratio of G” to G’,

tan 3=g (47)

The dynamic viscosity I-( is given by the equation,

q’=G” (48)
w

In dynamic measurements, one initially measures the variation of G*,


G’ and G” with strain amplitude, at a fixed frequency, to obtain the
linear viscoelastic region where the rheological parameters are inde-
pendent of the applied strain amplitude. Once this linear region is
established, measurements are then made at a fixed strain amplitude
(within the linear region) as a function of frequency.
To illustrate how the storage and loss moduli vary with frequency,
let us consider a simple case for the response of a Maxwell model with
a shear modulus G and viscosity n. G’and G” are given by the following
expressions,

G’ = wG2 (49)
1 + (%I2
G%
G”= (50)
1 + (W2
t, is the relaxation time that is equal to q/G. The response of the model
is shown in Fig. 17, where the reduced moduli are plotted as a function
of logarithm of the reduced frequency [391. It can be seen that at low
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97400 135

Fig. 17. Variation ofGIG, G”lG and q’lq with reduced kquency for a single Maxwell model.

reduced frequency G”/G is higher than G/G. However, G”lG increases


with increase of frequency passes through a maximum and at suffi-
ciently high frequency it falls to zero. G/G increases with frequency and
at a characteristic frequency value it becomes higher than G”/G and
eventually it reaches a plateau value to be denoted as G(m). At this point
the system behaves as an elastic solid and the spring constant in the
Maxwell model G = G(W). The cross-over point at which G’/G becomes
equal to G”/G (the storage and loss moduli are now equal to l/2 G(w)
and 6 = 45”) is ot, = 1, i.e. the relaxation time of thesystem is equal to
l/o*, where o* is the characteristic frequency. Finally, the curve of the
reduced dynamic viscosity q’/rl is also plotted in Fig. 17. It can be seen
that a limiting value of the dynamic viscosity is reached at low fre-
quency, so that the zero shear viscosity q(o) = q, the dashpot viscosity.
The above analysis is for a system with a single relaxation time,
which is clearly not the case with most practical systems. In the latter
case a broader response is obtained, indicating that there is more than
one relaxation time appropriate to the system. In this case the right
hand terms of Eqs. (49) and (50) should be replaced by a sum over all i
units. Alternatively, for an infinite number of elements the summation
could be replaced by an integral, i.e.,

G=l+ H(oti>2 d ln t

(51)
1 f (oti)2 ’
136 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

where H is the relaxation spectrum that is equal to the product of the


modulus with relaxation time In ti with the fractional weighting of that
part.
Two limiting cases may be considered. In the high frequency limit,
oti >> 1 and Eq. (51) reduces to,
+=

G(m)=1 Hdlnt (53)


-00

In the low frequency limit, on the other hand, oti KC 1 and Eq. (52)
reduces to,
+-
q(-)=j Htdlnt (54)
-ca

The above discussion illustrates the importance of dynamic measure-


ments in predicting the behaviour of the suspension. If measurements
can be made over a wide range of frequencies, one can obtain the
relaxation spectrum of the sample. Using the above equations it is
possible to obtain important rheological parameters such as the high
frequency modulus and the limiting (zero shear) viscosity, both of which
can be used to predict the properties of the suspension such as its
flocculation and sedimentation. As we will see later, by measuring the
viscoelastic parameters of the system as a function of frequency and
volume fraction of the suspension, one can obtain information on the
interaction forces between the particles.

4.3. Shear wave propagation

This experimental technique allows one to obtain the high frequency


modulus G(m) using a simple set-up. Basically the suspension is placed
in a cell that is fitted with two parallel metal or perspex plates, whose
distance of separation d can be changed using a micrometer attached to
one of the plates. Each plate is connected to a piezoelectric crystal (LiCl,
a transducer). An electric generator is connected to the bottom crystal
and is used to initiate a shear wave by a small amplitude torsional strain
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 137

(<< 10” rad). The time taken for the shear wave to travel a known
distance is measured using the piezoelectric crystal at the top plate,
which is connected to an oscilloscope or microcomputer. By measuring
the time t, for a shear wave maximum to go to a minimum, at a given
distance d between the plates, one can measure the shear wave velocity.
Let us consider the propagation of a shear wave through a block of
Hookian elastic material 1391.The wave moves through the material in
the x-direction with a velocity u = dWdt. The shear displacement occurs
orthogonally in they-direction with a velocity dy/dt. The resulting shear
strain y is then given by the following expression,
do dt
(55)
‘=dtii

If F is the force on the element of area A and density p, then the impulse
on an element which is equal to the momentum is given by,

Fdt=Adxpz

Since the force per unit area is the shear stress, then the shear rigidity
modulus, G, is given by the equation,
G=pv2 (57)
For a viscoelastic material, the modulus is complex with a storage and
loss components. For the estimation of these components one needs to
determine both the wavelength of the shear wave and attenuation
distance r, where r is equal to U~~LX,,
where X, is the distance travelled
by the wave during which its amplitude has fallen l/e of its initial value.
When r << 0.1, the shear rigidity modulus approaches G’ and this is
denoted as the high frequency modulus G,.

4.4. Steady-state measurements

Shear stress z - shear rate y measurements can be carried out using


rotational viscometers. Cone and plate or concentric-cylinder geometries
are the most convenient as the shear rate is close to a constant value. The
suspension is placed in the gap between the cone and plate or the concen-
tric cylinder and the shear rate is gradually increased by rotating the
cone (or the plate) or the outer (or inner) cylinder, while measuring the
torque on the other element (which allows one to calculate the stress).
138 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

Fig. 18. Flow curve for a pseudoplastic system.

The most common flow curve obtained with concentrated “structured”


or flocculated suspensions is shown in Fig. 18, usually described as
pseudoplastic flow curve. This curve is characterized by three main
parameters: the critical shear rate ycy,, above which the curve becomes
linear, the slope of this linear line qpl and the extrapolated yield stress
zp. Pseudoplastic flow may be analysed using three main models: the
power law model, i.e.,

2=171r- (58)
with n < 1, the Bingham model 1401,

T=Tp+rlplY (59)

and the Casson model [411,

$2 = ,$I2
c + qY2 r1’2 (f-30)

It should be mentioned that the flow curve in Fig. 17 represents the


case where there is no time dependence of stress versus strain. However,
many systems show time dependence, sometimes referred to as thixo-
tropy. In this case if the shear rate is gradually increased to an upper
limit and then gradually decreased to zero, the downward curve does
not coincide with the upward curve, i.e. the flow curve shows hysteresis.
This time dependent shear stress-strain relationship occurs with many
flocculated systems, since on shearing the sample the floe structure is
partially or completely broken down and when the shear rate is reduced
time is required to build up the original floe structure.
Th.F. Tadros lAdv. Colloid Interface Sci. 68 (1996) 97-200 139

5. Viscoelastic properties of concentrated suspensions

The viscoelastic properties of suspensions are determined by the


balance of three main forces: Brownian diffusion, hydrodynamic inter-
action and interparticle forces. The latter are the double layer repulsion,
the van der Waals attraction and steric interaction. The range of
interaction is determined by the volume fraction, @, and the particle size
(radius R) and shape distribution. The rheology of concentrated suspen-
sions is complex and various responses may be obtained depending on
the time scale of the experiment and the structure of the system (which
determines its relaxation time). In this respect, it is useful to use the
ratio of the relaxation time, t,, to the experimental time, t,, as a means
of classification of the various responses. This dimensionless group is
referred to as the Deborah number, D,, i.e.,

De=; (61)
e

If 0, >> 1, one obtains elastic deformation and the response is described


as “solid-like” behaviour. This, for example, is the case when measure-
ments are made at high frequency (t, is small) for a system with high
relaxation time. When 0, CC 1, one obtains viscous deformation and this
is referred to as “fluid-like” behaviour. This occurs at low frequencies (t,
is large) for systems with low relaxation times. Many colloidal systems
show a viscoelastic response with a D, in the region of 1, whereby the
experimental time scale of the measurement is comparable to the
relaxation time of the system. Thus, by varying the time scale of the
experiment (such as frequency in dynamic measurements) one can
obtain various responses and the results may be correlated with the
various interaction forces in the system. Alternatively, one can use a
fmed time scale regime (a fmed frequency range) and vary some parame-
ters of the system such as volume fraction, particle size and surface
forces to obtain information on the interaction in the system. This will
be illustrated below.
Four different systems may be distinguished: hard-sphere suspen-
sions, electrostatically stabilized suspensions (soft interactions), steri-
tally stabilized dispersions and flocculated (and coagulated) systems.
These systems increase in the order of the complexity of their rheology,
with the hard-sphere systems being the most simple and the flocculated
or coagulated systems being the most complicated both experimentally
140 Th.F. TadrosIAdv. Colloid Interface Sci. 68 (1996) 97-200

and theoretically. Progress on the rheology of concentrated suspensions


has been slow and only in recent years some advances have been made.
This is due to the development of modern rheological techniques. How-
ever, theoretical analysis of the rheological data is far from being
quantitative and in most cases the theories are based on many approxi-
mations. In spite of these limitations, viscoelastic measurements pro-
vide valuable information on the interactions in concentrated suspen-
sions and in some cases it is possible to obtain the magnitude of the
forces involved.

5.1. Suspensions with hard-sphere interactions

These are sometimes referred to as systems with neutral stability in


which both repulsion and attraction are screened. In other words, all
interactions are weak and the main forces responsible for flow are
hydrodynamic interaction and Brownian diffusion. The appropriate
dimensionless group is y x t,, where y is the shear rate and t, is the time
scale of Brownian diffusion that is given by the following expression,

6 nqoa3
t, = (62)
kT

The reduced shear rate, yred (= y&) is then given by the following
expression,

(63)

Model dispersions of hard spheres have been developed by Krieger and


his collaborators [42,431 who used monodisperse polystyrene latex dis-
persions with well characterised surface and particle radius. To mini-
mise repulsion, the double layer was compressed either by addition of
electrolyte or by replacement of water with a less polar medium such as
benzyl alcohol. Under these conditions, both attraction and repulsion
are minimized and the dispersion behaves as a hard-sphere system. To
check this, it is essential to plot the relative viscosity q, versus the
reduced shear rate (or reduced shear stress) at a particular volume
fraction, with several particle radii and fluid viscosities to see if the
relationship obeys a rheological equation of state of the form 142,431,
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 141

kT
Fig. 19. Reduced viscosity versus reduced shear rate for polystyrene latex dispersions
(I$= 0.4) with various particle radii.

As an illustration, Fig. 19 shows some results at a volume fraction of 0.4


for particles with various radii. All viscosity data for the various particle
size latices fall on the same line indicating that the systems behave as
hard-sphere suspensions.
The qr-Yredcurve shows two Newtonian regions at low and high shear
rate with a shear thinning region at intermediate values. This behaviour
can be clearly understood from a consideration of the balance between
the Brownian diffusion and hydrodynamic interaction. In the low shear
rate regime, the Brownian diffusion predominates over any hydrody-
namic interaction imposed by the shear flow. The suspensions show a
high viscosity, n(o), due to the random arrangement of particles (in some
form of a three dimensional random order). As the shear rate is in-
creased beyond a certain limit, the particles arrange themselves in
layers coincident with the plane of shear and the viscosity decreases
with increase of applied shear rate (shear thinning). In the high shear
rate regime, the layers can move freely past each other giving a low
Newtonian viscosity q(m). In this region the hydrodynamic interaction
prevails over Brownian diffusion.
142 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

If the above experiment is repeated for suspensions with various volume


fractions, a set of sigmoid curves will be produced with two Newtonian
regions. The values of r\(o) and n(-> decrease with decrease of +, as
expected. If q(o) or n(-1 is plotted versus $, a general behaviour is obtained
as illustrated in Fig. 19 whereby the high shear relative viscosity for
polystyrene latex suspensions is plotted versus 4. The relative viscosity
increases gradually with increase of +, but above a certain C)value it
shows a rapid increase with further increase of @reaching an asymptote
when @ reaches about 0.6. The slope of the curve in the limit + + 0, gives
the intrinsic viscosity, which for hard-spheres is equal to 2.5, whereas
the value of (I at the asymptote represents the maximum packing
fraction, $,, for the hard-sphere suspension (which is of the order of 0.64
for random packing). The relative viscosity-volume fraction curve
shown in Fig. 20 could be represented by the following equation [44,45],

(65)

Equation (65) is usually referred to as the Dougherty-Krieger equation,


A theory for the rheology of hard-sphere dispersions has been developed
by Batchelor [46] who considered the balance between hydrodynamic
interaction and Brownian diffusion, i.e.,

Fig. 20. Typical plot of qr versus +.


Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 143

I I I I I I
0.1 02 0.3 0.4 0.5
#
Fig. 21. Comparison of experimental qr versus $ results with theoretical calculations
(dashed line) based on Batchelor’s theory.

rlr = 1 + 2.5 Q + 6.2 o2 + 0 @3 (66)

The first two terms on the right hand side of Eq. (66) represent the
Einstein’s limit, whereas the third term (6.2 $2) is the contribution from
hydrodynamic interaction; the third term in $3 represents higher order
interactions. A comparison between the experimental r~/rl~results and
those calculated using Eq. (66) is shown in Fig. 21, which shows that
Bachelor’s theory is valid for 41c 0.2. Extension of the above theory to
more concentrated suspensions requires the introduction of higher order
interactions. No theory is, as yet, available that treats such a complex
problem. Computer simulation of the multibody interaction may offer a
starting point for predicting the viscosity of concentrated suspensions.

5.2. Stable systems with soft (electrostatic) interactions

These are systems with expanded double layers, i.e. at low electrolyte
concentrations, whereby the interaction is dominated by double layer
repulsion. The appropriate dimensionless group characterizing the proc-
ess (balance of viscous and repulsive force) is given by qoa2@$, where
v0 is the surface (or zeta) potential. The viscoelastic properties of these
systems can be investigated using constant stress (creep) or oscillatory
measurements. Figure 22 shows the creep compliance of electrostatically
144 Th.F. Tadros fAdv. Colloid Interface Sci. 68 (1996) 97-200

time/s

time/s

t & (cl
0.002 -
z

1 O.OOlL
D
2
o- - .

I I I I I II, 1
0 120 240 360

Fig. 22. Creep curves for polystyrene latex suspensions (a = 34.5 nm, C,,, = 5 x 1OA
mol dm3) at three volume fractions: (a) $ = 0.138; (b) (I= 0.142; (c) $ = 0.177.
Th.F. Tadros IAdu. Colloid Interface Sci. 68 (1996) 97-200 145

stabilized polystyrene latex dispersions (a = 34.3 nm) at three volume


fractions and constant NaCl concentration of 5 x lOA mol dm3 [47]. At
the lowest $ value (0.138), the viscous compliance is large compared with
the elastic component and the dispersion shows near Newtonian behav-
iour (Fig. 22a) at shear stresses < 0.3 Pa, with a limiting viscosity of 2.9
Pas. In contrast, at the highest volume fraction (0.177), the elastic
component dominates the viscous one and essentially an instantaneous
elastic response is observed (Fig. 22~). From a plot of instantaneous
strain versus shear stress a modulus of 590 Pa and a zero shear viscosity
of lo6 Pas are obtained. This clearly demonstrates the dramatic increase
in modulus and zero shear viscosity with a small increase in volume
fraction of the suspension. This increase in + from 0.138 to 0.177 results
in a change from “fluid-like” to “solid-like” response. This is the result
of the double layer overlap that occurs at the high volume fraction. Over
a narrow volume fraction range (0.14 c @ c 0.18) the viscous and elastic
components can be resolved as shown in Fig. 22b. A good fit to the creep
curve can be obtained using the equation,

J(t)=&[2-exp[-:)I+: (67)

where l/G, and t/n,, are the instantaneous elastic and viscous compli-
antes respectively. The term between the brackets is the contribution
from the retarded elastic compliance which in the present case can be
represented by a single relaxation time, t, = qJG, = 4.04 s.
The viscoelastic properties of electrostatically stabilized suspensions
depend on three parameters: volume fraction of the suspension, Q,
particle radius, a, and electrolyte concentration, C. As shown above,
increase in 9, at a given a and C, causes a change from “fluid-like”
behaviour with non-detectable elastic properties to a “solid-like” (“gel”)
behaviour with a high zero shear viscosity and an appreciable modulus
Go. On the other hand, increasing a at given C leads to an increase in +
above which significant viscoelasticity is obtained. This is illustrated in
Fig. 23, which shows the variation of G, with @ at four a values (26,34,
39 and 98 nm) and constant electrolyte concentration (lo3 mol dm3
NaCl). It can be seen that with increasing a, larger (I values must be
reached before a significant G,, is reached. Indeed as a is increased
significantly (> 200 nm), viscoelasticity is obtained when approaches the
close-packed limit. This is clear since the thickness of the double layer
is now relatively small compared to the particle radius. Alternatively,
146 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

Fig. 23. G,--$ curves for four a values: 0,26 nm; a,34 nm; Cl,39 nm; W, 98 nm. CNaCl=
10e3 mol dme3.

the double layer thickness can be decreased by increasing the electrolyte


concentration, while keeping + and a constant. In this case, viscoelas-
ticity is reached at higher $Ivalues.
As mentioned above, viscoelasticity of dispersions can also be inves-
tigated using oscillatory measurements [48,49]. As an illustration, Fig.
24 shows the variation of G*, G’and G” with frequency for monodisperse
polystyrene latex suspensions with a radius of 700 nm at 10” mol dm3
NaCl and at two volume fractions; $ = 0.475 and 0 = 0.524. Such
measurements were carried out at small strains (y,,= 0.0041, i.e. in the
linear viscoelastic region. In both cases, the suspensions show more
elastic than viscous response (G’ > G”), within the frequency range
studied. This is not surprising since at such low electrolyte concentration,
the double layer is extended (the double layer thickness UK = 100 nm)
Th.F. TadroslAdu. Colloid Interface Sci. 68 (1996) 97-200 147

162 104 1 5
w/Hz

162 10’1 1 5
WI Hz
Fig. 24. Variation of G’, G’and G” with frequency for polystyrene latex suspensions (a
= 700 nm) at 10” mol am3 NaCl and two 41values; 0, G’; 0, G’; A, G”.

and overlap between such layers is possible at such high volume frac-
tions. It should be noted that as $ is increased from 0.475 to 0.525, the
moduli values increase by about one order of magnitude. In addition,
with increase in @ G’becomes much larger than G” and it approaches
G* very closely. This reflects the strong double layer interaction at the
high volume fraction as the surface-to-surface distance becomes smaller
than twice the double layer thickness. This point will be discussed below.
As mentioned above, increasing electrolyte concentration results in
compression of the double layers and the surface-to-surface distance
becomes smaller than twice the double layer thickness at relatively
148 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

4=0-S 6
9
c, ,_r~ lo- mol dme3

I I I

10” 1 5
w/Hz

I
10 10 1 5
w/Hz
Fig. 25. Variation of G’, G’and G” with frequency for polystyrene latex suspensions (a
= 700 nm) at lo3 mol dm3 NaCl and two Q,values; 0, G’; 0, G’; A, G”.

higher $ values when compared to the results obtained at lower electrolyte


concentrations. This is illustrated in Fig. 25 for the same latex suspensions
but in the presence of 103 mol dm3 NaCl. The results show that G”’> G’
even at a @value of 0.566 (which is higher than that in the presence of 10-5
mol dm3 NaCl). The moduli values are also lower than those obtained
at the lower electrolyte concentration. This behaviour reflects the much
weaker repulsion obtained at the higher electrolyte concentration. At
any given + the latex shows much less elastic response in lo3 mol dm3
NaCl than that obtained in the presence of lo3 mol dm3 NaCl. This is
Th.F. TadroslAdu. Colloid Interface Sci. 68 (1996) 97-200 149

due to the smaller double layer thickness (10 nm) at the higher electro-
lyte concentration.
A good comparison between the results for the two electrolyte con-
centrations may be obtained from plots of G*, G’ and G” (at one
frequency, namely 1 Hz) versus the volume fraction @. The latter is the
core volume fraction, i.e. excluding the contribution from the double
layer. The results are shown in Fig. 26, where a logarithmic scale is used
for the moduli values. All the results were obtained at low strain values
to be as close as possible to the linear viscoelastic region. In 10” mol
dmW3NaCl, the moduli values show a rapid increase with increase in I$
within the range studies, namely 0.460-0.524. In addition, G’is always
greater than G” within this volume fraction range. At the higher Cp value
(0.524) G’approaches G* very closely and the suspension behaves as a
near elastic solid. In contrast, the results in lo3 mol dm3 NaCl, show
a rapid increase in G’and G” when $ exceeds 0.53. In addition, within
the volume fraction range studied (0.3-0.566) G’is either close to G” or
even lower than it.

I I
11
I

03 0.L I$ 0.5 06

Fig. 26. Variation ofG’, G’and G” (at v = 1 Hz) with $ at two electrolyte concentrations;
0, G*; 0, G’; A, G”.
150 Th.F. TadroslAdv. Colloid Interface Sci. 66 (1996) 97-200

The above trends can be adequately explained if one considers the


presence of the double layer around the particles. To a first approxima-
tion, the double layer thickness, l/~, should be added to the particle
radius to obtain the effective radius, u,~ At any given size, u,,~depends
on the electrolyte concentration, C, since the double layer thickness is
determined by C. For example, in the above case a,R is - 800 nm in lo5
mol dm3 NaCl, whereas aeff is - 710 in 10” mol dm3 NaCl. Since the
volume fraction scales with the cube of the radius, it is clear that, at any
given $, the effective volume fraction of the suspension is much higher
at the lower electrolyte concentration. &.Ris related to @by the following
expression,

(68)

In 10e5 mol dm3 NaCl, $eR= 1.5 @. Thus, at the lowest (I value studied,
namely 0.46, (peE= 0.7, which is above the maximum packing fraction
(0.64 for random packing). Under these conditions, the double layer
interaction is significant and some overlap of the double layers may
occur. This explains why the response at this electrolyte concentration
and volume fraction is predominantly elastic. At the highest Q value
studied, namely 0.524, &,fl= 0.79, which is significantly higher than the
maximum packing fraction. This results in significant double layer
overlap and the dispersion behaves as a near elastic solid with G’- G*.
In lo3 mol dmm3NaCl, however, $efi = 1.05 9 and up to the maximum
volume fraction studied (0.5651, &.Bis well below the maximum packing
fraction. This implies that there is insignificant overlap of the double
layers and the dispersions show more viscous than elastic response. To
achieve significant elastic response one has to increase the volume
fraction to values above 0.6.
The above results may be presented in another form by plotting G
versus h, the surface-to-surface distance between the particles in the
dispersion, i.e.,

(69)

for the present calculations of h a value of $, = 0.68 was used. The plots
of G’versus h are shown in Fig. 27 for lo5 and lo3 mol dm3 NaCl. At
the lower electrolyte concentration G’decreases rapidly with decrease
Th.F. Tadros JAdv. Colloid Interface Sci. 68 (1996) 97-200 151

50
h Inm
Fig. 27. Plots ofG;, and G’,_, at two electrolyte concentrations: lOA and lo3 mol dms NaCl.

of h, when the latter is below 200 nm (twice the double layer thickness),
reaching very high values (> 1000 Pa) when h is below 100 nm. As
mentioned above, this is the result of double layer overlap. In lo3 mol dm3
NaCl, however, such high values of G’are only reached when h is lower
than 50 nm. As mentioned above the double layer thickness is of the order
of 10 nm and hence overlap begins to occur when h < 20 nm. It is clear
from Fig. 27 that at any given h, G’in lo4 mol dm3 NaCl is several orders
of magnitude higher than the corresponding value in lo3 mol dm4.
As discussed by Goodwin et al. 150,511,it is possible to relate the high
frequency modulus, G,,, to the total energy of interaction between the
particles, VT, i.e.,

(70)
152 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

where a = (3132) +P n (n being the coordination number), d is the


centre-to-centre distance of separation between the particles (d = 2a +
h). The total energy of interaction (mainly the double layer repulsion)
is given by the expression,

4n=0Vi

b= d
exp [--K(d- 2u)l (71)

where E is the permittivity of the medium, E, is the permittivity of free


space and vd is the Stern potential. By differentiating Eq. (71) twice,
Goodwin and coworkers 150,511 obtained the following expression for
the theoretical shear modulus, Glth,

Gth= 4~atxoa2~~ [*d2>trd+2]exp-x(d-2a) (72)

Values of Glthwere calculated for the suspensions in 10” mol dm3 NaCl
(KU< 10) since these were highly elastic and the modulus showed little
dependence on frequency. In these calculations, a was taken to be 0.833.
The results of the calculation are shown in Fig. 27. It can be seen that
the theoretical Gth values increase less rapidly with decrease of h than
the experimental G’,, values. It should be mentioned, however, that
the calculation of G’, using Eq. (72) is based on a number of assumptions
which may not be completely valid for the present system. In addition,
the latex used for this study was fairly large (a = 700 nm) and although
the measurements were carried out at low strain, one is not sure that
the results lie perfectly in the linear viscoelastic region.
Another useful way of describing interactions in concentrated disper-
sions is to apply scaling laws for G’ versus $ results [481. Generally
speaking, the storage modulus of a dispersion scales with the volume
fraction with an exponent, n, i.e.,

G’=k+” (73)

The power n can be obtained from log-log plots of G’versus @ These


plots are usually linear above a critical @ value. In 10” mol dm3 NaCl,
n was found to be 20, whereas in lo3 mol dm3 NaCl n was found to be
30. The lower power at the lower electrolyte concentration reflects the
softness of the interaction as a result of the expanded double layer. At
the higher electrolyte concentration, the double layers are significantly
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 153

compressed and the suspension behaves as a near hard-sphere system.


It is clear from the above discussion that the rheological properties
of concentrated electrostatically stabilized suspensions can be control-
led by controlling three main parameters, namely the volume fraction
of the suspension +, the particle radius a and the electrolyte concentra-
tion C. For example, to achieve high $ values, without producing much
elasticity in the system, one needs to increase a and add small amounts
of electrolyte (well below the flocculation value) to compress the double
layers such that &.. - +. Alternatively, if suspensions are required with
high elasticity (order) at low volume fractions, the particle radius of the
suspension should be kept as small as possible and the electrolyte
concentration as low as possible (to produce expanded double layers).
These principles are applied in practice to many industrial suspensions,
whereby the mechanism of stabilization is electrostatic in nature, e.g.
charged latices.

5.3. Sterically stabilized suspensions

These are suspensions where particle repulsion results from interac-


tion between adsorbed or grafted layers of nonionic surfactants or
polymers. The appropriate dimensionless group characterizing the flow
(balance of viscous and steric repulsive forces) is given by r@2yVJ(V2 -
x)62. As discussed before, steric interaction is repulsive as long as x <
0.5. With short chains, the interaction may be represented by a hard-
sphere type with an effective radius aeff = a + 6. This is particularly the
case with non-aqueous suspensions with an adsorbed layer that is small
compared to the particle radius, and where any electrostatic interaction
is negligible. The rheology of such suspensions approach the hard-
sphere behaviour described before. This has been demonstrated by
Willey and Macosko [521 for sterically stabilized poly(viny1 chloride)
suspensions in several solvents. Plots of n,. versus reduced shear rate,
at any given volume fraction, showed the same values for all particle
sizes. However, with most sterically suspensions the adsorbed or grafted
layer has an appreciable thickness (compared to the particle radius) and
hence the interaction is “soft” in nature as a result of the longer range
of interaction. This is illustrated in Fig. 28 for suspensions of poly(viny1
acetate) stabilized by poly(2-ethyl hexyl methacrylate) and dispersed in
Isopar G (an isoparaffinic oil with a narrow cut aliphatic hydrocarbon)
[53]. In Fig. 28, both G’ and G” are plotted as a function frequency at
three 4 values, namely 0.18, 0.43 and 0.53. At @ = 0.18, G” > G’ at all
154 Th.F. TadroslAdu. Colloid Interface Sci. 68 (1996) 97-200

)’

IO'

0
c 0
??
0
\
9 O ?? G”

0
0
8 ” IO0 IO'
I I
IO0

I I

IO0 IO' IO2


w/rod s-1

Fig. 28. G’ and G” versus w at three C$of poly(viny1 acetate) in Isopar G. (a) 4 = 0.18; (b)
$I= 043; Cc) I$= 0.53.

frequencies (within the range studied), indicating a predominantly


viscous response. In addition, the moduli values are very small and they
show a rapid increase with increase in frequency. This is the typical
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 155

“fluid-like” behaviour. At such low volume fractions the surface-to-sur-


face separation is larger than twice the adsorbed layer thickness and
only occasional contact occurs between the particles. In other words, the
polymer layers do not undergo any significant overlap and the particles
diffuse freely in the suspension, As the volume fraction of the suspension
is increased from 0.18 to 0.43, the probability of particle-particle contact
becomes larger and this results in significant elastic interaction. This
results in an increase in the moduli values and G’becomes comparable
to G”. Also, the moduli show less dependence on frequency. Ultimately
when the volume fraction reaches sufficiently high values, the elastic
contribution becomes much larger than the viscous contribution. This
is illustrated in Fig. 28~ for $ = 0.53. Under these conditions, the
surface-to-surface distance becomes comparable (or even smaller) than
twice the adsorbed layer thickness and this promotes interpenetration
and/or compression of the chains. The moduli values increase by orders
of magnitude and they show much smaller dependence on frequency.
Results on aqueous sterically stabilized suspensions were obtained
by Tadros and collaborators 154,551.Polystyrene latex suspensions with
grafted polyethylene oxide (PEO) chains (M= 2000) were prepared using
the Aquersymer process [561. Methoxy-polyethylene oxide was grafted
onto a polystyrene backbone to form a block copolymer, which nucleated
to for a near monodisperse latex of solid particles. The process used a
double free radical initiator, azodiisobutyronitrile and a terminator to
reduce excess monomer. The reaction was carried out in methanol-
water mixture.
As an illustration, Fig. 29 shows the variation of qr with Q for latex
dispersions with three different particle radii (77.5, 306 and 502). For
comparison, the rl,-$ curve calculated using the Dougherty-Krieger
equation (see above) is shown on the same figure. The q,-$ curves for
the latex suspensions are shifted to the left as a result of the presence
of the grafted PEO. The experimental relative viscosity data may be
used to obtain the grafted polymer layer thickness. Using the
Dougherty-Krieger equation, one can obtain the effective volume frac-
tion of the particles provided a value can be assigned to the maximum
packing fraction and the intrinsic viscosity. The maximum packing frac-
tion may be obtained from a plot of l/(?Qv2 versus Q,and extrapolation to
l/($S” = 0, using the empirical procedure described before 1541,i.e.,

(74)
156 Th.F. TadroslAdu. Colloid Interface Sci. 68 (1996) 97-200

1
700 I

A
II, latex R=306nm

c. latex !+502nm A

d. Dougherty-krieger
2
0

z
.Pl
2 400 -
0”
.?I
5 6
U
2 300 - 0

10
H
2 0 A

(I
e
200 - D A

D A
O0
D A

0.4 0.45 0.5 0.55 0.6 0.65 0.7


Volume fraction of latex dispersions qJ
Fig. 29. Relative viscosity versus volume fraction for latex suspensions with various
particle sizes.

where K is the slope, which is constant for values of qr > 20. The values
of $, using the above empirical equation were found to be 0.592,0,624
and 0.644 for the particles with radii of 77.5, 306 and 502 nm respec-
tively. These values are reasonable, being very close to the maximum
random packing fraction. The intrinsic viscosity [q] was assigned a value
of 2.5.
Using the above calculations, the grafted PEO layer thickness was
calculated as a function of the volume fraction for the three latex
suspensions. The values are summarized in Table 1.
The results show a gradual decrease of 6 with an increase in the
volume fraction of the suspension. The adsorbed layer thickness also
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 157

Table 1

Adsorbed layer thickness (6) as a function of Q for latex dispersions containing grafted
PEO (M = 2000) with different particle radii

a = 77.5 nm a = 306 nm a = 502 nm

$ 6 (?? 6 Q 6

0.422 8.07 0.508 12.0 0.535 21.0


0.462 6.96 0.526 11.4 0.558 21.0
0.482 6.56 0.555 12:2 0.567 21.0
0.507 5.82 0.568 11.9 0.572 19.7
0.521 5.46 0.577 11.4 0.585 18.5
0.533 5.26 0.585 10.8 0.585 18.7
0.528 5.35 0.590 10.7 0.591 17.8
0.533 5.26 0.597 10.1 0.597 16.5
0.542 5.01 0.601 10.2 0.606 14.7
0.543 4.96 0.603 10.1

increases with increase in particle size. The particle size was also found
to influence the yield behaviour of the suspensions. Figure 30 shows the
Bingham yield value as a function of latex volume fraction for various
particle sizes, which shows the point at which the flow behaviour of the
suspension changes from near Newtonian to pseudoplastic flow. The
appearance of a yield value is an indication that the interaction between
the particles is beginning to dominate the Brownian diffusion. For the
small particles, the pseudoplastic flow occurred at @ > 0.41, while for the
larger particles (a = 5021, the pseudoplastic flow took place at (I > 0.53.
The above results clearly demonstrate a significant effect of particle
size (at the same grafted layer thickness) on the rheology of sterically
stabilized suspensions. This is reflected in the change of 6/a with
increase in a. This may arise from two effects [57]. First, a simple
geometric effect; if the area occupied per chain is constant for the three
particle sizes, then one would expect 6 to increase with increase in a.
This effect has been demonstrated for poly (vinyl alcohol) physically
adsorbed on polystyrene latex particles [221. Second, if the grafted
polymer density varies on the latex particles, then one would expect that
6 would decrease as the area per chain increases. The latex suspensions
used in the rheological measurements did show a systematic increase
158 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

100 ,

latex R=77.%m 0

II
??

latex R=306nm
A
A
latex R=502nm
Cl

0
0
0

6.4 0.45 0.5 0.55 0.6 0.65


Volume fraction of latex dispersions @
Fig. 30. Bingham yield value versus volume fraction of latex suspensions with various
particle sizes.

in the area per chain as the particle radius decreased. Thus, it seems
likely that the increase in 6 with increase in a may be due to the
combined effect of geometry and PEO graft density.
The question that needs to be answered is whether such suspensions
behave as hard-sphere systems. The problem in this system is that 6
depends on both (I and a, and would expect that in the limit when Wa is
very small, the suspension approximates a hard-sphere one. Unfortu-
nately, one cannot decrease Wa too much, otherwise weak flocculation
may occur.
In order to check the applicability of the hard-sphere type interaction,
it is essential to plot the relative viscosity, q,, versus the reduced shear
stress, z,, by measuring over varying shear rates at a particular volume
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 159

fraction, with several particle radii and fluid viscosities, to see if the
relationship obeys a rheological equation of state [42,43,58] of the form,

rlr = f (4&J (75)

where z, = za3/kT is a dimensionless shear stress. The results for the


three suspensions studied (at a constant volume fraction of 0.535) are
shown in Fig. 31. It can be seen that the data for a = 306 and 502 nm
superpose, indicating that these two suspensions approximate hard-
sphere systems. In contrast, the results for the a = 77.5 nm suspension
shows significantly higher rlrvalues over the whole reduced shear stress
studied. This illustrates the softness of interaction. This is not surpris-
ing since in this case 6/a has approximately twice the value of the other
two suspensions.
The viscoelastic behaviour of sterically stabilized latex suspensions
is illustrated in Fig. 32 which shows the variation of G*, G’and G” with
frequency ci)in Hz for latex suspensions at various volume fractions.
These suspensions have an average radius of 175 nm and the grafted
polymer layer is PEO with a molecular weight of 2000, i.e. of the same
nature as described above. At $ = 0.44, G” >> G’and all the moduli values

R=73.5nm

0
a
ooo
\
--I R&km

R-5i2nm
0

5 10 20 50 100 200
Reduced shear stress T&T/k1

Fig. 31. Relative viscosity versus reduced shear stress for latex suspensions with three
different radii at $ = 0.535.
160 Th.F. TadroslAdu. Colloid Interface Sci. 68 (1996) 97-200

Fig. 32. Variation of G’, c’, G” with w (Hz) for polystyrene latex suspensions (a = 175
nm), containing grafted PEO chains, at various volume fractions.

are very low. In other words the suspension behaves as a near viscous
fluid. This reflects the relatively weak interaction between the particles
at such volume fraction, since the surface-to-surface separation between
the particles is smaller than twice the grafted polymer layer thickness.
When Cpis increased to 0.465, G” is still higher than G’, but the moduli
values increase by about a factor of 2 compared to the values at (I = 0.44.
As the volume fraction is increased, the steric interaction between the
particles increases, since the surface-to-surface distance between the
particles becomes smaller. The surface-to-surface distance between the
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 161

particles at Q = 0.465 is still larger than 26 and hence the dispersion


shows a predominantly viscous response within the frequency range
studied. When $Iis increased to 0.5, G’ becomes now larger than G”
within the frequency range studied and the moduli values are increased
by an order of magnitude compared to the values obtained at $ = 0.44.
At $ = 0.5, the interparticle distance (assuming random packing) is of
the order of 30 nm, which is now smaller than 26 and elastic interaction
between the grafted chains becomes strong. As the volume fraction is
further increased above 0.5, such elastic interaction becomes stronger
and stronger and ultimately the suspension behaves as a near elastic
solid. This is illustrated by the results at $ = 0.575, which shows that G
77 G” (and G’ becomes close to G) and the moduli show much less
dependence on frequency. The interparticle separation distance at such
volume fraction is of the order of 12 nm, which is significantly smaller
than 26. This results in interpenetration and/or compression of the
grafted PEO chains resulting in very strong elastic interaction.
The exact volume fraction at which a suspension changes from predomi-
nantly viscous to predominantly elastic response may be obtained from
plots of G, G’and G” (at a fixed frequency) versus the volume fraction of
the suspension. This is illustrated in Fig. 33 which shows the results for
the above latex suspensions. It can be seen from Fig. 33 that G’= G” at (I
= 0.482 (sometimes referred to as the cross-over point), which corresponds
to $Q = 0.62, i.e. close to maximum random packing. At $ > 0.482, G
becomes progressively larger than G” and ultimately the value of G
approaches G* and G” reaches a relatively small value. For example at
(J= 0.585, G’- G* = 4.8 x lo3 Pa, whereas at @ = 0.62, G’- G* = 1.6 x lo5
Pa. Such high elastic moduli indicate that the suspensions at such high
volume fractions behave as “solids” (gels) as a result of the interpene-
tration and/or compression of the grafted chains.
The above discussion clearly illustrates the value of viscoelastic
measurements in studying the steric interaction between suspensions
containing grafted or physically adsorbed polymer layers. By following
the variation of the moduli with volume fraction, one may obtain the
onset of strong steric interaction. In the above example, such situation
occurred at a volume fraction of the core particles of - 0.482. As we will
see later, this volume fraction depends both on the particle radius and
the thickness of the grafted or adsorbed polymer layer.
The influence of particle size on the viscoelastic behaviour of steri-
tally stabilized suspensions is illustrated in Fig. 34 which shows the
log-log plots of G’ and G” versus the volume fraction of the latex
162 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

1C
1 I I I I I I I
.!A.O.f.6 OL8 0.50 052 0% 056 O-58
4
Fig. 33. Variations of G*, G’and G” (at o = 1 Hz) with Cpfor latex suspensions (a = 175
nm) containing grafted PEO chains.

dispersions for three particle radii, namely 77.5, 306 and 502 nm. The
general trend is the same as shown in Fig. 33, but the critical volume
fraction, @,, at which G’= G” (i.e., the cross-over point) increases with
increase in particle size. The exact volume fraction at which the disper-
sion changes from viscous to elastic is shown in Fig. 34, where G”/G’ =
tan 6 is plotted versus the volume fraction. The cross-over points at tan
6 = 1 show an increase in (I,, from 0.48 to 0.52 to 0.56 as a increases from
77.5 to 306 to 502 nm. This shows the more elastic nature of the smallest
particle dispersions.
Th. F. Tadros IAdv. Colloid Interface 5%. 68 (1996) 97-200 163

Fig. 34. Plots of moduli (at 1 Hz) as a function of latex volume fraction for three particle
radii (R = 77.5,306 and 502 nm).

A scaling law can be used to fit the elastic part of the log G’-log @
curves (see Eq. (73)). This gives a value of n = 35 for the smallest latex
(CZ= 77.5) and n = 110 for both latices of a = 306 and 502 nm. This implies
that the compressibility of the small particles is higher than that of the
larger particles. The compressibility of the sterically stabilized particles
may be described by the ratio of the grafted layer thickness to the particle
radius, i.e. 6/a, which decreases with increase in the particle radius.
164 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97400

Fig. 35. Plots of G”/G’ as a function of volume fraction for three latex dispersions: 0, a
= 77.5 nm; A, a = 306 nm; 0, a = 502 nm.

Table 2 shows the grafted layer thickness at +efl= 0.658, the compressi-
bility and power-law exponent for the three particle radii studied. The
correlation between the compressibility and the power-law exponent is
clear. The small-particle size dispersions behave as soft sphere systems,
whereas the larger particles approach hard-sphere systems.
The correlation of the rheology of concentrated sterically stabilized
dispersions with interparticle interactions has been recently investi-
gated by Costello et al. [59-621. Basically, one measures the energy
E(D)-distance D curves for a graft copolymer consisting of poly(methy1
methacrylate) backbone with PEO side chains (with similar molecular
weight as that used in the latex) which is physically adsorbed on smooth
mica sheets. The apparatus used to measure the forces between surfaces
bearing adsorbed polymer has been described in detail by Israelachvili and
Adams 1631and by Luckham [64]. In principle it consists of measuring
Th. F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 165

Table 2
The relationship between compressibility and particle size of the dispersions

Particle radius Adsorbed layer thickness Compressibility Power law fit


(nm) (nm) at $ = 0.658 (6/a) (n)

77.5 4.96 0.0640 35


306 10.1 0.0330 110
502 16.5 0.0329 110

the forces between mica sheets with molecularly smooth surfaces posi-
tioned with a cross cylinder geometry. The mica sheets are partially
silvered on the reverse side so that light interferometry can be used to
determine the surface separation. The force between the surfaces is
simply measured by monitoring the displacement of a leaf spring to
which one of the sheets is attached. Initially measurements are made
in the presence of the electrolyte solution used, say 10S2mol dm3 KNOB
and then in the same electrolyte solution but with the mica sheets
containing the adsorbed polymer layers. In this manner one can subtract
the double layer interaction from the total energy to obtain the contri-
bution from steric interaction. The forces between mica surfaces bearing
the copolymer are converted to interaction potential energy between flat
surfaces using the Derjaguin approximation for cross cylinders 1651,

E(D) = g (76)

where D is the surface separation and a is the cylinder radius. As an


illustration, Fig. 36 shows the energy-distance curve for mica sheets
covered by a graft copolymer of PMMA backbone and PEO side chains
with a molecular weight of 750. The same copolymer was used as a steric
stabilizer for latex dispersions (particle diameter 330 f 9 nm) for
subsequent rheological measurements. The results in Fig. 35 show a
monotonic and approximately exponential decrease of E(D) with increas-
ing surface separation. The exponential nature of the decay makes it
difficult to assess precisely the point at which the interaction begins,
although it falls below the detection limit of the instrument at about 25 nm.
Using de Gennes scaling theory 1571,it is possible to calculate the
energy of interaction between polymer layers,
166 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

10’

10'
T
E
‘;
ii
” 10'
i?
3
10’

IO’
IO 20

D lnm
D~srancc.

Fig. 36. Interaction energy E(D)against surface separation D, measured using the
macroscopic cross mica technique: (U) 1st compression, (Cl) 1st decompression, (0) 2nd
compression, (0) 2nd decompression, (solid line) calculated from Eq. (77).

E(D) = PkT w2.25 (77)


D1.75 ]-[%+%I
ss [ 1.25(D)i.25 + 1.75(2L)O.75

where L is the stabilizer thickness on each surface, s is the distance


between side chain attachment points, K is the Boltzmann’s constant, 2’
the absolute temperature and l3is a numerical prefactor.
In the above calculations, L was taken to be 12.5 nm and the solid
line in Fig. 36 shows the close agreement between theory and experiment.
The high frequency modulus, G’,, of latex dispersions was obtained
as a function of the volume fraction of the system as discussed above.
This high frequency modulus is related to the potential of mean force
V(R) by the expression (66),

(78)

where N is the number density of particles and g(R) is the radial


distribution function.
The assumption made in the derivation of the above expression that
the particle interactions involve only central pairwise additive potentials
only applies if the particles slip over each other without contact friction.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 167

Both short-range and long-range order have been observed in disper-


sions of monodisperse spheres, and it is likely that at least short-range
order is retained where the system is under oscillatory shear with a low
strain amplitude. Given that within the short-range domains a perfect
lattice arrangement exists, g(R) can be represented as a delta function
centered at the nearest neighbour spacing R. Following Buscall et al.
[51], the more elaborate description of Evans and Lips I671shows that
under these conditions, Eq. (78) reduces to,

G,=NkT+ (79)

where @, is the maximum packing fraction.


The above expression can be expressed in terms of interaction force
F(R), since by convention F = - dV(R)/dR. The data of Fig. 36 are given
as interaction energy between flat plates and this may be converted to
the force between two spheres E(D) using the Derjaguin approximation,

G,=NkT--
&+a
5R2
4E(D)+R7
dJw3
1 (80)

Using Eq. (77) for E(D) and assuming a reasonable value for L (12.5 nm)
and p (7 x lo”, the value giving the best fit to Eq. (77)), it is possible to
calculate G’, as a function of $ from the energy-distance curves. The
results of these calculations are shown in Fig. 37, together with the
measured values of G’,.
It can be seen from Fig. 37 that the form of the curve of G’, versus 0
is correctly predicted, although the calculated moduli values are about
two orders of magnitude higher than the experimental values when the
correct numerical prefactor in de Gennes expression has been used. By
adjusting this numerical prefactor (the solid line in Fig. 37), agreement
between theoretical and experimental moduli may be obtained.

5.4. Flocculated and coagulated systems

The rheology of unstable systems poses problems both from the experi-
mental and theoretical points of view. This is due to the non-equilibrium
nature of the structure at rest, resulting from the weak Brownian motion
[68]. For this reason, advances on the rheology of suspensions, where the
net force is attractive, have been slow and only of qualitative nature. On
168 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

0.35 0.45 0.55 0.65

Volume fraction

Fig. 37. High frequency limit of the storage modulus, G”,, against core volume fraction,
qeare:(D) experimental, (dotted 1ine ), calculated from Eq. (80) with j3 = 7 x 1O3, (solid
line), j3= 2 x 1O4.

the practical side, control of the rheology of flocculated and coagulated


systems is difficult, since the rheology depends not only on the magni-
tude of the attractive forces but also on how one arrives at the flocculated
or coagulated structure in question. As mentioned above, various struc-
tures may be formed, e.g. compact floes, weak and metastable struc-
tures, chain aggregates, etc. At high volume fractions of the suspension,
a flocculated network of particles is formed throughout the sample
whenever it is not being sheared. Under shear, however, this network
is broken into smaller units of flocculated spheres which can withstand
the shear forces [69]. The size of the units which generally survive will
be determined by a balance between the shear forces which tend to break
the units down and the forces of attraction which hold the spheres
together [70-721. The appropriate dimensionless group characterizing
this process (balance of viscous and van der Waals forces) is qou4y/A.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 169

Each flocculated unit is expected to rotate in the shear field, and it


is likely that these units will tend to form layers as individual spheres
do. As the shear stress increases, each rotating unit will ultimately
behave as an individual sphere and, therefore, a flocculated suspension
will show pseudoplastic flow, with the relative viscosity approaching a
constant value (pseudo-Newtonian) at high shear rates. The viscosity-
shear rate curve will also show a pseudo-Newtonian region at low and
high shear rates, as with stable systems, although the values of the
low and high shear rate viscosities (q, and rl__,)will of course depend
on the extent of flocculation in the system and the volume fraction. It
is also clear that such systems will show an apparent yield stress
(Bingham yield value, zP), normally obtained by extrapolation of the
linear portion of the z-y curve to y = 0. Moreover, since the structural
units in a flocculated system change with changes in shear, most
flocculated suspensions show thixotropy. Once shear is initiated,
some finite time is required to break the network of agglomerated
units into smaller units which persist under the shear forces applied.
As smaller units are formed, some of the liquid entrapped in the floes
is liberated, thereby reducing the effective volume fraction of the
solid. This reduction in eeff is accompanied by a reduction in neff and
this plays a major role in generating the thixotropy. Similar arguments
may be invoked to account for the shear thinning behaviour of floccu-
lated suspensions.
It is convenient to distinguish between two types of unstable systems,
depending on the magnitude of the net attractive force. When this is
relatively small, i.e. of the order of a few KT, the suspension is usually
referred to as weakly flocculated. This is the case, for example, with
suspensions that are flocculated in the secondary minimum or those
with a “thin” adsorbed layer. Weak flocculation also occurs when a “free”
(non-adsorbing) polymer is added to a stable suspension. However, in
this case the attractive energy may reach several tens kT units. The
second type of unstable suspensions are those where the net attraction
is large, as for example the case of flocculation in the primary minimum
(usually referred to as coagulated systems) or those flocculated by
reduction of solvency (to worse than a O-solvent) for the chains of a
sterically stabilized suspension. The attractive energy in these cases is
several hundred kT units. Below the rheology of these different systems
will be described.
170 Th. F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

5.4.1. Weakly flocculated systems


As mentioned above, weakly flocculated systems are produced by the
addition of free (non-adsorbing) polymer to a sterically stabilized sus-
pension. Several rheological investigations of such systems have been
carried out by Tadros and his collaborators [73-771.
As an illustration, Fig. 38 shows the variation of the extrapolated
Bingham yield value, zp,of polystyrene latex dispersions (a = 73.5 nm) with
the volume fraction of added PEO, $,, with three different molecular

a
140

: 60

; 60

E
9
g 40 -
C

53
20 -

00 3.06 0.08 0.1 0.12


0.02 0.04
The volume fraction of PEO1200001
_----

0
-
0 0.01 0.02 0.03 0.04 0.05 0.06
Volume fraction of PEO 135000)

Fig. 38. Caption opposite.


Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 171

weights (20000, 35000 and 100000). In these experiments, the volume


fraction of the latex, & was also changed. The latex used contained
grafted PEO chains and to eliminate the possibility of any bridging
flocculation, the grafted chains were made sufficiently dense and hence
the adsorption of free PEO on the latex particles is highly unlikely. An
indirect evidence (that this was the case) was obtained from a study of
the reversibility of flocculation [741.When a flocculated latex suspension
(that contained free polymer with concentration well above that is
required for inducing flocculation) was mixed with a stable latex at the
same volume fraction (but containing no free polymer) such that the free
polymer concentration was brought to values that are below the critical
concentration for flocculation, immediate reversibility was obtained. In
other words, the mixed latex gave a yield value that was identical to
that of the stable latex (containing no free polymer). In addition, all
flocculated latices gave a plastic viscosity that was similar to that of the
stable latex. Thus, there was little doubt that the flocculation obtained
was due to depletion rather than bridging and the floes obtained could
be destroyed by application of high shear.
The results of Fig. 38 show a rapid and linear increase in zp with
increase in $,, when the latter exceeds a critical value, +g . The latter is
the critical free polymer volume fraction for depletion flocculation. ($

50

0.005 0.01 0.015 0.02 3.025 0.03


Volume fraction of PEO (100000)
Fig. 38. The relationship between Bingham yield value and volume fraction of PEO for
various latex ((I = 73.5 nm) volume fractions; (a) M, = 20000; (b) M, = 35000; (c) M,.,
=
100000.
172 Th.F. TadrosIAdv. Colloid Interface Sci. 68 (1996) 97-200

decreases with increase of the molecular weight of the free polymer, M,,
as expected. However, there does not seem to be any dependence of $I,
on the volume fraction of the solid, except in the case of the lowest
molecular weight PEO (AI, = ZOOOO), where there is some indication of
a reduction in ($ with increase in 9,. Similar trends were obtained using
the other latex dispersions (with radii of 217.5 and 457.5 nm). However
there was a definite trend in the effect of particle size; at any given AI,,
the larger the particle size the smaller the value of ($ . This is illustrated

Volume frac:lon of PEO(200001

20 r ,
,
a =0.40 :
I
L ,
I
$=0.45 ,
/
I
(4=0.50 .: 0
I
/
/
j ii-$ /
/
_::::::: 2 A

a
0

D
._:::..-
0

b/
_ 1
IL I
-0 0.01 X.32 0.03 0.34 0.05 0.06
volume '-action of cEC);2~:00)

Fig. 39. Plot of Bingham yield value as a function of volume fraction of PEO (20000) for
various +, values; (a) a = 217.5 nm; (b) a = 457.5 nm.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 173

Table 3

Volume fraction of free polymer at which flocculation starts, $i

Particle radius (nm) M, (PEO)

73.5 20000 0.0150


73.5 35000 0.0060
73.5 100000 0.0055
73.5 20000 0.0150
217.5 20000 0.0055
457.5 20000 0.0050

in Fig. 39 which shows the results for PEO 20000 at the other two
particle sizes used.
A summary of @“p values for the various molecular weights and
particle sizes is given in Table 3.
The results in Table 3 show a significant reduction in @i when the
molecular weight of PEO is increased from 20000 to 35000, whereas
when M, is further increased from 35000 to 100000, the reduction in
+ is relatively smaller. Similarly, there is a significant reduction in
3 when the particle size is increased from 73.5 to 217.5 nm, with a
r:latively smaller decrease on further increase of a to 457.5 nm.
The straight line relationship between the extrapolated yield value
and the volume fraction of the free polymer can be described by the
following scaling law,

(81)

where K is a constant and 171is the power exponent in $, which may be


related to the flocculation process. By using Eq. (811, it is possible to fit
the zp versus $S at any given value of $, by proper choice of the value of
m. Typical plots for latex dispersions with a = 73.5 nm and PEO with
M, = 20000 at various $., values were shown 1761.Similar plots were
obtained at the other PEO molecular weights and the latices with larger
particle radii [76]. The values of 7n used for such fit are summarised in
Table 4.
It can be seen from Table 4 that m is nearly constant, being inde-
pendent of particle size and polymer concentration. An average value
for m of 2.8 may be assigned for such flocculated systems. This value is
174 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

Table 4

Power law fit for the Bingham yield value as a function of latex volume fraction for
various PEO molecular weights

Latex (a = 73.5 nm)

PEO (20000) PEO (35000) PEO (100000)

m m m
@P @P 4%

0.040 3.0 0.022 2.9 0.015 2.7


0.060 2.7 0.030 3.0 0.020 2.7
0.08 2.8 0.040 2.8 0.025 2.8
0.100 2.8 0.050 2.9 -

Latex (a = 217.5 nm) Latex (a = 457.5 nm)

PEO (20000) PEO (20000)

@!J m % m

0.020 3.0 0.020 2.7


0.040 2.9 0.030 2.7
0.060 2.8 0.040 2.8
0.080 2.8 0.050 2.7

similar to the exponent predicted for diffusion controlled aggregation


(3.5 I!I0.2). Ball and Brown [78] developed a computer simulation method
treating the floes as fractals which are closely packed throughout the
sample and they showed that the shear modulus scales with c)~,with an
exponent m that is equal to 4.4 + 0.2 for chemically limited aggregation
and 3.5 f for diffusion limited aggregation. Experimental results by
Buscall and Mill 1791 confirmed these theoretical predictions.
The near independence of $$ on $S can be explained on the basis of
the statistical mechanical approach of Gast et al. [37,801 which showed
such independence when the osmotic pressure of the free polymer
solution is relatively low and/or the ratio of the particle diameter to the
polymer coil diameter is relatively large (> 8-9). The latter situation is
certainly the case with the latex suspensions with diameters of 435 and
915 nm at all PEO molecular weights. The only situation where this
Th. F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 175

condition may not be fulfilled is with the smallest latex and the highest
molecular weight. However, even in this case, the dependence of @: on
@, is not pronounced (Fig. 3%).
The dependence of @i on particle size can be explained from a
consideration of the dependence of the free energy of depletion and van
der Waals attraction on particle radius. Both attractions increase with
increase of particle radius. Thus, the larger particles would require
smaller free polymer concentration at the onset of flocculation. This is
indeed the case as shown in Fig. 39 and Table 3.
It is possible, in principle, to relate the extrapolated Bingham yield
stress, zp, to the energy required to separate the floes into single units,
Esep 173,741,

3 0, n Esep
zp = (82)
ma3
where n is the average number of contacts per particle (the coordination
number). The maximum value for n is 12 which corresponds to hexago-
nal or face-centered cubic lattice. This maximumvalue is highly unlikely
for particle arrangement in a flocculated system. A more realistic value
for n is 8, corresponding to random arrangement of particles in a floe,
again assuming a compact structure. Again, this is probably unlikely
and a more realistic value, corresponding to an open floe structure would
be a significantly smaller value than 8.
In order to estimate Esepfrom ‘tP,a number of assumptions have to
be made. Firstly, one has to assume that all of the particle-particle
contacts are broken by shear. This may not always be realised, although
viscosity measurements showed that the high shear value for a floccu-
lated system is close to that of the latex before the addition of the free
polymer. This implies that, under these high shear conditions, most of
the particle-particle contacts are indeed broken. The second assumption
that has to be made is the value to be assigned for n. As mentioned above,
n becomes smaller the more open the floe structure is. It is, therefore,
possible that n may not remain constant, depending on the extent of
flocculation, which depends on the volume fraction of the free polymer
as well as its molecular weight. For the sake of comparison, values of n
varying between 4 and 12 were used and the results of the calculations
were tabulated by Liang et al. 1761.Some selected data (a = 73.5 nm)
with n = 4 are given in Table 5, together with the values of Gdepcalculated
from the Asakura and Oosawa (AO) 1301and Fleer, Scheutjens and
176 Th. F. T&-OS IAdv. Colloid Interface Ski. 68 (1996) 97-200

Table 5

Results of and
Esep Gdepcalculated using the A0 and FSV theories

@P ‘TP E =P Gdep
(N rn? (kT) MY

A0 model FSV model

(a) M, = 20000
0.04 12.5 0.30 8.4 18.2 78.4
21.0 0.35 12.1 18.2 78.4
30.5 0.40 15.4 18.2 78.4
40.0 0.45 18.0 18.2 78.4

(b) M, = 35000
0.03 17.5 0.30 11.8 15.7 78.6
25.7 0.35 14.8 15.7 78.6
37.3 0.40 18.9 15.7 78.6
56.8 0.45 25.5 15.7 78.6

(c) M, = 100000
0.02 10.0 0.30 6.7 9.4 70.8
15.0 0.35 8.7 9.4 70.8
22.0 0.40 11.1 9.4 70.8
32.5 0.45 14.6 9.4 70.8

Vincent (FSV) [31] theories on depletion flocculation. These theories


were discussed before and the equations for Gdepwere given (Eqs. (29)
and (30) respectively).
It can be seen from Table 5 that Es_, at any given +Pvalue, increases
with increase of +,. In contrast, the value of Gdepdoes not depend on the
value of $,. The theories on depletion flocculation only show a depend-
ence on $, and a. Thus, one cannot make a comparison between Esepand
G dep.The close agreement between E,, (assuming a value of n of 4) and
Gdepbased on Asakura and Oosawa’s theory 1291should only be consid-
ered fortuitous.
Using Eqs. (81) and (82), a general scaling law may be used to show
the variation of Esepwith the various parameters of the system,

(83)
Th.F. TadrosIAdv. Colloid Interface Sci. 68 (1996) 97-200 177

Equation (83) shows the four main parameters that determine I&_,:
the particle radius a, the volume fraction of the suspension $,, the
concentration of free polymer $,, and the molecular weight of the free
polymer, which together with a determines $i .
More insight on the structure of the flocculated latex dispersions was
obtained using viscoelastic measurements [771. As an illustration Fig.
40 shows plots of the storage modulus (G’) versus qp for PEO 20000, at
various latex (a = 73.5 nm) volume fractions (Q. Similar results were
obtained for the other PEO molecular weights. All the results show the

70

60

10

I
0’
0 0.02 0.W 0.06 0.12
Volume fraction of

Fig. 40. Plot of storage modulus (at 0.1 Hz) as a function of volume fraction of PEO
(20000) for latex (a = 73.5 nm).
178 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97400

same trend, namely an increase in G’with increase in Qp.However, the


results obtained differ from those obtained using steady state measure-
ments, namely the variation of G’with $, is not linear. In all cases, G
showed a gradual increase with increase in $, and in most cases it
reached a plateau at high 0, values. In contrast, zp showed a rapid and
linear increase with increase in $, with no sign of a plateau. This reflects
the difference in the two types of measurements. The elastic moduli were
obtained at low deformation which causes little perturbation to the
flocculated structure. Above $i flocculation occurs and thus G’ in-
creases in magnitude with further increase in $, until a three-dimen-
sional network structure is reached and G’reaches a limiting value. Any
further increase in polymer concentration may also cause a change in
the structure of the flocculated system, but not a significant increase in
the number of bonds between the units formed. G’ is predominantly
determined by the structure of the floes, i.e the number ofbonds between
flocculated units produced, although there is some contribution from the
strength of the units formed. The extrapolated yield value, on the other
hand, is determined more by the strength of the structural units formed.
In a shear flow these units become broken down and may reform as a
result of collisions between the particles. The stronger the particle-par-
ticle interaction, the higher the energy dissipation in shear flow and the
higher the extrapolated yield value.
The influence of particle size is shown in Fig. 41 which shows the
variation of the storage modulus of a 0.40 volume fraction dispersion
with Qpfor PEO 20000 at three particle diameters of 147,435 and 915
nm. It can be seen that the onset of flocculation does not show a clear
dependence on particle size. This is in contrast to the results obtained
using steady state measurements which showed a definite decrease in
ot: with increase in particle size. It is likely that the modulus measure-
ments are not sensitive enough to detect flocculation at low free polymer
concentration.
The results shown in Fig. 41 show that, at any given free polymer
concentration, the storage modulus increases with a decrease in particle
size. This effect can be understood from a consideration of the surface-
to-volume ratio which increases with a decrease in particle size. For a
given volume fraction of the dispersion, the total volume of the depletion
zone increases with decrease of particle size (a consequence of the larger
number of particles at a fured Q,). This results in an increase in the net
attraction and this is reflected in an increase of the storage modulus. As
mentioned above, the storage modulus is dominated by the number of
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 179

0.02 0.04 0.06 0.08 0.1 0.12


Volume fraction of PEO !Zp

Fig. 41. The relationship between the storage modulus (at 0.1 Hz) and free polymer (PEO
20000) suspensions of different particle sizes.

bonds in a flocculated system, and with a decrease in particle size the


number of such bonds increases resulting in a higher modulus.
Further evidence for the above effect of particle size was obtained
from plots of G*, G’and G” versus eS. The results are shown in Figs. 42
and 43 for two particle diameters of 147 and 915 nm, respectively. These
results were obtained from oscillatory measurements using PEO 35000 at
$ = 0.022 for D = 147 nm and PEO 20000 at $ = 0.02 for D = 915 nm. It
can be seen from the results of Figs. 42 and 43 that, at any given +,, the
moduli values are much higher for the smaller latex dispersions. In
addition, the cross-over point at which G’= G” occurs at 4, = 0.25 for
180 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

4 1
I I I I
1’
0.1 0.3 0.3 0.35 0.4 0.45 0.3
Volume fraclion of Ial~x (D=l4inm) 0,

Fig. 42. Plot of moduli (at 0.1 Hz) versus volume fraction of latex (D = 147 nm) with PEO
35000 at Qp= 0.022.

D = 147 and at +, = 0.55 for D = 915. The cross-over point may be taken
as the onset of flocculation. It is clear that the smaller the particle size,
the smaller the 4, value at the onset of flocculation.
Optical micrographs of the floe structure for the small latex particles
(D = 147 nm) 1771flocculated by PEO 20000 showed an open and discrete
floe structure at $, values just above the critical flocculation concentra-
tion (+r = 0.02). On increasing $, to 0.04, the floe structure was still
relatively open but the structure was relatively more compact when
compared to that at QS= 0.02, although there were several small floes in
existence. However, when $, was increased above 0.06, the floe structure
was further compacted and there was little difference in the floe struc-
ture at these high 8, dispersions. These observations together with the
larger floe size indicates that the floe strength at these high $ is higher
at these high $, values than at low $, values, as predicted. The similarity
in structure at $, > 0.06 explains the reason for the plateau value in G
at high ep values.
Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 181

J
0.35 0.4 0.4s 0.5 O.S5 0.6 0.6s
Volume fraction of latex (D=9lSnmtP

Fig. 43. Plot of moduli (at 0.1 Hz) versus volume fraction of latex (D = 915) with PEO
20000 at $p = 0.020.

5.4.2. Strongly flocculated systems.


These can be exemplified by flocculation of sterically stabilised sys-
tems produced by reduction of the solvency of the medium for the
stabiliser chain. For example, sterically stabilised latex dispersions
containing polyethylene oxide (PEO) can be flocculated by addition of
electrolyte, e.g. Na,SO, above a critical concentration (CFC). Alterna-
tively, at a given electrolyte concentration the latex dispersion becomes
flocculated above a critical temperature (CFT). The flocculation mecha-
nism was discussed above.
The flocculation of concentrated dispersions can be investigated
using rheological measurements [Sll. As an illustration, Fig. 44 shows
the variation of the extrapolated yield value as a function of Na2S0,
concentration at various latex volume fractions at 25°C. The latex had
a z-average particle diameter of 435 nm and it contained grafted PEO
with M = 2000. It is clear that when @, < 0.52, zB is virtually equal to
182 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

40 r

1
gs =0.40

g9=0.56
-

0.1 0.2 0.3 0.4 0.5 0.6


Concentration of Na,SO, (aq)/mol dn+
Fig. 44. Variation of Bingham yield value as a function of Na,SO, concentration at
various latex volume fractions at 25°C.

zero up to 0.3 mol dm3 Na$O, above which it shows a rapid increase
with further increase in Na$O, concentration. When 0, > 0.52, a small
yield value is obtained below 0.3 mol dm3 Na$O,, which may be
attributed to the possible elastic interaction between the grafted PEO
layers when the particle-particle separation is less than 26 (where 6 is
the grafted PEO layer thickness). However, above 0.3 mol dm3 Na2S0,,
there is a rapid increase in zp. Thus, the CFC of all the concentrated
latex dispersions is around 0.3 mol dm3. It should be mentioned that
at Na,SO, concentrations below the CFC, zp shows a measurable de-
crease with increase in Na,SO, concentration. This is due to the reduc-
tion in the effective radius of the latex particle as a result of the
reduction in solvency of the medium for the chains [82]. This accounts
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 183

60)

4,=0.56
5
UY
20 -

10 -

0
0 0.1’ 0.2 0.3 0.4 0.5 0.6
Na$O, (aq) concentration/ mol drne3
Fig. 45. Storage modulus as a function of Na2S0, concentration at various latex volume
fractions at 25°C.

for a reduction in the effective volume fraction of the dispersion which


is accompanied by a reduction in zP.
Figure 45 shows the results of the storage modulus as a function of
Na,SO, concentration. These results show the same trend as those
shown in Fig. 44, i.e. an initial reduction in G’due to the reduction in
the effective volume fraction, followed by a sharp increase above the
CFC (which is also 0.3 mol dm3).
Log-log plots of zP or G’versus 0, at various Na$O, concentrations
are shown in Figs. 46 and 47. All the data are described by the following
scaling equations,
184 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

in water
-

in 0.1 M
-

in 0.2 A4
-

in 0.3 M

in 0.4 M

in 0.5 M
1

0.03 L
I
0.35 0.4 0.45 0.5 0.55 0.6 0.65
Latex volume fraction qj
S

Fig. 46. Log-log plots of zPvemus I$~for latex suspensions at various Na.$O, concentrations.

Tp=k$F (84)

G’ = k’ 4; (0.35 < 0, < 0.53) (85)

where k and k’ are constants and m and n are exponents. The values of
m and n are listed in Table 6. These clearly show a sudden drop in 172
value from a value of -31 to -9.1 and of n from -30 to - 12 as the Na2S04
concentration is increased from 0.3 to 0.4 mol dm3. With further
Th.F. TadroslAdu. Colloid Interface Sci. 68 (1996) 97-200 185

50

20

10

$5
? in water
B -
EL
in 0.1 M
& -
s 2
in 0.2 M
G -

1 in 0.3 M

in 0.4 M
-
0.5
in 0.5 M
-

0.35 0.4 0.45 0.5 0.55 0.6 0.65


Latex volume fraction
ds

Fig. 47. Log-log plots of G versus & for latex suspensions at various Na,SO, concentrations.

increase in Na#O, concentration from 0.4 to 0.5 mol dm3, 7ndrops from
9.4 to 2.8, whereas n drops from 12 to 2.2. This low exponent is an
indication that an open network floe structure with low fractal dimen-
sions is formed.
The exponent of 2.8 or 2.2 at 0.5 mol dm3 NasSO, is just in the range
of the reported values in the literature. Many authors [78,79, 83-851
have reported that the exponent for flocculated suspensions is in the
range 2.0-4.5. However, the value of the exponent depends to some
extent on the treatment a coagulated suspension has been subjected to
before the measurements were made. Shi et al. [861 have reported that
186 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97400

Table 6

Scaling relationship between ~~ or G’ and (I~at various Na,SO, concentrations

Na,SO, concentration m n
(mol dm3)

0.0 47 27
0.1 44 34
0.2 34 31
0.3 31 30
0.4 9.4 12
0.5 2.8 2.2

the storage modulus exhibited a power law behaviour with respect to


particle concentration $ of colloidal gels, i.e. G’ - $4.1-4.2.The exponent
2.2 for G’ is comparable to that for colloidal silica suspensions [87]
assuming that the clusters comprising the network are fractals, al-
though in the case of the latex suspensions the flocculation is weaker.
Figure 48 shows the variation of G*, G’and G” with temperature at
0.2 mol dme3 NasSO,. The results show an initial systematic reduction
in the moduli values with increasing temperature up to 50°C. This is
the result in reduction in solvency of the chains with increasing tem-
perature. The latter increase causes a breakdown in the hydrogen bonds
between the ethylene oxide chains and water molecules. As discussed
above, reduction in solvency causes a reduction in the thickness of the
grafted PEO chains and hence a reduction in the effective volume
fraction of the suspension. This leads to a decrease in the modulus
values. However, above 50°C there is a rapid increase in the moduli
values with further increase in temperature. This temperature denotes
the onset of flocculation, i.e. the critical flocculation temperature (CFT)
at this electrolyte concentration. Similar trends were obtained at 0.3
and 0.4 mol dms3 Na,SO,, but in these cases the CFT was 35 and 15°C
respectively. Careful observation of the results showed some difference
in the trends obtained at various electrolyte concentrations. In 0.2 mol
dmm3Na2S04, G’was close to G” above the CFT, whereas for the other
two electrolyte concentrations, G’ was significantly greater than G”
above the CFT. A comparison between the three electrolyte concentra-
tions is shown in Fig. 49, where the ratio G/G” is plotted versus
temperature.
Th.F. Tadros lAdv. Colloid Interface Sci. 68 (1996) 97-200 187

25

10

20 30 40 50 60
Temperature / ??
C
Fig. 48. Variation of G’, G’ and G” with temperature for latex dispersions @, = 0.550)
in 0.2 mol dm3 Na$O,.

Figure 49 shows an initial reduction in the value of G’E with


increase of temperature, which as discussed above is due to the reduc-
tion of solvation of chains with increase of temperature. However, above
a critical temperature (the CFT) there is a rapid increase in this ratio.
The plots in Fig. 49 give a better location of the CFT.
Two models may be used to interpret the rheological results of the
above flocculated system. Both models have been developed by Hunter
and his collaborators [88-911. In the first model, Frith, Neville and
Hunter [EM]introduced a doublet floe structure model to deal with
sterically stabilised dispersions which have undergone flocculation.
They assumed that the major contribution to the excess energy dissipa-
tion in such pseudoplastic systems comes from the shear field which
188 Th.F. TadrosIAdv. Colloid Interface Sci. 68 (1996) 97-200

I I I
01 I I I

0 10 20 30 40 50 60 70
Temperature/‘C

Fig. 49. Plot of G’E’ versus temperature for latex dispersions (I$, = 0.550) at various
Na,SO, concentrations.

provides energy to separate contacting particles in the floes. Thus, the


extrapolated yield value can be expressed as,

Ttp =
#I E

(86)
2n2(a + Q3 sep
where QHis the hydrodynamic volume fraction of the particles, that is
equal to the effective volume fraction, i.e.,

(87)
Th.F. Tadros fAdv. Colloid Interface Sci. 66 (1996197-200 189

(a + S) is the interaction radius of the particle and Esep is the energy


needed to separate a doublet, which is the sum of van der Waals and
steric attractions,

Esep=&+G,
0
(88)

At a particle separation of - 12 nm (twice the grafted polymer layer


thickness), the van der Waals attraction energy is very small (1.66 KT)
and the contribution from G, to the attraction is significantly larger than
the van der Waals attraction. Therefore, Esepmay be approximated to
G,. Thus, from Eq. (87) one can estimate ESepfrom the extrapolated yield
value, zp, for the flocculated suspension. The results are summarised in
Table 7, which show an increase in Esepwith increase in QS,as expected.
The values are unrealistically high since the flocculation in the present
concentrated dispersions is not reversible within the conditions of the
rheological measurements. In addition, the assumptions made for the
calculation of Esepfrom ~~are not fully justified and the data shown in
Table 7 should only be considered as qualitative.
The second model used by Hunter and collaborators [89-911 was
based on the elastic floe structure. In this case, the structural units were
assumed to be small floes of particles (called flocculi) which are charac-

Table 7
Results ofE,,, calculated from zpfor flocculated sterically stabilised latex dispersions at
various latex volume fractions

0.4 mol dm3 Na,SO, 0.5 mol dm3 Na,SO,

zp (Nm? EsepWY) ~~(Nm-” Esep(KT)

0.429 1.3 97 0.246 3.5 804


0.452 2.4 165 0.288 5.4 910
0.511 3.3 179 0.330 7.4 961
0.535 5.3 262 0.371 11.4 1170
0.554 7.3 336 0.409 14.1 1190
0.569 9.1 397 0.440 17.0 1240
0.578 17.4 736 0.465 21.1 1380
0.487 23.1 1380
0.515 28.3 1510
190 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

terised by the extent to which the particle structure is able to entrap the
dispersion medium. A floe is made from an aggregate of several flocci.
These flocculi may range from loose open structures (if the attractive
forces between the particles are strong) to very close packed structures
with little entrained liquid (if the attractive forces are weak). In the
system of flocculated sterically stabilised dispersions, the structure of
the flocculi depends on the volume fraction of the suspension and how
far the system is from the CFC. Just above the CFC, the flocculi are
probably close packed (with relatively small floe volume), whereas far
above the CFC, a more open structure is found which entraps a consid-
erable amount of liquid. Both types of flocculi persist at high shear rates,
although the flocculi with weak attraction may become more compact
by maximising the number of interactions within the flocculus.
A satisfactory description of the kind of flow of the above flocculated
system (which is pseudoplastic) has been given by Hunter and collabo-
rators [89-911 by analysing the energy dissipation process as suggested
by Goodeve [92] and Gillespie 1931.These authors suggested that the
energy dissipation in flow can be separated into two components due to
the viscous flow of suspension medium around the flow units and the
energy dissipated in overcoming the interaction between the particles.
The total energy dissipated in flow, Etot, is then given by

E tot = 2p Y + Tpl f (89)

Since the flocculated systems mostly show a linear z-y above a critical
shear rate, Ycrit,it follows that nlplbecomes constant above Ycritand hence
the degree of openness of the flow units must remain constant above
Ycrit.In the elastic floe model, it is assumed that the flocculi established
at high shear rate are able to form floes which have an essentially
constant ratio of solid to trapped liquid. The degree to which the liquid
is trapped is measured by the floe volume ratio, C,,, given by the ratio
of e/Q, (where Qf is the volume fraction of floes and Qs that of the
particles). At high volume fraction, (Qand hence CFPmay be calculated
from the Krieger equation 1451,

(90)

where no is the viscosity of the medium, $“, is the maximum packing


Th. F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 191

Table 8

Floe radius calculated using the elastic floe model for flocculated sterically stabilised
polystyrene latex dispersions at various volume fractions in 0.4 mol dmm3Na,SO,

qlp1 rlr Qf c FP afloc ‘I,


(mPas) (Pm) (s-3

0.409 13.1 14.7 1.27 0.567 1.386 27 95


0.452 21.7 24.4 2.39 0.608 1.344 34 95
0.485 30.4 34.1 3.38 0.630 1.299 39 95
0.535 63.2 71.1 5.33 0.666 1.245 45 95
0.554 104.8 117.7 7.30 0.684 1.234 51 95
0.569 143.7 161.4 9.09 0.693 1.217 56 95
0.578 282.7 317.5 17.40 0.707 1.223 66 126

Table 9

Floe radius calculated using the elastic floe model for flocculated sterically stabilised
latex dispersions at various latex volume fractions in 0.5 mol dms Na,SO,

Qf c FP afloc
(pm)

0.246 9.8 11.0 3.5 0.538 2.185 46 156


0.288 14.2 15.9 5.4 0.574 1.993 47 187
0.330 16.6 18.6 7.4 0.588 1.781 51 187
0.371 22.8 25.7 11.4 0.612 1.650 54 218
0.409 27.0 30.4 14.1 0.623 1.524 57 218
0.440 33.6 37.8 17.5 0.636 1.446 60 218
0.465 42.5 47.7 21.1 0.648 1.394 60 249
0.515 62.3 69.9 29.0 0.666 1.292 62 280

fraction which may be taken as 0.74 and [ql is the intrinsic viscosity
taken as 2.5. The values of CFP calculated using Eq. (90) are listed in
Table 8.
Several other dissipation processes may be identified in the high
shear rate regime, i.e. orientation, stretching and compression of the
floes and transfer of flocculi from one floe to another. The most important
assumption of the elastic floe model is that all dissipative processes can
192 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

be simultaneously evaluated by considering in detail what happens


when two floes collide and are separated by the shear field. Detailed
examination of the process showed that only one contribution is impor-
tant in the energy dissipation process [94,95], namely the energy in-
volved in the movement of liquid into and out of the space between the
adjoining particles inside the floes. This leads to the following equation
for the Bingham yield value,

(91)

where cl0is the orthokinetic capture efficiency which depends weakly on


shear rate (a,, = p.18>, p is a constant (=27/5), h is a correction factor of
the order of unity, q0 is the viscosity of the medium, aflocis the radius of
the floes and A is the distance through which bonds are stretched inside
the floe by the shearing force. Thus, using Eq. (911, one can calculate
the radius of the floes from the experimental values of the extrapolated
yield value, assuming a0 = 1, h = l/3and A = 0.5 nm 1961.The results of
the calculation are given in Tables 8 and 9 at 0.4 and 0.5 mol drna
Na,SO, respectively.
Figure 50 shows the plot of the floe radius as a function of latex
volume fraction at the two Na2S0, concentrations studied. At any given
electrolyte concentration, the floe radius increases with increase in the

0.3 0.4 0.5 0.6


Latex volume fraction tin

Fig. 50. Plot of the radius of the floe as a function of latex volume fraction at above CFC.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 193

latex volume fraction, as expected. This can be understood by assuming


that the larger floes are formed by fusion of two floes and the smaller
floes by “splitting” of larger ones. From simple statistical arguments,
one can predict that aflocwill increase in 9, because in such a process
larger floes are favoured over small ones. In addition, at any given
volume fraction of latex, the floe radius increases with increase in
electrolyte concentration. This is consistent with the scaling results as
discussed above.
The above results show clearly the correlation of viscoelasticity of
flocculated dispersions with their interparticle attraction. These measure-
ments allow one to obtain the CFC and CFT of concentrated flocculated
dispersions with reasonable accuracy. In addition, the results obtained
can be analysed using various models to obtain some characteristics of

Fig. 51. Strain sweep (at a frequency of 1 Hz) for a coagulated polystyrene latex
suspension at various volume fractions.
194 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

the flocculated structure, such as the “openness” of the network, the


liquid entrapped in the floe structure and the floe radius. Clearly,
several assumptions have to be made, but the trends obtained are
consistent with expectations from theory.
Another example of strongly flocculated system is electrostatically
stabilised polystyrene latex suspensions coagulated by addition of elec-
trolyte, e.g. 0.2 mol dmm3NaCl. In this case coagulation into the primary
minimum occurs. The structure of such coagulated systems becomes
partially broken down above a critical strain (deformation) that depends
on the volume fraction of the suspension. This is illustrated in Fig. 51
which shows the variation of G*, G’and G” (at a frequency of 1 Hz) with
strain amplitude for such coagulated latex suspensions at various
volume fractions. It can be seen that G* and G’ initially remain inde-

0=0.X6

05 0.1 1
001
WI Hz
Fig. 52. Frequency sweep (in the linear viscoelastic region) for coagulated polystyrene
latex suspensions at various volume fractions.
Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200 195

loL 0

I I ol I I

s X10-2 lo-’ 2 xlcr’ 5m’


@

Fig. 53. Log-log plots of G’ versus $ for coagulated polystyrene latex suspensions.

pendent of the applied strain (the linear viscoelastic region), but above
a critical strain, yCYcr,
the moduli show a rapid reduction with further
increase in the strain amplitude. In contrast, G” (which is much lower
than G’) initially remains constant and then increases to a maximum
in the region of yCrand then decreases with further increase in yO.The
region above yCYcr
is the non-linear viscoelastic region whereby the floccu-
lated structure starts to become broken down with the applied shear.
Figure 52 shows the variation of G*, G’and G” (in the linear viscoelastic
region) with frequency, at various latex volume fractions. It can be seen
that in all cases G’ >> G” and the moduli show little dependence on
frequency. This behaviour is typical of a highly elastic (coagulated)
structure, whereby a “continuous” network structure is produced at such
high volume fractions. Scaling laws can be applied for the variation of the
modulus with the volume fraction of the suspension. The values of G’were
taken from the frequency sweep curves of Fig. 52 (which showed little
dependency on frequency). A log-log plot of G’versus 41for such coagulated
system is shown in Fig. 53. Such plots are linear and the variation of G
with $ can be represented by the following scaling equation,
196 Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200

y
5
2
10-l

10-2
I a

10-3

lo+
I
I I I I

10-l 2x10” 3m’ Lm


6
Fig. 54. Log-log plots of E, versus $.

G’= 1.98 x 101 (P (92)

The high power in $ is indicative of a relatively compact flocculated


structure. This is not surprising since the latex was coagulated at 0.2
mol dm4 NaCl, which is not much higher than the CFC of the latex
(-0.1 mol dn-?).
It is also possible to obtain the cohesive energy in the flocculated
structure from a knowledge of G’(in the linear viscoelastic region> and
ycr. The cohesive energy is related to the stress in the coagulated
structure by the following equation 1961,

E,=jody (93)
0

Since cs= y. G’, then,


Th.F. Tadros IAdv. Colloid Interface Sci. 68 (1996) 97-200 197

A log-log plot of E, versus (I is shown in Fig. 54 for such coagulated latex


suspensions.
The&-Q curve can be represented by the following scaling relationship,

E, = 1.02 x 103 q9.l (95)

Again, the high power in $ is indicative of the compact structure of these


coagulated suspensions.

References

111 R.H. Ottewill, in: J.W. Goodwin (Ed.), Concentrated Dispersions. Royal Society of
Chemistry Publication, No. 43, 1982, Chapter 9.
121 R.H. Ottewill, Concentrated dispersions, I. Fundamental considerations, in: G.W.
Poehlein, R.H. Ottewill and J.W. Goodwin (Eds.), Science and Technology of
Polymer Colloids, Vol. II. Martinus Nijhoff, Boston, The Hague, 1983, p. 503.
[31 Th.F. Tadros, Colloids Surfaces, 18 (1986) 137.
[41 P.N.J. Perrin, Brownian Movement and Molecular Reality. Taylor and Francis,
London, 1880.
[51 A. Einstein, Investigation on the Theory of Brownian Motion. Methuen, London, 1916.
[61 R.H. O&will, Ber. Bunsenges. Phys. Chem., 89 (1985) 517.
[71 I. Markovic, R.H. Ottewill, Th.F. Tadros and S.M. Underwood, Langmuir, 2 (1986)
625.
[81 R.H. Ottewill, in: Th.F. Tadros (Ed.), Solid/Liquid Dispersions. Academic Press,
London, 1987.
191 H.R. Kruyt, Colloid Science, Vol. I. Elsevier, Amsterdam, 1952.
1101 E.J.W. Verwey and J.Th.G. Gverbeek, Theory of Stability of Lyophobic Colloids.
Elsevier, Amsterdam, 1948.
I111 D.H. Napper, Polymeric Stabilization of Colloidal Dispersions. Academic Press,
London, 1981.
Ml P.J. Flory and W.R. Krigbaum, J. Chem. Phys., 18 (1950) 1086.
1131 P.J. Flory, Principles of Polymer Chemistry. Cornell University Press, Ithaca, NY,
1953.
1141 E.W. Fischer, Kolloid Z., 160 (1958) 120.
[151 R.H. Ottewill and T. Walker, Kolloid Z.Z. Polym., 227 (1968) 108.
1161 E.L. Mackor, J. Colloid Sci., 6 (1951) 490.
1171 E.L. Mackor and J.H. van der Waals, J. Colloid Sci., 7 (1951) 535.
1181 H.C. Hamaker, Physica (Utrecht), 4 (1937) 1058.
198 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

1191 F.Th. Hesselink, A. Vrij and J.Th.G. Overbeek, J. Phys. Chem., 65 (1971) 2094.
[20] Th.F. Tadros, in: Th.F. Tadros (Ed.), The Effect of Polymers on Dispersion
Properties. Academic Press, London, 1982, p. 1.
[21] M. Garvey, Th.F. Tadros and B. Vincent, J. Colloid Interface Sci., 49 (1974) 57.
[22] M. Garvety, Th.F. Tadros and B. Vincent, J. Colloid Interface Sci., 55 (1976) 440.
[23] von Smoluchowski, Z. Phys. Chem., 92 (1927) 129.
[24] N. Fuchs, Z. Physik., 89 (1934) 736.
[25] H. Reerink and J.Th.G. Overbeek, Disc. Faraday Sot., 18 (1954) 74.
[26] J. Gregory, in: Th.F. Tadros (Ed.), Solid/Liquid Dispersions. Academic Press, 1987,
Chapter 8.
[27] R.A. Ruehrwein and A. Ward, Soil Sci., 73 (1952) 485.
[28] J. Gregory, J. Colloid Interface Sci., 42 (1973) 448.
[29] S. Asakura and F. Oosawa, J. Chem. Phys., 22 (1954) 1255.
[30] S. Asakura and F. Oosawa, J. Polym. Sci., 33 (1958) 183.
[31] A. Vrij, Pure Appl. Chem., 48 (1976) 471.
[32] P.R. Sperry, J. Colloid Interface Sci., 87 (1982) 375.
[33] G.J. Fleer, J.H.M.H. Scheutjens and B. Vincent, ACS Symposium Ser., 240 (1984)
245.
[34] J.H.M.H. Scheutjens and G.J. Fleer, J. Phys. Chem., 63 (1979) 1619.
[35] J.H.M.H. Scheutjens and G.J. Fleer, Adv. Colloid Interface Sci., 16 (1982) 361.
[36] B. Vincent, J. Edwards, S. Emmett and A. Jones, Colloids and Surfaces, 18 (1986)
261.
[37] A.P. Gast, C.K. Hall and W. Russel, Faraday Disc. Chem. Sot. 76 (1983) 189.
[38] J.D. Ferry, Viscoelastic Properties of Polymers. Wiley, New York, 1980.
[39] J.W. Goodwin, in: Th.F. Tadros (Ed.), Solid/Liquid Dispersions. Academic Press,
London, 1987.
[40] R.W. Whorlow, Rheological Techniques. Ellis Hoorwood, Chichester, 1980.
[41] N. Casson, in: C.C. Mill (Ed.), Rheology of Disperse Systems. Pergamon Press,
Oxford (1959) p. 84.
[42] M.E. Woods and I.M. Krieger, J. Colloid Interface Sci., 34 (1970) 91.
[43] Y.S. Papir and I.M. Krieger, J. Colloid Interface Sci., 34 (1970) 126.
[44] I.M. Krieger and M. Dougherty, Trans. Sot. Rheol., 3 (1959) 137.
[45] I.M. Krieger, Adv. Colloid Interface Sci., 3 (1972) 111.
[46] G.K. Bachelor, J. Fluid Mech., 83 (1977) 97.
[47] R. Buscall, J.W. Goodwin, M.W. Hawkins and R.H. Ottewill, J. Chem. Sot.
Faraday Trans. I, 78 (1982) 2873.
[48] Th.F. Tadros, Langmuir, 6 (1990) 28.
[49] Th.F. Tadros and A. Hopkinson, Faraday Disc. Chem. Sot., 90 (1990) 41.
[50] J.W. Goodwin and A.M. Rhider, in: M. Kerker (Ed.), Colloid and Interface Science.
Academic Press, New York, 1976, Vol. 4, p. 529.
[51] R. Buscall, J.W. Goodwin, M.W. Hawkins and R.H. Ottewill, J. Chem. Sot.
Faraday Trans. I, 78 (1982) 2889.
[52] S.J. Willey and C.W. Macosko, J. Rheol., 22 (1978) 525.
[53] M.D. Croucher and T.H. Milkie, in: Th.F. Tadros (Ed.), The Effect of Polymers on
Dispersion Properties. Academic Press, London, 1982, p. 101.
[54] C. Prestidge and Th.F. Tadros, J. Colloid Interface Sci., 124 (1988) 660.
Th.F. Tadros lAdv. Colloid Interface Sci. 68 (1996) 97-200 199

1551 W. Liang, Th.F. Tadros and P.F. Luckham, J. Colloid Interface Sci., 153 (1992)
131.
WI C. Bromley, Colloids Surfaces, 17 (1985) 1.
I571 P.G. de Gennes, Adv. Colloid Interface Sci., 27 (1987) 189.
I581 G.N. Choi and I.M. Krieger, J. Colloid Interface Sci., 113 (1986) 101.
I591 B.A. de L. Costello, P.F. Luckham and Th.F. Tadros, Colloids and Surfaces, 34
(1988/1989) 301.
WI1 P.F. Luckham, M.A. Ansarifar, B.A. de L. Costello and Th.F. Tadros, Powder
Technology, 65 (1991) 371.
I611 B.A. de L. Costello, P.F. Luckham and Th.F. Tadros, J. Colloid Interface Sci., 152
(1992) 237.
[621 Th.F. Tadros, W. Liang, B. Costello and P.F. Luckham, Colloids and Surfaces, 79
(1993) 105.
[631 J.N. Israelachvili and G.E. Adams, J. Chem. Sot. Faraday Trans I, 74 (1978) 975.
[641 P.F. Luckham, Powder Technology, 58 (1989) 75.
[651 L.R. White, J. Colloid Interface Sci., 95 (1983) 286.
I661 R. Zwanzig and R.D. Mountain, J. Chem. Phys., 43 (1965) 4464.
I671 I.D. Evans and A. Lips, J. Chem. Sot. Faraday Trans., 86 (1990) 3413.
[681 W.B. Russel, J. Rheol., 24 (1980) 287.
[691 R.L. Hoffman, in: G.W. Poehlein, R.H. Ottewill and J.W. Goodwin (Eds.), Science
and Technology of Polymer Colloids. Martinus Nijhoff, Boston, The Hague, Vol.
II, 1983, p. 570.
[701 B.A. Firth and R.J. Hunter, J. Colloid Interface Sci., 57 (1976) 248.
I711 T.G.M. van de Ven and R.J. Hunter, Rheol. Acta, 16 (1976) 534.
I721 R.J. Hunter and J. Frayane, J. Colloid Interface Sci., 76 (1980) 107.
I731 D. Heath and Th.F. Tadros, Faraday Disc. Chem. Sot., 76 (1983) 203.
I741 C. Prestidge and Th.F. Tadros, Colloids and Surfaces, 31(1988) 325.
I751 Th.F. Tadros and A. Zsednai, Colloids and Surfaces, 49 (1990) 103.
I761 W. Liang, Th.F. Tadros and P.F. Luckham, J. Colloid Interface Sci., 155 (1993)
156.
I771 W. Liang, Th.F. Tadros and P.F. Luckham, J. Colloid Interface Sci., 160 (1993)
183.
I781 R. Ball and W.D. Brown, Personal communication.
I791 R. Buscall and P.D.A. Mill, J. Chem. Sot. Faraday Trans. I, 84 (1988) 4249.
I801 A.P. Gast, C.R. Hall and W.B. Russel, J. Colloid Interface Sci., 96 (1983) 251.
[811 W. Liang, Th.F. Tadros and P.F. Luckham, Langmuir, 9 (1993) 2077.
[821 Th. van ben Boomgaard, T.A. Ring, Th.F. Tadros, H. Tang and B. Vincent, J.
Colloid Interface Sci., 61(1978) 68.
1831 A. Zosel, Rheol. Acta, 21, 72 (1982).
[841 W.B. Russel, Powder Technol., 51, 15 (1987).
I851 R.C. Sonntag and W.B. Russel, J. Colloid Interface Sci., 116 (1987) 485.
1861 Wei-Heng Shih, Wan Y. Shih, Seong-11 Rim, Jun Liu and Ilhan, A. Aksay, Am.
Phys. Sot., 42 (1990) 4772.
1871 R. Buscall and P.D.A. Mills, J. Chem. Sot., Faraday Trans. I, 84 (1988) 4249.
I881 B.A. Frith, P.C. Neville and R.J. Hunter, J. Colloid Interface Sci., 49 (1974) 214.
I891 T.G.M. van de Ven and R.J. Hunter, J. Colloid Interface Sci., 68 (1979) 135.
200 Th.F. TadroslAdv. Colloid Interface Sci. 68 (1996) 97-200

[90] R.J. Hunter, Adv. Colloid Interface Sci., 17 (1982) 197.


[91] J.P. Friend and R.J. Hunter, J. Colloid Interface Sci., 37 (1971) 548.
[92] CF. Goodeve, Trans. Faraday Sot., 35 (1939) 342.
[93] T. Gillespie, J. Colloid Sci., 15 (1960) 219.
[94] R.J. Hunter, R. MaIarate and D.H. Napper, Colloids Surfaces, 7 (1983) 1
[95] T.G.M. van de Ven and R.J. Hunter, Rheol. Acta, 16 (1977) 534.
[96] J.D.F. Ramsay, J. Colloid Interface Sci., 109 (1986) 441.

You might also like