Belloni 2000

Home Search Collections Journals About Contact us My IOPscience
Colloidal interactions
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2000 J. Phys.: Condens. Matter 12 R549
(http://iopscience.iop.org/0953-8984/12/46/201)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 130.235.1.118
The article was downloaded on 11/10/2012 at 10:12
Please note that terms and conditions apply.

J. Phys.: Condens. Matter 12 (2000) R549–R587. Printed in the UK PII: S0953-8984(00)95363-3
REVIEW ARTICLE
Colloidal interactions
Luc Belloni
CEA/SACLAY, Service de Chimie Moléculaire, 91191-Gif-sur-Yvette Cedex, France
E-mail: [email protected]
Received 31 March 2000, in final form 27 September 2000
Abstract. The present topical review describes the concept of effective interactions between
spherical colloids in bulk solution. An introductive analysis based on the statistical mechanics
of liquids and mixtures derives the main classes of colloidal forces observed at various solvent–
solute compositions. A general survey of basic and advanced theories on ion-averaged interactions
between charged colloids ends with a critical examination of recent statements on the existence of
colloidal attraction and liquid–gas phase separation of pure electrostatic origin.
1. Introduction
As opposed to simple liquids, colloidal solutions are intrinsically complex systems containing
mesoscopic particles, the colloids, in the nanometre to micrometre size range, and small solvent
and solute molecules. The presence of solute components makes it possible to modify at will
and on a large scale the static and dynamic macroscopic properties of the whole solution. Many
organic/inorganic, natural, industrial and biological systems correspond to this schematic
picture. Micelles, mineral oxide particles, polymer latices, soluble proteins, clay platelets,
surfactant lamellae etc represent the colloid component. The stability of the solution with
respect to irreversible coagulation or reversible aggregation and phase transition as well as
its rheology are controlled by the addition of ions (in polar solvents), complexing molecules,
adsorbing or non-adsorbing polymers, . . . . Due to the large asymmetry in size, mass and time
scale between the colloidal particles and solvent/solute molecules, most of the experimental
techniques are directly sensitive only to the former. Examples are light and small angle x-
ray or neutron scattering and direct optical visualization. It is thus tempting to forget, at
least explicitly, about the small species and to consider the colloidal solution not as a mixture
but as a monodisperse system formally equivalent to a simple liquid, the colloids playing
the role of atoms. The central question in this one-component picture is the determination
of the effective interaction potential or force between colloids which implicitly accounts for
the solvent effects. The denomination ‘colloidal interactions’ refers to this indirect, solvent-
averaged quantity.
Any standard theory of simple liquids first assumes a basic, unambiguous form for
the potential of interaction between atoms or molecules (hard-core short range repulsion,
charge–charge and multipolar coulombic contribution, . . .), then solves the N -body statistical
mechanics problem. In this case, the central question is: for a given pair potential v(r) (energy
of interaction of an isolated pair in vacuum as a function of the separation r), what is the pair
distribution function g(r) (normalized probability to find a pair in the configuration r inside
the concentrated liquid) from which most of the equilibrium and structural properties can be
directly derived [1]? The situation in colloidal systems is somewhat reversed: one must first
0953-8984/00/460549+39$30.00 © 2000 IOP Publishing Ltd R549

R550 L Belloni
perform complex statistical averages, this time over the degrees of freedom of the solvent
(including solute) molecules, before deriving the effective colloidal interactions. When an
isolated spherical particle is immersed in the bulk, the solvent structure is locally perturbed
around it, up to a distance λ from its surface. When two such particles approach each other,
the modified solvent ‘shells’ begin to overlap and the local fluid structure becomes highly
directional. It is this spherical imbalance which results in an effective force F acting on
the colloids. The nature, the strength and even the sign of F are not obvious and depend
on the details of the solvent–solvent and solvent–colloid direct interactions. Assuming that
colloids in solution only interact by pairs, the effective or mean force pair potential v eff (r),
from which F is derived, then becomes the starting point in the standard, one-component
calculation of the colloidal correlations. This intermediate function v eff (r) is the key in the
powerful concept of colloidal interaction. Depending on the domain of interest, one focuses
on either the first or the second step in the averaging process. In the former case, the potential
is carefully determined as a function of solvent/solute composition and colloid nature and
dimension. (For large colloids of radius R λ, the interaction takes place only near contact.
In that regime of surface to surface separation h R, the surface curvature has a trivial effect:
according to the Derjaguin approximation, F /R is independent of R and can be expressed
as π E, where E is the interaction energy per unit area between two parallel, infinite, flat
colloidal surfaces separated by the distance h. Thus, most of the theories are developed in this
simpler planar geometry. For smaller colloids, the finite curvature must be taken explicitly
into account.) In the latter case, the richness of form in v eff (r) induces a wide spectrum of
structural and phase behaviour in concentrated colloidal fluids, much wider than in simple
liquids.
The situation becomes quite complicated if the colloids interact not only by pairs but also
by triplets, . . . . This happens when the solvent shell thickness λ is not negligible compared
to the dimension R and the mean distance between first colloid neighbours, as in highly
deionized charge-stabilized colloidal solutions (λ representing the Debye length). The overlap
of the spheres of interaction (radius ≈ R + λ) and the resulting colloidal interaction have a
multiple-body character and cannot be decomposed as a sum of pair contributions as before.
The N-body potential of mean force depends on the relative positions of all colloids inside the
solution. Although of formal interest, its practical use for illustrating the solvent effect and
calculating the colloid correlations is less trivial than in the pair-wise additive situation.
The experimental techniques used to determine colloidal interaction forces/potentials fall
in two complementary categories. For submicronic particles, the direct measurement of pair
interactions of two isolated colloids is not possible due to technical limitations in visualization,
separation control and force evaluation. On the other hand, the use of the thermodynamic
balance against the thermal energy kT (≈4 × 10−21 J at room temperature) and statistical
averages over N ≈ 1023 particles induce macroscopic behaviours from which interaction
curves can be indirectly extracted. The scattered intensity I (q) as a function of the scattering
vector q and the osmotic pressure are widely used quantities which allow quantitative
evaluations. Light and small angle x-ray and neutron scattering offer a q-range of 10−4 –
1 Å−1 which is ideally suited for the study of submicronic colloidal systems [2]. Due to
obvious experimental limitations (finite number of points in a restricted domain of scattering
angle, non-perfect precision of the data etc), the usual procedure consists of the following
steps: (i) assume pair-wise additivity for the colloidal interactions, (ii) start a priori with
a simple form for the effective potential v eff (r) including a few parameters (size, effective
charge, adhesiveness, . . .), (iii) solve accurately if not exactly the N -body problem of statistical
mechanics and (iv) compare the result with experimental spectra and adjust the parameters to
reach a reasonable agreement [3]. Concerning the osmotic pressure, the combination of various
Colloidal interactions R551
techniques covers the broad range 10−4 –10+9 Pa. Differences of absolute pressures between
solution and solvent reservoir through a membrane ( = 1–106 Pa), solvent vapour pressures
above the solution (105 –109 Pa) are measured directly or indirectly by equilibration with a
previously calibrated reference system [4–6]. Analysis of a single sedimentation (10−4 –1 Pa)
[7] or centrifugation (105 –107 Pa) [8] profile for heavy enough colloids also makes it possible
to determine the equation of state of the solution for a wide range of concentrations/pressures.
The theoretical interpretation of in terms of colloidal interactions requires the same statistical
treatment as before. A notable exception concerns lamellar systems where the osmotic pressure
is identified with the pair force per unit area (≡ − dE/dh) between infinite parallel plates.
The second category of experiments measures directly the force law between large enough
particles. With the surface force apparatus (SFA) [9] and atomic force microscope (AFM)
[10], two macroscopic curved surfaces (mean radius of curvature R) immersed in a liquid
are brought towards each other in a highly controlled way using piezoelectric crystals; the
separation distance h is measured through optical devices in the range 0–100 nm with a
precision of 1 Å; the force F is deduced from the deviation of cantilever springs of calibrated
spring constant. The ratio F /R is obtained in the range 10−6 –10−2 N m−1 . According
to the Derjaguin approximation, the resulting force law F /R(h) can be identified with the
corresponding law for large colloids immersed in a solution, at constant geometry, material
and solvent composition.
Modern techniques have been recently developed for direct measurement of colloidal pair
interactions in bulk. In paramagnetic oil droplet suspensions, the formation of aligned chains
of particles is observed under an applied magnetic field [11]. At equilibrium, the unknown
colloidal repulsion between two successive colloids is exactly balanced by a dipolar magnetic
attraction of known strength while the separation is measured through light diffraction.
With the total internal reflection microscopy (TIRM) [12], the fluctuating position of a
colloidal sphere (R 1 µm), subject to gravity, located above a glass plate, is analysed
in terms of the wall-sphere potential law. In other methods, video microscopy allows a direct
observation of colloidal configurations in bulk [13, 14] or near walls [15, 16]. The potential
of interaction is extracted from the measured pair distribution function [13, 15]. A related
technique catches with optical tweezers isolated pairs of colloidal particles [14, 16]. After
release of the external constraint, the relative Brownian motion of the interacting particles is
followed.
The main purpose of the present topical review is to convince non-specialists that the
variety of effective colloidal interactions can be understood in a systematic and quantitative
way using a simple albeit accurate description of the solvent/colloid correlations. The modern
general formalism of statistical mechanics of liquids and asymmetrical mixtures is recalled
in section 2 with a careful presentation of the notions of N -body potential of mean-force
and effective pair potential in the concentrated regime. Although most of the concepts are
valid for different geometries, the focus is on solutions of spherical particles. The specific
properties of elongated, rodlike or flexible colloidal systems are not addressed. Section 3
reviews the different classes of effective interaction between pairs of colloids expected in
neutral systems as a function of solvent–solvent and colloid–solvent affinities. Section 4 deals
with electrostatic interactions between charged colloids in concentrated solution. The general
survey includes advanced as well as basic, Debye–Hückel-like, theories. This somewhat
lengthy presentation leads to a critical examination of recent statements in the literature and
clarifies the important notion of ion-averaged attraction and liquid–gas phase separation of
pure electrostatic origin. Finally, section 5 evokes briefly the fundamental van der Waals
force as well as the miscellaneous interactions which result from complex couplings between
electrostatic, van der Waals and discrete solvent effects.
R552 L Belloni
2. Direct, mean force, effective potentials of interaction
The purpose of the present formalism is to transform solvent–colloid mixtures into colloidal
‘gases’. What are the effective, solvent-averaged interactions? To what extent do the text-book
concepts and formula of simple liquids [1], briefly recalled in 2.1, remain valid in the effective,
one-component approach?
2.1. Monodisperse systems

We consider N spherical particles in avolume V at the temperature T interacting via the
pairwise additive potential UN (r N ) = i<j v(rij ). v(r) represents the direct pair potential
between two isolated particles in vacuum and is, in principle, independent of T and of the
density ρ = N/V . The direct, pair force is f (r) = −v (r). Examples of v(r) are the
hard-sphere, soft-sphere, Lennard-Jones, square-well, . . . potentials.
In the canonical ensemble, the pair distribution function g(r) is given by

V2
g(r12 ) = exp[−βUN ] dr N −2 (1)
ZN

where β = 1/kT and ZN = exp[−βUN ] dr N is the partition function.
Virial and compressibility equations express the pressure P and the reduced
compressibility χ = (∂ρ/∂βP )T in terms of simple integrals of g(r):

ρ2
βP = ρ − g(r)βv (r)r dr (2)
6

χ = S(0) = 1 + ρ (g(r) − 1) dr . (3)
S(0) represents the infinite wavelength limit of the structure factor S(q).
It is always possible to write g(r) as
g(r) = exp[−βw(r)]. (4)
This defines the so-called pair potential of mean force w(r). Its name originates from the fact
that the force F (r) = −w (r) is the mean force felt by particles 1 and 2, fixed and separated by
the distance r, the average being taken over the configuration of the N − 2 remaining particles
3, 4, . . . :

(−∇1 UN ) exp[−βUN ] dr N −2
−∇1 w(r12 ) = = f1 (r1i ) = F1 (r12 ). (5)
exp[−βUN ] dr N −2 i=2,N
Like g(r), the mean force potential w(r) is of free energy nature. It depends on ρ and T and
implicitly contains many-body correlations. It is only at infinite dilution that the asymptotic
law g(r) = exp[−βv(r)] applies and that w(r) coincides with the direct pair potential v(r).
For the simple example of hard-sphere particles, v(r) is exactly zero for non-overlapping
spheres while w(r) presents an attractive well near contact and oscillations at intermediate
distances. This structure in w or in g is amplified as the density is increased and illustrates the
short-range liquid order which results from packing constraints.
2.2. Mixtures and McMillan–Mayer theory of solutions

We now investigate solutions containing solvent molecules and solute colloid particles. Such
mixtures are quite general: the solvent component may represent the organic or polar solvent
itself as well as small solute molecules (salt, polymer, micelle, . . .) which are added to the
solution in order to modify the colloid–colloid correlations. In the same way, the colloidal
sub-system could be composed of different types of particles. The main idea of any statistical
mechanical theory of solutions since McMillan–Mayer [17] consists in averaging over the
degrees of freedom of the solvent molecules and deriving solvent-averaged potentials of mean
force between solute particles. Note here that some particles like micelles in complex, mixed
colloidal suspensions can be considered as belonging either to the solvent components or to
the solute ones, depending on the degree of averaging and the level of analysis required.
To clarify the notation, we restrict ourselves here to binary mixtures composed of solvent
(s) and colloid (c) particles. Generalization to solutions containing many solvent and/or
solute species is straightforward. We assume that all particles interact via direct, bare,
state-independent, pair potentials, vss (r), vcs (r), vcc (r). The pair forces are represented
by fij (r) = −vij (r). The system contains Nc colloidal spheres. The colloidal density
Nc /V is denoted ρc . Rather than setting the solvent density ρs = Ns /V , it is better, both
conceptually and in practice, to set the solvent chemical potential µs or the solvent activity
zs = exp(βµs )/(3s . This is equivalent to setting the density ρs in a pure solvent reservoir which
is in osmotic equilibrium with the solution through a semi-permeable membrane, allowing
solvent molecules (but not colloids) to cross the membrane in order to equilibrate µs in both
compartments.
The different pair distribution functions gij (r) are expressed as statistical averages in this
semi-grand canonical ensemble. For example [1]:

V 2 zsNs
gcc (r12 ) = exp[−βUNs +Nc ] dr Ns dr Nc −2 (6)
) N s Ns !
where the semi-grand partition function is
z Ns
s
)= exp[−βUNs +Nc ] dr Ns dr Nc . (7)
Ns
N s!
The total energy U = Ucc + Ucs + Uss is the sum of all pair interactions in the colloid–solvent
mixture.
As for the monodisperse case, the definition
wcc = −kT ln(gcc ) (8)
introduces the colloid–colloid pair potential of mean force, the average being taken now on
the configurations of all solvent molecules and the Nc − 2 remaining colloids. wcc must be
viewed as a colloid + solvent-averaged pair potential.
Generalization of the virial (2) and compressibility (3) equations gives

1
βP = ρ − ρi ρj gij (r)βvij (r)r dr (9)
6 i,j
√ −1 √ρ i ρ j −1
ρi ρj −1
χ= Sij (0) = 1− ĉij (0) . (10)
i,j
ρ i,j
ρ
ρs must be understood as the averaged solvent density (which depends on ρc ) while ρ = ρc +ρs
represents the total density.
Sij−1 (q) is the element ij of the matrix S−1 (q), the inverse of the matrix S(q), composed
of the partial structure factors
√
Sij (q) = δij + ĥij (q) = δij + ρi ρj h̃ij (q). (11)
hij = gij − 1 is the total correlation function and ĥij its Fourier transform, normalized by the
√
density factor ρi ρj . In the same way, ĉij represents the normalized Fourier transform of the
R554 L Belloni
direct correlation function cij . Use of the Ornstein–Zernike (OZ) equation has been made in
the second equality of (10). This equation, which reads in matricial form,
S(1 − ĉ) = (1 + ĥ)(1 − ĉ) = 1 (12)
is the starting point of any theory based on integral equations. The exact closure to the OZ
equation can be formally written as [1]
gij (r) = exp[−βvij (r) + hij (r) − cij (r) + bij (r)] (13)
where bij represents the so-called bridge function. Neglecting bij defines the HNC integral
equation. Further linearization of the exponential factor beyond the contact (at distances larger
than the hard-sphere diameter σij ) leads to the MSA equation, cij (r) = −βvij (r); r > σij .
The McMillan–Mayer theory of solutions gives for the reduced osmotic compressibility
[18]
∂ρc
χosm = = Scc (0). (14)
∂β osm
represents the osmotic pressure, i.e. the difference between the absolute pressures in the
solution P and in the reservoir P . The thermodynamic derivative is expressed in the conditions
of osmotic equilibrium, namely at T and µs (or ρs ) constant. χosm depends explicitly only
on the colloid–colloid pair correlations, in contrast to the full compressibility χ (10), and is
formally equivalent to the compressibility (3) of the monodisperse gas of solute particles.
2.3. Potentials of mean force

In general, the colloidal particles are much larger and heavier than the solvent molecules. As
a consequence, most of the experimentally measurable quantities like the scattered intensity
are sensitive only to the correlations between colloids. It is thus tempting and fruitful to
formally regard the two-component mixture as a one-component system composed of colloids
interacting via an effective, solvent-averaged Nc -body potential of mean force WNc (r Nc )
[17, 19]. The idea of this approach is to decompose the full statistical averages of the mixture
in two successive steps [20, 21] (this decomposition is explicitly used in some numerical
simulations [22]): for a given configuration of the colloidal particles (r Nc ), the average over
the degrees of freedom of the solvent molecules is performed in the non-homogeneous space
imposed by the presence of the colloids, defining the solvent-averaged potential WNc (r Nc )
and the force felt by each colloid Fi (r Nc ) = −∇i WNc . Then, the statistical average over the
colloidal degrees of freedom is performed in the one-component system with the Boltzmann
factor exp(−βWNc ). Note that, in theory, as long as the static properties are concerned,
this one-component model (OCM) exactly reproduces the properties obtained with the two-
component approach, which treats in parallel and on an equal footing the colloid and the
solvent degrees of freedom (this would also be true for the dynamic properties but only in the
adiabatic limit where the solvent molecules relax instantaneously around the moving colloids).
In practice, this decomposition is a powerful way to exhibit fundamental concepts of interaction
and correlation in mixed systems and to derive simple, approximate but accurate expressions
for W used as inputs in the simple OCM. Different expressions for WNc (r Nc ) can be defined,
which differ only by an additive constant which is irrelevant for the colloid correlations or
forces but which is important for the thermodynamics of the solution.
Equating the semi-grand potential 1(T , V , µs , Nc ) = F − µs Ns = −P V + µc Nc (F is
the free energy) in the two- and one-component approaches gives the fundamental definition
[21, 23]
z Ns
s
exp[−βWNc ] = exp[−βUNs +Nc ] dr Ns = e−β1s exp[−β(Ucc + Ucs )]s (15)
Ns
N s !
where the first term in the right-hand side of (15) represents the partition function of the pure
solvent reservoir (of the same volume V ) with 1s = −P V and the second term (in brackets)
can be viewed as the mean Boltzmann factor exp[−βφNc ]s (with φ = Ucc + Ucs ) of a test
complex, Nc -body particle introduced randomly into the solvent system.
By construction, the thermodynamics and the spatial correlations (at any order) between
colloids in the OCM with the potential (15) are identical to those obtained in the original
two-component system. For example, the pair distribution function gcc (6) can be reproduced
with the OCM expression (1) with UN replaced by WNc . Note that WNc does not vanish in the
limit of very large inter-colloid distances. In other words, it contains familiar two-, three-, . . .
multi-body (colloid) terms as well as zero- and one-body terms which are important for the
thermodynamics of the solution [20, 21, 23]. Each term can be expressed as a diagrammatic
expansion in powers of zs . As far as the spatial correlations are concerned, the special zero- and
one-body terms can be discarded. The new definition, which differs from (15) by an additive
constant, independent of the colloidal positions, reads [24]
exp[−βφNc ]s
exp[−βWNc ] = . (16)
exp[−βφ1 ]N s
c
The term in brackets in the denominator represents the mean Boltzmann factor of a test isolated
colloid immersed in the solvent. At large inter-colloid distances, the mean Boltzmann factor in
the numerator can be factorized and the potential of mean force in (16) vanishes, as required.
At low colloidal density, the colloids interact mainly by pair and W can be restricted to the
sum of pair potentials W2 . In the limit of zero colloidal density, wcc in (8) coincides with W2
(we will see later how this simple law breaks down for salt-free charged colloidal systems).
At finite density, higher order terms become important and W is no longer pairwise additive.
From any of the two definitions for W , the mean force applied to a colloid i in the
configuration (r Nc ) can be decomposed as

Fi (r Nc ) = −∇i WNc = Fci + ρs (r )fsc (|r − ri |) dr . (17)
The first term is the direct force −∇i Ucc , i.e. the sum of pair forces fcc coming from the colloid
neighbours. The second term arises from the solvent particles whose local density inside the
inhomogeneous solution, around the colloids, is denoted by ρs (r ). In principle, expression
(17) for the mean forces does not bring more information than the potential of mean force W .
In practice, the free energy route and the force route are both useful and complementary to
fully understand the various couplings inside the solution.
Where the thermodynamics of the solution is concerned, the solvent contribution must be
taken into account through the zero- and one-body terms in the potential of mean force (15).
From the equivalent pressure point of view, the virial equation (9) of the solution becomes,
using the two successive steps in the averaging,

1 1 1
P = ρkT − ri ∇i UN = ρkT + ri · Fi + r · fs (r )ρs (r ) dr
3V i∈c,s s,c 3V i∈c c 3V c

1
+ |r − r |vss

(|r − r |)ρss (r , r ) dr dr (18)
6V c
while the OCM virial equation (2) would give

1
βP = ρc kT + ri · Fi . (19)
3V i∈c
Expression (18) for the absolute pressure contains the virial pressure of the solvent in the non-
homogeneous space characterized by (r Nc ). fs (r ) = −∇us (r Nc , r ) represents the external
R556 L Belloni

force due to all colloids, us (r Nc , r ) = i∈c vcs (|r − ri |) the external potential and ρss the two-
particle density. The osmotic pressure of the solution is rigorously given by the difference
between (18) and the corresponding expression for the pure solvent. To what extent is this
difference correctly recovered by the OCM expression (19)? This usually fruitful identification
of the colloidal ‘gas’ pressure (19) with the osmotic pressure of the solution is not exact
following the virial route and must be applied with caution. In particular, it fails to account
for the counter-ion contribution in charged colloidal systems (see section 4). This should
be opposed to the more universal OCM compressibility route, which always gives the exact
osmotic pressure through (14).
Closing this section, one could investigate the ideal case where the solvent molecules do
not interact with each other, vss = 0, but interact only with the colloids [21]. The (perfect
gas) pressure in the reservoir is P = ρs kT . Since the solvent molecules are uncorrelated, the
partition function of the solvent in the non-homogeneous fluid can be factorized. The final
result for the potential of mean force reads [21]
WNc = Ucc − P ϑ(r Nc ) (20)
where the interaction volume is
ϑ= e−βus (r Nc , r ) dr (21)
in the first definition (15), and is

Nc
ϑ = [e−βus (r ,r) − 1 − Nc (e−βus (r ,r) − 1)] dr
1
(22)
in the second definition (16). The local solvent density which monitors the mean forces (17)
is simply
Nc
ρs (r ) = ρs e−βus (r ,r )
. (23)
In the real case, expression (20) gives the correct behaviour only to first order in zs or ρs .
2.4. Effective, solvent-averaged, pair potential

In general, the Nc -body potential of mean force is not pairwise additive and depends on the
relative position of all colloids. This prevents a simple visualization and a clear understanding
of the solvent effects. For example, it is difficult to deduce, from the force felt by each colloid
in a given configuration, the extent to which the colloids ‘on average’ repel or attract one
another in solution. In order to overcome this difficulty, it is tempting to write formally WNc
as a sum of pair contributions

W N c (r N c ) = eff
vcc (rij ) (24)
i<j
eff
and introduce the so-called effective, solvent-averaged, colloid–colloid pair potential vcc (r).
Again, (24) is an approximation. A pair description alone is not able to reproduce the colloidal
forces and all colloid–colloid correlations resulting from the original full potential of mean
eff
force WNc , whatever the form of vcc (r). Nevertheless, it should be sufficient to reproduce
at least the pair correlations. This leads to the following rigorous definition which seems to
be the most natural and interesting one [25] and has been intensively applied with success in
charged systems since the early 1980s [26–29]:
eff
vcc (r) is the pair potential which would lead, within the OCM and the pairwise
additivity assumption, to the same colloid–colloid pair distribution function gcc (r)
as obtained within the two-component model, at identical temperature and colloidal
density. (25)
In practice, one must first solve the mixture problem characterized by the pair potentials vij (r)
or, equivalently, the OCM with the Nc -body potential of mean force. Then, the pairwise
additive OCM is solved in the reverse of the usual way, the (effective) pair potential being
extracted from the known pair correlations. Note that the uniqueness of the solution to this
inverse problem has been established [30]. The effective potential so defined is exactly the
potential extracted from standard fits of experimental scattering spectra I (q).
eff
In principle, there is no more information in vcc (r) than in gcc (r) or Scc (q). In contrast
eff
to WNc , vcc does not represent some intermediate function in a two-step averaging (one
must integrate over the colloidal degrees of freedom before deriving it). Nevertheless, in
practice, it is a powerful concept, which clearly illustrates how the solvent monitors the
eff
effective interaction between colloids. In particular, the sign of vcc indicates whether the
colloids feel a repulsion or an attraction from their neighbours, on average. Moreover,
eff
analytical, approximate albeit accurate expressions for vcc can be deduced in various
systems.
eff
vcc can be viewed as a solvent-averaged or semi-mean force pair potential. It differs
from W2 and depends on the colloidal density since it implicitly contains partially averaged
many-body interactions. By construction, it leads exactly to the same pair correlations as in
the original system but it is in general insufficient to give the correct n-body correlations of
(3)
higher order, gcc , . . . . Fortunately, this weakness does not affect the thermodynamics using the
compressibility route (14). In particular, if one approaches a liquid–gas spinodal line or critical
point in the phase diagram of the two-component system, χosm and Scc (0) present diverging
eff
values, which automatically imply the existence of attractions in vcc near this thermodynamic
state.
eff
We are now left with three kinds of colloid–colloid pair potential in the mixture, vcc , vcc
and wcc , which must not be confused. The two last depend on the thermodynamic state
eff
and differ by the degree of averaging. vcc is a solvent-averaged potential while wcc is the
solvent + colloid-averaged potential (of mean force). For example, in hard-sphere mixtures,
the former will present negative values near contact due to the osmotic pressure of the small
spheres against the two investigated colloids (this local osmotic pressure being perturbed by the
colloidal environment) while the latter will present negative values due to the osmotic pressure
of the solvent as well as colloid particles (two mixed depletion effects). In the absence of
eff
solvent, vcc and vcc coincide while wcc contains the usual one-component contributions of
eff
free-energy origin. In the opposite limit of zero colloidal density, vcc and wcc coincide (except,
again, for the important case of salt-free charged systems) and represent the interaction between
two isolated colloids immersed in the solvent.
In practice, the usual procedure for extracting the effective pair potential starts by mapping
the two-component OZ equation (12) onto a one-component one [25]. Identifying gcc in both
models gives for the OCM effective direct pair correlation function
eff ĉcs
2
ĉcc = ĉcc + . (26)
1 − ĉss
Once an integral equation within the OCM has been chosen, the effective potential is directly
eff
deduced from gcc (r) and ccc (r). In practice, the integral equations used in both models are
approximated and it is necessary and sufficient at this level of approximation to use the same
integral equation in the OCM as in the mixture [27–29] (this guarantees at least that direct and
effective potentials coincide in absence of solvent).
Comparing the diagrammatic expansions of the pair distribution function gcc in the two-
component mixture characterized by T , ρc , zs , vij and in the OCM characterized by T , ρc ,
eff
vcc , and identifying the terms of same order in ρc , gives the first diagrams listed in figure 1
R558 L Belloni
eff
Figure 1. First diagrams in the expansion of the colloid–colloid effective pair potential vcc (r).
The small black symbols are field solvent particles, weighted with solvent activity zs . The
solid lines represent
Mayer function links fij = exp(−βvij ) − 1. At this level of expansion,
ρs /zs = 1 + ρc fcs (r) dr. In the last diagram, the large black symbol is a field colloid particle
and the dotted lines represent 1 + fcc = exp(−βvcc ) links.
eff
for the effective potential vcc . Note the presence of diagrams involving colloid neighbours
which illustrates the ρc -dependence of this effective pair potential.
The leading, three-body (2 colloids+1 solvent) term is given by (20),(22) and corresponds
to the first diagram in figure 1:

eff
vcc (12) = vcc (12) − kT ρs (e−βvcs (13) − 1)(e−βvcs (23) − 1) dr3 . (27)
The corresponding expression using the force route (17) is

eff
−vcc (12) = fcc (12) + ρs fcs (13) e−β(vcs (13)+vcs (23)) dr3 . (28)
The integrals in (27) and (28) represent the first effect of the solvent molecules on the
potential/force between colloids. They can be derived and understood in an intuitive way:
the potential expression represents the difference in free energy of the inhomogeneous solvent
subsystem between configurations characterized by r12 and by r12 = ∞. The force expression
is even more obvious and represents the integrated force coming from the solvent molecules
which are located around the two colloids. Anticipating the discussion in the next sections,
one already notes that both cs repulsion and attraction lead at this level of approximation to an
effective attraction between colloids. In the former case, there are less solvent molecules in the
interstitial region between the colloids than in the opposite directions since, there, the solvent
is repelled by the two colloids at the same time. This asymmetry in solvent population results
in a force which pushes the colloids towards each other, namely in an effective attraction of
depletion origin. The analogous discussion in terms of potential is related to the convexity of
the exponential function: when the colloids are close to each other, the solvent molecules are
less frustrated by the repulsion coming from the colloids (compared to the large separation
configuration), their partition function in the inhomogeneous system is higher and their free
energy is lower, which means that the configuration is more favourable and the effective
potential is negative. In the latter case of cs attraction, the situation is reversed: more solvent
molecules accumulate in the interstitial region (simultaneous attraction from both colloids),
which results in an effective attraction between colloids, this time of bridging origin.
The leading diagram (27) is sufficient to understand the interaction between colloids at low
solvent density as it is sufficient to give the third virial coefficient in monodisperse systems.
At higher solvent activity, many body correlations take place, more diagrams must be included
and the situation becomes less obvious and requires careful analysis.
3. Colloidal interactions in neutral systems
3.1. Hard-sphere mixtures: depletion and structural forces

The simplest example of a colloid/solvent system is the hard-sphere mixture (diameters σs and
σc ), characterized by the pair potentials
βvij (r) = +∞ r < σij = (σi + σj )/2
=0 r > σij . (29)
This athermal system is governed by correlations of entropic (excluded volume) origin. The
main effect, the depletion attraction between large balls due to the osmotic pressure exerted
by the small ones, was first introduced by Asakura and Oosawa [31]. In the last few years, the
hard-sphere mixtures have been the subject of beautiful experimental and theoretical studies,
focusing on the force acting on the colloids as well as on the phase separation which could
result from the depletion attraction (see [32], [23] and references therein). Experiments mainly
concern mixtures of large neutral colloids and short non-adsorbing polymers immersed in theta
solvent. The polymer coils are seen as small spheres, more or less penetrable one to each other.
In the pure-hard sphere mixture, the particles feel each other only during collisions (at
contact). The force exerted by a solvent sphere on a colloid can be written as kT δ(r − σcs ).
Consequently, the volume integral which gives the solvent force acting on each colloid (17)
reduces here to a surface integral [33]

Fi (r Nc ) = − ρs (S )kT dS . (30)
Si
Si represents the sphere centred on ri , of diameter σcs . The surface vector dS points outwards
this sphere. Due to the presence of the colloidal neighbours around i, the solvent density
at contact ρs (S ) is not spherically symmetric. As a consequence, the unbalanced osmotic
pressure of the solvent molecules ρs (S )kT against the ith sphere results in a net force Fi .
To first order in zs or ρs (or in the ideal case where the solvent molecules do not feel each
other, σss = 0), the local density ρs (r ) coincides with ρs in the interstitial regions accessible
to the small balls and is zero elsewhere [21]. The surface force reduces to

Fi (r Nc ) = −ρs kT dS . (31)
Si ‘f ree’
Si ‘free’ represents the portion of surface Si accessible to the solvent molecules (not excluded
by the volume of the colloid neighbours). Equivalently, the Nc -body potential of mean force
reduces in the same limit to
WNc = −ρs kT ϑf ree (r Nc ) (32)
where the interaction volume coincides with the real free volume offered to the solvent
molecules in the fixed (r Nc ) configuration (minus the free volume at infinite separations).
The origin of the depletion attraction is now clear (figure 2(a)): when two colloids 1 and 2
approach each other at r12 < 2σcs = σc + σs , the two excluded volumes V1 and V2 , of radius
σcs , forbidden to the solvent, begin to overlap, which means that the free volume increases
by the quantity V1 ∩ V2 , and the entropic pair potential W2 is equal to −ρs kT V1 ∩ V2 . Using
the equivalent force picture, the presence of particle 2 in the vicinity of particle 1 prevents the
solvent molecules from going between 1 and 2. The existence of this depletion region breaks
R560 L Belloni
V2
V1
F
a
1 2
V1∩V2
V2
V1
1 2 b
Vs
V2∩Vs
V1∩Vs
Figure 2. Schematic representation of excluded volume correlations in asymmetrical hard-sphere

mixtures. The dotted zones are forbidden regions for the solvent particles, the solid arrows indicate
local solvent osmotic pressure forces, the dashed arrows represent effective forces. (a) Bare, first
order depletion attraction. (b) Second order structural repulsion: increase of the solvent density
near colloidal surfaces enhanced in the interstitial region (see text).
the osmotic pressure balance around 1 and the solvent molecules located in the direction
opposite to 2 push particle 1 against particle 2.
For two colloids in contact (r12 = σc ), the first diagram gives βW2 = −φs (1+ 23 σc /σs ) and
βF = −(π/4)ρs σs (2σc +σs ) [34]. At constant volume fraction occupied by the solvent φs , the
depletion attraction becomes shorter range and deeper at contact when the solvent molecules
become smaller [35].
In general, the free volume accessible to the solvent is not pair-wise additive. For
example, in the presence of three approaching colloids 1, 2 and 3, the gain in free
volume (compared to the infinite separation situation) is V1 + V2 + V3 − V1 ∪ V2 ∪ V3
≡ V1 ∩ V2 + V1 ∩ V3 + V2 ∩ V3 − V1 ∩ V2 ∩ V3 and, due to the extra, three-body term, is

smaller than the sum of the previous two-body volumes [36]. The existence of this three-
body repulsion indicates that the effective depletion pair attraction decreases in general
with increasing volume fraction (this shows up in the repulsive last diagram of√ figure 1,
involving three colloids). Meanwhile, for large size ratio σc /σs (larger than 2 3 + 3 ≈
6.5), the intersection volume V1 ∩ V2 ∩ V3 is empty, even for three colloids in contact,
and the depletion attraction is additive by pairs, at least to first order in the solvent
density.
When the volume fraction of the solvent sub-system becomes appreciable, the bare
depletion picture is not sufficient to describe the colloid interactions and diagrams of higher
order, involving 2, 3, . . . solvent molecules must be taken into account [37, 38]. The solvent-
solvent excluded volume (σss = 0) induces a short-range liquid order in the solvent liquid
which is perturbed near a colloidal surface. The main effect is the appearance of an ‘adsorbed’
solvent layer (of higher density than the bulk solvent density) around each colloid. The origin
of this layer is still entropic and can be explained as a depletion effect, this time between one
colloid and one solvent sphere. Following the same reasoning as before (see figure 2(b) for
a schematic picture), when one particular solvent sphere s is located near a colloidal wall 1
(r < σcs + σs ), a depletion region appears between them which is excluded to the solvent
neighbours, resulting in a net osmotic force which pushes the solvent sphere against the wall.
In other words, the free volume accessible to the solvent liquid is increased when the excluded
spherical volumes Vs and V1 of radius σs and σcs centred on the solvent and colloid positions
respectively, begin to overlap. The induced solvent layer has a thickness of σs . As a first effect
of the extra-solvent density, the bare depletion attraction between touching colloids increases.
More importantly and subtly, this accumulation of solvent near a surface is enhanced in the
interstitial region between two approaching colloids 1 and 2 (r12 < σc + 2σs ). Indeed, within
this region, the solvent sphere maximizes the gain in free volume for its solvent neighbours,
Vs ∩ V1 + Vs ∩ V2 . Or, using the pressure picture, the solvent sphere located at the entry of
the ‘funnel’ feels a net osmotic force along the mid-plane between 1 and 2 which pushes it
in the direction of the exit of the funnel. Both routes indicate that the local solvent density
around the two colloids is higher between them than in the opposite direction. According to
(30) for the mean force, this results in a repulsion between colloids which adds to the bare
depletion attraction (in other words, when two colloids approach each other, their adsorbed
solvent layers begin to overlap the hard cores and the partial destruction of them costs in free
energy). At separations σc + σs < r12 < σc + 2σs , the bare depletion vanishes and the new
repulsion dominates. This effect of order ρs2 corresponds to the diagrams of the second line
in figure 1.
The previous analysis helps us to understand why the interaction between large balls in
an ocean of small ones presents at high volume fraction a depletion attraction near contact
followed by oscillations of period σs between domains of repulsion and attraction (figure 3).
This behaviour, similar to what happens in monodisperse hard-sphere systems, is the signature
of the liquid order of the underlying fluid and of its alteration by the presence of one or two
colloids in the neighbourhood.
Direct measurements of hard-sphere depletion pair interactions have been performed
between curved mica surfaces in non-aqueous liquids [39] (SFA) and between polystyrene
colloids and flat glass surfaces in nonionic polymer solutions [40] (TIRM). Indirect
experimental evidence for the depletion is generally deduced from the stability of hard-sphere
mixtures with respect to phase separation [41]. Concerning the theoretical prediction of phase
diagrams, one must keep in mind that the strength of the depletion attraction determined at
zero colloidal density is in principle not sufficient to determine the precise location or even
R562 L Belloni
eff
Figure 3. Effective vcc (r) and mean force wcc (r) pair potentials between colloids in pure hard-
sphere mixtures: depletion + structural forces. σc /σs = 5. The curves are labelled with their
volume fraction composition 6s , 6c = 4%, 0%; 15%, 0%; 11.9%, 15%. The last two cases
correspond to the same solvent activity or solvent density in the reservoir 6s = 15%. BHP integral
equation [123] in the two-component and effective one-component models.
to guarantee the existence of the liquid–gas or fluid–crystal transition. Indeed, the effective
pair potential becomes less attractive in general for increasing colloidal volume fraction due
to the 3, 4, . . .-colloid contributions in the Nc -body potential of mean force (figure 3). At
high size ratio, the non-additivity character appears only at high solvent volume fraction
(the solvent molecules located in the gap between colloids 1 and 2 and those located in the
gap between colloids 2 and 3 are coupled only in diagrams of high order, involving many
solvent balls) and the pairwise additivity approximation should be valid at lower solvent
density. The reader interested by this issue is referred to a recent simulation work [23],
which unambiguously proves the existence of additive hard-sphere phase separation in this
regime.
Starting from the reference case of hard-sphere mixtures, we will now investigate step by
step the presence of extra, soft interactions between the various components. Four basic cases
can be distinguished.
3.2. Colloid–solvent repulsion: enhanced, longer-range, depletion

A soft repulsion, of range d σs , between colloid and solvent particles is added to the previous
hard-sphere interactions. As a consequence, a depletion layer of solvent develops around each
colloid (negative adsorption) which leads, following the same mechanism as before, to an
attraction between colloids. The interaction volume (22) cannot be simply identified with
a real free volume now but the resulting interaction is very similar to that obtained in the
pure hard-sphere case. Roughly speaking, the hard-spheres results can be applied with the
colloid–solvent contact distance σcs replaced by a renormalized value of the order σc /2 + d
(the case of pure hard-sphere mixtures with positive nonadditivity in diameter has recently been
considered in [42]). Thus, the enhanced depletion may be important even at low hard-sphere
volume fraction φs .
This simple analysis allow us to explain why two hydrophobic surfaces, carrying groups
which repel water molecules, attract each other [43]. The strength of this hydrophobic force
may be sensitive to the temperature [44]. Its range, linked to the range of the colloid–solvent
repulsion, is generally limited to a few nanometres in surface–surface separation.
Another illuminating application can be found in mixtures of large and small charged
objects of same sign, the small component playing the role of the solvent. The particles interact
via (ion-averaged) screened coulombic repulsions, which decay as exp(−κr)/r. The colloid–
solvent repulsion induces a deep attraction between colloids, which adds to and sometimes
dominates the direct repulsion [37]. This depletion of electrostatic origin has been observed
in various systems like mixtures of SDS micelles and large oil droplets stabilized by a SDS
surfactant layer [45, 11] (optical observation), crossed mica cylinders coated with an adsorbed
bilayer of CTAB and immersed in a CTAB micellar solution [46] (SFA experiments) and
polystyrene particle near a glass plate in presence of silica particles or charged polymers [47]
(TIRM). The depth of the attraction is monitored by the depletant concentration (in ρs to
leading order) while the range is related to the large Debye length κ −1 .
In the same way, the coulombic repulsion between large and small colloids in highly
deionized bidisperse charged latex suspensions induces a ‘charge ordering’ even at low volume
fraction which can be observed in x-ray or neutron scattering experiments [48, 49]. The contrast
match technique makes it possible to determine the pair correlations among particles of one
of the two components (viewed here as the ‘colloid’ component). The partial intensity curves
can be interpreted in terms of effective pair potentials (25) which present a clear electrostatic
depletion due to the hidden particles of the other (‘solvent’) component [49].
We next proceed with a slight digression. It happens in some charged colloidal systems
that, despite intensive cleaning through dialysis or ion exchange resin, small charged pieces
of material remain in the solution between the colloids. An important example is constituted
by latex systems where the surface of the lattices may release charged polymers of low mass
in the bulk [6]. These new particles are not small enough to cross a dialysis membrane.
Moreover, since the release seems to result from an equilibrium between surface and bulk, the
solution will always contain the extra component even after careful cleaning. The polymers
are negligible with respect to the mass fraction of colloids. They do not contribute (at least
directly) to any scattered intensity and cannot be seen by optical visualization. On the other
hand, they contribute to and even dominate the total number of particles as can be observed in
osmotic pressure experiments [6]. More importantly, in highly deionized solutions, the ionic
strength is kept very low, the Debye length is large, much larger than the real size of the small
molecules, and the effective, electrostatic, volume fraction of the charged polymers may be
large, inducing a strong depletion attraction between colloids. This effect could explain [50]
non-classical behaviours observed in highly deionised monodisperse latex systems [51].
3.3. Solvent–solvent repulsion: oscillating structural force

If the solvent molecules repel each other through a soft repulsion of range d σs , they tend
to accumulate near inert colloidal surfaces and a repulsion at contact followed by oscillations
will appear at high solvent density in the colloid–colloid effective potential. This structural
R564 L Belloni
force (in ρs2 to leading order) is similar to that described in hard-sphere mixtures although
the softness of the ss repulsion somewhat damps the oscillations of period d. It could be
clearly identified in mixtures of large neutral colloids and small charged micelles where the
absence of colloidal charge guarantees the absence of direct repulsion and indirect depletion
attraction between colloids. When the colloids become charged (again of the same sign as
the solvent), all effects add up and the resulting interaction between colloids presents complex
structures which depend on the size and charge ratios and on the solvent density, as seen in the
experiments mentioned above [45–49].
Before closing sections 3.2–3.3 on depletion and anticipating the analysis of charged
systems in section 4, it is important to note here that the description of charged solvent effects
solely in terms of electrostatic depletion attraction between colloids must be used with caution
at low ionic strength. In that regime, the effect of the solvent particles on the colloid-colloid
effective interaction is less of depletion than of screening nature. In practice, this means that the
diagrammatic or virial expansion is only slowly convergent and keeping only the first leading
terms could lead to predicting a strong attraction where the correct ion + solvent-averaged
interaction is purely repulsive (see the discussion at the end of 4.1).
3.4. Colloid–solvent attraction: steric, stabilizing repulsion + bridging attraction

What happens when the solvent molecules present an affinity for the colloidal surfaces? The
direct cs potential is now the sum of the hard-sphere repulsion and of a soft attraction.
As a consequence, the leading expression for the effective cc potential will contain three
contributions, (i) the bare hard-sphere depletion attraction, (ii) a steric repulsion and
(iii) a bridging attraction at larger distances. The physical picture is trivial: when the cs
attraction is turned on, an adsorbed solvent layer will appear around each colloid. When
two particles approach each other, the layers first begin to overlap. The local solvent density
increases, which induces a bridging attraction (ii) (note that if the adhesion attraction is highly
directional, the solvent molecules in the gap cannot stick to two surfaces at the same time and
the bridging effect disappears). The separation still decreasing, the layers are now perturbed
by the internal hard cores and are partially destroyed. The intersticial density decreases by
volume exclusion, there is a loss of attractive energy (cost in free energy) and this frustration
induces a so-called ‘steric’ repulsion at short distance (iii) which rapidly overcomes the bare
depletion (i), even at low affinity (see figure 4).
This kind of repulsion is the basis of any steric stabilization of colloidal suspensions due
to the presence of adsorbing molecules [9]. In the limit of strong binding, the adsorbed layer
can be viewed as an additional hard core which prevents the colloids from coming into true
contact. Examples of stabilizing agents are (adsorbing) polymers, small organic complexant
molecules or some organic or polar solvents. In the important case of water as a solvent,
hydrophilic surfaces are stabilized through the so-called hydration force. Before going into
detail in these examples, it is first necessary to complete the study with the last basic case of
direct interaction.
3.5. Solvent–solvent attraction: modulation of steric and hydration forces

In the case of solvent–solvent attractions of range d, the frustration felt by the solvent in the
vicinity of passive hard-core volumes induces a depletion layer around colloids. This depletion,
enhanced in the interstitial regions, induces as before an effective attraction between colloids
(figure 4). This means that passive surfaces behave as solvophobic when immersed in a solvent
with ss attraction. The range of the depletion attraction is usually of the order of d, except
eff
Figure 4. Effective colloid–colloid pair potential vcc (r) in hard-sphere mixtures with additional
colloid–solvent or solvent–solvent direct attraction: steric + bridging or solvophobic forces.
σc /σs = 5, 6s = 4%, 6c = 0. βvcs (r) and βvss (r) are Yukawa potentials of range d = σs /2 and
depth at contact Kcs and Kss , respectively. BHP integral equation.
in the limit of a solvent very close to its liquid–gas spinodal line or critical point where it is
related to the large correlation length of the fluid [52].
It is now possible to analyse the complex behaviours observed in real suspensions which
combine affinity of the solvent for the surfaces and solvent–solvent attraction or repulsion.
For macroscopic surfaces or micrometric colloids, steric stabilization involves most often
adsorbing polymers [53]. The steric layer is due to irreversible grafting or reversible physi-
or chemisorption. The polymer-average potential between colloids and the stability of the
solution will depend on the strength of the adhesion, on the temperature or on the quality of
the real solvent for the polymer (steric, entropic repulsion enhanced in good solvent, attraction
in bad solvent) and on the degree of coverage (bridging attraction at intermediate value) [9].
Although the precise form of the interaction involves the polymeric nature of the molecules,
the simple analysis described above is usually sufficient to qualitatively explain most of the
behaviours.
For nanometric colloids, polymers are less adapted due to the formation of bridging
necklaces [54] and to the important volume fraction occupied by the thick adsorbed
‘layers’ [55]. A better candidate for steric stabilization involves small complexant molecules.
The complexing shell plays the role of a ‘double sided tape’: the inner side is strongly adhesive
to bind to the colloidal surface and blocks its reactivity, while the outer side may induce a weak
attraction between covered stabilized particles [56].
The short-range, strong hydration repulsive force observed in colloidal and biological
aqueous solutions results from complex couplings between water–surface and water–
R566 L Belloni
water interactions (see the recent review [57]). The presence of dipole, quadrupole, . . . ,
polarizability in the water molecule induces highly directional (hydrogen bonds) water–
water attractions favouring tetrahedral icelike configurations. A surface can be considered as
hydrophilic only if ‘it likes water more than water likes water’. This is the case of materials like
silica or lecithin or surfaces carrying hydrophilic groups (H-bonding groups, bound ions, . . .).
The strong binding of H2 O molecules (surface hydration) guarantees a strong surface–surface
stabilizing hydration repulsion [43]. On the other hand, the cost in free energy (energy minus
orientation entropy) needed to disrupt the water network near passive surfaces leads to a
depletion of H2 O molecules (dewetting of the surface) and to a surface-surface hydrophobic
attraction. The effect is amplified if the surfaces carry hydrophobic groups (hydrocarbons,
fluorocarbons, . . .). The hydration force has been measured in numerous osmotic pressure
[58] or SFA experiments [9]. The data are usually adjusted with an exponentially decreasing
function of the surface–surface separation, the characteristic distance being of the order of
1–10 Å. From the theoretical point of view, refined discrete models of water fluids have been
studied in bulk or near various surfaces [59–61]. The different density profiles vary rapidly on
fractions of ångstroms due to the complex, non-spherical nature of the direct potentials. The
resulting force between surfaces presents a rich structure which strongly depends on the details
of the coulombic interaction parameters. In particular, the hydration force and the associated
stability seems to be very sensitive to the presence of ions and may vary drastically when
‘specific’ ion–solvent and ion–surface interactions are introduced [62]. The debate about the
quantitative explanation of the measured hydration force and about the precise role of the water
anisotropy on the apparent exponential behaviour is still open [57, 63].
4. Colloidal interactions in charged systems
We now focus on solutions of charged colloids and small ions. Within the so-called Primitive
Model (PM), the discrete nature of the polar solvent is forgotten and mixtures of charged hard
spheres (diameter σi , charge valence Zi , concentration ρi ) immersed in a continuous solvent of
dielectric constant ε are considered [19]. The PM pair potentials (solvent-averaged potentials
of mean force) are
βvij (r) = +∞ r < σij
= Zi Zj LB /r r > σij (33)
where the Bjerrum length LB = e2 /(4πε0 εkT ) is the single parameter which takes the solvent
implicitly into account. LB ≈ 7 Å in aqueous solutions at room temperature. It is implicitly
assumed in (33) that the interior of the charged particles has the same dielectric constant than
the solvent (no electrostatic images).
In such mixtures viewed as asymmetrical electrolytes, the colloidal spheres c are much
larger and more highly charged than the simple ions. Despite its crude, ‘primitive’-treatment
of the solvent (neglect of van der Waals and hydration forces), the PM contains most of the
characteristics of charged systems: (i) the infinite range of the 1/r coulombic potentials,
(ii) the simultaneous presence
of positive and negative charges in order to guarantee the
electroneutrality condition ρi Zi = 0, (iii) the coupling between ++, −− repulsions and
+− attractions. As before, the aim is to find out the effective, ion-averaged potential of mean
force between colloids, in a description where the small ions play the role of the solvent in
the general analysis. In contrast to what happened in neutral systems, the treatment in terms
of virial expansion is now useless because of the points (i)–(iii) and special approaches are
required.
Since the 1910s, the problem of charged systems has been attacked from different
directions, at different levels of approximation. The concept of electrostatic screening
introduced by Gouy, Chapman and Hückel in the 1910–20s, led to the famous DLVO screened
coulombic, effective potential between colloids in the 1940s [64]. Since the 1960s, numerous
studies of the Poisson–Boltzmann (PB) approach have derived the notion of ionic condensation
and effective colloidal charge, reminiscent of the Bjerrum concept of ionic association in the
1920s. The approach based on the numerical resolution of integral equations (especially
HNC-like closures) was developed for the PM of colloids in the early 1980s.
Concerning the ‘exact’, reference treatment based on numerical Monte Carlo (MC) or
molecular dynamics (MD) simulation, which could serve as tests of the validity of approximate
theories, only a few studies of bulk charged colloidal systems, with size and charge asymmetry
in the salt-free micellar regime, have been performed since the early 1980s [65–69]. The
relative lack of PM simulation data illustrates the technical difficulties related to the three
points described above: (i) the long-range nature of the potentials implies the heavy, time-
consuming calculation of Ewald sums at each colloid-ion configuration, (ii) the total number
of particles rapidly becomes large as the charge asymmetry increases since even a reasonable
number of colloids must be accompanied by a large number of counter-ions in order to maintain
the electroneutrality of the system and (iii) last but not least, the accumulation of counter-ions
in the vicinity of the colloidal surfaces implies the use of very small MC displacements or short
MD time steps in order to avoid colloid–counter-ion overlaps and thus an astronomic number of
steps to get a good statistic in the correlations (investigation of colloid–ion cluster moves seems
to be a good direction to follow in order to overcome this last difficulty [70]). These reasons
explain why only charge asymmetries of 1/−10 to 2/−20 have been published in the literature
in the last 20 years [65–67, 69] and why, as a consequence, unusual behaviours derived by
approximate theories at higher charge cannot be validated without ambiguity. Although very
recent publications support some optimism [68–71], one must admit that, even half a century
after DLVO, bulk colloidal systems are still not yet ‘exactly’ understood at high electrostatic
coupling, quantitatively as well as qualitatively.
While PM simulation is difficult in bulk, it becomes easier and has been intensively used
since the early 1980s in the fundamental, special case of two parallel, charged walls separated
by a simple electrolyte [72]. In that case, there are only two fixed colloidal surfaces (two
interacting double layers) and the previous difficulties disappear. The data are analysed in
pressure versus wall–wall separation curves. Thank to these complete simulation studies,
one could say that the behaviour of this particular geometry is exactly known since the
1980s.
In the last few years, one observes in the literature a return to the problem of charged
colloidal systems. Recent analyses based on modern theories like the density functional theory
(DFT) have brought fruitful points of view. They have also led to a few controversies, especially
about the fascinating issue of effective, purely electrostatic attraction between like-charged
colloids.
The purpose of this section is (i) to recall the correct equilibrium and structural properties
of the PM through well established but sometimes mis-appreciated approaches; (ii) to clarify
the controversies and discuss the new analyses in view of the traditional theories and (iii) to
describe the traps to avoid in the understanding of salt-free colloidal systems and criticize
statements which violate subtle but fundamental notions of electrostatics. For such purpose,
I have to begin at the trivial but sometimes forgotten bulk Debye–Hückel (DH) level. The
specialists could skip this introduction.
In the following, we will investigate cationic colloids (Zc > 0) of density ρc and
monovalent counter-ions Z− = −1 mixed with the monovalent +1/−1 salt of same counter-
R568 L Belloni
ion. Generalization to mixed valences is straightforward. If ρs and ρs represent the salt
concentration inside the solution and the salt reservoir, respectively, the ionic concentrations
in the three component system are imposed by the electroneutrality condition:
ρ− = Zc ρc + ρs
ρ+ = ρs
while ρ− = ρ+ = ρs in the reservoir. Ions can cross the semi-permeable membrane only
by neutral pairs. The salinity in the solution is identified with the co-ion concentration. It is
in general lower than the salinity imposed in the reservoir due to the Donnan, salt exclusion
effect. The equilibrium condition which sets the value of ρs is the equality of the chemical
potential of salt in both compartments:
µs = µ+ + µ− = µs = µ+ + µ− . (34)
The net osmotic pressure of the solution (against the electrolyte reservoir) is the difference
between the pressures of the solution P and of the reservoir P (within the PM, P and P
represent the pressures of the gas of charged particles; experimentally, there are themselves
osmotic pressures against pure water). The thermodynamic derivative (14) becomes
∂ρc
χosm = = Scc (0). (35)
∂β ρs
In the important salt-free case, ρs = ρs = 0, the two-component system colloid + counter-ion
verifies the local electroneutrality conditions:
Zc
χ = S−− (0) = Zc Sc− (0) = Zc Scc (0). (36)
Zc + 1
4.1. Debye–Hückel theory

In principle, the DH approximation gives the exact asymptotic (infinite dilution) law. Although
its domain of validity for colloids is in practice restricted to unrealistically high dilutions, its
analysis is useful to illustrate some fundamental concepts.
Within the bare DH approximation, all charged species, ions as well as colloids, are
considered as pointlike ions (σi = 0) and the correlations are treated at a linearized mean-field
level. Inserting the MSA direct correlation functions, cij (r) = −Zi Zj LB /r, r > 0, in the
PM–OZ equation (12) yields for the total correlations
Zi Zj LB −κ0 r
gij (r) = 1 − e (37)
r
where the total screening constant κ0 involves the contribution of the ions and the colloids:
κ0 = 4πLB [2ρs + Zc (Zc + 1)ρc ]. (38)
The PM virial (9) and compressibility (10) routes give different expressions for the pressure
of the solution:
βPc = ρ = 2ρs + (Zc + 1)ρc (39)
κ03
βPv = ρ − . (40)
24π
This thermodynamic inconsistency illustrates the approximate nature of the DH theory. While
the compressibility pressure (39) is reduced to the perfect gas law, the virial route (40) is richer
since it exhibits the first non-ideality correction in the equation of state. This behaviour is
common to all DH-like theories. Note that the extra term in (40) is negative and could induce,
in principle, a van der Waals loop in isotherm P (ρ) curves and a liquid–gas phase transition
in the phase diagram [73]. In fact, this would happen in concentration–charge regimes where
the bare DH equation of state (40) breaks down. For symmetrical electrolytes (restricted PM),
more analytical terms could be added to give reasonable predictions [74] compared to exact
MC phase diagrams [75]. For colloidal systems, this direction is less powerful [28].
Corresponding expressions are obtained for the pressure P in the reservoir. Using the
compressibility route, the net osmotic pressure is given by
β = ρc (1 + Zc ) + 2(ρs − ρs ). (41)
The relation between the values of salinity in the two compartments is
ρs (ρs + Zc ρc ) = ρs2 . (42)
This means that the DH ionic chemical potentials µi are identified with their ideal expressions
kT ln ρi (again within the compressibility route). Note that the resulting expression (41) for
could be directly obtained from the osmotic derivative (35).
Two important limited cases can be investigated. In excess of salt or at low colloidal
concentration (namely, Zc ρc ρs ):
Z c ρc
ρs ≈ ρs −
2
Zc2 ρc2
β ≈ ρc + . (43)
4ρs
This means that when one colloid with its Zc counter-ions is inserted inside the solution, Zc /2
pairs of salt ions leave the solution through the membrane. The net pressure is increased
by only one kT /V . On the other hand, for salt-free cases (Zc ρc ρs ), the salinity inside
the solution is negligibly small (ρs ρs ) and all counter-ions contribute to the perfect gas
pressure:
β = ρc (1 + Zc ). (44)
According to (37), the cc pair potential of mean force βwcc = − ln(gcc ) ≈ 1 − gcc takes the
screened coulombic, repulsive form
Zc2 LB −κ0 r
βwcc (r) = e . (45)
r
eff
On the other hand, the cc effective pair potential vcc extracted from the OCM–OZ equation (26)
coupled with the MSA closure reads
eff Zc2 LB −κr
βvcc (r) = e (46)
r
where the screening constant κ now involves only the ionic contributions:
κ= 4πLB [2ρs + Zc ρc ]. (47)
The fundamental difference in nature between these two potentials is illustrated in (45) and (46).
The colloids contribute to the electrostatic screening (κ02 = κ 2 +κc2 with κc2 = 4π LB ρc Zc2 ) only
eff
in the ion + colloid-averaged potential wcc and not in the ion-averaged potential vcc . In the
presence of salt, the two quantities coincide at zero colloidal concentration, the ionic strength
being dominated by the salt ions. This case follows the general behaviour described for neutral
systems. In the opposite (and more exciting), salt-free limit, κ contains only the counter-ion
eff
contribution and, due to the electroneutrality condition, vcc becomes ρc -dependent and is
(1 + Zc ) less screened than wcc . This behaviour remains valid even at high or infinite
1/2
R570 L Belloni
dilution of the colloids (in the mathematical sense that two potentials of different screening
differ even when both screening tend to zero). From this point of view, salt-free colloidal
solutions can never be considered as dilute. As ρc decreases, the ionic screening decreases in
parallel (≈(ρc Zc )1/2 ) and the colloids remain coupled at larger distances.
For salt-free systems, it is important to emphasize the mistake which consists in using
eff
without caution the OCM limiting law gcc = exp(−βvcc ) at zero density and thus identifying
eff
wcc with vcc . This would lead to the obviously wrong result χosm = Scc (0) = 1 − Zc . This
limiting law is valid in general except for the present salt-free case. The correct reasoning
eff eff
uses the limiting law ccc = −βvcc , which remains valid in all circumstances and which
leads, through the OCM–OZ equation, to expression (37) for gcc and to the correct, perfect
gas law χosm = 1/(1 + Zc ). Ignoring these subtleties and confusing effective and mean-force
potentials could lead to severe misinterpretations [76].
It is interesting to verify that the OCM compressibility route is able to exhibit the perfect
gas pressure of the counter-ions despite the non-explicit treatment of their degrees of freedom.
This illustrates the electroneutrality condition hidden behind the DH expression for gcc . On
the other hand, as noted in the general analysis, the OCM virial route is not so valid. Indeed,
introducing the effective potential (46) in the OCM virial equation (2) gives for the leading
term (no salt, ρc → 0) β ≈ ρc (1 + Zc /2)!
Before closing this section on the basic DH approach, we return to the electrostatic
depletion versus screening discussion in complex charged colloidal mixtures, already
mentioned in 3.2–3.3. The large colloids are now surrounded by ions (of screening constant
κ) as well as small ‘solvent’ colloids or polymer impurities (κs ). Starting from the PM
describing all species with coulombic potentials or from the intermediate model describing
large and small colloids only with ion-averaged screened coulombic potentials, the effective
interaction between large colloids averaged over all remaining components reads within the
DH approximation
eff Zc2 LB
βvcc (r) = exp[−(κ 2 + κs2 )1/2 r]. (48)
r
At this DH level of description, the small colloids are considered as pointlike multivalent
ions, only contributing to the total ionic strength. When the density of this component is low,
the total screening factor may be expanded. Keeping the first order in ρs only, the effective
potential becomes
Zc2 LB −κr κ2
eff
βvcc (r) ≈ e 1− s r . (49)
small ρs r 2κ
The negative contribution in the right-hand side of (49) is nothing else than the leading term
(27) in the virial diagrammatic expansion applied to the intermediate model (with linearization
of the cs correlations). Use of (49) without caution would lead to the incorrect conclusion
that the depletion attraction of electrostatic origin due to the solvent colloids dominates at
large distances, in contradiction with the pure repulsion in the original expression (48). The
importance of higher order terms in the expansion (the so-called ring diagrams [1]) illustrates
again the long-range and multi-body character of the electrostatic screening. A true depletion
attraction in charged mixtures may exist only beyond the DH linearized treatment of the cs
correlations [49].
4.2. Improved DH-like theories

The main failure of the bare DH theory is the neglect of the colloidal size and the linearized
treatment of the cc correlations. As a direct consequence, gcc (r) presents non-physical
negative values at short distances. Introducing a finite value for σc = 2a and replacing
the cc MSA closure by a more adapted, nonlinearized, approach (HNC integral equation) does
not qualitatively change the final DH results. Numerous analyses successfully followed this
direction in the early 1980s and derived important, somewhat forgotten, conclusions. As a first
step, we leave the cc MSA treatment, still neglecting the colloidal size. The present derivation
is very similar in spirit to the pioneer study by Beresford-Smith et al [20]. The MSA–DH cci
and cij (i, j = ions) are inserted in the PM–OZ equation (12) while no assumption is made
on gcc . Mapping the PM into the effective OCM (26) allows us to derive the same bare DH
eff
vcc (46) as before (inclusion of cc bridge functions beyond the HNC closure would affect this
effective potential at short distances). In parallel, gci and gij can be expressed as convolution
eff
products of gcc and vcc . Introducing these functions in the PM virial equation (9) leads to the
pressure

κ(κ 2 + 23 κc2 ) ρc κc2 ∞ κr e−κr
βP = ρ − + gcc (r) 1 − dr (50)
24π 24π 0 2 r
Note that gcc could be replaced by hcc without modifying the integral in (50). The DH virial
equation of state (40) could be recovered by inserting (37) in (50). The equation of state
(50) contains three terms: the perfect gas pressure of all particles, a negative contribution
which is independent of the colloidal positions and an integral term which depends on the cc
correlations. In the absence of salt, (50) becomes

κ 3 (1 + 23 Zc ) ρc κc2 ∞ κr e−κr
β = βP = ρc (1 + Zc ) − + gcc (r) 1 − dr . (51)
24π 24π 0 2 r
Using without caution the OCM virial equation (2) with the effective potential (46) leads to
(no salt)

Zc ρc κc2 ∞ e−κr
βP = ρc 1 + + hcc (r)(1 + κr) dr . (52)
2 24π 0 r
The already mentioned factor 21 in the counter-ionic perfect gas pressure is inadequate. One
must keep in mind that the OCM virial equation (2) was established for a density-independent
pair potential v(r). Generalization to ρ-dependent v(r) consists in formally replacing v (r)r/3
by v (r)r/3 − ∂v(r)/∂ ln ρ in (2). For the present screened coulombic potential, the modified
OCM virial equation becomes (no salt)

ρc κc2 ∞ κr e−κr
βP = ρc + gcc (r) 1 − dr . (53)
24π 0 2 r
Now, the perfect gas pressure of the counter-ions has completely disappeared. On the other
hand, the integrals in (53) and (51) coincide. The extra terms in the PM virial equation which
are not captured by the OCM virial approach illustrate the importance for the thermodynamics
of zero- and one-body terms in the full potential of mean force. For a discussion at the free
energy level, see [20]. A similar DH analysis has been recently proposed in [77].
If the finite colloidal size is taken into account, the previous approach remains analytical
and the DH-like conclusions remain valid (although the final expressions become much
heavier):
eff
(i) vcc keeps its screened coulombic, repulsive form (MSA or HNC closure for the cc
correlations) [26–29, 78]:
eff e−κr
βvcc (r) = Zc2 LB X 2 r > 2a. (54)
r
R572 L Belloni
Note again that the equality in (54) is restricted to the large separation regime when cc
bridge functions are introduced.
The factor X is a function of κa and of the cc correlations [29]:
eκa
X= ∞ . (55)
1 + κa + e−κa (sinh κa − κa cosh κa)ρc 0 dr gcc (r) e−κ(r−2a) /κr
When the last term in the denominator of (55) is discarded (which is valid at low volume
fraction), the resulting effective potential takes the famous DLVO form [64], originally
deduced from the linearized Poisson–Boltzmann approach (see 4.3) [26, 27, 78]:
eff Zc2 LB e−κ(r−2a)

βvcc (r) = r > 2a. (56)
(1 + κa)2 r
Due to the minimum contact distance a, the ions are rejected farther from the colloid centre,
their screening ability is weaker and the cc repulsion is stronger. At higher colloidal
density, the integral term in (55) is non-negligible and the ρc -dependent pair potential
between two colloids is perturbed by the presence of colloidal neighbours [26–29]. This
illustrates how the effective pair potential takes many-body correlations into account.
eff
(ii) Since vcc remains repulsive, this means that the osmotic compressibility χosm derived
from the compressibility equation is finite (and low) and the corresponding equation of
state is monotonic without special features.
(iii) On the other hand, the PM virial equation of state exhibits a negative electrostatic
contribution similar to the −κ 3 term in (40) or in (50) [79], which may induce van
der Waals loops and a liquid–gas phase transition. The simultaneous presence at
the same thermodynamic state of a diverging virial compressibility and of a purely
repulsive effective potential again illustrates the approximate nature and thermodynamic
inconsistency of DH-like treatments (only comparisons with more refined theories or exact
simulation data could say which DH route is the most accurate one for a given state). In a
correct approach, compressibility and virial PM osmotic pressures must coincide; if a PM
spinodal line is approached as near a critical point, the divergence of χosm implies high
values of Scc (q) at zero angle according to (35) and thus long-range positive correlations
in gcc (r). Since by definition these functions are identical within the effective OCM and
the PM, the effective pair potential associated with the OCM must present attractive values
in this region of the phase diagram (note that nothing can be concluded a priori about the
sign of the effective interaction far from the spinodal line).
The next step in an improved theory involves a treatment of the colloid–ion correlations at
a nonlinearized level. Using the integral equation approach, this can be done by replacing the
MSA with the HNC closure [80]. A similar but not fully equivalent approximation is based
on the Poisson–Boltzmann equation, which will be described first.
4.3. Poisson–Boltzmann theory

The objective of this approach is to calculate the local average ionic densities ρi (r) and the local
average electrostatic potential ψ(r ) around the colloids fixed in the configuration (r Nc ). The
inhomogeneous ionic fluid is in osmotic equilibrium with the salt reservoir and feels external
forces of excluded volume and electrostatic origin exerted by the macro-ions. The exact
Poisson equation relates ψ or the electric field E = −∇ψ to the average charge density ρel :

Aεψ = −div εE = −ρel /ε0 = −e(Zc ρc + Zi ρi )/ε0 . (57)
i
Dielectric discontinuities at the colloidal surfaces, |r − rc | = a, can be taken into account

in (57).
Within the Poisson–Boltzmann (PB) approximation, all ions are assumed to be pointlike
and the Nc -body potential of mean force (semi-grand potential of the inhomogeneous fluid) is
expressed as [22, 64, 81, 82]:

W N c (r N c ) = U − T S − µ i Ni
i

1 1
U= ρel (r )ψ(r ) dr = ε0 ε E 2 (r ) dr
2 2

−T S = kT ρi (r )(ln ρi (r ) − 1) dr
i

µi Ni = µi Ni = kT ln ρi ρi (r ) dr . (58)
The integral in the electrostatic energy U is over the fluid volume as well as the interior of the
colloids. The integral in the entropy S is over the fluid volume only. The last equation identifies
the ionic chemical potentials in the solution with their values in the reservoir and guarantees the
osmotic equilibrium. The zero of potential ψ is arbitrarily chosen in the reservoir. Expressions
(58) are valid for fixed colloidal charge (constant charge boundary condition) and must be
completed by extra terms [81] if instead the surface potential is fixed (constant potential) or
if the charging process of the surfaces results from chemical equilibrium between surface
sites and bulk ions (charge regulation [83]). The ionic profiles are given by the mean-field
Boltzmann approximation:
ρi (r) = ρi exp[−Zi eψ(r )/kT ] = ρi exp[−Zi ϕ(r )] r ∈ fluid (59)
where ϕ = eψ/kT = ψ/25 mV.
This law can be elegantly recovered within the density functional theory (DFT) formalism
[84] by considering WNc in (58) as a functional of the colloidal densities and by using the
equilibrium condition δWNc /δρi = 0 (functional derivative).
Insertion of these ionic profiles in the Poisson equation leads to the PB equation which,
for +1/−1 reservoir electrolyte, takes the well known form
Aϕ = κ 2 sinh ϕ r ∈ fluid (60)
while the Laplace equation Aϕ = 0 holds inside the particles. κ = 8π LB ρs represents the
screening constant inside the reservoir. Again, the PB equation (60) implicitly guarantees that
the salt chemical potential µs = µ+ + µ− = 2kT ln ρs is uniform everywhere in the ionic fluid
and in the reservoir. The constant charge boundary condition relates the external and internal
normal electric fields at the colloidal surfaces to the surface charge density Cc = Zc /(4π a 2 ):
∇n ϕext − εint /ε∇n ϕint = −4π LB Cc . (61)
This condition automatically expresses the global electroneutrality of the solution.
The salinity ρs inside the solution is obtained by averaging the co-ion profile. In
experiments performed without osmotic exchange, it is this quantity rather than ρs which
is fixed and known. In that case, ρs is formally considered as a parameter a posteriori adjusted
to give the correct ionic content in the solution. The salt-free limit is continuously reached
by simply investigating, in the scheme above, very low values of ρs or κ . As ρs → 0, ϕ is
shifted by the additive constant − ln ρs , the co-ion profile vanishes (sinh ϕ and cosh ϕ may
be replaced by exp(ϕ)/2) and the precise value of ρs becomes irrelevant for the counter-ion
profile.
R574 L Belloni
In a linearized version, ϕ is supposed to take low values (the local densities are close to
ρs ), sinh ϕ is expanded to first order in ϕ and the PB equation is replaced by the linearized PB
or DH equation:
Aϕ = κ 2 ϕ r ∈ fluid. (62)
This corresponds to expanding the functional S up to second order in the ionic profiles ρi [84].
While the nonlinearized PB equation is adapted to all salinity, (62) is restricted to salt excess
(Zc ρc ρs ). At low added salinity or in the absence of salt, the expansion must be performed
to first order in ϕ −ϕ0 rather than ϕ where ϕ0 is the potential somewhere to be chosen inside the
solution. The resulting linear equation contains ϕ0 -dependent screening and constant terms
and loses its general character.
The implicit approximations assumed by the PB theory are of mean-field nature. The
average of the product of the instantaneous charge density and electrostatic potential has been
replaced in U by the product of the averages of these quantities, ρel ψ (this would correspond
to replacing the two-particle density ρss (r , r ) by the product ρs (r )ρs (r ) in the general virial
equation (18), neglecting the inhomogeneous hss (r , r ) function [20]). The entropy S of the
ionic fluid has been replaced by its ideal expression. Equivalently, the local ionic profiles
ρi (r ) which are expressed in an exact theory as N -body statistical averages of the microscopic
Boltzmann factor exp(−βUN ) are replaced here by the exponential of the mean interaction,
exp(−βZi eψ). In other words, the ionic fluctuations or ion–ion correlations are neglected in
the PB mean-field theory.
The main task is to solve the PB equation (60) with the boundary condition (61) (or an
equivalent one). For a general colloid configuration, this requires a formidable numerical
calculation [22], even in the linearized version. The PB equation is a second-order differential
equation with partial derivatives. The fluid boundary at the colloidal surfaces where the
condition (61) applies is a complex surface in the three-dimensional solution. ‘Reflections’
of the electrostatic potential coming from one colloid onto the hard-sphere neighbours or
‘overlaps’ of the ionic layers around all colloids induce in general a complex, Nc -body map of
ϕ, which cannot be decomposed as a sum of one-body contributions. Nowhere in the solution
are ϕ and its gradient simultaneously known a priori Thus, a numerical calculation starts
with a guess for ϕ(r) and follows iterative procedures using Newton–Raphson techniques and
up-to-date numerical analysis to speed up the convergence [85].
Once the local potential ϕ(r) and ionic densities ρi (r) are obtained, many important
quantities relative to the colloids and to the thermodynamics can be deduced. The force acting
on colloid c is [81]

Fc = −∇c WNc = T dS . (63)
c
By convention, the vector dS points towards the ionic fluid. The tensor T contains an
electrostatic part Tel and a kinetic or thermal part Tth :
T = Tel + Tth = ε0 ε(EE − 21 E 2 I) − kT ρion I . (64)
Tel is the Maxwell tensor, which gives the electric force exerted on a charged surface (excluding
the self electric field due to the surface itself). Tth represents the perfect gas osmotic pressure
due to all ions of local total density ρion . Since the tensor T is divergence free (div T = 0), the
surface integral in (63) can be performed on the surface of colloid c as well as on any larger
closed surface surrounding the previous one and excluding colloidal neighbours. Physically,
this expresses the fact that each elementary ionic volume is at equilibrium, the electric force
being exactly balanced by the ion osmotic pressure gradient (thermal force). This freedom in
the choice of integration surface can be advantageously used to minimize numerical errors in
the force calculation. Indeed, the electric and thermal parts are high and opposite in sign near
the macro-ions and must be determined very accurately in order to extract the small difference
with a reasonable precision. It is thus preferable to choose a closed surface located farther from
the macro-ion, at mid-distance with the first neighbours, where the full tensor is dominated
by the thermal part. Additionally, this optimal choice allows us to deduce in some simple
geometry the sign of the force before any numerical calculation (see below).
Following the same route, the pressure P = −δWNc /δV can be obtained by investigating
a small increase δV of the volume V . In that case, the frontier of the solution (cell edge)
as well as the position of the Nc colloids are homothetically rescaled by the relative value
δr /r = δV /3V . The resulting increase in WNc gives [81]

1 1
P = ρc kT + rc Fc − r ρion kT dS . (65)
3V col c 3V cell c
The same expression can be elegantly recovered from the PM virial equation (18) by noting
that the electric force Zi eE applied on ion i is equal to the osmotic force kT ∇ ln ρi within the
PB approximation. The equation of state contains three terms, the perfect gas pressure of the
colloids, the colloid–colloid correlation term and the ionic contribution, which depends on the
local ionic density at the cell edge. If this density is uniform, the ions contribute as ρcell kT to
P [86, 87]. Note again that the OCM virial equation (19) would miss this last contribution.
Lastly, the osmotic pressure is obtained by subtracting the reservoir pressure P = 2ρs kT
from P .
We now apply these general PB expressions to specific configurations, starting from the
simplest case of one isolated colloid immersed in an electrolyte (Nc = 1, ρs = 0). On the
analogy of the linearized, DH solution, the long distance behaviour of the PB solution (60) is
written as [88] (r is measured from the centre of the colloid):
eff
Zc LB e−κ(r−a)
ϕ(r) ≈ . (66)
κ(r−a)1 (1 + κa) r
eff
The powerful concept of effective charge Zc or charge renormalization is very general [89]
and illustrates the strong accumulation or ionic condensation of counter-ions in the vicinity
of colloid surfaces. Because of this short-range, non-linear effect, the colloid can be regarded
far from its surface as if, at the linearized, DH level, it carried a weaker, partially neutralized
eff eff
charge, Zc < Zc , which depends on Zc and κa. Within the PB approximation, Zc
saturates to a constant value as the colloid becomes highly charged. In that regime, an increase
of Zc is counterbalanced by an equal increase in the condensed charge and the long-range
potential becomes independent of the details on the colloidal surface. More details on the
ionic condensation around spheres could be found in a recent review [90].
For two interacting colloids separated by the centre-to-centre distance D in an electrolyte
[91], the PB force is a priori always repulsive. Indeed, integrating the stress tensor on the closed
surface constituted by the mid-plane between the colloids, where E dS = 0 by symmetry, and
a hemisphere of infinite radius, where E = 0 and ρion = 2ρs , leads to the expression

ε0 ε 2
Fc = E + (ρion − 2ρs )kT dS (67)
mid-plane 2
which is obviously positive.
The configuration of two colloids has been extensively studied since the DLVO pioneer
works. For large colloids near contact (κ a 1, D − 2a a), it is sufficient to study the
simpler, one-dimensional geometry of two parallel charged plates (of same charge density than
the colloids) and use the Derjaguin approximation to deduce the colloid–colloid force. For
R576 L Belloni
a general configuration, the PB equation requires a more complex resolution. The potential
ϕ(r, θ ) depends on the distance r to one colloid and on the angle θ between r and the axis
joining the colloids. In the linearized DH version, ϕ can be analytically expressed as an
infinite expansion in Bessel functions in r and Legendre polynomials in cos θ [64, 92]. In the
nonlinearized PB case, bispherical coordinates are adapted to the numerical resolution [93].
Note that the complete and systematic numerical PB solution was obtained only in the early
1990s (and for moderate surface charge), 40 years after the first studies. This illustrates the
difficulty of the numerical problem, even for this fundamental and apparently simple geometry.
An analytical expression for the force or the effective, mean-force pair potential can be
derived in the asymptotic regime of large separations, κ(D − 2a) 1. The weak overlap
or linear superposition approximation applies and the potential far from the two particles,
in particular at the mid-plane, is the sum of the contributions due to each colloid, taken as
isolated. Inserting the long-range expression (66) for each contribution in (67) gives for the
pair interaction
eff 2
eff Zc LB e−κ(r−2a)
βvcc (r) = κ(2 − 2a) 1. (68)
(1 + κa)2 r
eff
In the linearized case, Zc is replaced by the bare value Zc and the famous DLVO potential
(56), already mentioned, is recovered. One must keep in mind that this expression is in principle
valid only at large separations. At shorter distances, the weak-overlap approximation and the
identification (66) break down, and departures from the asymptotic screened coulombic force
are observed, even in the linearized version [92].
The DLVO potential (56), (68) has been the starting point of thousands of studies and
has allowed a qualitative as well as quantitative understanding of the following phenomena:
(i) the stability of charged dispersions against irreversible aggregation [64]; (ii) the short-range
liquid order observed in scattering [2–6]; (iii) the fluid–crystal transition which appears at low
ionic strength [94]; (iv) the direct pair force measurements [9–16]; (v) the PM simulation data
[67]; . . . . The effective charge is usually considered as an adjustable parameter or is chosen
according to a priori theoretical predictions [88–90].
The PB resolution becomes very difficult in more complex geometries involving numerous
colloids (see the beautiful example in [22]). As far as the ionic profiles around macro-ions
in concentrated solutions are concerned, a simple alternative is to use the cell geometry. The
solution is divided into Nc spherical, overall neutral cells, each containing one centred colloid.
The cell radius R is related to the colloidal concentration, ρc 4π/3R 3 = 1. The spherically
symmetric PB equation is easily solved from r = a to r = R (where E = 0). The counter-ion
profiles are used to deduce effective charge values as a function of colloidal density, even in
the absence of salt [90]. According to (65), the PB osmotic pressure is given by the edge ionic
density [86], = (ρion (R) − 2ρs )kT , and successfully reproduces experimental equations of
state [5]. Of course, this simple cell geometry is not adapted to derive colloidal forces since
Fc = 0 by symmetry. Investigating eccentric positions of the macro-ion in the cell or the
presence of two colloids in the same cell seems to be a good direction to follow to get further
insights into the colloidal interactions at finite density [95].
Another geometry consists of two colloids confined near a planar wall or between two
parallel walls or inside a cylindrical pore. The two particles are at the same distance from
the wall. An elegant analytical proof has recently generalized the bulk result [96]: the force
felt by each colloid, projected along the wall, is always repulsive, whatever the charge state,
boundary condition and dielectric property of the walls and the colloids. This a posteriori
invalidates numerical PB data which predicted attraction between particles [97] and indicates
that experimental evidences of such attraction [15, 16, 98] must be interpreted with different
approaches.
In the last few years, the PB approach has been revisited in terms of the density functional
theory [82, 84, 99]. Instead of solving directly the PB equation, this new formulation focuses
on the strictly equivalent free-energy functional (58). Making the same kind of approximations
as in older PB or integral equation analyses (linearization, weak overlap approximations), it
becomes possible to recover in an elegant way the DLVO pair potential between particles as well
as the negative, DH-like, one-body term in the equation of state or free energy [100]. Attempts
have been made to go beyond this level of approximation and to introduce partial nonlinear
and multi-body corrections. The resulting expressions for the colloidal forces remain tractable
but must be used with caution since their domain of validity is not always well controlled. As
an illuminating example of its limitation, such analysis has recently predicted an attraction
between confined particles [101], in violation of the exact PB behaviour [96]. It is clear
that such analytical treatments are adapted to give the asymptotic, long-range forces between
charged objects but cannot replace the full numerical resolution of the PB equation at shorter
separations.
Before closing this section, we recall the formal analogies and differences between the
approach based on the bulk OZ equation with MSA–HNC closures and the approach based on
the PB equation. To identify both theories, one must first neglect the ionic size and use the
MSA for the ion–ion correlations in the OZ approach. Then, in the case of one isolated colloid
immersed in an electrolyte (ρc = 0), the PB (DH) ionic profiles, ρi (r), are rigorously identical
to those obtained with the HNC (MSA) colloid–ion closure, ρi gci (r). In the presence of many
colloids, this equivalence remains valid if the ensemble of particles is formally considered
as constituting one single object of complex geometry immersed at infinite dilution in the
supporting electrolyte. In that case, the colloid–ion functions lose the spherical symmetry
and the integral equation requires a non-trivial numerical treatment which could serve as an
alternative to the standard PB resolution [102]. On the other hand, the conventional OZ
approach described in 4.2, although formally exact, is more sensitive to the neglect of the
bridge functions. For example, at infinite dilution in colloids, the potential of mean force is
given by

βWcc = − ln(gcc ) = βvcc − ρi hci ⊗ cic − bcc (69)
ion
where the functions hci and cci represent the correlations of the ions with one isolated particle.
So far, no approximation is made. Neglecting the bridge function bcc implies that the force
between two colloids is related to the one-body ionic profiles. This is clearly equivalent to a
weak-overlap approximation, valid only at large separations. In that asymptotic regime, PB
(DH) and HNC (MSA) forces coincide. At shorter distances, the conventional HNC–MSA
theory misses the non-additive overlap corrections implicitly contained in bcc . This explains
in particular why the screened coulombic DLVO potential (56) is valid at all separations, up
to contact, using the OZ–MSA route and at large separations only using the PB–DH route.
From the thermodynamics point of view, the analogy between both approaches subsists.
Taking the linearized version of the PB equation (62), neglecting for simplicity the colloidal
size and inserting the linear superposition of one-body electrostatic potentials (66) in the virial
expression (65) makes it possible to recover almost completely the osmotic pressure (50). The
single difference is the disappearance of the −κ 3 /24π term which obviously results from the
neglect of the ion–ion correlation function hss (r , r ) [20] (zero-body term in the potential of
mean force (15)).
R578 L Belloni
4.4. Beyond DLVO—effective attraction between colloids
The existence of an effective attraction of pure electrostatic origin between like-charged

particles via their counter-ions is a fascinating phenomenon which was predicted by Oosawa
in the 1960s [103] and intensively studied in various geometry since the early 1980s. Many
controversies in the literature have accompanied its description. As said in the introduction,
the reason is certainly the intrinsic complexity of highly charged, asymmetrical mixtures and
the difficulty of controlling the validity of approximations more or less implicitly assumed
in theories in the absence of exact simulation data. It is clear that such attraction could be
predicted only beyond the mean-field PB–DH–MSA–DLVO picture by including in some way
ion–ion correlation effects. The studies which seem to invalidate this assertion by exhibiting
attraction within a mean-field treatment are not satisfying. Without being exhaustive, the
Sogami–Ise attraction derived within the DH approximation [104] (the form of this attraction is
similar to (49)) has been proven to be due to an incorrect thermodynamical treatment [20, 105].
Equivalently, the claim for a non-DLVO attraction from the PM–MSA solution results from
a confusion between mean-force and effective potentials [76]. In confined geometry, we
recall that the general PB proof of pure repulsion between two colloids [96] a posteriori
and definitely invalidates the attraction PB predictions, which may be due to insufficiently
precise numerical resolutions [97] or to approximations used outside their domain of validity
in analytical resolutions [101].
The situation is perfectly clear in the simple planar geometry of two identical, parallel,
homogeneously charged plates separated by a continuous solvent containing the counter-ions
and salt ions. As early as 1984, both Monte Carlo simulations [106] and inhomogeneous
HNC calculations [107] within the PM have proven without ambiguity the existence of
attraction between the plates (negative pressure) at high electrostatic coupling, e.g. at high
surface charge density, ionic valence or Bjerrum length. The attraction appears when the
two condensation shells strongly overlap at short separation resulting in non-obvious ion–ion
correlations [108] and non-monotonic ionic profiles [109]. Although it exists in principle for
all valences, it takes place in practice, for realistic charged plates in water, only in the presence
of divalent or more highly charged counter-ions. Since the pioneer works, the phenomenon
of attraction has been reproduced and analysed by numerous studies using different non-
mean-field theories like the partially inhomogeneous three-point extension integral equation
[110], the DFT with correlation contributions in the WDA version [111], the conventional
homogeneous HNC integral equation with addition of the first bridge diagram [112], the MC
simulation on a hypersphere [113], . . . . The attraction may be enhanced by the adsorption
of additional chemical counter-ion adsorption and/or by surface charge heterogeneity (‘long-
range hydrophobic’ attraction) [114]. In that last picture, the plates are considered as two-
dimensional electrolytes or ionic crystals which spontaneously adjust complementary to each
other. From the experimental point of view, attractive forces between charged plates immersed
in aqueous calcium solutions have been directly measured with SFA and AFM techniques [115]
and monitor the behaviour of clay suspensions and cements.
For bulk systems of spherical colloids, the situation is not so clear. Theoretical evidences
of effective attraction have been first obtained following the integral equation approach. Using
the bulk HNC closure for all pair correlations, Patey considered the case of two colloids
eff
immersed in an electrolyte and found attractive values in wcc ≡ vcc [116]. The same full
HNC equation reveals at finite colloidal density and high electrostatic coupling an instability
of the PM with respect to liquid–gas transition [80, 117]. As the colloidal or ionic charge
is increased, the osmotic compressibility and Scc (0) take higher and higher values and seem
to diverge (for the precise nature of the boundary line, see [118]). At the same time, the
eff
Figure 5. Effective vcc (r) (solid line) and mean force wcc (r) (dotted line) pair potentials between
colloids in salt-free, colloid + counter-ion primitive model. T = 298 K, ε = 78, σi = 100 nm/0,
Zi = +300/−1, 6c = 5%. Note the asinh scale on the vertical axis. Inset: corresponding pair
distribution function gcc (r). HNC integral equations in the PM and effective OCM.
free energy or virial routes do not present such instability, revealing a HNC thermodynamic
inconsistency, the inverse of the DH–MSA one [80]. The location of the two-phase region and
of the critical point in the density–temperature phase diagram depends on the charge asymmetry
and on the colloid–counter-ion contact distance σci [117]. Addition of salt tends to stabilize
the system and shifts the critical point towards lower temperature [28]. As expected from
the general analysis in sections 2.4 and 4.2, the state-dependent, OCM, effective potential
eff
vcc shifts from pure repulsion to partial attraction as the two-phase region is approached
[28]. An example is given in figure 5 for a typical salt-free colloidal system σi = 0/100 nm,
Zi = +300/−1. The attraction is enhanced in the presence of ionic adsorption onto the colloids
(charge regulation) [119]. To what extent could this behaviour be due to a failure of the HNC
equation? Numerous tests indicate that the phase transition is an intrinsic property of the PM
itself and that the quantitative HNC weakness is to overestimate the attraction and to predict
the transition too soon, at too low coupling. The planar geometry limit as well as the restricted
PM version (RPM) of symmetrical electrolytes [75, 118] constitute two extreme cases which
illustrate this general trend. In between, Patey’s attraction disappears for monovalent ions
but subsists for divalent ions when including the first bridge diagram [112]. The bulk two-
phase region is shifted towards higher charges [120] when the HNC closure is replaced by the
thermodynamically consistent ZH closure [121].
Simulation data are obviously necessary to quantify the strength of the attraction and to
determine the precise location of the critical point in the phase diagram. Since the early 1980s,
the few MC and MD simulations performed in the micellar regime (σc = 2–3 nm, Zc = 10–
20) with monovalent counter-ions [65–67] show a short-range liquid order among the colloids
R580 L Belloni
Figure 6. Colloid–colloid pair distribution function gcc (r) in salt-free, colloid + counter-ion
primitive model. T = 298 K, ε = 78.4, σi = 3 nm/0.4 nm, Zi = +20/−1 or −2, ρc = 0.02 M
(6c = 17%). MC data [69] (symbols), HNC (dotted lines) and BHP (solid lines) integral equation.
eff
Inset: corresponding effective colloid–colloid pair potentials vcc (r) extracted using the same
HNC and BHP integral equations in the effective OCM.
eff
which is consistent with a purely repulsive effective potential, although no vcc has been directly
extracted from the simulated gcc in practice. HNC [65–67] and advanced closures [122] have
been tested against these exact data (anecdotally, the first MC case published in the literature
[65] was located inside the HNC two-phase region). More recent MC works have shown how
this DLVO-like picture breaks down in the presence of divalent counter-ions [69]. Figure 6
presents MC gcc (r) for +20/−1 and +20/−2 systems at the same colloid volume fraction
(5%) [69]. In the former case, the data present a classical order peak located near the mean
separation distance between macro-ions, while in the latter case the peak is shifted near contact
and seems to reveal a short-range effective attraction. The same behaviour is predicted by the
HNC integral equation. A good quantitative fit of the MC data is obtained for both valences
using the powerful BHP integral equation [123], as shown in figure 6 [124]. This accurate
closure evaluates advanced bridge functions from approximated three-body direct correlation
functions. In order to get a definite and quantitative answer about the existence of effective
eff
attraction, it is necessary to extract the OCM potential vcc from the cc correlations. This is
hardly possible from the simulation data due to the statistical noise and to the limited r range,
which prevent precise Fourier transformations but can be done easily from the theory curves.
eff
The effective potentials vcc are plotted in figure 6. Considering the very good agreement
between MC and theory in gcc (r), these potential curves should be close to the exact one.
eff
As expected, vcc remains repulsive at all distances for monovalent ions and presents a clear
negative minimum at intermediate distances in the presence of divalent counter-ions. This
confirms the pioneer HNC predictions and demonstrates without ambiguity that, on average,
the spherical macro-ions attract each other through the counter-ions in highly charged bulk
solutions. The existence of a phase separation which could result from such attractions has
received partial confirmation in very recent powerful simulations [71]. Trivalent ions induce
the formation of large clusters containing many macro-ions inside the simulation box. The
observed heterogeneity is consistent with a condensation phenomenon or a liquid–gas phase
separation, analogous to that observed within the RPM. The complete structure of the PM
phase diagram and, in particular, the competition of condensation with crystallization, again
of pure electrostatic origin, remains to be elucidated.
An alternative to the difficult, time-consuming simulations in bulk is offered by the cell
approach where only one or two spherical macro-ions are considered in a box. In the original
case of one colloid located at the centre of the Wigner–Seitz spherical cell, MC data show a
non-PB behaviour at high coupling [87]. As the colloidal charge Zc is increased in the absence
of salt, the counter-ion concentration at the cell edge (and thus the osmotic pressure or the
effective charge) first increases, goes through a maximum and then decreases (MC or DFT–
WDA) [125], in contradiction with the saturation effect predicted by the PB approximation
[89]. This kind of instability indirectly announces the existence of attraction between highly
charged particles. Recent simulations have investigated the situation of two colloids inside the
same cell and derived pair potentials/forces as a function of the separation D [126–128]. As
usual in such geometry, the role of the cell boundary must be controlled with caution in order
to avoid undesirable effects. In presence of added salt, it is sufficient in principle to investigate
large enough cells, R κ −1 , D. In practice, for obvious reasons of computer limitations, κ R
is not very large and comparisons with simulations performed with periodic image condition
are necessary to interpret the data with confidence [128]. In any case, such simulations solve
exactly (at last!) the original DLVO problem of two isolated spherical colloids immersed in an
infinite electrolyte. The force remains repulsive for monovalent ions and becomes attractive
for divalent counter-ions [128]. On the other hand, interpretation of the MC results in salt-
free situations (κ = 0) [126, 127] is less obvious. The short-range attraction observed at
high coulombic coupling (divalent ions in solvent of low dielectric constant) [127] seems to
eff
be more related to the potential of mean force wcc rather than to the effective potential vcc .
Indeed, the measured force is partly due to colloidal neighbours which are hidden outside the
cell boundary. As said before, the colloidal density through the cell dimension is never an
irrelevant factor in salt-free systems. Even if the entropy of ions associated with the very large
cell volume wins over the colloid–ion attraction only at high dilution (logarithmic dependence
of S on ρc ) [89], correct sampling of the very few MC configurations with ions located near the
cell edge should be monitored during the simulation. Anyway, the present cell approach with a
fine control of boundary effects seems to be a promising simulation tool to derive information
on ion-mediated colloidal interactions, cheaper and complementary to the PM bulk approach.
Experimental evidence of ion-mediated effective attraction between spherical colloids is
still an open and controversial issue. While the DLVO picture including a pure, screened
coulombic repulsion between colloids has successfully explained most of the observed
behaviours for half a century, some particular phenomena have resisted such interpretation.
Electrostatic colloid crystals obtained at low concentration in highly deionized conditions
(monovalent counter-ions) have been studied by Ise et al by direct visualization using confocal
laser scanning microscopy and by ultrasmall-angle x-ray scattering [51]. Both techniques
reveal the presence of apparently empty regions or voids inside the crystal. This has been
claimed to be evidence for a long-range attraction [104]. A more recent DFT analysis imputes
the local demixion to the presence of the −κ 3 constant term in the DH-like virial equation of
state (51) [100]. This experimental observation has not been reproduced by similar studies
of highly deionized latex systems [5] and contradicts the direct force measurements between
R582 L Belloni
isolated particles [14]. As said in section 3.2, a possible explanation of the disagreement
could be the undesirable presence of charged polymer impurities released by the latex surfaces
in some solutions which would induce a depletion attraction of electrostatic origin between
colloids. In that mechanism, the voids would be full of undetectable polymers. In the same
way, the macroscopic liquid–gas-like phase separation experimentally observed at low ionic
strength [129] has been reanalysed in terms of ionic impurities [130].
These particular and controversial cases apart, there is no clear experimental confirmation
of the PM effective attraction between spherical macro-ions. This should be opposed to the
situation observed in other geometry as for parallel plates [115], parallel cylinders [131] or
in flexible polyelectrolytes [132] where attraction has been more or less directly measured in
various systems. The fact that the calculated and observed attraction is always coupled with the
presence of multivalent counter-ions is the key point which should incite the experimentalists
to replace the usual monovalent ions by divalent or trivalent ones in spherical colloidal systems.
Chemical destabilization seems to be the limiting factor in practice.
5. Van der Waals and miscellaneous interactions
So far the interior of the colloids has been considered as an inert, structureless material.
In reality, colloids are made of polar and polarizable molecules. Atoms themselves carry
instantaneous, fluctuating electronic dipoles. Interactions between all permanent and induced
dipoles belonging to the different colloids, averaged over the degrees of freedom of the polar
solvent, define the so-called van der Waals (VW) or dispersion forces [133, 134]. This universal
interaction due to correlations in polarization fluctuations is essentially attractive and tends to
destabilize colloidal suspensions towards irreversible coagulation. Its theoretical description
can be viewed as the ultimate task in the field of colloidal interactions due to its complexity
(the present VW section has been paradoxically rejected to the end of the review for this
reason). Indeed, a complete theory must involve electrodynamics, quantum mechanics and
statistical mechanics and faces the fundamental problem of non-pair-wise additivity of the total
interactions between all atoms/molecules/particles intrinsically associated with the presence
of polarizability. The case of an isolated pair of atoms was solved by London in 1930. Then,
summing all r −6 atom–atom pair interactions, Hamaker deduced simple forms for the pair
potential between colloidal bodies, which are widely used in the literature [9]. The Hamaker
constant, which characterizes the strength of the interaction, depends on the frequency-
dependent polarizability αi (ω) of the colloid and solvent materials. In order to go beyond this
pair-wise additivity approximation, valid in principle at low αc − αs only, one must account for
the multiple ‘reflexions’ of the electromagnetic waves between the different local, oscillating
dipoles. In the microscopic description, this would require us to consider colloids as collections
of fixed polarizable elementary volumes and solvent molecules as moving dipolar, polarizable
particles and to statistically average the multi-body interactions over the dipole fluctuations
and solvent spatial configurations. This is a formidable task even within classical mechanics
and electrostatics. An alternative, macroscopic approach, proposed by Lifshitz in 1956, views
colloids and solvent as continuous dielectric media, characterized by their dielectric response
function εi (ω) [133, 134]. Solving the Maxwell equations for the electromagnetic fields and
using quantum field theory makes it possible to derive the interaction between two dielectric
bodies immersed in a third medium (note the parallelism with the scattering problem which
involves the same concepts: in the microscopic Rayleigh–Gans theory, the intensity scattered
by a colloid is decomposed as a sum of contributions due to all its atoms, taken as isolated; the
macroscopic Mie theory accounts for multiple scattering inside the same colloid by solving the
Maxwell equations in continuous media). In practice, this can be done in simple cases only,
essentially in the planar geometry. The Hamaker constant in the formal Hamaker expression
is a complex function of the εi (ω) and can be calculated in principle from the experimental
absorption or dispersion spectra on the whole frequency domain [133].
From this introduction, it is clear that the classical DLVO decomposition of the total
colloidal interaction in sum of bare VW and extra (screened coulombic, hydration . . .)
independent contributions, though useful, may appear somewhat artificial and inconsistent.
All interactions are of long-range coulombic nature and thus must be coupled, revealing
miscellaneous interactions. (i) The hydration and hydrophobic forces (section 3) due to the
discrete solvent structure and to specific colloid–solvent affinity can be considered as a part of
the VW force at short separation. The solvent layer near the colloidal surface may be affected by
surface–solvent dipole–dipole VW coupling. Inversely, solvent depletion or adsorption should
alter the local dielectric constant in the London–Lifshitz theory. (ii) In the presence of uniform
or heterogeneous charges on the colloids and ions in the solvent, VW dipole fluctuations and
PM charge correlations become coupled. Thus, dispersion interactions and electrostatic images
due to dielectric discontinuities must be treated self-consistently. Ninham and Parsegian
generalized the Lifshitz theory by including charge fluctuations in the electrolyte at the DH
level [135]. The main result is that the static, ‘zero-frequency’ dispersion attraction becomes
screened as exp(−2κr). Advanced studies have been performed essentially in the planar
geometry [114, 136] with a few exceptions for spheres [119, 137]. (iii) Lastly, the ion binding
or physisorption to the colloidal surfaces, which drives a colloid–colloid enhanced attraction as
seen before, may be induced by dispersion interactions acting on ions themselves. This could
explain in a rational way [138] ion specific, non-generic effects like the lyotropic or Hofmeister
series for coagulation/crystallization efficiency, the salting in/salting out phenomenon and the
long-range hydrophobic attraction [139], experimentally observed for decades in protein or
polyelectrolyte systems and not yet fully understood. The self-consistent description of all
these miscellaneous interactions in bulk solutions of spherical colloids forms a challenge for
theoreticians in the near future.
Acknowledgments
This review follows a course given during the summer school of the European Network ‘Colloid
physics’, held in Varenna, Italy, in June 1998. It is a pleasure to thank the organizers of the
school, V Degiorgio and R Piazza for their invitation and for instigating the present paper.
References
[1] Hansen J P and McDonald I R 1986 Theory of Simple Liquids (London: Academic)
[2] Lindner P and Zemb Th (ed) 1991 Neutron, X-ray and Light Scattering: Introduction to an Investigative Tool
for Colloidal and Polymeric Systems (Amsterdam: North Holland)
[3] Corti M and Degiorgio V 1981 J. Phys. Chem. 85 711
Hayter J B and Penfold J 1981 J. Chem. Soc. Faraday Trans. I 77 1851
Chen S H 1986 Annu. Rev. Phys. Chem. 37 351
[4] Parsegian V A, Rand R P, Fuller N L and Rau D C 1986 Methods in Enzymology: Biomembranes, Protons and
Water, Structure and Translocation vol 127, ed L Packer (New York: Academic) p 400
Véretout F, Delaye M and Tardieu A 1989 J. Mol. Biol. 205 713
Goodwin J W, Ottewill R H and Parentich A 1990 Colloid Polym. Sci. 268 1131
Dubois M, Zemb Th, Fuller N, Rand R P and Parsegian V A 1998 J. Chem. Phys. 108 7855
[5] Reus V, Belloni L, Zemb Th, Lutterbach N and Versmold H 1997 J. Physique II 7 603
[6] Bonnet-Gonnet C, Belloni L and Cabane B 1994 Langmuir 10 4012
[7] Perrin J 1910 J. Physique 9 5
Piazza R, Bellini T and Degiorgio V 1993 Phys. Rev. Lett. 71 4267
Rutgers M A, Dunsmuir J H, Xue J Z, Russel W B and Chaikin P M 1996 Phys. Rev. B 53 5043
R584 L Belloni
[8] Sonneville-Aubrun O, Bergeron V, Gulik-Krzywicki T, Jönsson B, Wennerström H, Lindner P and Cabane B

2000 Langmuir 16 1566
[9] Israelachvili J N 1991 Intermolecular and Surface Forces (San Diego: Academic)
[10] Ducker W A, Senden T J and Pashley R M 1991 Nature 353 239
[11] Leal Calderon F, Stora T, Mondain Monval O, Poulin P and Bibette J 1994 Phys. Rev. Lett. 72 2959
[12] Prieve D C, Luo F and Lanni F 1987 Discuss. Faraday Chem. Soc. 83 297
Prieve D C and Walz J Y 1993 Appl. Opt. 32 1629
Flicker S G and Bike S G 1993 Langmuir 9 257
[13] Vondermassen K, Bongers J, Mueller A and Versmold H 1994 Langmuir 10 1351
[14] Crocker J C and Grier D G 1994 Phys. Rev. Lett. 73 352
[15] Kepler G M and Fraden S 1994 Phys. Rev. Lett. 73 356
[16] Crocker J C and Grier D G 1996 Phys. Rev. Lett. 77 1897
[17] McMillan W G Jr and Mayer J E 1945 J. Chem. Phys. 13 276
[18] Kirkwood J G and Buff F P 1951 J. Chem. Phys. 19 774
[19] Friedman H L and Dale W D T 1977 Modern Theoretical Chemistry 5 ed B J Berne (New York: Plenum) ch 3
[20] Beresford-Smith B, Chan D Y and Mitchell D J 1984 J. Colloid Interface Sci. 105 216
[21] Lekkerkerker H N W, Poon W C K, Pusey P N, Stroobants A and Warren P B 1992 Europhys. Lett. 20 559
Lekkerkerker H N W, Buining P, Buitenhuis J, Vroege G J and Stroobants A 1995 Observation, Prediction and
Simulation of Phase Transitions in Complex Fluids ed M Baus et al (Dordrecht: Kluwer) 53
[22] Fushiki M 1992 J. Chem. Phys. 97 6700
[23] Dijkstra M, van Roij R and Evans R 1999 Phys. Rev. E 59 5744
[24] One simple rigorous way to derive this expression is to define the potential of mean force from the Nc -body
colloid distribution function and take the limit of zero colloidal activity in the full grand canonical ensemble.
[25] Adelman S A 1976 J. Chem. Phys. 64 724
[26] Beresford-Smith B and Chan D Y C 1982 Chem. Phys. Lett. 92 474
Senatore G and Blum L 1985 J. Phys. Chem. 89 2676
[27] Belloni L 1986 J. Chem. Phys. 85 519
[28] Belloni L 1987 Thèse d’Etat Paris VI
[29] Khan S, Morton T L and Ronis D 1987 Phys. Rev. A 35 4295
[30] Henderson R L 1974 Phys. Lett. A 49 197
[31] Asakura S and Oosawa F 1954 J. Chem. Phys. 22 1255
[32] Poon W C K and Pusey P N 1995 Observation, Prediction and Simulation of Phase Transitions in Complex
Fluids ed M Baus et al (Dordrecht: Kluwer) p 3
[33] Attard P 1989 J. Chem. Phys. 91 3083
[34] Vrij A 1976 Pure Appl. Chem. 48 471
[35] Biben T and Hansen J P 1991 Phys. Rev. Lett. 66 2215
[36] Gast A P, Hall C K and Russel W B 1983 J. Colloid Interface Sci. 96 251
[37] Walz J Y and Sharma A 1994 J. Colloid Interface Sci. 168 485
[38] Mao Y, Cates M E and Lekkerkerker H N W 1995 Physica A 222 10
[39] Christenson H K and Horn R G 1985 Chem. Scr. 25 37
[40] Rudhardt D, Bechinger C and Leiderer P 1998 Phys. Rev. Lett. 81 1330
[41] Sanyal S, Easwar N, Ramaswamy S and Sood A K 1992 Europhys. Lett. 18 107
[42] Louis A A, Finken R and Hansen J P 2000 Phys. Rev. E 61 R1028
[43] Forsman J, Woodward C E and Jönsson B 1997 J. Colloid Interface Sci. 195 264
[44] Claesson P M, Kjellander R, Stenius P and Christenson H K 1986 J. Chem. Soc. Faraday Trans. I 82 2735
[45] Bibette J, Roux D and Nallet F 1990 Phys. Rev. Lett. 65 2470
[46] Richetti P and Kékicheff P 1992 Phys. Rev. Lett. 68 1951
[47] Sharma A and Walz J Y 1996 J. Chem. Soc. Faraday Trans. 92 4997
Sharma A, Tan S N and Walz J Y 1997 J. Colloid Interface Sci. 191 236
[48] Hanley H J M, Straty G C and Lindner P 1994 Langmuir 10 72
Ottewill R, Hanley H J M, Rennie A R and Straty G C 1995 Langmuir 11 3757
[49] Lutterbach N, Versmold H, Reus V, Belloni L, Zemb Th and Lindner P 1999 Langmuir 15 345
[50] Spalla O personal communication
[51] Dosho S et al 1993 Langmuir 9 394
Ise N, Konishi T and Tata B V R 1999 Langmuir 15 4176
[52] Attard P, Ursenbach C P and Patey G N 1992 Phys. Rev. A 45 7621
[53] Napper D H 1983 Polymeric Stabilisation of Colloidal Dispersions (London: Academic)
[54] Lafuma F, Wong K and Cabane B 1991 J. Colloid Interface. Sci. 123 9
[55] Spalla O, Nabavi M, Minter J and Cabane B 1996 Colloid. Polym. Sci. 274 555
[56] Peyre V, Spalla O, Belloni L and Nabavi M 1997 J. Colloid Interface Sci. 187 184
[57] Israelachvili J and Wennerström H 1996 Nature 379 219
[58] Parsegian V A, Rand R P and Fuller N L 1991 J. Phys. Chem. 95 4777
Ricoul F, Dubois M, Belloni L and Zemb Th 1998 Langmuir 14 2645
[59] Kusalik P G and Patey G N 1988 Mol. Phys. 65 1105
Torrie G M, Kusalik P G and Patey G N 1988 J. Chem. Phys. 88 7826
Attard P, Wei D, Patey G N and Torrie G M 1990 J. Chem. Phys. 93 7360
[60] Forsman J, Woodward C E and Jönsson B 1997 Langmuir 13 5459
[61] Weeks J D, Katsov K and Vollmayr K 1998 Phys. Rev. Lett. 81 4400
Lum K, Chandler D and Weeks J D 1999 J. Phys. Chem. B 103 4570
[62] Marcelja S 1997 Colloid Surf. A 129/130 321
[63] Marcelja S 1997 Nature 385 689
Israelachvili J N and Wennerström H 1997 Nature 385 689
[64] Verwey E J W and Overbeek J Th G 1948 Theory of Stability of Lyophobic Colloids (Amsterdam: Elsevier)
[65] Linse P and Jönsson 1982 J. Chem. Phys. 78 3167
[66] Vlachy V, Marshall C H and Haymet A D J 1989 J. Am. Chem. Soc. 111 4160
[67] Linse P 1990 J. Chem. Phys. 93 1376
[68] Delville A 1994 Langmuir 10 395
[69] Hribar B, Kalyuzhnyi Y V and Vlachy V 1996 Mol. Phys. 87 1317
Hribar B and Vlachy V 1997 J. Phys. Chem. B 101 3457
Lobaskin V and Linse P 1998 J. Chem. Phys. 109 3530
[70] Lobaskin V and Linse P 1999 J. Chem. Phys. 111 4300
[71] Linse P and Lobaskin V 1999 Phys. Rev. Lett. 83 4208
Hribar B and Vlachy V 2000 Biophys. J. 78 694
[72] Torrie G M and Valleau J P 1980 J. Chem. Phys. 73 5807
Jönsson B, Wennerström H and Halle B 1990 J. Phys. Chem. 84 2179
Megen W and Snook I 1980 J. Chem. Phys. 73 4656
[73] Langmuir 1938 J. Chem. Phys. 6 873
[74] Levin Y and Fisher M E 1996 Physica A 225 164
Zuckerman D M, Fisher M E and Lee B P 1997 Phys. Rev. E 56 6569
[75] Orkoulas G and Panagiotopoulos A Z 1994 J. Chem. Phys. 101 1452
Caillol J M, Levesque D and Weis J J 1996 Phys. Rev. Lett. 77 4039
Orkoulas G and Panagiotopoulos A Z 1999 J. Chem. Phys. 110 1581
[76] Chu X and Wasan D T 1996 J. Colloid Interface. Sci. 184 268
[77] Warren P B 2000 J. Chem. Phys. 112 4683
[78] Medina-Noyola M and McQuarrie D A 1980 J. Chem. Phys. 73 6279
[79] Blum L and Hoye J S 1977 J. Phys. Chem. 81 1311
[80] Belloni L 1985 Chem. Phys. 99 43
[81] Hoskin N E and Levine S 1955 Trans. R. Soc. A 248 433
Hoskin N E and Levine S 1955 Trans. R. Soc. A 248 449
Bell G M and Levine S 1957 Trans. Faraday Soc. 53 143
Bell G M and Levine S 1957 Trans. Faraday Soc. 54 785
[82] Löwen H, Hansen J P and Madden P A 1993 J. Chem. Phys. 98 3275
[83] Ninham B W and Parsegian V A 1971 J. Theor. Biol. 31 405
[84] Evans R 1993 Fundamentals of Inhomogeneous Fluids ed D Henderson (New York: Dekker)
Hansen J P 1995 Phase Transitions in Complex Fluids ed M Baus, L F Rull and J P Ryckaert (Dordrecht:
Kluwer)
[85] Houstis E N, Lynch R E, Rice J R and Papatheodorou T S 1978 J. Comput. Phys. 27 323
Houstis E N, Mitchell W F and Papatheodorou T S 1985 ACM Trans. Math. Software 11 379
Alvarez-Ramirez J, Martinez R and Diaz-Herrera E 1997 Chem. Phys. Lett. 266 375
Warszynski P and Adamczyk 1997 J. Colloid Interface Sci. 187 283
Bowen W R and Sharif A O 1997 J. Colloid Interface Sci. 187 363
Tomac S and Gräslund A 1998 J. Comput. Chem. 19 893
[86] Marcus R A 1955 J. Chem. Phys. 23 1057
[87] Wennerström H, Jönsson B and Linse P 1982 J. Chem. Phys. 76 4665
[88] Bell G M, Levine S and McCartney L N 1970 J. Colloid Interface Sci. 33 335
[89] Oosawa F 1971 Polyelectrolytes (New York: Dekker)
Manning G S 1969 J. Chem. Phys. 51 924
Alexander S, Chaikin P M, Grant P, Morales G J, Pincus P and Hone D 1984 J. Chem. Phys. 80 5776
R586 L Belloni
Ramanathan C V 1988 J. Chem. Phys. 88 3887

Kjellander R and Mitchell D J 1994 J. Chem. Phys. 101 603
[90] Belloni L 1998 Colloid Surf. A 140 227
[91] Again, the presence of the electrolyte (κ = 0) is essential in this situation. If not, the counter-ions which
equilibrate the charge of the two particles would leave the vicinity of the surfaces and occupy the infinite
volume, whatever the strength of the electrostatic attraction (in other words, the entropy always wins over
the energy at infinite dilution), and the force felt by the colloids would be the trivial, coulombic one.
[92] Marcelja S, Mitchell D J, Ninham B W and Sculey M J 1977 J. Chem. Soc. Faraday Trans. II 73 630
Glendinning A B and Russel W B 1983 J. Colloid Interface Sci. 93 95
Carnie S L and Chan D Y C 1993 J. Colloid Interface Sci. 155 297
[93] Hoskin N E 1955 Phil. Trans. A 248 433
Ledbetter J E, Croxton T L and McQuarrie D A 1981 Can. J. Chem. 59 1860
Carnie S L, Chan D Y C and Stankovitch J 1994 J. Colloid Interface Sci. 165 116
[94] Robbins M, Kremer K and Grest G S 1988 J. Chem. Phys. 88 3286
[95] von Grünberg H H and Belloni L 2000 Phys. Rev. E 62 2493
[96] Neu J C 1999 Phys. Rev. Lett. 82 1072
Sader J E and Chan D Y C 1999 J. Colloid Interface Sci. 203 268
[97] Bowen W R and Sharif A O 1998 Nature 393 663
[98] Carbajal-Tinoco M D, Castro-Roman F and Arauz-Lara J L 1996 Phys. Rev. E 53 3745
Larsen A E and Grier D G 1997 Nature 385 230
[99] Hansen J P and Löwen H 2000 Ann. Rev. Phys. Chem. at press
[100] van Roij R, Dijkstra M and Hansen J P 1999 Phys. Rev. E 59 2010
[101] Goulding D and Hansen J P 1999 Europhys. Lett. 46 407
[102] Sanchez-Sanchez J E and Lozada-Cassou M 1992 Chem. Phys. Lett. 190 202
[103] Oosawa F 1968 Biopolymers 6 1633
[104] Sogami I 1983 Phys. Lett. A 96 199
Sogami I and Ise N 1984 J. Chem. Phys. 81 6320
Smalley M V 1990 Mol. Phys. 71 1251
Sogami I S, Shinohara T and Smalley M V 1992 Mol. Phys. 76 1
[105] Overbeek J Th G 1987 J. Chem. Phys. 87 4406
Woodward C E 1988 J. Chem. Phys. 89 5140
Overbeek J Th G 1993 Mol. Phys. 80 685
Rosenfeld Y 1994 Phys. Rev. E 49 4425
Jönsson B, Akesson T and Woodward C E 1996 Ordering and Phase Transitions in Charged Colloids
ed A R Arora and B V R Tata (New York: VCH)
[106] Guldbrand L, Jönsson B, Wennerström H and Linse P 1984 J. Chem. Phys. 80 2221
Valleau J P, Ivkov R and Torrie G M 1991 J. Chem. Phys. 95 520
[107] Kjellander R and Marcelja S 1984 Chem. Phys. Lett. 112 49
Kjellander R and Marcelja S 1985 J. Chem. Phys. 82 2122
Kjellander R and Marcelja S 1986 J. Chem. Phys. 90 1230
[108] Kjellander R 1996 Ber. Bunsenges. Phys. Chem. 100 894
[109] Kjellander R, Akesson T, Jönsson B and Marcelja S 1992 J. Chem. Phys. 97 1424
[110] Lozada-Cassou M 1984 J. Chem. Phys. 80 3344
Lozada-Cassou M and Henderson D 1986 Chem. Phys. Lett. 127 392
Lozada-Cassou and Diaz-Herrera E 1990 J. Chem. Phys. 92 1194
[111] Stevens M J and Robbins M O 1990 Europhys. Lett. 12 81
Tang Z, Scriven L E and Davis H T 1992 J. Chem. Phys. 97 9258
[112] Attard P and Miklavic S J 1993 J. Chem. Phys. 99 6078
[113] Delville A, Pellenq R J M and Caillol J M 1997 J. Chem. Phys. 106 7275
[114] Attard P, Kjellander R, Mitchell D J and Jönsson B 1988 J. Chem. Phys. 89 1664
Podgornik R 1989 J. Chem. Phys. 91 5840
Miklavic S J, Chan D Y C, White L R and Healy T W 1994 J. Phys. Chem. 98 9022
Rouzina I and Bloomfield V A 1996 J. Phys. Chem. 100 9977
Forsman J, Jönsson B and Akesson T 1998 J. Phys. Chem. B 102 5082
[115] Kékicheff P, Marcelja S, Senden T J and Shubin V E 1993 J. Chem. Phys. 99 6098
[116] Patey G N 1980 J. Chem. Phys. 72 5763
[117] Belloni L 1986 Phys. Rev. Lett. 57 2026
[119] Belloni L and Spalla O 1997 J. Chem. Phys. 107 465

[121] Zerah H and Hansen J P 1986 J. Chem. Phys. 84 2336
[122] Rescic J, Vlachy V and Haymet A D J 1990 J. Am. Chem. Soc. 112 3398
Kalyuzhnyi Y V, Vlachy V, Holovko M F and Stell G 1995 J. Chem. Phys. 102 5770
[123] Barrat J L, Hansen J P and Pastore G 1988 Mol. Phys. 63 747
[124] Belloni L, to be published
[125] Groot R D 1991 J. Chem. Phys. 95 9191
[126] D’Amico I and Löwen H 1997 Physica A 237 25
Allahyarov E, Löwen H and Trigger S 1998 Phys. Rev. E 57 5818
[127] Allahyarov E, D’Amico I and Löwen H 1998 Phys. Rev. Lett. 81 1334
[128] Wu J, Bratko D and Prausnitz J M 1998 Proc. Natl Acad. Sci. 95 15 169
Wu J Z, Bratko D, Blanch H W and Prausnitz J M 1999 J. Chem. Phys. 111 7084
[129] Tata B V R, Rajalakshmi M and Arora A K 1992 Phys. Rev. Lett. 69 3778
[130] Palberg T and Würth M 1994 Phys. Rev. Lett. 72 786
[131] Bloomfield V A 1991 Biopolymers 31 1471
Leikin S, Parsegian V A, Rau D C and Rand R P 1993 Annu. Rev. Phys. Chem. 44 369
Sikorav J L, Pelta J and Livolant F 1994 Biophys. J. 67 1
Tang J X, Wong S, Tran P and Janmey P 1996 Ber. Bunsenges. Phys. Chem. 100 1
[132] Olvera de la Cruz M, Belloni L, Delsanti M, Dalbiez J P, Spalla O and Drifford M 1995 J. Chem. Phys. 103
5781
[133] Mahanty J and Ninham B W 1976 Dispersion Forces (New York: Academic)
[134] Hunter R J 1987 Foundations of Colloid Science vol 1 (Oxford: Clarendon) ch 4
[135] Ninham B W and Parsegian A V 1972 see chapter 7 of reference [133]
[136] Attard P, Kjellander R and Mitchell D J 1987 Chem. Phys. Lett. 139 219
Kjellander R and Marcelja S 1987 Chem. Phys. Lett. 142 485
Attard P, Mitchell D J and Ninham B W 1988 J. Chem. Phys. 89 4358
[137] Levin Y 1999 Physica A 265 432
[138] Ninham B W, Kurihara K and Vinogradova O I 1997 Colloid Surf. A 123 7
Ninham B W and Yaminsky V 1997 Langmuir 13 2097
[139] Spalla O 2000 Current Opinion Col. Interf. Sci. 5 5

Belloni 2000

Uploaded by

Copyright:

Available Formats

Belloni 2000

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Belloni 2000

Uploaded by

Copyright:

Available Formats

Home Search Collections Journals About Contact us My IOPscience

2000 J. Phys.: Condens. Matter 12 R549

Please note that terms and conditions apply.

Received 31 March 2000, in final form 27 September 2000

0953-8984/00/460549+39$30.00 © 2000 IOP Publishing Ltd R549

2. Direct, mean force, effective potentials of interaction

2.1. Monodisperse systems

2.2. Mixtures and McMillan–Mayer theory of solutions

2.3. Potentials of mean force

2.4. Effective, solvent-averaged, pair potential

The corresponding expression using the force route (17) is

3. Colloidal interactions in neutral systems

3.1. Hard-sphere mixtures: depletion and structural forces

Figure 2. Schematic representation of excluded volume correlations in asymmetrical hard-sphere

≡ V1 ∩ V2 + V1 ∩ V3 + V2 ∩ V3 − V1 ∩ V2 ∩ V3 and, due to the extra, three-body term, is

3.2. Colloid–solvent repulsion: enhanced, longer-range, depletion

3.3. Solvent–solvent repulsion: oscillating structural force

3.4. Colloid–solvent attraction: steric, stabilizing repulsion + bridging attraction

3.5. Solvent–solvent attraction: modulation of steric and hydration forces

4. Colloidal interactions in charged systems

4.1. Debye–Hückel theory

4.2. Improved DH-like theories

eff Zc2 LB e−κ(r−2a)

4.3. Poisson–Boltzmann theory

Dielectric discontinuities at the colloidal surfaces, |r − rc | = a, can be taken into account

4.4. Beyond DLVO—effective attraction between colloids

The existence of an effective attraction of pure electrostatic origin between like-charged

5. Van der Waals and miscellaneous interactions

[8] Sonneville-Aubrun O, Bergeron V, Gulik-Krzywicki T, Jönsson B, Wennerström H, Lindner P and Cabane B

Ramanathan C V 1988 J. Chem. Phys. 88 3887

[119] Belloni L and Spalla O 1997 J. Chem. Phys. 107 465

You might also like