J. Fluid Mech. (2000), vol. 402, pp. 57–88.

Printed in the United Kingdom 57

c 2000 Cambridge University Press

Contact-line dynamics of a diffuse

fluid interface
By D A V I D J A C Q M I N
NASA Lewis Research Center, Cleveland, OH 44135, USA

(Received 20 November 1997 and in revised form 25 July 1999)

An investigation is made into the moving contact line dynamics of a Cahn–Hilliard–

van der Waals (CHW) diffuse mean-field interface. The interface separates two in-
compressible viscous fluids and can evolve either through convection or through
diffusion driven by chemical potential gradients. The purpose of this paper is to show
how the CHW moving contact line compares to the classical sharp interface contact
line. It therefore discusses the asymptotics of the CHW contact line velocity and
chemical potential fields as the interface thickness  and the mobility κ both go to
zero. The CHW and classical velocity fields have the same outer behaviour but can
have very different inner behaviours and physics. In the CHW model, wall–liquid
bonds are broken by chemical potential gradients instead of by shear and change of
material at the wall is accomplished by diffusion rather than convection. The result
is, mathematically at least, that the CHW moving contact line can exist even with
no-slip conditions for the velocity. The relevance and realism or lack thereof of this
is considered through the course of the paper.
The two contacting fluids are assumed to be Newtonian and, to a first approxi-
mation, to obey the no-slip condition. The analysis is linear. For simplicity most of
the analysis and results are for a 90◦ contact angle and for the fluids having equal
dynamic viscosity µ and mobility κ. There are two regions of flow. To leading order
the outer-region velocity field is the same as for sharp interfaces (flow field indepen-
dent of r) while the chemical potential behaves like r−ξ , ξ = π/2/max{θeq , π − θeq },
θeq being the equilibrium contact angle. An exception to this occurs for θeq = 90◦ ,
when the chemical potential behaves like ln r/r. The diffusive and viscous contact line
singularities implied by these outer solutions are resolved in the inner region through
√
chemical diffusion. The length scale of the inner region is about 10 µκ –typically
about 0.5–5 nm. Diffusive fluxes in this region are O(1). These counterbalance the
effects of the velocity, which, because of the assumed no-slip boundary condition,
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

fluxes material through the interface in a narrow boundary layer next to the wall.
The asymptotic analysis is supplemented by both linearized and nonlinear finite
difference calculations. These are made at two scales, experimental and nanoscale.
The first set is done to show CHW interface behaviour and to test the qualitative
applicability of the CHW model and its asymptotic theory to practical computations
of experimental scale, nonlinear, low capillary number flows. The nanoscale calcu-
lations are carried out with realistic interface thicknesses and diffusivities and with
various assumed levels of shear-induced slip. These are discussed in an attempt to
evaluate the physical relevance of the CHW diffusive model. The various asymptotic
and numerical results together indicate a potential usefullness for the CHW model
for calculating and modelling wetting and dewetting flows.
 58 D. Jacqmin
1. Introduction
The idea that a fluid–fluid interface is diffuse, that is, has finite thickness, goes
back to Poisson (1831) and Gibbs (1876). Recently, diffuse interface modelling of
multi-phase mixtures has become fairly popular, primarily because of its usefulness
as a numerical device (Brackbill, Kothe & Zemach 1992; Unverdi & Tryggvason 1992;
Lafaurie et al. 1994; Sussman, Smeraka & Osher 1994; Kothe et al. 1996) but also as
a means to gain insight into the small-scale physics of fluids (Anderson & McFadden
1997). A review of much of this recent work is given by Anderson, McFadden &
Wheeler (1998).
Van der Waals (1893) introduced the first and probably still most used diffuse
interface model. Following Cahn (1961), it has been extensively used in the study
of the diffusion-dominated evolution of interfaces. The model’s application to fluid
dynamics has been recently proposed and discussed by Antanovskii (1995), Gurtin,
Polignone & Viñals (1996), Nadiga & Zaleski (1996), Anderson & McFadden (1997),
and Jacqmin (1996), among others, and it has been used in numerical calculations
of fluid mechanical problems by Jacqmin (1995, 1996), Chella & Viñals (1996), and
Jasnow & Viñals (1996).
This discussion in the main part of this paper is restricted to incompressible fluids.
The CHW (Cahn–Hilliard–van der Walls) model for this case takes the local free
energy density in a two-component fluid to be a function not only of composition but
also a quadratically increasing function of the local composition gradient. Infinitely
sharp interfaces are therefore forbidden. Van der Waals introduced the model in
order to discuss equilibrium fluid–fluid interfaces and to illustrate how and why these
interfaces have surface tension. The width, profile and surface tension of a CHW
interface in equilibrium is determined by the minimization of the free energy of the
interface.
If a CHW interface is not in equilibrium then two mechanisms exist to equilibrate
it. The first mechanism is diffusional: gradients of chemical potential can induce
countercurrents of the fluid’s two components. These can bring an interface to its
proper profile and, at micron-scale curvatures and below, even move an interface.
The second mechanism is advective: gradients of chemical potential produce surface
tension forcing of fluid momentum. The CHW model is similar to and a simplification
of van der Waals mean-field models of intermolecular forcings. For example, like
mean-field models based on Lennard–Jones integrals, the CHW model can be used
to model disjoining pressure effects. Because the CHW model includes diffusion it
can be used not only to examine interfacial instabilities due to disjoining pressures
but also any subsequent dissolution/solution of one fluid into the other.
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

A potential major application of continuum interface models is to the resolution of

the various flow singularities that arise from the classical sharp-interface treatment of
fluids. These include topological/stress singularities arising from droplet coalescences
and breakups (Lowengrub & Truskinovsky 1998). These problems are of basic physical
interest in themselves but they are also highly relevant to the question of the accuracy
of numerical methods that use diffuse interfaces.
This paper examines the applicability of the CHW model to moving contact
lines. Two points of view are taken, corresponding to two different meanings of
‘applicability’. The first question is: is the CHW model useful? The second question
is: are results from the CHW model relevant to real contact line physics? The answers
to both questions depend heavily on how the CHW model handles and resolves
the well-known inner-region contact-line singularity. Much of this paper therefore
 Contact-line dynamics of a diffuse fluid interface 59
concentrates on elucidating the asymptotic, small-scale structure of the CHW contact
line region.
The analysis in the paper is for a two-component, two-phase, incompressible fluid
system that obeys the Navier–Stokes equations with Newtonian stresses and that, to a
first approximation, obeys the no-slip boundary condition on walls. The usual contact
line model takes the fluid–fluid interface to be sharp and to be a material interface,
i.e. there is no flux through it. The further imposition of the no slip boundary condition
then leads to a contradiction: the flow field at the contact line is multi-valued and
the total shear force is logarithmically singular.
The obvious path to resolve the contradiction is to allow some slip, in, for example,
a small region surrounding the contact line, or equal to wall shear stress times a
small parameter. The physical import of shear-induced slip, as pointed out by Durbin
(1988), is that viscous transfer of momentum can break fluid–wall molecular bonds,
allowing the contact line to move.
The CHW model allows a different resolution. This resolution is possible – at least,
mathematically – even when perfect no-slip is assumed. The CHW interface is not a
material line and it can be moved diffusively as well as by advection. With no-slip,
Cahn–Hilliard chemical potential gradients right at the moving contact line become
such that diffusive fluxes there are O(1). These fluxes are competitive with convective
transport. Together with a diffusive wall boundary condition they allow the contact
line to be moved while simultaneously maintaining the diffuse interface’s form and
stability. The physical import of the diffusive boundary condition is that fluid–wall
molecular bonds are broken by potential energy related forces.
The CHW model appears to hold great promise (Jacqmin 1999) for numerical
applications; its diffuse interface allows the use of numerical methods that are more
general and powerful, though admittedly less accurate, than are available for sharp
interfaces. It can also be useful for analytic studies of two-phase flows involving
diffusive interfacial mass transfer, nucleation or dissolution, or disjoining pressures.
The question of its usefulness for moving contact line analyses comes down to the
details of how the CHW model resolves the contact line singularity. The first sub-
question is whether the resolution leads to correct macroscopic results. As expected
from Cox’s (1986) analysis, the answer is yes. The second sub-question is how
this resolution behaves asymptotically. This must be known in order to understand
and test non-asymptotic numerical or analytic CHW contact-line results or, when
developing numerical methods, to establish paths to numerical convergence. The
paper discusses some experimental-scale numerical calculations to show how results
from the asymptotic theory can be useful in understanding finite Reynolds number,
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

macroscopically thick, numerical interfaces.

The question of the physical relevance of the CHW diffusive resolution of the
contact-line singularity is difficult. It seems certain that some such effects must occur.
The high curvature of the interface near the contact line should affect the phase
behavior, meaning the mutual solubility and the density, of the fluids there. The
amelioration of high curvatures through diffusion is well known in crystal growth and
equation systems similar to the Cahn–Hilliard equation have had success in modelling
it. However, it turns out that for moving contact lines these effects become important
only at very small scales. The diffusive length scale of the CHW contact line will be
√
shown to be µκ, where µ is dynamic viscosity and κ is mobility. For a 90◦ equilibrium
contact angle the inner region – the region where the flow field is significantly affected
by chemical potential gradients – is about ten times this length scale. κ in liquids is
typically about 10−14 , in units of cm5 s−1 erg−1 . This results in a diffusive length scale
 60 D. Jacqmin
of somewhere between 1 Å and 1 nm and an inner region of 1–10 nm. The size of
small molecules is about 3 Å and interface thickness itself is about 1 nm. The CHW
theory applied to moving contact lines may thus be mildly inconsistent in that to
the degree it is applicable, or, at least, pointing in the right direction, it also calls
for a more careful consideration of subcontinuum effects, particularly the details of
wall–liquid molecular interactions, that it does not include. Also, at these scales it
may not be reasonable to assume shear stress/slip effects are negligible. An attempt
will be made in § 11 to evaluate chemical potential/diffusive versus shear stress effects
with some nanoscale calculations. These have been made with realistic diffusivities
and interface thicknesses and with various assumed levels of shear-induced slip.
Previous work in this area has been done by Seppecher (1996), who has presented
an analysis and some microscale calculations of diffuse contact lines in a model com-
pressible fluid. The Appendix discusses his work further. Works related to chemical
potential forcing of contact line flows include de Gennes, Hua & Levinson (1990),
who discuss the effect of van der Waals forces on low- or high-angle (near 0◦ or
180◦ ) moving contact lines; Blake (1993), who hypothesizes that activation free en-
ergies involved in breaking wall–liquid bonds are important in setting wetting rates;
Miller & Ruckenstein (1974) and Lopez, Miller & Ruckenstein (1976), who discuss
and evaluate van der Waals forcings of wetting fluid flows, as do Li & Slattery
(1991); Indeikina & Chang (1998), who discuss van der Waals forcings in contact line
movement of partially wetting fluids; and Wayner & Schonberg (1992), who discuss
spreading of a liquid film through evaporation and condensation caused by chemical
potential differences.
The following analysis and calculations will use the simplest possible Navier–
Stokes/CHW diffuse-interface flow model; they use a single composition variable
and, as mentioned above, the equations are incompressible. These simplifications
result in the neglect of some important effects. First, it would be better to represent
the system using two composition variables. Each could then diffuse independently
and each could vary independently in interfacial regions. Second, the neglect of
compressibility means that the chemical potential is a function only of mole or mass
fractions rather than densities. This analysis is most most likely to be physically
relevant to liquid–liquid systems where there is some degree of solubility between the
two components and where the solubility is a function of chemical potential. Also,
because of air’s relatively high volumetric solubility and diffusivity in liquids, it may
have qualitative relevance to air/liquid systems.
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

2. The diffuse interface Navier–Stokes equations

An interface’s thickness and the composition profile through the interface are set by
the competition between random molecular motion (thermal motion) and molecular
attractions. Van der Waals (1893) modelled this competition by introducing a free
energy density dependent on both composition and composition gradients:
f = 12 σα|∇C|2 + −1 σβΨ (C). (2.1)
Here, f is the free energy per volume and C is a ‘measure’ of phase. The first term
in (2.1) is the energy density due to phase gradients and the second is the bulk
energy density. Ψ (C) models the fluid components’ immiscibility and has two minima
corresponding to the fluid’s two stable phases. Van der Waals hypothesized R that
equilibrium interface profiles are those that minimize the free energy F = f dV .
 Contact-line dynamics of a diffuse fluid interface 61
From the calculus of variations, these profiles satisfy
σα∇2 C − −1 σβΨ 0 (C) = −φ = const.; (2.2)
where φ ≡ δF/δC is the fluid’s chemical potential. Equation (2.2) yields interfaces
with surface tension σ and interface thickness ; α and β are dimensionless O(1)
constants determined by the choice of Ψ and by a chosen definition of interface
thickness. The surface tension is equal to the integral of the free energy density
through the interface per unit area. For a flat interface with equilibrium profile
C = C(y) the surface tension is given by the integral
Z +∞  2
∂C
σ = σα dy. (2.3a)
−∞ ∂y eq
In other words,
Z +∞  2
∂C
α dy = 1. (2.3b)
−∞ ∂y eq
This determines α from a given profile.
Cahn (1961) extended van der Waals’ hypothesis to time-dependent situations by
approximating interfacial diffusion fluxes as being proportional to chemical potential
gradients. The Cahn–Hilliard equation
∂C
= κ∇2 φ = −κ∇2 (σα∇2 C − −1 σβΨ 0 (C)) (2.4)
∂t
models the creation, movement, and dissolution of diffusively controlled phase inter-
faces (Bates & Fife 1993); κ is called the mobility. The bulk phase chemical diffusivity
is given by −1 σβκΨ 00 (Cbulk phase ). The free energy of a system determined by (2.4)
decays monotonically:
Z
dF
= −κ (∇φ)2 dV . (2.5)
dt
The further extension of the van der Waals model to advection problems has been
recently carried out by, among others, Antanovskii (1995), Jacqmin (1995, 1996),
Chella & Viñals (1996), Nadiga & Zaleski (1996), and Anderson & McFadden
(1997). Advection can change free energy but must leave the total energy of a fluid
unchanged. The compensatory forcing of momentum by the free energy field must
therefore be Fi = −δF/δui = −C∇φ (Jacqmin 1996, 1997). A CHW two-phase,
viscous, incompressible fluid obeys
DC
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

= κ∇2 φ = −κ∇2 (σα∇2 C − −1 σβΨ 0 (C)), (2.6)

Dt
Dui
ρ(C) = −∇S + ∇ · τ visc,i − C∇φ, (2.7)
Dt
X ∂ui
= 0. (2.8)
i
∂xi
This gives the potential form of the surface tension forcing. S enforces incompress-
ibility. (However, it is not the pressure.) The τ visc,i are the viscous stresses. A stress
form of the surface forcing also exists. The momentum equations become
Dui X ∂τij
ρ(C) = −∇p + ∇ · τ visc,i + , (2.9)
Dt j
∂xj
 62 D. Jacqmin
where the surface stresses are
X  ∂C 2
τjj = σα , (2.10)
i6=j
∂xi
 
∂C ∂C
τij,i6=j = −σα , (2.11)
∂xi ∂xj
and p is the pressure.
The viscous boundary condition used in most of this paper and for all the asymp-
totic theory is no slip. Two boundary conditions are needed for C. The no-flux
condition for C is
∂φ
= 0. (2.12)
∂xn
The second boundary condition depends on the interface at the wall being at or near
local equilibrium. Postulating that the wall free energy is of the form
Z
Fw = γg(C) dA, (2.13)

that is, that the wall–fluid interfacial energy is a function only of the the fluid
composition right against the wall, then the resulting natural boundary condition,
which corresponds to a diffusively controlled local equilibrium at the wall, is
∂C
σα + γg 0 (C) = 0, (2.14)
∂xn
where xn is the direction normal to and into the wall. Equation (2.14) is analogous
to the classical contact angle condition in which the dynamic contact angle right
at the wall (the microscopic contact angle) is taken to be the same as the static
equilibrium angle. When g(C) is zero the equilibrium contact angle is 90◦ . A more
general condition that allows non-equilibrium is
 
DC ∂C 0
= Dw σα + γg (C) . (2.15)
Dt ∂xn
This results in the microscopic contact angle being a function of wall velocity. C
approaches the equilibrium condition as Dw increases to infinity. In the following
analysis Dw will be assumed to be at least O(1/). Then σαDw is at least O(1)†
and perturbations to the microscopic contact angle caused by low-capillary-number
contact line movement are, it will be shown, O(Ca), where Ca is the capillary number,
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

or smaller.
At this point, some discussion is needed about the physical meaning of the no-slip
condition. The reader has already been alerted (§ 1) that diffusive fluxes along the
wall in the vicinity of a moving contact line will be found to be O(1). It may thus
seem unreasonable that convection there is forced in this analysis to equal the wall
speed. Since the diffusion ‘flow’ seems to slip, why not the velocity?
† The assumption is justified by recent molecular dynamics calculations of fluid equilibration
against walls (Matsumoto, Maruyama & Saruwatari 1995; Wagner 1995) that indicate fluid–wall
equilibration times that are sub-nanosecond. Since σα∂C/∂xn and γg 0 (C) are each O(1), matching
this in (2.15) requires a Dw of about 109 . Fluid–fluid interfaces away from critical points are typically
nanometers wide, indicating an σα of about 10−7 . The choice σαDw > O(1) thus seems relevant
and physically reasonable. An exception to this general argument might be when the wall bonds of
the displaced fluid are very strong and thus, perhaps, slow to break.
 Contact-line dynamics of a diffuse fluid interface 63
A partial answer to this is that the diffusive flux is not a coherent, macroscopic
flow but is better viewed – at least for a small-molecule liquid–liquid system – as being
made up of numbers of microscopic-scale interchanges of molecular positions. The
result is two opposing countercurrants, one of one material, one of the other. At the
wall itself, these interchanges can be modelled as site interchanges: a molecule of one
fluid phase exchanges a wall site with a molecule of another phase, or a molecule
in the interior of the fluid grabs a wall site from a fluid molecule attached to the
wall (Blake 1993). These interchanges will tend to be energy downgradient, and their
general direction will be in response to changes in the surrounding fluid environment.
The energy required to make the interchanges is provided by random thermal motion.
In contrast, a slip velocity would involve a large-scale breaking of bonds with no
compensating release of free energy, as most of the convection would involve the
replacement at wall sites of like molecules by like. The exception to this is if slip
occurs only in the immediate vicinity of the contact line.
A hybrid model (Blake) is chemical-potential-gradient-induced slip:
!
∂C ∂C
(uj − Uj ) ∝ σα + γg 0 (C) , (2.16)
∂xn ∂xj

where Uj is the jth velocity component of the wall and uj − Uj is the slip velocity.
This slip velocity is proportional to the wall shear stress caused by chemical potential.
This allows slip only in the interface itself, is thus of very small scale, and would be
energy efficient.
The coexistence of no slip with along-wall diffusion is supported by at least two
molecular dynamics studies. Thompson & Robbins (1990) have carried out molecular
dynamic calculations of very small-scale Couette flow driven by crystalline walls. They
found, when wall molecule spacing and fluid molecule/fluid molecule separations are
roughly commensurate, that no slip occurs at a fairly low wall–fluid interaction energy.
At low and moderate wall-fluid energy the diffusion rate along the wall is essentially
the same as in the interior. They found that a high liquid–wall interaction energy (1.8
times that of the liquid–liquid interaction) was necessary before the near-wall liquid
solidified and self-diffusion was cut off. Koplik & Banavar (1998) have looked at
two-component nanoscale plane Poiseuille flow. The two components were miscible.
Though one of the components had a fairly weak wall interaction, they found that
each material individually obeyed the no-slip condition. They wrote: ‘The molecular
origin of no slip is that fluid molecules near the wall are, on the one hand, dragged
along in the direction of the net flow by their neighbors further away from the wall,
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

and, on the other hand, are pushed up against the wall by the crowding of molecules
in a dense liquid, where they interact with the effectively corrugated potential of the
wall. While the molecules are not literally stuck there, their translation speed is much
reduced as they interact with the almost-fixed wall atoms, and a typical trajectory is
a random walk with a drift in the direction of the applied force.’
Following Durbin (1988), it is possible to estimate the shear required for contact-
line slip from (more-or-less) first principles. This is done by finding the rate of work
needed to break or partially break wall–liquid bonds. Wall–liquid bonds have roughly
the same strength as liquid–liquid bonds, which in turn have about the same energy
as fluid–fluid surface energy (which indicates the energy needed for breaking of these
bonds). A typical wall–liquid bonding energy is from about 5–50 erg cm−2 . The
estimate of shear needed assumes that any slip would involve near-wall molecules
that would be generally trapped close to the wall by other liquid molecules and that,
 64 D. Jacqmin
as a result, their slip would require repeated movement from favourable near-wall
locations to unfavourable ones and back again. Each traversal of an energy barrier
would then give an energy dissipation that would be a significant fraction of the wall–
liquid bond strength–that is, it would essentially involve the partial breaking and then
remaking of a wall–liquid bond, in the course of this turning some work into heat. It
also assumes that wall molecule spacing and fluid molecule/fluid molecule separations
are roughly the same. The power per area needed to continuously break wall–liquid
bonds in this way is λσlw Vwall /4 × 10−8 , where σlw is the wall–liquid bonding energy
and λ is the fraction of wall–liquid bond energy that is dissipated at each traversal.
The assumed intermolecular spacing is 4 × 10−8 cm and Vwall /4 × 10−8 is the rate
of wall–liquid bond breaking at a particular location per second. The rate of work
2
per area done by shear stress is approximately equal to µVwall /L, where L is the slip
length (the near-contact-line inner scale) at the contact line and µ is the dynamic
viscosity. The work done and energy needed must be equal, giving an L of about
4 × 10−8 µVwall /λσlw = 4 × 10−8 (1/λ)(σ/σlw ) Ca. For a σ/σlw of about 2, a fairly typical
liquid–liquid capillary number of about 0.01, and assuming a λ of about 0.05, this
gives a slip length of about 2 Å.
This estimate appears to be in good agreement with molecular dynamics calcu-
lations of contact line flow. Results of Koplik, Banavar & Willemsen (1989) and
Thompson & Robbins (1989) both indicate barely distinguishable, sub-continuum
slip regions. Both these calculations are of liquid–liquid flow at Ca ' 0.01, with fairly
strong wetting of the wall by both fluids.
Along the same lines, though in a different direction, Barrat & Bocquet (1999)
have recently shown that non-wetting liquids can show much larger slip effects. This
is because of reduced levels of wall–liquid bonding. They carried out molecular
dynamics calculations of plane Poiseuille and Couette flow of liquids between walls
for which the liquids had equilibrium liquid/vapour contact angles greater than 90◦ .
They found slip lengths of about 6 Å for 100◦ contact angle fluid and about 3 nm for
a 145◦ contact angle.
Section 11 will consider slip further. Nanoscale calculations will be discussed that
include slip. Comparisons will be made between shear stresses and chemical potential
forces at the contact line to try to determine when each is dominant.

3. Experiment-scale calculations of CHW contact lines

The purpose of this section is to motivate the asymptotic analysis by first showing
the reader some ‘practical’ experimental-scale computational results. The interfaces
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

in these calculations are much wider than in reality but calculations using them
can nevertheless be qualitatively useful and they show the main features of diffuse-
interface asymptotic behaviour. The calculations are of the full nonlinear equations
(2.6)–(2.8) with Newtonian stresses and using the no-slip boundary condition. Some
of the important characteristics of CHW contact lines that are shown in this section
are (i) the CHW moving contact line is stable and stops most through-flow even with
no-slip conditions, (ii) through-flow is reduced as the interface becomes thinner, (iii)
the chemical potential field scales differently from interface thickness, and (iv) the
distortion of the interface profile due to convection becomes smaller as the interface
becomes thinner.
Even very crude calculations of the CHW diffuse-interface model seem to give
the essentials of what is known about dynamic contact lines. Figure 1 shows an
axisymmetric ‘gas’–liquid dynamic contact line in a steady frame of reference. The
 Contact-line dynamics of a diffuse fluid interface 65

Figure 1. Steady flow pattern at a diffuse-interface contact line. The upper fluid has
one-hundredth the density and one-thousandth the viscosity of the lower fluid.

‘gas’ has a specific density of 0.01, the liquid 1.00; the gas has one-thousandth the
viscosity of the liquid. The figure shows contours of the diffuse interface with the
velocity vector field superimposed; g(C) was set equal to zero. Equation (2.14) was
used for the C boundary condition so the meniscus right at the wall comes in at the
equilibrium 90◦ angle.
What is remarkable is the way the diffuse interface forces seem to almost completely
stop the downward motion of fluid just at the contact line. The computation was
made using a staggered grid so the leftmost vertical velocity that is shown is actually
1/2 grid cell from the true boundary. The computation is no slip in the sense that
the assigned wall velocity is used to compute the viscous shear forces on the leftmost
velocity. This condition is not relaxed at the meniscus. To within the resolution of
the computation the vertical flow in the cells next to the contact line appears to stop,
but just above and below the meniscus the flow is very close to wall speed.
The flow away from the contact line agrees in character with the wedge flow
solution of Huh & Scriven (1971). The gas, which is the upper fluid, is swept down
with the moving boundary but when it reaches the contact line it is forced out in a
jet at about a 30◦ angle to the meniscus. The viscous liquid flow is entrained along
the meniscus into the wall flow. The reversal of flow from outgoing gas to ingoing
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

liquid takes place in a thin boundary layer in the gas just above the meniscus.
Figures 2–5 show a set of calculations of a liquid–liquid meniscus in a driven cavity.
The calculations are of the full nonlinear equations. In order to isolate surface tension
effects, the two liquids were given the same density and viscosity. For symmetry, the
driven cavity was given two moving walls: the left wall is downward moving, the
right wall upward. The cavity size is 2 by 6 cm, wall speed is 0.4 cm s−1 , the surface
tension is 30 erg cm−2 , the viscosity is 1 P (1 g cm−1 s−1 ), and the specific gravity
is 1. The capillary number, µVwall /σ, of the flow is 0.01333. Ψ (C) is modelled as
(C − 1/2)2 (C + 1/2)2 . The equilibrium interface for this has a tanh profile and
 is defined to be the distance through the interface from C = −0.45
to +0.45.
From Ψ , equation (2.3b) and this definition, α = 3/ 2 tanh−1 (0.9) ' 1.019 and
β = 12 tanh−1 (0.9) ' 17.667. The equation for the equilibrium profile is C = 12 tanh(ζ),
 66 D. Jacqmin

(a) (b) (c)

Figure 2. The meniscus for the double-contact-line computation for three meniscus widths. Contours
of C are shown, for C from −1/2 to +1/2, at intervals of 0.01. The fluids above and below the
interface have identical densities and viscosities. The perturbation to interface structure becomes
smaller as the interfaces become thinner.

p
where ζ = −1 β/2α y = 2−1 (tanh−1 0.9) y. Figure 2 shows the meniscus for three
different interface widths: the width in (b) is half that of (a), and in (c) half that of
(b);  for the thickest interface is 0.6475 cm. The mobility κ was kept proportional to
the interface width so it too gets successively smaller. It is 2.65 × 10−5 cm5 s−1 erg−1
for the thickest interface. The equilibrium condition (2.14) with g 0 = 0 was used for
the C boundary condition.
Comparing figures 2(a)–2(c) we see that the interface becomes noticeably less
distorted as it becomes thinner. The source of the distortion is that there is some
flow through the meniscus next to the wall and this tends to drag the meniscus down
with it. But chemical diffusion limits the effects of this drag–the molecules of the
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

upper fluid migrate diffusively back to it. The interface is dragged down less as the
interface thins because diffusion, with its time scale proportional to O(κ/3 ), becomes
increasingly dominant over convection with its time scale of O(−1 ).
Figure 3 compares menisci centrelines. These are nearly identical, with a slight
increase of slope as the interface becomes thinner. The menisci are nearly straight
except right at the contact lines, where they curve in very sharply to the equilibrium
angle. The averaged slope of the thickest meniscus is 0.2842: the medium meniscus
has a slope of 0.2873, the thinnest a slope of 0.2972. The maximum slope (at the box
midpoint) of the thickest meniscus is about 0.331, of the thinnest about 0.336.
Figure 4 shows the stream function for each calculation. This shows some flow
through the menisci right at the contact lines but this flow becomes much narrower
and has much less volume as  and κ are reduced. The coarsest calculation also
 Contact-line dynamics of a diffuse fluid interface 67
0.3
(a)
0.2
Meniscus height

0.1

– 0.1

– 0.2

– 0.3
0 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
x
– 0.15
(b)
Meniscus height

– 0.20

– 0.25

– 0.30
0 0.125 0.250 0.375 0.500
x
Figure 3. The meniscus centreline for the menisci shown in figure 2. The solid line is for the thickest
meniscus, the dashed line for the intermediate, and the dashed-dotted line for the thinnest. (b) A
closeup at the left contact line.

shows a noticeable amount of return flow. This return flow is also present in the finer
calculations.
Figure 5 shows the chemical potential field for each calculation. The field shows a
‘half-dumbbell’ pattern about the contact lines but the size and intensity of this pattern
changes with  and κ. It roughly halves in size between (a) and (c). The chemical-
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

potential length scale thus decreases more slowly than the interface thickness. The
maximum value of the potential roughly doubles, from 84.1 to 187.7 erg cm−3 . The
length
√ scale of the near-contact-line
√ potential field is approximately proportional to
κ and its intensity to 1/ κ. The diffusive fluxes at the contact line, which are
proportional to κ times the magnitude of the potential divided by the length scale,
have approximately the same maximum intensity for all three interface widths.
Figure 6 shows the fine calculation potential in detail. The potential pattern is
centred on the contact line (not shown). As the contact line is approached the
potential pattern is increasingly dominated by diffusive as opposed to convective
effects; hence its increasing symmetry. The dumbbell pattern will be shown later to
correspond to a logarithmic term in the asymptotic solution that has zero amplitude
in the meniscus region but is large at the walls.
 68 D. Jacqmin
(a) (b) (c)

Figure 4. The stream function for the menisci shown in figure 2: (a) for the thickest meniscus
shows the most through-flow at the moving wall.

4. The linearized equations for 90◦ contact angle

We now seek to develop an analytic theory for the CHW contact line. The case of
a 90◦ equilibrium contact angle will be investigated first. To connect results to sharp-
interface Navier–Stokes theory the analysis is for  and κ asymptotically approaching
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

0. The situation that will be considered is that of a semi-infinite horizontal interface

bounded by a vertical wall. The wall is moving downward at a very low speed. It is
assumed, following Cox (1986), that the resulting flow can be analysed using linearized
equations and that the wall velocity is low enough so that the asymptotics can be
done using a two-region analysis. We seek to find the linearized steady-state solution
for the resulting moving contact line flow, for the chemical potential, the interface
slope, and for the perturbation to diffuse interface structure.
In order to analyse this dynamics in as simple a way as possible we consider, as
in § 3, a two-phase system in which the phases have the same viscosity and density.
They are contained in a half-plane bounded by a vertical wall located at x = 0. This
and the next few sections will consider the case of g(C) constant. The equilibrium
contact angle is then 90◦ and the equilibrium fluid interface obeys the one-dimensional
 Contact-line dynamics of a diffuse fluid interface 69

(a) (b) (c)

Figure 5. The chemical potential for the menisci shown

√ in figure 2. The size of a √chemical
potential pattern near a contact line scales with κ, its intensity scales as 1/ κ.

equation
d2 C̄
σα − −1 σβΨ 0 (C̄) = 0. (4.1)
dy 2
The interface is centred on y = 0. The analysis will be for general Ψ (C̄) except that
Ψ is symmetric about C̄ = 0 and scaled so that C̄(±∞) = ±1/2. Because of Ψ ’s
symmetry C̄ is antisymmetric about y = 0.
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

Now consider the perturbation to this interface caused by the wall being set
into very slow, steady, downward motion. The linearized, steady equations for this
perturbation are
κ∇2 φ = C̄y v, (4.2)
∂S ∂φ
µ∇2 u = + C̄ , (4.3)
∂x ∂x
∂S ∂φ
µ∇2 v = + C̄ , (4.4)
∂y ∂y
∂u ∂v
+ = 0, (4.5)
∂x ∂y
 70 D. Jacqmin

Figure 6. A close-up at the left-hand contact line of the chemical potential shown in figure 5(c).

σα∇2 C − −1 σβΨ 00 (C̄)C = −φ, (4.6)

∂φ
u = 0, v = −Vwall , =0 at x = 0, (4.7)
∂x
∂C
σαDw = C̄y v(x = 0) = −C̄y Vwall ; (4.8)
∂x x=0
C and φ indicate perturbation quantities. Base-state quantities are indicated by
overbars; µ is the dynamic viscosity.
Equations (4.2)–(4.8) decouple. The equations (4.2)–(4.5), (4.7) for the perturbation
potential and the velocities are self-contained. C can be found separately from (4.6)
and (4.8) after (4.2)–(4.5), (4.7) have been solved. Introducing the stream function ψ,
∂ψ/∂x = v, ∂ψ/∂y = −u, (4.9)
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

the non-dimensionalization
√ √ p
{x, y} → µκ{x, y}, ψ → µκVwall ψ, φ → µ/κVwall φ, (4.10)
and noting that C, φ and ψ are symmetric about y = 0, the C and momentum
equations become
∂ψ
∇2 φ = C̄y , (4.11)
∂x
∂φ
∇2 ∇2 ψ = −C̄y , (4.12)
∂x
with the x = 0 boundary conditions:
∂ψ ∂φ
ψ = 0, = −1, = 0, (4.13)
∂x ∂x
 Contact-line dynamics of a diffuse fluid interface 71
and the y = 0 symmetry conditions:
∂ψ ∂3 ψ ∂φ
= = = 0; (4.14)
∂y ∂y 3 ∂y
x, y, ψ, and φ now denote scaled quantities and C̄y denotes the derivative of C̄ with
respect to scaled y.

√
5. The asymptotic form of the {ψ, φ} equations as / µκ → 0
√
We now narrow consideration to the particularly simple case of / µκ → 0. The
√
case / µκ = O(1) has the same asymptotic structure and gives essentially the same
results. Adjustments needed to deal with this case will be discussed in § 8.
√
In the transformed coordinates the interface has thickness O(/ µκ). This goes
to zero as  and κ go to zero provided that κ is O(δ ), δ < 2. The quantities
φ and ψ are coupled only in the interfacial layer. If φ and ψ are slowly enough
varying then the forcing terms C̄y {∂ψ(x, y)/∂x, −∂φ(x, y)/∂x} can be approximated
√
by C̄y {∂ψ(x, 0)/∂x, −∂φ(x, 0)/∂x}. In the limit of / µκ → 0 these forcings become
delta functions. Integrating (4.11) and (4.12) through the layer yields
∂φ ∂φ ∂ψ
− = (5.1)
∂y 0+ ∂y 0− ∂x 0

and
∂3 ψ ∂3 ψ ∂φ
− =− . (5.2)
∂y 0+
3 ∂y 0−
3 ∂x 0
Because of the symmetry conditions at y = 0 these can be simplified to
∂φ 1 ∂ψ
=− (5.3)
∂y 0− 2 ∂x 0

and
∂3 ψ 1 ∂φ
= . (5.4)
∂y 0−
3 2 ∂x 0
The asymptotic form of equations (4.11)–(4.14) is thus
∇2 φ = 0, (5.5)

∇2 ∇2 ψ = 0, (5.6)
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

∂ψ ∂φ
ψ = 0, = −1, =0 at x = 0, (5.7)
∂x ∂x
∂ψ ∂3 ψ 1 ∂φ ∂φ 1 ∂ψ
= 0, = , =− at y = 0 (5.8)
∂y ∂y 3 2 ∂x ∂y 2 ∂x
for the domain x > 0, y 6 0.

6. Solution of the {ψ, φ} equations

The far-field solution consists of a forced component plus an infinite set of eigen-
functions. Schematically,
X X X
φ = φF + En φn , ψ = ψF + En ψn , q = qF + En qn , (6.1)
 72 D. Jacqmin
where q is the vorticity. The eigenfunctions are of the form
 
1 n+2 n 4n
{φn , ψn , qn } = n sin nθ, cos nθ + cos(n + 2)θ, − 2 cos (n + 2)θ (6.2)
r n+1 n+1 r
for n = 1, 3, 5 . . .. Here, θ = 0 corresponds to the interface and θ = −π/2 to the wall
for y 6 0. The eigenfunction amplitudes En are determined by matching to the inner
flow. The forced component is the dominant far-field component. To leading order,
the forced stream function is the wedge flow (Huh & Scriven 1971)

1
2
π 1
ψF,0 = r −A sin θ + Aθ cos θ + Bθ sin θ , A= 1 2

, B=− 1 2

. (6.3)
4
π − 1 4
π−1
The leading-order vorticity is
2
qF,0 = ∇2 ψ0 = (−A sin θ + B cos θ). (6.4)
r
The leading-order potential has a term proportional to ln r/r, plus 1/r terms:
 
8 ln r 4A 2
φF,0 = − A sin θ + θ cos θ + cos θ ; (6.5)
π r r π
φF,0 in turn forces higher-order far-field terms in ψF and qF . These are
   
16 ln r 3 1 1 32 ln r 1
ψF,2 = − A ( 4 cos θ+ 4 cos 3θ)+O , qF,2 = A 3 cos 3θ+O 3 . (6.6)
π r r π r r
Equations (6.3)–(6.6) approximate the far-field forced solution with a relative error of
O(1/r2 ).
The near-field solution must be found numerically. Equations (5.5)–(5.8) have been
solved using fourth-order-accurate finite differences and using the far-field solution as
a boundary condition at finite router . The accuracy and convergence of the calculated
solution has been confirmed by calculating on progressively finer grids along with
progressively greater values of router . The eigenfunction amplitude E1 was found to be
about 10.53.
Figure 7 shows the near-field stream function (a) and chemical potential (b). Because
the boundary condition is no slip, the velocity right next to the wall must be the
same as that of the wall. An inner no slip boundary layer is shown. This inner layer,
in which the velocity points downward though the interface, extends out to about
x = 5.1. The amount of downward volume flux through the interface is about −1.55
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

(the value of the stream function at x = 5.1). Expressed in dimensional units this
√ √
flux is −1.55 µκ Vwall , or O( µκ). Past this inner no-slip boundary layer, the velocity
through the interface is always upward. Its non-dimensional maximum value is 0.0618
at about x = 9.5.
The maximum value of the non-dimensional potential is found at the contact line
and is about 1.84. Since this drops to near zero over a non-dimensional distance of
O(1) the non-dimensional diffusive flux is O(1). The dimensional diffusive flux is given
in terms of non-dimensional {x, y, z, φ} by −Vwall ∇φ and is also O(1) (or O(Vwall )).
The chemical potential obeys the jump condition (5.1) at the interface, hence the
kinks in the contour lines there. The direction that a kink points to depends on the
direction of the velocity there: to the right for downward flow, to the left for upward.
The chemical potential in the outer regions of the plot is dominated by its sin θ ln r/r
far-field term.
 Contact-line dynamics of a diffuse fluid interface 73
(a) (b)
25.0

12.5

–12.5

–25.0
0 5 10 15 20 25 0 5 10 15 20 25

Figure 7. (a) The stream function and (b) the chemical potential for the asymptotic flow √at a
contact line. The asymptotic interface thickness is 0 in these units, because  → 0 faster than µκ.
The interface is located at the y = 0 axis. Both the stream function and the chemical potential are
symmetric about the interface. The wall is located at x = 0 and is moving downward with unit
speed. Because the boundary condition is no-slip, the velocity right next to the wall must be the
same as that of the wall. An inner no-slip boundary layer is clearly shown. This inner layer, in
which the velocity points downward though the interface, extends out to about x = 5.1. Past this
inner no-slip boundary layer, the velocity through the interface is always upward. The chemical
potential obeys the jump condition (5.1) at the interface, hence the kinks in the contour lines there.
The direction in which a kink points depends on the direction of the velocity there, to the right for
downward flow, to the left for upward.

Because of the far-field condition that ψ equals 0 along the meniscus the asymptotic
solution under consideration necessarily has no net through-flow. This seems to be
the only reasonable condition for the unbounded case, because there are no far-
field eigenfunctions with ψ not decaying to zero. However, the question arises as
to whether this unbounded solution is applicable to flow between walls. As a test,
numerical solutions have been found for the asymptotic equations (5.5), (5.6) for flow
in a sidewall-driven square box. As in § 3 the box is half-filled by each fluid. The
(asymptotically thin) interface runs horizontally at the box mid-height. The jump
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

conditions (5.1)–(5.2) are applied at the interface. Because of symmetry, only the flow
in the lower left-hand corner of the box needs calculation. Equations (5.5)–(5.8) were
solved together with the no-flux, no slip boundary conditions at the bottom wall:
∂ψ ∂φ
ψ = 0, = 0, = 0, (6.7)
∂y ∂y
and the symmetry conditions at the vertical midline:
∂ψ ∂3 ψ
= = φ = 0. (6.8)
∂x ∂x3
Figure 8 shows the resulting flow in a 100 × 100 (in non-dimensional, microscale
units) box. The square box results in a four-cell flow.
There are two contact lines in the cavity and, as is visible in figure 8, there can be
 74 D. Jacqmin

Figure 8. The stream function for the asymptotic flow in a 100 × 100 (in non-dimensional
microscale units) box.

net downward flow at one and a compensating flow upward at the other. Three cavity
sizes, 100 × 100, 200 × 200, and 400 × 400, were considered in an effort to estimate
the net flow for each contact line as the cavity size approaches infinity. This net flow
was found to be about 0.369 for the smallest box, about 0.233 for the middle box,
and about 0.146 for the largest box. These fluxes are given by the absolute value of
the stream function at the centre of the box. If the no-net-flux far-field solution is the
relevant solution for understanding the box flow then these stream function values
should become increasingly close to the stream function values of the unbounded
solution. The stream function at large r for the unbounded flow is given to good
accuracy by the ψF,2 + E1 ψ1 component of the far-field solution. |ψF,2 + E1 ψ1 | for the
unbounded condition is 0.4466 for {r, θ} = {50, 0}, 0.2611 for {100, 0}, and 0.1494 for
{200, 0}. The flux between r = 0 and rboxcentre of the cavity solution thus appears to
rapidly approach that of the no-net-flux unbounded solution as the cavity increases
in size. As a corollary, the no-net-flow condition for the unbounded solution appears
to be both relevant and correct.
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

7. The interface
This section discusses the solution to the C equation, equation (4.6). We determine
the perturbation to interface location and interface slope, thus connecting the present
theory to classical theory, and we also investigate the perturbation of the interface
profile. We show that the equation for interface slope matches in form that of sharp-
interface slip theory and that the perturbation to interface structure and to the energy
√
of the interface is O(/ µκ).
The C equation (4.6) and boundary condition (4.8) are non-dimensionalized by
√ p
x → µκ x, y → y, φ → µ/κ Vwall φ. (7.1)
The difference from the previous non-dimensionalization is that the y-direction is
 Contact-line dynamics of a diffuse fluid interface 75
stretched so that interface thickness becomes O(1). Equations (4.6), (4.8) become
 2 2  
∂2 C 00  ∂ C 
α 2 − βΨ (C̄)C + α √ =− √ Ca φ, (7.2)
∂y µκ ∂x2 µκ

2 σαDw ∂C
√ = −C̄y Vwall , (7.3)
µκ ∂x x=0

where Ca is the capillary number. In these coordinates, equation (2.3b) becomes

Z  2
∂C
α dy = 1. (7.4)
∂y eq
√
Since the non-dimensionalized φ is O(1), (7.2) suggests that C is O(/ µκ). This,
however, is false. The left-hand side of (7.2) has an eigenfunction solution C = C̄y
√
and this produces a near-singularity that gives C an O( µκ/) component. This
component gives the deflection and slope of the perturbed interface. Higher-order
terms then give the perturbation to the interface structure. The solution for C is of
the form
√  2 !
µκ  
C=− h(x)C̄y + √ C1 + √ C3 + · · · . (7.5)
 µκ µκ
The solution component C̄y gives a unit downward translation of the meniscus
√
without change of interface profile; µκ/ h(x) accordingly gives the perturbation to
the height of the interface in the present stretched coordinates. The interface slope
√ √
µκ/ dh/dx is O( µκ/) because of the differently stretched x- and y-coordinates;
dh/dx by itself gives the slope as viewed in the original dimensional coordinates.
The equation for C1 is
 
d2 C1 00 d2 h
α 2 − βΨ (C̄)C1 = αC̄y 2 − Ca φ . (7.6)
dy dx
Ry
The left-hand side of (7.6) has the homogenous solutions C̄y and C̄y 0 (1/C̄y2 ) dy.
The first is even, bounded at ∞, and an eigenfunction. The second is odd and
unbounded
R∞ at ∞. Because
of the eigenfunction, for a solution to exist the integral
2
0
C̄y α C̄ y (d h/dx2 ) − Caφ dy must equal zero. Using (7.4) and the fact that φ is
very slowly varying in y, this constraint leads to
d2 h
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

= Ca φ(x, 0). (7.7)

dx2
√
The wall boundary condition for h, found by substituting C = −( µκ/)h(x)C̄y into
(7.3), is
dh
σαDw = Vwall . (7.8)
dx x=0
The far-field solution for the meniscus slope is found by substituting (6.5) into (7.7).
The result, to lowest order, is
Vwall
hx = + Ca (4A ln r + const.) (7.9)
σαDw
where the constant must be determined from the inner solution. In dimensional
 76 D. Jacqmin
coordinates (7.9) becomes
Vwall √
hx = + Ca (4A ln r − 4A ln µκ + const.) (7.10)
σαDw
This is the same form as deduced by Cox (1986) for his two-region analysis. As
discussed in § 2, Dw is taken in this analysis to be at least O(1/). Vwall /σαDw is thus
O(Ca) or smaller. The dimensional form of Cox’s equation (5.4) (Cox 1986, p. 178) is
equivalent to
hx = Ca (f(π/2, 1) ln r − f(π/2, 1) ln s + const.). (7.11)
√
In (7.11) the slip length s replaces µκ in equation (7.10). Cox’s stress function
f(θeq , λ) evaluated for θeq = π/2 and a viscosity ratio of 1 replaces (and equals) 4A.
Cox assumes that the microscopic contact angle is the same as the equilibrium angle,
which is equivalent to setting Dw equal to ∞.
The constant in (7.9), (7.10) is determined from the {ψ, φ} near-field solution. From
this and equation (7.8) the undetermined constant is found to be about 0.643:
 
Vwall r
hx ' + Ca 4A ln √ + 0.643 . (7.12)
σαDw µκ
An equation relating h to the pressure field can also be derived. The linearization
of (2.9) gives
∂p ∂2 C ∂C
µ∇2 u = − σαC̄y + σαC̄yy , (7.13)
∂x ∂y∂x ∂x
∂p ∂2 C
µ∇2 v = + σαC̄y 2 . (7.14)
∂y ∂x
These are in dimensional form. Integration of (7.14) through the interface yields
Z
d2
σα 2 C̄y C dy = −(p+ − p− ). (7.15)
dx
Since u+ = u− and, from the continuity equation, ∂u/∂x|0+ = ∂u/∂x|0− implies
∂v/∂y|0+ = ∂v/∂y|0− , and since µ+ = µ− , the viscous terms drop out. The same
integration/reduction procedure as used to derive (7.7) then gives
d2 h
σ = p+ − p− . (7.16)
dx2
Since p is antisymmetric, this can be rewritten:
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

d2 h
= −2p− .
σ (7.17)
dx2
√ √ √
With the scalings x → µκ x, h → µκ h, p → 12 µVwall / µκ p, (7.17) becomes
d2 h
= −Ca p− . (7.18)
dx2
Thus along the interface, p+ = −p− = φ.
We return briefly to C1 . From (7.7), equation (7.6) becomes
d2 C1

α − βΨ 00 (C̄)C1 = Ca αC̄y φ(x, 0) − φ . (7.19)
dy 2

Since the two homogenous solutions are known the complete solution can be
 Contact-line dynamics of a diffuse fluid interface 77
written down using the method of variation of parameters. C1 gives the leading-
order perturbation to the meniscus profile. The solution is non-singular and behaves
like Ca φ/βΨ 00 (C̄) as y → ∞. The perturbation to interface structure isR uniformly
√
O(Ca / µκ). The leading-order perturbation to the surface tension is 2ασ C̄y C1,y dy,
√
in non-dimensional coordinates. This is also O(Ca / µκ).

√
8. The case / µκ = O(1)
√
The case δ = 2, / µκ = O(1) has the same two-region solution structure as δ < 2,
√
/ µκ → 0. The far-field solution is the same for both cases. The near-field solution
is different because for δ = 2 the details of the solution become dependent on the
shape of the interface profile.
We look first at the solution for the potential and velocity field. The far-field solution
√
and solution structure, equations (6.1)–(6.6), remains valid. With O( µκ) = O()
there is no longer a definite outer scale so it can be useful to write these equations
dimensionally. In particular, the dimensional chemical potential is
√
 !
8 ln r/ µκ 4A 2
φF,0,dim = µVwall − A sin θ + θ cos θ + cos θ . (8.1)
π r r π
The magnitude of the chemical potential is proportional to the viscous forcing µVwall .
√
There is a weak dependence on the diffusive length scale µκ through the equation’s
√
log term. Equation (8.1) is valid for r  . In the r = O() = O( µκ) region a
numerical solution must be carried out that includes the details of the interface
structure. The solution is the same in character and is carried out using the same
√
method as for the / µκ → 0 case.
√
When / µκ is O(1), φ in the inner region can no longer be considered to be
slowly varying on the O() length scale. The asymptotic expansion given by (7.5) is
therefore no longer useful in the inner region but it continues to be valid in the
far field. The local rate of convergence of the expansion is approximately /r. From
√
(7.2), the perturbation to C in the near field is O(Ca / µκ). Equation (7.10) for hx
therefore remains valid but with its constant now being a function of the interface
profile.

9. Non-90◦ contact angles

This section discusses some results for non-zero g(C). Results are primarily com-
putational. Their main purpose is to show the reader that the chief results from the
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

previous sections, in particular the ability of a diffuse interface contact line to stop
most through-flow and the increasing integrity of the interface as it becomes thinner,
carry over to the non-90◦ case.
First, however, an important fact that must be pointed out is that the wall free
energy γg(C) (see equation (2.13)) is not in general the same as the total wetting
energy. With non-zero g(C), layers are formed on walls that, in equilibrium, obey
d2 C̄
σα − −1 σβΨ 0 (C̄) = 0 (9.1)
dx2
with the boundary conditions
∂C̄
σα + γg 0 (C̄) = 0 at x = 0, C̄ = Cbulkphase at x = ∞. (9.2)
∂x
 78 D. Jacqmin
This minimizes the total of the layer energy plus the wall energy, which is the total
wetting energy. The layer – relative to the bulk phase at x = ∞ – is enriched with
one of the fluids, depleted in the other. These effects can be of very great interest,
see for example de Gennes’ (1985) discussion of critical wetting. However, we do not
want yet to consider such complicating effects. The wall free energy and the total
energy become the same and the wall layer disappears when g 0 (Cbulkphase ) = 0. The
calculations shown here are for only this simpler, but still important, situation.
With this limitation, the most significant differences between the 90◦ and non-90◦
degree cases are (i) a change in far-field eigenfunction structure and a resulting change
in the far-field behaviour of the chemical potential and (ii) a change in the structure
of the mathematical representation of the far-field solution. Equations (5.5)–(5.7)
continue to be valid. The jump conditions (5.1), (5.2) at the interface become
1 ∂φ 1 ∂φ ∂ψ
− = (9.3)
r ∂θ + r ∂θ − ∂r
and
1 ∂3 ψ 1 ∂3 ψ ∂φ
− 3 3 =− . (9.4)
r ∂θ + r ∂θ −
3 3 ∂r
We use the same r, θ coordinates as in § 6. The θ at which these conditions are applied
is designated θc , where θc is equal to the equilibrium wedge angle of the advancing
fluid minus π/2. The system is no longer symmetric so the entire domain x > 0
must be considered. The quantities ψ, ∂ψ/∂θ, ∂2 ψ/∂θ2 , and φ are continuous at the
interface.
As with the 90◦ (θc = 0) case the far-field solution can be represented in terms of
a forced solution plus an infinite set of far-field eigenmodes. The stream function
component of the forced solution is, again, the wedge flow solution of Huh & Scriven
(1971). The resulting forced chemical potential is of the form
1
ΨF ' sin θ/ sin θc . (9.5)
r
This solution form fails for θc = 0, i.e. for the 90◦ contact angle case already
considered. Unlike for the solution for θc = 0 there is no cascade to higher-order
terms. This is because ΨF is analytic: there is no jump in its derivative at the interface.
The far-field eigenfunctions divide into two subsets. In the first set the leading-order
term of the chemical potential is identically zero in the advancing fluid and of the
form
1
1 2
π
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

sin ξn (θ − θc ), ξn = n 1
, n = 1, 3, 5, . . . . (9.6)
rn
ξ
2
π − θc
in the receding; ξn is chosen so that boundary conditions
∂φ
φ(θc ) = 0, =0 (9.7)
∂θ π/2

are satisfied. The stream function eigenfunction is forced by the jump condition
(9.3). It also behaves like 1/rξn . The chemical potential eigenfunctions in the second
set are zero in the receding fluid, behave like (9.6) in the advancing, and have
ξn = nπ/2/(π/2 + θc ), n = 1, 3, 5, . . . . In general, the leading-order potential behaves
very differently in the two fluids. For θc = −π/4 its receding fluid behaviour is like
r−2/3 . In the advancing fluid its far-field behaviour is dominated by φF . The leading-
order eigenfunction there behaves like r−2 . As θc approaches 0, the forced mode and
 Contact-line dynamics of a diffuse fluid interface 79

(a) (b)

Figure 9. Equilibrium (motionless) thick and thin interfaces. The left-hand √ wall is preferentially
wetted by the lower fluid, the right-hand wall by the upper. g(C) is +15 2(3C − 4C 3 )/2 for the
left-hand wall, the opposite for the right. The interfaces are at the expected 45◦ angle.

the leading-order receding fluid and advancing fluid eigenfunctions merge to give the
ln r/r behaviour found in § 6.
The calculations of § 3 have been repeated for the case of a 45◦ contact angle
(θc = −45◦ ).√All parameters are the same as in that section except for the addition of
g(C) = ±15 2(3C − 4C 3 )/2. C varies from −1/2 to +1/2 √ so the jump in wall energy
from the relatively wetting fluid to the other is 30/ 2 erg cm−2 . Since the surface
tension is 30 dynes/cm the indicated equilibrium contact angle is 45◦ . Figure 9 shows
two diffuse interfaces, one thick, one thin, in their equilibrium state. Their thicknesses
are the same as the thick and thin interfaces in § 3. The sidewalls have been assigned
opposite g(C) – one is preferentially wet by the lower fluid, the other by the upper.
The result is a straight meniscus running from one sidewall to the other. This allows
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

the interface slope and perturbations to it to be easily measured. Figure 10 shows

the same interfaces at Ca = 0.01333. Each suffers about the same perturbation to its
average slope. As in § 3 we see that the thinner interface suffers much less structural
perturbation. Figure 11 shows the chemical potential for the thick interface. It is
qualitatively in agreement with analysis; the potential decays much more rapidly in
the wetting fluid (because of its narrower wedge at the contact line) than in the
non-wetting.
Figures 10 and 11 show results for a wetting flow, i.e. the more wetting fluid is
advancing. The analysis in this paper is linear so it does not distinguish between
wetting and dewetting. Preliminary calculations have been made for this fluid system
for dewetting flow. These show a very different behaviour than for wetting. The
calculations show complete contact line failure at a Ca of about −0.008 and a
transition to film formation (a residue of the more wetting fluid left on and near
 80 D. Jacqmin

(a) (b)

Figure 10. Same as figure 9 but with the left wall moving down, the right wall up. Ca = 0.01333.
The macroscopic slope is changed but the interface slope right at the walls remains 45◦ .

the receding wall) at about Ca = −0.006. For dewetting, linear theory appears to be
accurate down to a Ca of only about −0.003.
Calculations have also been made for a 30◦ contact angle. These give a transition
to film flow at about Ca = −0.003.

10. Comparison of numerical calculations to theory

This section looks at the nonlinear numerical calculations of § 3 and compares

them to the theoretical results of § 6 and § 7 and the linear box calculations of § 6.
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

The calculations in § 3 have interfaces with very large thicknesses. Related to that,
√
their ratio / µκ is large, contrary to the regime dealt with by the theory. Can their
results nevertheless be related to the asymptotic theory? We show here that these ‘fat
interface’ results appear to approach those of asymptotic theory according to O(2 ).
According to (7.12) if κ is halved then the slope of an interface with a moving
contact line increases by 2Ca A ln 2. For the calculations in § 3, with Ca = 0.01333,
this equals about 0.0198. These calculations are for a µ of 1. For the thick meniscus
κ is 2.65 × 10−5 cm5 s−1 erg−1 and for the thin meniscus it is 6.625 × 10−6 . In non-
dimensional units the driven cavity has a size of 388.3 by 1164.9 for the thick meniscus,
√
twice that for the thin; / µκ for the thin meniscusis is about 62.9.
For these parameters, equation (7.12) with r set equal to 1 (the left-to-right midline
of the box) gives an hx for the thick meniscus calculation of about 0.309. For the thin
meniscus (7.12) gives 0.348. The numerical calculations in § 3 give, respectively, 0.331
 Contact-line dynamics of a diffuse fluid interface 81

Figure 11. The chemical potential for the thick meniscus of figure 10. The potential decays much
more rapidly in the wetting fluid (because of its narrower wedge against the wall) than in the
non-wetting.

and 0.336. The numerical results are qualitatively right, but the expected increase in
slope with increasing container size is lacking.
Relating the calculations in § 3 to the asymptotic theory involves two adjustments.
The first is for the finite thickness of the interfaces, the second is for the effects
of the box enclosure. To judge the first effect the calculations in § 3 have been
redone with interfaces with different thicknesses. Halving the interface thickness for
the 388.3 × 1164.9 calculation results in a slope of 0.283; halving it again yields
0.268. Halving the interface thickness for the 776.6 × 2329.8 enclosure yields a slope
of 0.314 while doubling it gives 0.377. Extrapolating these results using a second-
order extrapolation yields  = 0 slopes of 0.259 for the small container and 0.292
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

for the large. Extrapolating using a third-order approximation and assuming an

O(2 ) dependence of the slope as  → 0 yields 0.262 and 0.304. These adjusted
results, particularly those for the second extrapolation, demonstrate close to the same
increase in slope with increase in domain size that is predicted by theory. However,
the difference of these results from the asymptotic theory has now increased.
This difference is accounted for by effects of the top and bottom walls of the box
enclosure. A calculation using the asymptotic theory (equations (5.5)–(5.6) with the
jump conditions (5.1)–(5.2) has been made for the smaller enclosure. This yields a
midpoint slope of 0.260, in good agreement with the extrapolations. An asymptotic
calculation for the larger enclosure is, unfortunately, too expensive. However, to
further study enclosure effects nonlinear calculations have also been made for the
box enclosure discussed in § 6. The linearized box calculations of § 6 yield midpoint
slopes of 0.143 for the 100 × 100 box, 0.182 for the 200 × 200 box, and 0.222 for the
 82 D. Jacqmin
400 × 400 box. hx increases at almost exactly the same rate as given by the asymptotic
result for the single contact line but the square box enclosure changes the slope from
that of the semi-infinite interface by a constant roughly equal to 0.08. Nonlinear
calculations made for the 400 × 400 box for the three different interface widths of § 3
give a midpoint hx for the thickest interface of 0.284, for the middle, 0.238, and for
the thinnest, 0.225. The convergence towards the result of the asymptotic linearized
calculation appears to be O(2 ).
Another measure of the distance/closeness of macroscopic interface and asymptotic
results is found in the potential field. Since the spatial non-dimensionalization is
isotropic the potential field in the asymptotic and § 3 calculations should have the
same appearance. Comparing figure 6 to figure 7(b), we see that the outer regions of
the two fields match well – the logarithmic term in equation (6.5) explains the necking
down at the meniscus. Right at the contact line, however, the two are in disagreement.
The asymptotic result shows the potential gradient oriented roughly parallel to the
wall while the results in § 3 show it perpendicular. Again, this disagreement seems
to be due to differences in meniscus width. Another easy point of comparison is
the potential maximum, which occurs right at the contact line. The calculations are
in agreement with theory in that the maximum potential is very close to inversely
proportional to the square root of the mobility. The coarse meniscus calculation
has a maximum potential 59% of that predicted by the asymptotic theory, the fine
calculation increases to 66%. Some additional linearized calculations have been made
that indicate continuing slow increase in this ratio with increasingly thin menisci.

11. Nanoscale calculations

The results so far have not addressed the question of whether or not realistic
diffusivities and levels of interdiffusion could possibly play a role in regularizing
moving contact lines. In this section we look at results from direct simulations of
realistically scaled interfaces with realistic liquid–liquid diffusivities. Liquid–liquid
interfaces away from critical points are about 1 nm thick (10−7 cm), about three or
four molecular lengths. A typical molecular diffusion rate is about 10−5 cm2 s−1 . The
macroscopic interface calculations presented in § 3 had interface thicknesses about
106 larger than realistic and a diffusivity about 103 too large.
The calculations presented in this section are for the same geometry, wall velocities,
surface tension, and viscosity as in § 3, but with the cavity size drastically scaled
down. Most calculations are of a 16 × 48 nm cavity but some have been made at
twice and four times that, in order to investigate the domain-size dependence of
the results. The calculations are at about four to sixteen times the scale of moving
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

contact line molecular dynamics calculations by Koplik et al. (1989) and Thompson
& Robbins (1989). As in § 3, Ψ (C) is modelled as (C − 1/2)2 (C + 1/2)2 . Its α, β
and tanh equilibrium profile and a definition of its interface thickness are given in
§ 3. Its relationship between diffusivity and mobility is D = −1 σβκΨ 00 (Cbulkphase ) =
24 tanh−1 (0.9)−1 σκ ' 35.33−1 σκ.
Diffuse-interface Stokes flow is invariant to a transformation that can be used to
connect the calculations of § 3 to nanoscale calculations. If the domain of flow is
shrunk or expanded by a given factor, if  is changed by the same factor, and if
the mobility is changed by the factor squared, viscosities, surface tension, densities,
and velocity boundary conditions remaining the same, then steady-state flow fields
and interface positions are the same at each scale. The thin interface calculation
in § 3 has an  of about 0.16 cm and a mobility of 6.6 × 10−6 . The corresponding
 Contact-line dynamics of a diffuse fluid interface 83
thickness and mobility in a 16 by 48 nm box (scale factor 8 × 10−7 ) are 1.29 nm and
4.2 × 10−18 . The diffusivity in the calculations in § 3 is about 0.043 cm2 s−1 . In the
rescaling the diffusivity is changed by the scale factor and it becomes about 3.5×10−8 .
The calculations in § 3 are at finite Reynolds number (Re = 0.8) so the transformation
in this case does not hold perfectly. Still, considering the low interfacial distortion
shown in the thin interface results of § 3 and the extremely low diffusivity at realistic
interface widths that those calculations correspond to, it appears that diffusion is
definitely a candidate for being a mechanism for contact line regularization.
Nanoscale calculations have been carried out for a wide range of diffusivity and
slip. The slip model used was
∂v ∂v
(v − Vwall ) = −Sµ = −L (11.1)
∂xn ∂xn
where xn is the direction into the wall and µ∂v/∂xn is the wall shear stress. L
is the inner scale or slip length that would be set by shear in the absence of
chemical potential effects. The interface thickness was set to 1.2 nm. With no slip,
and under these conditions, it was found that diffusion can maintain the interface
profile with no noticeable perturbation down to about D = 10−7 . At D = 10−8 the
contact line is maintained but the interface profile is distorted. The chemical potential
loses its characteristic half-dumbbell shape and begins to align itself parallel to the
convectively thickened/thinned interface. At D = 10−9 the contact line begins to fail.
With D = 10−5 the mobility is about 1.1 × 10−15 †, giving a diffusive length scale of
about 3.3 × 10−8 . In agreement with the analysis, the inner no-slip velocity layer has
a width of about 1.4 nm.
The effects of slip were investigated for D = 10−6 . At S = 10−6 (µ = 1 P, so L = S
in all the following calculations) the maximum velocity reached in the 16 by 48 box is
only 0.12 cm s−1 (Vwall = 0.4). Slip effects are dominant and extend the whole length
of the enclosure. For the no-slip case, the maximum chemical potential for D = 10−6
is 4.7 × 107 erg cm−3 . This corresponds to a local dissolution of one phase in the
other of about 1%. The maximum chemical potential force on the fluid is about
5 × 1014 erg cm−4 . With S = 10−6 the chemical potential keeps its half-dumbbell shape
but is greatly reduced in amplitude, to a maximum of about 3.2 × 106 . Diffusive fluxes
are therefore only about one fourteenth that of the no-slip case. With S = 10−7 ,
slip and diffusion are competitive in effect. The maximum velocity is 0.34 and the
maximum chemical potential is 1.7×107 . With S = 10−8 diffusion effects are dominant.
The maximum chemical potential is about 85% that of the no-slip case.
The no-slip D = 10−6 case was also calculated for the 48 by 144 nm box. The
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

average slope of the interface in the large box is 0.253. In the small box it is 0.144,
for a difference of 0.109. (The theoretical difference, from equation (7.12) for r → ∞,
is 0.079.) Except for this, there is little change from one calculation to the other.
The maximum chemical potential in the large box is 2% greater than in the small.
The no-slip inner velocity layer is in each case about 5 Å. The minimum radius of
curvature of the interface, found at the contact line, is in both cases about 10 nm.
The above calculations have been rather conservative in that they have considered
both small diffusivities and large slip lengths. The molecular dynamics calculations
of Koplik et al. and Thompson & Robbins indicate slip lengths of at most about
3 Å. The yield-stress analysis of Durbin (see § 3) gives close to the same result. Barrat

† This mobility is an order of magnitude below generally accepted values. D = 10−5 is typical
for ideal solutions but may be low for non-ideal diffusion near interfaces.
 84 D. Jacqmin
& Bocquet also agree, provided that the fluids involved each individually partially
wet the wall. L = 10−6 therefore seems to be completely unrealistic. From Barrat
& Bocquet, L = 10−7 appears to be close to the upper limit for slip length for
fluid systems with fluid–vapour contact angles less than 100◦ . The implication of the
calculations in this section is therefore that chemical potential effects are competitive
with and perhaps dominant over shear-induced slip.

12. Conclusions
The CHW interface model is a very approximate but nevertheless widely used
fluid interface model. Though it has generally been used in the context of diffusion
processes there has been an increasing interest in the condensed matter, critical
fluids, and computational fluid dynamics communities in using it for fluid dynamics
modelling.
This work was begun to find out if the CHW model could validly and usefully be
used for contact line flows. This question can be looked at from either a macroscopic
or microscopic point of view.
The macroscopic results are that the CHW contact line has the same far-field
behaviour as the more traditional sharp-interface slip models of contact line flow. A
computational fluid dynamicist interested in only macroscopic results can thus use
either model. Certain difficult to know parameters enter into both: in the case of slip
models it is the material-dependent slip length, in the case of the CHW model it is
the material-dependent mobility κ. If κ or the slip length are unknown the accuracy
of the results will be only qualitative unless some empirical matching is made to
experiments. Fortunately, the sensitivity of the macroscopic contact angle to these
unknowns, being logarithmic, is unlikely to be ruinous.
The microscopic results have shown that chemical potential effects may be dominant
over shear and that it may be possible for a moving contact line to be regularized even
with no-slip boundaries. We have looked at the case of regularization achieved via
chemical diffusion driven by intermolecular potential gradients. Chemical potential
induced compressibility (Seppecher 1996) probably also plays an important role. The
diffusion model used in this paper is the standard model introduced by Cahn.
The results herein have been derived for general bulk free energy Ψ (C). They
therefore apply both to near-critical fluid mixtures and to mixtures far from critical
for which deep-quench or double-obstacle free energy models would be appropriate.
The applicability of the CHW model to realistically thick interfaces with realistic
mobilities has been shown by the nanoscale calculations in § 11. A difficulty is that
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

the diffusion length scale can be very small, indicating that subcontinuum effects may
have to be carefully modelled.
A related question is which model of contact line regularization – shear-slip or
chemical-potential driven fluxes – is likely to further our understanding of contact line
flow? It seems worthwhile to look at CHW dewetting and try to determine if it matches
to experiments. The preliminary results reported in § 9 show that CHW dewetting
is very different from wetting and that the CHW model gives a dewetting failure
at very low capillary number. Another flow problem that could be looked at is the
instability of a thin film lying on a solid. Unlike one-dimensional lubrication models,
Navier–Stokes/CHW calculations of its evolution could be followed to complete
rupture – the appearance of a contact line – and beyond. The Navier–Stokes/CHW
model would be more difficult to solve than a one-dimensional lubrication model,
but much easier than models that rely on two-dimensional mean-field integrals of
 Contact-line dynamics of a diffuse fluid interface 85
molecular attractions (see Indeikina & Chang 1998). A third problem that the CHW
model could be useful for is the fluid dynamics involved in phase nucleation against
solids.
This paper has not argued that the CHW model is in detail correct. In fact, it
has been known almost since its inception that a basic assumption underlying it –
that fluid interfaces are thick enough so that composition gradients across them can
be validly time-averaged and then differentiated – is, except perhaps for near-critical
fluids, not true. Given the large diffusive fluxes, the simple diffusion model that
this paper has used is also only qualitatively correct. It is hoped that, nevertheless,
the CHW model points toward the right physics, which are very basic: (i) that the
contact line kinematic and stress singularities can be resolved by the two phases
having differing directions of flux; (ii) that the contact line composition profile at the
wall can be maintained by the diffusive interchange of particles; and (iii) that bond
breaking at the wall is caused by chemical potential forces.

Appendix. Linearized version of Seppecher’s analysis

The following may help the interested reader compare Seppecher’s work to the
work in this paper. Seppecher discusses a model compressible Cahn–Hilliard fluid
that is isothermal, has the free energy density
f = 12 σα|∇ρ|2 + −1 σβΨ (ρ) (A 1)
and has no separate p–V –T relationship, so that the fluid’s thermodynamics is
completely described by (A 1). The equations of motion for this fluid are
Dui
ρ = ∇ · τ visc,i − ρ∇φ, (A 2)
Dt
∂ρ X ∂ρui
=− . (A 3)
∂t i
∂xi
The steady, dynamic contact line equations linearized about a 90◦ equilibrium
contact angle are
∇ · τ linvisc,i = ρ̄∇φ, (A 4)
X ∂ρ̄ui
= 0. (A 5)
∂xi
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

i
Outside the O() thick interface, these equations are incompressible:
∂φ
µ̄± ∇2 u = ρ̄± , (A 6)
∂x
∂φ
µ̄± ∇2 v = ρ̄± , (A 7)
∂y
∂u ∂v
+ = 0; (A 8)
∂x ∂y
µ̄ and ρ̄ take on different constant values above and below the interface. In these
regions ρ̄φ is the same, to within a constant, as the pressure.
 86 D. Jacqmin
Away from the contact line (A 4) can be integrated through the interface to yield
   
∂u ∂v ∂u ∂v
µ̄+ + = µ̄− + , (A 9)
∂y ∂x 0+ ∂y ∂x 0−
Z
∂v ∂v
2µ̄+ − 2µ̄− = ρ̄+ φ(x, 0+ ) − ρ̄− φ(x, 0− ) − ρ̄y φ(x, y) dy. (A 10)
∂y 0+ ∂y 0−
Equation (A 9) plus the conditions u+ = u− and v+ = v− = 0 are used as boundary
conditions to find the far-field hydrodynamic solution. This is the same solution as
given by Cox (1986). In the case of µ− = µ+ , which was the case considered by
Seppecher, the far-field solution is given by (6.3). A far-field potential is forced by the
hydrodynamic solution through (A 6)–(A 8).
The meniscus equation is found by the same integration/reduction as in § 7. The
linearized relationship between density and the chemical potential is
σα∇2 ρ − −1 σβΨ 00 (ρ̄)ρ = −φ. (A 11)
The integration/reduction procedure yields
Z
d2 h
σ 2 = ρ̄y φ(x, y) dy (A 12)
dx
which, from (A 10), becomes
!
d2 h ∂v ∂v
σ 2 = ρ̄+ φ(x, 0+ ) − ρ̄− φ(x, 0− ) − 2µ̄+ − 2µ̄− . (A 13)
dx ∂y 0+ ∂y 0−

The meniscus profile forced by the far-field flow is of the usual form:
hx = Ca(4A∗ ln r + const.); (A 14)
A∗ is the variable viscosity generalization of the coefficient A given by (6.1b).
Near the contact line, the contradiction between the far-field solution (v− = v+ = 0)
and the flow at the wall (v− = v+ = Vwall ) is resolved in an O()-scale near field. There
the full equations (A 4) and (A 5) must be solved, with the far-field solution used as
the r → ∞ boundary condition. The near-field solution determines the constant in
√
(A 14). Like the / µκ = O(1) case discussed in § 8, this constant depends on the
details of the meniscus profile, in other words, on the choice of Ψ .

REFERENCES
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

Anderson, D. M. & McFadden, G. B. 1997 A diffuse-interface description of internal waves in a

near-critical fluid. Phys. Fluids 9, 1870–1879.
Anderson, D. M., McFadden, G. B. & Wheeler, A. A. 1998 Diffuse-interface methods in fluid
mechanics. Ann. Rev. Fluid Mech. 30, 139–165.
Antanovskii, L. K. 1995 A phase field model of capillarity. Phys. Fluids 7, 747–753.
Barrat, J.-L. & Bocquet, L. 1999 Large slip effects at a nonwetting fluid–solid interface. Phys. Rev.
Lett. 82, 4671–4674.
Bates, P. W. & Fife, P. C. 1993 The dynamics of nucleation for the Cahn–Hilliard equation. SIAM
J. Appl. Maths 53, 990–1008.
Blake, T. D. 1993 Dynamic contact angles and wetting kinetics. In Wettability (ed. J. C. Berg),
pp. 251–309. Marcel Dekker.
Brackbill, J., Kothe, D. B. & Zemach, C. 1992 A continuum method for modeling surface tension.
J. Comput. Phys. 100, 335–354.
Cahn, J. W. 1961 On spinodal decomposition. Acta Metallurgica 9, 795–801.
 Contact-line dynamics of a diffuse fluid interface 87
Chella, R. & Viñals, J. 1996 Mixing of a two-phase fluid by cavity flow. Phys. Rev. E 53, 3832–3840.
Cox, R. G. 1986 The dynamics of the spreading of liquids on a solid surface. Part 1. Viscous flow.
J. Fluid Mech. 168, 169–194.
Durbin, P. 1988 Considerations on the moving contact-line singularity, with application to frictional
drag on a slender drop. J. Fluid Mech. 197, 157–169.
Dussan V., E. B., Rame E. & Garoff, S. 1991 On identifying the appropriate boundary conditions
at a moving contact line: an experimental investigation. J. Fluid Mech. 230, 97–116.
Gennes, P. G. de 1985 Wetting: statics and dynamics. Rev. Mod. Phys. 57, 827–863.
Gennes, P. G. de, Hua, X. & Levinson, P. 1990 Dynamics of wetting: local contact angles. J. Fluid
Mech. 212, 55–63.
Gibbs, J. W. 1878 On the equilibrium of heterogenous substances. Trans. Conn. Acad. 3, 108–248,
343–524. Reprinted in The Scientific Papers of J. Willard Gibbs, pp. 55–371. Oxbow Press.
Gurtin, M. E., Polignone, D. & Viñals, J. 1996 Two-phase binary fluids and immiscible fluids
described by an order parameter. Math. Models Meth. Appl. Sci. 6, 815–831.
Huh, E. & Scriven, L. E. 1971 Hydrodynamic model of steady movement of a solid/liquid/fluid
contact line. J. Colloid Interface Sci. 35, 85–101.
Indeikina, A. & Chang, H.-C. 1998 A molecular theory for dynamic contact angles. In IUTAM
Symp. on Non-Linear Singularites in Deformation and Flow, pp. 321–337. Kluwer.
Jacqmin, D. 1995 An energy approach to the continuum surface tension method: application to
droplet coalescences and droplet/wall interactions. Presented at the 1995 ASME IMECE, San
Francisco, CA.
Jacqmin, D. 1996 An energy approach to the continuum surface tension method. AIAA Paper
96-0858.
Jacqmin, D. 1997 A variational approach to deriving smeared-interface surface tension models. In
Barriers and Challenges in Computational Fluid Mechanics (ed. V. Venkatakrishnan & M. D.
Salas), pp. 231–241. Kluwer.
Jacqmin, D. 1999 Calculation of two-phase Navier–Stokes flows using phase-field modeling.
J. Comput. Phys. 155, 96–127.
Jasnow, D. & Viñals, J. 1996 Coarse-grained description of thermo-capillary flow. Phys. Fluids 8,
660–669.
Koplik, J. & Banavar, J. R. 1998 The no-slip condition for a mixture of two liquids. Phys. Rev.
Lett. 80, 5125–5128.
Koplik, J., Banavar, J. R. & Willemsen, J. F. 1989 Molecular dynamics of fluid flow at solid
surfaces. Phys. Fluids A 1, 781–794.
Kothe, D. B., Rider, W. J., Mosso, S. J. & Brock, J. S. 1996 Volume tracking of interfaces having
surface tension in two and three dimensions. AIAA Paper 96-0859.
Lafaurie, B., Nardone, C., Scardoveli, R., Zaleski, S. & Zanetti, G. 1994 Modeling merging and
fragmentation in multiphase flows with SURFER. J. Comput. Phys. 113, 134–147.
Li, D. & Slattery, J. C. 1991 Analysis of the moving apparent common line and dynamic contact
angle formed by a draining film. J. Colloid Interface Sci. 143, 382–396.
Lopez, J., Miller, C. A. & Ruckenstein, E. 1976 Spreading kinetics of liquid drops on solids.
J. Colloid Interface Sci. 56, 460–468.
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press

Lowengrub, J. & Truskinovsky, L. 1998 Quasi-incompressible Cahn–Hilliard fluids and topological

transitions. Proc. R. Soc. Lond. A 454, 2617–2654.
Matsumoto, S., Maruyama, S. & Saruwatari, H. 1995 A molecular dynamics simulation of a
liquid droplet on a solid surface. Proc ASME-JSME Thermal Engng Joint Conf., Maui, Vol. 2,
pp. 557–562.
Miller, C. A. & Ruckenstein, E. 1974 The origin of flow during wetting of solids. J. Colloid
Interface Sci. 48, 368–373.
Nadiga, B. T. & Zaleski, S. 1996 Investigations of a two-phase fluid model. Eur. J. Mech. B/Fluids
15, 885–896.
Ngan, C. G. & Dussan V., E. B. 1982 On the nature of the dynamic contact angle: an experimental
study. J. Fluid Mech. 118, 27–40.
Poisson, S. D. 1831 Nouvelle Théorie de l’Action Capillaire. Paris: Bachelier.
Rowlinson, J. S. & Widom, B. 1982 Molecular Theory of Capillarity. Oxford University Press.
Seppecher, P. 1996 Moving contact lines in the Cahn–Hilliard theory. Intl J. Engng Sci. 34, 977–992.
 88 D. Jacqmin
Sussman, M., Smeraka, P. & Osher, S. J. 1994 A level set approach for computing solutions to
incompressible two-phase flow. J. Comput. Phys. 114, 146–159.
Thompson, P. A. & Robbins, M. O. 1989 Simulations of contact-line motion: slip and the dynamic
contact angle. Phys. Rev. Lett. 63, 766–769.
Thompson, P. A. & Robbins, M. O. 1990 Shear flow near solids: epitaxial order and flow boundary
solutions. Phys. Rev. A 41, 6830–6837.
Unverdi, S. O. & Tryggvason, G. 1992 A front-tracking method for viscous, incompressible,
multi-fluid flows. J. Comput. Phys. 100, 25–37.
Waals, J. D. van der 1893 The thermodynamic theory of capillarity flow under the hypothesis of a
continuous variation of density. Verhandel. Konink. Akad. Weten. Amsterdam 1 (English transl.
by Rowlinson, J. S. in J. Statist. Phys. 20 (1979), 197–244).
Wagner, L. 1995 Molecular dynamics simulations of polymer droplets. Phys. Rev. E 51, 499–503.
Wayner, P. C. & Schonberg, J. 1992 Spreading of a liquid film on a substrate by the evaporation–
adsorption process. J. Colloid Interface Sci. 152, 507–519.
https://doi.org/10.1017/S0022112099006874 Published online by Cambridge University Press