Computers & Fluids 82 (2013) 110–121

Contents lists available at SciVerse ScienceDirect

Computers & Fluids

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m p fl u i d

Some experiences with the numerical simulation of Newtonian

and Bingham fluids in dip coating
Abdelkader Filali a, Lyes Khezzar a,⇑, Evan Mitsoulis b
a
Department of Mechanical Engineering, The Petroleum Institute, P.O. Box 2533, Abu Dhabi, United Arab Emirates
b
School of Mining Engineering & Metallurgy, National Technical University of Athens, Zografou 157 80, Athens, Greece

a r t i c l e i n f o a b s t r a c t

Article history: The dip-coating process is simulated numerically for Newtonian and Bingham fluids with a particular
Received 18 September 2012 emphasis on finding the free surface location under different sets of conditions. The main focus of the
Received in revised form 12 April 2013 simulations is the evaluation of the Finite-Volume Method (FVM) combined with the Volume-Of-Fluid
Accepted 22 April 2013
(VOF) technique embedded in the commercial code (Fluent), while some results are also obtained with
Available online 9 May 2013
the Finite-Element Method (FEM) using another commercial code (Polyflow). The objective is to check
how well the results compare with previous results regarding the free surface location for both axisym-
Keywords:
metric and planar geometries, and also provide new results. The numerical results were first validated
Dip coating
Free surfaces
against previous experimental data for the Newtonian limit case. Then numerical results were compared
VOF method favourably with previous simulations available in the literature for both Newtonian and Bingham fluids.
Inertia The influence of the coating fluid properties as well as surface withdrawal speed for Newtonian fluids and
Surface tension of the yield stress for Bingham fluids are discussed. The effect of inertia for the planar case is investigated,
Newtonian fluid and the formation of the cusp and wavy shape of the free surface is analyzed.
Bingham fluid Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction
Coating flows are liquid flows forming on solid surfaces. Dip-
coating is a process that belongs to the family of self-metered coat-
ing processes, wherein a substrate is dipped into a liquid coating
solution and then is withdrawn from the solution at a controlled
speed (Fig. 1). The combination of substrate speed, liquid proper-
ties (such as density, viscosity, and surface tension), and the pro-
cess geometry influence the thickness of the coated film. The
produced film has to satisfy thickness and uniformity require-
ments of the final product design.
Dip-coating processes are found in a wide field of practical
applications. These coatings are used for decorative, protective
and functional applications that include galvanized steel, magnetic
information storage systems and manufacture of semi-conductor
components. Although coating processes have been developed
mainly through empirical means, see Ruschak [1] and Weinstein
and Ruschak [2], appropriate design and manufacture calls for a
detailed understanding, knowledge and ability to predict the coat-
ing flow process. Computational Fluid Dynamics (CFD) methods,
based on the solution of the governing conservation equations,
combined with interface-tracking techniques, such as the Vol-
⇑ Corresponding author. Tel.: +971 50 820 1536.
E-mail address: lkhezzar@pi.ac.ae (L. Khezzar).
ume-Of-Fluid (VOF) technique of Hirt and Nichols [3], are in prin-
ciple not limited in their range of applicability, and offer a strong
viable alternative. The present work uses such techniques in the
study of dip-coating flows with the objective of finding their limi-
tations and of gaining experience in applying them in cases where
the free surfaces are highly curved.
Three non-dimensional numbers emerge to describe this flow.
These are:
(i) the capillary number, Ca, which represents the ratio of vis-
cous to surface tension forces and is defined by:
lV 0
Ca ¼ ; ð1Þ
r
where l is the viscosity, r is the surface tension, and V0 is a charac-
teristic speed (in dip coating the substrate speed);
(ii) the Reynolds number, Re, which represents the ratio of iner-
tia to viscous forces and is defined by:
qV 0 lc
Re ¼ ; ð2Þ
l
where lc represents the characteristic length, defined as lc = (r/qg)1/2,
where g is the acceleration of gravity;
(iii) the fluid property number, P0, which is equal to the square
root of the ratio Ca/Re with the capillary rise (r/qg)1/2 used
as the length scale, is defined by:

0045-7930/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compfluid.2013.04.024
 A. Filali et al. / Computers & Fluids 82 (2013) 110–121
Fig. 1. Schematic representation of the dip-coating problem with reference axes
and boundary conditions.
 1=4 rffiffiffiffiffiffi
g Ca
P0 ¼ l ¼ :
qr3 Re
It is to be noted that the fluid property number is related to the
Morton number, Mo ¼ P40 .
The dip-coating flow is usually laminar in nature, initially tran-
sient and tends towards a steady state, and has a free surface the
dynamics and final shape of which need to be predicted accurately,
first during the transient and then steady-state periods. The loca-
tion of the free surface and the fluid rheology in the case of non-
Newtonian fluids presents particular challenges to the numerical
simulation of such flows.
Following the initial pioneering work of Landau and Levich [4]
and Derjaguin [5], several experimental investigations were car-
ried out, such as by Lee and Tallmadge [6], Tallmadge and Soroka
[7], Groenveld [8], and Spiers et al. [9]. Kizito et al. [10] carried
out experiments in a broad parametric range, including extended
Ca and Re numbers. Using laser-induced fluorescence, they identi-
fied the phenomena leading to the formation of an asymptotic
meniscus profile, which eventually develops a cusp at the interface
when inertia forces are important (Re > 7). Based on the property
number, P0, two phenomena of free coating are identified. When
P0 is larger than about 0.5, the non-dimensional final film thickness
becomes constant beyond Ca  1. For Ca < 1, the data shows that
the effect of Re is negligible and the film thickness depends only
on the capillary number. However for Ca > 1, the viscous force pre-
dominates, and surface tension has a negligible influence in that
situation. When P0 is less than about 0.1, the non-dimensional final
film thickness depends on Ca and Re but it becomes constant be-
yond a Weber number (We = Ca Re) of about 0.2. In that case Re
is important, and the surface tension force becomes negligible
compared with the inertia force. In both cases, the non-dimen-
sional final film thickness becomes constant as the effect of surface
tension on the meniscus becomes relatively unimportant.
Numerical simulations based on the Finite Element Method
(FEM) have been carried out by Tanguy et al. [11,12], who modeled
the dip-coating process for Newtonian and non-Newtonian fluids.
Numerical results were presented for both flat-plate coating (pla-
nar geometry) and wire coating (axisymmetric geometry), and
compared successfully with experimental data by Lee and Tall-
madge [13]. Most previous numerical works used FEM. Tanguy
et al. [12] used the Augmented Lagrangian Method developed by
Fortin and Glowinski [14]. An iterative solution based on FEM
was used by Reglat et al. [15] to predict the free-surface flow dis-
placement of a Newtonian fluid by a cylindrical substrate. The
111
method was based on a prediction/projection fixed-point iterative
scheme and a combination of an original nodal displacement
scheme in combination with B-spline smoothing and remeshing
to update the free surface. Hurez and Tanguy [16] carried out a
FEM analysis using Newtonian and non-Newtonian Bingham fluids
with an iterative method based on the Arbitrary Lagrangian–Eule-
rian (ALE) approach in conjunction with mesh smoothing. It was
shown that the influence of the viscosity and the withdrawal
velocity must be studied separately, and that the capillary number
is not sufficient to describe the meniscus flow dynamics, so that as
mentioned above, a fluid property number P0 is also needed.
Yamamoto et al. [17] used FEM to simulate transient viscoelastic
dip-coating flows around a rod with three Phan–Thien–Tanner
(PTT) fluids that differ in the stretch-thickening property. Recently,
Jin et al. [18], using the FIDAP software (Ansys, Inc.), confirmed the
asymptotic Newtonian value of the flow rate of 0.582 for large cap-
illary numbers (Ca ? 1) and observed wavy wiggles on the free
surface for large values of capillary number (Ca  3.5) with the
property number P0 = 0.83. Jenny and Souhar [19] used FEM with
a semi-implicit front-tracking method for the moving free surface
and identified steady and wavy regimes based on the property P0
ð3Þ
and capillary numbers Ca for high Reynolds number Re > 1. It
was concluded that beyond a critical Re  5, the flow becomes
wavy producing fluctuations of the order of 10% of the mean value
of the film thickness.
It can be said that on the experimental side, most of the previ-
ous studies were based on flow-visualization techniques and con-
sisted of observing and measuring the free-surface conditions.
These efforts, successful as they were, were limited in the amount
of quantitative information that can be obtained. On the other
hand, CFD as a simulation tool allows access to useful quantitative
information on such flows, which would be difficult to obtain using
experimental tools. It transpires also from the above literature re-
view that most numerical studies used FEM with remeshing tech-
niques. Present-day numerical methods based on the Finite
Volume Method (FVM) and combined with an interface tracking
technique, such as VOF, are capable of simulating dip-coating flows
and can be used for analysis and design purposes.
In the present work, an analysis of the dip-coating process for a
vertical plane and a cylindrical rod is carried out using FVM
embedded in ANSYS – FLUENTÒ code version 12.1 [20]. Both New-
tonian and non-Newtonian Bingham fluids are used. The location
of the free surface at equilibrium, in the presence of surface tension
and external gravity force, is predicted starting from an initial li-
quid bath at rest with a horizontal free surface, using VOF for inter-
face capturing. The present study provides an evaluation of the
VOF algorithm and FVM in predicting the behavioral details of such
flows. The numerical work is contrasted with the results obtained
in the works by Kizito et al. [10] and Hurez and Tanguy [16]. Par-
ticular attention is given to the existence of cusps on the free sur-
face under certain flow conditions and the resulting flow field
structure for dip coating in planar geometries.
2. Mathematical modeling
Since the focus of this study is an evaluation of the technique
based on the FVM-VOF combination, its mathematical model is de-
scribed in some detail, while a short description of the FEM mod-
eling approach is also given for completion.
2.1. Governing equations
The field conservation equations for mass and momentum for
transient, incompressible, laminar flow are given below and are
shared by the liquid and gas phases:
112 A. Filali et al. / Computers & Fluids 82 (2013) 110–121
r  u ¼ 0;
Du
q ¼ rp þ r  s þ qg þ F S ;
Dt
where u represents the velocity vector, p the pressure, s the extra
stress tensor, and Fs is the surface tension force (see Brackbill
et al. [21]), which will be discussed later.
For a generalized Newtonian fluid, the rheological constitutive
equation that relates the extra stress tensor to the rate-of-strain
tensor is given by
s ¼ gc_ ;
where g is the apparent viscosity and c_ represents the rate-of-strain
tensor given by.
c_ ¼ ru þ ruT ;
where ru is the velocity-gradient tensor and the superscript T is
the transpose of a tensor. The apparent viscosity g is a function of
the magnitude of the rate-of-strain tensor
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 Courant numbers (defined by: Cr = Dt/DxcellVfluid) was below 2.
c_ ¼ jc_ j ¼ ðc_ : c_ Þ;
2
When the apparent viscosity is a constant g = l, then Eq. (6)
represents the Newtonian fluid.
For viscoplastic non-Newtonian fluids having a yield stress s0,
the rheological constitutive equation is described by the Her-
schel–Bulkley model [20], which is a generalization of the Bingham
model and combines the effects of the yield stress and the power
law. For stresses below a yield stress s0, the material behaves like
a rigid solid. As the stresses increase beyond the yield stress
threshold, s0, the material behaves as a fluid described by a power
law. The Herschel–Bulkley model is characterized by three param-
eters: yield stress, consistency index K, and power-law index n.
The ideal Herschel–Bulkley model is discontinuous at the yield-
stress point and thus very difficult for computations. This difficulty
is circumvented by using the bi-viscosity model, introduced by
Tanner and Milthorpe [22], which makes use of a critical shear rate,
c_ c , below which the material becomes very viscous, approaching
the behavior of a rigid solid. Then, the modified Herschel–Bulkley
model obeys Eq. (6) for fluids with the apparent viscosity given
by the following relation [23]:
 n1
s c_
g ¼ _0 þ K _ ; for c_ > c_c
c cc
 
ð2  c_ =c_c Þ c_
g ¼ s0 þ K ð2  nÞ þ ðn  1Þ ; for c_ < c_c
c_c c_c
As mentioned above, the critical shear rate marks the transition
from solid-like (unyielded) to liquid-like (yielded) behavior and
takes a very small value (equal to 105 in the present work), while
c_ is the magnitude of the rate-of-strain tensor given by Eq. (8). In
the present work, we restrict our simulations to the case of Bing-
ham fluids for which n = 1 and K = l.
2.2. Method of Solution and numerical schemes
The conservation equations are integrated numerically using
the FVM embedded in ANSYS – FLUENTÒ code version 12.1 [23]
in combination with the VOF model to predict the liquid–gas inter-
face. The fluid is considered to be a generalized Newtonian fluid
obeying Eq. (6) either with a constant viscosity (g = l, Newtonian
fluid) or with the Bingham model relationships of Eq. (9) with
n = 1. The first order implicit time integration scheme and the third
order QUICK interpolation scheme is used for both, the momentum
and the volume-fraction equations, and coupling of the pressure
ð4Þ and velocity is achieved using the well-known SIMPLE (Semi-Im-
plicit Method for Pressure-Linked Equations) algorithm [24]. The
convergence criteria for the inner iteration at each time step were
ð5Þ
set to 106 for all the relative (scaled) residuals.
In transient dip-coating fluid motion, two characteristic times
can be identified, a characteristic time scale associated to the first
2
Stokes problem [25] can be defined as tSt  qlc =g where lc repre-
sents the characteristic length scale defined earlier (Eq. (2)). The
second one is associated with the stretching or formation of new
interface in the meniscus region, defined as tSt  lc/V0. In the pres-
ent work and for low capillary numbers (Ca < 0.5), the smallest
ð6Þ characteristic time scale is the one calculated from the first Stokes
2
problem tSt  qlc =g. Hence, a minimum non-dimensional time step
4
of 10 , defined by the ratio of the real time step to the character-
istic time scale, is used to avoid the appearance of unphysical wig-
ð7Þ gles on the free surface. For relatively high capillary numbers
(Ca > 0.5) and to avoid divergence, a non-dimensional time step
equal to 104 is initially used and then as the film is created, it is
increased gradually to a maximum value of 102 to reduce the
overall calculation time. In the present simulations, 95% of the cell
ð8Þ Solving the transient governing equation to obtain a steady-state
solution of the flow at very low capillary numbers takes more time
than at a high capillary numbers (which corresponds to negligible
surface tension effects, hence a merely viscous flow). At very low
capillary numbers, one simulation may take up to 8 days of com-
puting time.
The FEM technique uses a remeshing algorithm to track the
fluid interface. Numerically, the interpolation schemes of quadratic
co-ordinates, quadratic velocity and linear pressure were em-
ployed, and a Picard iteration scheme was used to improve conver-
gence [26]. For the FEM computations using POLYFLOWÒ, all cases
converged. The convergence criterion, based on the relative differ-
ence between successive updates used in our simulations, was
105. For each type of field, the modification at every node be-
tween two successive iterations was compared to the maximum
value of the field at the current iteration. It is to be noted that
the FEM technique is rather expensive in terms of memory and
CPU time requirements, in contrast to the FVM-VOF combination.
Although the remeshing technique of FEM is intrinsically more
accurate than the VOF method for modest mesh sizes, this advan-
tage is offset by using sufficiently refined meshes with the VOF.
ð9aÞ 2.3. Free surface treatment and interface location
The simulation of dip coating consists of finding the free surface
ð9bÞ shape and location from an arbitrary initial position. The FVM in
FLUENTÒ uses the VOF algorithm of Hirt and Nichols [3] to capture
the air–liquid interface. This method has been shown to be more
flexible and efficient than other methods for tracking liquid–gas
interfaces see, e.g. [27,28]. This formulation uses a single phase
(one fluid) set of conservation Eqs. (4) and (5) and constitutive
Eqs. (6) and (9), and differences in material properties and surface
tension are accounted for by solving a convection equation for the
volume fraction of the secondary phase aw, assumed to be the li-
quid in the present work. Assuming incompressible flow, this
equation, in the absence of mass transfer, takes the following form:
Daw @ðaw Þ
¼ þ u  raw ¼ 0 ð10Þ
Dt @t
The volume fraction for the air aa, the primary phase; will be
computed using the following constraint:
aw þ aa ¼ 1 ð11Þ
In addition to the solution of Eq. (10), the VOF uses an accurate
algorithm for advecting the volume fraction so as to maintain con-
 A. Filali et al. / Computers & Fluids 82 (2013) 110–121
servation of mass. In this work, the geometric reconstruction
scheme is used to represent the interface between the fluids using
a piecewise linear interface construction belonging to the family of
methods introduced by Youngs et al. [29]. The resulting velocity
field from Eqs. (4) and (5) is shared among the phases. The volume
surface tension force FS depends on the interface topology. It is
modeled using the Continuum Surface Force (CSF) method of
Brackbill et al. [21], thus:
qjraw
FS ¼ r qw þqa ð12Þ
2 the bath. It was concluded that this approach of imposing the flow
where qa and qw are respectively the air and the liquid densities
and j is the curvature defined in terms of the unit normal to the
interface n, which in turn is calculated using the volume fraction
and its gradients:
 
raw
j¼rn¼r ð13Þ
kraw k
In the one-fluid model formulation volume-fraction-averaged
properties are calculated using the following relation:
/ ¼ aw /w þ ð1  aw Þ/a ð14Þ
where / can represent density or viscosity.
The film thickness h0 at exit is calculated at every time step, and
the final free surface location is reached when h0 become constant.
It is well known that the VOF method [3] produces a diffused inter-
face zone corresponding to artificial mixtures created by numerical
diffusion. To reduce the numerical diffusion zone and improve on
the accuracy of the final film thickness calculations, a well refined
mesh, in particular at the interface and near the moving plate
where the film is developing, has been used. 20 grid cells were
used across the film thickness in the axisymmetric geometry
where Ca was relatively large. For the planar case as many as 40
grid cells were used across the film thickness.
The diffusion zone at the interface is found to be very small with
the meshes used. At exit, its extent was less than 6% of the final
film thickness. In the VOF results, the location of the free surface
is hence found by creating an iso-surface where the liquid vol-
ume-fraction is equal to 1. The position of this iso-surface is then
extracted easily. The implication is that the position of the inter-
face is known with an error of less 6%.
2.4. Boundary and initial conditions
Two geometrical configurations have been used in this study;
the first one concerns dip coating of a rod in an axisymmetric
geometry (r, z) [16], and the second one a planar geometry (x, y)
of dip coating on an infinite plane [10]. As already mentioned,
the solution domain and boundary conditions for the axisymmetric
geometry are shown in Fig. 1.
A no-slip condition is adopted on the two walls and the moving
rod/plate. Thus the substrate moves at v = V0, while the non-mov-
ing walls have zero velocities. The upper edge of the air-filled do-
main was modeled as a zero gauge pressure boundary condition.
In the case of the FVM-VOF combination, the free surface profile
is not known a priori, and the problem is solved as a transient one
starting with an initial horizontal free liquid surface. The position
of the free surface is tracked with time until a steady-state condi-
tion is reached when the film thickness reaches a constant final va-
lue. Hence, initially, at t = 0, the liquid is at rest and the
computational domain is filled with the air–liquid mixture with
the liquid level set at a horizontal level z = z0. In contrast, the
FEM calculations use a remeshing method to track the free-liquid
interface, starting with a necessary ‘‘good’’ initial guess of the final
free surface position.
113
In order to maintain a near constant bath level, the domain con-
tains a small inlet flow to compensate for the liquid withdrawal
from the domain. The flow rate at the inlet boundary condition
was estimated from a relationship provided by Wilson [30].
Although this relationship is limited to Newtonian fluids it was
also used for the Bingham fluid cases. The variation of the liquid
bath level was monitored during the calculations. The bath level
was tracked at every time step during the simulation, and the re-
sults showed a difference of less than 2% between the initial and
final liquid bath levels with no impact on the flow structure in
rate had no influence on the present results.
The dimensions of both the axisymmetric and planar geometry
correspond to an experimental setup on which data for the coating
process were obtained. For the axisymmetric geometry the same
configuration and dimensions of Hurez and Tanguy have been con-
sidered [16] for the FEM calculations, while an extended domain
has been used for the FVM calculations. For the planar geometry,
the dimensions were taken from the experimental setup of the
coating applicator described by Kizito et al. [10]. Due to the sym-
metry, only half of the geometry is investigated. It consists of a ver-
tical flat plate being withdrawn with constant speed V0, from a
liquid bath of effectively infinite depth and semi-infinite horizontal
dimension. Thus, the side wall and the outlet boundary conditions
have a negligible effect on the film thickness values and the free
surface shape. As discussed by Kizito et al. [10], the dimensions
of the applicator depth considered has negligible influence on the
data obtained in their work [10].
3. Results and discussion
3.1. Mesh refinement studies
Before we embark on the effect of the different parameters, it is
instructive to present a validation of previous results for the axi-
symmetric [16] and planar case [10]. The fluid properties are listed
in Tables 1 and 2.
For the axisymmetric case, Fig. 2 shows the calculation domain
used with the mesh and boundary conditions. The fluid domain
considered for the FVM calculations has a width of 0.047 (m) and
a depth of 0.08 (m), whereas for the FEM calculations the same
dimensions of the fluid domain available in [16] were used with
0.047 (m) width and 0.063 (m) depth, starting from a guessed free
surface shape with an initial bath level at z = z0 = 0.048 (m). A
structured mesh consisting of non-uniform, rectangular cells, with
refinement near the regions of greatest interest, i.e., near the verti-
cal-wall regions, the free surface, and the expected vertical film re-
gion was used.
Grid sensitivity tests were conducted with different numbers of
cells to calculate the location of the free surface for the case in the
literature [16] with Ca = 6.03, using Fluid N1. For the FVM-VOF
method, three mesh sizes were considered: M1 with 18,000 cells,
M2 with 34,875 cells and M3 with 57,000 cells, respectively. The
Table 1
Physical properties of fluids [16].
Parameters Newtonian Bingham
Fluid N1 Fluid N2 Fluid B1 Fluid B2
Viscosity, l (Pa s) 2.9 100 2.9 2.9
Yield stress, s0 (Pa)   1.0 4.0
Density, q (kg/m3) 910 910 910 910
Surface tension, r (N/m) 0.0359 0.0359 0.0359 0.0359
Substrate speed, V0 (m/s)
V0 = 0.0241 (m/s) Ca = 1.95 Ca = 67.13
V0 = 0.0747 (m/s) Ca = 6.03
114 A. Filali et al. / Computers & Fluids 82 (2013) 110–121

Table 2
Physical properties of fluids for the planar geometry [10].

Fluid Density, q (kg m3) Surface tension, r (N m1) Viscosity, l (Pa s) Fluid property, P0
5 cs Silicone oil 916 0.0197 0.00458 0.03
20 cs Silicone oil 949 0.0206 0.01898 0.1
100 cs Silicone oil 964 0.0209 0.09640 0.56
200 cs Silicone oil 967 0.021 0.19340 1.1
1000 cs Silicone oil 970 0.0212 0.97000 5.5

Pressure outlet

Initial bath level

Wall
Axis of Moving
symmetry rod / plate

Wall Fig. 4. Grid sensitivity test with the FEM-remeshing method. Free surface profile
Fluid for a Newtonian fluid for Ca = 6.03 (axisymmetric case).

Fig. 2. Finite volume mesh with reference axes and boundaries.

meshes M2 and M5 are used in all simulations to optimize the

computation memory and time.
For the planar case, a fluid domain of 0.15 (m) width and 0.35
(m) depth is used, and the flow configuration is the one docu-
mented experimentally in [10]. For very low capillary numbers in
the range of 0.01 < Ca < 0.1, the exit film thickness h0 is very thin
requiring a fine mesh along the interface and solid boundaries.

Fig. 3. Grid sensitivity test with the FVM-VOF numerical method. Free surface
profile for a Newtonian fluid for Ca = 6.03 (axisymmetric case).

location of the free surface is plotted for the three different meshes
in Fig. 3. For the FEM-remeshing method, three mesh sizes were
considered: M4 with 2425 cells, M5 with 4850 cells and M6 with
7300 cells, respectively. The location of the free surface is plotted
for the three different meshes in Fig. 4. As the difference between Fig. 5. Grid sensitivity test for the 2-D planar case using the FVM-VOF numerical
M2, M3 and M5, M6 respectively for both methods, is insignificant; method. Free surface profile for a Newtonian fluid for P0 = 1.1 and Ca = 1.
 A. Filali et al. / Computers & Fluids 82 (2013) 110–121 115

a b

Fig. 6. Variation of the free surface shape with capillary number Ca for Newtonian fluids: (a) Ca = 1.95, (b) Ca = 6.03 and (c) Ca = 67.13 (axisymmetric case). Dimensional
results with different methods.

Numerical tests were conducted for different Newtonian fluids
with their physical properties listed in Table 2 [10].
Grid sensitivity tests were conducted with different numbers of
cells to calculate the location of the free surface for a Newtonian
fluid with P0 = 1.1 and Ca = 1. Four mesh sizes were considered:
M7 with 216,550 cells, M8 with 313,950 cells, M9 with 414,050
cells and M10 with 673,200. The location of the free surface is plot-
ted for the four different meshes in Fig. 5. Since for low capillary
numbers the film thickness is very small and needs more refine-
ment near the walls and the interface zones, mesh M6 is used in
all simulations.
3.2. Axisymmetric dip coating – Newtonian fluids
Numerical results are validated with experimental data on fluid
N1 (Table 1) used in dip coating experiments [16] with the FVM-
VOF combination. Additional numerical results based on the
FEM-remeshing technique using Polyflow software [26] are pre-
sented for comparison between both approaches in terms of free-
surface profiles, final film thickness and streamline patterns.
In the FEM-remeshing method, the initial free surface shape is
guessed, which is not the case for the FVM-VOF method where
the initial free surface corresponds to the position of the fluid at
rest, i.e., z = z0 = constant. Fig. 6b illustrates the comparison of the
calculated free-surface shape with the experimental data of [16].
Good agreement is shown between the present work and the
experimental data of Hurez and Tanguy [16] for Ca = 6.03 with a
difference of 5% for the final film thickness and a maximum of
8% in the dynamic meniscus region.
The comparison with numerical results obtained by Hurez and
Tanguy [16] for the shape of the free surface for different capillary
numbers Ca can also be seen in Fig. 6a–c respectively. For the VOF
model, the difference between the present results and those of
Hurez and Tanguy [16] is about 5% for the final film thickness
and a maximum of 7% in the dynamic meniscus region for
Ca = 1.95, see Fig. 6a, and about 7% for the final film thickness
and a maximum of 10% in the dynamic meniscus region for
Ca = 6.03, see Fig. 6b. For the FEM the difference is about 2% for
the final film thickness and 9% in the dynamic meniscus region
for Ca = 1.95, and about 4% for the final film thickness and 7% in
the dynamic meniscus region for Ca = 6.03.
However, Fig. 6c for high Ca = 67.13 shows a clear disagreement
when compared with the Hurez and Tanguy results [16]. By taking
a closer look it is clear that the results of the present work obtained
with the two different numerical methods (FVM and FEM) appear
to be closer in terms of final film thickness, location of stagnation
point, and free surface. The disagreement is confined to the menis-
cus region and is on the order of 30%. There are potentially two
116 A. Filali et al. / Computers & Fluids 82 (2013) 110–121
Newtonian Fluid
Hurez and Tanguy [16]
FVA - VOF
FEM-Remeshing
Experimental [16]
Line for visual aid
Fig. 7. Variation of the dimensional final film thickness h0 with capillary number Ca
for Newtonian fluids (axisymmetric case). Results with different methods.
main reasons for this discrepancy at this high Ca number. First the
number of elements used in [16] was rather modest (around 200)
compared to the present mesh sizes. When Ca is high, the displace-
ment of the free surface during the calculations with the reme-
shing technique is important, and using a relatively coarse mesh
leads to the distortion of the cells and eventually to divergence
as we have experienced it. A high Ca leads to a relatively larger film
thickness, and this required us to extend the meshed domain in the
vertical direction, thus providing an exit boundary condition far
away from the liquid bath free surface. It appears that the meshed
domain used by Hurez and Tanguy [16] for Ca = 67.13 was not long
enough in the vertical direction to have an exit boundary far re-
moved from the liquid bath surface to allow a good resolution of
the final film thickness. This could be the second reason to explain
the discrepancy between the present results and those of [16] as
seen in Fig. 6c.
The variation of the final film thickness h0 with capillary num-
ber Ca in a system of logarithmic coordinates is presented in Fig. 6.
The numerical results for both FVM and FEM techniques are com-
pared with experimental and numerical data of Hurez and Tanguy
[16]. As shown in Fig. 7, the final film thickness increases linearly
in log scale with capillary number. For Ca = 1.95 and 6.03, the
FEM results are closer to the numerical and experimental results
of Hurez and Tanguy [16] than the FVM results, whereas for high
Ca (Ca = 67.3), the FVM and FEM techniques agree in terms of the
final film thickness but disagree with the numerical results of Hur-
ez and Tanguy [16] as discussed above.
As an illustration of the behavior of the free surface in dip coat-
ing, streamline patterns superimposed onto contour plots of the
volume fraction are shown in Figs. 7a and 8b for Ca = 6.03 for the
FVM-VOF and the FEM-remeshing technique, respectively. The fi-
nal meniscus shape and flow field show the sharp separation of
the flow into two regions: in the upper part, the fluid is entrained
along the web while in the lower part the fluid returns to the res-
ervoir and generates relatively intense recirculation.
As a comparison between the two methods for Ca = 6.03 in
Fig. 8a and b, respectively, and as shown previously in Fig. 6b,
the results are close in terms of the free surface shape, streamline
patterns, and the position of the stagnation point in the x- and y-
directions, with a difference less than 3%. There is no discernible
difference (less than 0.5%) in the location of the eye of the flow
recirculation zone inside the bath between the FVM and FEM
calculations.
3.3. Axisymmetric dip coating – Bingham fluids
For the Bingham fluid, numerical simulations are conducted
using two fluids B1 and B2 for different values of the yield stress
s0 and/or the dimensionless Bingham number, which is defined
a
Stagnation
point
Liquid VFr = 1
Air VFr = 0
z (m)
r (m)
b
Stagnation
point
z (m)
r (m)
Fig. 8. Streamlines patterns superimposed onto contour plots of the volume
fraction of the Newtonian fluid N1 (VFr = 1) for Ca = 6.03: (a) FVM-VOF, (b) FEM-
remeshing technique (axisymmetric case).
as Bn = s0/(lV0/h0). Fig. 9 presents a comparison between the pres-
ent results and numerical results available in the literature [16] for
Bn = 0.173 and 0.7, respectively. Sensitivity to the critical shear rate
has been tested by using a smaller critical shear rate c_ c ¼ 108 s1
than the value of c_ c ¼ 105 s1 , which was the standard value used
in all of our computations. It was found that there was no discern-
ible effect on the final film thickness and the position of the stag-
nation point, except for a small difference of less than 1% in the
dynamic region of the free surface.
For Bn = 0.173 in Fig. 9a, results of the present work obtained
with the two different numerical methods (FVM and FEM) are very
similar to each other. There are, however, differences with the
numerical results of [16]. Thus, although a fairly good agreement
on the location of the free-surface shape in the upper region of
the meniscus in which the film thickness becomes constant exists,
and also close to the surface of the bath with a difference of 6%, a
relatively large difference of about 20% exists in the meniscus dy-
namic region in which the film thickness varies with the height
above the bath level and in which viscous, gravity, and surface ten-
sion have all to be taken into account. Results for Bn = 0.7 in Fig. 9b
show that the free surface shape using the FVM-VOF technique
agree with the numerical results obtained by Hurez and Tanguy
[16], while for the FEM-remeshing technique the results show a
difference of 15% in the dynamic region of the free surface but
agree well with the FVM-VOF and Hurez and Tanguy [16] results
in terms of the final film thickness.
 A. Filali et al. / Computers & Fluids 82 (2013) 110–121 117

Fig. 10 represents the variation of the free-surface shape for the

a
Bingham Fluid (Bn = 0.173)
two Bingham fluids B1 and B2 for different values of the yield
stress s0 or Bn as well as the Newtonian limit case (Bn = 0). It can
Hurez and Tanguy [16]
be noticed from Fig. 10 that entrainment of the free surface shape
FVA - VOF
FEM-Remeshing for Bingham fluids, Bn = 0.173 and 0.7 respectively, exists only in
the dynamic region. In contrast to the Newtonian fluid, the entrain-
ment of the free surface for Bn = 0.173 is around 4% and for Bn = 0.7
is 6%.
Of interest is the extent and shape of the yielded/unyielded (un-
shaded/shaded) regions separated by the contour |s| = s0, where |s|
is the magnitude of the stress tensor. When the Bingham number is
successively increased from Bn = 0.173 to Bn = 0.7 to Bn = 10, the
results are given in Fig. 11 [31]. It is shown that increasing the yield
stress affects the free-surface shape and tends to increase the
entrainment of the coating liquid. The comparison among these
cases shows that increasing the Bn number changes the shape
and extent of the yielded/unyielded regions. In particular increas-
ing Bn (a dimensionless yield stress) leads to a progressive increase
of the unyielded (shaded) region, as expected since the material
becomes more solid-like [31]. Most of the unyielded regions are
b found away from the meniscus and the free surface, in the upper
Bingham Fluid (Bn = 0.7)
area of the reservoir, where the flow is slow and not changing very
Hurez and Tanguy [16]
much. There are yielded areas in the reservoir near the entry and
FVA - VOF
the walls, where shearing occurs for the lower Bn numbers of
FEM-Remeshin g
Bn = 0.173 and 0.7. On the other hand, near the meniscus and the
free surface for the same Bn numbers, yielding occurs due to curva-
ture and shearing on the moving web. However, at higher Bn num-
bers (Bn P 10) the unyielded region is more important and covers
most of the flow domain, even under the free surface and the
meniscus, because of its small curvature, and a very uniform
plug-like velocity profile (Fig. 11c).
By considering a constant fluid viscosity (K = l, n = 1) and a con-
stant withdrawal velocity of the web, the position of the stagnation
point is not affected, but it is clear that increasing the yield stress
tends to increase entrainment and to influence the velocity field as
shown in Fig. 12. This figure represents the streamline patterns
superimposed onto contour plots of the volume fraction of the
Bingham fluid B2. It is seen that the fluid velocities tend to vanish
in the bath, limiting the zone of influence of the moving wall to the
Fig. 9. Predicted free surface location and results by Hurez and Tanguy [16] for
Bingham fluids: (a) Bn = 0.173, (b) Bn = 0.7 (axisymmetric case). Results with near field, where the fluid has a tendency to ‘‘stick’’ to the web
different methods. causing the effect of ‘‘swelling’’, as it has been mentioned in previ-
ous work [16].

3.4. Planar dip coating – Newtonian fluids

Bingham Fluid (FVM-VOF) We now turn our attention to the case of the 2-D planar geom-
Bn = 0 etry with Newtonian fluids, which was studied experimentally in
Bn = 0.173 [10]. Fig. 13 represents the variation of the dimensionless thickness
2
Bn = 0.7 T0 ½T 0 ¼ ðqgh0 =lV 0 Þ1=2  as a function of Ca for several Newtonian
fluids with different Property numbers P0. The comparison of the
Stagnation point
present results with experimental data [6,10] displays a good
agreement with a maximum difference of 12%. For Ca < 1, the
dimensionless thickness T0 increases steeply with increasing Ca,
and the effect of Re (or inertia forces) becomes negligible. For P0
values of 0.03 and 0.1, the steep increase of T0 with Ca is captured
correctly together with a slight overshoot beyond the asymptotic
value. The asymptotic value is reached earlier as P0 decreases.
The correlation developed by Wilson [30] valid for very low values
of Ca  1 is also represented, and it is interesting to see that only
partial agreement with it exists for experimental values of P0 = 0.1.
When P0 = 0.03, both experimental and numerical results exhibit
higher values for T0 than the Wilson line and a steeper variation
with P0. The numerical results of Jenny and Souhar [19] are how-
ever closer to the Wilson line albeit at low values of Ca. But overall
Fig. 10. Variation of the free surface shape with the Bingham number Bn for the departure from the Wilson line was also observed by Jenny and
Bingham fluids using the FVM-VOF numerical method (axisymmetric). Souhar [19]. For P0 = 5.5 experiments and calculations display a
118 A. Filali et al. / Computers & Fluids 82 (2013) 110–121

Fig. 11. Contour plots of the yielded (white) and unyielded (black) regions separated by the line (|s| = s0) for (a) Bingham fluid B1 with Bn = 0.173, (b) Bingham fluid B2 with
Bn = 0.7 and (c) for Bn = 10 (axisymmetric case).

2
Fig. 13. Comparison of the non-dimensional final thickness T0 ½T 0 ¼ ðqgh0 =lV 0 Þ1=2 
b as a function of Ca for different Newtonian fluids with experimental [10] and
numerical [19] and Wilson‘s model [23].

underpredicting the asymptote. For P0 equal to 1.1 the calculations

Fig. 12. Streamline patterns superimposed onto contour plots of the volume
fraction of the Bingham fluid B2 (VFr = 1) obtained with different methods: (a) FVM-
VOF, (b) FEM-remeshing technique (axisymmetric case).
good agreement showing an increase followed by a levelling-off to-
wards the asymptotic value around Ca = 1 with the calculations
follow closely the experimental values up to Ca = 0.5, beyond
which the calculation reveal a monotonic increase towards the
asymptotic value in contrast to the experiments which show a sud-
den earlier jump, albeit with some scatter. Thus when P0 > 0.5, T0
becomes constant beyond Ca = 1, which means that the surface
tension forces are no longer influential in this case. The present cal-
culations, shown in Fig. 13, indicate an asymptotic value of
T0 = 0.664 ± 0.03 as Ca is increased, which is in agreement with
the asymptotic value of T0 = 0.67 ± 0.02 obtained by Kizito et al.
[10] and the value of T0 = 0.68 obtained by Ruschak [1].
As a demonstration and validation of the present numerical re-
sults, Fig. 14 represents the comparison of the interfacial profile
between the present work and experimental results by Kizito
et al. [10] as well as experimental results by Lee and Tallmadge
[6] for P0 = 1.1 and three different values of Ca and Re. For
Ca = 0.95, the numerical predictions correlate very well with the
experimental values of [6,10], with the values of [10] being closer
to the calculations. For Ca = 2 and 2.5 it is clear that both calcula-
tions and experiments confirm that beyond Ca = 2 surface tension
forces become negligible when compared with other forces so that
 A. Filali et al. / Computers & Fluids 82 (2013) 110–121
Fig. 14. Comparison of the free surface location with experimental results of Kizito
et al. [10] and Lee and Tallmadge [6] for P0 = 1.1. Effect of capillary number, Ca
(planar case).
the interfacial profile becomes independent of Ca. Looking closely
at the profiles, pronounced discrepancies of the order of 6% be-
tween calculations and experiments exist in the meniscus region.
3.5. Inertia effects
During high-speed withdrawal of low-viscosity liquids, inertial
forces become important and influence the meniscus shape near
the liquid pool line. These forces have been neglected in most with-
drawal theories and numerical studies known to the authors
[1,5,7–9]. Relatively few works focused on the effect of inertia
forces in dip coating, e.g., Kizito et al. [10]. In order to gain a better
understanding of the effect of inertia, additional numerical simula-
tions within a range of Reynolds numbers (Re = 7.6 and 19.5) are
conducted and discussed below.
Fig. 15. Comparison of the interfacial profiles with experimental data of Kizito et al.
[10] for high Ca and high Re (planar case).
119
Fig. 15 compares interfacial profiles between the present
numerical results and the experimental data obtained by Kizito
et al. [10] in the range where inertia forces are important
(We > 2) and the film thickness is in the asymptotic range. In both
cases considered (P0 = 1.1, Ca = 4.45, Re = 7.6) and (P0 = 0.1,
Ca = 0.4, Re = 19.5). In these two cases the interface in the meniscus
exhibits a depression displaying a cusp [10]. Although the differ-
ence between the two experimental interfaces is slight, it is still
observable. The numerically predicted interfaces show a more pro-
nounced depression, and for the case with Re = 19.5 they display an
overshoot beyond the mean horizontal liquid pool line before join-
ing it again, in contrast to the case with Re = 7.9, which shows a
smooth variation of the interface.
In dip coating, a moving plate being withdrawn out of a liquid
entrains a film with a certain thickness and creates a stagnation
point located on the interface of the air–liquid, thus separating
the flow entrained up the plate and the flow back into the bath.
As shown in Figs. 16a and 17a for the velocity magnitude and
Figs. 16c and 17c for the velocity vectors, respectively, below the
stagnation point and for high Re, there exists a locally large accel-
eration of the fluid. As a consequence, the pressure decreases lo-
cally (Figs. 16b and 17b) in the flow direction and the liquid
meniscus dips below the liquid pool line. The interface then be-
comes eventually horizontal when the pressure recovers as the
meniscus goes deeper into fluid, the fluid particles decelerate and
the meniscus dips back towards the fluid. The decrease and later
increase of the pressure, as shown in Figs. 16b and 17b, creates a
cusp at the bottom of the depression. This cusp can be conducive
to air entrainment and hence is not desirable. Fig. 17c reveals com-
Velocity a
magnitude (m/s)
Static pressure
(Pa)
b
Stagnation
point c
z = z0
Fig. 16. (a) Contours of the velocity magnitude |u|, (b) static pressure, and (c)
velocity vectors, P0 = 1.1, Ca = 4.45, Re = 7.6 (planar case).
120 A. Filali et al. / Computers & Fluids 82 (2013) 110–121

Velocity
magnitude (m/s)
a

Static pressure
(Pa)

Fig. 19. Location of the wavy and non-wavy state in the log (Ca) and log (1/m).
Stagnation
point
dip coating process. The benefit of one-fluid formulation is in pro-
c viding information on the air–fluid motion near the interface as it
is a natural part of the solution. This can help in the interpretation
and prevention of undesirable interface effects.
Figs. 17a and 18b show the streamline patterns superimposed
z = z0 onto contours of the volume fraction for both cases exhibiting a
cusp, P0 = 1.1, Ca = 4.45, Re = 7.6 and P0 = 0.1, Ca = 0.4, Re = 19.5,
respectively. In both cases Re = 7.6 and 19.5, two different behav-
iors are expected in line with the results by Kizito et al. [10]. In
the first the meniscus profile goes back smoothly after the dip, in
Fig. 17. (a) Contours of the velocity magnitude |u|, (b) static pressure, and (c) contrast with the second case for high Re (Re = 19.5), where the
velocity vectors, P0 = 0.1, Ca = 0.4, Re = 19.5 (planar case). meniscus goes above and below the horizontal pool line with
incipient interface folding before it becomes eventually horizontal.
It should be noticed also that no recirculation zones were predicted
plex air motions adjacent to the free surface. These recirculation
in the hump below the interface as mentioned by Kizito et al. [10],
zones generate shearing on the interface, leading eventually to
and the fluid particles move smoothly through the curved shape of
possible air entrainment in the liquid, which is not desirable in
the free surface as confirmed by the velocity vectors in Figs. 16c,
17c and 18, respectively.
From the different simulations considered, it is found that for a
variety of the parameters examined (property number, capillary
a number and Reynolds number), the shape of the free surface can
be wavy depending on the range of the above parameters. Fig. 19
Stagnation represents the location of the wavy and non-wavy state cases for
point
the different fluids considered in this study. According to [18],
the separation line defined by 1/m  0.12Ca1.35 (m is a fluid prop-
erty number defined in [19] as m = qr3/l4g) is an approximation
of the limit between the wavy and non-wavy states. Clearly and
as expected, the cusp cases for (P0 = 0.1, Ca = 0.4 and P0 = 1.1,
Liquid V Fr = 1 Ca = 4.45) lie within the wavy region according to the separation
b Air V Fr = 0 defined by Jenny and Souhar [19].
Stagnation
point
4. Conclusions

A combined FVM-VOF numerical method was successfully ap-

Fig. 18. Streamline patterns superimposed onto contour plots of the volume
fraction (VFr = 1) using FVM-VOF for both cusp cases: (a) P0 = 1.1, Ca = 4.45, Re = 7.6,
and (b) P0 = 0.1, Ca = 0.4, Re = 19.5 (planar case).
plied to the computation of dip-coating flows under several dy-
namic conditions. The numerical simulation provides a means to
obtain details and quantitative information on the flow field that
can be difficult if not impossible to obtain using experimental
means.
Numerical calculations were performed using Newtonian and
Bingham fluids for both flat plate coating and wire coating and
 A. Filali et al. / Computers & Fluids 82 (2013) 110–121
compared successfully with experimental and for some cases with
numerical data available in the literature.
The effect of high inertia forces on the free surface are ade-
quately captured with a clear exhibit of surface waviness and a
tendency to ingest air for cases displaying a cusp. Calculations also
demonstrate that the recirculation region below the cusp men-
tioned by Kizito et al. [10] does not exist. However, air recirculation
above the interface in the meniscus region can exist and lead to air
ingestion in the wavy regime.
Overall, the successful completion of various different simula-
tion tests under different conditions suggests that the VOF ap-
proach can be useful for coating process design and optimization
studies.
Acknowledgement
Financial support for Mr. A. Filali towards his graduate studies
from the Petroleum Institute is gratefully acknowledged.
References
[1] Ruschak KJ. Coating flow. Annu Rev Fluid Mech 1985;17:65–89.
[2] Weinstein SJ, Ruschak KJ. Coating flows. Annu Rev Fluid Mech 2004;36:29–53.
[3] Hirt CW, Nichols BD. Volume of fluid (VOF) method for the dynamics of free
boundaries. J Comput Phys 1981;39:201–25.
[4] Landau LD, Levich BV. Dragging of a liquid by a moving plate. Acta Physico-
Chim USSR 1942;17:42–4.
[5] Derjaguin BV. On the thickness of a layer of liquid remaining on the walls of
vessels after their emptying, and the theory of the application of
photoemulsion after coating on the cine film. Acta Physico-Chim USSR
1943;20:349–52.
[6] Lee CY, Tallmadge JA. Meniscus shapes in withdrawal of flat sheets from liquid
baths. Dynamic profile data at low capillary numbers. Ind Eng Chem Fund
1974;13:356–60.
[7] Tallmadge JA, Soroka AJ. The additional parameter in withdrawal. Chem Eng
Sci 1969;24:377–82.
[8] Groenveld P. High capillary number withdrawal from viscous Newtonian
liquids by flat plates. Chem Eng Sci 1970;25:33–40.
[9] Spiers RP, Subbaraman CV, Wilkinson WL. Free coating of a Newtonian liquid
onto a vertical surface. Chem Eng Sci 1974;29:389–96.
121
[10] Kizito JP, Kamotani Y, Ostrach S. Experimental free coating flows at high
capillary and reynolds number. Exp Fluids 1999;27:235–43.
[11] Tanguy P, Fortin M, Choplin L. Finite element simulation of dip coating, I:
Newtonian fluids. Int J Num Meth Fluids 1984;4:441–57.
[12] Tanguy P, Fortin M, Choplin L. Finite element simulation of dip coating, II: non-
Newtonian fluids. Int J Num Meth Fluids 1984;4:459–75.
[13] Lee CY, Tallmadge JA. Description of meniscus profiles in free coating II –
analytical expressions. AIChE J 1973;19:403–7.
[14] Fortin M, Glowinski R. Méthodes de Lagrangien augmente. Application a la
résolution numérique de problèmes aux limites. Paris: Dunod-Bordas; 1982.
[15] Reglat O, Labrie R, Tanguy PA. A new free-surface model for the dip coating
process. J Comput Phys 1993;109:238–46.
[16] Hurez P, Tanguy PA. Finite element analysis of dip coating with Bingham
fluids. Polym Eng Sci 1990;30:1125–32.
[17] Yamamoto T, Nojima N, Mori N. Numerical analysis of flow behavior in
meniscus region of viscoelastic dip coating flows. J Textile Eng 2005;51:21–8.
[18] Jin B, Acrivos A, Münch A. The drag-out problem in film coating. Phys Fluids
2005;17:103603.
[19] Jenny M, Souhar M. Numerical simulation of a film coating flow at low
capillary numbers. Comput Fluids 2009;38:1823–32.
[20] Herschel WH, Bulkley R. Consistency measurements of rubber benzene
solutions. Kolloid-Zeitschrift 1926;39:291–300 [in German].
[21] Brackbill JU, Kothe DB, Zemach CA. Continuum method for modeling surface
tension. J Comput Phys 1992;100:335–54.
[22] Tanner RI, Milthorpe JF. Numerical simulation of the flow of fluids with yield
stress. In: Taylor C, Johnson JA, Smith WR, editors. Num meth lam turb flow.
Proc 3rd Int Conf, Seattle. Swansea: Pineridge Press; 1983. p. 680–90.
[23] ANSYS FLUENT 12.0/12.1 Documentation; 2009. <www.ansys.com>.
[24] Patankar SV, Spalding DB. Calculation procedure for heat, mass and
momentum transfer in 3-dimensional parabolic flows. Int J Heat Mass
Transfer 1972;15:1787–806.
[25] Schlichting H. Boundary-layer theory. New York: McGraw-Hill; 1968.
[26] ANSYS POLYFLOW 12.1 user’s guide; September 2009.
[27] Lakehal D, Meier M, Fulgosi M. Interface tracking towards the direct
simulation of heat and mass transfer in multiphase flows. Int J Heat Fluid
Flow 2002;23:242–57.
[28] Rider WJ, Kothe DB. Reconstructing volume tracking. Technical report LA-UR-
96-2375. Los Alamos Laboratory Repot; 1996.
[29] Youngs DL, Morton KW, Baines ML. Time-dependent multimaterial flow with
large fluid distortion. In: Numerical methods for fluid dynamics. New
York: Academic Press; 1982. p. 273–85.
[30] Wilson SDR. The drag-out problem in film coating theory. J Eng Math
1982;16:209–21.
[31] Mitsoulis E. In: Binding DM, Hudson NE, Keunings R, editors. ‘‘Flow of
viscoplastic materials: models and computations’’ invited review in Rheology
Reviews 2007. London, UK: Brit. Soc. Rheol.; 2007. p. 135–78.