Iranian Journal of Materials Science and Engineering, Vol.

5, Number 2, Spring 2008

MODELING SURFACE TENSION AND WALL ADHESION IN MOLD

FILLING PROCESS
N. Hatami1, R. Babaei1and P. Davami2
[email protected]
Date of Receive: January 2008 Date of Acceptance: May 2008
1
Razi Metallurgical Research Center, Tehran, Iran
2
Department of Material Science and engineering, Sharif University of Technology, Tehran, Iran

Abstract: In this study an algorithm for mold-filling simulation with consideration of surface
tension has been developed based on a SOLA VOF scheme. As the governing equations, the
Navier-Stokes equations for incompressible and laminar flows were used. We proposed a way of
considering surface tension in mold-filling simulation. The proposed scheme for surface tension
was based on the continuum surface force (CSF) model; we could confirm the remarkable
effectiveness of the surface tension by experiment which concluded in very positive outcome.

Keywords: Simulation, Surface Tension, SOLA-VOF.

1. INTRODUCTION rather than as a boundary value condition on the

interface. In 1994, Sussman, Osher and Smereka
Molecules of Fluid on, or near liquid surfaces extended the LSM method to a compressible
experience uneven molecular forces of two-phase flow. In 1998, M. W. Williams et al.
attraction. This causes the liquid surfaces to proposed the CST (Continuum Surface Tension)
possess an elastic skin (surface tension). Surface method [16]. This method generated better than
tension is an inherent characteristic of material second-order accurate approximations to the
interfaces because abrupt changes in molecular curvature of circular and spherical interfaces. In
forces occur when fluid properties change 2002, Marianne M. Francois proposed the GFM
discontinuously. Surface tension results in a (Ghost Fluid Method) and compared it with CSF
microscopic localized "surface force". These method [12]. In the same year, Berthelsen
forces exert themselves on fluid elements at showed that the LSM and CSF methods are
interfaces in both the normal and tangential equivalent [15].
directions. Fluid interfacial motion induced by In this study, the CSF method was selected as the
surface tension plays a fundamental role in many numerical method because previous methods
natural and industrial phenomena. For example, have suffered from difficulties in modeling
capillarity, low-gravity fluid flow, hydrodynamic topologically complex interfaces which have
stability, surfactant behaviour, cavitation, and surface tension.
droplet dynamics in clouds and in fuel sprays
used in internal combustion engines [1-10] are 2. PHYSICAL MODEL
examples. A Detailed analysis of these processes
typically involves the use of numerical models to The surface stress boundary condition at an
aid in understanding the resulting non-linear interface between two fluids (labeled 1 and 2) is
fluid flows. In 1988 Sethian and Osher [13] [17]:
proposed an LSM (Level Set Method). In this wV
( P1  P2  VK )nˆi (W 1ik  W 2ik )nˆ k  (1)
method, a continuous function is introduced over wxi
the whole computational domain. This function
Where ı is the fluid surface tension coefficient
has the properties of a distance function
(in units of force per unit length), PĮ is the
indicating the shortest distance to the interface.
pressure in fluid Į for Į = 1; 2, IJĮik is the viscous
In 1992, Brackbill et al. proposed the CSF
method for modeling surface tension [11]. This stress tensor, n i is the unit normal (into fluid 2)
model interprets surface tension as a continuous, at the interface, and K is the local surface
three-dimensional effect across an interface, curvature R11  R 21 , where R1 and R2 are the

1
N. Hatami, R. Babaei and P. Davami

principal radii of curvature of the surface. V can because they have a constant surface tension
only have a surface gradient; this would be, coefficient. This force has only normal
perhaps, more clearly indicated by replacing wV components; therefore, the surface force per
wx i interfacial unit area can then be written as:
& & & &
FSa ( xS ) Vk ( xS )nˆ ( xS )
in (1) by (G ik  nˆ i nk ) wV . The gradient along a
(5)
&
wx k Where K ( x S ) is the curvature considered
direction normal to the interface, ’N , is: positive if the center of curvature is in fluid 2,
& &
’N nˆ (nˆ.’) (2) and n̂ ( x S ) is the unit normal to A at x S ,
The surface tension, V , may vary along the assumed to point into fluid 2 (Fig. 1). Consider
interface and its gradient tangent to the interface two fluids, fluid 1 and fluid 2, separated by an
is defined using the differential surface operator, interface at time t. Two fluids are distinguished
’S , &
by some characteristic function, C( x S ) ,
’S ’  ’N (3) ­C1 InFluid1
& ° (6)
In this study, a one–Phase fluid flow model was C ( xS ) ®C2 InFluid 2
used, the fluid labeled 2 is empty and its density ° C ! (C  C ) / 2 Interface
¯ 1 2
is zero. Projecting (1) along the unit normal, n̂ , that changes discontinuously at the interface.
and tangent, tˆ , results in scalar boundary The CSF Method originally considered replacing
conditions for the fluid pressure in directions the discontinuous characteristic function with a
both normal and tangent to the interface ~ &
smooth variation of fluid color C( x ) from C1 to
respectively. While the normal stress boundary C2 over a distance of - (h) where h is a length
condition can be satisfied at the interface
comparable to the resolution afforded by a
between the two fluids that at rest, the tangential
computational mesh with spacing 'x . This
stress boundary condition requires the fluid to be
replaces the boundary-value problem at the
in motion. Surface tension manifests itself in the
interface with an approximate continuous model,
normal direction as a force, Vk , whitch drives
which mimics the problem specification in a
fluid surfaces towards a minimal energy state
numerical calculation, where one specifies the
characterized by the configuration of a minimum
values of c at the grid points and interpolates
surface area. Spatial variations in the surface
between them. It is no longer appropriate to
tension coefficient go along the interface ( wV ) , apply a pressure jump induced by the surface
ws
tension at an interface. Rather, surface tension
because fluids flow from regions of lower to should be considered to act everywhere within
higher surface tension. In our model, the normal the transition region. Consider the volume force,
boundary condition for interfaces is modeled & &
FSv ( xS ) , that gives the correct surface tension
where the surface tension coefficient is constant. & &
This condition is reduced to Laplace’s formula force per interfacial unit area, FSa ( x S ) , as h o 0 .
for the surface pressure (PS) where the fluid We identify this volume force for finite h as
pressure jumps across an interface under surface & & ~ &
tension, FSv (x ) = Vk ( x& ) ’C ( x ) (7)
[C ]
PS { P2  P1 Vk (4)
Surface pressure is therefore proportional to the
curvature (k) of the interface. Since surface
tension results in a net normal force directed
towards the centre of curvature of the interface,
the highest pressure is in the fluid medium on
the concave side of the interface.
3. MATHEMATICAL MODEL (CSF METHOD)
Surface tension contributes to surface pressure
(4), which is the normal force per interfacial unit
area. We consider interfaces between fluids Fig. 1. Mathematical method.

2
Iranian Journal of Materials Science and Engineering, Vol. 5, Number 2, Spring 2008

Where [c] is the jump in color, [c] = C 2  C1 . surfaces of constant color, this normal is the
The reader is reffered to ref[11] for a detailed gradient of the mollified color function,
& & ~ &
discussion of CSF method. n ( x ) ’F ( x ) (13)
The unit normal is
4. NUMERICAL MODEL ~ &
& ’F ( x ) (14)
nˆ ( x ) ~ &
4.1. Color Function ’F ( x )
For tracing the free surfaces, VOF technique is ~ &
Therefore, K .’F ( x ) is needed to evaluate the
used. Also F (0<F<1) is used as the characteristic
surface volume force, which is given by,
function in the CSF method. When ~ &
computational cells are full, F=1 becomes zero K .’F n (’.nˆ ) (15)
~
since there is no fluid in the cell. In other Since ’F is not at zero in the transition region,
research references [11] for example, density the surface volume force is also not at zero in the
functions have been the chosen characteristic transition region.
function. Consider at grid points,
~ & & 4.3. Discrete Equations
C( x ) F( x ) (8)
We have used the MAC method to discrete
Therefore the volume force is still given by (6). equations. In this method the F Function resides
The transition region thickness is then of the at cell's centers. The curvature K therefore will
order of the grid spacing, and at the points also be cell-centered. We also chose to locate
~ & &
outside the transition region, C( x ) has the FSV at cell centers. The normal vectors at the
values 0, 1 in fluids 1, 2, respectively. The
cell centers must be interpolated from nearby
interface between the fluids is given by the
cell faces in the MAC method
surface F( x& ) 1 (1  0) 1  F ! .
2 2
5. WALL ADHESION (BOUNDARY
One can multiply the integrand on the right side CONDITION
~ &
of (6) by the function g( x& ) C( x ) because of
C! The effects of wall adhesion on fluid interfaces
& &
the interface x xS and g( x& ) 1 . For in contact with rigid boundaries in equilibrium
~ & & can be estimated easily within the framework of
incompressible flow, we use C( x ) F( x ) ,
the CSF model in terms of șeq, the equilibrium
therefore g( x& ) is given by contact angle between the fluid and wall. The
&
&
g(x)
F (x) (9) angle șeq is called the static contact angle
F! because it is experimentally measured when the
And the volume force in (6), when multiplied fluid is at rest. In Fig. 2, if 0< șeq <90, the fluid
by g( x& ) , becomes: will wet the wall and if 90< șeq <180, it will also
& & & &
& & & ’F ( x ) F ( x ) separate itself from the wall. To calculate the
FSV ( x ) Vk ( x ) (10)
static contact angle we can write (Fig. 2)
[F ]  F !
With this modification, fluid acceleration due to V 12 CosT eq  V 31 V 32
surface tension is modeled as a volume force V 32  V 31 InEquilibrium (16)
Ÿ CosT eq
density. Thus, if this force is substituted into the V 12
Navier-Stoks formulation, we have:
&
du Vk’F .F (11)
{
dt [ F ]  F ! U

4.2. Evaluation of Curvature

&
The curvature of a surface A at x S , k, is
calculated from
K (’.nˆ ) (12)
where, n̂ is the unit normal to the surface. In the
CSF model, the interface is replaced by nested Fig. 2. Contact angle.

3
N. Hatami, R. Babaei and P. Davami

Where, V 12 ,V 31 and V 32 are surface tension 100012

100011
coefficients between materials labeled from 1
100010 Numerical

Pressu re [Pa]
to 3. The equilibrium contact angle is not simply
100009
a material property of the fluid. It also depends 100008
Exact
on the walls smoothness and geometry. 100007
The interfaces normal at points on the wall is 100006
nˆ nˆ wall CosT eq  nˆt SinT eq (17) 100005
Where n̂t lies in the wall and is normal to the 100004

contact line between the interface and the wall, 0 10 20 30 40 50

Distance[mm]
and n̂t is the unit wall normal directed into the Fig. 3. Comparison between numerical results and
wall. The unit normal n̂t is computed by using exact Pressure of Static liquid drop.
(13). Wall adhesion boundary conditions are 50 u 50 u 50 computational grid ( 'x =2mm) are
more complex when the contact lines are in compared with (17). The fluid drop radius is
motion, i.e., when the fluid in contact with the R = 10 and 15 (cm), density = 1000 (Kg/m3),
wall is moving relative to the wall. The background density = 0 (Kg/m3), and surface
equilibrium of the wall adhesion boundary tension coefficient = 0.07275 (N/m). The
condition in (15) may have to be generalized by pressure jump is 100005+ KV (N/m2). This value
replacing șeq with a dynamic contact angle, șd, is compared with the mean computed drop
that depends on local fluid and wall conditions. pressure obtained with the CSF model. The sum
is done over the computational cells lying within
6. STABILITY the drop that has fluid. The relative error
between the theoretical and computed drop
The explicit treatment of surface tension is stable
pressure is given by,
¦ P / N  KV
when the time step resolves the propagation of N
capillary waves [19], i 1 i
% ERR (20)
1/ 2
§  U ! ('x) 3 · KV
't s  ¨¨ ¸¸ (18) where, N is the number of cells within the drop.
© 2SV ¹
Table 1 illustrates computational errors in two
Where,  U ! ( U1  U 2 ) / 2 value of drop radius to the mesh size ratio
This condition should be added to other time Table 1. Computational Errors
steps in the limitation conditions in the algorithm
Radius /mesh size %Error
of fluid flow modeling.
10 0.219
7. NUMERICAL RESULTS 15 0.1357
To illustrate the flexibility and accuracy of the
Fig. 3 illustrates variation of theoretical and
model, we present the results of several standard
numerical drop pressure through the drop
static and dynamic problems with surface
diameter when the drop radius to the mesh size
tension.
ratio is 10 and simulation time is 0.2 s.
7.1. Static Liquid Drop Test
In the absence of viscous, gravitational, or other 7.2. Square Drop Test
external forces, surface tension causes a static When a drop is initially square, it responds to
liquid drop to become spherical. Laplace's unbalanced surface tension forces. The mesh
formula for a drop surrounded by a background size, computational grid and liquid properties are
fluid at 100005 (Pa) pressure, (4), gives the the same as in the previous test. Gravity is
internal drop pressure defined by neglected and the Square length is 32 mm.
2V Results are shown at a sequence of times, t=0,
P KV (19)
R 0.15, 0.2, 0.3, 0.4, 0.5, 0.7, 0.8, and 1.1 s in fig.4.
Where, R is the drop radius. Results in the At t=1.1 s, the drop is nearly circular in cross
Cartesian geometry using a tree-dimensional section (minimum energy state).

4
Iranian Journal of Materials Science and Engineering, Vol. 5, Number 2, Spring 2008

Fig. 4. Variation of Square drop shape.

understanding of the molten metal flow with
7.3. Mold filling test
surface tension. The numerical method is used to
In this test, we used pure melt Mg
solve for the velocity and pressure and the
(density = 1580[kg/m3], surface tension
advection of free surface is described. We have
coefficient = 0.59 [N/m]). The computational
detailed the boundary condition used and
grid is 77×14×64 and the mesh size is 4 mm.
discussed numerical stability issue. Many types
The contact angle equals 150q. As shown in Fig.
of free surface problems can now be solved with
5, the effects of surface tension were small in the
the aid of our program, as can be seen one of
mold filling, but modeling surface tension makes
them is casting. Our numerical results were
for better results. The free surface is flatter when
compared with experiments and have good
surface tension is modeled (minimum energy
agreements.
state) and fluid flow modeling results are better
as well. ACKNOLEDGMENTS
8. CONCLUSIONS This project was supported by the Razi
Metallurgical Research Center. The authors
Through a detailed study of the properties of would like to acknowledge fruitful conversation
CSF method for modeling of surface tension at with our colleagues A. Mirak, M Dadashzadeh,
fluid-gas interface, we have a deeper K. Asgharie.

5
N. Hatami, R. Babaei and P. Davami

T =0.55 Sec.

T=0.8 Sec.

T=1.0 Sec.

T=1.2 Sec.
a b c
Fig. 5. Comparison between simulation and experiments in mold filling, a-simulation with surface tension
model, b- experimental, c- simulation without surface tension model.
3. Ostrach, S: Ann. Rev. Fluid Mech. 14, 313
REFERENCES (1982).
4. Myshkis, A. D., Babskii, V. G.,
1. Levich, V. G.: Physicochemical Kopachevskii, N. D., Slobozhanin, L. A.,
Hydrodynamics. Prentice-Hall, Englewood and Tyuptsov, A. D.: Low-Gravity Fluid
Cli_s, NJ, 1962. Mechanics. Springer-Verlag, New York,
2. Lamb, H.: Hydrodynamics. Cambridge 1987.
University Press, Cambridge, 1932, 6th 5. Drazin, P. G. and Reid, W. H.:
edition.

6
Iranian Journal of Materials Science and Engineering, Vol. 5, Number 2, Spring 2008

Hydrodynamic Stability. Cambridge University of Technology, 2003.

University Press, Cambridge, 1981. 21. Babaei, R., Abdollahi, J., Homayonifar, P.,
6. Oguz, H. N. and Sadhal, S. S.: J. Fluid. Varahram, N. & Davami, P.: Improved
Mech. 194, 563 (1988). Advection Algorithm of Computational
7. Gaver, D. P., III, and Grotberg, J. B.: J. Modeling of Free Surface Flow Using
Fluid. Mech. 213, 127 (1990). Structured Grids. Computer Methods in
8. Batchelor, G. K.: an Introduction to Fluid Applied Mechanics and Engineering, V.195,
Dynamics. Cambridge University Press, 2006, PP.775-795.
Cambridge, 1967.
9. Pruppacher, H. R. and Klett, J. D.:
Microphysics of Clouds and Precipitation.
Reidel, Dordrecht, Holland, 1978.
10. Oran, E. S. and Boris, J. P.: Numerical
Simulation of Reactive Flow. Elsevier,
NewYork, 1987.
11. Brackbill, J. U., Kothe, D. B., and Zemach,
C.: A Continuum Method for Modeling
Surface Tension. Journal of Computational
Physics 100, 1992, pp.335-354.
12. Francois, M. M., Kothe, D. B., Cummins, S.
J.: Modeling Surface Tension Using a Ghost
Fluid Technique Within a Volume of Fluid
Formulation. Los Alamos, National
Laboratory, NM 87545, U.S.A.
13. Osher, S. & Sethian, J. A.: Fronts
propagating with curvature-dependentspeed
(Algorithms based on hamilton-jacobi
formulations). Journal of Computational
Physics 79, 1988, pp.12–49.
14. Sussman, M., Smereka, P. & Osher, S: A
level set approach for computing solutions to
incompressible two-phase .Flow. Journal of
Computational Physics. 1994, pp.146–159.
15. Berthelsen, P. A: A Short Introduction to the
Level Set Method and Incompressible Two-
Phase Flow. A Computational Approach,
Department of Applied Mechanics, NTNU,
A rough draft, 2002.
16. Williams, M. W., Kothe, D. B., Puckett, E.
G.: Accuracy and Convergence of
Continuum Surface Tension Models. Los
Alamos, National Laboratory, NM 87545,
U.S.A.
17. Landau, L. D. & Lifshitz, E. M.: Fluid
Mechanics. Pergamon Press, New York,
1959.
18. Levich, V.G. & Krylov, V. S.: Ann. Rev.
Fluid Mech. 1, 293 ,1969.
19. De Boor, C.: A Practical Guide to Splines.
Springer-Verlag, New York, 1967.
20. Tavakoli, R.: fluid flow simulation in low
pressure die cast. Msc thesis, Sharif