Mechanics Research Communications 63 (2015) 26–32

Contents lists available at ScienceDirect

Mechanics Research Communications

journal homepage: www.elsevier.com/locate/mechrescom

An original algorithm for VOF based method to handle wetting effect

in multiphase flow simulation
Romain Guillaument ∗ , Stéphane Vincent, Jean-Paul Caltagirone
Université de Bordeaux, Institut de Mécanique et Ingénnierie – TREFLE, UMR CNRS 8508, Site ENSCPB, 16 Avenue Pey-Berland, 33607 Pessac Cedex, France

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

Article history: The numerical modeling of contact angle and wetting surface properties is investigated with volume
Received 10 October 2013 of fluid (VOF) interface tracking method. A method is proposed to model the wetting properties with
Received in revised form 9 September 2014 an existing continuum surface force (CSF) approach by introducing an original smooth volume of fluid
Accepted 1 November 2014
function (SVOF) resulting from the solving of a penalized Helmholtz equation. For the diffusion equation
Available online 4 December 2014
proposed here, the interest of the penalty method is to account implicitly for wetting effects and, to
avoid introducing sliding laws for the velocities at the wall or modifying by hand the interface shape on
Keywords:
the solid boundaries. The SVOF model is validated and evaluated on published experiments concerning
Volume of fluid
Two-phase flow simulation
droplet impacts on flat and inclined surfaces with various wetting properties.
Contact angle model Published by Elsevier Ltd.
Droplet impact
Hydrophilic and hydrophobic surfaces

1. Introduction built by integrating the Navier–Stokes equations in each phase and

by defining jump relations at the interface to ensure mass and
The numerical simulation of two-phase flows involving wet- momentum conservation [3]. Considering fixed structured grids,
ting effects is of interest for numerous fundamental and applied the interface is not explicitly discretized on the mesh used to solve
researches in fluid mechanics applied to material selection, sur- the Navier–Stokes equations. Kataoka [4] introduce a phase func-
face treatment, heat exchangers, energy production or lubrication tion variable C to define the local volume fraction in each grid cell.
in industrial processes. The unsteadiness and complex optical prop- The Navier–Stokes equations in each phase are generalized to the
erties of two-phase flows make the experimental measurements whole two-phase medium by convolving the conservation equation
difficult to achieve and have encouraged research efforts to build by C and integrating them among all the control volumes. Finally,
efficient physical models and numerical methods. a single fluid model, also called 1-fluid model, is obtained, which
Among the wide variety of physical approaches, two main contains a new force in the momentum equations accounting for
fundamental modeling strategies exist, based on microscopic con- the surface tension effects. This single fluid model requires the solv-
siderations or continuous fluid mechanics. Molecular dynamics ing of an advection equation on C to describe the time and space
of multiphase flows involve molecular interfacial forces [1,2], evolutions of the interface. A wide variety of methods have been
which act on a smaller scale than the minimum numerically proposed over the last 20 years to numerically treat the advection
resolved structures and are responsible for an interface energy of C, such as the Front Tracking method [5], the level set technique
excess. Microscopic models integrate molecular forces into con- [6], or the volume of fluid (VOF) approach [7]. We choose to use
tinuous models. These latter models introduce diffuse interfaces the single fluid model and to improve VOF methods to simulate
and contain a complete description of capillary effects – includ- wetting effects as these approaches are among the most popular
ing wetting-through thermodynamically consistent equations. The in the field of two-phase flow simulation. In the VOF methods, the
resolution of diffuse interfaces can only be applied to problems surface tension force is classically approximated with the contin-
involving small ranges of interfacial scales to keep reasonable com- uum surface force (CSF) of Brackbill et al. [8], which formulates the
putational grids in three dimensions. interfacial force as a function of the gradients of C. An extension
The models based on continuous fluid mechanics are the most of the CSF approach is proposed in this work to treat the dynamic
commonly used in academic and industrial CFD tools. They are contact angle in VOF methods. The reader can refer to the work
of Manservisi and Scardovelli [9] for a review of two-phase flow
models dedicated to contact angle on fixed grids.
∗ Corresponding author. Tel.: +33 5 40006337. First, the filtered 1-fluid model is presented briefly. An original
E-mail address: guillaument@icmcb-bordeaux.cnrs.fr (R. Guillaument). extension of the CSF method, which uses an auxiliary diffuse VOF

http://dx.doi.org/10.1016/j.mechrescom.2014.11.002
0093-6413/Published by Elsevier Ltd.
 R. Guillaument et al. / Mechanics Research Communications 63 (2015) 26–32 27

With respect to these definitions, a single set of Navier–Stokes

equations for a two-phase flow is designed in the framework of the
1-fluid formation [4,10]:
 
∂u
 + u · ∇u = −∇ p + g + ∇ · ([∇ u + ∇ T u]) + FST (3)
∂t
∇ ·u = 0 (4)
∂C
+ u · ∇C = 0 (5)
∂t
where FST is the surface tension force depending on the interface
location. This force is modeled by a continuum surface force (CSF)
approach based on the work of Brackbill et al. [8]:
 
∇C
FST (C) = nıi =  ∇ ·   ∇C (6)
∇ C 
Fig. 1. Definition sketch of the droplet impact on a solid boundary according to the
fictitious domain approach. The notation  denotes the local mean curvature of the inter-
face,  the constant surface tension coefficient, n the normal to
the interface and ıi the Dirac function indicating the interface.
The single-field representation of the flow (3)–(5)) relies on sev-
variable, is then detailed and the numerical implementation of the
eral assumptions. First, the advection Eq. (5) on phase function
single fluid CSF model for wetting effects is explained. Several test
C describes the topological modifications of the interface with-
cases are done to show the advantages of this model. Then, the
out phase change. Through this equation, the color function C is
regularized CSF approach on the impact of a droplet on a dry surface
advected with the local fluid velocity u and consequently, the evo-
for various wetting conditions is illustrated. These numerical tests
lution of both phases (fluid 0 and 1) is known at each physical time
are used to validate the physical meaning of the numerical model
and at each position in the calculation domain. The single velocity
for wetting. Finally, concluding remarks are provided.
field of the interfacial flow is incompressible, i.e. ∇ · u = 0 through
Eq. (4), due to the fact that each phase is incompressible and no
sliding between the phases occurs on a macroscopic point of view
2. Two-phase model and numerical methods
[11].
2.1. The 1-fluid model
2.2. A smooth boundary interface model for modeling capillary
effects
As previously explained, the modeling and the simulation of
two-phase flows involving separated phases, i.e. the characteris-
One of the main limitations of recent VOF methods is associ-
tic interfacial length scale is larger than the smallest spacing of
ated with the generation of artificial deformations of the interface,
the computational grid, is classically achieved by introducing the
and the creation of small artificial blobs [12–17] when the charac-
Navier–Stokes equations in each phase [10] and by insuring the
teristic length-scale of interfaces is comparable to the local grid
connection of the velocity field at the interface by defining jump
size. Moreover, if we refer to the modeling of the surface ten-
conditions on the velocity and the stress tensor [3]. Another point
sion term (6), then the discretization of the surface tension force
of view, chosen in this work, consists in defining a unified flow
and in particular the obtention of a curvature estimate requires
model valid in all phases and at the interface. This single fluid
the approximation of second order partial derivatives of the VOF
model of Kataoka [4], also called 1-fluid, is widely used to solve in
function, whose gradients are restricted to one numerical control
problems involving three-dimensional problems and strong inter-
volume. As a consequence, the compact support of the discrete sur-
face deformations. In this modeling framework, the two-phase flow
face tension force, relying on centered schemes following the work
representation is restricted to non miscible, incompressible and
of Brackbill [8], is incomplete, as the VOF function only varies on
isothermal fluids. Moreover, a constant surface tension is assumed.
one cell. Our main idea is to build an auxiliary Smooth VOF (SVOF)
The required jump equations valid at the interface [3] are directly
function called CS , which will be obtained thanks to C, in order to
integrated in the 1-fluid model by convolving the single phase con-
estimate the curvature and the normal to the interface required in
servation equations with a phase function C describing the interface
Eq. (6). The use of additional functions the surface tension force
evolutions over time through a material advection equation. As
has been proposed by Popinet [18] and Gueyffier [19], who intro-
described in Fig. 1, this color function C is assumed to behave like
duces height functions and analytical curvature estimates in order
a Heaviside function as the following:
to calculate on a discrete level the local interface curvature.
 The building of a smooth color function was also proposed by
1 if x ∈ phase k = 1 Rudman [20] who implemented the Kernel functions and applied
C(x, t) = (1)
0 if x ∈ phase k = 0 the surface tension force formulation to the smooth C variable. The
main difference between the various smooth function approaches
The interface  between phases 0 , i.e. C = 0, and 1 , i.e. C = 1, previously proposed and our Smooth-VOF technique, is that our
is defined by the isosurface C = 0.5. Using the color function, local approach is consistent with the spreading of the phase function in
quantities such as the density  or the dynamic viscosity  are, for the normal direction of the interface.
example, defined by numerical mixture laws such as: The smooth VOF will not explicitly replace the sharp VOF func-
tion C, in order to keep the correct mass conservation properties
 = C1 + (1 − C)0 brought by the PLIC numerical algorithm [21] used in this work.
(2) The SVOF function is a post-treatment of C which should match the
 = C1 + (1 − C)0 interface position, i.e. the position M such as C(M) = 0.5 is verified in
28 R. Guillaument et al. / Mechanics Research Communications 63 (2015) 26–32

the same cells as those in which CS (M) = 0.5. In this way, the aver-
aging procedures that are required to build  and  will be based
on CS and will characterize the same fluid sub-domains as C. The
SVOF averaging and surface tension forces will be detailed in the
next sections.
The numerical procedure used to obtain CS is first explained.
It has been presented briefly by Pianet et al. [22] concerning bub-
ble dynamics in liquid columns. The parameterization of the SVOF
equation has been carried out by analogy with the unsteady diffu-
sion equation for scalar concentration transfers:

∂
+ ∇ · a∇ = 0 (7)
∂t
where is the scalar concentration and a the diffusivity coefficient.
√
It is known that the diffusion length ı is equal to a [11]. By choos-
ing a = 1, it is demonstrated that the characteristic time scale of
diffusion is equal to ı2 . In the zones where C is constant, it is
required that CS = C. By discretizing in time, with n the time index
corresponding to time n t, and t the numerical time step, we
obtain: Fig. 2. Initialization of a circular phase function – comparison of VOF function C,
and Smoothed VOF function CS obtained with D = 1.
∇· ∇ n+1
+ n+1
= n
(8)

If Eq. (8) is solved N times, with 1 ≤ n ≤ N − 1 and N t = , it tracked, it consists only in a post-processing of the real volume
ensures that T diffuses on a length equal to ı and that = C far from fraction.
the diffusion zone, i.e. the zone where concentration gradients are
null. 2.3. Management of contact angle with the SVOF boundary
By analogy with the previous concentration developments, the interface model
SVOF method consists in building a smooth VOF function CS by solv-
ing an Helmholtz equation such as (8) with a source term initially By using Eq. (9), we are able to define a smooth volume fraction
equal to the sharp VOF function C as follows: which preserves the mean position of the interface, defined as the
C = 0.5 isosurface. Maintaining a contact angle at a liquid–gas–solid
−∇ · D∇ C S,n+1 + C S,n+1 = C S,n (9)
interface consists in imposing the angle made by C = 0.5 with the
where the diffusion coefficient D is equal to Li h2 .
This parameter solid boundary. In the literature, most of the existing models for
is fixed in order to ensure that the VOF function CS spreads on a interface tracking methods on fixed grids rely on solving the con-
distance Li on each side of the interface C = 0.5. Li is set the by user, servation and interface equations without contact angle, and on
with values generally in the range between 0.2 and 2 of mesh step. correcting the interface position or slip velocity manually in the
The coefficient h is equal to the local characteristic size of the grid cells cut by the fluid interface at the fluid/solid interface, in order
cell. From a numerical point of view, Eq. (9) is implicitly discretized to satisfy the contact angle [24]. Related works are reviewed for
with finite volume and a centered scheme, to be consistent with example in the article of Manservisi and Scardovelli [9].
the Navier–Stokes equation approximation. The linear system is Our objective is to extend the use of a smooth VOF function
solved with a BICGSTAB II solver with ILU preconditioning [23]. If in order to implicitly impose the contact angle directly in the
we define * as Li h2 /N, then the numerical algorithm for obtaining Helmholtz equation dedicated to CS . Taking benefit of a twenty
CS is year experience related to previous works dedicated to fictitious
CS,1 = C domains and penalty methods [25–29], we know that it is possi-
For k = 1 . . . N − 1, solve ble to implicitly impose, in any conservation or transport equation,
∗
the value of the solving variable or its gradients by adding a vol-
−∇ · ∇ C S,k+1 + C S,k+1 = C S,k (10) ume or surface penalty term in the initial equation. This specific
At the end of the smoothing algorithm (10), i.e. for k = N − 1, we term is a control term, which acts on a numerical point of view on
have CS at the new time step that is equal to CS,k+1 = CS,N . An example the order of magnitude of the discretization matrix coefficients in
of SVOF function obtained on a grid sample is proposed in Fig. 2. order to satisfy locally the initial equation or, on the contrary, the
This figure compares functions C and CS showing that C depends new relation associated with the penalty formulation.
on mesh, whereas CS smoothes the form. It is demonstrated that Starting from the Helmholtz equation (9) on CS , we introduce a
CS = 0.5 is smoother than C = 0.5 and that the diffusion zone of CS volume penalty term depending on the penalty parameter BS . The
around the interface is controlled. Its thickness is proportional to reference value CS depends on the value of the macroscopic contact
D. Choosing D = 1 spreads the smooth interface by one grid cell on angle near the wall, noted Cp . In semi discrete form, it reads:
each side of this interface. −∇ · D∇ C S,n+1 + C S,n+1 + BS (C S,n+1 − C S ) = C S,n (12)
The key point in the SVOF approach is to use CS for modelling
the surface tension force as follows: For small values of BS , i.e. BS
→ 0, the standard smooth VOF equa-
  tion is recovered whereas for large values of BS , i.e. BS → + ∞, it is
∇CS imposed that CS,n+1 = CS . This way, if an explicit relation is found
F ST (C) =  ∇ ·   ∇CS (11)
∇ C S  between CS and the contact angle, Eq. (12) will implicitly allow the
maintaining of the contact angle while at the same time saving
Eq. (11) is a generalization of the CSF method of Brackbill [8], the smoothing of C. In fact, CS does not directly impose the con-
comparable to the one used by Sussman et al. [6] with smooth Level tact angle. The use of the penalized CS in the surface tension force
Set approaches. In our case, the smooth function is not directly through Eq. (11) makes the relation between Cp and C implicit.
 R. Guillaument et al. / Mechanics Research Communications 63 (2015) 26–32 29

Fig. 3. Definition of droplet diameter, droplet height for a droplet impact on a flat surface (a) and on an oblique surface (b)

Taking into account a dynamic wetting macroscopic scale, it is was done for the impact of various liquid droplets and different
difficult by imposing a geometric equation of a contact angle, which meshes. The numerical law therefore allows us to define the value
takes enormous time to recalculate interfaces and curvatures. Thus of CS , called Cp , to be imposed on the solid surface in the SVOF as
we show that mastering a diffusion equation of the interface is suf- shown in Fig. 5, in order to satisfy a prescribed contact angle:
ficient to obtain good behavior of fluids on the surface. So, in the
next section, an explicit relation between Cp and C is shown in Cp ( C) = 0.5[− tanh(0.025( C − 90)) + 1] (13)
order to close the equation system (11) and (12). A hyperbolic tangential behavior allows us to determine the
bijection between Cp and C when 15◦ ≤ C ≤ 170◦ , while Cp is con-
3. Numerical tests and validations stant outside this range of C .

The impact of a droplet on a flat surface is associated to complex

3.2. Impact on flat and oblique surfaces
two-phase flow phenomena involving dramatic interface deforma-
tions, which mainly depends on contact angle and surface tension.
The study of a water drop impacting a hydrophilic surface
The fluid dynamics are directly dependent on the physical and
(wetting) or a hydrophobic surface (non-wetting) is characterized
chemical properties of interfaces through capillary effects. The wet-
by simulating the conditions of Fujimoto [30] and Wang experi-
ting properties of the target solid surface, the height and diameter
ments [31] (summarized in Table 1). Dimensionless parameters are
of the droplet during time, as well as the contact angle between
introduced in order to analyse the numerical results: a dimension-
the droplet and the surface are interesting parameters to measure
less time t* equal to tv0 /dp (v0 : initial velocity of droplet).
in order to characterize the unsteady free-surface flow involved.
Three-dimensional simulations are realized on sixteen pro-
The relevant physical parameters (droplet diameter d and droplet
cessors of Xeon cluster SGI with MPI procedures. The calculation
height h), and the sequence of possible droplet shapes are presented
domain is given in Table 1.
in Fig. 3.
The impact of a droplet on an oblique surface, making an
In case (a), the wetting case of a droplet impacting on a flat
angle p with the horizontal direction, is now undertaken. The
surface is provided. The non-wetting configuration will be associ-
orientation of the surface is accounted for by changing the
ated to the same notations, with a droplet alternatively spreading
and receding over time. The subfigure (b) presents the impact of
a droplet on an oblique surface. The three first figures depict the
wetting configuration whereas the last one corresponds to the final
shape of the droplet when the surface is assumed hydrophobic. The
contact angle between the local tangent to the droplet interface at
the contact between the liquid and the solid surface is defined as
C.

3.1. Numerical law for SVOF contact angles

As previously presented in the section devoted to models and

numerical methods, it is possible to impose the wetting property
of a surface by using a penalty method in the diffusion equation
(9). If CS = 0 on the surface, a non-wetting effect is imposed while
CS = 1 provides a wetting property. The main problem of the numer-
ical method is to establish a numerical law to determine a relation
between the numerical values of CS to be used and the related con-
tact angle C to be imposed between 0◦ and 180◦ . The case of the
spreading of droplets made of various fluids on a flat surface is cho- Fig. 4. Evolution of the numerically measured static contact angle C according to
sen to determine the correlation law between CS and C . In these imposed values of Cp on a flat surface for various fluids. A correlation (black line)
is found, which is expressed by Eq. (13). Blue dots and red triangles respectively
simulations, no initial velocity is considered in the droplet, whose
correspond to oil and water fluids. Blue diamond, red square and green triangle
radius is 1.5 mm. A numerical law correlating CS with C is deduced correspond to 128, 256, 512 mesh. (For interpretation of the references to color in
regardless of the type of fluid and the mesh as shown in Fig. 4. It this figure legend, the reader is referred to the web version of the article.)
30 R. Guillaument et al. / Mechanics Research Communications 63 (2015) 26–32

Fig. 5. Imposition of Cp on the solid surface, two cases: left picture Cp = 0.8( c = 62◦ ) and right picture Cp = 0.5( c = 90◦ ).

Fig. 6. Relaxation of dimensionless droplet wetting diameter d/dp and droplet height h/dp for a water droplet impinging onto a surface – left picture (dp , vo , p , c ) = (2.15 mm,
0.514 m/s, 0◦ , 30◦ ) – right picture (dp , vo , p , c ) = (2.15 mm, 0.514 m/s, 0◦ , 126◦ ).

Fig. 7. Visualization of water drop deposition on a hydrophilic surface (column 1) and a hydrophobic surface (columns 2 and 3).

Table 1
Simulation parameters: p angle of inclined plan, c static contact angle, dp initial diameter of droplet.

Case p c dp (mm) Domain (X, Y, Z) Exponential mesh Smallest spacing (␮m)

Wang wet 0◦ 30◦ 2.15 6dp , 1.5dp , 6dp 120 80 120 20

Wang nowet 0◦ 126◦ 2.15 6dp , 1.5dp , 6dp 120 80 120 20
Fujimoto 45◦ 30◦ 1.12 6dp , 2dp , 1.5dp 200 100 200 12
 R. Guillaument et al. / Mechanics Research Communications 63 (2015) 26–32 31

Fig. 8. Time evolution of spreading length and height during liquid droplet impact
on an inclined wall for (dp , vo , p ) = (1.12 mm, 2.8 m/s, 45◦ ) – impact on a wetting
surface with C = 30◦ as in the experiments of Fujimoto et al. [30].
Fig. 9. Convergence study according to grid density for the case of the impact of a
droplet on a non-wetting surface – comparison to experiments of Wanget al. [31].

orientation of the gravity. The time step is chosen, such that the
Courant–Friedrichs–Lewy CFL condition is less or equal to 0.2. 4. Concluding remarks
An exponential mesh was used to keep Neumann boundaries
condition far away. A new Eulerian model based on a Volume Of Fluid has been pro-
posed to handle capillary effects, in particular wetting properties
of solid surfaces. The model relies on a regularized VOF function
3.3. Validations of the simulation for estimating the surface tension forces including the contact line
with a Continuum Surface Force model of Brackbill et al. [8], called
The simulated results proposed in Figs. 6 and 8, concern the “Smooth VOF”. The imposition of the contact angle at the solid-fluid
interfacial shapes of the droplet during time. interfaces uses a penalty method applied to the Smooth VOF func-
Under hydrophilic conditions, the time evolution of the dimen- tion, which is finally obtained by solving a penalized Helmholtz
sionless droplet height during time is well predicted by the SVOF equation. A bijection has been showed between the value of the
approach, as presented in Figs. 6 and 8. Concerning the dimension- SVOF imposed on the walls and the physical contact angle.
less diameter, a very good agreement is also noticed, even if small The potential of the SVOF method to model wetting effects is
differences are observed between experiments and simulations for easy to set up and leaves the free interface a near wall, whereas the
a dimensionless time between 0.6 and 1.5. These discrepancies are existing models implement a geometric building approach with
probably due to the definition of the diameter and the experimental contact angle adjustment for every iteration, or rely on the defi-
technique used to measure it, which are not described in the publi- nition of sliding velocity conditions on solid surfaces. The model
cation of Wang et al. [31]. Numerically, the maximum diameter of is validated by simulating a set of Wang et al. [31] experiments
the droplet is undertaken over time. when droplets impact on flat surfaces and the measurements of
In Fig. 6, numerical and experimental droplet dimensionless Fujimoto et al. [30] concerning the impact of droplets on inclined
diameter and height are plotted as function of time. As for the wet- walls. A very good agreement concerning the unsteady behavior of
ting case, an overall very satisfactory agreement is demonstrated the droplet height and diameter has been demonstrated in all con-
between the simulations and the measurements of Wang et al. [31] figurations. In addition, the convergence of the SVOF model with
for t* ≤ 2. After this time, discrepancies appear on the droplet diam- respect to the mesh point density has been obtained. Future works
eter, probably due to the way it is measured. In the hydrophobic will be devoted to parametric investigative study of the wetting
case, the droplet spreads on the flat surfaces until t = 4 ms (dimen- surface properties effect over droplet spreading. In addition, the
sionless time of t* = 1), while a receding motion is observed for concept of dynamic contact angle will be analyzed in terms of local
t ≤ 4 ms corresponding to a droplet’s oscillation during impact. flow dynamics, namely, we proposed the dynamic contact angle
In Fig. 7, through the instant picture of 2D plan, the 3D simula- to be assimilated to an instant snapshot of the static contact angle
tions of Wang’s experiments depending on the dimensionless time deformation under local flow conditions at the wall. Moreover, this
have been shown. model has been used in complex simulations for industrial config-
urations with a good agreement [32,33]. Finally, the correlation of
CS according to C is under improvement, by using accurate estima-
3.4. Convergence study and sensitivity to numerical parameters tions of the interface curvature, thanks to specific procedures used
in computer graphic libraries.
The sensitivity of the contact angle model to grid refinement is
investigated in Fig. 9, which concerns the impact of a water droplet
on a wetting surface. Three different constant Cartesian grids are Acknowledgements
considered, with 60 × 40 × 60, 120 × 80 × 120 and 180 × 120 × 180
grid cells respectively. Compared to experiments of Wang et al. We acknowledge the Aquitaine Regional Council for the finan-
[31], it is demonstrated that the simulations converge to the cor- cial support dedicated to a 432-processor cluster investment,
rect solution and that a 120 × 80 × 120 grid is sufficiently refined located in the I2M Institute. This work was also granted access
to capture the dynamics of the droplet impact. The decrease of the to the HPC resources of CCRT, CINES and IDRIS by GENCI (Grand
droplet height is not presented in Fig. 9 as the grid density has little Equipement National de Calcul Intensif) under reference number
effect on this parameter. x2009026115 and x2010026115.
32 R. Guillaument et al. / Mechanics Research Communications 63 (2015) 26–32

References [19] D. Gueyffier, J. Li, A. Nadim, R. Scardovelli, S. Zaleski, Volume of fluid interface
tracking with smoothed surface stress methods for three dimensional flow, J.
[1] D. Jamet, D. Torres, J. Brackbill, On the theory and computation of surface ten- Comput. Phys. 152 (1999) 423–456.
sion: the elimination of parasitic currents through energy conservation in the [20] M. Rudman, A volume-tracking method for incompressible multifluid flows
second-gradient method, J. Comput. Phys. 182 (2002) 262–276. with large density variations., Int. J. Numer. Meth. Fluids 28 (1998) 357–378.
[2] Y. Yamamoto, T. Ito, T. Wakimoto, K. Katoh, Numerical simulations of spon- [21] D. Youngs, K. Morton, M. Baines, Time-dependant multimaterial flow with large
taneous capillary rises with very low capillary number using front tracking fluid distorsion, in: Numerical Method for Fluid Dynamics, Academic Press,
method combined with generalized navier boundary condition, Int. J. Multiph. New York, 1982.
Flow 51 (2013) 22–32. [22] G. Pianet, S. Vincent, J. Leboi, J.P. Caltagirone, M. Anderhuber, Simulating com-
[3] J. Delhaye, Jump conditions and entropy sources in two-phase systems. Local pressible gas bubbles with a smooth volume tracking 1-fluid method., Int. J.
instant formulation, Int. J. Multiph. Flow 1 (1974) 395–409. Multiph. Flow 36 (2010) 273–283.
[4] I. Kataoka, Local instant formulation of two-phase flow, Int. J. Multiph. Flow 12 [23] I. Duff, P. Amestoy, J.Y. L’Excellent, Multi-frontal parallel distributed symmet-
(5) (1986) 745–758. ric and unsymmetric solvers, Comput. Methods Appl. Mech. Eng. 184 (2000)
[5] S. Unverdi, G. Tryggvason, A front-tracking method for viscous, incompressible, 501–520.
multi-fluid flows, J. Comput. Phys. 100 (1992) 25–37. [24] M. Bussmann, S. Chandra, J. Mostaghimi, Modeling the splash of a droplet
[6] M. Sussman, P. Smereka, S. Osher, A level set approach for computing solutions impacting a solid surface., Phys. Fluids 12 (2000) 3121–3132.
to incompressible two-phase flow, J. Comput. Phys. 114 (1994) 146–159. [25] K. Khadra, S. Parneix, P. Angot, J. Caltagirone, Fictitious domain approach for
[7] C. Hirt, B. Nichols, Volume of fluid (VOF) methods for the dynamics of free numerical modelling of Navier–Stokes equation., Int. J. Numer. Meth. Fluids 34
boundaries, J. Comput. Phys. 39 (1981) 201–225. (2000) 651–684.
[8] J. Brackbill, B. Kothe, C. Zemach, A continuum method for modelling surface [26] S. Vincent, T. Randrianarivelo, G. Pianet, J. Caltagirone, Local penalty methods
tension, J. Comput. Phys. 100 (335) (1992) 335–354. for flows interacting with moving solids at high reynolds numbers., Comput.
[9] S. Manservisi, R. Scardovelli, A variational approach to the contact angle dynam- Fluids 36 (2007) 902–913.
ics of spreading droplets, Comput. Fluids 38 (2009) 406–424. [27] T. Randrianarivelo, G. Pianet, S. Vincent, J. Caltagirone, Numerical modelling of
[10] R. Scardovelli, S. Zaleski, Direct numerical simulation of free-surface and inter- solid particle motion using a new penalty method, Int. J. Numer. Meth. Fluids
facial flow, Ann. Rev. Fluid Mech. 31 (1999) 567–603. 47 (10–11) (2005) 1245–1251.
[11] J. Magnaudet, Simulation d’écoulements en présence d’interfaces. Lecture [28] S. Sarthou, S. Vincent, J. Caltagirone, P. Angot, Eulerian–Lagrangian grid cou-
Notes, Ecole de Printemps de Mécanique des Fluides Numérique, 1997. pling and penalty methods for the simulation of multiphase flows interacting
[12] J. López, J. Hernández, P. Gómez, F. Faura, A volume of fluid method based with complex objects., Int. J. Numer. Meth. Fluids 56 (2008) 1093–1099.
on multidimensional advection and spline interface reconstruction, J. Comput. [29] D. Lacanette, A. Gosset, S. Vincent, J. Buchlin, E. Arquis, Macroscopic analysis of
Phys. 195 (2004) 718–742. gas-jet wiping: numerical simulation and experimental approach, Phys. Fluids
[13] J. López, J. Hernández, P. Gómez, F. Faura, An improved PLIC-VOF method for 18 (2006), 042103:1-14.
tracking thin fluid structures in incompressible two-phase flows, J. Comput. [30] H. Fujimoto, Y. Shiotani, A. Tong, Three-dimensional numerical analysis of the
Phys. 208 (2005) 51–74. deformation behavior of droplets impinging onto a solid substrate., Int. J. Mul-
[14] E. Puckett, A. Almgren, J. Bell, D. Marcus, W. Rider, A high-order projection tiph. Flow 33 (2007) 317–332.
method for tracking fluid interfaces in variable density incompressible flows, [31] M. Wang, F. Lin, J. Ong, S. Lin, Dynamic behaviors of droplet impact and spread-
J. Comput. Phys. 130 (1997) 269–282. ing – water on glass and paraffin, Colloids Surf. A: Physicochem. Eng. Aspect
[15] E. Puckett, A. Almgren, J. Bell, D. Marcus, W. Rider, Reconstructing volume 339 (2009) 224–231.
tracking, J. Comput. Phys. 141 (1998) 112–152. [32] R. Guillaument, A. Erriguible, C. Aymonier, S. Marre, P. Subra-Paternault,
[16] R. Scardovelli, S. Zaleski, Interface reconstruction with least-square fit and split Numerical simulation of dripping and jeting in supercritical fluids/liquid micro
Eulerian–Lagragian advection., Int. J. Numer. Meth. Fluids 41 (2003) 251–274. coflows., J. Supercrit. Fluids 81 (2013) 15–22.
[17] D. Harvie, D. Fletcher, A new volume of fluid advection algorithm: the stream [33] R. Guillaument, S. Vincent, J.P. Caltagirone, M. Laugier, P. Gardin, Plate-out mod-
schemes, J. Comput. Phys. 162 (2000) 1–32. elling for cold-rolling systems lubricated with oil-in-water emulsions., Proc.
[18] S. Popinet, An accurate adaptive solver for surface-tension-driven interfacial Inst. Mech. Eng. Part J: J. Eng. Tribol. 225 (9) (2011) 905–914.
flows, J. Comput. Phys. 228 (2009) 5838–5866.