3D Phase Field Modeling of Electrohydrodynamic Multiphase Flows
Transcript of 3D Phase Field Modeling of Electrohydrodynamic Multiphase Flows
Accepted Manuscript
3D Phase Field Modeling of Electrohydrodynamic Multiphase Flows
Qingzhen Yang, Ben Q. Li, Yucheng Ding
PII: S0301-9322(13)00093-1
DOI: http://dx.doi.org/10.1016/j.ijmultiphaseflow.2013.06.006
Reference: IJMF 1918
To appear in: International Journal of Multiphase Flow
Received Date: 14 January 2013
Revised Date: 12 June 2013
Accepted Date: 15 June 2013
Please cite this article as: Yang, Q., Li, B.Q., Ding, Y., 3D Phase Field Modeling of Electrohydrodynamic Multiphase
Flows, International Journal of Multiphase Flow (2013), doi: http://dx.doi.org/10.1016/j.ijmultiphaseflow.
2013.06.006
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1
(Submitted to Journal of Multiphase flows, January 2013; revised June 2013)
3D Phase Field Modeling of Electrohydrodynamic Multiphase
Flows
Qingzhen Yanga,b
, Ben Q. Lib,*
, Yucheng Dinga
aState Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong
University, Xi’an, Shaanxi 710049, P. R. China
bDepartment of Mechanical Engineering, University of Michigan, Dearborn, Michigan
48128, U.S.A
*Corresponding author: Ben Q. Li (Email: [email protected]).
ABSTRACT: A 3D phase field model is developed to investigate the
electrohydrodynamic (EHD) two phase flows. The explicit finite difference method,
enhanced by parallel computing, is employed to solve the coupled nonlinear governing
equations for the electric field, the fluid flow field and free surface deformation.
Numerical tests indicate that an appropriate interpolation of densities within the interface
is critical in ensuring numerical stability for highly stratified flows. The 3D phase field
model compares well with the Taylor theory for the deformation of a single dielectric
droplet in an electric field. Computed results show that the deformation of a leaky
dielectric droplet in an electric field undergoes various stages before it reaches the final
oblate shape. This is caused by the free charge relaxation near the fluid-fluid interface.
The coalescence of four droplets in an electric field illustrates a truly 3D deformation
behavior and a complex evolving fluid flow field associated with the participating
droplets. The coalescence is a result of combined actions produced by the global electric
force, the circulatory flows generated by the local electrohydrodynamic stress and the
2
electrically-induced deformation. The 3D phase field model is also applied in modeling
of an electrohydrodynamic patterning process for manufacturing nanoscaled structures, in
which complex 3D flow structures develop as the electrically-induced deformation
evolves.
Keywords: phase field method; leaky dielectric model; electrohydrodynamics; two phase
flow; 3D modeling
1. Introduction
The earliest account of electrohydrodynamic (EHD) research is attributed to Gilbert
(1958) in the 17th century, who observed the conical shape upon placing a charged rod
above a sessile drop. Later, Rayleigh (1882) studied the deformation and the bursting of
charged drops in an electric field. The experiments conducted by Allan and Mason (1962)
eventually motivated Taylor (1966) to propose the concept of “leaky dielectrics”, which
provides an accurate account for the electric charge behavior in a poorly conducting
liquid. Within the framework of the leaky dielectrics, Taylor (1966) developed a
theoretical model for the deformation of a single drop immersed in a host liquid of
different electrical properties in a uniform electric field. Over the last decade, the EHD
flow has received much attention in research community because of its widespread
industrial applications, which, among others, include coating flows (Fermin, 2001),
transport of small liquid samples in microfluidics (Stone et al., 2004), various
microfluidics applications that use the concept of the lab-on-a-chip (Zeng and Korsmeyer,
2004) and micro- or nano-structure formation in soft lithography (Chou et al., 1999;
Chou and Zhuang, 1999; Schaffer et al., 2000; Wu and Russel, 2009; Yang et al., 2013).
As with other complex flow problems, only a limited number of the EHD flow
3
problems within an idealized setting can be solved analytically. Numerical simulations,
more often than not, are required to obtain solutions for the flow systems involving two
fluids with different properties. In numerical solution of two phase flow problems, the
presence of the free surface requires a special treatment. Different numerical schemes are
now available, and the popular ones include the level set method (Adalsteinsson and
Sethian, 1995; Sussman et al., 1994), the volume of fluid method (Hirt and Nichols, 1981;
Nichols et al., 1980; Puckett et al., 1997), and the front tracking method (Tryggvason et
al., 2001). The phase field method, which is based on the coarse-grain averaging concept
from statistical mechanics, has recently emerged as a useful tool for the study of the two
phase flow problems (Anderson et al., 1998; Jacqmin, 1999; Badalassi et al., 2003). A
salient feature of this method lies in its treatment of a moving or free surface as a thin
molecular diffuse layer, thereby enabling the molecular or nanoscaled description of the
collective behavior of molecules within the interface layer. Thus, within the framework
of the phase field modeling, a sharp fluid-fluid interface is represented by a narrow layer
in which the fluids may mix. Tracking of the interface is realized by a conserved order
parameter (or the phase field parameter) that varies continuously over the thin interfacial
layer but otherwise mostly uniform in the bulk fluid phases (Kim, 2012). The concept of
a diffuse interface was first proposed by Van de Waals (1893) long ago, but it has gained
popularity only in recent years as a tool for numerical simulations of two phase flows
(Carlson et al., 2010; Villanueva and Amberg, 2006).
Some numerical studies on the problems of two phase flows in electric fields have been
presented by various researchers. Feng (1996) calculated the equilibrium shape of a leaky
dielectric droplet in an electric field by employing the Galerkin finite element method,
4
coupled with the spine-parameterization of the free surface. The scheme works well for a
single droplet but can be difficult to apply for a system involving multiple droplets (Li,
2006; Huo and Li, 2006). Sherwood (1988) developed a boundary integral approach to
simulate the large deformation of a droplet with finite conductivity in a creeping flow.
The boundary integral model is capable of describing the drop breakup caused by either
the electric stress or the mechanical stress due to the fluid motion. The EHD two phase
flows have also been studied by the level set and volume of fluid methods (Tomar, 2007),
and by a front tracking/finite volume method (Hua, 2008). The latter also includes
various electric models (i.e., leaky dielectric, perfect dielectric and constant charged
models) for the fluids. Recently, Lin et al. (2012) presented a phase field model for the
2D axisymmetric EHD flows and studied the electro-coalescence of two droplets in an
electric field.
This paper presents a 3D phase field model for the EHD two phase flows. The thrust for
the work is derived from the need to develop a comprehensive numerical model for
complex, evolving two phase flows and free surface deformation in the presence of an
external electric field. Such a numerical model is not only of crucial importance to our
fundamental understanding of the physics governing the fluid flow and droplet formation
but also of engineering significance in developing practical guidelines for scalable top-
down fabrication processes for mass production of functional micro- or nano-structures.
For many of these applications and for others such as the electro-coalescence of multiple
droplets, a full 3D representation is required. While considerable amount of work has
been done on the 2D (and/or axisymmetric) EHD two phase flows, there appears to have
been little work reported on the full 3D phase field modeling for the EHD two phase flow
5
problems. An extension of 2D phase field to 3D phase field code can be involved in
practice, though theoretically it may appear straightforward. The data management and
CPU efficiency can be much more demanding for 3D coding and modeling. The
complexity is further compounded by the fact that for a 3D phase field based EHD model,
variables describing multiphysics needs to be solved simultaneously. In what follows, the
mathematical formulation and numerical development of the phase field model for the 3D
EHD flows of leaky dielectric fluids are described. Numerical and parallel computing
performances are discussed and the model validation is carried out by comparing with the
celebrated Taylor theory for droplet deformation in an electric field. The computed
results of 3D EHD two flows are selected from the examples of the electro-coalescence
of multiple droplets and of the EHD patterning processes for micro- or nano-structure
formation.
2. Mathematical Formulations
2.1 Phase field equations
In phase field the free energy density is represented by the phase parameter C, the form
of energy f : [0, 1] → R is taken as
2 21 21 1
12 4
f C C C C , ΩT : =Ω ×(0,T) (1)
where Ω is a bounded domain in R3, with a Lipschitz boundary ∂Ω, and C is the phase
parameter with the property that values of C=1 and C=0 correspond to the two distinctive
phases. Also, in the above equation, γ stands for the surface tension, ξ the interface
thickness and 6 2 a constant (Ding et al., 2007). The first term in Eq. (1) accounts for
the excess free energy due to the inhomogeneous distribution of volume fraction in the
6
interface region, whereas the second term represents the bulk energy density (Cahn and
Hilliard, 1958).
The Cahn-Hilliard equation with convection term, which is a result of the minimization
of the free energy, is employed here to describe the evolution of the phase field parameter
C (Jacqmin, 1999),
0C
u C Mt
, ΩT (2)
where u represents the fluid velocity, M is the phase field mobility and is considered as a
constant in this study, is the chemical potential which is defined f
C
. In this
equation, both convection and diffusion effects are included.
2.2 Electric field equations
For the EHD problems, the Poisson equation may be used to describe the electric field
distribution,
0e
r C V ΩT (3)
where ε0 is the permittivity of vacuum, εr(C) the dielectric constant, V the electric
potential, and ρe the free charge density. In the equation, the magnetic effect is neglected
since for the system under consideration, the electric field changes rather slowly, in
comparison with the diffusion of the electric field.
The free charge conservation can be expressed as (Saville, 1997),
e
eu C Et
ΩT (4)
where ζ(C) denotes the electrical conductivity, and E V is the electric strength. The
second term on the left represents the transport of the free charges by convection, and the
7
term on the right stands for the transport by electromigration. The diffusion effect is
ignored here (Saville, 1997).
There are two limiting cases for Eq. (4). First, if the time scale of charge relaxation
0 rt is much smaller than that of the flow c c ct L U (Lc being the length scale and
Uc the characteristic velocity), then Eq. (4) reduces to
0C V ΩT (5)
Thus, for fluids having a high electrical conductivity, Eq. (5) may be used (Hua et al.,
2008; Lin et al., 2012; Saville, 1997; Tomar et al., 2007). Second, for poorly conductive
materials, tσ>>tc. As a result, the free charge may be neglected and a pure dielectric
model is adequate to describe the electric field (Hua et al., 2008; Lin et al., 2012). For
this limiting case, Eq. (4) becomes redundant, and Eq. (3) recovers to the Laplace
equation. In this study, Eq. (4) is employed since it is a general description of electrical
charge conservation in an electric field.
2.3 Fluid flow equations
The flow field is described by the governing equations of mass and momentum
conservation with the momentum sources originating from the electric field and
interfacial energy. Both phases are considered to be incompressible. Hence the mass
conservation over the whole simulation domain, including both fluid phases and their
interface, can be expressed as (Ding et al.,2007)
0u ΩT (6)
In the EHD system, the electric and the surface tension forces on the interface should be
added. The modified Navier-Stokes equation for variable density and viscosity can be
written as
8
+ e
uC C u u f f
t
ΩT (7)
where the stress tensor П, the electric force ef due to polarization charge and free charge,
and the surface tension force f are given by the following expressions,
,
T
i jp C u u (8)
2 20 0 , 0
1 1
2 2
ee i j rf E E E E E C
(9)
f c (10)
In Eqs. (7) - (10), ρ(C) represents the density of fluid, p the pressure, μ(C) the viscosity,
δi,j identity matrix, and the chemical potential.
2.4 Nondimensionalization
The governing equations above may be nondimensionalized with the characteristic
length Lc, the velocity Uc, and the time tc= Lc/Uc. Here, Uc is determined by balancing the
electrical force and viscous force, Uc=ε0E02Lc/µ0 (Saville, 1997), where µ0 is the
characteristic viscosity which maybe µ1 or µ2 depending on the detail problem. The
nondimensionalization process produces the following dimensionless parameters,
0
0
c cU LRe
0 cU
Ca
c cL U
PeM
20 0r c
E
E LBo
c
CnL
(11)
The well-known Reynolds (Re) number describes the relative importance between the
inertial and the viscous force, the Capillary (Ca) number the relative magnitude of
viscous force and interface tension force, the Peclet (Pe) number the ratio of the
convective over diffusive mass transport, and the electrical Bond number (BoE) the ratio
of the electric over the surface tension force. The Cahn number (Cn) describes the ratio of
9
the interface thickness over the characteristic length. Cn is a reflection of how the
interface is resolved (Kim and Lu, 2006; Lin, 2012). When Cn →0, the phase field model
approaches the sharp interface limit (Jacqmin, 1999). The density, electrical permittivity,
conductivity, and viscosity ratios of the two fluids are: 1 2 , 1 2 , 1 2 ,
1 2 , respectively.
2.5 Numerical Method
The above equations are discretized and solved numerically along with appropriate
boundary conditions for specific problems given below. The finite difference numerical
scheme with explicit time-marching (Zikanov, 2010) is employed to obtain numerical
solutions. The first order terms are discretized by the upwinding scheme and all the
variables are defined at the collocated mesh points (Johnston and Liu, 2002). To speed up
the calculations, the code is enhanced by implementing a parallel computing algorithm.
After discretization, the Poisson equation (Eq. (3)) can be written in the form Ax=b,
where A coefficient matrix, x the unknown, b the force vector, and is solved by the
successive over-relaxation (SOR) method. The iteration follows the scheme:
x(k+1)
=(D+ωL)-1
ωb-[ωU+(ω-1)D]x(k)
, where D, L and U are the diagonal, lower and
upper triangular sub-matrices decomposed from matrix A. To speed up the convergence
rate, the relaxation coefficient is set at ω>1. The governing equation of free charge
conservation (Eq. (4)) is solved by the explicit finite difference scheme. The calculation
starts with the initial volume charge set to zero everywhere. The Navier-Stokes equation
(Eq. (7)) is solved by the standard projection method (Li, 2006; Zikanov, 2010), where
the resulting Poisson equation is solved also by the SOR method. The forth order finite
difference scheme is employed to solve the Cahn-Hilliard equation (Eq. (2)). It worth
10
pointing out that the material parameters εr(C), σ(C), ρ(C), and μ(C) may be taken as a
linear function of phase parameter C within the interface region. Numerical experience
suggests, however, that the harmonic interpolation yields a better, more consistent
approximation, when a large difference in density across the interface exists (Kim, 2012),
viz.
1 2 1
1 1 1 1=
2i i i
(12)
This is because in general, the harmonic interpolation gives a more accurate account of
flux balance over a cell within which an abrupt change in properties exits (Patankar,
1980).
3. Results and Discussion
3.1 Computational Details
Before the numerical results are presented, computational aspects concerning the
numerical accuracy warrant some discussion. Here, the mesh sensitivity is discussed first,
for which the case of the single leaky droplet deformation under the electric field is
chosen. The electric field configuration and the computational domain are shown in Fig.
1. For the droplet of radius R, a domain of (6R×6R×6R) is chosen for computation. Since
the deformation is symmetric about the planes of x=0 and z=0, a 1/4 domain is needed
(see Fig. 1b). For the problem, the characteristic length is chosen to be the radius of the
droplet, and the parameters for computations are: Re=1, Ca=0.2, Pe=1800, Cn=0.025,
BoE=0.2, λρ=1, λε=2, and λµ=1. For the results presented below, the droplet is labeled as
liquid 1 and the host medium liquid 2. The dimensionless parameters are scaled by the
11
properties of the fluid 2, with Lc=R, the radius of the sphere. This remains true henceforth
unless otherwise indicated.
The computed results are listed in Table 1 for various meshes. Apparently, a mesh of
81×161×81 gives a reasonably good accuracy and is thus used for the computations
discussed below. Studies suggested that a spontaneous shrinkage of the drop occurs if the
mesh is too coarse (Yue et al., 2007; Donaldson et al., 2011). This was also observed
with the simulations in the present study. For the given parameters above, the problem of
artificial droplet shrinkage will not occur with the mesh of 81×161×81.
For the finite difference solution of a two phase flow problem, the variable density
across the fluid-fluid interface can be modeled as either a linear function (Badalassi et al.,
2003; Kim, 2005) or a harmonic function (Khatavkar et al., 2007; Liu and Shen, 2003),
that is,
1 2 1C C C (D1), 1 2
1 1=
C C
C
(D2).
For the convenience of discussion below, D1 and D2 will be used to represent the linear
and harmonic approximations, respectively. Also one of the following interpolation
schemes is used to obtain the value of the density at the center of an element may be
interpolated from the values at the two side points,
1 2 1
1
2i i i (I1),
1 2 1
1 1 1 1=
2i i i
(I2),
where ρi and ρi+1 are the densities at the nodal points. Again for the convenience, I1 and
I2 will be used to denote the two interpolations, respectively.
It is known that for two phase flows with a large density ratio, the local imbalance
between the pressure gradient and the surface tension leads to spurious currents (Lafaurie
12
et al., 1994; Ni, 2009). For some cases, the spurious flow currents gradually disappear
with the time marching forward, as the diffusion effect eventually dominates within the
interface. However, if the surface tension effect is strong, these spurious currents will
exacerbate to cause a disruptive instability in flow simulations. This is further aggravated
if the density ratio of the two fluids is very large and the problem of spurious currents
will not simply disappear just with further refinement of mesh size. To assess the
numerical accuracy of different approximations, four different combinations, D1-I1, D1-
I2, D2-I1 and D2-I2, of the above interpolations are studied. As a testing case, a droplet
of perfect sphericity is considered, with both gravity and the external electric fields set to
zero and with R=0.18mm. Theoretically, the velocity should remain zero at t ≥ 0 for these
conditions. Numerically, however, the spurious currents may occur depending upon the
interpolation functions used for approximating the density variation in the fluid-fluid
interface region. Tests show that for a large density ratio, e.g., ρ1/ρ2=1000, the D1-I1
approximation scheme becomes unstable with time, leading to a disruption in
computation. Clearly, simple arithmetic interpolations do not bode well for a large
density ratio problem. All the other three schemes use the harmonic interpolations,
though differing in details; they give better results. Schemes D2-I1 and D2-I2 create
disturbances in velocity (i.e., spurious currents) initially but these disturbances eventually
disappear as simulation is allowed to continue for sufficiently long time. The D1-I2
approximation, however, performs the best and the spurious current is well suppressed
after a few time steps. These results are shown in Fig. 2.
3.2 Deformation of a Single Droplet (Model Validation)
It is well known that when a droplet and its surrounding fluid are both leaky dielectrics,
13
the droplet can deform into either a prolate ellipsoid or an oblate ellipsoid in a DC
electric field (Taylor, 1966). For the configuration shown in Fig. 1, an analytical
expression can be obtained by Taylor’s theory, which now is widely used as a benchmark
test for numerical computations (Hua et al., 2008; Lin et al., 2012; Tomar et al., 2007).
By this theory, the deformation of the droplet by the electric forces obeys the following
relation (Taylor, 1966; Torza et al., 1971),
1 2 1 2 1 2
2
1 2
/ , / , /9
16 2 /
dEfL B Bo
DL B
(13)
where L and B are the length of the parallel and perpendicular axes of the ellipsoid with
respect to the applied electric field, and the subscripts 1 and 2 denote the droplet and the
surrounding host fluid, respectively. The function, 1 2 1 2 1 2/ , / , /df , represents the
discriminating function,
2
1 21 1 1 1 1 1 1
2 2 2 2 2 2 2 1 2
2 33, , 1 2
5 1df
(14)
where the droplet takes a spherical shape for fd=0, a prolate shape for fd <0 and an oblate
shape for fd >0.
Computations were performed on a cube Ω = [0,l ]x[0,l ]x[0,l ] where l = 6R, and the
boundary conditions are taken as the same as in Lin et al. (2012), the difference being
that here the 3D calculations are considered. The boundary conditions are listed as
follows,
0 0 0,V V u y l T
0 0 0 0,V u y T
0 0 0 0 0,2 2
V l lu n x x z z T
n
14
0 0 0 0,p C
Tn n n
where ∂Ω is the boundary of computational domain. Initially the velocity is set to zero
and the phase field is C=1 for the droplet and C=0 elsewhere. Because of symmetric
conditions on the z=l/2 and x=l/2 planes, only a quarter of the domain needs to be
discretized. The calculations below used a mesh of 81x161x81. The original radius of the
droplet R=l/6, and it is also used as the characteristic length. Other parameters are: Re=1,
Ca=0.2, Pe=1800, Cn=0.025, BoE=0.2, λρ=1, λε=2 and λµ=1.
Fig. 3 compares the results obtained from the present 3D phase field calculations and
from the analytical solution by Taylor’s theory. As it is seen, for small droplet
deformations, the agreement between the 3D numerical and the analytical solutions is
excellent. For large deformations, however, the numerical model deviates from the
analytical solution. This is expected in that the Taylor theory is based on the small
perturbation analysis, which is valid only for small droplet deformations.
3.3 Transient Behavior of Electrically-induced Droplet Deformation
While the analytical solution is for a steady state deformation, the phase field model
presented is fully 3D and transient, which allows us to capture the transient behavior of a
droplet as it undergoes deformation. Knowledge on the transient development of droplet
deformation can be of great value in analyzing the underlying physics governing the
electrohydrodynamics of a deforming droplet. One interesting case is presented in Fig. 4,
where the final (or equilibrium) droplet shape is oblate. The droplet deformation,
however, for this case undergoes several different phases as the droplet deformation
transits in time to reach its equilibrium shape from its initial condition. The transient
15
development of the deformation experiences the following stages: it starts with a
spherical shape, evolves to a prolate ellipsoid, returns to spherical, and finally becomes
oblate. The detailed flow velocity field, flow streamlines and droplet deformation are
plotted in Fig. 4, displaying different stages of droplet deformations. The parameters used
for the calculations are: Re=4, Ca=0.8, Pe=7200, Cn=0.025, BoE=0.8, λρ=1, λε=5, λµ=1
and λζ=0.2. The conductivity of droplet and surrounding liquid are 4.0×10-11
S/m and
2.0×10-10
S/m, respectively. Analysis of the simulation data reveals that this transition in
deformed shapes is a result of electric charge interacting with the fluid flow field. At the
onset of the deformation process, there is no free charge and, the whole system behaves
like pure dielectric materials. The droplet deformation is dominated by the dielectric
forces, which stretches the droplet along the electric field and thus leads to a prolate
ellipsoidal shape. This has been shown by both experimental measurements and
theoretical analyses (Hua et al., 2008; Tomar et al., 2007). As the time goes by, free
charges starts to accumulate at the droplet-fluid interface as demanded by Eq. (4), and the
additional Coulomb force gradually comes into play, which acts to compress the droplet
along the two poles. As a result, the droplet returns to the spherical shape, and then
evolves continuously and eventually into the final oblate shape. It is important to stress
that the transition will not occur if Eq. (5) were used to model the free charge distribution.
This is because by Eq. (5) the free charge distribution is assumed to be established
instantaneously and no charge relaxation will occur.
3.4 Electro-Coalescence of Multi Droplets
The phenomenon of a dielectric droplet or particle moving in a non-uniform electric
field is referred to as dielectrophoresis (Pohl, 1978). By the theory of dielectrophoresis,
16
two dielectric droplets placed in the direction parallel to the electric field would attract
each other and coalesce. Two leaky dielectric droplets, however, can respond to an
applied electric field in different ways. They may be pushed apart or pulled together
depending upon the detailed flow and electric field conditions (Baygents et al., 1998).
The electro-coalescence of two droplets was simulated numerically by the phase field
model (Lin et al., 2012), assuming a 2D axisymmetry. In this study, the 3D phase field
simulation was conducted of the coalescence of four identical liquid droplets of radius R
immersed in a host fluid of different material properties under the influence of electric
field. The computational domain was (l×l×l/2) with l = 6R and initially the four droplets
were placed in the plane of z=l/4, as shown in Fig. 5. Two adjacent droplets were
separated by 0.5R in the x direction and R in the y direction. Because of symmetry, only
1/4 of the domain or (l/2×l×l/4) was used and discretized into a mesh of 101×201×51.
For the case shown in Fig. 5, λζ =0.25 and λε=2 were chosen, with the electrical
conductivities of the two phases being 5*10-9
S/m and 2*10-8
S/m, respectively. Other
parameters, and the boundary and initial conditions are the same as given in the previous
section.
The electro-coalescence involves the following three different interactions (Baygents et
al., 1998). First, the global electrical force on droplet exists due to the dielectrophoresis
effect, even though the droplets are free of native charge. This force causes the droplet
mass center to move in translation. Second, the flow circulation outsides the droplet
accompanying the internal circulation, which is produced by the distributed tangential
electric stress, will produce a hydrodynamic force. It either enhances or impedes the
coalescence process depends on the flow circulation pattern. Third, the local electric field
17
induces the droplet deformation and thus changes the distance of droplets.
Fig. 5 shows Computed results for the coalescence of four liquid droplets in an electric
field applied in the y (or vertical) direction. The droplets first deform into an oblate shape
(t*=54) by the leaky dielectric effect (Eq. (14)). The deformation elongates the droplets in
the x (or horizontal direction) and causes the collision of the two droplets in the same row
(t*=84). As a result, the four droplets are combined into two bigger ones (t
*=90), each of
which continues to deform (t*=120). These resulting two large droplets then gradually
move together (t*=348) and eventually coalesce into a single large droplet (t
*=354, 360,
420).
The above dynamically evolving events in essence are a manifestation of different
mechanisms driving the electrically-induced coalescence in the parallel and normal
directions of the applied electric field. Let us first consider the coalescence in the normal
direction. By symmetry, the like electric charges accumulate at the two droplets in the
same row. As the like electric charges repel each other, the two droplets are pushed away
from each other by the global electric force. The circulatory flows outside a droplet move
away from the droplet, which produces an action to push apart the two droplets. The only
action that favors the coalescence is the droplet deformation or elongation in the
horizontal direction, which reduces the separation between the droplets in the same row
and eventually causes them to collide with each other.
For the electro-coalescence in the direction parallel to the electric field, however, the
situation is different. The global electric force favors translation of the drops towards
each other, because the opposite charges induced on the nearest surfaces of two droplets
produce an attraction between the two. Since λζ < λε, the circulatory flow outside the
18
droplet, which is caused by local tangential electric stress, moves towards the droplet and
thus enhances the coalescence. One the other hand, by Eq. (14), the droplets deform into
an oblate ellipsoid and this oblate deformation takes place along the horizontal direction,
which increases the clearance between the two droplets aligned in the vertical direction
(or parallel to the electric field) and thus further separates the droplets apart, an action
unfavorable for coalescence. The droplets eventually coalesce under a combined action
of the global electric force and the circulatory flow, which overpowers the unfavorable
effect caused by deformation.
Fig. 6 shows the 3D flows at the dimensionless times t*= 84 and t
*=348 at which the
coalescence is taking place in the normal and parallel directions, respectively. For a
single droplet at equilibrium, the circulatory flows are present inside and outside the
droplet. Similar flow structures are also observed in Fig. 4c, and also in 2D calculations
(Hua et al., 2008; Lin et al., 2012; Taylor, 1966; Tomar et al., 2007). This is caused by
the tangent electric stress (Saville, 1997) and occurs with the droplets placed far from
each other. The corresponding flow structures also evolve, as the droplets are pulled
together. As shown in Fig. 6a and its accompanying 3D views, when two droplets are in
contact (t*= 84), the circulatory flows inside the droplets are diminished. The flow
streamline illustrates that the two droplets move towards each other and the coalescence
is manifested by the exchange of the fluid between the two droplets. Fig. 6b shows the
evolution of the 3D flow structures both inside and outside the droplets just before these
two larger droplets coalesce together. It is important to note that in order to observe the
coalescence in the direction normal to the applied field, the droplets need to be placed
close enough to each other so that deformation-caused collision occurs to affect the
19
coalescence. On the other hand, as it was demonstrated by simulation, if the droplets are
placed horizontally (or normal to the applied field) at a distance of R, they will drift apart
from each other from their initial positions.
3.5 Electrohydrodynamic Patterning
The EHD Patterning is a novel technique to produce the controllable polymer
micro/nano-patterns (Chou et al., 1999; Chou and Zhuang, 1999; Schaffer et al., 2000;
Wu and Russel, 2009; Yang et al., 2013). In this process, the polymer/air or
polymer/polymer film is sandwiched between a pair of electrodes as shown in Fig. 7a,
with the upper-electrode (i.e. template) being patterned (e.g. Fig. 7b). The polymer/air
interface deforms according to a modulated electrical force resulting from the patterned
structure, thereby producing a micro- or nano-structure that conforms well with the
template. Usually the scale of EHD patterning is on the order of hundreds of nanometers
or larger. At this scale, the continuum assumption is still valid (Koplik and Banavar,
1995). The 3D phase field model, presented above, can be used to simulate the EHD
patterning process (Kim and Lu, 2006a; Kim and Lu, 2006b; Tian et al., 2011). The
characteristic length is chosen as the distance d0 of the two electrodes (see Fig. 7a). The
properties of the air and polymer differ dramatically, and for modeling the air/polymer
system, λρ=1000, λε=3, λµ=1000 and λζ=104 are chosen. The conductivity for polymer is
taken as 10-5
S/m. The template is patterned with a periodic array of cubic protrusions, as
shown in Fig. 7b. The patterned feature has a periodic length l0 of 200nm, with
d0=100nm, Δd=50nm, and the cross section of the protrusion being 100nm*100nm.
Taking the advantage of the periodicity, the 3D simulations were conducted within a
computational domain of l0×(d0+Δd)×l0, which is the size of one extruded cubic structure.
20
The periodic conditions are applied in the horizontal directions. The equi-potentials and
the no slip flow conditions are set at the electrodes, with the bottom electrode acting as
the ground electrode. The simulation used a mesh of 101*76*101.
Fig. 8 illustrates the transient development of structures as they occurs in the EHD
patterning process. The initial polymer/air interface is flat and the interface starts to
deform under the action of the electric force between the polymer surface and the
template, which is generated by the electric voltage applied between the two electrodes.
This is shown in Fig. 8b. The onset of the perturbation in the interface continues to grow
under the action of the electric force, as illustrated in Figs. 8c and 8d, forming the
structure in conformation with the pattern on the template. The structure is eventually
pulled by the electric force until the top of the polymer structure touches the template and
then flattened out by the template (see Figs. 8e and 8f). In Figs 8d and e, four ridges on
each patterned structure can be observed, each of which corresponds to one corner of the
cubic extrusion on the template. Fig. 8f is the top-down view of Fig. 8e, which shows the
EHD patterned structure conforms well with the template. Simulations also show that the
growing rate of the structure height accelerates with time, which is attributed to the fact
that the electric force becomes stronger when the polymer/air interface is pulled closer to
the template.
Detailed flow pattern during the process is depicted in Fig. 9, where the electric force
pulls up the polymer layer towards the upper electrode. As the polymer fluid rises up in
the middle, the fluid far away from the extruded electrode is depressed because of
requirement for mass conservation. As is evident by the flow steam lines, a large rotating
eddy entrains the liquid to the center region. Fig. 9b shows the corresponding 3D flow
21
structure, which indicates that a full 3D model is necessary in faithfully representing the
EHD patterning process for the micro- or nano-structure formation.
3.6 Parallel Computing
The 3D phase field model for the EHD flow problem described above can be
computationally very intensive. Parallel computing is applied to improve the computing
efficiency. Parallel computing performance of the code was evaluated with the EHD
patterning problem as the testing case. The in-house Fortran code for the 3D phase field
modeling was parallelized with the openMP paradigm. The test used the same mesh of
101*76*101 and other parameters are also kept the same as in the previous section.
Computations were done on a AMD Opteron Linux platform, which has 48
microprocessors with 64GB Ram capable of high performance parallel computing. The
same platform was also used to obtain the computed results presented above.
The parallel computing performance for the 3D phase field model calculations is shown
in Fig. 10. Clearly, the more numbers of CPUs are used, the less time is required for
computation. However, it appears that the most drastic time reduction occurs with the
increase up to 10 CPUs, which give an improvement of 6 times in computational
efficiency. A further increase in number of the CPU processors becomes less effective,
largely because the CPU time is increasingly consumed by the inter-communication and
data-exchange between different processors. By using 10 CPUs, one EHD patterning case
would take several days to finish.
22
4. Conclusions
This paper has presented a phase field computational methodology for the numerical
solution of 3D electrohydrodynamic two phase flow problems. The mathematical
formulation consists of a set of coupled nonlinear equations, including the Navier-Stokes
equations for fluid motion, the convective Cahn-Hilliard equation for the free surface and
phase field, and the Poisson equation for the electric field distribution in pure and/or
leaky dielectric fluids. These equations are discretized by the finite difference method
along with the explicit time marching scheme, enhanced by a parallel computing
algorithm. For the problems of two fluids with a large disparity in fluid densities, an
appropriate interpolation of densities across the free surface is essential to ensure the
numerical stability. Numerical efficiency can be drastically improved via parallel
computing, but the most gains are obtained with the use of up to 10 CPUs, when the code
is parallelized with openMP.
The 3D model is validated by comparing with the Taylor model for a droplet
deformation in an electric field. It is found that while a leaky dielectric droplet deforms
into an oblate ellipsoid as its final equilibrium shape in an electric field, the transient
development in deformation entails the transition from the initial spherical to the prolate
and then to the intermediate spherical and eventually the final oblate shape. This is
different from the deformation behavior of a pure dielectric droplet, the difference being
attributed to the relaxation of free charges in the interface region. The simulation for the
coalescence of four droplets in an electric field reveals a truly 3D deformation and fluid
flow field. Different mechanisms of electro-coalescence in the directions parallel and
perpendicular to the electric field are discussed. The usefulness of the 3D phase field
23
model is further illustrated by its application in modeling the electrically-induced two
phase flow phenomena associated with an electrohydrodynamic patterning process for
micro- or nano-structure formation.
Acknowledgments
This work is supported by NSFC (Grant no. 90923040) and the National Basic Research
Program of China (Grant no. 2009CB724202). The first author also acknowledges the
partial support of the University of Michigan at Dearborn as a visiting Ph.D. student.
Helpful discussion with Professor H. A. Stone of Princeton University on the subject of
electro-coalescence is also gratefully acknowledged.
References
Adalsteinsson, D., Sethian, J., 1995. A fast level set method for propagating interfaces. J. Comput. Phys.
118, 269–277.
Allan, R. S., Mason, S. G., 1962. Particle behavior in shear and electric fields. I. Deformation and burst of
fluid drops. Proc. R. Soc. A 267, 45–61.
Anderson, D. M., McFadden, G. B., Wheeler, A. A., 1998. Diffuse-interface methods in fluid mechanics.
Ann. Rev. Fluid Mech. 30, 139–165.
Badalassi, V.E., Ceniceros, H.D., Banerjee, S., 2003. Computation of multiphase systems with phase field
models. J. Comput. Phys. 190, 371–397.
Baygents, J. C., Rivette, N. J., Stone, H. A., 1998. Electrohydrodynamic deformation and interaction of
drop pairs. J. Fluid Mech. 368, 359–375.
Cahn, J. W., Hilliard, J. E., 1958. Free energy of a nonuniform system. I. Interfacial Free Energy. J. Chem.
Phys. 28, 258–267.
Carlson, A., Do-Quang, M., Amberg, G., 2010. Droplet dynamics in a bifurcating channel. Int. J.
Multiphase Flow 36, 397–405.
Chou, S. Y., Zhuang, L., Guo, L., 1999. Lithographically induced self construction of polymer
microstructures for resistless patterning. Appl. Phys. Lett. 75, 1004–1006.
Chou, Y., Zhuang, L., 1999. Lithographically induced self-assembly of periodic polymer micropillar arrays.
J. Vac. Sci. Technol. B 17, 3197–3202.
24
Ding, H., Spelt, P. D. M., Shu, C., 2007. Diffuse interface model for incompressible two-phase flows with
large density ratios. J. Comput. Phys. 226, 2078–2095.
Donaldson, A. A., Kirpalani, D. M., Macchi, A., 2011. Diffuse interface tracking of immiscible fluids:
Improving phase continuity through free energy density selection. Int. J. Multiphase Flow 37, 777–
787.
Feng, J. Q., Scott, T. C., 1996. A computational analysis of electrohydrodynamics of a leaky dielectric
drop in an electric field. J Fluid Mech. 311, 289–326.
Fermin, R. J., 2001. Electrohydrodynamic coating flows. Ph.D. thesis, University of Minnesota.
Gilbert, W., de Magnete (Dover, New York, 1958), Vol. II, Chap. II, p. 89, first published ca. 1600,
translated by P. F. Mottelay in 1893.
Hua, J., Lim, L. K., Wang, C.-H., 2008. Numerical simulation of deformation/motion of a droplet
suspended in viscous liquids under influence of steady electric field. Phys. Fluid 20, 11302–11317.
Hirt, C. W., Nichols, B.D., 1981. Volume of fluid (VOF) method for the dynamics of free boundaries. J.
Comput. Phys. 39, 201–225.
Huo Y., Li, B. Q., 2006, Surface deformation and convection in electrostatically-positioned droplets of
immiscible liquids under microgravity, J. Heat Trans-T ASME ASME 128, 520-529.
Jacqmin, D., 1999. Calculation of two-phase Navier-Stokes flows using phase-field modeling. J. Comput.
Phys. 155, 96–127.
Johnston, H., Liu, J.-G., 2002. Finite Difference Schemes for Incompressible Flow Based on Local
Pressure Boundary Conditions. J. Comput. Phys. 180, 120–154.
Khatavkar, V. V., Anderson, P. D., Duineveld P. C., Meijer, H. E. H., 2007. Diffuse-interface modelling of
droplet impact. J. Fluid Mech. 581, 97–127.
Kim, D., Lu, W., 2006. Interface instability and nanostructure patterning. Comput. Mater. Sci. 38, 418–425.
Kim D., Lu, W., 2006. Three-dimensional Model of Electrostatically Induced Pattern Formation in Thin
Polymer Films. Phys. Rev. B 73, 035206.
Kim, J., 2012. Phase-Field Models for Multi-Component Fluid Flows. Commun. Comput. Phys. 12, 613–
661.
Kim, J. S., 2005. A continuous surface tension force formulation for diffuse-interface models. J. Comput.
Phys., 204, 784–804.
Koplik, J., Banavar J. R., 1995. Continuum Deductions from Molecular Hydrodynamics. Annu. Rev. Fluid
Mech. 27, 257-92
Lafaurie, B., Nardone, C., Scardovelli, R., Zaleski, S., Zanetti, G., 1994. Modelling merging and
fragmentation in multiphase flows with SURFER. J. Comput. Phys. 113, 134–147.
Li, B. Q., 2006, Discontinuous Finite Elements in Fluid Dynamics and Heat Transfer, Springer Press, New
York, NY.
Lifshitz, I. M., Slyozov, V. V., 1961. The kinetics of precipitation from supersaturated solid solutions. J.
Phys. Chem. Solids 19, 35–50.
25
Lin, Y., Skjetne, P., Carlson, A., 2012. A phase field model for multiphase electro-hydrodynamic flow. Int.
J. Multiphase Flow 45, 1–11.
Liu C., Shen, J., 2003. A phase field model for the mixture of two incompressible fluids and its
approximation by a Fourier-spectral method. Phys. D, 179, 211–228.
Ni, M.-J., 2009. Consistent projection methods for variable density incompressible Navier–Stokes
equations with continuous surface forces on a rectangular collocated mesh. J. Comput. Phys. 228,
6938–6956.
Nichols, B. D., Hirt, C. W., Hotchkiss, R.S., 1980. SOLA-VOF: A solution algorithm for transient fluid
flow with multiple free boundaries, Los Alamos National Lab Report LA-8355.
Pohl, H. A., 1978. Dielectrophoresis. Cambridge University Press, UK
Puckett, E. G., Almgren, A. S., Bell, J. B., Marcus, D. L., Rider, W. J., 1997. A high-order projection
method for tracking fluid interfaces in variable density incompressible flows. J. Comput. Phys. 130,
269–282.
Patankar, S. V., 1980. Numerical heat transfer and fluid flow. Taylor & Francis,
Rayleigh, L., 1882. On the equilibrium of liquid conducting masses charged with electricity. Phil. Mag. Ser.
14,184–186.
Saville, D. A., 1997. Electrohydrodynamics: The Taylor-Melcher Leaky Dielectric Model. Annu. Rev.
Fluid Mech. 29, 27–64.
Schaffer, E., Thurn-Albrecht, T., Russell, T. P., Steiner, U., 2000. Electrically induced structure formation
and pattern transfer. Nature (London) 403, 874–877.
Sherwood, J. D., 1988. Breakup of fluid droplets in electric and magnetic fields. J. Fluid Mech. 188, 133–
146.
Stone, H. A., Stroock, A. D., Ajdari, A., 2004. Engineering flows in small devices: Microfluidics toward a
lab-on-a-chip. Annu. Rev. Fluid Mech. 36, 381–411.
Sussman, M., Smereka, P., Osher, S., 1994. A level set approach for computing solutions to incompressible
two-phase flow. J. Comput. Phys. 114, 146–159.
Taylor, G., 1966. Studies in Electrohydrodynamics. I. The Circulation Produced in a Drop by Electrical
Field. Proc Roy. Soc. Lond. Ser. A, Math. Phys. Sci. 291, 159-166.
Tian, H., Shao, J., Ding, Y., Li X, Li, X., 2011. Numerical studies of electrically induced pattern formation
by coupling liquid dielectrophoresis and two-phase flow. Electrophoresis 32, 2245–2252.
Tomar, G., Gerlach, D., Biswas, G., Alleborn, N., Sharma, A., Durst, F., Welch, S.W.J., Delgado, A., 2007.
Two-phase electrohydrodynamic simulations using a volume-of-fluid approach. J. Comput. Phys. 227,
1267–1285.
Torza, S., Cox R. G., Mason S. G., 1971. Electrohydrodynamic Deformation and Burst of Liquid Drops.
Phil. Trans. R. Soc. Lond. A 269. 295-319.
26
Tryggvason, G., Bunner, B., Esmaeeli, A., Juric, D., Al-Rawahi, N., Tauber, W., Han, J., Nas, S., Jan, Y.-j.,
2001. A front-tracking method for the computations of multiphase flow. J. Comput. Phys. 169, 708–
759.
Van der Waals, J. D., 1893. Verh. K. Akad. Wet. Amsterdam Section vol. 1 No. 8 (Dutch), Translation of
Van der Waals, J. D. (The thermodynamic theory of capillarity under the hypothesis of a continuous
density variation), 1979. J. Stat. Phys. 20, 200-244.
Villanueva, W., Amberg, G., 2006. Some generic capillary-driven flows. Int. J. Multiphase Flow 32, 1072–
1086.
Wu N., Russel, W. B., 2009. Micro- and nano- patterns created via electrohydrodynamic instabilities. Nano
Today 4, 180–192.
Yang, Q., Li, B. Q., Ding, Y., 2013. A numerical study of nanoscale electrohydrodynamic patterning in a
liquid film. Soft Matter 9, 3412-3423.
Yue, P., Zhou, C., Feng, J. J., 2007. Spontaneous shrinkage of drops and mass conservation in phase-field
simulations. J. Comp. Phys. 223, 1–9.
Zeng, J., Korsmeyer, T., 2004. Principles of droplet electrohydrodynamics for lab-on-a-chip. Lab Chip 4,
265–277.
Zikanov, O., 2010, Essential Computational Fluid Dynamics. John Wiley & Sons.
27
Table 1. The mesh sensitivity of leaky droplet deformation*
Mesh Deformation
(D) Error
51×101×51 0.062658 12.7%
61×121×61 0.070084 2.36%
81×161×81 0.070191 2.21%
91×181×91 0.071029 1.04%
101×201×101 0.071779
* The case 101×201×101 is used as a base for comparison
Fig. 1. Schematic illustration of an EHD two phase flow problem: (a) a droplet in an
electric field generated by two electrodes and (b) the computational domain.
28
Fig. 2. Spurious currents at time t*=10
-3 as calculated by different approximation
schemes. The parameters used for calculations (1–droplet and 2–host fluid): λρ=1000,
λµ=1000, Ca=0.18. All these figures are shown at the cross section z=0. It should be
mentioned that the velocity is characterized by Uc=µ1/ρ1R for this case.
Fig. 3. Comparison of the deformation D between the analytical solutions and the 3D
phase field numerical results for BoE=0.2, λε=ε1/ ε2=2 and λµ=µ1/ µ2=1.
29
Fig. 4. Transient development of droplet deformation, where the cross section at z=0 is
shown rather the 3D shape along with the fluid flow field and phase field. The initially
spherical shape is spherical (not shown). The deformation undergoes through the prolate
first (a), then the spherical (b), and then eventually the oblate shape (c). The
dimensionless time for (a-c) are t*=1.8, 8.46 and 360, respectively. The corresponding
deformation parameters are: D=0.14, D≈0 and D=−0.26.
Fig. 5. Snap shots of the coalescence process of four droplets in an electric field (applied
in the vertical direction) at times: t*= 0, 54, 84, 90, 120, 348, 354, 360, and 420. The
parameters used for computations: Re=4, Ca=0.8, Pe=7200, Cn=0.025, BoE=0.8, λρ=1,
λε=2, λµ=1 and λζ=0.25.
30
Fig. 6. 3D view of the flow field distribution as the four droplets undergo electro-
coalescence and at two different times: t*= 84 (a) and t
*=348 (b).
31
Fig. 7. (a) Configuration for the EHD patterning and (b) the shape of the template.
Fig. 8. Numerical simulation of the EHD nanopatterning process – evolution of the
polymer/air interface as it deforms by an applied electric field. The corresponding
dimensionless times for (a) to (e) are: t*=0, 8, 12, 16, and 20.
32
Fig. 9. The flow field and the steamline for the EHD nanopatterning at t*=16, as
computed from the 3D phase field model: (a) the cut-pane view at the middle plane, and
(b) the 3D flow structure.
Fig. 10. Parallel computing performance of the 3D phase field model for EHD two
phase flow problems: computing time per time step vs. number of CPUs. Inset shows
the log-log plot.
Highlights
We present a 3D phase field model for two phase electrohydrodynamic flows.
The transient behavior of a leaky droplet deformation under electric field is investigated.
We study the four droplets electro-coalescence
We compute electrically-induced two phase flow associated with nanostructure patterning.
*Highlights (for review)