Engineering Applications of Computational Fluid

Mechanics

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tcfm20

Finite element simulation for multiphase fluids

with different densities using an energy-law-
preserving method

Yanhai Lin & Yongyue Jiang

To cite this article: Yanhai Lin & Yongyue Jiang (2020) Finite element simulation for multiphase
fluids with different densities using an energy-law-preserving method, Engineering Applications of
Computational Fluid Mechanics, 14:1, 642-654, DOI: 10.1080/19942060.2020.1756413

To link to this article: https://doi.org/10.1080/19942060.2020.1756413

© 2020 The Author(s). Published by Informa Published online: 05 May 2020.

UK Limited, trading as Taylor & Francis
Group

Submit your article to this journal Article views: 765

View related articles View Crossmark data

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tcfm20
ENGINEERING APPLICATIONS OF COMPUTATIONAL FLUID MECHANICS
2020, VOL. 14, NO. 1, 642–654
https://doi.org/10.1080/19942060.2020.1756413

Finite element simulation for multiphase fluids with different densities using an
energy-law-preserving method
Yanhai Lina and Yongyue Jiangb
a Fujian Province University Key Laboratory of Computation Science and School of Mathematical Sciences, Huaqiao University, Quanzhou,

People’s Republic of China; b School of Mathematics and Physics, University of Science and Technology Beijing, Beijing, People’s Republic of
China

ABSTRACT ARTICLE HISTORY

This paper reports the results of a numerical study on the dynamics of isothermal multiphase flu- Received 8 November 2019
ids using an energy-law-preserving method. A phase-field model is taken into account. Different Accepted 4 April 2020
to previous studies, a continuous finite element technique is used to simulate the Navier-Stokes- KEYWORDS
Cahn-Hilliard coupled model. A modified discrete energy law of the numerical simulation is derived Multiphase fluids;
in detail. A penalty formulation is applied for continuous conditions to ensure the stability of the energy-law-preserving
pressure. The coalescence of two kissing bubbles and the rising of a lighter drop are simulated as method; sharp interface
numerical examples, and the estimated orders of the velocity gradient are computed to examine limit; phase-field model;
the accuracy of the numerical solution. The paper shows that in the computing process the energy continuous finite element
law is preserved for each time step and that the errors in the discrete energy law equal less than 10−8 . technique; numerical
simulation

1. Introduction
The study of multiphase fluids is very important in
biomedicine, the production of chemicals, mathemati-
cal physics and engineering practice: for example, in the
preparation of nanofluid within cooling systems; the flow
of human blood within targeted drug therapy processes;
and the evaporation of liquid within vapor–liquid sep-
arators (Jiang et al., 2015; Jiang & Lin, 2019; Jiang &
Yang, 2018; Lin & Jiang, 2018). To satisfy the needs of
rapidly developing technologies, increasing attention is
being paid to the mechanism of multiphase fluid flows
and heat transfers. Numerical simulations of multiphase
fluids using mathematical modeling, numerical calcula-
tions, and subsequent analysis allow the flows, interface
changes, and heat transfer to be analyzed intuitively and
easily. The difference between numerical solutions and
practical physics should be the main theme of numer-
ical simulations because of exactly how the mathemat-
ical model has been developed and how the numerical
method reflects this development.
Both phase-field and sharp-interface models are
extensively applied in the study of multiphase fluids with
a clear contact surface. The sharp-interface model (Meir-
manov, 1992; Visintin, 1996) was developed earlier than
the phase-field model, and its viewpoint is closer to that
of practical physics. In the simulation processes of the
sharp-interface models, all phases are simulated by a
single-phase method and contact surfaces are treated as a
jump-boundary condition. Because of the different prop-
erties of the different phases, the boundary is complex
(e.g. jump-boundary conditions and discontinuous solu-
tions). Additionally, the sharp-interface model can entail
a difficult computational process. Despite the jump-
boundary conditions, many researchers still study mul-
tiphase fluids using the sharp-interface model because
its basis is more realistic. Zhang et al. (2014) developed
a finite difference model to describe the viscous incom-
pressible Navier-Stokes (N-S) equations, and used a two-
step projection algorithm to discretize the N-S equations
on a fixed Cartesian grid. Han et al. (2014) presented
an adaptive multi-resolution technology to investigate
compressible multi-phase fluids by applying the sharp-
interface model and a pyramid data structure. Using the
finite difference method, Zhang and Ren (2016) investi-
gated moving contact lines dynamics within a two-phase
fluids system with a polymer additive effect under the
consideration of polymer stress modeled by the FENE-P.
Unlike the sharp-interface model, the phase-field
model (or diffuse-interface model) introduces the hypoth-
esis of continuity across a contact surface, which should
therefore not be considered as a jump-boundary. It intro-
duces a thickness in the interface of different phases, and

CONTACT Yanhai Lin linyanhai999@hqu.edu.cn

© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.
 ENGINEERING APPLICATIONS OF COMPUTATIONAL FLUID MECHANICS
physical properties (e.g. density, viscosity, and thermal
conductivity) change continuously within the region of
contact surface. The phase-field model treats the mul-
tiphase fluids as a single fluid with different material
properties, avoiding the appearance of jump-boundary
conditions. A new order parameter is used for iden-
tifying different phases, and this is mostly uniform in
the bulk energy. Qian et al. (2006) proposed a modified
slip-boundary condition based on molecular dynamics
theory. They relaxed the no-slip conditions by introduc-
ing a slip length that allows the boundary conditions to
slip near the contact line, but retain the no-slip property
in other areas. Through scaling analysis and numeri-
cal computations, Yue et al. (2010) proposed a criterion
for attaining the sharp-interface limit within the calcu-
lation of moving contact lines. It indicated that no-slip
boundary conditions are not suitable for calculations of
moving contact lines. By using the level-set method and
asymptotic theories, Sui and Spelt (2013) proposed a
test model to simulate the moving contact lines. These
results exhibit a good convergent property, even with a
coarse mesh. If the diffuse-interface model is reasonable,
a sharp-interface limit should be achieved or the thick-
ness of the interface should be at the nanometer scale,
which is a challenge for numerical methods. Liu and Ding
(2015) proposed a generalized diffuse-interface model to
simulate multi-phase flows over a curved substrate by
using an immersed boundary technique. Luo et al. (2016)
proposed a new curvature-boundary condition to solve
the contact-angle problem of multiphase flows within
moving contact lines.
In the last two decades, many simulation processes
from phase-field models have achieved better accuracy
than those from sharp-interface models: for example,
the drop coalescence, the bubble rising, the flow defor-
mation, and so on. The order parameters have been
improved through modifications with different mixtures
of energy functions, and increasingly efficient numerical
techniques have been applied to solve problems within
a phase-field model. For example, Spelt (2006) investi-
gated the influences of droplet size and contact-angle
on the shear flow past a two-dimensional (2D) droplet
by using the finite difference technique. By using the
same numerical technique, Li et al. (2018) proposed a
phase-field model to study the thermocapillary effect on
the melting flow of metallic powders. The alternating
direction implicit scheme was also used to solve non-
linear terms issued in the phase-field equations. Zhang
and Tang (2007) developed a phase-field model to sim-
ulate the mixing process of incompressible fluids using
the adaptive moving mesh technique. They noted that the
adaptive technique is used to solve the three-dimensional
(3D) case conveniently. Furthermore, Tsai et al. (2010)
643
developed an adaptive 3D phase-field model for the sim-
ulation of alloy directional solidification. They also found
that 3D simulations can be performed efficiently using
the adaptive moving mesh technique. Chen and Shen
(2013) solved anisotropic Cahn–Hilliard systems using
a spectral-Galerkin method. Zhao et al. (2016) explored
a phase-field model to simulate metal-form processes
within a square cavity using an implicitly finite volume
technique on coupled uniform grids. Wang et al. (2017)
explored two- and three- dimensional phase-field mod-
eling to simulate crystal growth process, and Aihara
et al. (2019) presented a novel multi-phase field model
for expressing multiphase flows that consist of an arbi-
trary number of phases using a conservative Allen-Cahn
equation. Zhao et al. (2019) numerically simulated the
energy-scaling characteristics of a time fractional phase-
field model for the power-law scaling dynamics within
coarsening processes. Zhang et al. (2019) investigated
the influence of a forced flow on the permeability of the
mushy zone during solidification using the phase-field
lattice Boltzmann technique coupled with an adaptive
mesh scheme, whereas Ren et al. (2019) presented vec-
tor valued phase-field models to simulate polycrystalline
coarsening solidification processes using an operator
splitting technique. In addition, Akbarian et al. (2018),
Mosavi et al. (2019), and Salih et al. (2019) presented
more real case studies by using the applications of CFD
models.
Recently, the finite element technique was extensively
applied to solve phase-field models of simulations. By
using the numerical technique, Guo et al. (2014) pre-
sented a phase-field diffuse-interface model of the N-S
equation with a liquid–liquid interface between fluids
with different densities. Furthermore, Jiang et al. (2015)
also computed a phase-field model with the moving con-
tact lines. Murmatsu et al. (2018) presented a phase-field
model to study the formation process of ferroelastic crys-
tal grains, and used the finite element technique to solve
phase transformations and strain distributions. By com-
bining with the user defined element method, Zhang et al.
(2018) presented a modified phase-field model to simu-
late cohesive fracture processes. Furthermore, Aldakheel
et al. (2018) explored an extended phase-field model to
simulate the brittle fracture processes with a low-order
virtual element technique. Based on an adaptive finite
element scheme coupled with a non-uniform grid, Zhu
et al. (2019) investigated a multi-grain growth phase-field
model of directional alloy solidifications and seaweed-
like micro-structure evolution.
Motivated by the above-mentioned studies (particu-
larly Guo et al. [2014] and Jiang et al. [2015]), this article
describes an energy-law-preserving technique coupled
with a continuous finite element scheme for solving the
644 Y. LIN AND Y. JIANG

multiphase flow of fluids with different densities, specif- and then we have
ically the model proposed by Abels et al. (2012). The
continuity condition is treated by a penalty method, as ∂ρi φi
+ v · (ρi φ) + ∇ · Pi = 0, (6)
in previous studies (Jiang et al. 2015), and the discrete ∂t
energy law is derived by an extended temporal form ∂φ
+ v · ∇φ + ∇ · P = 0, (7)
based on the work in Guo et al. (2014). In Section 2, ∂t
we briefly introduce some base equations of a two-phase ∂ρ ∂(ρ1 + ρ2 )
model and concretely give and derive a weak form of the = = −∇ · (ρ1 v + P1 + ρ2 v + P2 )
∂t ∂t
problem. In Section 3, we derive the fully discrete scheme = −∇ · (ρv + P1 + P2 ), (8)
and the discrete energy law with an extended temporal
scheme. The coalescence of two droplets and the case of a
rising bubble are considered as examples in Section 4, and where Pi = ρi vi − ρi v and P = Pρ22 − Pρ11 .
the sharp interface limit is proved numerically. Finally, The model for multiphase fluids with various densities
some important results and outlooks are summarized. proposed by Abels et al. (2012) can be written as:

∇ · v = 0, (9)
2. Two-phase model, weak form, and energy
law ∂(ρv)
+ ∇ · ((ρv + J) ⊗ v) = −∇p + ∇ · (2μD(v))
∂t
The mass conservation equation in local form is given by
+ ρg + λ∇φ, (10)
∂ρi
+ ∇ · (ρi vi ) = 0, (1) ∂φ
∂t + v · ∇φ = ∇ · (M∇λ), (11)
∂t
where ρi , vi , i = 1, 2 represent the fluid density and veloc-
λ = σ ε −1 ψ  (φ) = −σ ε φ, (12)
ity offluid i. The total mass density ρ is obtained by
ρ = 2i=1 ρi and the mass concentration hi is denoted
 and the boundary conditions as v = 0, ∂n λ = 0, and
as hi = ρρi with 2i=1 hi = 1. It is assumed that the vol-
∂n φ = 0. Here, v = (vx , vy ) is velocity vector and vx , vy
ume of each fluid after mixing remains constant, in other
are velocity components in x,y directions. In particular, φ
words, the excess volume caused by mixing is 0. We use ρi
denotes the order parameter for defining different phases
to represent the constant specific density of fluid
 i before (φ = 1: fluid A; φ = −1: fluid B). p denotes the hydro-
mixing, and introduce φi = ρρii . We have 2i=1 φi = 1
static pressure. D(v) = ∇v+∇v
T
from the assumption that the excess mixing volume is 2 is the viscous part of the
zero, and this expression implies that ρ = ρ1 φ1 + ρ2 φ2 . stress tensor. μ is the fluid dynamic viscosity. λ is the
Using the mass concentration difference φ = φ2 − φ1 , chemical potential. σ is the physical surface tensor. M is
the total density of this mixture is defined as phenomenological mobility coefficient. ε is the interface
thickness. The total mass density of the mixture, ρ, can
1−φ 1+φ be expressed as ρ = ρ2 +ρ + ρ2 −ρ ρ2 −ρ1
2 φ. J = − 2 M∇λ,
1 1
ρ = ρ1 φ1 + ρ2 φ2 = ρ1 + ρ2 . (2) 2
2 2 ψ(φ) = 14 (φ 2 − 1)2 denote the mass flux due to the var-
In Lowengrub and Truskinovsky (1998), the mass ious densities (Lowengrub & Truskinovsky, 1998)
  and
averaged/barycentric velocity v defined as ρv = ρ1 v1 + double-well potential, respectively. g = 0 −9.8 rep-
ρ2 v2 . Unlike that, the volume-averaged mixture velocity resents acceleration due to gravity.
(Boyer, 2002) is adopted here: We rewrite the chemical potential λ as λ = ω +
ρ1 ρ2 cφ, where the constant c enhances the stability of the
v = φ1 v1 + φ2 v2 = v1 + v2 . (3)
ρ1 ρ2 computations in the phase field model (Chen & Shen,
According to this expression, we have that 2013). is the bounded domain of the system in R2 .
      is the boundary of the computational domain. We
ρ1 ρ2 ∂ ρ1 ρ2 denote W1,3 ( ) = (W1,3 ( ))2 , L2 ( ) = (L2 ( ))2 , and
∇ · v =∇ · v1 + ∇ · v2 = − +
ρ1 ρ2 ∂t ρ1 ρ2 L20 ( ) = p ∈ L2 ( ), pdx = 0 . Thus, the weak form
∂ as: Find v ∈ W1,3 ( ), p ∈ L20 ( ), φ, and ω ∈ W 1,3 ( )
= − (φ1 + φ2 ) = ∂t 1 = 0. (4) such that
∂t
The mass conservation equation can be rewritten in  
∂φ
local form with the velocity v as ϕ + (v · ∇φ) + M∇(ω + cφ) · ∇ϕ dx = 0,
∂t
∂ρi
+ ∇ · (ρi v) + ∇ · (ρi vi − ρi v) = 0, (5) ∀ϕ ∈ W 1,3 ( ) (13)
∂t
 ENGINEERING APPLICATIONS OF COMPUTATIONAL FLUID MECHANICS 645

the full energy law that decreases with time, because the
((ω + cφ)χ − σ ε−1 ψ  χ − σ ε∇φ · ∇χ)dx = 0,
term ρg · v refers to the work done by gravity (G). From
the numerical examples (presented in section 4), it will be
∀χ ∈ W 1,3 ( ) (14)
shown that the total energy (E + G) increases with time
in the numerical computational process.
(∇ · v)qdx = 0, ∀q ∈ W 1,3 ( ) (15)
3. Fully discrete system and discrete energy law
 ∂(ρv) 
∂t · u + ∇ · ((ρv + J) ⊗ v) · u + μ∇v : ∇u In this section, a fully discrete weak form of the prob-
−λ∇φ · u − ρg · u − p(∇ · u) lem is presented. A discrete energy law equality similar to
dx = 0, ∀u ∈ W1,3 ( ) (16) Equation (21) is also derived. We use the special numer-
ical scheme in Guo et al. (2014) to discretize the weak
Taking ϕ = ω + cφ, v = u, q = p, and χ = ∂φ ∂t in form of the problem in Equations (13)–(16).
Equations (13)–(16), the continuous weak form becomes First, we apply a penalty formulation to the compu-
 tation of pressure by rewriting Equation (10) as ∇ · v +
∂φ
(ω + cφ) + (v · ∇φ)(ω + cφ) δp = 0, where δ = 10−6 (Altmann & Heiland, 2015; Lin,
∂t 1997). Let

+M|∇(ω + cφ)|2 dx = 0, (17)
W = W1,3 ( ) × L20 ( ) × W 1,3 ( ) × W 1,3 ( )
 
∂φ −1 ∂ψ σε ∂ 2 and Wh = Uh × Ph × Hh × Hh ⊂ W is a finite-
(ω + cφ) − σε − |∇φ| dx = 0,
∂t ∂t 2 ∂t dimensional sub-space of W. It is obtained by finite ele-
(18) ment discretizations of . The time step Δt > 0 and
(vhn , pnh , φhn , ωhn ) ∈ Wh is an approximation of v(t n ) =
(p∇ · v)dx = 0, (19)
v(nΔt), p(t n ) = p(nΔt), φ(t n ) = φ(nΔt), and ω(t n ) =
ω(nΔt). (vhn+1 , pn+1 n+1 n+1
h , φh , ωh ) is the approximation

∂(ρv) at time t n+1 = (n + 1)Δt. Then, the discrete weak form
+ ∇ · ((ρv + J) ⊗ v) · v − p(∇ · v)
∂t of the system can then be written as
  
+μ∇v : ∇v − (ω + cφ)∇φ · v − ρg · v dx √ n+1 n+ 1
⎛  |v|2  ∂ρ |v|2 ⎞ ∇ · ρv h + δph 2 qdx = 0, (22)
∂
∂t ρ 2 + ∂t 2 + ∇ · (ρv)|v|2
⎜ ⎟
⎜ +ρv · ∇ |v|2 + J · ∇ |v|2
2 2
⎟
= ⎜ ⎟ dx ⎛  ⎞
⎝ +(∇ · J)|v|2 − p(∇ · v) + μ|∇v|2 ⎠ √ √ n+1
n+ 12 n+ 1 √ n+1 n+ 1
ρh ρv t · u + ρh 2 ρv h + Jh 2
−λ∇φ · v − ρg · v ⎜ ⎟
⎜ ⎟
     ⎜ √ n+1 n+ 12 ⎟
∂ |v|2 |v|2 |v|2 ⎜ · ρv h · u − ph (∇ · u) ⎟
= ρ + (∇ · J) + ∇ · ρv ⎜ ⎟
⎜   ⎟
∂t 2 2 2 ⎜ n+ 1 
√  n+ 1  √  n+ 1 ⎟
 ⎜ 1
+ 2 ∇ · ρh 2 n+1
ρv h + Jh 2
ρv h · ⎟
2
2 ⎜ ⎟
−p(∇ · v) + μ|∇v| − λ∇φ · v − ρg · v dx ⎜ ⎟
⎜ √ n+1 ⎟
    ⎜ n+ 12 ⎟
∂ |v|2 ⎜ u + μ∇ ρv h : ∇u − ρh g · u ⎟
= ρ − p(∇ · v) + μ|∇v|2 − λ∇φ · v − ρg · v ⎜   ⎟
∂t 2 ⎝ n+ 12 n+ 12 n+ 12 ⎠
  − ωh + cφh ∇φh · u
∂ |v|2 |v|2
× dx + ρv ·n+ J · n ds
∂t 2 2 × dx = 0, (23)
   
∂ |v|2
= ρ − p(∇ · v) + μ|∇v|2 − λ∇φ · v − ρg · v
∂t 2 
√ n+ 21 n+ 1
× dx = 0 (20) φtn+1 ϕ + ρv h · ∇φh 2 ϕ
where we have adopted the expression ∂t ρ + ∇ · (ρv +   
n+ 1 n+ 1
J) = 0. Furthermore, we obtain the continuous energy +M∇ ωh 2 + cφh 2 · ∇ϕ dx = 0, (24)
law by solving Equations (17)–(20):
 
dE n+ 1 n+ 1
= ρg · vdx − μ||∇v||2L2 − M||∇λ||2L2 (21) ωh 2 + cφh 2 χ − σ ε−1 g(ωhn+1 , φhn )χ
dt
 
where E = ρ2 ||v||2L2 + σ2ε ||∇φ||2L2 + σ ε−1 ψdx is the −σ ε∇φh
n+ 21
· ∇χ dx = 0, (25)
total free energy. The energy law in Equation (21) is not
646 Y. LIN AND Y. JIANG

where ⎛   ⎞
n+ 12 √ n+1 n+ 1 √
n+1 √ n+1
 ⎜ ρh ρv h + Jh 2 ·∇
h · ρvρv
h ⎟
 n ⎜   ⎟
ρhn+1 + ρh ⎝ 1 n+ 1 √ n+1 n+ 12 √ n+ 12 √ n+1 ⎠
√ n+ 21 + 2 ∇ · ρh 2 ρv h + Jh ρv h · ρv h
ρh = ,
2
ρ1 + ρ2 ρ2 − ρ1 n+1 dx = 0,
ρhn+1 = + φh ,
2 2
n+ 1 Jn+1 + Jnh
Jh 2 = h , 2
2 (|φhn+1 | + |φhn |2 − 2)(φhn+1 + φhn )φtn+1
 
√ n+1 ρhn+1 vhn+1 − ρhn vhn 2
= (|(φhn+1 ) − 12 |)t .
ρv t = ,
Δt
Here, E is the total free energy at some point in time.
n+ 1 ρ n+1 + ρhn n+ 12 ωn+1 + ωhn
ρh 2 = h , ωh = h , G is the cumulative work done by gravity up to that time.
2 2 The iterative scheme (s = 1, 2, . . . ) is applied at each
 
√ n+1 ρhn+1 vhn+1 − ρhn vhn time-point tn+1 , in other words, search for φ s , ωs , vs , and
ρv h =   , ps that satisfy the following formulation:
ρhn+1 − ρhn  
√ n+1 p + pnh
n+ 12 φhn+1 + φhn φhn+1 − φhn ∇· ρv h,s + δ s qdx = 0, (29)
φh = , φtn+1 = , 2
2 Δt
1 2 √ √ √ √
g(φhn+1 , φhn ) = (|φhn+1 | + |φhn |2 − 2) ⎛ ρ s + ρhn ρ s vs + ρhn vhn
⎞
4  Δt  ·u
⎜ 2 ⎟
pn+1 + pnh ⎜ ρ s +ρhn √ n+1 Js +Jnh √ n+1 ⎟
n+ 1 ⎜ + ρv h,s−1 + 2 · ∇ ρv h,s−1 · u ⎟
× (φhn+1 + φhn ), ph 2 = h . ⎜
⎜
2 ⎟
⎟
2 ⎜ −
ps +pnh
(∇ · u) + 1
∇ ⎟
⎜  2 2 ⎟
√ n+1 ⎜ ρ s +ρhn √  Js +Jhn  
√ n+ 2 1 ⎟
⎜ · ρv
n+1
+ ρv · u ⎟
Similar to Equation (21), we let u = ρv h , ϕ = ⎜
⎜
2 h,s−1 2 h,s−1 ⎟
⎟
⎜ √ n+1 ⎟
n+ 1 n+ 1 n+ 21 ⎝ √ √ +μ∇ ρv : ∇u ⎠
ωh 2 + cφh 2 , χ = φtn+1 , q = ph , and then we h,s
ρ s + ρhn 1 φ +φ n
obtain the discrete energy law − 2 g · u − 2 (ωs + cφ s + ωh + cφh )∇ 2 · u
n n s h

dx = 0, (30)
En+1 − En + Gn+1 − Gn
t
     
 √ n+1   n+ 1    
= − μ |∇ ρv h  |2L2 + δ |ph 2  |2L2 φ s − φhn √ n+ 21 φ s−1 + φhn
ϕ + ρv h,s−1,s−1 · ∇ ϕ
  Δt 2
 n+ 21  2
+ M |∇λh  |L2 , (26)   
1
  +M∇ (ωs + cφ s + ωhn + cφhn ) · ∇ϕ dx = 0,
1   σε 2
i i 2
E = | ρh vh  |L2 +
i ||∇φhi ||2L2 (31)
2 2
σ 2 2
+ (|φhi | − 1) dx, (27) 
4ε 1
(ωs + cφ s + ωhn + cφhn )χ − σ ε−1 g(φ s−1 , φhn )χ

i−1   2
k+ 1 √ k+1 
Gi = − Δtρh 2 g · ρv h dx, (28) φ s + φhn
k=0 −σ ε∇ ∇χ dx = 0, (32)
2
where some additional manipulation gives:
where
 
φhn+1 + φhn φhn+1 − φhn  
n+ 21 √ n+1 ρ s vs + ρhn vhn
∇φh · ∇φtn+1 = ∇ ·∇ ρv h,s =   ,
2 Δt
ρ s + ρhn
2 2
(∇φhn+1 ) − (∇φhn )2 (|∇φhn+1 | )t ωs + cφ s
= = , Js = −(ρ 2 − ρ 1 )M∇ ,
2Δt 2 2
 ENGINEERING APPLICATIONS OF COMPUTATIONAL FLUID MECHANICS 647

 
√ n+1 ρ s vs−1 + ρhn vhn (30) and (31), which significantly reduces the computa-
ρv h,s−1 =   , tional cost.
ρ s + ρhn
ωhn + cφhn
Jnh = −(ρ 2 − ρ 1 )M∇ ,
2 4. Numerical examples, results, and discussion
 
√ n+1 ρ s−1 vs−1 + ρhn vhn In this section, we use Equations (22)–(25) to re-compute
ρv h,s−1,s−1 =   . the examples in Guo et al. (2014), for the coalescence of
ρ s−1 + ρhn
two kissing bubbles and a lighter drop rising. We per-
First, we solve Equation (32) and to get φ s and ωs , then form the computations under the FreeFem++ platform
obtain vs and ps by inserting φ s and ωs into Equations (Hecht et al., 2007) and present the results with the help

Figure 1. Interface structure of two kissing bubbles at t = 0.00, 0.03, 0.07, 0.10, 0.15, and 0.25.
648 Y. LIN AND Y. JIANG
of Matlab and Tecplot. In the computation process, the
P2 finite element space is used for v, φ, and ω, and the
pressure p is defined in the P1 finite element space.
4.1. Coalescence of two kissing bubbles
We set two heavier kissing bubbles in a lighter fluid (den-
sity ratio is 1 : 10). We ignore the effect of gravity in this
example. The computation region ω is defined as
(x, y) ∈ [−0.5, 0.5] × [−0.5, 0.5].
Figure 2. Velocity field for two kissing bubbles at t = 0.03, 0.07, 0.10, 0.15, and 0.25 with φ = 0.
The computation starts with a relatively coarse 64 × 32
grid, but uses an adaptive mesh (which adapts every 100
time steps). The parameter values in this example are set
as follows:
ρ1 = 1, ρ2 = 10, μ = 10−2 , σ = 1.96, ε = 10−2 ,
M = 5 × 10−4 , c = 0.50, Δt = 10−5 .
For simplicity, p = 0 is set as the initial value of the
fixed point method. Figure 1 shows the evolution of the
bubbles at different moments t = 0.00, 0.03, 0.07, 0.10,
 ENGINEERING APPLICATIONS OF COMPUTATIONAL FLUID MECHANICS 649

0.15, 0.25, and the different colors in the vicinity of

the interface represent the continuous change from one
phase to the other (−1 to 1). The jagged interface at t =
0.00 is due to the relatively coarse grid (64 × 32). The two
kissing bubble interfaces become smooth by using the
preserving-energy law technique and adapted mesh. Fur-
thermore, two bubbles coalesce to a big bubble because of
the overlapped interface. Then it becomes elliptical under
the surface tension. As the order parameter φ changes
continuously from −1 to 1, we choose the φ = 0 con-
tours to show the interface structure and the velocity of
the fluid in Figure 2. The accuracy of the numerical solu-
tion for two kissing bubbles is validated, as the numerical
results agree well with those in Guo et al. (2014).
We examine the total free energy given by Equation
(27) and the energy law derived by Equation (26). It
should be noted that the total free energy is also the
total energy in the coalescence of the two kissing bub-
bles, because we have ignored the effect of gravity in this
case. The errors in the discrete energy law are shown in
Figure 3, and the decrease with time matches that given
by Equation (26). In the computation process, the dis-
crete energy law is preserved for each time step and the
computational accuracies are effective because the error
becomes less than 10−10 .

4.2. Dynamics of a rising droplet

We consider a lighter drop set in a heavier medium with
two density ratios: (1) 1 : 5 and (2) 1 : 50 (ρ1 = 1, ρ2 = Figure 3. (a) Evolution of the total energy; (b) Error in the discrete
5, 50). Moreover, we set the other parameters in this energy law.
example as follows:

μ = 10−2 , σ = 1.96, ε = 10−2 , M = 5 × 10−4 ,

c = 0.50, Δt = 10−6 .
We choose the computing domain as [0, 4] × [0, 4] with-
out considering that the moving contact line when the
rising drop touches the top of the computation region
should obey the slip-boundary conditions (Jiang et al.,
2015). The initial value of the order parameter φ is
φ(x, y, t = 0)
⎛ ⎛  ⎞⎞
1⎝ 0.3 − (x − 2)2 + (y − 0.85)2
= 1+tan h ⎝ √ ⎠⎠ .
2 2 2ε
We discretize the square domain with a relatively
coarse grid (64 × 64) without adapting the mesh. The
evolution of the interface structures of the droplets are
presented in Figure 4 (1 : 5) and Figure 5 (1 : 50). In the
case 1 : 5, the drop attains a mushroom shape at about
t = 0.15. The deformation is slower than for the higher
density ratio 1 : 50, which forms a mushroom shape at
about t = 0.12. As time goes on, the up-welling motion
causes the drop to change into a horseshoe shape at about
t = 0.2 and the droplet pinches off between t = 0.3 and
t = 0.35. These behaviors are similar to those reported in
Guo et al. (2014), which used ∇ · u = 0 and a 128 × 128
grid.
The accuracy of the numerical solutions in Aihara
et al. (2019) and Jiang et al. (2015) is examined
 by com-
puting ||∇ · (u − uh )|| = ||∇ · uh || =
n n |∇ · unh |2 dx
at each time step. The estimated order γ for ∇ · u is com-
∇·(u−uh )
puted as γ = lg12 × lg ∇·(u−u 0.5h )
. The results in Tables
1–3 show that the value of γ is close to 2.
For the case of 1 : 50, evolutions of the total free energy
given by Equation (27) and the total energy (E + G) given
Equation (27) + Equation (28) is shown in Figure 6(a).
The energy-law errors at discrete form introduced in
Equation (26) are also shown in Figure 6(b). They show
that the discrete energy law is preserved for each time step
650 Y. LIN AND Y. JIANG

IsoValue IsoValue IsoValue

0.5 0.5 0.5

IsoValue IsoValue IsoValue

0.5 0.5 0.5

IsoValue IsoValue
IsoValue 0.5
0.5 0.5

Figure 4. Evolution of rising drop with ratio 1 : 5 at t = 0.00, 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35, and 0.40.

and the computational accuracies are effective because
the error becomes less than 10−8 .
The sharp interface limit for the phase-field model
refers to the numerical solution that remains constant
when the thickness of the interface reduces to a given
value. From the results of Jiang et al. (2015) and Guo et al.
(2014), the sharp interface limit for a Cahn–Hilliard-
type phase-field model can be obtained by decreasing
the thickness of the interface ε, although the numerical
solution is useless unless the thickness of the interface
is at the nanometer scale. However, the general require-
ment for the length of the grid is at least one discrete-grid
element through the contact region. For example, if the
thickness of the interface is set to ε = 0.01, the length
1
of the grid should be at best 100 . If the sharp interface
limit does not exist in this mathematical model, the true
 ENGINEERING APPLICATIONS OF COMPUTATIONAL FLUID MECHANICS 651

IsoValue IsoValue IsoValue

0.5 0.5 0.5

IsoValue IsoValue IsoValue

0.5 0.5 0.5

IsoValue IsoValue IsoValue

0.5 0.5 0.5

Figure 5. Evolution of rising drop with ratio 1 : 50 at t = 0.00, 0.05, 0.12, 0.15, 0.20, 0.25, 0.30, 0.35, and 0.40.

numerical solution should be obtained by reducing the
thickness of the interface to the nanometer scale. This
requires the length of the grid to be on the nanometer
scale at least, which makes the computation more chal-
lenging. Thus, the sharp interface limit is very important
for phase-field models. We present an examination of the
sharp interface limit for the case 1 : 50 in Figure 7. We
consider four different values of the interface thickness:
ε = 0.03, ε = 0.02, ε = 0.01, ε = 0.005. Only three of
the four bubbles can be distinguished in Figure 7. This
result shows that the interface converges to a unique
shape for the cases where ε = 0.005 and ε = 0.01, but
not for the other two cases (ε = 0.02 and ε = 0.03). Thus,
the sharp interface limit exists in this model and the criti-
cal value of the interface thickness for the sharp interface
limit is ε ∈ [0.01, 0.02).
652 Y. LIN AND Y. JIANG

Table 1. Value and estimated order of ∇ · u for the kissing bub-

ble.
Mesh 32 × 16 64 × 32 128 × 64
∇ · (u − uh ) 0.000013587 0.000003743 0.000000879
Estimated order - 1.86 2.09

Table 2. Value and estimated order of ∇ · u for the rising drop

with ratio 1 : 5.
Mesh 32 × 32 64 × 64 128 × 128
∇ · (u − uh ) 0.002461438 0.000748203 0.000168579
Estimated order - 1.72 2.15

Figure 7. Interface structure for the rising drop with ε = 0.03,

Table 3. Value and estimated order of ∇ · u for the rising drop ε = 0.02, ε = 0.01, ε = 0.005, at t = 0.12.
with ratio 1 : 50.
Mesh 32 × 32 64 × 64 128 × 128
5. Conclusion
∇ · (u − uh ) 0.00857227 0.00305162 0.00069953
Estimated order - 1.49 2.13 We investigated a multiphase of different densities using
the model proposed by Abels et al. (2012). A penalty for-
mulation was applied for continuous conditions to ensure
stability of pressure. We used a continuous finite element
technique coupled with a new time form to treat the
Navier-Stokes-Cahn-Hilliard model. A modified discrete
energy law of the numerical simulation was derived. The
results using a relatively coarse grid and this numerical
scheme are the same as previous results with a finer grid,
demonstrating the higher efficiency of our approach. The
estimated order for ∇ · u in the numerical example of a
rising lighter drop is close to 2, validating the accuracy of
the numerical solution. The sharp interface limit of the
model presented in this paper was examined using the
numerical example of a rising lighter drop. The results
show that this simulation truly approaches a sharp inter-
face limit and the critical value of the interface thickness
is between 0.01 and 0.02. We do not consider the effect
of gravity and we will consider this in future work. Also,
we will continue our work on multiphase fluids with a
moving contact line. Indeed, we notice that there is little
research on deriving the equation of the discrete energy
law for the moving contact line. The work in Jiang et al.
(2015) has shown that we derived the equation of the
discrete energy law and one kind of numerical scheme
solving the moving contact line problem and preserv-
ing the energy law with no effect of different densities.
We will improve our numerical method and change the
numerical scheme to treat more complex boundary con-
ditions with the problem of moving contact lines with
different densities.

Acknowledgments
Figure 6. (a) Evolution of the total free energy (E) and the total
energy (E + G); (b) Error in the discrete energy law. Yanhai Lin was supported by the National Natural Sci-
ence Foundation of China (No. 11702101), the Fundamental
 ENGINEERING APPLICATIONS OF COMPUTATIONAL FLUID MECHANICS
Research Funds for the Central Universities and the Promo-
tion Program for Young and Middle-aged Teachers in Science
and Technology Research of Huaqiao University (No. ZQN-
PY502), the Natural Science Foundation of Fujian Province
(No. 2019J05093), and the Quanzhou High-Level Talents Sup-
port Plan.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Funding
Yanhai Lin was supported by the National Natural Sci-
ence Foundation of China (No. 11702101), the Fundamental
Research Funds for the Central Universities and the Promo-
tion Program for Young and Middle-aged Teacher in Science
and Technology Research of Huaqiao University (No. ZQN-
PY502), the Natural Science Foundation of Fujian Province
(No. 2019J05093), and the Quanzhou High-Level Talents Sup-
port Plan.
References
Abels, H., Carcke, H., & Grün, G. (2012). Thermodynami-
cally consistent, frame indifferent diffuse interface models
for incompressible two-phase flows with different densities.
Mathematical Models and Methods in Applied Sciences, 22(3),
1150013. https://doi.org/10.1142/S0218202511500138
Aihara, S., Takaki, T., & Takada, N. (2019). Multi-phase-
field modeling using a conservative Allen-Cahn equation
for multiphase flow. Computers and Fluids, 178, 141–151.
https://doi.org/10.1016/j.compfluid.2018.08.023
Akbarian, E., Najafi, B., Jafari, M., Ardabili, S. F., Shamshir-
band, S., & Chau, K. W. (2018). Experimental and compu-
tational fluid dynamics-based numerical simulation of using
natural gas in a dual-fueled diesel engine. Engineering Appli-
cations of Computational Fluid Mechanics, 12(1), 517–534.
https://doi.org/10.1080/19942060.2018.1472670
Aldakheel, F., Hudobivnik, B., Hussein, A., & Wriggers,
P. (2018). Phase-field modeling of brittle fracture using
an efficient virtual element scheme. Computer Methods
in Applied Mechanics and Engineering, 341, 443–466.
https://doi.org/10.1016/j.cma.2018.07.008
Altmann, R., & Heiland, J. (2015). Finite element decompo-
sition and minimal extension for flow equations. Esaim:
Mathematical Modelling and Numerical Analysis, 49(5),
1489–1509. https://doi.org/10.1051/m2an/2015029
Boyer, F. (2002). A theoretical and numerical model for the
study of incompressible mixture flows. Computers & Flu-
ids, 31(1), 41–68. https://doi.org/10.1016/S0045-7930(00)
00031-1
Chen, F., & Shen, J. (2013). Efficient energy stable schemes with
spectral discretization in space for anisotropic Cahn-Hilliard
systems. Communications in Computational Physics, 13(5),
1189–1208. https://doi.org/10.4208/cicp.101111.110512a
Guo, Z., Lin, P., & Lowengrub, J. S. (2014). A numerical
method for the quasi-incompressible Cahn-Hilliard-Navier-
Stokes equations for variable density flows with a discrete
energy law. Journal of Computational Physics, 276, 486–507.
https://doi.org/10.1016/j.jcp.2014.07.038
653
Han, L. H., Hu, X. Y., & Adams, N. A. (2014). Adap-
tive multi-resolution method for compressible multi-phase
flows with sharp interface model and pyramid data struc-
ture. Journal of Computational Physics, 262, 131–152.
https://doi.org/10.1016/j.jcp.2013.12.061
Hecht, F., Pironneau, O., Hyaric, A. L., & Ohtsuka, K. (2007).
FreeFem++ (Version 2.17-1). http://www.freefem.org/ff+
+/ftp/freefem++doc.pdf
Jiang, Y., & Lin, P. (2019). Improvement of numerical sim-
ulation for GMAW based on new model with thermo-
capillary effect. Journal of Computational and Applied
Mathematics, 356, 37–45. https://doi.org/10.1016/j.cam.2019.
01.039
Jiang, Y., Lin, P., Guo, Z., & Dong, S. (2015). Numerical simula-
tion for moving contact line with continuous finite element
schemes. Communications in Computational Physics, 18(1),
180–202. https://doi.org/10.4208/cicp.170314.160115a
Jiang, Y., & Yang, W. (2018). Numerical simulation for multi-
phase fluids considering different densities with continuous
finite element method. Journal of Physics: Conference Series,
1053, 012014. https://doi.org/10.1088/1742-6596/1053/1/
012014
Li, J., Fan, T., Taniguchi, T., & Zhang, B. (2018). Phase-field
modeling on laser melting of a metallic powder. Interna-
tional Journal of Heat and Mass Transfer, 117, 412–424.
https://doi.org/10.1016/j.ijheatmasstransfer.2017.10.001
Lin, P. (1997). A sequential regularization method for time-
dependent incompressible Navier-Stokes equations. SIAM
Journal of Numerical Analysis, 34(3), 1051–1071. https://
doi.org/10.1137/S0036142994270521
Lin, Y., & Jiang, Y. (2018). Effects of Brownian motion and
thermophoresis on nanofluids in a rotating circular groove:
A numerical simulation. International Journal of Heat
and Mass Transfer, 123, 569–582. https://doi.org/10.1016/
j.ijheatmasstransfer.2018.02.103
Liu, H., & Ding, H. (2015). A diffuse-interface immersed-
boundary method for two-dimensional simulation of flows
with moving contact lines on curved substrates. Journal
of Computational Physics, 294, 484–502. https://doi.org/
10.1016/j.jcp.2015.03.059
Lowengrub, J., & Truskinovsky, L. (1998). Quasi-incompressible
Cahn-Hilliard fluids and topological transitions. Proceedings
of The Royal Society A – Mathematical Physical and Engineer-
ing Sciences, 454(1978), 2617–2654. https://doi.org/10.1098/
rspa.1998.0273
Luo, J., Hu, X., & Adams, N. A. (2016). Curvature bound-
ary condition for a moving contact line. Journal of Com-
putational Physics, 310, 329–341. https://doi.org/10.1016/
j.jcp.2016.01.024
Meirmanov, A. M. (1992). The Stefan problem. De Cruyter.
Mosavi, A., Shamshirband, S., Salwana, E., Chau, K. W., &
Tah, J. H. M. (2019). Prediction of multi-inputs bubble col-
umn reactor using a novel hybrid model of computational
fluid dynamics and machine learning. Engineering Appli-
cations of Computational Fluid Mechanics, 13(1), 482–492.
https://doi.org/10.1080/19942060.2019.1613448
Murmatsu, M., Yashiro, K., Kawada, T., & Terada, K. (2018).
Simulation of ferroelastic phase formation using phase-
field model. International Journal of Mechanical Sciences,
146–147, 462–474. https://doi.org/10.1016/j.ijmecsci.2017.
12.027
654 Y. LIN AND Y. JIANG
Qian, T., Wang, X., & Sheng, P. (2006). A variational approach
to the moving contact line hydrodynamics. Journal of Fluid
Mechanics, 564, 333–360. https://doi.org/10.1017/S0022112
006001935
Ren, J., Chen, Y., Xu, B., Sun, M., & Li, D. (2019). A vector-
valued phase field model for polycrystalline solidification
using operator splitting method. Computational Materi-
als Science, 163, 37–49. https://doi.org/10.1016/j.commatsci.
2019.02.045
Salih, S. Q., Aldlemy, M. S., Rasani, M. R., Ariffin, A. K.,
Ya, T. M. Y. S. T., Ansari, N. A., Yaseen, Z. M., &
Chau, K. W. (2019). Prediction of multi-inputs bubble col-
umn reactor using a novel hybrid model of computational
fluid dynamics and machine learning. Engineering Appli-
cations of Computational Fluid Mechanics, 13(1), 860–877.
https://doi.org/10.1080/19942060.2019.1652209
Spelt, P. D. M. (2006). Shear flow past two-dimensional droplets
pinned or moving on an adhering channel wall at moder-
ate Reynolds numbers: A numerical study. Journal of Fluid
Mechanics, 561, 439–463. https://doi.org/10.1017/S002211
2006000954
Sui, Y., & Spelt, P. D. M. (2013). An efficient computa-
tional model for macroscale simulations of moving con-
tact lines. Journal of Computational Physics, 242, 37–52.
https://doi.org/10.1016/j.jcp.2013.02.005
Tsai, Y. L., Chen, C. C., & Lan, C. W. (2010). Three-dimensional
adaptive phase field modeling of directional solidifica-
tion of a binary alloy: 2D–3D transitions. International
Journal of Heat and Mass Transfer, 53(9–10), 2272–2283.
https://doi.org/10.1016/
j.ijheatmasstransfer.2009.11.040
Visintin, A. (1996). Models of phase transitions. PNLDE 28.
Birkhauser.
Wang, X., Zhang, H., Zhou, W., & Ouyang, J. (2017). A
3D phase-field model for simulating the crystal growth
of semi-crystalline polymers. International Journal of Heat
and Mass Transfer, 115, 194–205. https://doi.org/10.1016/
j.ijheatmasstransfer.2017.08.016
Yue, P., Zhou, C., & Feng, J. (2010). Sharp-interface limit of
the Cahn-Hilliard model for moving contact lines. Journal
of Fluid Mechanics, 645, 279–294. https://doi.org/10.1017/
S0022112009992679
Zhang, P., Hu, X., Wang, X., & Yao, W. (2018). An iteration
scheme for phase field model for cohesive fracture and its
implementation in Abaqus. Engineering Fracture Mechan-
ics, 204, 268–287. https://doi.org/10.1016/j.engfracmech.
2018.10.006
Zhang, Z., Kang, J., Guo, Z., & Han, Q. (2019). Effect of the
forced flow on the permeability of dendritic networks: A
study using phase-field-lattice Boltzmann method. Interna-
tional Journal of Heat and Mass Transfer, 131, 196–205.
https://doi.org/10.1016/j.ijheatmasstransfer.2018.11.036
Zhang, C., Lin, N., Tang, Y., & Zhao, C. (2014). A sharp interface
immersed boundary/VOF model coupled with wave gener-
ating and absorbing options for wave-structure interaction.
Computer & Fluids, 89, 214–231. https://doi.org/10.1016/
j.compfluid.2013.11.004
Zhang, Z., & Ren, W. (2016). Simulation of moving contact lines
in two-phase polymeric fluids. Computers and Mathematics
with Applications, 72(4), 1002–1012. https://doi.org/10.1016/
j.camwa.2016.06.016
Zhang, Z., & Tang, H. (2007). An adaptive phase field
method for the mixture of two incompressible fluids. Com-
puter & Fluids, 36(8), 1307–1318. https://doi.org/10.1016/
j.compfluid.2006.12.003
Zhao, J., Chen, L., & Wang, H. (2019). On power law scal-
ing dynamics for time-fractional phase field models dur-
ing coursing. Communications in Nonlinear Science and
Numerical Simulation, 70, 257–270. https://doi.org/10.1016/
j.cnsns.2018.10.019
Zhao, Y., Zhao, C. Y., Xu, Z. G., & Xu, H. J. (2016). Mod-
eling metal foam enhanced phase change heat transfer in
thermal energy storage by using phase field method. Inter-
national Journal of Heat and Mass Transfer, 99, 170–181.
https://doi.org/10.1016/j.ijheatmasstransfer.2016.03.076
Zhu, C., Xu, S., Feng, L., Han, D., & Wang, K. (2019). Phase-
field model simulations of alloy directional solidification
and seaweed-like microstructure evolution based on adap-
tive finite element method. Computational Materials Science,
160, 53–61. https://doi.org/10.1016/j.commatsci.2018.12.058