c Pleiades Publishing, Ltd., 2016.

ISSN 0021-8944, Journal of Applied Mechanics and Technical Physics, 2016, Vol. 57, No. 5, pp. 937–948. 
c C.-T. Lee, C.-C. Lee, J.-E. Lin, and M.-L. Liu.
Original Russian Text 

ON NUMERICAL MODELING OF CAPILLARY FLOW DYNAMICS

IN A MICROCHANNEL MODEL BY USING THE FINITE ELEMENT METHOD

C.-T. Leea , C.-C. Leeb , J.-E. Linc , and M.-L. Liud UDC 532.5

Abstract: A simple microchannel model with submillimeter-scale geometries is proposed for study-
ing capillary flows and investigating the dynamics in the channel. The finite element method incor-
porating surface tension and two-phase flow characteristic is applied. Velocity and pressure fields in
the microchannel are presented. It is shown that the capillary-phase front in the microchannel is
stirred, suffering small oscillations and retreating from the previous position before traveling again.
Such a phenomenon is caused by nonlinear interaction of the capillary flow, surface tension, and
boundary conditions.
Keywords: two-phase capillary flow, microchannel, surface tension, finite element method.
DOI: 10.1134/S0021894416050229

INTRODUCTION

Capillary microfluidics systems have attracted considerable attention of researchers in various disciplines,
such as chemistry, biology, biomedicine, and material science [1]. In general, engineers use high-resolution charge
coupled devices (CCDs) and high-speed cameras to visualize capillary flows in microchannels.
Capillarity is a natural phenomenon caused by surface tension effects where the liquid wets the interior of
the channel wall according to adhesion and cohesion forces [2]. In practical applications, it is often referred to
microfluidic chips that can process or manipulate small amounts of fluids, typically of the order of nanoliters or
smaller. Various substrates [glass, polymer, or polydimethylsiloxane (PDMS)] are used to fabricate microchannels
by means of laser machining, mask and photoresist, lithography and etching methods [3]. Capillary flows found
numerous applications in microelectromechanical systems (MEMS) mainly because of their power-free and low-
cost characteristics [4–8]. However, neither the pressure distribution in the microchannel was obtained, nor the
conditions that give rise to a capillary flow were found. The research on the capillary flow, either numerically or
theoretically, is still relatively scarce with only few illustrative issues being treated [9–12].
To solve this issue, we resort to the numerical simulation, which is an indispensable tool for studying many
nonlinear wave processes described by the nonlinear Navier–Stokes equations. A capillary flow driven by gravity in
an L-shaped microchannel with a submillimeter scale is numerically simulated. The finite element analysis and the
level set method are applied to solve the Navier–Stokes equations incorporating surface tension and two-phase flow
characteristics.

a
Beijing Institute of Technology-Bryant College, Zhuhai, Guangdong, P.R. China; scat1440@yahoo.com.
b
Department of Chemistry, Simor Fraser University, British Columbia, BC, Canada; kdv1890@yahoo.co.uk.
c
Department of Mathematical Sciences, George Mason University, Fairfax, Virginia, USA; jelin@gmu.edu.
d
Department of Physics, Shantou University, Shantou, Guangdong, P. R. China; 12mlliu@gmail.com. Translated
from Prikladnaya Mekhanika i Tekhnicheskaya Fizika, Vol. 57, No. 5, pp. 199–211, September–October, 2016.
Original article submitted December 2, 2013; revision submitted October 10, 2014.
0021-8944/16/5705-0937 
c 2016 by Pleiades Publishing, Ltd. 937
 1

AX
gsl gsa
Y
g la 2 h
X
gsl gsa
AT

Fig. 1. Microchannel geometry: (1) reservoir; (2) microchannel.

g la

1 3

2 gsl gsa

Fig. 2. Configuration of the liquid droplet, surface tension forces, and capillary forces
on the interfaces between the liquid (1), solid (2), and air (3).

1. MODEL DEFINITION

The present model (Fig. 1) consists of a capillary channel with a radius of 10 mm and a length of 100 mm
attached to a water reservoir with a radius of 15 mm and a length of 100 mm (h is the channel height, AX is the
area of the wetted region along the x coordinate, AT is the area of the entire region of length l, and H is the height
of the water reservoir counted from the axis of symmetry of the microchannel). Water is set to flow freely into the
outlet on the top of the reservoir. Initially the horizontal flow channel is filled with air. Gravity and wall adhesion
cause the water in the reservoir to creep into the channel and move along its boundary. The deformation of the
water surface induces surface tension at the air/water interface, which, in turn, creates a pressure jump across the
interface. The pressure variation pushes the water and air to move, and the motion of the fluids is maintained due
to capillary and gravity forces. The configuration of the liquid droplet and the surface tension and capillary forces
are illustrated in Fig. 2 (θ is the contact angle and γij is the surface tension between the phases i and j). As both
the reservoir and the channel are cylinders, we solve an axisymmetric problem in the plane rz (r and z are the
cylindrical coordinates).
938
2. MATHEMATICAL MODEL OF THE WATER/AIR INTERFACE

The capillary flow is a two-phase and laminar flow, and the interface is defined by using a level set function
ϕ(t, x). The zero contour ϕ(t, x) = 0 defines the interface. The interface motion is described by the transport
equation

ϕt + ∇ · (ϕu) = 0. (1)

The present model not only emphasizes the capillary flow in a microchannel, but also reflects a steady flow. In
addition, for finding the velocity and pressure fields, we have to solve the Navier–Stokes equations that describe
the transport of mass and momentum for fluids of constant density incorporating the capillary effects and surface
tension:
∂u  
ρ + ρ(u · ∇)u = ∇ · pI + η((∇u) + (∇u)t ) + Fst + ρg; (2)
∂t

∇ · u = 0. (3)

Here ρ [kg/m3 ] is the density, η [N · s/m2 ] is the dynamic viscosity, u [m/s] is the velocity, p [Pa] is the pressure,
g [m/s2 ] is the gravity acceleration vector, and Fst is the surface tension force acting at the air/liquid interface,
which is represented as
 ∇ϕ 
Fst = σκ∇ϕ = −σ ∇ · ∇ϕ.
|∇ϕ|
Here κ is the mean curvature of the surface described by the level set function ϕ as
 ∇ϕ 
κ =− ∇· ,
|∇ϕ|
σ [N/m] is the surface tension coefficient, and n = ∇ϕ/|∇ϕ| is the normal vector to the interface.
The “jumps” in the density and viscosity across the interface are smoothed by the formulas

ρ = ρair + (ρwater − ρair )ϕ, μ = μair + (μwater − μair )ϕ,

where ρair , ρwater , μair , and μwater are the dimensionless densities and viscosities of water and air, respectively. Such
an approach ensures smooth computations of the entire flow.
For incompressible flows, the velocity field is divergence-free, i.e., ∇ · u = 0; therefore, Eq. (1) is equivalent
to the conservation law

ϕt + u · ∇ϕ = 0.

We would resort to the finite element method to solve the equation of motion of the interface. Such a method
should be conservative, no spurious oscillations should occur during the computations, and the profile of ϕ should
be kept constant during the calculations. For the one-dimensional problem, the divergence-free condition implies
that the velocity is constant. However, for several space dimensions, the divergent-free condition does not really
imply a constant velocity, which causes variations in the velocity and distorts the shape of the function ϕ across
the interface. To handle this problem numerically, we make some adjustments by introducing additional terms
 ∇ϕ 
ϕt + u · ∇ϕ = γ∇ · ε∇ϕ − ϕ(1 − ϕ) , (4)
|∇ϕ|
where ε is the parameter that determines the thickness of the interface and γ is the reinitialization parameter. This
is due to the fact that most numerical methods can easily introduce some artificial diffusion, which smears the
profile of the interface during the numerical computation. Therefore, the value of γ has to be carefully chosen in
order to prevent this smearing and maintain the interface profile. In our approach, we take γ = 1 and ε = hc /2
(hc is the characteristic finite element mesh size in the domain across the interface). The procedure of obtaining a
steady-state solution of Eq. (4) is referred to as the reinitialization.
939
3. SURFACE TENSION

Surface tension makes the surface of a liquid act as an elastic sheath, which minimizes the surface area of
the liquid so as to minimize the energy of the entire fluidic system. The surface tension force applied to the liquid
can be deduced from the derivative of the surface energy with respect to the spatial coordinate. It is seen in Fig. 2
that there are three surface forces, namely, γla , γsl , and γsa , acting at the liquid/solid/air interface and satisfying
Young’s law [13, 14]:
γsa = γsl + γla cos θ.
The total surface energy of the capillary channel is composed of four parts (see Fig. 1). The first one is the vacant
area (AT − AX ) multiplied by γsa . The second part is the wetting area AX multiplied by γsl . The third part is the
surface energy E0 stored in the filling reservoir (E0 hardly changes due to the infinitesimal amount of the liquid
entering the capillary). The fourth part is the complex surface of the capillary meniscus front multiplied by γla (the
fourth term is neglected because the meniscus front area is very small as compared to other surfaces).
Then the total energy in the capillary channel (see Fig. 1) is expressed as
Es = E0 + [AT γsa + AX (γsl − γsa )].
Assuming that the cross section of the capillary channel is a rectangle with a width w and a height h, the total
energy can be expressed as
Es = E0 + 2(h + w)[lγsa − x(γsa − γsl )]. (5)
Taking the derivative of Eq. (5) with respect to x, we obtain the equivalent capillary force Fs applied to the liquid
column along the x direction:
dEs
Fs = − = 2(h + w)(γsa − γsl ) = Δpla wh.
dx
The pressure drop Δpla across the liquid/air interface is, therefore, deduced under the assumption that the
channel height h is much smaller than the channel width w [15]:
2(h + w)(γsa − γsl ) 2(γsa − γsl )
Δpla =  . (6)
wh h
Equation (6) can be rewritten as the so-called Laplace pressure drop for the capillary tube by replacing the hydraulic
radius rh = Dh /2 = wh/(w + h) of the rectangular microchannel with the inner radius r of the capillary tube:
2(γsa − γsl ) 2γla cos θ
Δpla =  . (7)
r r
As it follows from Eqs. (6) and (7), the smaller the channel/capillary tube size, the greater the pressure drop
across the liquid–air interface. For a water-filled glass microchannel in air under standard laboratory conditions,
the surface tension is γla = 0.0728 N/m, ρ = 1000 kg/m3 , and θ = 67.5◦ in our simulations. For these values, the
pressure gradient across the meniscus is
2 · 0.0728 cos 67.5◦
Δpla   5.57. (8)
10−4
It should be noted that the traditional statement for the considered problem is to pose the no-slip condition
on the wall and the full kinematic and dynamic conditions on the free surface. Further the statement can be
simplified with the help of the asymptotic approach [16–21].

4. FINITE ELEMENT ANALYSIS

We use the finite element method to discretize the Navier–Stokes equations (2) and (3) with surface tension
and convection of the capillary flow interface. Accordingly, we would need to set these equations in a form of what
we call the weak form of the problem, which is the basis for constructing our finite element solutions by multiplying
the differential equations with test functions, which are zero at fixed boundaries, but otherwise arbitrary, and then
940
integrate over the domain of consideration. In order to apply our method to complex geometries, we use adaptive
grids in space and a finite difference scheme in time [22–24].
We first define the finite-dimensional space of test functions on the set of piecewise-linear functions f (x) as
Vh = {f (x): f (x) = 0 ∀x ∈ Λ ⊂ ∂Ω},
where Ω is the computational domain, ∂Ω represents the boundary of Ω, and Λ is the part of the boundary ∂Ω
where the Dirichlet boundary conditions apply.
There is also a vector-valued function space defined as

f (x) = [f1 (x), . . . , fn (x)]t ∀x ∈ Ω,
Wh =
fi (x) = 0 ∀x ∈ Λi ⊂ ∂Ω, i = 1, . . . , n
[fi (x) is a piecewise-linear function]. Solving Eq. (1) is now formulated as finding a function ϕ ∈ Vh such that
  
vϕt dx − ∇v · (ϕu) dx + vϕu · v̌ dS = 0 ∀v ∈ Vh , (9)
Ω Ω ∂Ω

where v̌ is the normal vector on the channel walls. If the boundaries are wetted walls (u · v̌ = 0), then the last term
in Eq. (9) vanishes.
The spatial discretization of the reinitialization equation (4) reduces to the problem of finding ϕ ∈ Vh such
that
  
vϕt dx − ∇v · (−f + γε∇ · (∇ϕ)) dx + v(f − γε∇ · (∇ϕ)) · v̌ dS = 0 ∀v ∈ Vh , (10)
Ω Ω ∂Ω

where f = γϕ(1 − ϕ). To avoid any flow crossing through the boundaries, the integrals over the boundary should
be set to zero.
The temporal discretization of Eq. (9) is performed by using the explicit Euler method. Let ϕn = ϕ(tn ) at
the time step tn . Before proceeding, an intermediate value ϕn+1
c ∈ Vh has to be calculated first as follows:
 
ϕ n+1
−ϕ n
v c dx − ∇v · (ϕn un ) dx = 0 ∀v ∈ Vh . (11)
dt
Ω Ω

In addition, an intermediate value of the normal vector to the interface n̂n+1

c ∈ Wh has to be approximated at the
same time:
 
∇ϕn+1
v· c
dx = v · n̂n+1
c dx = 0 ∀v ∈ Wh . (12)
|∇ϕn+1
c |
Ω Ω

After calculations by Eqs. (11) and (12), we use a second-order accurate discretization in time for Eq. (4). We
start by choosing m = 0 and ϕ0l = ϕn+1
c ; then, for m = 1, 2, . . . , we can determine ϕm+1
l ∈ Vh by the formula
   m+1 
ϕm+1
− ϕlm
ϕ + ϕl m
v l dx + γ l − ϕm+1
l ϕm
l ∇v · n̂n+1
c dx
dt 2
Ω Ω
  ϕm+1 + ϕm 
− εγ ∇ l l
· n̂n+1
c (∇v · n̂n+1
c ) dx = 0 ∀v ∈ Vh . (13)
2
Ω
The iterations are terminated when
ϕm+1
l − ϕml 
< ζ.
dt
Here we set ζ = 0.001 in our numerical approach as a tolerance error. We expect only a few time steps to fulfill
condition (13), and finally we set ϕm+1
l = ϕn+1 for a new updated value in Eq. (10). In our numerical computation,
we set ε = hc /2.
After obtaining an updated value of the function ϕ from Eq. (13), we apply the finite element method to
solve the incompressible Navier–Stokes equations (2) and (3) with surface tension. The spatial discretization of the
Navier–Stokes equations is to find u ∈ Wh and p ∈ Vh such that
941
     
(ρu)t · v dx − (u · ∇v)ρu dx = (∇ · v)p dx − η ∇vi · (∇ui + uxi ) dx
Ω Ω Ω Ω i

+ v · (ρg + Fst ) dx ∀v ∈ Wh ,
Ω

q(∇ · u) dx = 0 ∀q ∈ Vh .
Ω
To obtain numerical results, one needs to approximate the surface curvature ϕ = const and the gradient of
the function ϕ. The approximation (∇ϕ)n+1 is determined from the equation
 
(∇ϕ)n+1 · v dx = ∇(ϕn+1 ) · v dx ∀v ∈ Wh .
Ω Ω
The mean curvature κ n+1 is then calculated by the formula
 
(∇ϕ)n+1
vκ n+1 dx = ∇v · dx ∀v ∈ Wh .
|(∇ϕ)n+1 |
Ω Ω
The intermediate velocity un+1
c can now be calculated by taking the pressure term p explicitly. We have to find
un+1 ∈ W h such that
c
 
1
(ρn+1 un+1
c − ρ u
n n
c ) · v dx − (un · ∇v) · ρun+1
c dx
dt
Ω Ω
  
= (∇ · v) · pn dx − η n+1 ∇vi · (∇un+1
ci + unxi ) dx
Ω Ω i

+ v · (ρn+1 g + Fstn+1 ) dx ∀v ∈ Wh ,
Ω
where
Fstn+1 = κ n+1 (∇ϕ)n+1 .
Then, we find a pseudovalue of un+1 , but with the pressure term being taken implicitly. We have to find
u n+1
∈ Wh such that
 
1
(ρ u
n+1 n+1
− ρ u ) · v dx − (un · ∇v) · ρun+1 dx
n n
dt
Ω Ω
  
= (∇ · v) · pn+1 dx − η n+1 ∇vi (∇un+1
ci + unxi ) dx
Ω Ω i

+ v(ρn+1 g + Fstn+1 ) dx ∀v ∈ Wh .
Ω
In order to find un accurately, the updated pressure pn+1 is required, i.e., we have to find pn+1 ∈ Vh such
that
 
1 ∇q · ∇(pn+1 − pn )
− q∇ · un+1
c dx = dx ∀q ∈ Vh .
dt ρn+1
Ω Ω
Finally, we can find an updated value of un+1 , i.e., un+1 ∈ Wh such that
 
un+1 − un+1 ∇(pn+1 − pn )
v· c
dx = − v · dx ∀v ∈ Wh .
dt ρn+1
Ω Ω
The Navier–Stokes solver from the software package, CFD-ACE+, is used in the present computations.
The method of mapped meshes is used in the computational domain. The number of mesh points is 871,
and the number of quadrilateral elements is 780 (Fig. 3).
942
 y .10-4, m
8

_2
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 x .10-3, m

Fig. 3. Mapped mesh structure in the computational domain.

5. INITIAL AND BOUNDARY CONDITIONS

In our numerical approach, the function ϕ has to be initiated at t = 0. Therefore, we assign the air/liquid
interface at ϕ = 0.5 (ϕ = 0 in the air and ϕ = 1 in the liquid). A smeared Heaviside step function is used to
describe the function ϕ in the domains 0 ≤ ϕ ≤ 0.5 and 0.5 < ϕ ≤ 1.0:
⎧
⎪
⎨ 0, ϕ < −λ,
−1
H(ϕ) = 1/2 + ϕ/(2λ) + (2π) sin (πϕ/λ), −λ ≤ ϕ ≤ λ,
⎪
⎩
1, ϕ > λ.
The hydrostatic pressure p = ρgH is prescribed at the inflow boundary. Only water enters through the inlet
so that the level set function in this area is ϕ = 1.
At the outlet, the pressure is set to be p = 5.57 according to Eq. (8) for the fixed control volume model of
the L-shaped microchannel, and no other conditions are imposed on the level set function ϕ.
The conditions on the channel walls are vanishing of the normal velocity component
u · nwall = 0
and the frictional boundary force
η
Ff r = − u,
β
where β is the slip length. The boundary conditions also allow us to specify the contact angle θ (in our example,
θ = 67.5◦ ), and the slip length equals the mesh element size.
The computations were performed for the following parameters: θ = 3π/8 rad, p = 105 Pa, T = 293 K,
g = 9.81 m/s2 , ρwater = 1000 kg/s3 , ρair = 1.293 kg/s3 , μwater = 8.90 · 10−4 Pa · s, μair = 1.85 · 10−6 Pa · s, and
σ = 7.2 · 10−2 N/m.

6. RESULTS AND DISCUSSION

The time evolution of the capillary interface during the first five seconds of the process is shown in Fig. 4.
The surface tension imposed by the surface curvature begins to bend the water front and draw the water through
the boundaries. After the liquid has filled the inlet, the liquid front acquires the shape of a meniscus in the channel.
The contact angle remains unchanged during the numerical computation, showing that our numerical method is
reliable and accurate.
The outlet pressure is considered to be p = 5.57 on the boundary with no viscous stress, and the walls are
subjected to the slip boundary condition with the contact angle θ = 67.5◦. The flow is regarded as isothermal, and
the flow conditions are considered to be laminar.
943
 y .10-4, m (a) u, m/s y .10-4, m (b)
8 1.0 8
0.9
6 0.8 6
0.7
4 0.6 4
0.5
2 0.4 2
0.3
0 0.2 0
0.1
_2 0 _2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m 0 0.2 0.4 0.6 0.8 1.0 1.2

y .10-4, m (c) y .10-4, m (d)

8 8

6 6

4 4

2 2

0 0

_2 _2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m 0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m

y .10-4, m (e)
8

_2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m
Fig. 4. Capillary flow and interface (meniscus) location at different times: t = 0 (a), 5 · 10−5 (b),
10−3 (c), 2.5 · 10−3 (d), and 5 · 10−3 s (e).

944
 y .10-4, m (a) y .10-4, m (b)
8 8

6 6

4 4

2 2

0 0

_2 _2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m 0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m

y .10-4, m (c) y .10-4, m (d)

8 8

6 6

4 4

2 2

0 0

_2 _2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m 0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m

y .10-4, m (e)
8

_2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m
Fig. 5. Velocity field of the capillary flow at different times: t = 0 (a), 5 · 10−5 (b), 10−3 (c),
2.5 · 10−3 (d), and 5 · 10−3 s (e).

945
 y .10-4, m (a) u, m/s y .10-4, m (b) u, m/s
8 0 8 100
_100
6 6 50
_200

4 _300 4 0
_400
2 _500 2 _50

_600
_100
0 _700 0
_800 _150
_2 _2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m 0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m

y .10-4, m (c) u, m/s y .10-4, m (d) u, m/s

8 10 8 0
0
6 _10 6 _20
_20
_40
4 _30 4
_40 _60
2 _50 2 _80
_60
0 _70 0 _100
_80
_120
_90
_2 _2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m 0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m

y .10-4, m (e) u, m/s

8
0
6
_50
4
_100
2
_150
0
_200
_2
0 0.2 0.4 0.6 0.8 1.0 1.2 x .10-3, m
Fig. 6. Pressure distribution of the capillary flow at different times: t = 0 (a), 5 · 10−5 (b), 10−3 (c),
2.5 · 10−3 (d), and 5 · 10−3 s (e).

946
 x .10-4, m

0 1 2 3 4 5 t.10-3, s

Fig. 7. Time evolution of the interface/wall contact point position in a capillary flow moving in a
microchannel with a velocity u = 0.116 m/s.

The velocity field at different times are recorded in Fig. 5. It shows that the liquid is jostled to the interface
of the capillary flow at the beginning and then squeezed to the channel bottom, oscillating slightly and moving
forward when the capillary action takes place. There is a bounce-back of the flow near the turn (corner) of the
model due to gravity, making the liquid cram into the bottom area of the channel, and then the capillary action
takes over the situation to push the liquid to creep along the channel. The maximum velocity is located at the
channel bottom near the interface, featuring the strong interaction between surface tension and the wetted wall
boundary.
The surface plot of the pressure distribution is shown in Fig. 6, which is the first time it has been reported
for a capillary flow in a microchannel. At t = 0, the horizontal channel is filled with air and the reservoir is filled
with water, creating a static pressure gradient in the model. The intense pressure mark is set foot on the inner
middle wall near the reservoir/channel intersection. At t = 5 · 10−4 s, the first sign that something is unusual comes
a few moments before the capillary and gravity being set to take action. At the final time of the calculation, the
pressure jump across the liquid/air interface reaches 300 Pa. Such a jump is caused by surface tension, which forces
the water and air to move through the horizontal channel and overcome the flow resistance. The result also suggests
that the high pressure gradient dominates the flow behavior in the channel.
Furthermore, we calculate the position of the interface/wall contact point by integrating the level set function
along the horizontal channel. Figure 7 shows the position of the contact point as a function of time. The slight
oscillations of the curve suggest that the water front might be stirred initially when the capillary action takes place.
This phenomenon also shows a nonlinear effect between surface tension and the wetted wall boundaries.

CONCLUSIONS

We design a microscale capillary model and investigate the capillary flow and interface behavior in a channel
by a numerical method. The flow dynamics is analyzed. The velocity and pressure fields are obtained. A bounce-
back of the flow caused by capillary and gravity forces is observed near the reservoir/channel junction. The position
of the interface/wall contact point is calculated as a function of time.
Numerical results obtained on the basis of the proposed mathematical model provide new information about
capillary flows in microchannels and can be used to improve technologies based on this phenomenon, which are
applied in chemistry, biomedicine, and engineering.
947
REFERENCES

1. D. J. Beebe, G. A. Mensing, and G. M. Walker, “Physics and Applications of Microfluidics in Biology,” Annual
Rev. Biomed. Eng., No. 4, 261–286 (2002).
2. E. W. Washburn, “The Dynamics of Capillary Flows,” Phys. Rev. No. 17, 273–282 (1921).
3. E. Kim, Y. Xia, and G. M. Whitesides, “Polymer Microstructures Formed by Moldings in Capillaries,” Nature
376, 581–584 (1995).
4. M. K. Schwiebert and E. H. Leong, “Underfill Flow As Viscous Flow between Pparallel Plates Driven by
Capillary Action,” IEEE Trans. Components, Packaging Manufactur. Technol. 19, 133–137 (1996).
5. Q. Weilin, G. M. Mala, and D. Q. Lee, “Pressure-Driven Water Flows in Trapezoidal Silicon Microchannels,”
Int. J. Heat Mass Transfer 43, 353–361 (2000).
6. E. Kim and G. M. Whitesides, “Imbibition and Flow of Wetting Liquids in Non-Circular Capillaries,” J. Phys.
Chem. 101, 855–863 (1997).
7. Y. Zhu and K. Petkovic-Duran, “Capillary Flow in Microchannels,” Microfluid. Nanofluid., No. 8, 275–282
(2010).
8. W. R. Jong, T. H. Kuo, S. W. Ho, et al., “Flows in Rectangular Microchannels Driven by Capillary Force and
Gravity,” Int. Comm. Heat Mass Transfer 34, 186–196 (2007).
9. G. Mason and N. R. Morrow, “Effect of Contact Angle on Capillary Displacement Curvatures in Pore Throats
Formed by Spheres,” J. Colloid Interface Sci. 168, 130–141 (1994).
10. R. Turian and F. Kessler, “Capillary Flow in a Noncircular Tube,” AIChE J. 46, 695–702 (2000).
11. D. Ericson, D. Li, and C. B. Park, “Numerical Simulations of Capillary-Driven Flows in Nonuniform Cross-
Sectional Capillaries,” J. Colloid Interface Sci. 250, 422–430 (2002).
12. W. B. Young, “Analysis of Capillary Flows in Non-Uniform Cross-Sectional Capillaries,” Colloids Surfaces,
A: Physicochem. Eng. Aspects 234, 123–128 (2004).
13. J. N. Israelachvili, Intermolecular and Surface Forces (Academic, London, 1985).
14. P. G. de Gennes, “Wetting: Statics and Dynamics,” Rev. Mod. Phys. 57, 827–890 (1985).
15. N. Tas, “Nanofluidic Bubble Pump Using Surface Tension Directed Gas Injection,” Anal. Chem. 74, 2224–2227
(2002).
16. O. V. Voinov, “Hydrodynamics of Wetting,” Fluid Dyn. No. 11, 714–721 (1976).
17. C. Baiocci and V. V. Pukhnachev, “Problems with One-Sided Constraints for Navier–Stokes Equations and
the Dynamic Contact Angle,” J. Appl. Mech. Tech. Phys. 31 (2), 185–197 (1990).
18. V. E. B. Dussan and S. H. Davis, “On the Motion Fluid–Fluid Interface along a Solid Surface,” Fluid Mech.
65, 71–95 (1974).
19. V. E. B. Dussan, “On the Spreading of Liquids on Solid Surfaces: Static and Dynamics Contact Lines,” Annual
Rev. Fluid Mech. 11, 371–400 (1979).
20. V. V. Pukhnachev and V. A. Solonnikov, “On the Problem of Dynamic Contact Angle,” J. Appl. Math. Mech.
46, 771–779 (1982).
21. O. V. Voinov, “Dynamics of Wetting of a Solid by Liquid: Movement of Thin Film,” in Encyclopedia of Surface
and Colloid Science (Marcel Dekker Inc., New York, 2002), pp. 1546–1559.
22. W. N. Gordon and C. A. Hall, “Construction of Curvilinear Coordinate Systems and Application to Mesh
Generation,” Int. J. Numer. Methods Eng. 7, 461–477 (1973).
23. W. N. Gordon and L. C. Thiel, “Transfinite Mappings and Their Application to Grid Generation,” in Numerical
Grid Generation (Elsevier, New York, 1982), pp. 171–192.
24. Y. C. Liou and Y. N. Jeng, “A Transfinite Interpolation Method of Grid Generation Based on Multipoints,”
J. Sci. Comput. 13, 105–114 (1998).

948