See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/230852271

Numerical Simulation of Droplet-Based Microfluidics - A Review

Article in Micro and Nanosystems · September 2010

DOI: 10.2174/1876402911002030193

CITATIONS READS
30 1,486

2 authors:

Jing Liu Nam-Trung Nguyen

Nanyang Technological University Griffith University
19 PUBLICATIONS 456 CITATIONS 949 PUBLICATIONS 33,063 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Nam-Trung Nguyen on 29 May 2014.

The user has requested enhancement of the downloaded file.

 Micro and Nanosystems, 2010, 2, ????-???? 1

Numerical Simulation of Droplet-Based Microfluidics - A Review

Jing Liu and Nam-Trung Nguyen*
School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore,
639798
Abstract: This paper discusses the features and applications of interface tracking techniques for modeling droplet-based
microfluidics. The paper reviews the state of the art of methods for tracking and capturing the interface. These methods
are categorized as the implicit and the explicit methods. The explicit methods need to reconstruct the interface by recon-
necting the markers on the interface. The implicit techniques implicitly describe the interface as a simple function. Thus,
complicated topological changes of interface can be handled naturally and automatically. The implicit methods reviewed
in this paper are the volume-of-fluid method, the phase-field method, the Lattice-Boltzmann method and the level-set
method. The explicit methods mainly include the boundary-integral methods and the tracking methods.
Keywords: Droplet-based microfluidics, boundary-integral method, tracking methods, volume-of-fluid method, level-set
method.

1. INTRODUCTION Recently, different techniques were proposed for inter-

face prediction. Traditionally, the interface that separates the
Over the last few decades, modeling immiscible fluids two phases can be solved using volume-tracking method,
such as oil and water have been a classical research topic. front-tracking method, fractional volume of fluid (VOF),
Droplet-based microfluidics presents a unique platform for phase-field method, lattice Boltzmann method, the level-set
mixing, reaction, separation, dispersion of drops and numer- method (LSM), and others. Slavov and Dimova categorized
ous other functions [1-4]. Droplet-based microfluidics is of these methods into two types: explicit tracking methods and
great interest for biological research, chemical synthesis, implicit tracking methods [9]. Explicit methods include
drug delivery and medical diagnostics. Droplet-based micro- boundary integral method, front tracking method and im-
fluidics refers to devices and methods for controlling the mersed interface method. Implicit ones include phase-field
fluid flows in length scales smaller than one millimeter. method and level-set method. The present review will be
Monodisperse droplets in microfluidic devices have been conducted to these categories. Some researchers classified
generated using different microchannel configurations such the methods into interface tracking methods and interface
as T-junction [5-6], flow focusing [7], or co-flowing [8]. The capturing methods based on whether the computational mesh
droplet size and the generation frequency can be accurately or massless markers is attached on the interface [10-11]. The
controlled by adjusting the flow rate ratio of the two phases, former includes boundary-integral methods, finite-element
their viscosities and interfacial tension. In a microfluidics methods and immersed boundary methods. The latter in-
system, the effects of volume-based inertia and gravity do cludes lattice Boltzmann method, constrained-interpolation-
not play an important role as compared to the macroscale profile, level-set, volume-of-fluid, coupled level-set, volume-
case. The surface-based interfacial tension and the viscosity of-fluid and partial-miscibility-model and phase-field meth-
dominate the behavior of multiphase flow in microscale. ods. Garrioch and Baligo proposed to categorize these tech-
Numerical simulation is a powerful tool for understanding niques as the moving font mesh to tack the interface and the
these complex multiphysics phenomena and the dynamics of fixed mesh to capture the interface location [12]. Each cate-
a microfluidic system. gorization approach summarizes one of characters of the
In general, the Navier-Stokes equations are solved nu- numerical methods of the interface prediction. For example,
merically to obtain the velocity and pressure of fluid based the level set method is an implicit, interface capturing and
on the established discretisation schemes such as the finite employing the fixed grid method.
difference method (FDM), the finite volume method (FVM),
or the finite element method (FEM). Recently, the Lattice 2. PROBLEMS WITH MULTIPHASE SYSTEMS
Boltzmann method (LBM) is developed to solve the Navier-
Stokes equations. However, the challenge in solving multi- Droplets Formation
phase fluid flow problems is handling the interface separat-
Menech applied the phase-field method to model the
ing the fluids involved and its evolution over time. The inter-
breakup of a big droplet in a T-shaped junction [13]. The
face shape deforms continuously as two phases significantly
affect each other as well as the boundary conditions on them. numerical results demonstrated that the droplet breakup re-
Nevertheless, special treatments need to deal with the prop- gion depends on the Capillary number (Ca), viscosity ratio
erty jump (or gradient) across the interface. A precise inter- (
= μ d / μ c ), and the initial drop dimension. Menech dem-
face location needs to be determined. onstrated that the numerical method can accurately describe
the dynamics of a droplet. The same numerical method was
used to investigate the different droplet formation regimes:
*Address correspondence to this author at the School of Mechanical and squeezing, dripping and jetting. The regimes are distin-
Aerospace Engineering, Nanyang Technological University, 50 Nanyang guished by the capillary number Ca and the viscosity ratio
.
Avenue, Singapore, 639798; Tel: (+65) 67904457; Fax: (+65) 67911859;
E-mail: [email protected]
The results show that the droplet volume is strongly affected

1876-4029/10 $55.00+.00 © 2010 Bentham Science Publishers Ltd.

2 Micro and Nanosystems, 2010 Vol. 2, No. 3 Liu and Nguyen

by the viscosity ratio. In a similar T-junction channel, lattice ment between the level-set method and the diffuse interface
Boltzmann method was used to study the behavior of the method to solve the drop spreading problem [38].
droplet formation process [14-15]. By employing volume of
fluid model in three-dimension (3D), the impact of the vis- 3. EXPLICIT INTERFACE TRACKING METHOD
cosity of continuous phase on the droplet size and the fre-
quency of the drop formation was studied [16]. The numeri- The interface is explicitly tracked either with a moving
cal results agree well with the analytical results. Lattice grid or massless markers seeded at the interface. Either a
Boltzmann method can also be used to numerically investi- moving grid or a moving grid coupled the fixed grid is em-
gate the dynamics of droplet formation in a cross-junction ployed. Thus, the explicit method includes boundary integral
configuration [17]. The interfacial tension between two im- method and front tracking method.
miscible phases is calculated using the continuum surface
force model (CSF). The simulation results of flow pattern, Boundary-Integral Method
droplet size and velocity vectors agree well with the experi- In the boundary integral method, the evolution of a de-
mental results. formed droplet is calculated by time integrating the fluid
velocity of a set of marker points on the interface [19]. The
Droplets Behavior marker velocities are obtained by solving a boundary-
The transport, merging, rupture, separating and colliding integral equation instead of solving the velocity field. Thus,
the interface is explicitly tracked. The flow solution is de-
of droplets have also been modeled numerically [18-24].
duced from the information of the discrete points along the
Zheng and Zhang (2000) used level-set method to simulate
interface. The details of boundary integral method were re-
the droplet spreading problems on the solid surface with so-
viewed by Pozrikidis [39]. Its application and development
lidification, which also involves a phase change process [25].
in multiphase fluid flow have been summarized in the litera-
Recently, the phenomenon of a droplet falling on the liquid ture [40]. Their review also discussed the numerical instabil-
film with the phase field method was calculated [26]. A ring- ity and the limited accuracy due to the involvement of the
shaped vortex was captured during the droplet merging. The interfacial tension. However, the presented theory only based
motion of three coupling phases i.e. particle encapsulated on a two-dimensional (2D) model. Actually, the previous 3D
droplets is readily to simulate with the level set method [27]. boundary integral method relying on fixed grid with uniform
marker point density and cannot resolve the extreme changes
Actuated Droplets of the interface during droplet breakup. To address this prob-
lem, an alternative method employing remeshing algorithms
The numerical methods also play an import role in the was proposed. These algorithms involve local mesh refine-
problems of droplet actuated by external forces. For exam- ment and reconnection. Another method is utilizing the adap-
ple, the droplet can be controlled by a magnetic field or a tive dicretization algorithm to resolve the interface to fit the
temperature field described by the Maxwell equation or the deformed drop shape [19]. The method was later refined
energy equation, respectively. The droplet can also be based on the minimization of the mesh configurational en-
squeezed or stretched by optical radiation pressure [46]. In ergy as the surface evolves [41]. The method was validated
the presence of a magnetic field, Korlie used a VOF model as accurate for test cases of drop breakup and coalescence.
to simulate the bubbles rising in a ferrofluid and magnetic Their work is followed by many papers with more 3D prob-
droplet falling down in a non-magnetic fluid [28]. A similar lems. The limit of numerical instability was overcome by the
problems were solved by Ki using level-set method [29]. method proposed [42]. A special mesh stabilization method
Boundary element method and the finite element method was utilized to solve the 3D extremely deformed interface.
were used to determine the equilibrium surface shapes of the Further developments showed that this method is very suc-
ferrofluid drop [30]. The volume of fluid method is applied cessful in solving single droplet breakup [43], droplet
to catch the shape of a droplet which was driven by an exter- squeezing among spheres [44], and optical stretching and
nal permanent magnet in a cylindrical computational domain squeezing of a sessile droplet [45].
[31]. Among thermally mediated actuation problems, many
available models were numerical studied such as the rising Front Tracking Methods
process of multiple bubbles [32] and droplet breakup behav-
ior [33]. The front tracking method was developed by Glimm’s
group and Tryggvason’s group. Glimm presented a 3D front
Moving Contact Line Problems tracking algorithm to solve Rayleigh-Taylor instability prob-
lems [46]. Unverdi and Tryggvason investigated the rising of
The droplet evolution on a solid surface involving a triple one and two bubbles [47]. This method needs connected
phase line (gas, liquid and solid contact line) problems can markers to reconstruct the interface grid dynamically in the
also be solved numerically. Under a shear flow, Spelt nu- calculation process. The front grid moves through the fixed
merically investigated a droplet rolling on a solid surface by grid giving the precise location and the geometry of the in-
adding the contact line velocity in the redistance function of terface. The interfacial tension is computed on the front and
level-set approach [34-35]. The level-set method coupled transfer to the fixed global grid. As discussed in [47], a sepa-
with the immersed interface method [36] or the volume-of- rate unstructured triangular grid was introduced to mark the
fluid method [37] can describe the motion of a sessile droplet interface position while the computational domain is discre-
on a planar surface. Ding and Spelt reported a good agree- tized by a regular fixed rectangular grid, Fig. (1). Naturally,
Numerical Simulation of Droplet-Based Microfluidics Micro and Nanosystems, 2010 Vol. 2, No. 3 3

the Navier-Stokes equations are solved on the fixed grid to reconstruct the interface, while the implicit method need
control the motion of the multiphase system. It assumed that resolve n-dimensional equation on the whole domain. The
the fluid properties are constant in each phase. The interfa- advantage of the implicit method is that it can solve more
cial jump condition is added to the momentum equations via completed problems with topological changes. They include
a discrete delta function along the smooth interface. This volume of fluid method, phase field method, lattice Boltz-
multi-grid method shows that the front-tracking method is mann method, and level set method.
complex. Furthermore, this method can not solve thin film
resulting from the droplet rupture. A topological change al- Volume of Fluid Method
gorithm near the interface needs to be considered. Problems
with topological changes involving breakup of drops and jet A review of volume of fluid method is given by Scar-
[48]. A pressure driven cell was modeled as a Newtonian dovelli and Zaleski [55]. Hirt and Nichols detailed the VOF
microdroplet with constant surface tension. The topological techniques for interface capturing in non-uniform mesh [56].
change of the interface was added by Torres and Brackbill In the VOF method, the volume fraction function is intro-
(2000) to solve the problem of coalescence of two spherical duced and defined in a volume fraction field throughout the
drops [23]. Esmaeeli and Tryggvason extended this approach whole computational domain. The volume fraction function
to cases with phase change [49, 50]. 


is + u • 
= 0 can be solved on a fixed-grid. Where

Recently, Mao showed with 2D conservative front- t
tracking method that the 2D discontinuity curves can be represents the volume fraction and has a value between 0 and
tracked in a 1D fashion [51]. The front tracking algorithm 1. Fig. (2a) shows a typical distribution of the volume frac-
can also be applied uniformly to N dimensions [52]. A de- tion values. The values are reconstructed to find the interface
veloped front-tracking method is advecting the unconnected location as shown in Fig. (2b). The interface lies within
points set [53]. The point set was used to mark the interface. computational cell with a volume fraction function of 0<

The three-dimensional deformation of a spherical drop was <1, and the location of interface is 0.5. For a volume fraction
modeled as the test case with topological changes. The im- function of unity, the computational cell is completely filled
provement of the front tracking method was studied in the with the primary phase. The computational cell is completely
volume conservation and multiphase problems [54].
filled with the secondary phase in the case of volume frac-
tion function of zero. To solve the function, the initial ap-
proximate interface position needs to be found. In addition,
interface reconstruction should be carried out to determine
the weighted density and viscosity for the computational
cells and compute the volume flux for the convective terms
in the governing equations. The crude reconstruction can
generate a large error even with a simple velocity field such
as translation or solid body rotation. The reconstruction
methods include but not limit to Simple Line or Piecewise
Linear Interface Construction (SLIC) [57-58], least-square fit
with split Eulerian-Lagrangian advection [59] and Parabolic
Reconstruction of Surface Tension (PROST) employing
piecewise parabolic curves [60]. Rudman introduced flux-
corrected transport (FCT) method for volume capturing and
did not use interface reconstruction [61].
After the interface reconstruction, the interface motion by
the underlying flow field must be modeled by a suitable ad-
vection algorithm. The discretization of advection term is the
main difficulty to guarantee the physical volume fraction
Fig. (1). The computational mesh. The 3D rectangular domain uses distribution and the sharpness of the interface. The lower
the uniform structure grid, and the interface uses a separate unstruc- order scheme can smear the interface due to the numerical
tured 2D grid. diffusion. A higher order scheme can result in numerical
oscillations. Gopala et al. reviewed and discussed the advan-
tages and limitations of several volume advection techniques
4. IMPLICIT INTERFACE CAPTURING METHOD
[62] such as flux-corrected transport algorithm [63], com-
The explicit method needs the stored points on the inter- pressive interface capturing scheme (CICSAM) [64], gamma
face in order to reconstruct the interface. The interface reso- differencing scheme [65] and Lagrangian piecewise-linear
lution depends on the number of points. In the implicit interface reconstruction (L-PLIC) method [66]. The com-
method, a simple function is defined on a fixed grid to cap- parison was based on two practical test cases of sloshing of a
ture the interface. The interface location is approximated by liquid wave and the Rayleigh–Taylor instability problem.
the indicator function. Thus, both methods can not tell us Recently, a high resolution differencing schemes, Normal-
where the exact interface location is. For n-dimensional ized Variable Diagram (NDV), was reported to preserve the
problems, the explicit method use n-1 dimensional grid to sharpness of interface and the boundedness of volume frac-
tion [67].
4 Micro and Nanosystems, 2010 Vol. 2, No. 3 Liu and Nguyen

(b)
(a)
Fig. (2). (a) Volume fraction values; (b) Contour of volume fraction.
=0.01,
=0.21,
=0.5, and
=0.88.

The advantage of the VOF method is its superior volume tion is governed by the Cahn-Hilliard equation [73], a fourth
or mass conservation of each fluid over other methods such order nonlinear parabolic diffusion equation, it is


as the level-set method. The interface could be captured im- t + (u ) =
[
f ( )]. The fnction f (
) is a polyno-
plicitly, thus topological changes can be handled automati-
cally. Problems with interfacial geometrical quantities such mial of
[74]. The method was used to model coalescent of
as the unit normal and the curvature can be encountered. two bubbles. Another alternative governing equation of the
These quantities are calculated from the volume fraction phase function is the Allen-Cahn type [75-76]. Because the
function which is nearly discontinuously distributed, and are phase-field method is a first order approximation to the sharp
important in the calculation of surface tension. It is not interface model, a fine grid is required around the interface.
straight forward to calculate these quantities accurately. For-
tunately, the level set method can fix this problem. As re- The advantages and disadvantages of phase-field method
sults, a number of researchers recently proposed a hybrid are discussed by Slavov and Dimova [9]. This method can be
algorithm of coupling the level-set function with the VOF easily extended to a three-dimensional problem, and has a
methods (CLSVOF) [68-70]. The VOF was used to recon- straight-forward extension to arbitrary multi-phase systems
struct the interface, while the level-set function is used to [77]. Moreover, the method can handle the topological
changes of the interface automatically. Adaptive moving
calculate the interface and curvature. However, these meth-
mesh algorithm can be use to increase the efficiency and
ods are not easy to extend to three dimensions. Therefore,
accuracy of the method [76]. This method requires an as-
NDV was used for the advection of the volume fraction
ymptotic analysis to be performed in order to obtain a map-
function, and extended the CLSVOF method to multi-
ping between the parameters of the phase field equations and
dimensional unstructured grid although the method is rela-
the sharp interface equations. As a result, the phase field
tively complex [67]. The three-dimensional shape of a rising
model only reproduces the dynamics of the sharp interface
bubble in a liquid was tested.
equations with the limitation that the density difference is
The VOF method can exhibit the high accuracy in the sufficiently small. Also a refined grid is needed to resolve
calculation of three dimensional flows. A pinching pendant the interface. In contrast, the level-set approach can be used
drop was modeled and displayed a good agreement with the to exactly locate the interface in a simple fashion.
experimental results [58]. The problems of bubble rising in Recently, application of the method include multiple
the liquid [70-71], droplets adhering to a vertical wall [67], pinch-offs of a long cylindrical thread at small Reynolds
and a water drop impact onto a deep water pool with wave number [78], droplet pinch-off in liquid/liquid jet configura-
breaking and plunging [72] were modeled successfully. The
tion [79], and two drops coalescence evolving four phases
commercial software Fluent also employed VOF method to
[80].
calculate multi-phase flow.
Phase-Field Method Lattice Boltzmann Method
One of the implicit methods for capturing the interface is The lattice Boltzmann method (LBM) is a relatively new
phase-field method, which is also called the diffuse-interface simulation technique. The difference between phase field
method. In phase field method, a phase function
( x, t ) is method and LBM only lies in the discretization to the Na-
introduced to represent the interface
= 0 . Away from the vier-Stokes equations [26]. The method is successful in deal-
ing with the interface tension and the complex boundaries
interface, the function values of two phases are
= +1 and with a large density ratio between the two phases. The lattice
 =
1 , respectively. The evolution of the phase-field func- Boltzmann method constructs the kinetic equation of the
Numerical Simulation of Droplet-Based Microfluidics Micro and Nanosystems, 2010 Vol. 2, No. 3 5

discrete-velocity distribution function utilizing discrete lat- signs distinguish the two different fluids. Fig. (3) shows the
tice and discrete time. Thus, it is viewed as a special finite typical contour of the level-set function. The values on the
difference scheme for the kinetic equation. But LBM distin- curves are the distance from the interface. By taking the time


guishes itself from the conventional numerical schemes derivative of
( x , t ) = 0 , the convection equation (level set
which are based on the discritization of the continuum equa-

tions. The LBM is a derivative of the lattice gas theory. A equation)
t + u  
= 0 is used for calculating the motion
lattice gas theory consists of a regular lattice with particles of the interface. Fluid properties can be calculated from the
residing on the nodes. As a result, the N-S equation can be level-set function through a smoothed Heaviside function.
derived from the lattice Boltzmann equation. The macro-
Currently, LSM has been used successfully for handling
scopic dynamics of fluid is the behavior of the particles in
moving interface and free boundary problems in microflu-
the system.
idics [87]. Topological changes can be captured in a straight-
An overview of the LBM was presented and discussed by forward fashion, and thus the methods are readily imple-
Chen and Doolen [81]. This review discussed the different mented in both two and three spatial dimensions. We can
methods for modeling interfacial tension in a multiphase conclude that LSM is a powerful tool to solve the compli-
fluid flow: the free energy model and the interparticle inter- cated deformation and to construction of implicit surfaces on
action potential model. Lamura employed the first approach a fixed rectangular grid. The challenging problems such as
to simulate oil-water-surfactant mixtures [82]. The latter coalescence/rupture of droplet and droplet falling down un-
method was first proposed by Shan and Chen [83]. Except der difficult conditions can be simulated easily with the level
for above methods, another one is the He-Shan-Doolen ap- set method.
proach [84]. Its limits and advantages were discussed by
Numerically, it is desirable to keep the level set function
comparing with front tracking method based on the 2D bub-
close to a signed distance function to the interface. However,
ble rising model [85]. The results showed that the BLM is
the fact is the level set function will generally deviate from a
not appropriate for low Mach number. Recently, LBM
signed distance function. Moreover, the discretization of the
method becomes a popular modeling platform for droplet-
level set equation can cause numerical dissipation and inac-
based microfluidics. The droplet formation process at a mi-
curate identification of interfaces. Flat or steep regions often
crofluidic T-junction [14-15] and a microfluidic cross-
occur near the interface. Generally, a reinitialization proce-
junction [17] was investigated with LBM.
dure is needed to reset the level set function to be a signed
distance function to the interface with some degree of accu-
Level-Set Method
racy [21].
Osher and Sethian demonstrated a new algorithm of level Normally, the level set and the reinitialization equations
set method to track the moving interface on the fixed-grid

are evolved numerically by higher order advection schemes.
system [86]. The level set function
( x , t ) was introduced They are two examples of a general Hamilton-Jacobi (HJ)
over the whole domain or near the interface and was defined equation
t + H (
) = 0 . The detailed solution of this
as a signed shortest normal distance from the interface. The equation is well documented in the book by Osher and Fed-
interface is implicitly represented by the zero level-set,


kiw [88], which also include other higher order numerical

( x , t ) = 0 . The values of
( x , t ) are positive inside the discretiztion methods such as essentially non-oscillatory
interface, and negative outside the interface. The different (ENO) and weighted essentially nonoscillatory (WENO).
The gain in accuracy of on the higher order schemes is com-
promised by the long time integration, which is not desirable.
Therefore, new techniques need to be developed to reduce
the computational time. Solving the level set function in the
global domain wastes computing resources and times, be-
cause only the domain near the interface is of interest. An
approach was introduced to only solve the level-set function


within a band
(x ) <  near the interface, where
is width
of the interface [89]. This method does not compromise the
accuracy since the level-set value is important only around
the interface. Fig. (4) shows the marked region of the inter-
face, the computational domain of the level set function and
the reinitialize function. This tremendous time saving can
reduce one order of computational effort.
Even using the higher-order advection schemes to evolve
the level-set equation, the mass error still cannot be avoided.
Fig. (5) depicts the typical examples of mass error. The error
is caused because the sharpness of the interface is destroyed
by under-resolved regions during the discretisation process.
This mass conservation problem can be solved by many
Fig. (3). The contour of level set function.
methods. One method employs the re-initialization equation
6 Micro and Nanosystems, 2010 Vol. 2, No. 3 Liu and Nguyen

interface where the level set method failed to preserve mass.

Traditionally, the particles are placed throughout the fluid
domain and move with the local velocity [92-93]. This
method is successful in solving the free surface problems i.e.
falling free surface, splashing of water, and water sloshing in
a tank. Later, an approach of seeding Lagrangian marker
particles within a narrow band around the interface was pro-
posed [94]. First, the initial particles seeded randomly into
the cells. Then these particles are attracted to its correspond-
ing region after a simple attraction procedure. While the
negative particles are attracted to the
< 0 region, the posi-
tive particles are attracted to the
> 0 region as shown in the
example of Fig. (6). It is necessary to perform particle cor-
rection strategies to prevent that the particles escape across
the interface [94-96]. The last procedure to be carried out is
particle reseeding, when there are not enough particles in
some region to resolve the interface well after few iterations
[95]. Fig. (7) shows one case of the particle poor distribution
Fig. (4). The band of computational domain for level set equation
around the interface. However, reseeding algorithm is not a
and reinitialization equation.
perfect method to smooth the interface. It is necessary to
randomly seed and attract the particles repeatedly when the
as mentioned above. Another different reinitialization proce- reseeding algorithm failed to maintain the smoothness of the
( )
dure is to solve the equation t + A
P +   = 0 to the
interface.
Recently, level-set methods have been used quite suc-
steady state [90] where
A is the difference of the mass be- cessfully in moving interface as well as multiphase flow in
tween initial mass and the mass in the droplet deforming two- or three-dimensional problems: particle-encapsulated
process. P is constant, and  is curvature. Unfortunately, this droplet transporting in the fluid flow [87], droplet spreading
improvement in the mass conservation cannot solve the mass and solidification [25], bubble rising and growth in a station-
imbalance problem of the level set method [90]. Therefore, a ary fluid [97], droplet falling down[98], and bubble adhering
global mass correction technique is needed [27, 91]. A cor- to the solid surface in the fluid [99], topological changes of
rection is added to the level-set equation in order to preserve the interface such as droplet pinching and connection [100],
the initial mass. Mass is conserved in a global sense. Actu- droplet collision and membrane break [101], breakup into
ally, most researchers agree that employing the particle smaller droplets [102]. In the field of the droplet based mi-
level-set method can improve mass conservation. A number crofluidics, numerical simulation with level-set method has
of particles are seeded into the grid cell to track the charac- the potential to play an important role.
teristic information, and thus can accurately reconstruct the

Fig. (5). Mass error in the level set method: (a) mass loss; (b) mass surplus.
Numerical Simulation of Droplet-Based Microfluidics Micro and Nanosystems, 2010 Vol. 2, No. 3 7

tal microfluidic circuits. J. Microelectromech. Syst., 2003, 12(1),

70-80.
[3] Fidalgo, L.M.; Abell, C.; Huck, W.T.S. Surface-induced droplet
fusion in microfluidic devices. Lab Chip, 2007, 7(8), 984-986.
[4] Lehmann, U.; Hadjidj, S.; Parashar, V.K.; Vandevyver, C.; Rida,
A.; Gijs, M.A.M. Two-dimensional magnetic manipulation of mi-
crodroplets on a chip as a platform for bioanalytical applications.
Sens. Actuators, B, 2006, 117(2), 457-463.
[5] Garstecki, P.; Fuerstman, M.J.; Stone, H.A.; Whitesides, G.M.
Formation of droplets and bubbles in a microfluidic T-junction -
Scaling and mechanism of break-up. Lab Chip, 2006, 6(3), 437-
446.
[6] Xu, J.H.; Li, S.W.; Tán, J.; Wang, Y.J.; Luo, G.S. Preparation of
highly monodisperse droplet in a T-junction microfluidic device.
AIChE J., 2006, 52(9), 3005-3010.
[7] Nguyen, N.T.; Ting, T.H.; Yap, Y.F.; Wong, T.N.; Chai, J.C.K.;
Ong, W.L.; Zhou, J.; Tan, S.H.; Yobas, L. Thermally mediated
droplet formation in microchannels. Appl. Phys. Lett., 2007, 91(8).
[8] Xu, J.H.; Li, S.W.; Lan, W.J.; Luo, G.S. Microfluidic approach for
rapid interfacial tension measurement. Langmuir, 2008, 24(19),
11287-11292.
[9] Slavov, V.; Dimova, S. Phase-field versus level set method for 2D
dendritic growth. In, 2007.
[10] Cristini, V.; Tan, Y.C. Theory and numerical simulation of droplet
dynamics in complex flows - A review. Lab Chip, 2004, 4(4), 257-
Fig. (6). Some of the particles with radius assigned after attraction
264.
procedure. [11] Yap, Y.F. Numerical modeling of multiphase flows in microchan-
nels. PhD Thesis, Nanyang Technological University: Singapore,
2007.
[12] Garrioch, S.H.; Baliga, B.R. A PLIC volume tracking method for
the simulation of two-fluid flows. Int. J. Numer. Methods Fluids,
2006, 52(10), 1093-1134.
[13] De Menech, M. Modeling of droplet breakup in a microfluidic T-
shaped junction with a phase-field model. Phys. Rev. E - Statistic.
Nonlinear, Soft Matter Phys., 2006, 73(3), 1-9.
[14] Gupta, A.; Kumar, R. Effect of geometry on droplet formation in
the squeezing regime in a microfluidic T-junction. Microfluid.
Nanofluid., 2010, 8(6), 799-812.
[15] Van Der Graaf, S.; Nisisako, T.; Schroën, C.G.P.H.; Van Der
Sman, R.G.M.; Boom, R.M. Lattice Boltzmann simulations of
droplet formation in a T-shaped microchannel. Langmuir, 2006,
Fig. (7). Particle reseeding. Gray ones are positive particles, and 22(9), 4144-4152.
[16] Sang, L.; Hong, Y.; Wang, F. Investigation of viscosity effect on
black ones are negative particles. droplet formation in T-shaped microchannels by numerical and
analytical methods. Microfluid. Nanofluid., 2009, 6(5), 621-635.
CONCLUSION [17] Wu, L.; Tsutahara, M.; Kim, L.S.; Ha, M. Three-dimensional lat-
tice Boltzmann simulations of droplet formation in a cross-junction
microchannel. Int. J. Multiphase Flow, 2008, 34(9), 852-864.
This paper reviews the theory and applications of the dif- [18] Coulliette, C.; Pozrikidis, C. Motion of an array of drops through a
ferent interface tracking and capturing methods for droplet- cylindrical tube. J. Fluid Mech., 1998, 358, 1-28.
based microfluidics. In a multipase fluid flow, it is difficult [19] Cristini, V.; B
awzdziewicz, J.; Loewenberg, M. Drop breakup in
to numerically simulate the interface because of the jump of three-dimensional viscous flows. Phys. Fluids, 1998, 10(8), 1781-
properties and the moving interface. Moreover, the capillary 1783.
[20] Quan, S.; Lou, J.; Schmidt, D.P. Modeling merging and breakup in
force, shear stress, interfacial tension, or other external the moving mesh interface tracking method for multiphase flow
forces are strongly coupled. Thus, the interface movement simulations. J. Comput. Phys., 2009, 228(7), 2660-2675.
needs to be linked with the discretization of the velocity field [21] Sussman, M. A level set approach for computing solutions to in-
and the properties of fluids to model challenging problems compressible two-phase flow. J. Comput. Phys., 1994, 114(1), 146-
such as droplet breakup and coalescence. We reviewed here 159.
[22] Sussman, M.; Fatemi, E. Efficient, interface-preserving level set
the successful methods in handling the complicated interface redistancing algorithm and its application to interfacial incom-
deformation. The review would help the readers to have an pressible fluid flow. SIAM J. Sci. Comput., 1999, 20(4), 1165-1191.
overview on the available techniques for modeling droplet- [23] Torres, D.J.; Brackbill, J.U. The point-set method: front-tracking
based microfluidics. without connectivity. J. Comput. Phys., 2000, 165(2), 620-644.
[24] Zhang, H.; Zheng, L.L.; Prasad, V.; Hou, T.Y. A curvilinear level
set formulation for highly deformable free surface problems with
REFERENCES application to solidification. Numer. Heat Transfer, Part B, 1998,
34(1), 1-20.
[1] Burns, M.A.; Mastrangelo, C.H.; Sammarco, T.S.; Man, F.P.; Web- [25] Zheng, L.L.; Zhang, H. An adaptive level set method for moving-
ster, J.R.; Johnson, B.N.; Foerster, B.; Jones, D.; Fields, Y.; Kaiser, boundary problems: Application to droplet spreading and solidifi-
A.R.; Burke, D.T. Microfabricated structures for integrated DNA cation. Numer. Heat Transfer, Part B, 2000, 37(1-4), 437-454.
analysis. Proc. Nat. Acad. Sci. U.S.A., 1996, 93(11), 5556-5561. [26] Takada, N.; Misawa, M.; Tomiyama, A. A phase-field method for
[2] Cho, S.K.; Moon, H.; Kim, C.J. Creating, transporting, cutting, and interface-tracking simulation of two-phase flows. Math. Comput.
merging liquid droplets by electrowetting-based actuation for digi- Simul., 2006, 72(2-6), 220-226.
8 Micro and Nanosystems, 2010 Vol. 2, No. 3 Liu and Nguyen

[27] Yap, Y.F.; Chai, J.C.; Wong, T.N.; Toh, K.C.; Zhang, H.Y. A method in N-dimensions. J. Sci. Comput., 2007, 31(1-2), 213-236.
global mass correction scheme for the level-set method. Numer. [53] Gois, J.P.; Nakano, A.; Nonato, L.G.; Buscaglia, G.C. Front track-
Heat Transfer, Part B, 2006, 50(5), 455-472. ing with moving-least-squares surfaces. J. Comput. Phys., 2008,
[28] Korlie, M.S.; Mukherjee, A.; Nita, B.G.; Stevens, J.G.; Trubatch, 227(22), 9643-9669.
A.D.; Yecko, P. Modeling bubbles and droplets in magnetic fluids. [54] Dijkhuizen, W.; Roghair, I.; Annaland, M.V.S.; Kuipers, J.A.M.
J. Phys. Condens. Matter, 2008, 20(20). DNS of gas bubbles behaviour using an improved 3D front tracking
[29] Ki, H. Level set method for two-phase incompressible flows under model-Model development. Chem. Eng. Sci., 2010, 65(4), 1427-
magnetic fields. Comput. Phys. Commun., 2010, 181(6), 999-1007. 1437.
[30] Lavrova, O.; Matthies, G.; Mitkova, T.; Polevikov, V.; Tobiska, L. [55] Scardovelli, R.; Zaleski, S. Direct numerical simulation of free-
Numerical treatment of free surface problems in ferrohydrodynam- surface and interfacial flow. Annu. Rev. Fluid Mech., 1999, 31,
ics. J. Phys. Condens. Matter, 2006, 18(38), S2657-S2669. 567-603.
[31] Afkhami, S.; Renardy, Y.; Renardy, M.; Riffle, J.S.; St Perre, T. [56] Hirt, C.W.; Nichols, B.D. Volume of fluid (VOF) method for the
Field-induced motion of ferrofluid droplets through immiscible dynamics of free boundaries. J. Comput. Phys., 1981, 39(1), 201-
viscous media. J. Fluid Mech., 2008, 610, 363-380. 225.
[32] Tryggvason, G.; Bunner, B.; Esmaeeli, A.; Juric, D.; Al-Rawahi, [57] Noh, W.; Woodward, P. SLIC-simple line interface calculation.
N.; Tauber, W.; Han, J.; Nas, S.; Jan, Y.J. A front-tracking method International Conference on Fluid Dynamics, 1976, 59, 330-340.
for the computations of multiphase flow. J. Comput. Phys., 2001, [58] Gueyffier, D.; Li, J.; Nadim, A.; Scardovelli, R.; Zaleski, S. Vol-
169(2), 708-759. ume-of-fluid interface tracking with smoothed surface stress meth-
[33] Yap, Y.F.; Tan, S.H.; Nguyen, N.T.; Murshed, S.M.S.; Wong, ods for three-dimensional flows. J. Comput. Phys., 1999, 152(2),
T.N.; Yobas, L. Thermally mediated control of liquid microdroplets 423-456.
at a bifurcation. J. Phys. D: Appl. Phys., 2009, 42(6). [59] Scardovelli, R.; Zaleski, S. Interface reconstruction with least-
[34] Spelt, P.D.M. A level-set approach for simulations of flows with square fit and split Eulerian-Lagrangian advection. Int. J. Numer.
multiple moving contact lines with hysteresis. J. Comput. Phys., Methods Fluids, 2003, 41(3), 251-274.
2005, 207(2), 389-404. [60] Renardy, Y.; Renardy, M. PROST: A parabolic reconstruction of
[35] Spelt, P.D.M. Shear flow past two-dimensional droplets pinned or surface tension for the volume-of-fluid method. J. Comput. Phys.,
moving on an adhering channel wall at moderate Reynolds num- 2002, 183(2), 400-421.
bers: A numerical study. J. Fluid Mech., 2006, 561, 439-463. [61] Rudman, M. Volume-tracking methods for interfacial flow calcula-
[36] Hunter, J.K.; Li, Z.; Zhao, H. Reactive autophobic spreading of tions. Int. J. Numer. Methods Fluids, 1997, 24(7), 671-691.
drops. J. Comput. Phys., 2002, 183(2), 335-366. [62] Gopala, V.R.; van Wachem, B.G.M. Volume of fluid methods for
[37] Yokoi, K.; Vadillo, D.; Hinch, J.; Hutchings, I. Numerical studies immiscible-fluid and free-surface flows. Chem. Eng. J., 2008,
of the influence of the dynamic contact angle on a droplet impact- 141(1-3), 204-221.
ing on a dry surface. Phys. Fluids, 2009, 21(7). [63] Boris, J.P.; Book, D.L. Flux-corrected transport. I. SHASTA, a
[38] Ding, H.; Spelt, P.D.M. Inertial effects in droplet spreading: A fluid transport algorithm that works. J. Comput. Phys., 1973, 11(1),
comparison between diffuse-interface and level-set simulations. J. 38-69.
Fluid Mech., 2007, 576, 287-296. [64] Ubbink, O.; Issa, R.I. A Method for Capturing Sharp Fluid Inter-
[39] Pozrikidis, C. Interfacial dynamics for Stokes flow. J. Comput. faces on Arbitrary Meshes. J. Comput. Phys., 1999, 153(1), 26-50.
Phys., 2001, 169(2), 250-301. [65] Jasak, H.; Weller, H.G.; Gosman, A.D. High resolution NVD dif-
[40] Hou, T.Y.; Lowengrub, J.S.; Shelley, M.J. Boundary integral ferencing scheme for arbitrarily unstructured meshes. Int. J. Numer.
methods for multicomponent fluids and multiphase materials. J. Methods Fluids, 1999, 31(2), 431-449.
Comput. Phys., 2001, 169(2), 302-362. [66] Van Wachem, B.G.M.; Schouten, J.C. Experimental validation of
[41] Cristini, V.; Blawzdziewicz, J.; Loewenberg, M. An adaptive mesh 3-D Lagrangian VOF model: Bubble shape and rise velocity.
algorithm for evolving surfaces: simulations of drop breakup and AIChE J., 2002, 48(12), 2744-2753.
coalescence. J. Comput. Phys., 2001, 168(2), 445-463. [67] Park, I.R.; Kim, K.S.; Kim, J.; Van, S.H. A volume-of-fluid
[42] Zinchenko, A.Z.; Rother, M.A.; Davis, R.H. Cusping, capture, and method for incompressible free surface flows. Int. J. Numer. Meth-
breakup of interacting drops by a curvatureless boundary-integral ods Fluids, 2009, 61(12), 1331-1362.
algorithm. J. Fluid Mech., 1999, 391, 249-292. [68] Son, G. Efficient implementation of a coupled level-set and vol-
[43] Janssen, P.A.; Anderson, P.D. Boundary-integral method for drop ume-of-fluid method for three-dimensional incompressible two-
deformation between parallel plates. Phys. Fluids, 2007, 19(4). phase flows. Numer. Heat Transfer, Part B, 2003, 43(6), 549-565.
[44] Zinchenko, A.Z.; Davis, R.H. Boundary-integral algorithm for [69] Sussman, M.; Puckett, E.G. A coupled level set and volume-of-
drop squeezing through a granular material. In: 9th International fluid method for computing 3D and axisymmetric incompressible
Conference on Multiscale and Functionally Graded Materials, two-phase flows. J. Comput. Phys., 2000, 162(2), 301-337.
FGM IX, Oahu Island, HI, October 15-18, 2008. [70] Sussman, M.; Smith, K.M.; Hussaini, M.Y.; Ohta, M.; Zhi-Wei, R.
[45] Chraïbi, H.; Lasseux, D.; Arquis, E.; Wunenburger, R.; Delville, A sharp interface method for incompressible two-phase flows. J.
J.P. Stretching and squeezing of sessile dielectric drops by the opti- Comput. Phys., 2007, 221(2), 469-505.
cal radiation pressure. Phys. Rev. E - Statistic. Nonlinear, Soft Mat- [71] van der Pijl, S.P.; Segal, A.; Vuik, C.; Wesseling, P. A mass-
ter Phys., 2008, 77(6). conserving Level-Set method for modelling of multi-phase flows.
[46] Glimm, J.; Grove, J.W.; Li, X.L.; Shyue, K.M.; Zeng, Y.; Zhang, Int. J. Numer. Methods Fluids, 2005, 47(4), 339-361.
Q. Three-dimensional front tracking. SIAM J. Sci. Comput., 1998, [72] Wang, Z.; Yang, J.; Koo, B.; Stern, F. A coupled level set and
19(3), 703-727. volume-of-fluid method for sharp interface simulation of plunging
[47] Unverdi, S.O. A front-tracking method for viscous, incompressible, breaking waves. Int. J. Multiphase Flow, 2009, 35(3), 227-246.
multifluid flows. J. Comput. Phys., 1992, 100(1), 25-37. [73] Cahn, J.W.; Hilliard, J.E. Free energy of a nonuniform system. III.
[48] Agresar, G.; Linderman, J.J.; Tryggvason, G.; Powell, K.G. An Nucleation in a two-component incompressible fluid. J. Chem.
adaptive, cartesian, front-tracking method for the motion, deforma- Phys., 1959, 31(3), 688-699.
tion and adhesion of circulating cells. J. Comput. Phys., 1998, [74] Liu, C.; Shen, J. A phase field model for the mixture of two incom-
143(2), 346-380. pressible fluids and its approximation by a Fourier-spectral method.
[49] Esmaeeli, A.; Tryggvason, G. Computations of explosive boiling in Physica D, 2003, 179(3-4), 211-228.
microgravity. J. Sci. Comput., 2003, 19(1-3), 163-182. [75] Liu, C.; Walkington, N.J. An eulerian description of fluids contain-
[50] Esmaeeli, A.; Tryggvason, G. Computations of film boiling. Part I: ing visco-elastic particles. Arch. Ration. Mech. Anal., 2001, 159(3),
Numerical method. Int. J. Heat Mass Transfer, 2004, 47(25), 5451- 229-252.
5461. [76] Zhang, Z.; Tang, H. An adaptive phase field method for the mix-
[51] Mao, D.k. Towards front-tracking based on conservation in two ture of two incompressible fluids. Comput. Fluids, 2007, 36(8),
space dimensions II, tracking discontinuities in capturing fashion. 1307-1318.
J. Comput. Phys., 2007, 226(2), 1550-1588. [77] Kim, J.; Kang, K.; Lowengrub, J. Conservative multigrid methods
[52] Liu, J.; Lim, H.K.; Glimm, J.; Li, X. A conservative front tracking for Cahn-Hilliard fluids. J. Comput. Phys., 2004, 193(2), 511-543.
 Numerical Simulation of Droplet-Based Microfluidics Micro and Nanosystems, 2010 Vol. 2, No. 3 9

[78] Kim, J. A diffuse-interface model for axisymmetric immiscible [91] Lakehal, D.; Meier, M.; Fulgosi, M. Interface tracking towards the
two-phase flow. Appl. Math. Comput., 2005, 160(2), 589-606. direct simulation of heat and mass transfer in multiphase flows. Int.
[79] Kim, C.H.; Shin, S.H.; Lee, H.G.; Kim, J. Phase-field model for the J. Heat Fluid Flow, 2002, 23(3), 242-257.
pinchoff of liquid-liquid jets. J. Korean Phys. Soc., 2009, 55(4), [92] Harlow, F.H.; Welch, J.E. Numerical calculation of time-dependent
1451-1460. viscous incompressible flow of fluid with free surface. Phys. Flu-
[80] Kim, J. A generalized continuous surface tension force formulation ids, 1965, 8(12), 2182-2189.
for phase-field models for multi-component immiscible fluid flows. [93] Chen, S.; Johnson, D.B.; Raad, P.E.; Fadda, D. The surface marker
Comput. Meth. Appl. Mech. Eng., 2009, 198(37-40), 3105-3112. and micro cell method. Int. J. Numer. Methods Fluids, 1997, 25(7),
[81] Chen, S.; Doolen, G.D. Lattice Boltzmann model for fluid flows. 749-778.
Annu. Rev. Fluid. Mech., 1998, 30, 329-364. [94] Enright, D.; Fedkiw, R.; Ferziger, J.; Mitchell, I. A hybrid particle
[82] Lamura, A.; Gonnella, G.; Yeomans, J.M. A lattice Boltzmann level set method for improved interface capturing. J. Comput.
model of ternary fluid mixtures. Europhys. Lett., 1999, 45(3), 314- Phys., 2002, 183(1), 83-116.
320. [95] Wang, Z.; Yang, J.; Stern, F. An improved particle correction pro-
[83] Shan, X.; Chen, H. Lattice Boltzmann model for simulating flows cedure for the particle level set method. J. Comput. Phys., 2009,
with multiple phases and components. Phys. Rev. E, 1993, 47(3), 228(16), 5819-5837.
1815-1819. [96] Tran, L.B.; Udaykumar, H.S. A particle-level set-based sharp inter-
[84] Nourgaliev, R.R.; Dinh, T.N.; Theofanous, T.G.; Joseph, D. The face cartesian grid method for impact, penetration, and void col-
lattice Boltzmann equation method: Theoretical interpretation, nu- lapse. J. Comput. Phys., 2004, 193(2), 469-510.
merics and implications. Int. J. Multiphase Flow, 2003, 29(1), 117- [97] Son, G. A numerical method for incompressible two-phase flows
169. with open or periodic boundaries. Numer. Heat Transfer, Part B,
[85] Sankaranarayanan, K.; Kevrekidis, I.G.; Sundaresan, S.; Lu, J.; 2001, 39(1), 45-60.
Tryggvason, G. A comparative study of lattice Boltzman and front- [98] Ni, M.J.; Abdou, M.; Komori, S. A variable-density projection
trancking finite-difference methods for bubble simulations. Int. J. method for interfacial flows. Numer. Heat Transfer, Part B, 2003,
Multiphase Flow, 2003, 29(1), 109-116. 44(6), 553-574.
[86] Osher, S.; Sethian, J.A. Fronts propagating with curvature- [99] Son, G. A level set method for incompressible two-fluid flows with
dependent speed: Algorithms based on Hamilton-Jacobi formula- immersed solid boundaries. Numer. Heat Transfer, Part B, 2005,
tions. J. Comput. Phys., 1988, 79(1), 12-49. 47(5), 473-489.
[87] Yap, Y.F.; Chai, J.C.; Wong, T.N.; Nguyen, N.T.; Toh, K.C.; [100] Nochetto, R.H.; Walker, S.W. A hybrid variational front tracking-
Zhang, H.Y.; Yobas, L. A procedure for the motion of particle- level set mesh generator for problems exhibiting large deformations
encapsulated droplets in microchannels. Numer. Heat Transfer, and topological changes. J. Comput. Phys., 2010, 229(18), 6243-
Part B, 2008, 53(1), 59-74. 6269.
[88] Osher, S.; Fedkiw, R.P. Level set methods and dynamic implicit [101] Tanguy, S.; Berlemont, A. Application of a level set method for
surfaces. Springer-Verlag: New York, 2003. simulation of droplet collisions. Int. J. Multiphase Flow, 2005,
[89] Peng, D.; Merriman, B.; Osher, S.; Zhao, H.; Kang, M. A PDE- 31(9), 1015-1035.
Based Fast Local Level Set Method. J. Comput. Phys., 1999, [102] Croce, R.; Griebel, M.; Schweitzer, M.A. Numerical simulation of
155(2), 410-438. bubble and droplet deformation by a level set approach with sur-
[90] Chang, Y.C.; Hou, T.Y.; Merriman, B.; Osher, S. A level set for- face tension in three dimensions. Int. J. Numer. Methods Fluids,
mulation of Eulerian interface capturing methods for incompressi- 2010, 62(9), 963-993.
ble fluid flows. J. Comput. Phys., 1996, 124(2), 449-464.

View publication stats