Content uploaded by Vittorio Cristini
Author content
All content in this area was uploaded by Vittorio Cristini
Content may be subject to copyright.
Theory and numerical simulation of droplet dynamics in
complex flows—a review†
Vittorio Cristini‡ and Yung-Chieh Tan
Department of Biomedical Engineering, University of California, Irvine, REC 204, Irvine, CA
92697-2715, USA. E-mail: cristini@math.uci.edu; Fax: 949 824 1727; Tel: 949 824 9132
Received 3rd March 2004, Accepted 12th May 2004
First published as an Advance Article on the web 1st July 2004
We review theoretical and numerical studies and methods for droplet deformation, breakup and coalescence in flows
relevant to the design of micro channels for droplet generation and manipulation.
Introduction
Microscopic droplet formation and manipulation are key processes
(e.g., see ref. 1) in DNA analysis,
2–4
protein crystallization,
5,6
analysis of human physiological fluids,
7
and cell encapsulation.
8
Droplets are often used to either meter or mix small volumes of
fluids to improve significantly the mixing efficiency and allow
more complexity in chemical processing. For example, nanoliters
of reagents can be added (through droplet coalescence) and
removed (breakup) from a droplet leading to desired chemical
reaction rates and times. Droplet generation and manipulation using
immiscible flows,
9–18
has the advantage that generation is fast and
provides flexibility for the amount of reagent in each droplet, and
the cost of the micro channels is low, compared to electrowetting-
based devices. In addition, repeated fission and fusion of a single
stream of droplets at a bifurcating junction can be used to produce
parallel streams of smaller droplets containing a fraction of the
original chemical concentrations. Reactions of large volumes of
solution can be rapidly analyzed by splitting the solution into tens
or hundreds of smaller droplets.
10
This method can be applied to
studying protein conformations,
19
determining reaction kinetics,
20
and characterizing single molecule polymerase chain reactions.
21
For droplet systems controlled by external flows, the design of
the channel geometry is used to control the forces that create,
transport, split and fuse droplets. To our knowledge, droplet
generation in microfluidic circuits using flow was first demon-
strated (Fig. 1) by Thorsen et al.
14
Song et al.
11
were the first to
demonstrate droplet fission in microfluidic channels, and to show
that the channel walls, while significantly hindering droplet motion,
can improve the chemical mixing efficiencies by rotating the
droplets. Tan et al.
10
first demonstrated sorting and separation of
satellite droplets from droplets of larger sizes in microfluidics by
controlling the shear gradient at the bifurcating junction (Fig. 2),
† Lab on a Chip special issue: The Science and Application of Droplets in
Microfluidic Devices.
‡ Also at the Department of Mathematics.
Vittorio Cristini is currently an assistant professor of Bio-
medical Engineering and Mathematics at the University of
California, Irvine. He is also a member of the Chao Family
Comprehensive Cancer Center there. He earned his Ph.D.
degree in Chemical Engineering (2000) from Yale University,
where he also received a M.S. and a M.Phil. He earned his
degree of Dottore in Ingegneria Nucleare (Nuclear Engineer-
ing) from the University of Rome, Italy in 1994. Dr Cristini is an
expert in the field of multiphase flows and droplet and bubble
dynamics, where he has pub-
lished about twenty papers. For
his important work on droplet
breakup in laminar and turbu-
lent flows, Dr Cristini was the
first recipient of the “Andreas
Acrivos Dissertation Award in
Fluid Dynamics” from the
American Physical Society—
Division of Fluid Dynamics. Dr
Cristini has also published sev-
eral papers in the fields of
crystal and tumor growth.
Fig. 1 Control of droplet sizes by the pressure of the water-in-oil flow
(Fig. 4 from ref. 14 reproduced with permission from the American Physical
Society).
Fig. 2 Creation, sorting and collection of satellite droplets. (a) Satellite
droplets are created with larger primary droplets. (b) Primary droplets are
sorted towards the lower daughter channel while satellite droplets move into
the loop region of the sorter. (c) The channel collecting the satellite droplets
is free of primary droplets (data from ref. 10).
This journal is © The Royal Society of Chemistry 2004
MINIATURISATION FOR CHEMISTRY, BIOLOGY & BIOENGINEERING
DOI: 10.1039/b403226h
257Lab Chip, 2004, 4 , 257–264
and droplet fusion control using droplet focusing designs over a
variety of droplet sizes, numbers and generation speeds (Fig. 3).
Control of both the droplet generation speed
9
and size
9,16,18
has
been achieved in novel microfluidic designs (Figs. 4, 5 and 6). Tan
et al.
9
were successful in generating monodispersed pico- to
femtoliter sized droplets in a PDMS based microfluidic device by
controlling both the magnitude and type of flow in the micro
channels. Flow magnitude and type are controlled by varying the
difference between oil flow rates
9
and the difference between oil
and water flow rates,
9,16
in co-flowing streams at the droplet
generation site, and by varying the geometry and thus the flow
resistance of the channels at various junctions downstream.
10,18
Both designs lead to effectively varying the amount of straining of
the flow and thus provide finer control of the droplet sizes.
To control and optimize the creation, mixing, sorting, fusion and
splitting of droplets, droplet dynamics, and droplet breakup and
coalescence criteria and rates in 3-D flow and channel geometry
(e.g., in channel junctions and in non symmetric flow-focusing
streams) must be understood and controlled. The relative fluid
motions induced by the droplets, the detailed characteristics of the
droplet–solid contact angles, the interfacial tension at the droplet–
liquid interface, the flow induced by the movement of the droplets,
and the shear force generated by the carrier flow around the droplets
affect the design. The current design is often empirical and not
founded on a rigorous study of the flow characteristics. Thus, there
is room for considerable improvements by using theory and direct
flow calculation.
Breakup and coalescence transitions are the result of the complex
interplay of viscous, inertial, capillary, Marangoni, electrostatic
and van der Waals forces over a wide variety of spatial and
timescales. Asymptotic (e.g. lubrication) theories describe the
limiting behaviors (e.g. pinching or film-drainage rates) of the
geometrical and field variables at the onset of these transitions
based on simplified assumptions on geometry (e.g. one-dimension-
ality), fluid and flow conditions. Numerical methods are capable of
describing transitions accurately and efficiently in simulations of a
variety of flow geometries (2-D and 3-D) and conditions where
interfaces deform significantly in principle without relying on
simplifying assumptions. A weakness of numerical simulations is
computational expense because the transition regions characterized
by small length- (and time-) scales need to be resolved by the
numerical discretization requiring a very large number of computa-
tional elements. Thus it is often practice to incorporate asymptotic
theories in numerical simulations, limiting direct numerical
solution to the larger scales.
Fig. 3 Fusion of large droplet slugs and free flowing droplets is controlled
by the external flow, which changes depending on the geometry and surface
properties of the walls (data from ref. 10). Flow rates are in ml min
21
.
Fig. 4 Two different scalings are shown for droplet generation (from data
in ref. 9). D/Di is the ratio of droplet diameter to the width of the orifice. Qo
with units of ml min
21
, is the flow rate of the oil phase. The flow rate of the
water phase is 0.5ml min
21
.
Fig. 5 Phase diagram for drop formation in flow focusing design (Fig. 3
from ref. 16 reproduced with permission from the American Institute of
Physics). For increasing oil flow rates (top-to-bottom) and increasing oil
flow rates relative to the water flow rate (left-to-right), water droplet sizes
decrease.
Fig. 6 Passive breakup at T-junctions (from Figs. 1 and 2 from ref. 18
reproduced with permission from the American Physical Society): by
varying the relative resistance (length) of the two side arms, the split droplet
volumes also vary. Bottom: the line dividing breaking and non-breaking
droplets is given by eqn. (2).
Lab Chip, 2004, 4 , 257–264258
In this article, we review recent theoretical and numerical
investigations of droplet dynamics in flow and we provide an
overview of numerical methods that can be applied to describe
breakup and coalescence phenomena in microfluidic design.
Computational accuracy and efficiency and the potential impact of
mesh adaptivity will also be discussed.
Theoretical and numerical investigations of
droplet breakup and coalescence
Drop breakup
Droplet generation in flow-focusing design, and breakup in flow,
are the result of the competition of viscous stresses associated with
the imposed flow field, and capillary stresses due to surface tension
between the two phases. In the flow-focusing design of water-in-oil
emulsions (e.g. see refs. 9 and 16), the shearing comes from the
relative magnitude of the co-flowing streams of oil and water. At a
channel T-junction,
18
droplets are sheared and extended as they
travel through the junction and can be split. It is necessary to know
under which conditions for the microphysical parameters and flow
type breakup occurs and at what rates, or droplets are stable (non
breaking), and the extent of droplet deformation.
Consider a drop of size (e.g. undeformed diameter) D, in a matrix
fluid of viscosity m undergoing flow with characteristic magnitude
G of the local velocity gradient, and with surface tension s. Viscous
stresses scale as mG, and extend a drop on a timescale
22
(1 + l)/G,
where l is the ratio of drop-to-matrix viscosities, taking equally
into account the contributions to friction coming from the fluid
viscosities. Surface tension s tends to relax a deformed drop back
to spherical. From a dimensional analysis, the capillary relaxation
velocity scales as s/m, thus the drop relaxation timescale is (1 + l)
mD/s, and capillary stresses scale as s/D. The relevant dimension-
less number is the capillary number Ca = mGD/s (assuming that
inertia is negligible), the ratio of viscous-to-capillary stresses, or
equivalently, the inverse ratio of the two corresponding timescales.
When Ca = O(1), viscous stresses deform the droplet significantly
and breakup may occur. The velocity gradient in a micro channel of
hydraulic diameter ~ D
i
and flow rate Q
o
can be estimated as G ~
Q
o
/D
i
3
and, under conditions of Ca = O(1), this gives a simple
relationship between generated droplet size and imposed flow
rate:
D ~sD
i
3
/mQ
o
(1)
Thus, the larger the flow rate, the smaller the droplet size. Using
9
s ~ 10
23
N m
21
, D
i
= 40 mm, m ~ 30 3 10
23
kg ms
21
and Q
o
= 12 ml min
21
, formula (1) gives a droplet size of D ~ 10mm (and
Re ~ 0.01) and is in agreement with the data in Fig. 4. These and
similar scalings have been employed to generate monodispersed
emulsions of controlled droplet sizes by Thorsen et al.,
14
Anna et
al.
16
and by Tan et al.
9
as illustrated in Figs. 4 and 5. Fig. 4 reveals
that droplet sizes fall in the theoretical scaling (1) when D < D
i
,
that is, droplets are small enough that the hydrodynamic forces
exerted by the channel walls are not important and breakup fully
relies on the straining of the imposed flow. When drops are large,
wall effects are dominant over the stresses directly imposed by the
flow, and the dependence of droplet sizes on flow rate is weaker.
These considerations are corroborated by the trend observed in Fig.
5. Similar scalings for breakup criteria have also been successfully
applied (Fig. 6) to the design of channel T-junction geometry and
inlet and outlet flow rates.
18
Interestingly, they found that the
capillary number above which droplets passively break at a T-
junction scales as
Ca ~e
0
(1/e
0
2/3
2 1)
2
(2)
where e
0
~ l
0
/w
0
, the ratio of length-to-width of the droplets in the
mother channel are upstream of the junction. Downstream of the
junction, split droplet volumes scale inversely with the lengths of
the side arms.
Experimental, theoretical and numerical studies of drop breakup
in imposed flows have been reviewed by Rallison
23
, Stone
24
and
Guido and Greco
25
(see also the review by Basaran
26
for jets).
Criteria for breakup were investigated experimentally (e.g. by
Bentley and Leal
27
) and analytically (e.g. by Navot
28
and
Blawzdziewicz et al.
29,30
). The distribution of drop fragments
resulting from breakup in shear flow was studied (e.g. see refs. 22
and 31). Numerical simulations have been developed (e.g. see refs.
22,32–34). Emulsification is typically promoted using surfactants
that decrease surface tension on the droplet interfaces thus
favouring drop and jet breakup (see refs. 35–42 and the review by
Maldarelli and Huang
43
). Methods for producing controlled micro
sized droplets were developed for shear flow,
22,31,44–47
using co-
flowing streams (see ref. 48 and the recent microfluidics literature
listed above) and extrusion flow.
49
Monodispersed emulsions of
large numbers of droplets with controlled sizes were generated in
flow using the tip-streaming phenomenon due to redistribution of
surfactants to localized end caps on the drop interface.
50,51
Cristini
et al.
22
reported a study on the deformation and breakup of drops in
shear flow demonstrating that nearly bi-disperse emulsions of large
numbers of microscopic droplets of controlled sizes and generation
times can be achieved even without surfactants (Fig. 7). Inter-
estingly, the two sizes alternate (as also found by Tan et al.
9
). The
reason for this lies perhaps in the asymmetrical evolution of the
drop interface near the pinch-off region into cones with different
angles during the latest stages of pinch-off (Fig. 8, top), as
described by the theories of Blawzdziewicz et al.
52
and Lister and
Stone.
53
It was also found
22
that the breakup times have a non-
monotonic dependence on the capillary number, and have a (broad)
minimum corresponding to moderately supercritical shear rates.
This information can be used to optimize emulsification times.
It has been shown,
52,53
that during pinch-off of a thin liquid
thread under zero-Reynolds-number conditions (Fig. 8), the
thinning rate becomes asymptotically constant in time. Thus pinch-
off occurs in a finite time. As the thread thickness h
min
decreases,
viscous stresses mu/h
min
(u = 2dh
min
/dt is the thinning rate)
balance the capillary pressure : s/h
min
: mu/h
min
~s/h
min
. Thus the
neck pinching velocity u ~s/m and is (asymptotically) constant.
This is illustrated in Fig. 8 (bottom).
53
The axial curvature H’’ of
the thread (rescaled with h
min
(t)) at the minimum h
min
(t) is also
found to be asymptotically constant as h
min
? 0, thus revealing
self-similarity of the shape in the transition region. These scalings
for the capillary-driven pinch-off can be used when estimating drop
formation times in a micro channel. In a flow-focusing design
9,16
for water-in-oil emulsions, the time for accumulation of enough
water for one droplet is given by D
3
/Q
w
(the volume of the droplet
divided by the water flow rate). Capillary driven pinch-off time can
Fig. 7 Generation of a nearly bi-dispersed emulsion (From Figs. 11 and 12
from ref. 22 reproduced with permission from the American Institute of
Physics). The cumulative distribution of droplets of alternating sizes (inset)
formed from breakup of a mother drop in shear flow is shown. The droplet
sizes ¯a are rescaled with the maximum stable size in the flow.
Lab Chip, 2004, 4 , 257–264 259
be estimated as (1 + l) mD/s. Droplet generation time will be
determined by the larger of these two times. For droplets that are
smaller than the channel width (D < D
i
), the droplet size D is given
by the scaling (1), and thus the ratio of the two times is O(D/D
i
)
3
indicating that droplet generation time is dominated by the capillary
time for small droplets.
Drop coalescence
Coalescence rates and efficiencies between droplets depend on the
local dynamics of the fluid drainage in the near contact region
between the two approaching fluid interfaces. Asymptotic theories
and numerical simulations of film drainage between coalescing
fluid–fluid interfaces,
54–63
including surfactant effects,
64–70
pro-
vide useful information such as the rates of drop–drop approach.
Two drops approaching each other trap a thin film of the
continuous phase between their interfaces. At small enough gaps
the hydrodynamic forces overcome capillarity and the drop
interfaces deform and often acquire a dimpled shape that traps more
fluid thus opposing coalescence.
55,62,69
In flow-driven drop
interactions, coalescence occurs for capillary numbers smaller than
a critical value such that the drop interaction time is larger than the
drainage time for the fluid trapped in the gap.
57
For sub-critical
capillary numbers, at sub-micron separations, van der Waals forces
become dominant leading to rapid coalescence (film rupture).
Surfactants (through Marangoni stresses, surface viscosity, Gibbs
elasticity, surface and/or bulk diffusivity and intermolecular forces)
can have a significant effect and stabilize emulsions by increasing
deformation and causing surface tension gradients (Marangoni
stresses, see the book by Edwards et al.
71
) that resist radial flow in
the gap and interface–interface approach thus preventing coales-
cence.
68–70,72,73
Chesters and Bazhlekov
69
studied numerically the
axi-symmetric film drainage and rupture between two drops
approaching each other under a constant force in the presence of an
insoluble surfactant. For drops in the millimeter size range the
influence of surfactant diffusion is typically negligible. Drainage is
virtually unaffected by the presence of surfactants down to a film
thickness at which high Marangoni stresses and a transition to
immobile interfaces set in, leading to drainage orders of a slower
magnitude. These phenomena are illustrated in Fig. 9
69
for the
minimum film thickness h
min
. Analytical approximations for the
drainage time were derived.
69
For smaller drops the influence of
diffusion alleviates gradients in surfactant concentration reducing
the effect of surfactants. Yeo et al.
70
(see also ref. 73) using theory
and numerical simulations focused on constant approach velocity
collisions and highlighted the differences in dynamics that arise.
They showed that adding even a slight amount of insoluble
surfactant results in the immobilization of the interface. Three
regimes of drainage and possible rupture exist depending on the
relative magnitudes of the drop approach velocity and the van der
Waals interaction force: nose rupture, rim rupture, and film
immobilization and flattening. The possibility of forming secon-
dary droplets by encapsulating the continuous phase film into the
coalesced drop at rupture was also quantified.
Recent experimental
74,75
and numerical investigations
76
of flow
driven drop–drop coalescence with surfactants revealed that there is
a non-monotonic dependence of the critical capillary number for
coalescence on the surfactant coverage (Fig. 10). The critical
capillary number has a minimum for intermediate coverage due to
Marangoni stresses, and increases at high coverage (close to the
maximum packing of molecules on the interface) perhaps as a
consequence of interface immobilization and reduced deformation.
Fig. 8 Top: initial (dashed) and final (solid) shapes during pinch-off (Fig.
2 from ref. 53). Liquid threads separate large daughter drops and have nearly
conical shapes near the locations of minimum thickness. Bottom (Fig. 6
from ref. 53): constant thread-thinning velocity and shape self-similarity as
h
min
? 0. Reproduced with permission from the American Institute of
Physics.
Fig. 9 Schematic illustrating the effects of surfactants and transition to
immobilized interfaces and slow drainage (Fig. 13,
69
reproduced with
permission from Elsevier).
Fig. 10 Coalescence occurs for capillary numbers below the critical curve.
Curves are shown (data from ref. 76) as a function of surfactant coverage c
of the interface for insoluble surfactants and for a surfactant that is soluble
in the bulk fluids, for different values of a dimensionless van der Waals
force, indicating non-monotonic dependence on the surfactant coverage.
Lab Chip, 2004, 4 , 257–264260
These findings demonstrate that while much progress has been
made in the theoretical understanding of film drainage, there are
still open questions.
Numerical methods
The two main approaches to simulating multiphase and multi-
component flows are interface tracking and interface capturing.
Interface tracking methods
In interface tracking methods (or sharp-interface methods) the
computational mesh elements lay in part or fully on the fluid–fluid
interfaces. Such methods, including boundary-integral meth-
ods,
32,33,34,70,77–80
finite-element methods
81–83
and immersed-
boundary methods,
84,85
are very accurate for simulating the onset of
breakup and coalescence transitions but have difficulties in
simulating through and past the transitions.
In boundary-integral methods, the flow equations are mapped
from the immiscible fluid domains to the sharp interfaces
separating them thus reducing the dimensionality of the problem
(the computational mesh discretizes only the interface). In finite-
element methods, the fluid domains are discretized by a volume
mesh and thus the dimensionality is not reduced. Both these
approaches lead to accurate and efficient solution of the flow
equations because the interface is part of the computational mesh
and the equations and interface boundary conditions are posed
exactly. In immersed-boundary methods, the interfacial forces are
calculated on a surface mesh distinct from the computational
volume mesh where the flow equations are solved and thus in
addition, interpolation onto the volume mesh is needed. A three-
dimensional boundary integral simulation of the onset of drop
breakup in simple shear flow is shown in Fig. 11 (top).
33
In the three
frames, the evolution of a drop towards breakup and the formation
of a thinning liquid thread separating two large daughter drops are
shown. The labels report the dimensionless time from breakup. The
calculated drop shapes (computational mesh) are compared to an
experiment (solid contour) by Dr Guido and coworkers (University
of Naples, Italy, personal communication), demonstrating the high
accuracy of the numerical method. These simulations used an
adaptive triangulated mesh.
34
In Fig. 12 a boundary-integral
simulation of the flow of deformable droplets through a channel is
shown.
86
In Fig. 13 an application of the finite-element method to
high-Reynolds-number satellite production from a jet is illus-
trated.
83
Cross sections of the computed evolution in time (top)
compare well with the experiments (bottom) (corresponding to
different initial conditions). The computational accuracy allowed
the authors to recover features of the evolutions such as capillary
waves (a)–(c), overturning of the top and bottom of the satellite in
(f), spade-shaped profiles, (g) and (h), and spawning of a
subsatellite. In Fig. 14, immersed-boundary simulations of bubbles
interacting and coalescing are shown.
84,87
Near breakup and coalescence transitions, sharp interface models
break down because of the formation of singularities in flow
variables
88
and complex ad hoc cut-and-connect algorithms have
been employed
34,84,89
to change the topology of the meshes and
continue simulating through a transition. (A method to automat-
ically reconnect sharp interfaces has been recently developed by
Shin and Juric
85
). In the simulation in Fig. 14 (bottom), Nobari et
al.
87
reconnected the interfaces as their separation fell under a
prescribed value but noticed that the flow can depend sensitively on
the time at which the interface reconnections are performed.
Reconnection conditions based on asymptotic theories describing
liquid thread pinch-off and film drainage
52,53,90–92
are also used to
extrapolate to the instant of breakup or coalescence, as illustrated in
Fig. 11 (bottom), where Cristini et al.
34
simulated the evolution of
the drop in shear after the pinch-off transition by employing a cut-
and-connect mesh algorithm after the asymptotically linear thin-
ning of the liquid thread
52,53
had been established. This allows
simulations to be continued past the transition while preserving
Fig. 11 Top: adaptive 3-D boundary-integral simulation of the onset of
drop breakup in shear flow (Fig. 1 from ref. 33 reproduced with permission
from the American Institute of Physics). Bottom: the simulation is
continued past the transition by reconnecting the computational mesh (Fig.
6 from ref. 34 reproduced with permission from Elsevier).
Fig. 12 Boundary-integral simulation of 3-D droplet motion through a
channel (Fig. 8 from ref. 86 reproduced with permission from Cambridge
University Press).
Fig. 13 Finite-element simulation (top) of satellite formation and
dynamics from a jet compared to an experiment (bottom) (Figs. 2 and 3 from
ref. 83 reproduced with permission from the American Institute of
Physics).
Lab Chip, 2004, 4 , 257–264 261
accurate information on breakup time (Fig. 11, top) and fragment
sizes.
Interface-capturing methods
Simulations through breakup and coalescence transitions using
interface capturing methods, i.e. lattice-Boltzmann and lattice-
gas,
93–98
constrained-interpolation-profile,
99
level-set,
100
volume-
of-fluid,
101
coupled level-set and volume-of-fluid
102
and partial-
miscibility-model and phase-field methods,
99,103–114
do not require
mesh cut-and-connect operations because the mesh elements do not
lay on the interface, but rather the interface evolves through the
mesh. The fluid discontinuities (e.g. density, viscosity) are
smoothed and the surface tension force is distributed over a thin
layer near the interface to become a volume force (surface tension
being the limit as the layer approaches zero thickness). Interface-
capturing methods are then ideal for simulating breakup and
coalescence in immiscible two-fluid systems (and the effect of
surfactants) and for three or more liquid components, and can be
especially powerful for micro channel design. Lattice-Boltzmann
methods are based on a particle distribution function and on
averaging to capture the macroscopic behavior. The constrained-
interpolation-profile, level-set and volume-of-fluid methods de-
scribe the macro scale directly and use auxiliary functions advected
by the flow (e.g. level-set, volume fraction, and color functions) to
mark the different fluid domains. In Fig. 15, we reproduce a
simulation of drop breakup in shear flow using a volume-of-fluid
method;
115
(see also refs. 116–118). The drop is strongly stretched
in the supercritical flow leading to rupture into numerous fragments
of alternating sizes (see for comparison refs. 9 and 22). The diffuse-
interface (phase-field) approach is based on free-energy functionals
and uses chemical diffusion in narrow transition layers between the
different fluid components as a physical mechanism to smooth flow
discontinuities and to yield smooth evolution through breakup and
coalescence. Through the energy formulation, van der Waals
interactions, electrostatic forces and components with varying
miscibilities can be described. An example of simulation of jet
pinch-off using the partial-miscibility model is shown in Fig. 16 (J.
Lowengrub, U. C Irvine E. Longmire and U. Minnesota, personal
communications, see also refs. 111 and 112). The liquid–liquid
interface and the axial velocity profiles during jet pinch-off from
the diffuse-interface simulation and from an experiment show good
agreement.
Mesh adaptivity
Accurate numerical simulations of multiphase flow and topology
transitions require the computational mesh to resolve both the
macro (e.g. droplet size, channel geometry) and micro scale where
pinch-off or coalescence occur, and to capture local interface
curvature, interface–interface separation, van der Waals forces,
surfactant distributions and Marangoni stresses. Adaptive mesh
algorithms greatly increase accuracy and computational efficiency
in boundary-integral,
34
finite element,
81,82
immersed boundary,
84
and interface capturing
119–127
methods. In Fig. 17 level-set
simulations of drop coalescence in 2-D (top) and drop breakup in
3-D (bottom) under Stokes flow conditions are reported using novel
unstructured adaptive meshes.
34,127
The adaptive mesh algorithm
automatically imposes a mesh element size proportional to the
distance from the interface. As the interfaces deform, approach or
pinch-off, the mesh dynamically maintains accurate resolution of
the flow near the interface. This allows a simulation to recover the
Fig. 14 Top: 3-D front-tracking simulation of rising bubbles (Fig. 13 from
ref. 84 reproduced with permission from Elsevier). Bottom: axi-symmetric
front-tracking simulation of droplet coalescence (Fig. 13 from ref. 87
reproduced with permission from the American Institute of Physics).
Fig. 15 3-D volume-of-fluid simulation (Fig. 16 from ref. 115 reproduced
with permission from the American Institute of Physics) of drop breakup in
shear flow (view along the velocity gradient).
Fig. 16 Comparison between a diffuse-interface simulation (right) (J.
Lowengrub, U. C Irvine) and an experiment (left) (E. Longmire, U.
Minnesota, personal communications) of a liquid–liquid jet pinch-off.
Lab Chip, 2004, 4 , 257–264262
lubrication forces that resist drop approach and delay coalescence
(top). The log-linear plot in the inset reports gap thickness history
for increasing mesh resolution (left to right) and is compared with
lubrication theory (straight line), that in 2-D predicts exponential
gap thinning and thus an infinite coalescence time when only
hydrodynamics are considered.
127
The simulation results converge
to the lubrication theory. Below a gap thickness that decreases with
increased resolution, numerical coalescence occurs. A 3-D simula-
tion (bottom) accurately describes drop breakup during retraction
of a previously elongated drop.
Conclusions
Theoretical considerations relevant to controlled droplet genera-
tion, breakup and coalescence in micro channels were discussed.
Theoretical and numerical investigations of droplet breakup and
coalescence in imposed flows were reviewed. Numerical methods
were reviewed that can be applied for directly simulating droplet
generation, dynamics, breakup and coalescence in micro channels.
The magnitude and type of flow are both important in determining
generation, breakup and coalescence times and droplet sizes. The
local velocity gradient imposed on a droplet is a function of position
in the micro channel, and depends strongly on the channel
geometry. In microfluidics design, channel geometry and flow
conditions can be optimised using (adaptive) simulations. Im-
portant issues are still open in the literature. In particular, one topic
where theoretical and numerical work has been limited and further
investigation is highly desirable in the near future is drop–wall
interactions. For droplets of size comparable to the channel width,
the hydrodynamic forces exerted by the wall on the drop may
exceed those exerted by the imposed channel flow. Under these
conditions, droplet deformation and droplet breakup and coales-
cence criteria and rates may be strongly affected.
V. C. is grateful to John Lowengrub for the many enjoyable and
useful discussions on numerical methods, and acknowledges
support from the National Science Foundation.
References
1 mTAS, 7th International Conference on Micro Total Analysis Systems,
Proceedings of mTAS, ed. M. A. Northrup, K. F. Jensen and D. J.
Harrison, Squaw Valley, California, USA, 2003, Transducers Re-
search Foundation, San Diego, CA.
2 M. A. Burns, B. N. Johnson, S. N. Brahmasandra, K. Handique, J. R.
Webster, M. Krishnan, T. S. Sammarco, P. M. Man, D. Jones, D.
Heldsinger, C. H. Mastrangelo and D. T. Burke, Science, 1998, 282,
484.
3 M. G. Pollack, P. Y. Paik, A. D. Shenderov, V. K. Pamula, F. S.
Dietrich and R. B. Fair, 7th International Conference on mTAS, 2003,
619–622.
4 S. Kaneda and T. Fujii, 7th International Conference on mTAS, 2003,
1279–1282.
5 M. Hirano, T. Torii, T. Higuchi, M. Kobayashi and H. Yamazaki, 7th
International Conference on mTAS, 2003, 473.
6 B. Zheng, L. S. Roach and R. F. Ismagilov, J. Am. Chem. Soc., 2003,
125, 11170–11171.
7 V. Srinivasan, V. K. Pamula, M. G. Pollack and R. B. Fair, 7th
International Conference on mTAS, 2003, 1287–1290.
8 B. Schaack, B. Fouque, S. Porte, S. Combe, A. Hennico, O.
Filholcochet, J. Reboud, M. Balakirev and F. Chatelain, 7th Inter-
national Conference on mTAS, 2003, 669–672.
9 Y. C. Tan, V. Cristini and A. P. Lee, Sens. Actuators, 2004, in
review.
10 Y. C. Tan, J. Fisher, A. I. Lee, E. Lin, V. Cristini and A. P. Lee, Lab
Chip, 2004, DOI: 10.1039/b403280m.
11 H. Song, J. D. Tice and R. F. Ismagilov, Angew. Chem., Int. Ed., 2003,
42(7), 767–772.
12 J. S. Go, E. H. Jeong, K. C. Kim, S. Y. Yoon and S. Shoji, 7th
International Conference on mTAS, 2003, 1275–1278.
13 S. Sugiura, M. Nakajima, S. Iwamoto and M. Seki, Langmuir, 2001,
17, 5562.
14 T. Thorsen, W. R. Robert, F. H. Arnold and S. R. Quake, Phys. Rev.
Lett., 2001, 86, 4163.
15 T. Nisisako, T. Torii and T. Higuchi, Lab Chip, 2002, 2, 24.
16 S. L. Anna, N. Bontoux and H. A. Stone, Appl. Phys. Lett., 2003, 82,
364.
17 J. D. Tice, D. A. Lyon and R. F. Ismagilov, Anal. Chim. Acta, 2004,
507, 73.
18 D. R. Link, S. L. Anna, D. A. Weitz and H. A. Stone, Phys. Rev. Lett.,
2004, 92, 1178–1190, Art. 054503.
19 M. Kakuta, D. A. Jayawickrama, A. M. Wolters, A. Manz and J. V.
Sweedler, Anal. Chem., 2003, 75, 956–960.
20 C. de Bellefon, N. Pestre, T. Lamouille, P. Grenouillet and V. Hessel,
Adv. Synth. Catal., 2003, 345(1 + 2), 190–193.
21 M. Nakano, J. Komatsu, S. Matsuura, K. Takashima, S. Katsura and A.
Mizuno, J. Biotechnol., 2003, 102, 117–124.
22 V. Cristini, S. Guido, A. Alfani, J. Blawzdziewicz and M. Loewenberg,
J. Rheol., 2003, 47, 1283.
23 J. M. Rallison, Annu. Rev. Fluid Mech., 1984, 16, 45.
24 H. A. Stone, Annu. Rev. Fluid Mech., 1994, 26, 65.
25 S. Guido and F. Greco, in Rheology Reviews, ed. D. M. Binding and K.
Walters, British Soc. Rheol. 2004, in press.
26 O. Basaran, AIChE J., 2002, 48, 1842.
27 B. J. Bentley and L. G. Leal, J. Fluid Mech., 1986, 167, 241.
28 Y. Navot, Phys. Fluids, 1999, 11, 990.
29 J. Blawzdziewicz, V. Cristini and M. Loewenberg, Phys. Fluids, 2002,
14, 2709.
30 J. Blawzdziewicz, V. Cristini and M. Loewenberg, Phys. Fluids, 2003,
15, L37.
31 V. Schmitt, F. Leal-Calderon and J. Bibette, Colloid Chem. II (Book
Series: Topics in current chemistry), EDP Sciences, Les Ulis, France,
2003, vol. 227, p. 195.
32 A. Z. Zinchenko, M. A. Rother and R. H. Davis, J. Fluid Mech., 1999,
391, 249.
33 V. Cristini, J. Blawzdziewicz and M. Loewenberg, Phys. Fluids, 1998,
10, 1781.
34 V. Cristini, J. Blawzdziewicz and M. Loewenberg, J. Comput. Phys.,
2001, 168, 445.
35 H. A. Stone and L. G. Leal, J. Fluid Mech., 1990, 220, 161.
36 W. J. Milliken, H. A. Stone and L. G. Leal, Phys. Fluids A, 1993, 5,
69.
37 W. J. Milliken and L. G. Leal, J. Colloid Interface Sci., 1994, 166,
275.
38 Y. Pawar and K. J. Stebe, Phys. Fluids, 1996, 8, 1738.
39 C. D. Eggleton, Y. Pawar and K. Stebe, J. Fluid Mech., 1999, 79,
385.
40 X. Li and C. Pozrikidis, J. Fluid Mech., 1997, 341, 165.
41 M. Siegel, SIAM J. Appl. Math., 1999, 69, 1998.
Fig. 17 Adaptive unstructured meshes of triangles (top) and tetrahedra
(bottom) maintain computational accuracy during simulations (data from
ref. 127). Some tetrahedra may appear skewed (bottom) as a result of
projecting the 3-D mesh onto the plane of the figure.
Lab Chip, 2004, 4 , 257–264 263
42 Y.-J. Jan and G. Tryggvason, in Proceedings of the Symposium on
Dynamics of Bubbles and Vortices Near a Free Surfaces,ed. Sahin and
Tryggvason, ASME, NY, 1991, vol. 119, 46–59.
43 C. Maldarelli and W. Huang, in Flow particle suspensions. ed. U.
Schaflinger, CISM Courses and Lectures, Springer-Verlag, New York,
1996, vol. 370, p. 125.
44 T. G. Mason and J. Bibette, Langmuir, 1997, 13, 4600.
45 A. J. Abrahamse, R. van Lierop, R. G. M. van der Sman, A. van der
Padt and R. M. Boom, J. Membr. Sci., 2002, 204, 125.
46 W. L. Olbricht and D. M. Kung, Phys. Fluids A, 1992, 4, 1347.
47 W. G. P. Mietus, O. K. Matar, C. J. Lawrence and B. J. Briscoe, Chem.
Eng. Sci., 2001, 57, 1217.
48 P. B. Umbanhowar, V. Prasad and D. A. Weitz, Langmuir, 2000, 16,
347.
49 I. Kobayashi, M. Yasuno, S. Iwamoto, A. Shono, K. Satoh and M.
Nakajima, Colloids Surf., 2002, 207, 185.
50 R. A. de Bruijn, Chem. Eng. Sci., 1993, 277, 48.
51 C. D. Eggleton, T. M. Tsai and K. J. Stebe, Phys. Rev. Lett., 2001,
87.
52 J. Blawzdziewicz, V. Cristini and M. Loewenberg, Bull. Am. Phys.
Soc., 1997, 42, 2125.
53 J. R. Lister and H. A. Stone, Phys. Fluids, 1998, 10, 2758.
54 G. P. Neitzel and P. Dell’Aversana, Annu. Rev. Fluid Mech., 2002, 34,
267.
55 S. G. Yiantsios and R. H. Davis, J. Fluid Mech., 1990, 217, 547.
56 S. G. Yiantsios and R. H. Davis, J. Colloid Interface Sci., 1991, 144,
412.
57 A. K. Chesters, Chem. Eng. Res. Des., 1991, 69, 259.
58 P. D. Howell, J. Eng. Math., 1999, 35, 271.
59 R. H. Davis, J. A. Schonberg and J. M. Rallison, Phys. Fluids A, 1989,
1, 77.
60 D. Li, J. Colloid Interface Sci., 1994, 163, 108.
61 E. Klaseboer, J. Ph. Chevaillier, C. Gourdon and O. Masbernat, J.
Colloid Interface Sci., 2000, 229, 274.
62 M. A. Rother and R. H. Davis, Phys. Fluids, 2001, 13, 1178.
63 J. Eggers, J. R. Lister and H. A. Stone, J. Fluid Mech., 1999, 401,
293.
64 G. Singh, G. J. Hirasaki and C. A. Miller, J. Colloid Interface Sci.,
1996, 184, 92.
65 D. Li, J. Colloid Interface Sci., 1996, 181, 34.
66 K. D. Danov, D. S. Valkovska and I. B. Ivanov, J. Colloid Interface
Sci., 1999, 211, 291.
67 D. S. Valkovska, K. D. Danov and I. B. Ivanov, Colloids Surf., A, 2000,
175, 179.
68 V. Cristini, J. Blawzdziewicz and M. Loewenberg, J. Fluid Mech.,
1998, 366, 259.
69 A. K. Chesters and I. B. Bazhlekov, J. Colloid Interface Sci., 2000,
230, 229.
70 L. Y. Yeo, O. K. Matar, E. S. P. de Ortiz and G. E. Hewitt, J. Colloid
Interface Sci., 2003, 257(1), 93–107.
71 D. A. Edwards, H. Brenner and D. T. Wasan, Interfacial Transport
Processes and Rheology, Butterworth–Heinemann, London, 1991.
72 J. Blawzdziewicz, V. Cristini and M. Loewenberg, J. Colloid Interface
Sci., 1999, 211, 355.
73 L. Y. Yeo, O. K. Matar, E. S. P. de Ortiz and G. F. Hewitt, J. Colloid
Interface Sci., 2001, 241, 233.
74 Y. T. Hu, D. J. Pine and L. G. Leal, Phys. Fluids, 2000, 12, 484.
75 J. W. Ha, Y. Yoon and L. G. Leal, Phys. Fluids, 2003, 15, 849.
76 H. Zhou, V. Cristini, C. W. Macosko and J. Lowengrub, Phys. Fluids,
in review.
77 C. Pozrikidis, Boundary Integral and Singularity Methods for
Linerarized Viscous Flow, Cambridge University Press, Cambridge,
1992.
78 A. Prosperetti and H. N. Oguz, Philos. Trans. R. Soc. London, 1997,
355, 491.
79 T. Y. Hou, J. S. Lowengrub and M. J. Shelley, J. Comput. Phys., 2001,
169, 302.
80 C. Pozrikidis, Eng. Anal. Bound. Elem., 2002, 26, 495.
81 E. D. Wilkes, S. Phillips and O. Basaran, Phys. Fluids, 1999, 11,
3577.
82 R. Hooper, V. Cristini, S. Shakya, J. Lowengrub, C. W. Macosko and
J. J. Derby, in Computational Methods in Multiphase Flow, ed. H.
Power and C. A. Brebbia, Wessex Institute of Technology Press, 2001,
vol. 29.
83 P. K. Notz, A. U. Chen and O. A. Basaran, Phys. Fluids, 2001, 13,
549.
84 G. Tryggvason, B. Bunner, A. Esmaeeli, D. Juric, N. Al-Rawahi, W.
Tauber, J. Han, S. Nas and Y.-J. Jan, J. Comput. Phys., 2001, 169,
708.
85 S. Shin and D. Juric, J. Comput. Phys., 2002, 180, 427.
86 C. Coulliette and C. Pozrikidis, J. Fluid Mech., 1998, 358, 1.
87 M. R. Nobari, Y.-J. Jan and G. Tryggvason, Phys. Fluids, 1996, 8,
29.
88 T. Y. Hou, Z. Li, S. Osher and H. Zhao, J. Comput. Phys., 1997, 134,
236.
89 N. Mansour and T. Lundgren, Phys. Fluids A, 1990, 2, 1141.
90 J. B. Keller and M. J. Miksis, SIAM J. Appl. Math., 1983, 43, 268.
91 J. Eggers and T. F. Dupont, J. Fluid Mech., 1994, 262, 205.
92 J. Eggers, Phys. Fluids, 1995, 7, 941.
93 D. H. Rothman and S. Zaleski, Lattice Gas Cellular Automata,
Cambridge University Press, Cambridge, 1997.
94 S. Chen and G. D. Doolen, Annu. Rev. Fluid Mech., 1998, 30, 329.
95 R. R. Nourgaliev, T. N. Dinh, T. G. Theofanous and D. Joseph, Int. J.
Multiphase Flow, 2003, 29, 117.
96 T. Watanabe and K. Ebihara, Comput. Fluids, 2003, 32, 823.
97 K. Sankaranarayanan, I. G. Kevrekidis, S. Sundaresan, J. Lu and G.
Tryggvason, Int. J. Multiphase Flow, 2003, 29, 109.
98 A. Lamura, G. Gonnella and J. M. Yeomans, Europhys. Lett., 1999, 45,
314.
99 T. Yabe, F. Xiao and T. Utsumi, J. Comput. Phys., 2001, 169, 556.
100 S. Osher and R. Fedkiw, J. Comput. Phys., 2001, 169, 463.
101 R. Scardovelli and S. Zaleski, Annu. Rev. Fluid Mech., 1999, 567.
102 M. Sussman and E. G. Puckett, J. Comput. Phys., 2000, 162, 301.
103 J. S. Lowengrub and L. Truskinovsky, Proc. R. Soc. Lond. Ser. A,
1998, 454, 2617.
104 J. Lowengrub, J. Goodman, H. Lee, E. Longmire, M. Shelley and L.
Truskinovsky, in Free boundary problems: theory and applications,
ed. I. Athanasopoulos, M. Makrakis and J. F. Rodrigues, CRC Press,
London, 1999 vol. 221.
105 D. M. Anderson, G. B. McFadden and A. A. Wheeler, Annu. Rev. Fluid
Mech., 1998, 30, 139.
106 D. Jacqmin, J. Comput. Phys., 1999, 155, 96.
107 D. Jamet, O. Lebaigue, N. Coutris and J. M. Delhaye, J. Comput. Phys.,
2001, 169, 624.
108 H. Lee, J. S. Lowengrub and J. Goodman, Phys. Fluids, 2002, 14,
492.
109 H. Lee, J. S. Lowengrub and J. Goodman, Phys. Fluids, 2002, 14,
514.
110 V. E. Badalassi, H. D. Ceniceros and S. Banerjee, J. Comput. Phys.,
2003, 190, 371.
111 J. S. Kim, K. Kang and J. S. Lowengrub, J. Comput. Phys., 2004, 193,
511.
112 J. S. Kim, K. Kang and J. S. Lowengrub, Commun. Math. Sci., 2004,
2, 53–77.
113 L. Q. Chen, Annu. Rev. Mater. Res., 2002, 113, 32.
114 G. Patzold and K. Dawson, Phys. Rev. E: Stat. Phys., Plasmas, Fluids,
Relat. Interdiscip. Top., 1995, 52, 6908.
115 J. Li, Y. Y. Renardy and M. Renardy, Phys. Fluids, 2000, 12, 269.
116 Y. Y. Renardy, V. Cristini and J. Li, Int. J. Multiphase Flow, 2002, 28,
1125.
117 Y. Y. Renardy and V. Cristini, Phys. Fluids, 2001, 13, 7.
118 Y. Y. Renardy and V. Cristini, Phys. Fluids, 2001, 13, 2161.
119 G. Agresar, J. J. Linderman, G. Tryggvason and K. G. Powell, J.
Comput. Phys., 1998, 143, 346.
120 M. Sussman, A. Almgren, J. Bell, P. Colella, L. Howell and M.
Welcome, J. Comput. Phys., 1999, 148, 81.
121 N. Provatas, N. Goldenfeld and J. Dantzig, J. Comput. Phys., 1999,
148, 265.
122 O. Ubbink and R. I. Issa, J. Comput. Phys., 1999, 153, 26.
123 H. D. Ceniceros and T. Y. Hou, J. Comput. Phys., 172, 609, 2001.
124 J. H. Jeong, N. Goldenfeld and J. A. Dantzig, Phys. Rev. E: Stat. Phys.,
Plasmas, Fluids, Relat. Interdiscip. Top., 2001, 64, Art. 041602 Part
1.
125 J. H. Jeong, J. A. Dantzig and N. Goldenfeld, Metall. Mater. Trans. A,
2003, 34, 459.
126 I. Ginzberg and G. Wittum, J. Comput. Phys., 2001, 166, 302.
127 X. Zheng, A. Anderson, J. Lowengrub and V. Cristini, J. Comput.
Phys., in review.
Lab Chip, 2004, 4 , 257–264264
