Chemical Engineering Science 54 (1999) 17591774

Dynamics of drop formation in viscous ows

Xiaoguang Zhang
Air Products and Chemicals, Inc., 7201 Hamilton Boulevard, Allentown, PA 18195-1501, USA
Received 22 April 1998; accepted 21 December 1998
Abstract
This paper presents numerical results of the dynamics of a viscous liquid drop that is being formed directly at the tip of a vertical,
circular tube and breaks into an ambient, viscous uid. A model is developed to predict the evolution of the drop shape and its
breakup based on the volume-of-uid/continuum-surface-force method, which is a solution algorithm for computing transient,
two-dimensional, incompressible uid ow with surface tension on free surfaces of general topology (Richards et al. (1995 Physics of
Fluids, 257, 111145)). The full NavierStokes systemis solved by using nite-dierence formulation on a Eulerian mesh. The mesh is
xed in space, with the ow and interface moving through it to ensure accurate calculations of complex free surface ows and
topology, including surface breakup and coalescence. The nonlinear dynamics of drop growth and breakup are simulated for
describing and predicting the universal features of drop formation. The focus here is on dynamic eects of a quiescent or owing
ambient uid on drop breakup and the subsequent generation of satellite droplets. The eects of nite inertial, capillary, viscous, and
gravitational forces are accounted for in order to classify drastically dierent formation dynamics and to elucidate the fate of satellite
droplets. The numerical predictions are compared with experimental measurements for a typical system of 2-ethyl-1-hexanol drops
forming and breaking into quiescent water, and the results show excellent agreement. 1999 Elsevier Science Ltd. All rights
reserved.
Keywords: Drops; Satellites; Evolution; Deformation; Breakup; Dynamics; Free-surface ows; Volume-of-uid; Continuum-surface
force
1. Introduction
A common way of dispersing a liquid in an immiscible
uid is to ow it continuously through a nozzle or an
orice plate from which it emerges into the ambient uid
and breaks into drops. At low ow rates, the liquid being
ejected emanates from the nozzle as discrete drops under
its own weight. At high ow rates, the liquid is ejected
from the nozzle as a jet that subsequently breaks up into
small drops because of well-known Rayleigh instability
(Clift et al., 1978). The formation of liquid drops from
nozzles has long been a topic of interest because of its
occurrence in a wide variety of engineering applications,
such as distillation and extraction processes and spraying
and emulsifying technologies, among others. The pre-
vious applications have demonstrated the importance of
a fundamental understanding of the dynamics of drop
formation in designing and controlling these processes to
obtain certain desired drop characteristics. The dynamics
of a viscous liquid drop forming from a capillary tube
and breaking into a quiescent, ambient uid that is
inviscid and dynamically inactive has been extensively
investigated in an earlier study by Zhang (1999). The
evolution with time of surface prole and internal ow
of the drop is simulated by using a numerical method
based on algorithms of volume of uid (VOF) (Hirt and
Nichols, 1981) and continuum surface force (CSF)
(Brackbill et al., 1992). In the study by Zhang (1999), the
eects of nite inertial, capillary, viscous, and gravi-
tational forces are accounted for in order to classify
drastically dierent formation dynamics and, in particu-
lar, to elucidate the feature of satellite generation. The
present paper follows this earlier work very closely and
extends the results to the situation where a viscous liquid
drips from a circular capillary tube at low ow rates and
breaks up as drops into another immiscible, viscous
liquid. The focus here is on the dynamic eects of the
ambient uid on drop formation. Despite the fact that
the overall process of drop formation may not be quali-
tatively altered, the dynamic eects of an ambient uid
bring about an additional hydrodynamic force on the
forming drop, which may change the surface prole of
0009-2509/99/$ see front matter 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 9 9 ) 0 0 0 2 7 - 5
the drop and features of drop breakup and satellite
creation. Moreover, an external ow of the ambient uid
has been found to lead to a smaller-volume breako drop
and a longer drop length prior to its breakup (Oguz and
Prosperetti, 1993; Zhang and Stone, 1997).
During liquid dripping from a capillary tube, the vol-
ume of a drop that emerges fromthe tube increases by the
continuous addition of the drop liquid. When the volume
of the forming drop exceeds a critical value, the drop
necks and a large portion of it starts to fall rapidly and
eventually breaks o from the tube. The most interesting,
yet complicated, phenomena in the drop formation are
the evolution of a liquid thread that connects the detach-
ing drop and the remainder of the liquid on the tube
during the drop breakup and generation of satellite drop-
lets subsequent to the thread breakup. Athorough know-
ledge of the drop evolution and breakup is indispensable
for complete predictions and designs of practical pro-
cesses that involve interfacial contact and ows of two
liquids.
Drop formation from a nozzle or an orice has been
studied extensively from theoretical and experimental
perspectives. Much of the previous theoretical, as well as
experimental, work has been aimed at predicting the size
of drops breaking from a nozzle as a function of uid
properties, nozzle geometry, and ow rate of the liquid
through the nozzle (Clift et al., 1978; Kumar and Kuloor,
1970). The theoretical analyses for this purpose have been
based primarily on macroscopic force balances and have
assumed that drop formation occurs in two stages. The
rst stage takes place with a pure static growth of the
drop, which ends with a loss of equilibriumof forces. The
second stage corresponds to the necking and breaking of
the drop from the nozzle (Scheele and Meister, 1968;
Heertjes et al., 1971). These studies have concluded that
the volume of the drops that are so formed depends on
not only the nozzle size and liquid properties, as is the
case with static pendant drop (Michael, 1981), but also
the liquid ow rate through the nozzle. Although the
previous studies dier slightly in their approaches to
analyzing the second stage of drop formation, these sim-
plied models can, at best, approximate reality when the
liquid ow rate is vanishingly small and the predictions
of the size of breako drops exhibit deviations from
experimental measurements, with errors around 20%
(Clift et al., 1978).
Although the previously cited studies of drop forma-
tion in the dripping mode have captured some of the
gross features of the phenomenon, they have done little
to elucidate the fundamental uid mechanics of the pro-
cess, viz., the evolution of the shape of the forming drop
in particular, the development, extension, and breakup of
the liquid thread and satellite generation. Complete
simulations of drop formation involve the solution of the
NavierStokes system with specied boundary condi-
tions. Cram (1984) and Eggers and Dupont (1994) have
derived and solved one-dimensional equations of mass
and axial momentum conservations to simulate drop
dripping. These approximate equations have been
derived from the NavierStokes system by either
(1) neglecting the radial component of the velocity
and variation of the axial component of the velocity
and pressure in the radial direction (Cram, 1984) or (2)
extending the velocity and pressure variables in a Taylor
series in the radial direction and retaining only
the lowest-order terms in these expansions (Eggers
and Dupont, 1994). Notwithstanding the failure to
prole actual velocity elds observed inside growing
drops and to describe the dynamics of drop breakup
(Schulkes, 1993), these models are surprisingly successful
in predicting the evolution with time of the drop shapes,
as made evident by the qualitative comparison of cal-
culated shapes with experimental observations in a few
specic situations when the liquid owrate is vanishingly
small.
Qualitative features of the dynamics of the thread
breakup have been explored experimentally by Peregrine
et al. (1990) and, in more detail, by Shi et al. (1994). These
studies show the details of the evolution with time of
a liquid thread at the time preceding, at the instant of,
and at the time following drop detachment. Peregrine et
al. (1990) have documented photographic sequences of
events occurring during breakup of the thread in their
paper. Their results exhibit the process of double break-
age of the liquid thread. Under the weight of a detaching
drop underneath, the thread necks and breaks at its
lower end, where the thread joins with the falling drop, to
form a free primary drop. Because of unbalanced capil-
lary forces on the thread after its rst breakup, the thread
recoils; secondary breakup then occurs at its upper end,
leading to the generation of satellite droplets. These ex-
perimental observations have been well modeled by
Schulkes (1994), who numerically integrated a potential
ow formulation of Eulers equations for the formation
of inviscid drops. The primary contribution of Shi et al.
(1994) is the demonstration of the breakup of drops with
a large range of viscosities by experiment and computa-
tion, using the one-dimensional model developed by Egg-
ers and Dupont (1994). Shi and his coworkers show that
liquid threads can spawn a series of smaller necks with
ever-thinner diameters prior to breakup. In spite of their
demonstration of these phenomena during drop detach-
ment, the authors do not provide any quantitative evalu-
ation of the process or systematic description of the roles
of operating conditions on the drop formation and
thread breakup. Recently, Zhang and Stone (1997) have
developed a numerical model to simulate the formation
of viscous drops from a capillary tube using a boundary
integral method. These researchers consider the evolu-
tion and breakup of a drop to assess quantitatively
the eects of viscous, buoyant, and capillary forces.
In particular, they reveal the important role of the
1760 X. Zhang/Chemical Engineering Science 54 (1999) 17591774
viscous eects and imposed ows of an ambient uid on
the dynamics of drop deformation and breakup. In the
low-Reynolds-number ow limit, this model cannot
identify the importance of the inertial force, which has
been found to be eective as the ow rate is increased
(Zhang and Basaran, 1995; Zhang, 1999).
It is noteworthy that Richards et al. (1995) have de-
veloped a dynamic simulation method based on the
VOF/CSF numerical technique to investigate the full
transient of liquid drop and jet formation from startup to
breakup. In their study, a transition from dripping to
jetting of an emerging liquid is obviously identied when
the liquid ow rate exceeds a critical value. Results of
their numerical simulations show signicantly more ac-
curacy than previously simplied analyses in predicting
the jet dynamics, including the jet evolution, velocity
distribution, and volume of breako drops. The novel
feature of this method is the use of a Eulerian volume-
tracking approach designed to simulate ows with free
interfaces of arbitrarily complex topology (e.g., merging
and breakup), which are of particular interest to us in our
research. However, Richards and coworkers have fo-
cused, in their paper (Richards et al., 1995), on predicting
the volume of breako drops under the conditions cor-
responding to liquid ow rates near and above the
formation of a jet with the Reynolds number exceeding
400. A detailed understanding of liquid dripping into
another immiscible, viscous uid at the level of Zhang
(1999) is still lacking.
The major goal of this study, along with the sub-
sequent study of numerical solutions, is to remedy these
and related inadequacies so as to complete our under-
standing of drop formation. The transient NavierStokes
equation has been solved for the axisymmetric free-
boundary problemof a Newtonian liquid that is dripping
vertically and breaking as drops into another immiscible
Newtonian liquid. The VOF/CSF-based numerical ap-
proximations used by Richards et al. (1995) have been
extended to simulate the complete process of drop forma-
tion from the time a drop emerges from the tube to its
breako from the tube with continuous feeding of the
drop liquid at a certain ow rate. In contrast to most of
the previous studies, the only assumptions involved in
the present numerical model are (1) Newtonian uids of
drop and ambient phases with constant physical proper-
ties and (2) laminar ows. The special feature of
VOF/CSF, which allows free surfaces to cross the com-
putational mesh smoothly, ensures that the calculations
pass the point of necking followed by natural breakup of
drops without interruption, which is a major incentive
for using the VOF/CSF method in this study. We pay
particular attention to the dynamic eects of an ambient
uid on breakup of a liquid thread and generation of
satellite droplets. The numerical results are compared
with the available experimental data, which are obtained
using an ultra-high-speed motion analysis and video
system. Section 2 presents the numerical model, which
includes the problem denition and formulation as well
as a brief discussion of the VOF/CSF algorithm. A more-
detailed description of the VOF/CSF and its validation
on several test problems can be found elsewhere
(Richards, 1994; Hirt and Nichols, 1981; Brackbill et al.,
1992). The approach used to solve the governing equa-
tions and associated boundary conditions is also
described in Section 2. Section 3 briey describes the
experimental apparatus and methods of data acquisition
and analysis. Typical experiments with 2-ethyl-1-hexanol
(2EH) drops forming in distilled water are performed,
and the resulting data have been compared with the
numerical simulation for dierent conditions. The com-
putational results and analyses of the ndings are
the subject of Section 4. Most of the results are presented
in dimensionless forms and show the importance of
dynamic eects of an ambient uid and inertial,
viscous, capillary, and gravitational forces on drop
formation. For illustration and verication purposes,
comparisons of numerical simulations and experimental
measurements for typical processes are also presented
in Section 4. The concluding remarks are provided in
Section 5.
2. Mathematic formulation and numerical method
The system of interest is an axisymmetric drop of an
incompressible Newtonian liquid, density and viscosity
zj, forming into an immiscible, incompressible New-
tonian uid, density j#Aj and viscosity j, at the tip of
a circular cylindrical capillary tube. The drop liquid is
injected at a constant owrate, Q, as shown in Fig. 1. The
tube has inner and outer radii, R and R
M
, respectively,
and its axis lies along the direction of the gravity vector,
g; therefore, an axisymmetric free-boundary problem is
imposed. The ambient uid is contained in a cylindrical
tank that has an inner radius, R
R
, and is coaxial with the
capillary tube. The surface tension, o, of the liquidliquid
interface is spatially uniform and constant in time. It is
convenient to dene a cylindrical coordinate system
r, z, , whose origin lies at the center of the outlet plane
of the tube and where r denotes the radial coordinate; z is
the axial coordinate measured in the opposite direction
of gravity, g; and is the azimuthal angle. For the
axisymmetric conguration of interest in this study, the
problem is independent of the azimuthal coordinate.
Isothermal, transient ows of liquids of the drop and
ambient phases are governed by the NavierStokes
system
V) v"0, (1)
cv
ct
#V) (vv)"!
1
j
Vp#
1
j
V) t#g#
1
j
F
@
, (2)
X. Zhang/Chemical Engineering Science 54 (1999) 17591774 1761
Fig. 1. Schematic of a drop forming from a capillary tube and breaking
into an ambient uid.
where v is the velocity vector, t is time, p is the scalar
pressure, and F
@
denotes body forces that may be present
in the system. The viscous stress tensor, , is dened as
follows for the Newtonian liquid:
t"j[(Vv)#(Vv)2]. (3)
The transient NavierStokes system (Eqs. (1) and (2)) is
solved by using a nite-dierence formulation on a
Eulerian mesh, which is xed in space with the ow and
interface moving through it. The basic algorithm is to
break up a time descretization of the momentum equa-
tion into two steps. In the rst step, a velocity eld is
computed from incremental changes resulting from vis-
cosity, advection, gravitational acceleration, and body
forces. In the second step, the velocity eld is projected
onto a zero-divergence vector eld, resulting in a single
Poisson equation for the pressure eld. The details of the
overall solution scheme can be found in a study by Kothe
and Mjolsness (1992).
A nonconventional approach, referred to as the CSF
method (Brackbill et al. 1992), is used to represent the
eect of the surface tension at free surfaces. It interprets
the surface tension as a continuous, three-dimensional
eect across free surfaces and incorporates it as a localiz-
ed volume force in the NavierStokes equation rather
than as a boundary-value condition. The volume force,
which is nonzero only within free surfaces, is given in the
CSF model by
F
@
"F
QT
"oVF, (4)
where o is the surface tension, is the local free surface
curvature, and F denotes a VOF volume function that is
used to track the prole of free surfaces.
Free surfaces are represented using the VOF technique
pioneered by Hirt and Nichols (1981). The VOF method
provides a means of following uid regions through a
Eulerian mesh of stationary cells and enables a nite-
dierence representation of free surfaces that are arbit-
rarily oriented with respect to the computational mesh.
The scalar function, F, is dened as the fractional volume
of the drop uid in the respective cells of the computa-
tional mesh and is given by
F(x, z, t)"

1, in the drop,
'0,(1, at the free surface,
0, in the ambient fluid.
(5)
The free surface position in the respective cells is divided
from its neighboring cells and is governed by
cF
ct
#(v ) V)F"0. (6)
This equation states that the volume function, F, moves
with the uid and provides the information necessary to
reconstruct the free surfaces.
The NavierStokes system (Eqs. (1) and (2)) is solved
for drop formation beginning at the instant at which the
free surface of the drop is at, situated at the tip of the
tube with the entire system being at rest at t)0, subject
to the following boundary conditions. The three-phase
contact line, where the drop liquid, the ambient uid, and
the solid surface meet, remains pinned to the sharp inner
edge of the tube surface for all times t*0:
r
R
"1, at
z
R
"0. (7)
In contrast to liquidgas systems, when a liquid drop
forms from a tube into another immiscible liquid (the
focus of the present paper), whether the contact line pins
on the inner edge or the outer edge depends on the
relative wettabilities of the two liquids on the at surface
of the tube (Berg, 1993). For simplicity, the contact line is
assumed to pin to the inner edge of the tube and a tube
having a wall suciently thin is used in this study. Never-
theless, the simulations can also be applied for systems in
which the contact line pins to the outer edge of the tube
as long as the tube wall is suciently thin that its eects
can be neglected (Zhang, 1999). The xed contact line
eliminates the troublesome determination of the contact
1762 X. Zhang/Chemical Engineering Science 54 (1999) 17591774
angle in this problem. Far upstream of the tube outlet,
the ow inside the tube is fully developed,
v
BP
"0, v
B X
"2
Q
R
1!

r
R

, (8)
as
z
R
P!R, for 0)
r
R
)1,
where v
B P
and v
BX
are the radial and axial components of
the velocity of the drop liquid, respectively. In computa-
tions reported in this paper, the tube length has been set
to be twice of the inner radius of the tube. Greater lengths
have been tested, and the calculated results have shown
no noticeable dierence. The ambient uid may be qui-
escent or may be set in motion by continuously supplying
the ambient uid at a constant ow rate, Q
M
. In the latter
case, a fully developed owin the annulus far upstreamof
the outer container is imposed (Bird et al., 1960),
v
MP
"0,
v
MX
"2
Q

(R
R
!R

)
;

1!

r
R
R

#
1!(R

/R
R
)
ln(R
R
/R

)
ln(r/R
R
)

, (9)
as
z
R
P!R, for
R

R
)
r
R
)
R
R
R
,
where v
MP
and v
MX
are the radial and axial components of
the velocity of the ambient uid, respectively. Along
surfaces where the liquid and solid come in contact, the
liquid obeys the conditions of no slip and no penetration:
v"0, at
z
R
"0, for 1)
r
R
)
R
M
R
, (10)
at
r
R
"1 and,
r
R
"
R
M
R
, for !R(
z
R
)0,
at
r
R
"
R
R
R
, for !R(
z
R
(R.
A continuative outow boundary condition is used at
the top of the computational domain. The length in #z
direction is set to be suciently longer than the estimated
drop breako length (i.e., the length of a drop at instant
of its breakup). Greater lengths have been tested, and the
calculated results have shown no noticeable dierence in
features of drop breakup and satellite generation.
Nondimensionalizing the governing NavierStokes
system and boundary conditions, using R as the length
scale and the average velocity of liquid inside the tube
u
G
"Q/R as the velocity scale, yields three dimension-
less parameters that describe the uid mechanics of drop
formation:
Re"
ju
G
R
j
, Ca"
ju
G
o
and G"
AjRg
o
.
The Reynolds number, Re, measures the importance of
inertial forces relative to viscous forces; the capillary
number, Ca, measures the importance of viscous forces
relative to surface tension forces; and the gravitational
Bond number, G, measures the importance of gravi-
tational forces relative to surface tension forces. The
viscosity ratio of the drop liquid and ambient uid, z, is
introduced to account for the dynamic eects of the
ambient uid. The wall eects of the ambient uid con-
tainer are described by the ratio of the inner radii of the
container and the capillary tube, R
R
/R. If an axial ow is
imposed in the ambient uid surrounding the capillary
tube with a average velocity u
M
"Q
M
/(R
R
!R
M
), a di
mensionless characteristic velocity, u
M
/u
G
, enters the prob-
lem description.
The numerical solutions for the axisymmetric free-
boundary problem of drop formation are closely fol-
lowed the previous studies (Richards, 1994; Zhang, 1999).
For a comprehensive description and analysis of the
numerical method, the reader is referred to Richards
(1994). The VOF/CSF algorithms are used to solve the
governing NavierStokes system and the boundary con-
ditions on a Eulerian rectangular, staggered mesh in
cylindrical geometry. In order to achieve large cell-wise
resolution of the free surface change and high computa-
tional eciency, a ner local mesh is used around the
drop region, whereas a coarser mesh is used in the ambi-
ent region. The complete process of drop formation is
simulated from the time a drop emerges from the tube
until its detachment from the tube, with continuous feed-
ing of the drop liquid at a specied ow rate.
3. Experimental approach
The experiments have been designed to obtain quanti-
tative information on the dynamics of drop formation of
a typical liquidliquid system of 2-ethyl-1-hexanol (2EH)
drops forming and breaking into distilled water in order
to compare the measurements with and verify the predic-
tions of the numerical model presented earlier. In the
experiments, attention is particularly paid to the evolu-
tion with time of the shape of a growing drop as it necks
and then breaks up and also to the creation of satellite
droplets subsequent to drop breakup.
The apparatus used to form drops has been described
elsewhere (Zhang and Basaran, 1995). It consists essen-
tially of a ne-capillary tube through which the liquid
used to form drops is delivered at a constant volumetric
ow rate by means of a liquid syringe pump (ATI Orion
M361). A liquid drop is formed at or emerges directly
from the tip of the tube. The capillary tubes used in the
experiments are chosen to have suciently large ratios of
inner and outer radii, with R/R
M
'0.8 that the wall
eects of the tubes can be neglected (cf. Zhang, 1999). The
capillary tubes are submerged in a water container
X. Zhang/Chemical Engineering Science 54 (1999) 17591774 1763
Fig. 2. Time sequence of shapes of a 2EH drop forming in quiescent water from a tube of R"0.16 cm at the liquid ow rate Q"5 ml/min, where t is
the time of the drop formation measured from the instant that the drop emerges from the tube.
having a suciently large inner radius as compared with
the tube radius with their ratio R
R
/R"62.5 so that the
wall eects of the container can be neglected.
An ultra-high-speed video camera by Kodak (Ektapro
Electronic Memory Motion Analyzer Model EM1012),
which is used for continuously capturing images of the
drop formation process, and the associated hardware for
recording, storing, and analyzing drop-shape data are
essential to the experimental study. The camera system is
composed of an intensied imager, which can record
1000 full images or 12,000 partial images per second and
allows rapid and accurate determination of the loci of
instantaneous interface proles from which the various
measures of drops are evaluated. The entire apparatus is
placed on a vibration isolation table from Newport.
The drop liquid is 2-ethyl-1-hexanol with the viscosity
and density of 0.089 g cm\ s\ and 0.83 g cm\, respec-
tively; and the ambient liquid is distilled water which
has a viscosity and density of 0.01 g cm\ s\ and
1.0 g cm\, respectively. The interfacial tension is meas-
ured to be 13.2 g s\ (Harris and Byers, 1989). All experi-
ments are performed at room temperature of 22$0.5C.
In a typical run, a steady ow with a desired rate is
established through the capillary and the ambient water
is quiescent. The water and 2EH are mutually saturated
prior to each experiment. The system is allowed to run
for about 5 mins before measurements are taken. A peri-
odic ow situation is then reached in which the drops
form, grow, and detach from the outlet of the capillary.
Since, in this dripping regime, primary and satellite drops
of uniform size are continuously created, this technique
provides a reliable and repeatable illustration of the
dynamics of drop formation. Reproducibility of results
for the drop shape and volume has been found to be
within 5% by making measurements under the same
conditions, but at dierent times.
4. Results and discussion
This section presents the results of an investigation of
all major eects governing the dynamics of drop forma-
tion, which are represented by Re, Ca, G, z, R
R
/R, and
u
M
/u
G
. Calculations are performed by systematically vary-
ing one parameter while keeping the others xed. The
quantitative results to be reported have thus been made
over wide ranges of the governing parameters to provide
insight into the dynamics and to classify drastically
dierent formation processes. In the nature of its im-
portance and complexity of drop breakup and satellite
generation, studies directed towards parameter correla-
tions and/or maps to characterize satellite generation are
underway in our laboratory. Whereas it is straightfor-
ward to vary any one of the dimensionless governing
parameters while holding all others xed in numerical
modeling, it is not possible to do so in laboratory experi-
ments. Therefore, a typical case of 2EH drop formation
in quiescent water is considered for comparative and
verication purposes. The experimental investigation for
drop formation has been performed by varying the vol-
umetric ow rate, Q, to cover certain ranges of the
Reynolds number and the capillary number: that is, the
Reynolds number varies as 10(Re(70, the capillary
number varies as 7;10\(Ca(4;10\, and the
gravitational Bond number and viscosity ratio are xed
at G"0.32 and z"8.9, respectively. Besides the surface
prole of forming drops, two measurements are used to
characterize the features of drop evolution and breakup:
the maximum length which a drop can reach prior to its
breakup,
B
, measured along the tube axis from the tip of
the tube to the apex of the drop (see Fig. 1); and the
volume of a primary (or breako) drop, . As shown in
Section 4.1, the excellent results obtained by comparisons
of numerical simulations and experimental measure-
ments give condence as to the accuracy of the numerical
model and its predictions.
4.1. Typical case
Fig. 2 shows a time sequence of interface proles ob-
tained by the numerical model (presented by the curved
images on the right side of each interface prole) and
experiments (the black images on the left side) for a 2EH
drop forming in quiescent, ambient water from a tube of
radius R"0.16 cm, at a constant liquid ow rate of
Q"5 ml/min. As the volume of the drop increases
1764 X. Zhang/Chemical Engineering Science 54 (1999) 17591774
successively with the continuous addition of the drop
liquid, the drop rises on the tip of the tube. The geometry
of the drop gradually transforms from nearly spherical to
pear shaped during its growth period. As the drop vol-
ume becomes suciently large that the buoyancy force
on the drop exceeds that of the surface tension (at
t'1.21 s), the drop necks rapidly and a large portion
starts to detach irreversibly from the tube. During the
necking sequence, the drop elongates rapidly and con-
tracts in the middle to form a liquid thread. The liquid
thread, which connects the detaching drop and the re-
mainder of the liquid on the tube, is stretched along with
rapid rise of the detaching drop and eventually breaks at
its upper end at t"1.2712 s, resulting in a free primary
drop. Immediately after the thread breaks at the upper
end, the newly freed end of the thread recoils due to the
unbalanced force of the surface tension. Meanwhile, the
cone-shaped liquid remaining on the tube tends to re-
trieve its apex and become spherical because of the capil-
lary force. A large curvature, therefore, develops at the
joining point of the liquid cone and the thread, leading to
a large capillary pressure at and high velocities of the
liquid out of that point (Zhang, 1999). Then, in a similar
fashion to the initial breakup, the thread breaks again,
this time at its lower end at t"1.275 s. As a result of the
thread rupturing at both ends sequentially, a satellite
droplet is created, lying between the primary drop and
the cone-shaped liquid remaining on the tube. In fact, the
satellite droplet is found to be very small, typically hav-
ing a volume less than 1% that of the primary drop.
Comparison of the numerical prediction with the experi-
mental measurements shows excellent agreement. The
numerical model accurately predicts the evolution of the
interface prole of the drop, particularly, the double
breakage of a liquid thread and generation of a satellite
droplet subsequent to the thread breakup.
The dynamic response of drop formation is especially
notable during the period of drop breakup. Fig. 3 shows
a summary of breakup features in terms of the maximum
lengths that a drop can attain at the instant when the
primary drop is about to detach (or the limiting lengths),

B
, and the volume of the primary drop, , as functions of
the ow rate, Q. The drop lengths and volumes presented
in Fig. 3 are obtained for 2EH drops forming in water
from a tube of the inner radius, R"0.16 cm. Three
interface proles of drops at the instant they are about to
break are included in Fig. 3 for ow rates of Q"5, 10,
and 15 ml/min. The corresponding experimental results
of
B
/R, /R, and photographic proles of drops are
also shown in Fig. 3 by points and black images on the
left, respectively. Again, excellent agreement between the
numerical simulation and experimental measurements
has been demonstrated in Fig. 3. As the ow rate in-
creases, the pressure at the tube exit and the mass (or
volume) of liquid accumulating in the cone directly above
the tube rise rapidly, due in part to the large axial
Fig. 3. Dimensionless limiting length and breako volume of 2EH
drops forming into quiescent water as a function of the ow rate from
a tube of R"0.16 cm.
momentum of the entering liquid and also to the resist-
ance that the thread exerts on the large amount of liquid
exiting the tube during the necking and drop detachment
steps. As a consequence of the pressure and mass buil-
dup, a larger elongation of the liquid cone occurs as
Q increases. As shown in the inserted proles in Fig. 3,
the increased deformation of the liquid cone retards, or
even abolishes, development of the liquid thread at the
larger Q, a determining factor that changes the features of
thread breakup and satellite generation. The behavior of
thread development and breakup is so interesting and
important in drop formation, particularly in the creation
of satellite droplets, these features will be discussed fur-
ther in a separated part. Fig. 3 also shows that the
breako volume of a drop increases with the ow rate, Q.
The rate of increase in the breako volume of a drop
decreases and tends to level o as Q becomes suciently
large. Indeed, as the ow rate continues to increase, over
a critical value, transition to jetting occurs and then the
volume of a primary drop decreases (Richards et al.,
1995). Moreover, in spite of the fact that the elongated
liquid cone and enlarged detaching drop may contribute
to the increase in the limiting length of a drop, the results
in Fig. 3 indicate only a slight increase in the limiting
length as Q increases. The primary reason for the weak
dependence of the limiting length on the ow rate is the
oblate deformation of the detaching drop caused by the
viscous eects of the ambient uid. Figure 3 depicts
clearly that the additional deformation or attening of
the top of a detaching drop, brought about by the ambi-
ent uid, increases as Qincreases because of the increased
X. Zhang/Chemical Engineering Science 54 (1999) 17591774 1765
Fig. 4. Variation of interaction time with the ow rate for 2EH drops
forming in quiescent water from a tube of R"0.16 cm. The times for
the three images in inserts of Q"5 ml/min are t"1.271, 1.2748, and
1.277 s, respectively. The times for the three images in inserts of Q"10
ml/min are t"0.643, 0.646, and 0.65 s, respectively.
Fig. 5. Dimensionless limiting length and breako volume of drops as
a function of the Reynolds number at Ca"0.01, G"1, z"1, and
u
M
/u
G
"0.
forming rate of the drop; therefore, it tends to counteract
the elongation of the liquid cone.
As a result of the double breakage of a liquid thread,
a satellite droplet is predicted by the numerical model
and is observed in experiments to be created under cer-
tain operating conditions (see Fig. 2). Detailed examina-
tion of the thread breakage and satellite creation reveals
the importance of deformation of the liquid cone that is
sessile on the tube and joins the lower end of the thread.
A quantitative measurement is dened to describe the
breaking feature of the liquid thread as the time interval
that elapses between the breakup of the two ends of the
thread, commonly referred to as interaction time. Fig.
4 shows the variation in the time interval, At, with the
ow rate obtained by the numerical model (curve) and
experiments (points) for 2EH drops forming into water
from a tube of R"0.16 cm. Two series of interface
proles of breaking drops, one obtained from the model
(the right side) and the other from experiments (the left
side), are inserted in Fig. 4 for ow rates of Q"5 and 10
ml/min. These inserted proles demonstrate typical situ-
ations of drop formation with and without satellite
generation. Evidently, as the ow rate increases, the in-
creased axial momentumof the entering liquid causes the
liquid cone to sustain a large axial elongation. It is more
dicult for the surface tension force to restore the cone-
shaped liquid mass to near-spherical form when compet-
ing with increasing inertial force and then to develop
a localized, large curvature at the point where the cone
joins with the thread. Therefore, more time is required for
the secondary rupture of the thread at the joining point.
As shown in Fig. 4, the interaction time, At, increases
with elevated ow rate. When the ow rate is suciently
high, the secondary breakup of the thread can no longer
occur before the thread recoils and coalesces with the
liquid cone. In that case, no satellite droplet is created
and the interaction time becomes innity, as indicated in
Fig. 4. Apparently, whether or not double breakage of
a thread occurs and a satellite droplet is created are
determined substantially by the relative importance of
the inertial and capillary forces, which can be measured
by a Weber number, e"ReCa. For the present study
of the 2EH drops forming into water at dierent ow
rates, the results from numerical model and experi-
ments show that the threshold of the ow rate is
8.5$0.5 ml/min, which corresponds to a critical Weber
number of 0.031.
4.2. Eect of the Reynolds number
Fig. 5 demonstrates computational predictions of the
variation of the dimensionless limiting length,
B
/R, and
the volume of a breako drop, /R, as a function of
the Reynolds number, while holding other governing
parameters xed as Ca"0.01, G"1, and z"1. For
numerical results shown in Sections 4.24.5, a quiescent
ambient uid is considered with a outer container of the
xed size of R
R
/R"3. An obviously smaller value of R
R
/R
has been used in numerical calculations in comparison
with that in experiments in order to attain a high com-
putational eciency. In fact, as shown in Section 4.6, the
wall eects of an ambient uid container can be neglected
as R
R
/R*2.5. Moreover, the boundary condition of
rigid free slip is used in calculations shown in Sections
4.24.5 for the container wall to further minimize the wall
eects. The eects of axial ow of the ambient uid are
shown in Section 4.7. Four proles of interface shapes of
1766 X. Zhang/Chemical Engineering Science 54 (1999) 17591774
Fig. 6. Evolution with time of interface proles of detaching drops at
Reynolds numbers of (a) Re"1 and (b) Re"10, with Ca"0.01,
G"1, z"1, and u
M
/u
G
"0.
drops at the instant they are about to detach are inserted
in Fig. 5 for Re"0.1, 1, 10, and 100. As expected, the
limiting length,
B
, of a detaching drop increases mono-
tonically as the Reynolds number increases, due prim-
arily to the increase in the inertial force to elongate the
liquid cone remaining on the tube (see Fig. 5). Moreover,
the larger axial momentum of the liquid leaving the tube
forces more liquid to squeeze into the drop and causes
the volume of the detaching primary drop to increase as
Re increases. At a certain critical Re, transition to jetting
occurs, at which time the volume of the primary drop
attains its maximum, as shown in Fig. 5.
As the Reynolds number increases, the dynamic eects
of the ambient uid becomes pronounced on a detaching
drop because of the increased speed of the drop rising in
the medium. The increased tangential viscous stress on
a detaching drop from the ambient uid causes the drop
to change its shape from prolate (at Re)1) to oblate
(at Re&10) and, then, to dimpled ellipsoidal
(at Re*100), as shown in inserted drop proles in Fig. 5.
Fig. 6 demonstrates breakup features of drops at
Reynolds numbers of Re"1 and 10, while holding other
governing parameters xed as Ca"0.01, G"1, z"1,
and u
M
/u
G
"0. At low Reynolds numbers (Fig. 6a), the
liquid thread ruptures at both ends sequentially because
of the stretching under the weight of the detaching pri-
mary drop and development of large local mean curva-
tures at the joining points of the thread with the two
bodies of liquid. Peaks of capillary pressure are then
induced inside the thread at its two ends and force the
liquid out of these breaking points, accelerating thread
rupture (see also Zhang, 1999). Consequent to the double
breakage of the thread, a satellite drop is generated. In
contrast, at the large Reynolds number (Fig. 6b), the
liquid thread is no longer obvious as a result of the
increasing elongation of the liquid cone on the tube
during the drop necking sequence. Following breako of
the primary drop, the tip of the liquid cone rebounds and
forms a bulbous head under unbalanced capillary forces.
Eventually, as shown in Fig. 6b, the bulbous head merges
into the liquid cone and a new drop emerges. No satellite
droplet is generated at this large Reynolds number.
4.3. Eect of the capillary number
The eects of the capillary number on drop breakup, in
terms of the dimensionless limiting length,
B
/R, and
detaching volume of the drop, /R, are shown in Fig.
7 with other parameters xed at Re"1, G"1, z"1,
and u
M
/u
G
"0. Four drop proles at the instant of detach-
ment are inserted for Ca"0.001, 0.01, 0.05, and 0.1.
Similar to the cases of drop forming into air (Zhang,
1999), the limiting length,
B
, and the volume of the
primary drop, , are shown to increase monotonically as
Ca increases.
Noteworthy, as the capillary number becomes su-
ciently large (such as Ca"0.1 in Fig. 7), the capillary (or
surface tension) force cannot restore the detaching pri-
mary drop to spherical shape and, therefore, a large
curvature can no longer be developed at the joining point
of the detaching drop and the liquid thread. Moreover,
the increased viscous force relative to the capillary force
holds a relatively long liquid thread prior to its rupture.
Therefore, in contrast to drop formation at small capil-
lary numbers, interface disturbances may cause the
thread to rupture at its middle portion prior to the
capillary-driven breakup of the thread at its upper end
joining the detaching primary drop, as shown in Fig. 8
for a series of interface proles of a breaking drop under
conditions of Ca"0.1 and all other conditions the same
as those in Fig. 7. As a result of the sequential rupture of
the thread at its top portion, a small satellite droplet is
created. The rest of the thread then recoils rapidly and
merges into the liquid cone. A similar breakage feature of
a drop of a highly viscous liquid has been observed in
experiments with glycerol drops (Zhang and Basaran,
X. Zhang/Chemical Engineering Science 54 (1999) 17591774 1767
Fig. 7. Dimensionless limiting length and breako volume of drops as
a function of the capillary number at Re"1, G"1, z"1, and
u
M
/u
G
"0.
Fig. 8. Evolution with time of interface proles of detaching drops at
Re"1, Ca"0.1, G"1, z"1, and u
M
/u
G
"0.
1995). At intermediate range of the capillary number, the
eects of the viscous force is no longer suciently great
to cause a long liquid thread, and the surface tension
force is still small relative to the inertial force, leading to
drop formation at a small Weber number. Similar to the
situation of large Reynolds numbers, the development of
a thread is suppressed by the large elongation of the
liquid cone on the tube, and obviously, the double break-
age of the thread and satellite generation will no longer
occur (cf. Fig. 6b). At a very small capillary number, the
capillary force becomes dominant and causes the double
Fig. 9. Dimensionless limiting length and breako volume of drops as
a function of the gravitational Bond number at Re"1, Ca"0.01,
z"1, and u
M
/u
G
"0.
breakage of a thread at its two ends and subsequent
satellite generation (cf. Fig 6a).
4.4. Eect of the gravitational Bond number
The gravitational Bond number, G, has been varied in
calculations by changing the magnitude of the density
dierence between the two phases, Aj. Fig. 9 shows the
variation of the dimensionless volume of a primary drop,
/R, and the limiting length of the drop at the instant of
its detachment,
B
/R, as functions of the gravitational
Bond number with the other parameters xed at Re"1,
Ca"0.01, z"1, and u
M
/u
G
"0. In Fig. 9, instantaneous
interface proles of detaching drops are inserted for
G"0.1, 1, 3, and 10. Fig. 9 shows monotonical decrease
in the volume of the detaching drop and the limiting
length with increasing values of G. In particular, the
volume of the primary drop is shown to be a strong
function of the magnitude of the gravitational force, or G,
as G(2. The rate of decrease in the breako volume
with increasing Gfalls and tends to level o as Gbecomes
suciently large and the volume of the primary drop is
suciently small, indicating the diculty of reducing the
volume of breako drops. Because of the large driving
force presented to pull the liquid out of the tube at a large
value of G, the drop is stretched immediately after its
emergence from the tube and breaks rapidly. The large
pulling eect of gravity signicantly reduces the volume
of liquid remaining on the tube during drop breako, and
the development of a liquid thread is no longer as evident
1768 X. Zhang/Chemical Engineering Science 54 (1999) 17591774
Fig. 10. Dimensionless limiting length and breako volume of drops as
a function of the viscosity ratio at Re"1, Ca"0.01, G"1, and
u
M
/u
G
"0.
as with small G values. Obviously, at such large values of
G, the double breakage of the thread and satellite genera-
tion cannot be expected, as predicted by the numerical
model. As a result of increasing velocity of drop de-
tachment at large G, the increasing viscous force of the
ambient uid causes the detaching drop to deform to
a dimpled ellipsoidal shape; a similar phenomenon
occurred as Re is increased.
4.5. Eect of the viscosity ratio of two phases
Despite the fact that liquid viscosities of drop and
continuous phases have little inuence on the volume of
a breako drop (Kumar and Kuloor, 1970), the viscous
force has been found to aect drop necking and breakup
greatly, particularly the development, extension, and
breakup of a liquid thread and generation of satellite
droplets, as briey discussed in Section 4.3. To address
the viscous eects on drop formation, the viscosity ratio
of the drop liquid to the ambient uid, z, is used and has
been changed by varying the viscosity of the drop liquid
alone in present numerical simulations. Fig. 10 shows the
variation of dimensionless breako volume, /R, and
limiting length,
B
/R, as a function of z with other gov-
erning parameters xed at Re"1, Ca"0.01, G"1, and
u
M
/u
G
"0. Instantaneous interface shapes when drops are
about to break are also displayed in Fig. 10 for z"0.01,
1, 5, and 10. As the viscosity ratio (or the viscosity of the
drop liquid) increases, the dimensionless limiting length,

B
/R, increases substantially due to the signicant in-
crease in lengths of the liquid thread and cone, as shown
in the inserted drop proles in Fig. 10. Viscosity of the
drop liquid plays a great role in stabilizing the drop
interface, which makes possible greater thread elongation
and extension by damping, and even eliminating, oscilla-
tions of the interface. As a result of the increased exten-
sion of the thread, a longer elapsed time is required for
drop breakup and, thus, more liquid is retained in the
outlet of the tube, resulting in a longer liquid cone on the
tube. In spite of its weak dependence on the viscosity
ratio, z, the volume of breako drops varies with z in
a peculiar way it exhibits a maximum in the neighbor-
hood of z"1, as shown in Fig. 10, for the ranges of Re,
Ca, and G studied. This feature of nonmonotonical vari-
ation of the volume of breako drops with the liquid
viscosity is similar to that observed in early experiments
(Kumar and Kuloor, 1970).
Along with the increased elongation and extension of
a liquid thread at large viscosity ratio, the creation of
satellite droplets subsequent to the thread break is
Fig. 11. Evolution with time of interface proles of detaching drops at
viscosity ratios of (a)z"0.01 and (b) z"5, with Re"1, Ca"0.01,
G"1, and u
M
/u
G
"0.
X. Zhang/Chemical Engineering Science 54 (1999) 17591774 1769
expected to be interesting and complicated. Fig. 11 shows
two series of interface proles of drops. Each takes a close
look at thread breakup and creation of satellite droplets,
with viscosity ratios of z"0.01 and 5 and all other
conditions the same as those in Fig. 10. Apparently, the
stabilizing eects of the larger viscosity of the drop liquid
(or the viscosity ratio) leads to the production of a longer,
thicker liquid thread, as shown in Fig. 11b. Immediately
after its rst breakup at the upper end, the thread recoils
under the unbalanced capillary force to form a round
knob at its top; this leads to the formation of a main,
large satellite droplet following the secondary breakup of
the thread. As a result of the large viscous force at the
large viscosity ratio, a noticeably large interaction time
(or the time interval between the two breakups), At, is
observed, namely that At"14 milliseconds for z"5
(Fig. 11b) as compared with At"1.6 milliseconds at the
small viscosity ratio of z"0.01 (Fig. 11a). Therefore, as
shown in Fig. 11b, waves have sucient time to develop
and propagate on the thread surface. Along with the
thread recoil and the secondary breakage, these waves
grow and eventually break the thread when, at certain
points, the amplitude of the waves is compatible with the
radius of the thread. This leads to the creation of a sub-
satellite droplet that has a much smaller size and lies
between the main satellite and the cone-shaped liquid
remaining on the tube. In contrast, the development and
extension of a liquid thread are no longer so evident at
the small viscosity ratio (Fig. 11a). As a result, a consider-
ably smaller satellite droplet is created subsequent to
double breakage of the thread. Thus, we expect the cre-
ation of satellite droplets to cease or to be unmeasurable
as the viscosity ratio becomes extremely small, for
example, in the case of bubble generation in liquids (cf.
Oguz and Prosperetti, 1993).
4.6. Wall eects of the outer container
When drops are formed in a bounded medium, the
dynamics of their deformation and breakup is aected by
the wall of the ambient uid container. Wall eects have
always been present, to a greater or lesser extent, as
a result of disturbances of the ambient uid by forming
drops. Fig. 12 depicts computational results of the wall
eects on drop formation, which are described by the
variation of dimensionless breako volume, /R, and
limiting length,
B
/R, of a drop as a function of the ratio
of the inner radius of the container to that of the capillary
tube, R
R
/R. The results in Fig. 12 are obtained by holding
other governing parameters xed, with Re"1,
Ca"0.01, G"1, z"1, u
M
/u
G
"0, and the boundary
condition of rigid no slip for the container wall. Four
interface proles of detaching drops are inserted in
Fig. 12 for R
R
/R"1.35, 1.4, 2, and 4. Quantitative
measurements of /R and
B
/R in Fig. 12 indicate
clearly that the wall eects become signicant and the
Fig. 12. Dimensionless limiting length and breako volume of drops as
a function of the dimensionless inner radius of the ambient uid
container at Re"1, Ca"0.01, G"1, z"1, and u
M
/u
G
"0.
container radius, R
R
, becomes a determining factor gov-
erning the deformation and breakup of a forming drop at
small values of R
R
/R. The presence of the wall causes an
increased deformation of a detaching drop, with elonga-
tion occurring in the axial direction to yield an approx-
imately prolate ellipsoid shape (see Fig. 12). At a value of
R
R
/R suciently small that the drop size is compatible
with the outer boundary, the drop lls most of the con-
tainer cross section and a slug ow regime results.
Apparently, the increased viscous eects of the ambient
uid delay drop breakup considerably, resulting in a sig-
nicant increase in the breako volume of a drop as R
R
/R
decreases. As a result of increased volume of the detach-
ing drop and its axial elongation, the limiting length of
the drop increases dramatically when the slug ow oc-
curs at R
R
/R(1.5, as shown in Fig. 12. It is expected that
as the distance between the wall and a drop, or R
R
/R,
increases, the wall eects will decrease. Beyond a certain
value of R
R
/R so that the drop no longer sees the
container wall, the wall eects become negligible. As
illustrated in Fig. 12, when the radius ratio R
R
/R exceeds
a value of about 2.5, the variation of breako volume and
limiting length of the drop with R
R
/R becomes virtually
unmeasurable, indicating negligible eects of the con-
tainer wall.
Besides the change in shape of a detaching drop with
R
R
/R, the inserted interfacial proles of breaking drops in
Fig. 12 indicate clearly that the shape of the liquid cone
remaining on the tube varies as well. As R
R
/R decreases,
1770 X. Zhang/Chemical Engineering Science 54 (1999) 17591774
Fig. 13. Evolution with time of interface proles of detaching drops at
Re"1, Ca"0.1, G"1, z"1, R
R
/R"1.35, and u
M
/u
G
"0.
the increased viscous eects delay the drop breakup, in
a manner similar to that of increasing z (discussed in
Section 4.5). Therefore, more liquid is accumulated in the
liquid cone, causing the cone shape to change from con-
cave to convex prior to drop breakup. Fig. 13 depicts this
breakup feature of a drop at the slug ow regime with
R
R
/R"1.35 and all other conditions the same as those in
Fig. 12. Unlike drops forming into an unbounded me-
dium (cf. Fig. 6a), no measurable thread is developed
between the breako drop and the liquid cone and hence
the drop breaks sharply at its middle portion, as the
result of the accumulation of a large volume of liquid in
and convexity of the liquid cone at such a small value of
R
R
/R. Obviously, in such a situation, a satellite droplet
can no longer be created subsequent to the drop breakup.
Moreover, a comparison of the breakup features of drops
within a tightly bounded medium (Fig. 13) with those in
an innite medium (Fig. 6a) indicates clearly that the
breakup time of the drop, along with its breako volume
and limiting length, is signicantly increased by the wall
eects as R
R
/R decreases.
4.7. Eect of the external ow
Flowing an ambient uid concurrently with the dis-
persed phase has been used as an eective and practical
method for reducing the volume (or size) of breako
drops by accelerating drop detachment with an
Fig. 14. Evolution with time of the dimensionless length of drops
forming in an ambient uid which ows at dierent characteristic
velocities of u
M
/u
G
"0, 50, and 200 at Re"1, Ca"0.1, G"1, z"1,
and R
R
/R"2.
additional viscous shear force (Clift et al., 1978). Besides,
an external uniform ow is found to vary the interface
prole and breaking length of a drop during drop forma-
tion considerably at innitesimal inertial eects (Zhang
and Stone, 1997). Fig. 14 presents the entire history of the
formation of three drops that are each subjected to an
external ow of dierent velocities. The drop evolution is
presented quantitatively in terms of the drop length, ,
nondimensionalized by the inner radius of the tube, R.
The results shown in Fig. 14 are calculated for the three
situations of drop forming in an ambient uid for which
fully developed ows are applied concurrently at dier-
ent average velocities of u
M
/u
G
"0, 50, and 200, respec-
tively. Other controlling parameters are held constant:
Re"1, Ca"0.01, G"1, z"1, R
R
/R"2, and R
M
/R
"1.1 (see Fig. 1). Evidently, the external ow of the
ambient uid plays an important role in drop evolution
and breakup, as clearly depicted by the two series of
instantaneous proles of drops at u
M
/u
G
"0 (the lower
series) and 200 (the upper series), respectively, in Fig. 14.
When injected into a container with a specied ow rate
(or average velocity), the ambient uid interacts with
a forming drop and brings about an additional hy-
drodynamic force on it. Initially, at the growing stage, the
drop of small volume grows and rises slowly at the tip of
the tube. The entering ambient uid, if possessing a su-
ciently high velocity as shown in Fig. 14 with u
M
/u
G
"200,
passes over the drop rapidly and expands, resulting in
X. Zhang/Chemical Engineering Science 54 (1999) 17591774 1771
a reduction in its velocity and, consequently, an increase
in the ambient pressure (Batchelor, 1967). As a result of
the large pressure on its top, the drop deforms and
attens at its tip; then the drop length, , is measured to
be shorter than that in quiescent or slowly owing me-
dium, as shown in Fig. 14 at t(0.12 s. Along with the
increase in the interface area of the drop at longer time,
the viscous force on the drop interface induced by the
external ow becomes important and stretches the drop
into a spheroid in the axial direction, whereas the drop is
seen to be pear shaped under the gravitational force
when the external ow is vanishing, as shown in the
inserted proles in Fig. 14. Diering from body forces,
which pull the drop liquid away from the tube, such as
the gravitational force (cf. Fig. 8), the external ow acts
on the drop interface as a viscous shear stress, which
stretches and elongates the drop and, therefore, leads to
a great increase in the drop length. At the extremely high
velocity of the external ow, such as that of u
M
/u
G
"200 in
Fig. 14, a thread about three time longer than the tube
radius is developed during necking and breakup of the
drop. It is expected that a large satellite droplet, or
multiple satellites, will be created subsequent to the
thread breakup. This phenomenon is similar to that
observed in the formation of conducting or insulating
drops driven by an external electric eld, which acts with
electric charges and induces an electrostatic force on
drop surfaces (Zhang and Basaran, 1996).
Fig. 15 shows variations of breako volume and limit-
ing length of drops as a function of the average velocity of
an external ow, u
M
, normalized by the average velocity of
the drop liquid ejecting from the tube, u
G
under condi-
tions that are otherwise the same as those used in Fig. 14.
Three typical interface proles of drops at the instant
when they are about to detach are inserted for external
ows of u
M
/u
G
"1, 100, and 200. As expected, the external
ow, acting as an additional force assisting to break up
drops, reduces the breako volume of the drops. As the
velocity of the external ow (or its dimensionless term
u
M
/u
G
) increases, the volume of breako drops decreases
and, correspondingly, the dripping rate increases mono-
tonically. It is noteworthy that since the volume of
breako drops decreases as the velocity of an external
ow increases, the increase in the limiting length,
B
/R,
with u
M
/u
G
(shown in Fig. 15) is obviously attributed to an
increase in the thread length. The concurrent increase in
drop elongation and decrease in the volume of breako
drops greatly aect the dynamics of thread rupture and
the subsequent generation of satellite droplets. Fig. 16
depicts the feature of a drop breaking into a owing
medium with u
M
/u
G
"200 and all other conditions the
same as those in Fig. 14. Because of the large shear stress
on the drop interface generated by the external ow, the
drop produces a long, thick thread prior to its breakup.
Immediately following the detachment of a primary
drop, the newly freed tip of the thread tends to retract
Fig. 15. Dimensionless limiting length and breako volume of drops as
a function of the dimensionless velocity of the owing media at Re"1,
Ca"0.01, G"1, z"1, and R
R
/R"2.
Fig. 16. Evolution with time of interface proles of detaching drops at
Re"1, Ca"0.1, G"1, z"1, R
R
/R"2, and u
M
/u
G
"200.
downward under the unbalanced capillary force. How-
ever, owing to the great opposition from the external
ow, the tip of the distorted thread reaches a minimum
height and then starts to ascend at t"0.154 s. As it does
so, a bulbous head develops at the tip of the thread
and continues to expand. A secondary neck then be-
gins to develop immediately below the head at
t"0.17 s. Eventually, a second breakup occurs to create
1772 X. Zhang/Chemical Engineering Science 54 (1999) 17591774
a secondary drop with the volume almost 10% of that of
the primary drop. This is in contrast to the situation with
quiescent or slowly owing ambient uid where the sec-
ondary (or satellite) drop has a volume less than 1% of
that of the primary drop (cf. Fig. 6(a)). The large dier-
ence in the thread breakup is also indicated by dierence
in the interaction time. Namely, in a quiescent ambient
uid, the time interval that elapses between the two
breakups is about 3.6 ms as shown in Fig. 6a, while with
an external ow of u
M
/u
G
"200, it takes more than 17 ms
(see Fig. 16). Hence, for the latter case, the secondary
breakup can be regarded as a breakup entirely separate
from the rst one; evidently the liquid jetting occurs to
generate drops of nonuniform sizes.
5. Concluding remarks
Through both theoretical and experimental means, we
have investigated the eects of inertial, viscous, gravi-
tational, and surface tension forces on drop formation.
The emphasis has been on determining the importance of
the dynamic eects of an ambient viscous uid on drop
evolution and breakup. The VOF/CSF numerical algo-
rithm, on which our numerical model is based, allows
calculations to pass the breaking point during drop
formation continuously without numerical modications
to overcome the singular nature of the interface rupture.
This feature makes it possible to characterize numerically
the double breakage of a liquid thread and the sub-
sequent generation of satellite droplets under dierent
conditions. Reassuringly, the numerical predictions show
excellent agreement in comparison with experimental
measurements for a typical liquidliquid system of 2EH
drops forming in water.
According to the results obtained in the present study,
the maximum (or limiting) length that a drop attains
prior to its breakup and the volume of a drop that breaks
o from the tube increase signicantly with increased
Reynolds and capillary numbers and with decreased
gravitational Bond number of the drop liquid. During
necking and breakup of a drop, a liquid thread, which
connects to a detaching drop and a cone-shaped liquid
mass sessile on the tube, develops and stretches. Under
certain conditions, the thread breaks at both ends se-
quentially, resulting in satellite droplets. When drops
form into a quiescent ambient uid, whether the satellite
droplet is created subsequent to the thread breakup is
substantially dependent on the relative importance of the
inertial and surface tension forces. Dynamic eects of an
ambient uid have been shown to play an important role
in drop formation. In particular, the increasing wall ef-
fects of an ambient uid container cause a signicant
increase in the volume of breako drops and a large
reduction in the length of the liquid thread a factor that
may alter the possibility of creation of satellite droplets.
In contrast, an external ow of the ambient liquid leads
to a great increase in the thread length and radius but to
a decrease in the volume of breako drops. Moreover,
as a result of viscous damping of the interface distur-
bances, a long, thin thread is developed when the ratio of
the viscosity of the drop liquid to that of the ambient
uid becomes suciently large, resulting in the genera-
tion of multiple satellite droplets subsequent to drop
breakup.
Acknowledgments
This work was supported by the Division of Chemical
Sciences, Oce of Basic Energy Sciences, U.S. Depart-
ment of Energy under contract DE-AC05-96OR22464
with Lockheed Martin Energy Research Corp. The
author performed this study when he was working at
Oak Ridge National Laboratory, Tennessee. The author
thanks Dr. J. R. Richards of Dupont for providing the
details of his calculations.
References
Batchelor, G.K. (1967). An introduction to uid dynamics, New York:
Cambridge at the University Press.
Berg, J.C. (1993). ettability. New York: Marcel Dekker.
Bird, R.B., Stewart, W.E., & Lightfoot, E.N. (1960). ransport phe-
nomena. New York: Wiley.
Bogy, D.B. (1979). Drop formation in a circular liquid jet. Annual
Review of Fluid Mechanics, 11, 207228.
Brackbill, J.U., Kothe, D.B., & Zemach, C. (1992). A continuummethod
for modeling surface tension. Journal of Computational Physics 100,
335354.
Clift, R., Grace, J.R., & Weber, M.E. (1978). Bubbles, drops, and par-
ticles. New York: Academic Press.
Cram, L.E. (1984). A numerical model of drop formation. In J. Noye,
& C. Fletcher (Eds.), Computational techniques and applications:
CAC-83. Amsterdam: Elsevier.
Eggers, J., &Dupont, T.F. (1994). Drop formation in a one-dimensional
approximation of the NavierStokes equation. Journal of Fluid
Mechanics 262, 205221.
Harris, M.T., & Byers, C.H. (1989). An advanced technique for inter-
facial tension measurements in liquidliquid systems. ORNL/TM-
10734.
Heertjes, M.P., de Nie, L.H., & de Vries, H.J. (1971). Drop formation
in liquidliquid systems I Prediction of drop volumes at
moderate speed of formation. Chemical Engineering Science 26,
441449.
Hirt, C.W., & Nichols, B.D. (1981). Volume of uid (VOF) method for
the dynamics of free boundaries. Journal of Computional Physics 39,
201225.
Kothe, D.B., & Mjolsness, R.C. (1992). Ripple: A new model for
incompressible ows with free surfaces. AIAA Journal 30,
26942700.
Kumar, R., & Kuloor, N.R. (1970). The formation of bubbles and drops.
Advances in Chemical Engineering 8, 255368.
Michael, D.H. (1981). Meniscus stability. Annual Review of Fluid Mech-
anics 13, 189215.
X. Zhang/Chemical Engineering Science 54 (1999) 17591774 1773
Oguz, H.N., & Prosperetti, A. (1993). Dynamics of bubble growth and
detachment froma needle. Journal of Fluid Mechanics 257, 111145.
Peregrine, D.H., Shoker, G., & Symon, A. (1990). The bifurcation of
liquid bridge. Journal of Fluid Mechanics 212, 2539.
Richards, J.R. (1994). Fluid mechanics of liquid-liquid system. Ph.D.
dissertation, University of Delaware.
Richards, J.R., Beris, A.N., & Lenho, A.M. (1995). Drop formation in
liquidliquid systems before and after jetting. Physics of Fluids, 7,
26172630.
Scheele, G.F., & Meister, B.J. (1968). Drop formation at low velocities
in liquidliquid systems: Part I. Prediction of drop volume.
A.I.Ch.E. Journal 14, 919.
Schulkes, R.M.S.M. (1993). Nonlinear dynamics of liquid columns:
A comparative study. Physics of Fluids A, 5, 21212130.
Schulkes, R.M.S.M. (1994). The evolution and bifurcation of a pendant
drop. Journal of Fluid Mechanics 278, 83100.
Shi, X.D., Brenner, M.P., & Nagel, S.R. (1994). A cascade of structure in
a drop falling from a faucet. Science, 265, 219222.
Zhang, D., & Stone, H.A. (1997). Drop formation in viscous ows at
a vertical capillary tube. Physics of Fluids 9, 22342242.
Zhang, X. (1999) Dynamics of growth and breakup of viscous
pendant drops into air. Journal of Colloid and Interface Science,
In press.
Zhang, X., & Basaran, O.A. (1995). An experimental study of dynamics
of drop formation. Physics of Fluids 7, 11841203.
Zhang, X., & Basaran, O.A. (1996). Dynamics of drop formation from
a capillary in the presence of an electric eld. Journal of Fluid
Mechanics 326, 239263.
1774 X. Zhang/Chemical Engineering Science 54 (1999) 17591774