You are on page 1of 12

Chemical Engineering Science 59 (2004) 21892200

www.elsevier.com/locate/ces
Modelling of droplet breakage probabilities in an oscillatory
baed reactor
Dimitri Mignard, Lekhraj Amin, Xiong-Wei Ni

Centre for Oscillatory Baed Reactor Application (COBRA), Chemical Engineering, School of Engineering and Physical Sciences,
Heriot-Watt University, Edinburgh EH14 4AS, UK
Received 22 May 2003; received in revised form 25 February 2004; accepted 25 February 2004
Abstract
The breakage of droplets dispersed in a continuous aqueous phase determines the performance of many mixing devices and reactors that
rely on eective contact between two phases, e.g. emulsion mills, liquidliquid extraction columns, stirred tank reactors and Oscillatory
Baed Reactors. Quantitative knowledge of the mechanisms involved in the breakage provides parameters for design and prediction. In
the work presented here, oil was dispersed in water in a continuous OBR, and a High Speed Camera was used to record the events of
breakage of individual oil droplets and probabilities of breakage were estimated. It was conrmed that breakage was more sensitive to
changes in the amplitude of oscillation than in the frequency of oscillation. A novel integral model was developed based on an analysis
of the total work eected on the deforming droplet in order to interpret the results. The quantitative results from direct observation were
compared to the model predictions. The model with tted parameters was nally extrapolated to smaller diameters, in an attempt to predict
the critical drop diameter for breakage.
? 2004 Elsevier Ltd. All rights reserved.
Keywords: Droplet breakage probability; Oscillatory Baed Reactor; Dynamic balance; Dispersion; Stretch rates
1. Introduction
There has been much research on the general subject of
breakage of oil droplets, and a number of reviews can be
found in the literature (e.g. Briscoe et al., 1999). Gourdon
et al. (1994) also provided a useful overview of droplet
breakage within the context of Population Balance Mod-
elling for solvent extraction column. The following is a short
summary of the results from papers that are relevant to the
present work.
1.1. The form of the ow eld
Pioneering work on the subject was rst performed by
Taylor (1934): his four roll-mill machine produced sim-
ple and dened 2-D ow elds, such as simple shear
and plane convergent hyperbolic ow. Simple tangential
shear was found to induce deformation of droplets, but no

Corresponding author. Tel.: 44-131-451-3781; fax: 44-131-451-3129.


E-mail address: x.ni@hw.ac.uk (X.-W. Ni).
URL: http://www.cobra.hw.ac.uk
continuous stretch and no breakage was possible for sys-
tems with a high viscosity ratio z 4 (z =j
d
}j
c
, in which
j
d
and j
c
, are the viscosities for the dispersed phase and
the continuous phase, respectively). However, more com-
plex 2-D ow could always provoke continued stretch and
breakage, provided that the strain rates were above a criti-
cal value. In connection with this threshold rate, a critical
capillary number Ca
cr
has since then be dened for dier-
ent viscosity ratios, e.g. Grace (1982), Bentley and Leal
(1986), and Stone and Leal (1989b). One of the main prob-
lems that investigators have been facing ever since is that
this critical number varies with the nature and history of
the ow for the particular industrial application. Mietus
et al. (2002) used a transparent horizontal Couette ow cell
to distinguish between the eects of extensional and shear
strains, an approach motivated by the fact that most disper-
sion and emulsion devices rely on a complex combination
and timing of dierent ow patterns. One of their main nd-
ings was that, although the apparatus would provide capil-
lary numbers well in excess of the critical values obtained
in steady ow, the resulting large deformations within these
regions of strong ow would not directly lead to break-
age if the residence time was not long enough. However,
0009-2509/$ - see front matter ? 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2004.02.012
2190 D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200
deformation induced there may then enable the droplet to be
broken up by weaker ow, and therefore an understanding
of the kinetics of deformation in steady ow is needed, both
for near-spherical droplets and for elongated ones.
Attempts at introducing such a time-dependent deforma-
tion of spherical droplets for the theoretical prediction of
burst were made by a few researchers, usually by expressing
the shape and the rate of deformation tensor as expansions
to rst or even second order in a small capillary number Ca.
For example, Barth es-Biesel and Acrivos (1973) combined
this approach with linear stability theory, and succeeded
to some extent depending on the particular type of ow
(3-D extensional, hyperbolic, and simple shear). Hakimi and
Schowalter (1980) used an expansion in Ca to investigate
the steady-state shapes reached by the droplets rather than
break-up, but they distinguished between vorticity and shear
within the ow, which allowed them to investigate the drop
orientation independently from the deformation. An eccen-
tric disk-rheometer could generate ows with independently
adjustable levels of shear and vorticity, and the resulting ex-
perimental data was reasonably matched by the predictions
of shapes and orientation from the theory for values of Ca
as high as 0.40.7. Their results also showed that vorticity
alone can aect shape in a complex way. One may presume
that it acts indirectly by presenting a dierent orientation of
the elongated drop to the ow velocity gradients. It is also
known that periodical reorientation of a droplet in a sim-
ple shear ow may considerably increase the stretching of a
droplet, as reported by Elemans et al. (1993) and others. An-
other strand of research focussed on the breakup of threads
in already very elongated droplets: an initial study follow-
ing Taylors work (1934) was done by Tomotika (1935),
who showed that capillary perturbation waves developing
along a thread could grow to the extent that break-up would
occur.
1.2. Unsteady ow
All the results mentioned so far were mostly concerned
with steady ows, yet it has been known since Taylors work
(1934) that eects of transient ow may be very important
too. Insightful studies on the subject were made by Stone
et al. (1986) with their computer controlled four-roll mills.
They generated ow of a preset form to any degree between
pure extensional shear and parallel shear. A droplet was sub-
jected to increasing velocity gradients until deformation oc-
curred when the critical capillary number was reached. The
velocity gradient was then maintained at the critical value
(with the droplet Reynolds number still of the order of 10
4
only), until the ow was suddenly stopped. The droplet then
relaxed and sometimes broke apart, with droplet ends rais-
ing up and separating. Internal ow within the droplet was
probably driven by surface tension gradients (variations in
the local curvature radius and hence in the Laplace pres-
sure), and this ow was invoked to explain this breakage by
end-pinching. The authors provided diagrams giving the
critical length to diameter ratio L}d of the deformed drop,
for which breakup would spontaneously happen when the
extensional ow was stopped. This critical ratio was found
to depend only on the viscosity ratio. Capillary wave theory
was found to be relevant only when the droplet length was
initially extended to a suciently high value, allowing the
time scale for capillary wave propagation to catch up with
that required for the relaxation of the droplet thread-like
fragments. This explanation was later supported by numeri-
cal simulation of the relaxation of elongated droplets (Stone
and Leal, 1989a). In addition, further experiments and simu-
lations were also performed, in which the ow was changed
in a stepwise manner, either from critical to subcritical, or
with a change in the amount of shear or vorticity, or with
both types of change combined (Stone and Leal, 1989b).
Elemans et al. (1993) also investigated the inuence of tran-
sient ow in a Couette ow apparatus that generated simple
shear ow. For Newtonian uids dispersed in a Newtonian
continuous phase, breakup was accounted for by the Stone
and Leal model (1989a) if 6 L}d15, and the Tomotika
model (1935) if L}d15. Their paper presented a very use-
ful procedure to calculate the L}d ratio achieved and the
time required for break-up in simple shear ows above Ca
cr
,
depending on the residence times, shear rates and ratios of
the local capillary numbers to Ca
cr
. Although no such re-
lationship would be available for more complex ow, their
technique might nevertheless prove useful as a shortcut one.
1.3. Turbulence
One particular case of unsteady ow is turbulence, in par-
ticular isotropic turbulence. Theory there has been applied
to stirred tanks, and also to mixing devices such as per-
forated sieve columns of either the pulsed or reciprocated
type. The advantage of applying the theory of isotropic tur-
bulence lies in its generality, and the relationship it estab-
lishes between spatial scales and available kinetic energy
dissipated in motion at each scale. However, the applicabil-
ity of isotropic turbulence theory may have to be checked:
in particular, the equilibrium range has to be present. More-
over, a distinction then arises between viscous dominated
sub-range of the equilibrium range on one hand, and the mo-
mentum dominated sub-range that appears for Re 10, 000
on the other. In order to describe the ratio of deforming
forces to cohesion force, the viscous sub-range will require
a Capillary number, and the inertia sub-range a Weber num-
ber. Coulaloglou and Tavlarides (1977) considered that the
breakage of a droplet of diameter d in a stirred tank resulted
from the collision with eddies of suitable wavenumbers k
in the inertial sub-range, and developed a model which was
later improved on by Tsouris and Tavlarides (1994). Mod-
els developed for stirred tanks have been used in simula-
tion of pulsed ow devices, e.g. Jare s and Proch azka (1987)
used Coulaloglou and Tavlaridess model (1977) for a Karr
D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200 2191
column, and distinguished between the viscous and the in-
ertial sub-range. Tjaberinga et al. (1993) combined Bentley
and Leals results (1986) on transient ow breakage with
isotropic theory in the viscous and in the inertial sub-range
to predict breakage in emulsiers, with simulation for dis-
persions owing past a contraction. Ni et al. (2002) used
Tsouris and Tavlaridess model (1994) for a continuous os-
cillatory baed reactor. However, in such a system, the ex-
ploitation of the isotropic theory would require suitably de-
veloped turbulence for droplet sizes ranging from several
millimeters to a hundred micrometers. In an Oscillatory Baf-
ed Reactor (OBR) of diameter of a few centimeters, large
scale anisotropic eddies are of almost the same magnitude
as the largest drops. In such a case, a dierent approach is
needed if one wishes to evaluate the frequency of break-
age of a droplet. One such approach considers a dynamic
balance of the forces applied on the droplets at the places
within the reactor where they are most likely to break.
1.4. Forces inuencing the breakup of droplets
Forces in question may include drag forces in the general
sense, combining friction and dynamic pressure, and also
inertia. The generality of denition for drag means that
the term encompasses both the viscous and the momentum
dominated regimes. Many of the studies have been restricted
to low Re number, which is applicable to many industrial
situations (e.g. very viscous molten polymer blend). How-
ever, many systems fall outside this category, including the
one presented in this paper for which the viscosity ratio was
near 5, and the oscillatory Reynolds number in the continu-
ous phase of the order of 6006000. Again, in this situation
the viscosity would be absorbed within a drag force coef-
cient, which should be considered alongside with velocity
gradients and the associated Weber numbers instead of a
capillary number. Inertia is contributing to the deformation
of droplets (as exemplied with the attening of deceler-
ated drops observed by Mietus et al., 2002), and it should be
expected to be a signicant force in varying velocity elds.
Pietzch and Pilhofer (1984) used such a time-averaged,
dynamic approach to estimate the mean Sauter diameter of
droplets d
32
in a pulsed sieve-plate extraction column. They
found good agreement with their experimental data and oth-
ers, not only for pulsed sieve-plate columns, but also for
reciprocating plate columns. Their correlation spanned a
wide variation range for the plate geometry and the intensity
of the oscillations: orice diameter D
o
= 212.7 mm, plate
free area : = 0.0440.55, oscillation velocity x
o
[ = 2.5
75 mm s
1
, where : is the ratio of bae to column area, x
o
is the oscillation amplitude (mm) and [ the oscillation fre-
quency (Hz). Most interestingly, their correlation was able
to cover all ow-regimes, matching not only the (correct)
predictions for d
32
given by the isotropic theory in turbulent
conditions, but also the experimental observations in milder
ow. A discussion was provided on the basis of the domi-
nance of the inertia forces at low pulsation intensity, and of
the drag forces at higher intensity.
1.5. Droplet breakage in an OBR
The work reported in this paper is focussed on droplet
breakage in a continuous and horizontal OBR. OBRs are a
viable and simple alternative to conventional stirred tanks
for liquidliquid systems. The impeller is replaced by the
combination of pulsed ow and stationary orice baes.
This arrangement promotes the formation of eddies suitable
for mixing, fractionation of the dispersed phase and mass
transfer. Advantages of OBRs over stirred tanks include
the existence of more homogeneous conditions throughout
the reactor. This should facilitate the suitability of models
for the size distribution in the dispersed phase, and help
the control and prediction of reactor performance in an in-
dustrial context. OBRs also share similarities with pulsed
perforated-sieve columns, and could be considered as merely
being one such column with a unique perforation per plate.
However, OBRs present a unique ow pattern, as shown in
Fig. 1, which details the ow eld between two baes dur-
ing the rst half-cycle of uid oscillation. In this case, the
oscillation is generated at the base of the reactor. The dark
lines at both the top and bottom of the cell represent the
positions of the orice baes. Fig. 1 was obtained from
Digital Particle Image Velocimetry (DPIV), and was pro-
duced by A.W. Fitch (Fitch, 2003). Details of the method
and more results can be found in Fitch et al. (2001). The ar-
rows on the gure represent velocity vectors averaged over
a few cycles, and superimposed on them are the coloured
contour of shear rate. It can be seen that a tore-shaped vor-
tex develops at the start of the cycle behind the orice and
around the incoming jet. This vortex is pushed o as the jet
velocity reaches its maximum, and weakens as the forward
phase terminates. It then acts as an obstacle to the down-
ward incoming jet that develops during the down-stroke of
the cycle (not shown here) before decaying completely. In
terms of shear rates, the areas most aected are again con-
centrated around the orice edge and immediately behind it.
Less intense but nevertheless still strong gradients also ap-
pear all along the boundaries of the incoming jet, and also
at the boundary between the jet and the tore-shaped vor-
tex. Fig. 2 shows the extensional (axial) strain rate during
these forward phases of the cycle. It is most intense across
the orice, still with strong values inside the jet and at the
convergence between the tore-shaped vortex and the jet. It
should be noted that DPIV cannot provide velocity data at
the baes themselves, where one would expect strong shear
and strain.
Therefore, a droplet passing through an orice during the
forward phase of the cycle will rst be subjected to a combi-
nation of shear and extensional strain, with the highest shear
if it comes near the edge, i.e. the rst region. After pass-
ing the orice, the extensional stretch decreases but remains
2192 D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200
Fig. 1. Results from DPIV measurements in an OBR cell, rig diameter = 40 mm, : = 0.33, x
o
= 4 mm and [ = 3 Hz. Velocity vectors are shown on
some of the nodes. Shear strain rates are represented by shading, as shown on the scale on the right-hand side. Slide a marks the start of the forward
phase of the oscillation cycle and slide b denotes the start of the downward phase.
piston
Dispersed
phase inlet Outlet
Continuous
phase inlet
4 Image capture windows, P
1
-P
4
Injection port window, P
inj

Fig. 2. Schematic diagram (top-view) of the continuous oscillatory baed
reactor.
strong, and similarly with the shear. This second region of
deforming ow will be referred to as the jet and vortex
zone. From what was mentioned in the introduction, it is
clear that the initial strong ow at the orice may deform
some droplets suciently so that weaker ow in the jet and
vortex zone may stretch further the elongated droplets and
break them apart. On the other hand, the droplets may even
already be breaking apart at the orice, even though the ori-
ce zone is not very extended and residence times there are
short.
Another interesting observation concerning the breakage
of droplets in OBRs is the comparatively larger inuence
of small changes of oscillation amplitude when compared
with the changes in oscillation frequency (Ni et al., 2002),
a fact that should be reected in a model for the probability
of breakage of droplets. However, in the model by Jare s and
Proch azka (1987), the amplitude and frequency were not
separated but kept together as a product, i.e. the oscillation
velocity x
o
[. This suggests that further theoretical devel-
opments are necessary. In this paper, an integral approach
was adopted, based on the total amount of work eected on
the droplet for its deformation. This method is consistent
with our goal of predicting probabilities of breakage, as it
accounts for the history of individual drops and of the ow
conditions that the drops encountered. Therefore, drops that
may be stretched following the transient exposure to the
critical ow are considered to break when they have sub-
jected to enough deformation work from the drag and inertia
forces. This model should also take into account the pres-
ence of edges and strong velocity gradients at the orice, as
well as the time of passage at the bae during the cycle.
The amount of the remaining time for the forward phase of
cycle would also be used as an estimate for the residence
time of the droplets within the jet and vortex zone. Vorticity
eects (e.g. those associated with the tore-shaped vortex)
were deemed too complex and were neglected. Stretches
normal to the radial direction were also neglected, although
their value in a tore-shaped vortex might be signicant as
the continuous phase spreads away from the centerline. The
probability model could then be used as a basis to predict
breakage rates in a continuous OBR, as it has been found
that these rates were set by the frequency of passage through
baes for a range of frequencies and amplitudes of oscilla-
tions (Mignard et al., 2003).
2. Materials and method
2.1. Experimental procedure and materials
The continuous OBR used in this work is shown schemat-
ically in Fig. 3. It consisted of seven horizontal glass tubes,
each 40 mm in inside diameter and 1.5 m in length, and
connected to each other by U-bends having an outlet for
D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200 2193
Fig. 3. Example of a sequence of pictures taken with the high-speed
camera at 750 FPS. For clarity, only pictures taken every 1}75th s are
shown here. A droplet is crossing the bae on the right-hand edge and
breaking. The observation window was at P
inj
as in Fig. 2. (Conditions:
[ = 1 Hz, x
o
= 6 mm, Q = 2 l min
1
.)
draining. The total length of the reactor was 12.1 m. The
112 orice baes were spaced at a distance equal to 1.8
times the inside tube diameter, and their orice diameter
of D
o
= 23 mm corresponded to a free area : = 0.33. The
net ow was provided by a LOWARA pump, and the ow
rates were monitored by a ow meter. Fluid oscillation was
achieved by means of a crank-piston arrangement driven by
a helical geared motor through a frequency inverter, pro-
viding a frequency range of 010 Hz. Peak-to-peak oscilla-
tion amplitudes of 060 mm could be obtained by adjust-
ing the o-centre positions of the crank in a ywheel. A
non-return valve in the ow inlet would reduce any propa-
gation of oscillation upstream. Sample ports and optical ob-
servation windows were installed at ve dierent positions,
P
inj
, P
1
, P
2
, P
3
and P
4
, along the length of the column. The
continuous phase was water and the dispersed phase was
coloured silicone oil, which was injected at the rst tube as
shown in Fig. 3. The silicone oil had the following proper-
ties: density = 915 kg m
3
, viscosity = 4.6 cp and surface
tension = 0.0211 N m
1
. All the experiments were done at
room temperature, and without the use of any surfactants.
The reactor was rst ushed with water at high ow rates
to remove any air or oil deposits. The net ow rate was
then adjusted at its required value (0.52 l min
1
), and the
oscillations were started with the preset amplitude and fre-
quency (in the range 13 Hz, and 57.5 mm centre-to-peak,
respectively). The dispersed phase was fed into the system
at a xed ow rate. The images were captured and stored
in PC from ve dierent optical windows along the ow
path as shown in Fig. 3 using an Ektapro Kodak 4540mx
0
50
100
150
200
250
0 20 40 60 80 100 120
x
o
(mm.s
-1
)
S
t
r
a
i
n

(
s
-
1
)
Fig. 4. Maximum strain rates observed from DPIV data, rig
diameter = 40 mm, : = 0.21, uid = water. Open symbols: extensional
strain; lled symbol: shear strain.
high-speed camera tted with a Micro Nikkor objective lens
(55 mm diameter, 1:28 focal length ratio). The camera was
operated at 7501125 frames-per-second (FPS). The light
source and the camera were kept at about 0.3 and 0.15 m
respectively from the image window. The hold-up for the
dispersed phase during the experiments was low (e.g. a few
percent), enabling the observations of individual passages
of drops past the bae and the break-up events. Wetting of
the glass tube wall was not observed in the straight section.
Some wetting of the metal baes did occur to a small extent,
but the images of droplets attached to the bae were sim-
ply ignored. The images captured on PC were analysed us-
ing Droplet Detector v3 Image Analysis Software. For high
amplitudes and frequencies more droplets were observed in
each individual image and vice versa, therefore an appro-
priate number of images were needed in order to represent
the experimental conditions, and was found to be around
between 800 and 1000 per location.
The high-speed digital camera allowed direct observation
of breakage of droplets, which provided a sound basis for
the correlation between rate constants and droplet diame-
ters in OBRs. Droplets were counted as they passed by the
bae, and breakage was recorded when it happened. The
average probability of breakage P
br
for droplets of a given
size was then estimated by a simple ratio. The sizes of all
droplets, including that of daughter droplets generated from
break-up, were estimated from direct manual measurement
on the image.
2.2. Preliminary observations
Fig. 4 is a photographic sequence showing a droplet as
it was deforming and breaking just after passing the baf-
e on the right-hand side. The droplet was moving from
right to left on the picture, following the direction of the
continuous ow superimposed with ow oscillation at that
time. In this particular instance, the droplet was rst re-
tained behind the bae for a moment (Figs. 4d and e),
2194 D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200
while part of it was being stretched past the orice (Fig. 4f).
The elongated droplet was eventually dislodged from the
bae (Fig. 4g), and breakage occurred soon afterwards
(Figs. 4hj). In other cases, the droplet deformation was still
initiated at the orice, but the whole droplet was ung for-
ward immediately. Further deformation and possible break-
age would then occur well within the jet and vortex zone.
Such high-speed camera footages conrmed the breakage
mechanism in OBRs that was presented in the introduction.
The droplets were stretched and even broken by steep ve-
locity gradients at the orice edge and also within the jet
and vortex region. The forward phases of the cycle in the
continuous OBR, which corresponds to the pulsed ow fol-
lowing the direction of the net ow, was also seen to ac-
count for all breakage events in the 38 mm diameter range.
In fact, breakage was observed at lower than such a diam-
eter range. In initial observations made on a batch OBR
with the camera zooming in the bae orice, droplets as
small as 1.6 mm were observed to break up. However, the
gathering of statistically meaningful amounts of data for a
given volume range of droplets was not practical for diam-
eters smaller than this size. For such sizes, the probability
of breakage per passage through a single bae was simply
too small, which suggested that the passage through a large
number of baes present within the continuous OBR was
responsible for the signicant breakage rates seen for small
droplets. Another factor that restricted the usable range of
sizes was the low resolution of the high speed-camera, with
respect to the conicting requirements of estimating the size
of a small droplet and of keeping it within the view eld as
it broke. These factors limited the usable droplet diameter
to a practical minimum of 3 mm, and results for lower sizes
would have to be extrapolated. The hold-up, which was re-
stricted to low values of a few percent, was observed to have
no eect on the droplet size distributions (the results to be
published in a subsequent paper), and hence this factor was
not taken into account for the modelling work.
In order to ensure that a statistically meaningful amount
of data was collected for each drop size during one exper-
iment, a methodology had to be established. The standard
method consists of calculating P
br
from an increasing num-
ber of observed droplets, until any further count of droplet
does not aect P
br
anymore (Haverland, 1988). However,
for the lack of an adequate injection system on the horizontal
OBR used here, the data available had to be obtained from
a limited number of droplets of uncontrolled sizes that hap-
pened to cross the bae. Typically, numbers of the order of
50 could be obtained, but sometimes less, especially for the
larger droplets. The idea was to consider the breakage events
as a Bernoulli random process of probability P
br
, with each
new event being independent from the previous one. The
theoretical variance of the process is then var =P
br
(1P
br
).
If n droplets are observed and P
EXP
br
is the breakage proba-
bility estimated from this sample, then the variance of the
sample around its mean value is var
2
n
= var
2
}n (Davenport,
1970). If we assume that the means of samples would have
a normal distribution around P
br
, it is then possible to put
error bars on experimental plots for P
br
.
3. Model development
For breakage rate constants, a model based on direct ex-
perimental observation was developed specically for con-
tinuous OBRs. Its inputs were the drop diameters, the fre-
quency and amplitude of oscillation, and the net ow rate,
as well as up to ve adjustable parameters.
3.1. Breakage rate constants
Ni et al. (2002) studied the changes in DSDs for oil
droplets dispersed in water in a continuous OBR. They
operated at net ow Reynolds numbers of Re
n
=2501000,
combined with an oscillatory Reynolds number of Re
o
=
2100 7900 (frequency [ = 13 Hz, centre-to-peak ampli-
tude x
0
=57.5 mm). It was observed that the DSDs in the
0.1 2.6 mm diameter range would not be aected by the
net ow rate of the continuous phase, at least in the condi-
tions tested. Mignard et al. (2003) remarked that this fact
could be explained if passage through the baes was re-
sponsible for breakage, and the direct observations reported
in the previous section conrmed this view. Therefore, the
average breakage rate q (s
1
) for droplets of a given diam-
eter d, was linked to the breakage probability per passage
at a bae, P
br
, and the average residence time within a cell,
t
c
(s):
q
i
=
P
br
t
c
, (1)
where P
br
is the object of this paper, and deriving an ex-
pression for it would be of use in population balance mod-
elling because of its connection to q
i
through Eq. (1). It is
the product of two other probabilities: the probability P
edge
that the droplet touches the edge (or the region of intense
shear in its vicinity), and the probability P
energy
, conditional
to the droplet touching the edge, that it is stretched to the
breaking point by shear and extensional strains:
P
br
= P
edge
P
energy
. (2)
3.2. Probability of contacting the edge
A droplet of diameter d that is about to cross a bae has
a chance of contacting the edge of the orice if its centre
is situated within a distance d}2 of it. This area can be
represented as an annulus of surface area }4[(D
o
+d}2)
2

(D
o
d}2)
2
]=}2 D
o
d, in which D
o
is the orice diameter.
Hence, if we assume a uniform probability for the position
of the droplet within the cross-sectional area of the orice,
as well as a uniform plug-ow:
P
edge
k
1

}2 D
o
d
}4 D
2
o
= k
1

2d
D
o
. (3)
D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200 2195
This simple approximation takes into account events such
as the one recorded in Fig. 4. The parameter k
1
1 allows
for the region of intense shear extending some distance away
from the edge.
3.3. Probability of breakage after contacting the edge
It was frequently observed that the large droplets that
managed to escape breakage were those crossing the bae
at low velocity, especially towards the end of the forward
cycle. Therefore, expressions for P
energy
must take into ac-
count the time of passage of the droplet during the cycle,
t
E
, and the ratio of the instant capillary number to the criti-
cal capillary number, Ca(t
E
)}Ca
cr
. The time t elapsed after
passage will also be a factor, in line with the introduction
presented earlier. Hence, the interval of time available for
breakage to occur will be [t
E
, t]. Moreover, if one assumes
that the instant owrate of droplets is proportional to that of
the continuous phase:
P
energy
=
_
1}4[
1}4[

en
x(t
E
) dt
E
_
1}4[
1}4[
x(t
E
) dt
E
(4)
in which
en
is the probability of breakage for a droplet
touching the edge at t
E
, 1}4[ is the start of the forward
phase of the cycle and 1}4[ the end, x = 2[x
o
cos(2[
t
E
)}: + x
n
is the instant velocity at the orice, and x
n
is
the constant velocity associated with the net ow. After
some rearranging and setting a dimensionless entry time
0
E
= 2[t
E
:
P
energy
=
1
2
_
}2
}2

en
cos(0
E
) d0
E
, (5)

en
will take values 1 or 0 only. It is assumed that a droplet
will be stretched to signicant extent and susceptible to
breaking only if Ca(t
E
)}Ca
cr
1. Hence,
if Ca(t
E
)}Ca
cr
1,
en
= 0. (6)
Calculating Ca requires some knowledge of the shear rate
. This was approximated by assuming direct proportional-
ity between and the instant ow rate. Admittedly, this is a
crude, ad hoc assumption. Fortunately, the DPIV data men-
tioned in the introduction provides some degree of validity
to this claim. In these DPIV maps, the regions of the max-
imum shear are concentrated near the orice, and plotting
the observed maximum tangential shear rate and the max-
imum extensional (axial) shear rates as a function of the
pulsation intensity should be of some relevance. Such plots
are presented in Figs. 5 and 6, respectively. It can be seen
that both plots can broadly be brought around a straight line
for the operating conditions given (Re
o
= 8407500), al-
beit with some deviation. It must be noted that the straight
lines representing the strain rates do not intercept the origin.
However, the DPIV data did not cover the orice itself, and
some discrepancies might arise from this. For the shear rate,
0
0.2
0.4
0.6
0.8
1
0 1 2 3 4 5 6 7 8 9
Diameter (mm)
P
b
r
1Hz, 6mm - exp. 1Hz, 6mm - model
1.5Hz, 6mm - exp. 1.5Hz, 6mm - model
2Hz, 6mm - exp. 2Hz, 6mm - model
Fig. 5. Estimated breakage probability per passage through bae, P
br
, at
constant x
o
=6 mm. Plain lines with lled symbols represent experimental
results; dotted lines and open symbols represent model predictions, with
optimal t of k
1
= 1.87, k
2
= 2339, k
3
= 2.304, k
4
= 0.261, k
5
=0.198.
(Error bars on experimental curves represent 95% condence interval).
0
0.2
0.4
0.6
0.8
1
0 1 2 3 4 5 6 7 8 9
Diameter (mm)
P
b
r
2Hz, 5mm - exp. 2Hz, 5mm - model
2Hz, 6mm - exp. 2Hz, 6mm - model
2Hz, 7.5mm - exp. 2Hz, 7.5mm - model
Fig. 6. Estimated breakage probability per passage through bae, P
br
, at
constant [=2 Hz. Plain lines with lled symbols represent experimental
results; dotted lines and open symbols represent model predictions, with
optimal t of k
1
=1.87, k
2
=2339, k
3
=2.304 and k
4
=0.261, k
5
=0.198.
it was assumed that

s
= k
s
x
D
o
(7)
while for the extensional strain rate:

e
= k
e
x
D
o
. (8)
Eqs. (7) and (8) enable to re-write Eq. (6) by introducing
an adjustable parameter k
2
:
Ca}Ca
cr
= j
c
(c
1

e
+ c
2

s
)d}o = k
2
d
D
o

j
c
x
o
(9)
in which j
c
is the continuous phase viscosity, o the interfa-
cial tension, c
1
and c
2
some constants that were temporarily
introduced to combine the two forms of strain in a linear
expression for the total strain rate.
2196 D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200
If Ca(t
E
)}Ca
cr
1, an estimate of the energy spent on
stretching the drop has to be made.
3.4. Energy balance on the drop
We denote E
break
o
the energy required for splitting the drop.
This energy is approximated by being proportional to the
energy required to increase the surface area of a spherical
drop of diameter d, to that of two equal sized daughter
droplets of diameter d}2
1}3
. Hence,
E
break
o
=
_
2 o
d
2
2
2}3
o d
2
_
= (2
1}3
1) od
2
. (10)
A balance of the forces applied to a droplet shows that
only the inertia force F
I
and the drag force F
D
need being
considered: the eect of gravity was neglected. If as a rst
approximation, we consider one end of the droplet to re-
main within the vicinity of the orice and the other one to
experience the full amount of velocity x =2[x
o
cos(2[
t
E
)}: + x
n
, then the amount of work done on the droplet
deformation within [t
E
, t] should be
E
o
=
_
t
t
E
(F
D
+ F
I
) s dt (11)
with s being the stretch rate for the elongated droplet. As t
increased between t
E
and 1}4[, breakage was considered to
occur if and as soon as
k
o
E
o
}E
break
o
1 (12)
with a new adjustable parameter, k
3
.
3.5. An estimate for the instant stretch rate
Since Ca(t
E
)}Ca
cr
1, the droplet will elongate inde-
nitely as long as it does not break apart or the ow stops
and changes direction at the end of the forward cycle. The
following relationship was assumed
s(t) = k
stretch
(t) s(t) (13)
with s(t
E
) = d. Hence,
s(t) = dexp
__
t
t
E
k
stretch
(t) dt
_
(14)
and
s(t) = d k
stretch
(t) exp
__
t
t
E
k
stretch
(t) dt
_
. (15)
Using Eqs. (7) and (8) and introducing another adjustable
parameter k
3
:
s(t) = k
3

x(t)
0.6

d
D
o
exp
_
k
3

x(t) x(t
E
)
D
o
_
. (16)
Afactor 0.6 was introduced to account for the jet contraction.
3.6. Drag force
Following Pietzsch and Pilhofers approach (1984) with
averaged r.m.s. velocities, Hu and Kintners work (1955)
was used to calculate the drag force F
D
here. One problem
is that the relative velocity between the droplet and the con-
tinuous phase was not known. Let x
d
be the droplet velocity,
and x
c
= x}0.6 the jet velocity. Assuming that the droplet is
at the boundary of the jet, the relative velocity is not actually
known. Pietzch and Pilhofer (1984) set x
d
= x
c
(with aver-
age values over half a cycle), which assumes no lag from
encountering the edge of the orice. They further assumed
that the continuous phase outside the jet was at rest, which
is not really the case in OBRs according to the DPIV data.
Fig. 4 also shows that droplets are not necessarily carried
away by the jet. However, the complexity of the problem re-
quires such simplications whenever possible. In this work,
we noted that for a droplet with a side at rest near the orice
and the other side subjected to the jet, the relative velocity
may be of similar intensity to that of a droplet pushed by the
jet into a region of low velocity. Hence, an approximation
suited to both situations would be
x
d
= k
lag
x}0.6 (17)
and
x
c
= (1 k
lag
) x}0.6 (18)
for which k
lag
1 would give the scenario shown in Fig. 4,
and k
lag
=1 would give Pietzsch and Pilhofers scenario. In
both cases, the relative velocity is x}0.6. Hence,
F
D
= C
D

d
2
4

j x
2
2 (0.6)
2
. (19)
From Hu and Kintner (1955), a dimensionless quantity Y
was rst calculated:
Y =
4
3

_
Re
o
(d, t)
P
0.15
+ 0.75
_
1.275
for 2 6Y 670, (20)
Y = 0.045
_
Re
o
(d, t)
P
0.15
+ 0.75
_
2.37
for Y 70 (21)
with Re
o
(d, t) = d xj
c
}0.6j
c
being the oscillatory Reynolds
number for the droplet; and P =(j
c
o
3
}qj
4
c
) (|j
c
j
d
|}j
c
)
a modied uid number. The drag coecient was found
to be C
D
= Y}[We(t) P
0.15
], with We(t) the instant Weber
number. The drag force is then
F
D
=

4
o d
Y(t, d)
P
0.15
. (22)
For Y 2, Stokes law is applied.
3.7. Inertia force
The inertia force is given by
F
I
=j
c

_
j
d
j
c
+ 0.5
_

d
3
6
x
d
(23)
D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200 2197
which is
F
I
= k
lag
j
c

_
j
d
j
c
+ 0.5
_

d
3
6

(2[)
2
x
o
0.6 :
sin(2[ t). (24)
In the scenario encountered in Fig. 4, the inertia force
would stabilise the droplet when accelerating, but it would
help stretching when the droplet was decelerating. From
an energy standpoint, we would have E
o
= E
D
|E
I
| dur-
ing acceleration of the droplet, with E
o
and E
D
0. This
corresponds to energy imparted from the drag force being
shared between accelerating the droplet, and deforming the
droplet. Similarly, energy from deceleration is added to E
D
and contributing to deformation. By contrast, in Pietzsch
and Pilhofers scenario (1984), inertia is always a destabil-
ising force. This may happen when the droplet is not partly
retained by the bae. In this case, the inertia force takes a
direction opposed to the drag when the acceleration is pos-
itive, and hence the droplet may be stretched apart. This
other mechanism was allowed simply by changing the sign
of the inertia term.
3.8. Procedure for calculating P
br
with parameter
optimisation
After substituting in Eq. (11) the expressions obtained
in Eqs. (10), (16), (22) and (24), the full expression for
E
o
}E
break
o
is
E
o
E
break
o
=
_
0
0
E
_
k
3
x
o
2.4:D
o

Y(0, d)
P
0.15
+
k
I
6
_
1 +
0.5j
c
j
d
_

d
D
o
We
max
(d) sin 0
_

e
k
3
(x
o
(sin 0sin 0
E
)}0.6:D
o
)
2
1}3
1
cos 0 d0 (25)
with 0 =2[t and 0
E
=2[t
E
; k
I
=k
lag
k
3
is an adjustable
parameter, which may also hide a moment of inertia or
take a negative value; We
max
(d)=(j
d
d}o)(2[x
o
}0.6:)
2
is
a Weber number for the droplet. At this stage in the present
work, the adjustable parameters k
4
= k
3
k
o
and k
5
= k
I
k
o
were introduced, which eliminated the need for k
o
and k
I
when E
o
}E
break
o
was substituted into Eq. (12). k
o
E
o
}E
break
o
in
Eq. (12) was then computed using a RungeKutta method.
Therefore, for a set of ve parameters k
1
k
5
and a given
diameter d, it was possible to calculate P
br
with Eqs. (2),
(3), (5), (6), (9), (10), (12), (20), (21) and (25).
k
1
. . . k
5
were adjusted so as to match the experimen-
tally observed curves for P
br
. The minimisation scheme
was Fletchers version of the Marquardt method (Fletcher,
1971). The objective function to be minimised by the pro-
gramme was
,
2
=

d,[, x
o
_
P
br
(d, [, x
o
) P
EXP
br
(d, [, x
o
)
P
EXP
br
(d, [, x
o
)
_
2
(26)
with P
EXP
br
being the experimentally observed value for P
br
.
This formula used the relative rather than the absolute resid-
uals, which allowed comparable weighting to the results
from dierent diameters. This approach was preferred be-
cause low values were of signicant for the purpose of ex-
trapolation to small diameters. However, the occurrence of
some nil values for P
EXP
br
could prevent it from working, and
it was decided that Eq. (26) had to be used with a modica-
tion: an absolute dierence term would replace any relative
dierence term for which P
EXP
br
0.05.
3.9. Extrapolation to small diameters, and comparison
with Pietzch and Pilhofers equation
Once the optimal set of parameters was found, the model
could be extrapolated to small diameters in order to deter-
mine the minimum diameter for breakage d
min
for which
P
br
= 0. This value could be compared to that returned by
the simple dynamic model that was proposed by Pietzch and
Pilhofer (1984). Their equation was applied with the oscilla-
tory velocity being set at its maximum (in the middle of the
forward phase of a cycle). Since the OBR was horizontal,
the terms related to gravity were dropped, and the equation
became
d
min
=

0.9: o

2
j
d
x
o
[
2
+
1
2.56
_
C
D
x
o
j
c
:j
d
_
2

1
1.6
C
D
x
o
j
c
:j
d
. (27)
4. Results and discussion
Experimental curves for breakage probabilities and
least-squares match using Eq. (25) are presented in
Figs. 5 and 6. In both gures, the solid lines represent the
experimental data, and the dotted lines the model results af-
ter the parameters of k
1
k
5
were adjusted. The signicance
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10
Parameter value

2
k1
k2 .1e-3
k3
10xk4
-10xk5
Fig. 7. Eect on ,
2
of the variation of one parameter at a time around
the optimised solution.
2198 D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200
of the k-values found for the optimised parameters can be
checked at a glance in Fig. 7, which presents the variation of
the objective function ,
2
as a function of the values of pa-
rameters, with one parameter at a time being varied around
the optimal solution. It can be seen that a clear local min-
imum is achieved for each of the ve tting parameters,
indicating a high degree of condence in the minimisation
methodology. It should be noted that the parameter k
5
was
less signicant, since it could vary to some extent without
aecting the match to the experimentally observed curves.
This means that inertia forces had less of an inuence than
drag for the conditions tested.
Two features can be extracted from Fig. 5: rstly, the
probability of breakage P
br
increases with the increase of
droplet diameters for both experimental and modelling meth-
ods. Partly as a result of using the relative residuals in Eq.
(26), the predicted curves in Fig. 5 lie mostly below the
experimental ones, but attempts at correcting this problem
would lead to discrepancies at lower diameters and an in-
creased ,
2
. This highlights the diculty in ensuring that the
model can be extrapolated to small diameters, while also
account for the behaviour of large drops. Secondly, the pre-
dicted trend of P
br
(the dotted lines) is that P
br
always in-
creases when the oscillation frequency increases. This is in
line with the general trend reported in the previous para-
graph for the experimental data, but does not account for the
crossing-over of the curves of 1 and 1.5 Hz cases. Similar
phenomena can be seen in Haverlands work: for droplets of
a given diameter and when increasing oscillation frequency,
the breakage probability sometimes decreased to a mini-
mum before increasing again (Haverland, 1988). Neverthe-
less, the similar trends with similar magnitudes of breakage
probability are clear for both experimental and modelling
methods.
In Fig. 6, the two features showing in Fig. 5 are also appli-
cable here. In this case, the modelling predictions are closer
to the experimental data than those shown in Fig. 5, with
the exception that there was no further increase of P
br
when
the oscillation amplitude was raised from 6 to 7.5 mm. In
fact, the curves for x
o
= 6 and 7.5 mm are intertwined in
Fig. 6, crossing over 4 times. No obvious explanation for
this was present. However it is worth noting that some of the
Haverlands gures for breakage probability showed a cer-
tain degree of meandering, and also the duplicate measure-
ments in those gures seemed to demonstrate a spread of re-
sults (Haverland, 1988). Therefore, it might be the case that
in both this work and the Haverlands study, some unknown
factors were not controlled, however inuenced the repro-
ducibility or the shape of the breakage probability curves.
In Fig. 6 it can be seen that the three predicted curves (the
dotted lines) for three dierent amplitudes at a xed oscil-
lation frequency are in parallel, whereas the experimental
curves show a large increase in P
br
from x
o
= 5 to 6 mm,
while there is little change fromx
o
=6 to 7.5 mm. The former
may be interpreted as a consequence of an exponential-like
stretch of the droplets, and this was reected in the model
Table 1
Minimum diameters for breakage extrapolated from the model, and com-
pared with that using the Pietzch and Pilhofers equation
[ (Hz) x
o
(mm) d
min
(mm) d
min
from
Pietzch and Pilhofer (mm)
1 6 3.52 5.66
1.5 6 2.79 3.76
2 6 2.53 2.84
2 5 3.43 3.31
2 7.5 1.75 2.32
as a suitable expression for the stretch rate s. However, ex-
plaining the latter eect is more dicult. In order to do so,
it was rst assumed in the model that all drops touching the
edge of the bae would break anyway if x
o
was greater than
6 mm. This would imply that Ca(t
E
)}Ca
cr
1, and that Eq.
(12) be always satised for x
o
6 mm and [ =2 Hz. This
would then impose a value for k
1
of around 1.4 in order to
match the curves for 6 and 7.5 mm. However, an optimal
set of parameters k
1
k
5
could not be found with these con-
ditions.
Comparing the experimental curves of Fig. 5 with Fig. 6,
it shows that a 20% increase in the oscillation amplitude x
o
from 5 to 6 mm produced an increase of P
br
of the order of
0.20.35. This eect was bigger than that produced when
doubling the frequency from 1 to 2 Hz, where the maximum
increase of P
br
was only of 0.2. This suggests that the oscil-
lation amplitude has stronger eect on breakage probability
than that of the frequency, which was the rationale behind
the development of the model presented here.
4.1. Extrapolation to smaller diameters
The minimum drop diameters for breakage are presented
in Table 1 and were extrapolated fromEq. (25) together with
the optimised tting parameters. The values obtained from
the Pietzch and Pilhofers equation (27) are also included
for comparison. Table 1 shows that the extrapolated mini-
mum drop diameters were signicantly smaller than those
obtained from Eq. (27), with one exception of x
o
= 5 mm
and [ = 2 Hz. The discrepancies are due to that Eq. (27)
used an approximation for the average inertia force com-
bined with a value for the maximum drag force, whereas
the model considered the instant values of the drag force
and the inertia force. Therefore, the model developed in this
paper should, in principle, be more accurate than Eq. (27).
Table 1 also shows that the predicted minimumdiameters are
all larger than 1.75 mm, while for the same operating condi-
tions, the mean drop sizes in the continuous OBR (Ni et al.,
2002) are much smaller than these minimum drop diameters
extrapolated here. This discrepancy could be explained by
the fact that large drops are much more susceptible to defor-
mation (since they correspond to higher capillary numbers)
and hence have very dierent drag coecients and inertia
D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200 2199
moments. Therefore, the model may not extrapolate very
well to sizes below the minimum diameter of drops observed
in this work (ca. 3 mm). Moreover, dynamic models like the
one used here ignore the turbulent eddies in the universal
equilibrium range that may arise on the smaller scales. Tur-
bulent eddies may have sizes comparable to those of small
droplets of a few hundred micrometre, and may break those
small drops. Although the probability of breakage may be
very low for these small drops, the cumulative eect of a
large number of baes would bring about some visible ef-
fects. A more appropriate model there would be the one
presented by Tjaberinga et al. (1993), which involved Kol-
mogorov eddies in the viscous or inertial subrange that are
generated behind a contraction.
5. Conclusion
The breakage probabilities of large droplets (3.48 mm
diameter) in a continuous OBR were measured using the
High Speed Camera. A model of the force balance was de-
veloped with the help of the direct visual observation of
the stretching and breakage process, and the parameters of
the model were adjusted to match best the observed prob-
abilities. The trends for the predictions were mostly in line
with the observations, with the amplitude of the oscillations
having a much bigger eect on breakage probability than
the frequency. Extrapolations to the minimum diameters for
breakage was in line with other models, but seemed to con-
tradict observations in OBRs where smaller diameters were
usually aected after a number of baes. The change in
drop shape at large diameter was a possible cause, and the
inuence of turbulent eddies at the sub-millimetric scale was
also invoked. Further work will make use of the isotropic
theory for the small drop sizes.
Notation
Ca(t
E
) droplet capillary number at time t
E
Ca
cr
critical droplet capillary number
C
D
drag coecient
D
o
orice diameter, m
d droplet diameter, m
d
min
minimum diameter for breakage, m
E
D
drag force work during stretch, J
E
o
interfacial tension work during stretch, J
E
break
o
energy required for splitting a drop, J
F
D
drag force, N
F
I
inertia force, N
[ frequency of oscillation, Hz
q average breakage rate, s
1
k
1
k
5
adjustable parameters
P modied uid number, P=(j
c
o
3
}qj
4
c
) (|j
c

j
d
|}j
c
)
P
br
probability of breakage when passing a bae,
from model
P
EXP
br
probability of breakage when passing a bae,
from experiment
P
edge
probability that the droplet touches the orice
edge

en
probability of breakage for a droplet touching
the edge at t
E
P
energy
probability of breakage, conditional to the
droplet touching the edge
Re
n
net ow droplet Reynolds number
Re
o
oscillatory droplet Reynolds number
s droplet stretch rate, m s
1
t time, s
t
E
initial time for droplet passing the orice, s
x instant velocity at the orice, m s
1
x
c
jet velocity, m s
1
x
d
droplet velocity, m s
1
x
o
amplitude of oscillations, m
We(t) instant Weber number for the drop
We
max
maximum Weber number for the drop,
We
max
(d) = j
d
d(2[x
o
)
2
}o}(0.6:)
2
Y a dimensionless quantity dened in Eqs. (20)
and (21)
Greek letters
: plate free area ratio
shear rate, s
1
0 time
0
E
dimensionless entry time
z viscosity ratio j
d
}j
c
j
c
dispersed phase viscosity, kg m
1
s
1
j
d
continuous phase viscosity, kg m
1
s
1
o interfacial tension coecient, N m
1
t
c
average residence time within a cell, s
,
2
sum of squared residuals (merit function)
References
Barth es-Biesel, D., Acrivos, A., 1973. Deformation and burst of a liquid
droplet freely suspended in a linear shear eld. Journal of Fluid
Mechanics 61 (1), 121.
Bentley, B.J., Leal, L.G., 1986. An experimental investigation of drop
deformation and breakup in steady, two-dimensional linear ows.
Journal of Fluid Mechanics 167, 241283.
Briscoe, B.J., Lawrence, C.J., Mietus, W.G.P., 1999. A review of
immiscible uid mixing. Advances in Colloid and Interface Science
81, 117.
Coulaloglou, C.A., Tavlarides, L.L., 1977. Description of interaction
processes in agitated liquidliquid dispersions. Chemical Engineering
Science 32, 12891297.
Davenport, W.B., 1970. Probability and Random Processes, International
Student Edition, McGraw-Hill Kogakusha Ltd., pp. 328332.
Elemans, P.H.M., Bos, H.L., Janssen, J.M.H., 1993. Transient phenomena
in dispersive mixing. Chemical Engineering Science 48 (2), 267276.
Fitch, A.W., 2003. Characterisation of ow in an oscillatory baed
column using digital particle image velocimetry and laser induced
2200 D. Mignard et al. / Chemical Engineering Science 59 (2004) 21892200
uorescence. Ph.D. Thesis, Heriot-Watt University, School of Physical
Sciences and Engineering, Edinburgh, UK.
Fitch, A.W., Ni, X., Stewart, J., 2001. Characterisation of exible baes
in an oscillatory baed column. Journal of Chemical Technology and
Biotechnology 76, 10741079.
Fletcher, R., 1971. A modied Marquardt subroutine for non-linear
least squares. AERE-R 6799, Atomic Energy Research Establishment,
Harwell, England.
Gourdon, C., Casamatta, G., Muratet, G., 1994. Population balance based
modelling of solvent extraction columns. In: Godfrey, J.C., Slater,
M.C. (Eds.), LiquidLiquid Extraction Equipment. Wiley & Sons Ltd.,
New York, pp. 136226 (Chapter 7).
Grace, H.P., 1982. Dispersion phenomena in high viscosity immiscible
uid systems and application of static mixer as dispersion
devices in such systems. Chemical Engineering Communications 14,
225277.
Hakimi, F.S., Schowalter, W.R., 1980. The eects of shear and vorticity
on deformation of a drop. Journal of Fluid Mechanics 98 (3), 635645.
Haverland, H., 1988. Untersuchungen zur Tropfendispergierung
in ussigkeitpulsierten Siebboden-Extractionskolonnen. Dissertation,
Technische Universit at Clausthal, Germany, pp. 124138.
Hu, S., Kintner, R.C., 1955. The fall of single liquid drops through water.
A.I.Ch.E. Journal 1 (1), 4248.
Jare s, J., Proch azka, J., 1987. Break-up of droplets in Karr reciprocating
plate extraction column. Chemical Engineering Science 42 (2),
283292.
Mietus, W.G.P., Matar, O.K., Lawrence, C.J., Briscoe, B.J., 2002.
Droplet deformation in connced shear and extensional ow. Chemical
Engineering Science 57, 12171230.
Mignard, D., Amin, L.P., Ni, X., 2003. Population balance modelling of
droplets in an oscillatory baed reactor, using direct measurements
of breakage rate constants. Journal of Chemical Technology and
Biotechnology 78 (2/3), 364369.
Ni, X., Mignard, D., Saye, B., Johnstone, J.C., Pereira, N., 2002. On the
evaluation of droplet breakage and coalescence rates in an oscillatory
baed reactor. Chemical Engineering Science 57, 21012114.
Pietzch, W., Pilhofer, T.H., 1984. Calculation of the drop size in pulsed
sieve-plate extraction columns. Chemical Engineering Science 39 (6),
961965.
Stone, H.A., Leal, L.G., 1989a. Relaxation and breakup of an initially
extended drop in an otherwise quiescent uid. Journal of Fluid
Mechanics 198, 399427.
Stone, H.A., Leal, L.G., 1989b. The inuence of initial deformation on
drop breakup in subcritical time-dependent ows at low Reynolds
numbers. Journal of Fluid Mechanics 206, 223263.
Stone, H.A., Bentley, B.J., Leal, L.G., 1986. An experimental study of
transient eects in the breakup of viscous drops. Journal of Fluid
Mechanics 173, 131158.
Taylor, G.I., 1934. The formation of emulsions in denable elds of ow.
Proceedings of the Royal Society, A 146, 501523.
Tjaberinga, W.J., Boon, A., Chesters, A.K., 1993. Model experiments and
numerical simulations on emulsication under turbulent conditions.
Chemical Engineering Science 48 (2), 285293.
Tomotika, S., 1935. On the instability of a cylindrical thread of a viscous
liquid surrounded by another viscous liquid. Proceedings of the Royal
Society, A 150, 322337.
Tsouris, C., Tavlarides, L.L., 1994. Breakage and coalescence models for
drops in turbulent dispersions. A.I.Ch.E. Journal 40 (3), 395406.

You might also like