Phase Inversion Studies in Liquid-Liquid Dispersions

MICHAEL A. NORATO‘, COSTAS TSOURIS2 and LAWRENCE L. TAVLARIDES’*

Department of Chemical Engineering and Materials Science, Syracuse University, Syracuse, NY 13244-1 190

Chemical Technology Division, Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, TN 37831-6226
A study of the phase inversion behavior of liquid-liquid dispersions in stirred vessels is performed for liquids of various
physical properties at various operating conditions. Physical parameters studied are density (867 to 1180 kg/m3), vis-
cosity (0.00096 to 0.00378 Pas), and interfacial tension (0.0089 to 0.0323 N/m). Ambivalence region plots are presented
and compared with results reported in the literature. Experiments are performed to examine the effects of impeller type
and impeller-to-tank diameter ratio (DIT) on the ambivalence behavior. Also, phase inversion time experiments are per-
formed to investigate the time required for complete phase inversion under various dynamic conditions. The traditional
method of plotting the organic phase volume fraction at phase inversion against the agitation speed at that condition is
compared with the method of plotting the initially dispersed phase volume fraction at phase inversion against the agita-
tion speed at that condition. A hysteresis phenomenon is shown in phase inversion from OIW to WIO and WIO to OIW
dispersions. Also, it is shown that, depending on the physical properties of the dispersed and continuous phases, phase
inversion may occur when the agitation speed is increased or decreased.

On a men6 une Ctude du comportement d’inversion de phases de dispersions liquide-liquide dans des rCservoirs agitCs
pour des liquides aux propriCtCs physiques diverses et dans diffkrentes conditions de fonctionnement. Les paramktres
physiques Ctudiks sont la masse volumique (867 51 1180 kg/m3), la viscositk (0,00096 9 0,00378 Pas) et la tension inter-
faciale (0,0089 B 0,0323 Nlm). Des relevCs des rigions d’ambivalence sont prCsent6s et compares aux risultats de la lit-
tCrature scientifique. Des expCriences sont menCes afin d’examiner les effets du type de turbine et du rapport du diamktre
de la turbine au diambtre du rCservoir (017) sur le comportement d’ambivalence. De m&me, des expkriences sur les
temps d’inversion de phase sont menCes pour Ctudier le temps requis pour une inversion de phase complbte dans des con-
ditions dynamiques variies. La mCthode traditionnelle de track de la fraction volumique de la phase organique a I’in-
version de phase par rapport B la vitesse d’agitation I? cette condition est comparCe B la mCthode de track de la fraction
volumique de phase disperske initiale h l’inversion de phase par rapport B la vitesse d’agitation ii cette condition. Un
phknombne d’hystCrbse est montrk dans I’inversion de phase des dispersions OIW B WIO et WIO B OIW. On montre
Cgalement que, selon les propriCtks physiques des phases disperske et continue, l’inversion de phase peut survenir
lorsque la vitesse d’agitation est augmentke ou diminde.
Keywords: liquid-liquid dispersions, phase inversion, ambivalence region, hysteresis.

I n a stable liquid-liquid dispersion there is a dynamic equi-

librium between the two competing phenomena of drop
breakage and drop coalescence. The breakage of drops is
due to turbulence created by the impeller. Cutter (1966)
found that the fraction of the small eddies which cause drop
breakage was largest in the immediate vicinity of the
impeller, and a study by Tsouris and Tavlarides (1994) led
to the conclusion that drop breakage is confined to the
impeller region. Drop coalescence is thought to occur every-
where in the vessel (Calabrese et al., 1986).
If the system deviates from this equilibrium in a manner
which causes the drop coalescence rate to greatly exceed the
drop breakage rate, there will be a rather sudden coalescence
of the dispersed phase drops into a single continuum; and
the phase which was previously continuous will become dis-
persed. This phenomenon usually occurs within a few sec-
onds and is known as phase inversion.
The post phase inversion dispersion may be quite differ-
ent from the initial dispersion with respect to dispersed and
continuous phase physical properties, but the rates of drop
breakage and coalescence are once again at equilibrium.
Phase inversion is among the least understood phenomena
in liquid-liquid contacting. An understanding of phase
inversion is important to the design of many industrial con-
tactors such as those used in solvent extraction and hetero-
geneous reaction engineering. Efficient operation of these
*Author to whom correspondence may be addressed. E-mail address:
rrdewey@syr.edu
contactors often dictates that they be operated at high dispersed
phase volume fractions and a high degree of agitation to ensure
sufficient interfacial contact between the phases. Operation
under such conditions, if not properly controlled, is always
performed at the risk of the occurrence of phase inversion.
The phase inversion behavior is affected by the physical
properties of the liquids which constitute the dispersion as
well as by geometric factors such as the type and number of
impellers and materials of construction of the agitator and
vessel. Effects of various physical properties, such as density,
viscosity, interfacial tension, and droplet surface charge,
have been examined by several investigators (Quinn and
Sigloh, 1963; Selker and Sleicher, 1965; Luhning and
Sawistowski, 1971; McClarey and Mansoori, 1978;
Guilinger et al., 1988; and Tobin and Ramkrishna, 1992).
Other investigators have studied the effects of geometric
parameters. For example, Gilchrist et al. (1989) showed that
the spacing between the baffles and vessel wail can affect
drop coalescence at low agitation speeds. Kumar et al.
(199 1) examined the effects of the material of construction
of the impeller on phase inversion behavior. They found
that, if the materials of construction of the impeller were
preferentially wetted by the dispersed phase, the drops could
coalesce on the impeller and form a thin film. If this film
were to break on or around the impeller, the phase inversion
characteristics of the system could be changed.
Phase inversion time delay has also been studied to some
extent (Gilchrist et al., 1989; Kato et al., 1991; and Pacek
et al., 1994). Gilchrist et al. (1989) examined the delay time
associated with phase inversion by making experimental use

486 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, JUNE, 199X
 I 1 1 I
Aqueous Dispersed I

-c
._
. Aqueous Dispersed C 0 R9 i Dlrectm of OW to WIO Phasn Inversion

P
U
W

-5
9 -
-

-
\ Ambivalence Region
.
2 b.

W
m
c - +
Direction of WIO lo O W Phase Inversion
a a

i
Organic Dispersed

':
Organic Dispersed

does not necessarilyreturn

900 400 600 800 1000 1200 1400 1600 1800

Agitation Speed [rpm] Agitation Speed [rprn]

Figure 1 -General ambivalence behavior for an immiscible liquid-
liquid system.
of two rules of thumb. The first rule states that which ever
phase is present in less than about 0.3 volume fraction will
be the dispersed phase, and the second states that the phase
in which the impeller is immersed at initiation of agitation
will be continuous. Gilchrist et al. (1989) created WIO dis-
persions of approximately 40% chlorobenzene hy placing
the impeller in the organic phase before starting agitation.
These dispersions are then inverted to OIW dispersions after
some delay time. They found that over a narrow range of
concentrations the delay time varied from zero (instantane-
ous inversion) to infinity (inversion never occurred). The
delay time was found to depend on fluid dynamic features
and liquid physical properties.
Kato et al. (1991) briefly studied dynamic phase inversion
by subjecting a steady-state dispersion of 56% n-hexane in
water (OIW) to an instantaneous increase in agitation speed.
They reported that following the agitation speed increase,
the dispersion inverted after a delay time of 1.2 min. They
were then able to reverse the experimental path by instantane-
ously decreasing the agitation speed and the WIO dispersion
inverted back to an OIW dispersion after a delay of 3.4 min.
Pacek et al. (1 994) employed a video technique to follow
drop size distributions and viewed secondary dispersions
while performing phase inversion delay time experiments.
They found that the delay time of phase inversion in their
system depended upon whether the impeller was started
from rest or the agitation speed was changed in a step fashion.
The traditional method of plotting organic phase volume
fraction versus the agitation speed at phase inversion
(Figure 1) is useful in that it clearly defines the conditions
under which only a single dispersion morphology may be
maintained. However, an alternative method of displaying
ambivalence behavior (Kumar et al., 1991) plots the volume
fraction of the initially dispersed phase at phase inversion,
regardless of which phase was dispersed, versus the agitation
speed at phase inversion (Figure 2). This method is useful in
that it more clearly shows the hysteresis effect, and thus,
demonstrates the irreversibility of the phase inversion process.
In this study, a number of behavioral features relevant to
phase inversion in liquid-liquid dispersions are explored.
Phase inversion (ambivalence) curves are experimentally
determined for liquid-liquid systems of various physical
properties, and the curves are compared to ascertain the
effects of these properties. Also, the effects of some geometric
Figure 2 - Toluene and water ambivalence data obtained in this
work and plotted by the method of Kumar et al. (1991). This plot
also reveals the hysteresis loop and shows the schematic of exper-
imental pathways used in the phase inversion time experiments.
TABLE1
Diameters and DIT Ratios of Impellers Used in the Phase
Inversion Experiments
Impeller Diameter
Impeller Type (m) DIT
Six Blade Rushton Turbine 0.04234 0.417
Six Blade Rushton Turbine 0.06294 0.619
Six Blade Rushton Turbine 0.05061 0.498
Marine Propeller 0.05080 0.500
parameters, such as impeller type and impeller-to-tank diam-
eter ratio (Dm, on phase inversion behavior are examined.
The delay time required for a given dispersion to completely
invert from one morphology to another is examined by two
different experimental pathways. Results of electrical con-
ductivity measurements are presented which show that
phase inversion occurs rather rapidly. Unique experimental
results are also shown which suggest that droplet surface
charge may induce the formation of secondary dispersions
(similar to those observed by Pacek et al. (1994)). Inversion
curves are plotted both by the traditional method and by that
employed by Kumar et al. (1991). The plots created by the
latter method are examined, and a new alternative interpre-
tation to that given by Kumar et al. (1991) is offered.
Experimental
APPARATUS
All experiments were performed in a 0.833 L cylindrical
vessel of Rushton geometry. The vessel was constructed of
a glass cylinder mounted vertically between two stainless
steel plates. The cylinder was of 0.102 m I.D. (T, by 0.102
m in height (H), yielding a TIH ratio of unity. The vessel
was baffled with four stainless steel, equally spaced, vertical
baffles.
A single impeller was centrally located in the vessel. Two
impeller geometries were used, a six blade Rushton turhine
and a marine propeller. The dimensions of the impellers and
the corresponding D/T ratios are shown in Table 1.
Agitation was provided by an Electro-Craft Corporation
Model 158335 DC servo motor controlled by a Lambda

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, JUNE, 1998 4x7
 TABLE2
Physical Properties of the Liquid-Liquid Systems Used in the Phase Inversion Experiments
Liquid-Liquid System
Toluene /Water 867
Toluene + CCl, (4.54 wt%) / Water 900
Toluene + CCl, (23.8 wt%) /Water 1040
Toluene + CCl, (43.I wt%) / Water 1180
Toluene /Water + Glycerol (29.7 wt%) 865
Toluene / Water + Glycerol (35.2 wt%) 864
Methyl Isobutvl Ketone /Water 805
Electronics Corporation Model LM DO-32 regulated power
supply or a Lightnin Labmaster IITM Model TS 2510
portable laboratory mixer. All fluids were pumped into the
vessel using an Ismatek SA micro metering pump. All tub-
ing was made of TeflonR. Electrical conductivity measure-
ments in the time delay experiments were made using a
CDM3 Radiometer-AIS Electrical Conductivity Probe con-
nected to either a Linseis Model L6512 chart recorder or a
LeCroy Model 93 10A Portable Digital Oscilloscope.
LIQUID-LIQUID
SYSTEMS
All liquid-liquid systems used in the phase inversion
experiments with density and viscosity as the parameters
consisted of an organic phase (toluene and carbon tetrachlo-
ride mixture) and an aqueous phase (water and glycerol
mixture). The addition of carbon tetrachloride allowed the
density of the organic phase to be varied without drastically
affecting the other pertinent physical properties, and the
addition of glycerol to the aqueous phase had a similar
effect on the aqueous phase viscosity.
Methyl isobutyl ketone (MIBK) was used in place of
toluene as the organic phase when the effects of the interfa-
cial tension were studied. MIBK was chosen because the
density and viscosity are comparable to those of toluene at
20° C, but the interfacial tension with water at that temper-
ature differs from that of toluene and water by approximately
a factor of 4.
The organic phase in the delay time experiments was
toluene, and the aqueous phase was water. In one set of
experiments, however, food grade sodium chloride (NaCl)
was added to the aqueous phase at a concentration of
approximately 1 mol/L. This addition was performed in
order to increase the ionic strength of the aqueous phase to
eliminate the effect of surface charge on the coalescence of
dispersed phase drops. After the addition of NaCl in water,
the viscosity, density, and interfacial tension with toluene
were measured. The differences from pure water properties
were within experimental error and considered insignificant.
A listing of the various liquid-liquid systems and the
accompanying physical properties is shown in Table 2.
EXPERIMENTAL
PROCEDURE
All experimental apparati were thoroughly scrubbed with
detergent and tap water. The equipment were then rinsed
seven times with both tap water and distilled water and
allowed to air dry. Care was taken to monitor all equipment
and solvents for contamination.
Creating dispersions with high dispersed phase volume
fractions (0.7 or 0.8) from rest was done as follows: these
dispersions were formed by semi-batch filling of the vessel
488 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, JUNE, 1998
Po,
0
(Pa*s) (Pass) (N/m)
996 0.00057 0.00096 0.0323
995 0.00059 0.00101 0.0299
996 0.00065 0.00097 0.0227
996 0.00073 0.00100 0.0263
1080 0.00057 0.00297 0.0278
1100 0.00058 0.00378 0.0275
977 0.00061 0.00102 0.0089
with an initial dispersion of a lower phase fraction and stir-
ring the dispersion at a fixed rate while adding only dis-
persed phase until the desired volume fraction was achieved.
The flow was then stopped, and the dispersion was stirred
for approximately 30 s. If phase inversion had not occurred
in the interim, the agitation speed was increased (for oil dis-
persed in water, OIW) by 20 rpm (at constant volume frac-
tion); and the dispersion was stirred for another 30 s. The
procedure was repeated until phase inversion was visually
observed. Phase inversion could be detected visually
because there would be a marked change in the light reflec-
tive qualities of the dispersion. It is important to note that
the duration of phase inversion was of the order of 1 s. The
vessel contents were measured after each experiment to
determine the actual volume fraction.
Two experiments were also conducted in which, after a
step change in the agitation speed, a longer time of 10 min
was allowed for the phase inversion to occur. In the first
experiment the initial dispersion was toluene dispersed in
water, while in the second experiment, the initial dispersion
was water dispersed in toluene. In both experiments, phase
inversion occurred only a few seconds after a step change
was made in the agitation speed. Before phase inversion
conditions were reached, the dispersion was stable for the
10 min time period between changes in the agitation speed.
These experiments showed that, for the fluids used in this
study, 30 s was enough time for phase inversion to occur
after a perturbation in the agitation speed.
The procedure for the WIO (water dispersed) ambivalence
experiments was similar to that of the OIW ambivalence
experiments, but differed in that the initial dispersions con-
sisted of large organic phase volume fractions and high agi-
tation speeds. The dispersed phase volume fraction was
adjusted by adding pure aqueous phase, and the agitation
speed was incrementally decreased until phase inversion
was observed. The above methods were employed at several
dispersed phase volume fractions for each type of dispersion
and each liquid-liquid system or impeller geometry.
Two types of phase inversion delay time experiments
were also performed. In these experiments, the electrical
conductivities of various toluene and water dispersions
(both directions, WIO to OIW and OIW to WIO dispersions
were examined) were monitored continuously through
phase inversion. Either the dispersed phase volume fraction
was varied at constant agitation speed or the agitation speed
was varied at constant volume fraction. The time required
for the electrical conductivity of a dispersion to make the
marked change from that indicating one dispersion mor-
phology to another was also examined.
In the first set of delay time experiments, a dispersion
(either initially OIW or WlO) was created at conditions which
were in the ambivalent region, and the volume fraction was
 1 1
E 0.8
O9 t 09t
01
+ + Org. Phase Density = 1040 kg/rn3
Agitation Speed [rprn]
Figure 3 - Ambivalence behavior of toluene
varying organic phase density.
u
1400 1600
+ CCI,-water
1800
with
varied at constant agitation rate. Pure dispersed phase was
stepwise pumped into the vessel (water for initially WIO dis-
persions and toluene for initially OIW dispersions), forcing
dispersion to exit the vessel and thus, increasing dispersed
phase volume fraction. This procedure was carried out at a
fixed agitation rate until phase inversion was observed both
visually and by a marked change in electrical conductivity.
These experimental pathways are shown schematically in
paths l(a) (for initially O N dispersions) and l(b) (for ini-
tially WIO Dispersions) in Figure 2. The quantity of dis-
persed phase added each time to the system was very small
as compared to the volume of the dispersed phase existing
in the vessel. After each addition, the system was agitated
for at least 10 min before the next addition. Immediately
after phase inversion, the liquid flow and agitation were ter-
minated. Once the phases separated, the vessel was drained,
and the volume fractions of the two phases were measured.
In the second set of delay time experiments, a dispersion
was created at conditions in the ambivalence region, and the
agitation speed was varied at constant volume fraction. The
initial dispersion was created at a dispersed phase volume
fraction and agitation speed which assured the desired initial
dispersion morphology. Then, pure dispersed phase was
introduced into the vessel until the desired dispersed phase
volume fraction (within the ambivalence region) was
achieved. The dispersed phase flow was then terminated,
and the dispersion was agitated for a minimum of 10 min
before the agitation speed was incrementally changed. The
agitation speed was incrementally increased in the case of
initially OIW dispersions (pathway 2(a), Figure 2) and
decreased in the case of initially WIO dispersions (pathway
2(b), Figure 2). The agitation speed was varied until phase
inversion was observed both visually and by marked change
in conductivity. The 10 min waiting period before incre-
menting the agitation speed was determined to be sufficient
as after that time period the electrical conductivity of the
dispersion had not changed from its value at the previous
agitation speed change.
Both of these experimental pathways were somewhat dif-
ferent from those employed by previous investigators.
Gilchrist et al. (1989) varied impeller position to ensure ini-
tial dispersion morphology in systems of chlorobenzene and
water. Pacek et al. (1994) either used an experimental path-
way which was similar to that of Gilchrist et al. (1989) or,
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, JUNE, 1998
6 0.2-
o.l - /.
w w
0 MlBWater (Low Pnterfacial Tension) 1
TolueneNVater (Hi h Interfacial Tension)
was in the form of a step change in agitation speed such as
that employed by Kato et al. (1991). The experimental path-
ways used in this work explored dynamic phase inversion
by either incrementally changing the agitation speed at con-
stant volume fraction or by incrementally changing the volume
fraction at constant agitation speed.
Results
RESULTSOF AMBIVALENCE EXPERIMENTS
Figure 3 shows the ambivalence behavior with varying
organic phase density in toluene + CCldwater systems. It is
shown that the density difference between the two liquid
phases does not appreciably affect the ambivalence behav-
ior. These results agree with those of Selker and Sleicher
(1965) and McClarey and Mansoori (1978), who also
showed that the density difference between the phases does
not greatly affect the ambivalence behavior provided that
the difference is not very large. In the event of a large den-
sity difference, i.e., an order of magnitude, there would be a
larger effect due to the difficulty in maintaining a complete
dispersion at lower agitation speeds.
Interfacial tension has a strong effect on the ambivalence
behavior of dispersions as shown in Figure 4 where the
results of the MIBK and water ambivalence experiments are
compared to those of the pure toluene and water system. The
upper ambivalence curve is not significantly different from
that of toluenelwater, but the lower ambivalence curve is
shifted downward significantly. In general, as the interfacial
tension decreases, the difficulty in inducing a phase inver-
sion increases, and the ambivalence region widens.
A lower interfacial tension yields higher drop breakup
rates and lower drop coalescence rates (Coulaloglou and
Tavlarides, 1977). Both effects suggest that higher volume
fraction of the drop phase is needed for phase inversion to
occur. Figure 4 shows that for the lower interfacial tension
system of MIBWwater, the phase inversion volume fraction
of the dispersed phase is higher than that of the toluene1
water system for phase inversion from WIO to OIW (aqueous
phase is initially dispersed in this case). This observation is
in agreement with the expected behavior. However, the
phase inversion volume fraction of the dispersed phase of the
MIBWwater system is lower than that of the toluene/water
489
 g
._
4-

u.
09

08-

07-
I
.. ' + . . 0 - z
0.9-

5 0.8-

5 06-
a,
-

0
1: L
'
0
,
Rushton Turbine
Marine Propeller
, . , . , , ' ' , . . . , . . I ' ' """' ' ' " "

$00 400 600 800 1000 1200 1400 1600 1800 2000 10" 100 10' 102 10'
Agitation Speed [rprn] Energy Dissipation Ratelunit Mass [rn2/s3]

Figure 5 - Ambivalence behavior of toluene-water + glycerol Figure 7 -Comparison of pure toluene-water ambivalence behavior
with varying aqueous phase viscosity. (with respect to energy dissipation rate per unit mass) for two dif-
ferent impeller types. Both impellers have DIT approximately
equal to 0.5.

1
0.05
09-

&- 0 8 -
4-

? 07-
LL
a,
06-

a,
00.15-
-
I

-0.2-
u
'z 0.3 -
m
-0.25 -

D/T = 0.498
-0.3
I 1
water Continuous

0 pr-l- - I
10' 100 10' 102 14 15 16 17 18 19 20 21 22
Energy Dissipation Ratelunit Mass [m2/s3] Time [s]

Figure 6 - Ambivalence behavior (with respect to energy dissipa- Figure 8 - Electrical conductivity trace showing OIW to WIO
tion rate per unit mass) of pure toluene-water with varying phase inversion time behavior.
impeller-to-tank diameter ratio.

system for inversion from OIW to WIO (organic phase is ini-
tially dispersed in this case). This observation is contrary to
the expected behavior and suggests that other factors beside
density, viscosity, and interfacial tension play a role in
phase inversion behavior.
The viscosity of a given phase plays a very important role
in the ambivalence behavior, regardless of whether that
phase is continuous or dispersed. Figure 5 shows that the
upper ambivalence curve moves upward as the continuous
phase viscosity is increased. However, this movement
appears to reach some upper limit as there is little apparent
difference between the upper ambivalence curves for a con-
tinuous phase viscosity of 0.00297 Pa.s and that of a contin-
uous phase viscosity of approximately 0.00378 P a s . This
upper limit, however, may be attributed to the difficulty in
obtaining the points of the ambivalence curve at these high
dispersed phase volume fractions due to the inability to
homogeneously disperse the system throughout the agitation
speed range of the experiment. The lower ambivalence curve
shifts downward with increasing dispersed phase viscosity.
This trend coincides with the findings of Selker and Sleicher
(1965); in that as the viscosity of a phase increases, its ten-
dency to be dispersed increases.
The phase inversion study also included an investigation
of some geometric parameters, such as the effect of chang-
ing the ratio of the impeller diameter to the tank diameter
(Dmon a toluene-water system using six blade Rushton
turbines of three different diameters over a range of agita-
tion speed from 480 to 1250 rpm. The resulting ambivalence
curves for the three impellers are plotted against energy dis-
sipation rate per unit mass in Figure 6. The curves are quite
close considering the scatter in the data. Because the three
impellers were so close in D R , and there is considerable
scatter in the data; no definite conclusions can be drawn
with regard to the effect of D R ratio. A wider range of
impeller diameters would have to be interrogated to yield
more conclusive results.
The effect of impeller type was also briefly investigated.
Two impellers were used in a toluene and water system: a six-
blade Rushton turbine and a marine propeller, both possessing

400 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, JUNE, 199X
 01
Organic Phase Volume Fraction after Phase Inversion = 0 35
-0 1
2-031
-0 8 1
-09:
10 12 14 16 18 20
Time [s]
Figure 9 - Conductivity trace showing large oscillations after
phase inversion.
D/T EL 0.50 and operated over a range of 500 to 1700 rpm.
When the organic phase volume fraction at phase inversion
is plotted against the energy dissipation rate per unit mass
(Figure 7), the ambivalence data for the two impeller types
are relatively close. The widths of the two ambivalence
regions match quite well; but it should be noted that the con-
stant used in calculating the energy dissipation for the
marine propeller was based on D/T = 0.40, as no power
number versus Reynolds number curves corresponding to
marine propellers with D/T = 0.50 were found in the litera-
ture.
Some ambivalence experiments were repeated to deter-
mine the reproducibility of results. The results for two
experiments generating the upper ambivalence curve for a
neutrally buoyant system of toluene + CCl,/water agreed to
within a 10% relative error, and the results for a similar set
of experiments with a toluene + CC1,Iwater system of
organic phase density equal to approximately 1200 kg/m3
yielded results which agreed to within 11% relative error.
RESULTSOF INVERSION TIME DELAY EXPERIMENTS
The results of the phase inversion time experiments
showed that the final catastrophic instability transient which
resulted in the complete phase inversion of the dispersion
occurred in a time span of only a few seconds. These results
are similar to those obtained by earlier investigators using
different experimental pathways and suggest that, although
the overall delay time between an operational parameter
change and phase inversion depends on the experimental
pathway, the time of the final catastrophic transient is essen-
tially independent of experimental path.
Figure 8 shows an example conductivity trace which cor-
responds to OIW to WIO phase inversion. As can be seen in
this figure, phase inversion was characterized by marked
changes in electrical conductivity. In several experiments
the conductivity trace after phase inversion from WIO to
OIW dispersions, by varying agitation speed, showed rather
large oscillations relative to the oscillations prior to phase
inversion (Figure 9). This behavior could be caused by the
formation of secondary dispersions as described by Pacek
et al. (1994). This observation was confirmed by the fact
that when agitation was terminated after phase inversion, the
two liquid phases took an excessive amount of time to sep-
THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, JUNE, 1998
01
0-
-0 1 -
-0 2 -
i E -0 3 -
al
:,-04-
-0
c
> -05-
-0 6 -
-0 7
22
i
1
24
-
-0.g8 10 0 12 14 816 18 20
8
Time [s]
Figure 10 - Conductivity after termination of agitation.
arate. Also, even for long times after bulk phase separation,
the aqueous phase remained turbid indicating a suspension
of very small toluene droplets in the aqueous phase. Figure
10 shows a conductivity trace for the above described experi-
ment after agitation was stopped. The oscillations in conduc-
tivity decrease markedly as the dispersions begin to disengage,
and the disengagement takes a considerable amount of time.
The reason for the formation of droplets within drops is
not well understood. Pacek et al. (1994) saw the formation
of droplets within drops of glyceroYwater in chlorobenzene
just prior to phase inversion. They video taped the formation
of very large dispersed phase drops just prior to phase inver-
sion. This drop formation resulted in a marked increase in
conductivity fluctuations and coincided with the observation
of drops of chlorobenzene in all but the smallest drops of
water. Following phase inversion, the drops of chloroben-
zene in water were not seen.
It was thought that electrostatic repulsion due to double
layer forces might have an effect on the formation of a sec-
ondary dispersion during phase inversion from WIO to 0 1W.
Tobin and Ramkrishna (1992) reported that surface charge
on drops can substantially inhibit drop coalescence in agi-
tated liquid-liquid dispersions. To examine this possibility,
we repeated the phase inversion experiments in which we
observed the large conductivity fluctuations and excessive
phase separation times with food grade sodium chloride
(NaCI) dissolved in the aqueous phase at a concentration of
about 1 m o l L Conductivity traces for two experiments are
shown in Figure 11. Curve (a) is the conductivity trace for a
dispersion without NaCl in the aqueous phase, and curve
(b). is that for a dispersion with NaCl dissolved in the aque-
ous phase. From Figure 11, it is clear that there were no
large fluctuations in the conductivity after phase inversion
in the NaCl added system, and the fluctuations prior to
phase inversion are of the same order of magnitude as in the
experiment where no NaCl was used. Also, it was observed
that the aqueous phase which was initially turbid due to the
presence of small toluene droplets was clear after phase
inversion, and the individual phases separated quite quickly
after agitation ceased.
It is apparent that the increase in the ionic strength, by the
addition of 1 m o m NaCl, eliminated the secondary disper-
sion. This observation, however, needs to be proved by visual
observation of the drops. The presence of ions, in general,
49 I