Powder Technology 101 Ž1999.

142–150

Experimental determination of bubble size distributions in bubble

columns: prediction of mean bubble diameter and gas hold up
´
P.L.C. Lage ) , R.O. Esposito
´
Programa de Engenharia Quımica, COPPEr UFRJ, C.P. 68502, 21945-970, Rio de Janeiro, RJ, Brazil
Accepted 17 June 1998

Abstract

Bubble diameters were measured photographically in a bubble column, which was operating in the homogeneous regime with air and
aqueous isopropanol solutions. The bubble size data were determined for several values of the superficial gas velocity, and used to fit
bubble size distributions. The gas hold up was measured under the same conditions and its values were calculated from the bubble size
distributions, using the theory for the conservation of the bubble size distribution function. The experimental results were compared to
bubble size predictions given by models of bubble formation in orifices and to results from well-known correlations for the gas hold up. It
has been verified that the models for bubble formation in orifices give reasonable predictions of the mean bubble diameter, while the
correlations for gas hold up lead to poor agreement with the experimental data. The indirect measurement of gas hold up from the bubble
size distribution was considerably more accurate. q 1999 Elsevier Science S.A. All rights reserved.

Keywords: Bubbles; Size distribution; Size measurement; Gas hold up

1. Introduction In the homogeneous regime, with no mass or heat

transfer, the mean diameter of the bubbles in the column
Bubble columns can be employed in many mass trans-
can be obtained from several models which can predict the
fer processes. However, the lack of a more complete
volume of bubbles formed on submerged orifices under
knowledge on the bubble column fluid dynamic behavior
isothermal conditions, a subject reviewed recently by Pinto
in its various regimes causes several operational difficul-
w1x. If the gas hold up is experimentally determined, the
ties and design uncertainties, which include poor predic-
mean diameter of the bubbles can also be estimated using
tions of the mean bubble diameter, gas hold up and
a correlation developed by Molerus and Kurtin w2x for the
interfacial area.
homogeneous regime, which relates the mean diameter to
The goal of this present work was to analyze the
the gas hold up and the gas and liquid superficial veloci-
experimental determination and the theoretical prediction
ties.
of two fundamental fluid dynamics parameters in the
The bubbles usually present a non-uniform size distribu-
project of a bubble column: the gas hold up and the bubble
tion inside the bubble column. Moreover, as the bubble
diameter Žor size distribution. for the homogeneous bub-
size varies, the bubble shape and the bubble ascension
bling regime. These parameters permit the determination
velocity also change. In the homogeneous bubbling regime,
of the interfacial area, which is of fundamental importance
the bubbles may be assumed to rise unperturbed Žlittle
in the computation of the mass transfer rates between the
liquid circulation. leading to a direct relationship between
disperse Žgas. and continuous Žliquid. phases.
the bubble ascension velocity and bubble residence time.
The homogeneous regime of bubbling is the simplest of
This facilitates the use of experimental data to determine
such regimes, and can be obtained at low superficial
the bubble size distribution function inside the column
velocities of the gas, using a carefully designed gas-distri-
using its conservation equation.
bution system.
The gas hold up can be measured directly from the
expansion of the liquid column. Of the correlations found
)
Corresponding author. Fax: q55-21-290-6626; E-mail: in the literature for the gas hold up prediction in bubble
paulo@peq.coppe.ufrj.br columns, those developed by Hikita et al. w3x and by Akita

0032-5910r99r$19.00 q 1999 Elsevier Science S.A. All rights reserved.

PII: S 0 0 3 2 - 5 9 1 0 Ž 9 8 . 0 0 1 6 5 - X
 ´
P.L.C. Lage, R.O. Espositor Powder Technology 101 (1999) 142–150
and Yoshida w4x are considered the most accurate, with a
mean error of 11% w5x.
This work has experimentally determined the bubble
size distribution in a bubble column operating in the
homogeneous regime. The theory for the conservation of
the bubble size distribution function w6x was used to derive
an expression to calculate the gas hold up inside the
column, using the measured bubble size distributions.
These results were compared with those obtained using
several prediction methods for mean bubble diameters and
gas hold up.
2. Experimental procedure
For a bubble column in a semi-batch operation in the
homogeneous regime, the experimental measurements nec-
essary to characterize its fluid dynamics are gas flow rate,
height of the liquid column and bubble diameter. The
experimental system was built in the Thermofluid Dynam-
ics Laboratory of the Programa de Engenharia
´
QuımicarCOPPE, and is shown schematically in Fig. 1.
Depending on the desired range of gas flow rate, the gas
from the compressed air line passed through one of three
calibrated rotameters arranged in parallel. During each
experimental run, the valve upstream of the rotameter was
kept totally open, and the gas flow rate controlled com-
pletely by the needle valve before the distributor. The
distributor consisted of a sintered-glass plate, which was
used to guarantee a uniform distribution of the gas flow,
and by a perforated copper plate, in whose orifices the
bubbles were formed.
The internal bubble column diameter was Ž7.20 " 0.02.
cm. The pressure gauge of the feed line was kept at 4
kgfrcm2 at all experimental points, in such a way that the
pressure drop of the distribution system was maintained
approximately constant. The perforated plate, with a diam-
eter of 6.4 cm, had 88 orifices uniformly distributed over
its area, each with a diameter of 0.5 mm. The diameter of
Fig. 1. Experimental setup.
143
the sintered-glass plate was also 6.4 cm. The thickness and
degree of sintering were high enough for most of the
pressure drop throughout the distribution system to be due
to the sintered plate. This, added to the limited volume of
the chamber localized before the distributor, guaranteed
bubble formation under the constant flow rate regime w1x.
The liquid column heights before and during the opera-
tion were obtained with the help of a graduated scale
attached to the column, varying between 50 and 80 cm.
These measurements allowed for the direct determination
of the gas hold up, whose experimental error varied be-
tween 10 and 40%, becoming greater as the air flow rate
increased, due to froth formation. Care was taken not to
allow leakage of liquid through the distribution system by
pouring the liquid solution together with a small flow of
gas. The gas flow was then stopped in order to begin each
experimental run.
The diameters of the bubbles were determined from
photographs of the operating column, 40 cm above the gas
distributor. The dimensions of the bubbles, which were
supposed to be ellipsoids, were measured directly from the
photographs, with the help of a millimetre. The correction
to the real size was based on the graduated scale attached
to the column, which was at the same focal distance as the
measured bubbles. In fact, the focus was adjusted on the
graduated scale and only the well focalized bubbles were
measured. In order to reduce the optical errors, only those
bubbles situated in the central region of the column were
measured. Assuming an ellipsoidal form, the equivalent
bubble diameter was obtained, which is the diameter of the
sphere whose volume is that of the bubble. About 50 to
100 bubbles were measured at each experimental point,
using 2–3 photographs. The error in the measurement of
each axis of the ellipse corresponding to the cross sectional
area of the bubble in the photograph, is about 6%. Thus, if
the imprecision of the ellipsoidal bubble hypothesis and
the optical error are added to this error, it is possible to
estimate the experimental error in the determination of the
equivalent bubble diameter in the range from 10 to 15%.
Under the homogeneous regime hypothesis and for the
short columns used, it will be shown that the bubble size
distribution does not vary appreciably along the column.
Thus, the size distribution obtained in the column region
shown in the photographs, is approximately valid for the
entire column.
The liquids used in the experimental runs were aqueous
solutions of isopropanol, a surfactant agent which hinders
the phenomenon of bubble coalescence, making it easier to
keep the column operating in the homogeneous bubbling
regime ŽFig. 2.. The measurements were made for three
different isopropanol concentrations: 0.5%, 1% and 2%
vrv.
For all experimental runs, the ambient temperature was
in the 23–268C range. This small variation in temperature
hardly affects the physical properties of the system. In the
following calculations, the physical properties of the gas–
144 ´
P.L.C. Lage, R.O. Espositor Powder Technology 101 (1999) 142–150

Fig. 2. Bubble column Ž u G s 0.65 cmrs, aqueous 2% vrv isopropanol solution..

liquid system were evaluated at 258C. The properties of the
liquid phase were considered to be those of the pure water,
except for the interfacial tension, which was estimated
using the correlation of Szyszkowshi w7x for each iso-
propanol solution. The values obtained for the 0.5%, 1%
and 2% vrv isopropanol solutions were 67, 63 and 58
dynrcm, respectively.
distributions are almost identical, which facilitates the
experimental determination of the bubble distribution func-
tion.
According to the above cited hypotheses, the steady-state
conservation equation for the size distribution function of
the bubbles can be obtained by a development similar to
that used for perfectly mixed reactors w8,9x:

f Ž d e . NF Ž d e . s n I I Ž d e . Ž 1.
3. Conservation of the bubble distribution function
where d e is the equivalent bubble diameter, N and F Ž d e .
The bubble size distribution inside a bubble column are, respectively, the total number and the size distribution
depends on the size distribution of the injected bubbles function of the bubbles inside the bubble column, f Ž d e . is
Žformed at some submerged orifice., on the coalescence the bubble escape frequency, and n I and I Ž d e . are the
and break-up phenomena, on the bubble escape frequency, number rate of injected bubbles and their size distribution
on the gas density changes with hydrostatic pressure and function, respectively.
on a possible existing mass transfer process between the Using the normalization condition for the two size
liquid and gas phases. distribution functions, N can be obtained by:
For a short bubble column, the effect of the hydrostatic
pressure can be neglected. In addition, for the absorption ` I Ž de .
of a slightly soluble gas, the mass transfer effect on the N s nI H0 d Ž d e . or N
f Ž de .
bubble distribution can also be neglected, and for a non-
coalescent system at low gas superficial velocities, the y1
`
coalescence and break-up phenomena are negligible. In s nI H0 f Ž de . F Ž de . d Ž de . Ž 2.
this case, it will be shown that the bubble local and global
 ´
P.L.C. Lage, R.O. Espositor Powder Technology 101 (1999) 142–150 145

From F Ž d e ., the gas hold up can be calculated from its the bubble drag coefficient, given as a function of the
definition: Reynolds number, Re s Ud erÕ. The asymptotic value for
N p `
C D for large Re was modified to 0.988, a value obtained
´s 3
H d F Ž d . dŽ d .
e e e Ž 3. by Karamanev w10x himself from a correlation of bubble
VT 6 0 ascension experimental data using 32 different gas–liquid
where VT is the total volume of the liquid–gas system. systems. This value was applied for Re G 126, in order to
Since n I is equal to the gas flow rate divided by the mean render the correlation for C D continuous. The d erd h ratio
volume of the injected bubbles, Eqs. Ž1. and Ž2. lead to: is better represented by the Vakrushev and Efremov corre-
lation w12x, which was therefore used in this work.
`
3
H0 d F Ž d . d Ž d .
e e e
´ s uG `
Ž 4.
3 4. Bubble diameter determination
HH0 d ef Ž de . F Ž de . d Ž de .
4.1. Correlation of Molerus and Kurtin
where u G is the superficial gas velocity and H is the
liquid–gas system height inside the bubble column. In order to verify if the bubbling regime was homoge-
In the homogeneous bubbling regime, the escape bubble neous, the correlation developed by Molerus and Kurtin w2x
frequency can be determined using only the absolute bub- was used to interpret the experimental results with respect
ble ascension velocities. Let L be the upper limit for the to the superficial velocity of the gas and the gas hold up.
equivalent bubble diameter for a given distribution. If For semi-batch operations, the superficial velocity of the
Õ Ž L. is the absolute bubble ascension velocity of the liquid phase is zero. In this case, this correlation provides
largest bubble, that is, it is the largest ascension velocity, the gas hold up, ´ , as a function of a dimensionless
then during the D t s HrÕ Ž L. interval, all bubbles in the relative velocity of the gas, defined by:
bubble column with d e s L will leave it. For d e - L, the
1r3
fraction of the existing bubbles leaving the bubble column r L y rG u G3
during this time interval is D zrH, where D z s D t Õ Ž d e ..
Thus, the fraction of bubbles with diameter d e that leave
gs
ž rL
P
ÕPg / Ž 7.

the column per unit of time, which is the bubble escape in the following form
frequency, is given by D zrŽ H D t . s Õ Ž d e .rH, which is r0 r0 2
independent of the values of D t and L used in the above b 3 s 18 P Re P 1 q 0.347 P P žd/ q 3 Re1.5 q Re 2
deduction. Accordingly: d
r0
f Ž d e . s Õ Ž d e . rH Ž 5.
From Eqs. Ž1., Ž2. and Ž5., it is clear that F Ž d e . does
ž
P 0.3 q 0.424 P
d / Ž 8.

not depend on H, but does depend on Õ Ž d e ., while N is where

proportional to H. Eqs. Ž4. and Ž5. show that ´ is not 1r3
r L y rG d B3 g
directly proportional to u G because I Ž d e . and F Ž d e . de-
pend on u G .
bs
ž rL
P
Õ2 / Ž 9.

For bubble columns in semi-batch operation or operat- uG P d B

ing with low liquid flow rates, Õ Ž d e . f UŽ d e ., where Re s Ž 10 .
´PÕ
UŽ d e . is the relative bubble ascension velocity, which is a
y1
function of d e . Neglecting any effect of the bubble popula- r0 0.9
tion on the bubble drag coefficient, UŽ d e . may be consid-
ered as the terminal ascension velocity of an isolated
d
s
ž 3
'´
y1
/ Ž 11 .

bubble with an equivalent diameter d e .
Recently, Karamanev w10x developed a correlation for
bubble ascension in contaminated liquids, which incorpo-
rates the effect of the geometric shape of the bubble
through the Tadaki and Maeda w11x correlation, which
relates d e to the projected diameter of the deformed bubble
Žhydraulic diameter., d h . Karamanev w10x correlation is
given by:
1r2
U Ž d e . s Ž 4 gD r d er3 r C D . Ž d erd h .
where g is the gravity, r is the liquid density, D r is the
difference between the liquid and gas densities and C D is
Representing graphically the curves obtained from the
correlation ŽEqs. Ž7. – Ž11.., for various bubble sizes, and
the experimental data of ´ = g , it is possible to estimate
the mean bubble diameter. One such graph is shown in
Fig. 3 for the 2% vrv isopropanol solution.
4.2. Bubble formation models
For low superficial gas velocities in short bubble
Ž 6.
columns in the presence of surfactants and when there is
no mass or heat transfer, the coalescence and bubble
breakage phenomena are not relevant. Thus, the bubble
146 ´
P.L.C. Lage, R.O. Espositor Powder Technology 101 (1999) 142–150

mental data for the equivalent bubble diameters. About 8

to 12 values of accumulated fraction versus diameter were
used at each experimental point to obtain the bubble size
distribution function. Previous work showed that the log
normal distribution is adequate to represent the size distri-
bution for these experimental data w6x, which is given by:
2
1 ln Ž d erd 50 .
F Ž de . s
'2p de ln z exp y
½ '2 ln z 5 Ž 15 .

The values for the d 50 and z parameters were calcu-

lated by minimizing the sum of the squared errors w6x. Fig.
4 presents one of the size distributions obtained and the
corresponding experimental data. The agreement is quite
good.
The determination of the F Ž d e . distribution directly
from the experimental data is possible, because the differ-
ence between the F Ž d e . and I Ž d e . distributions are negli-
gible under the conditions analyzed. Fig. 5 exemplifies the
small difference between F Ž d e . and I Ž d e ., when I Ž d e . is
Fig. 3. Diagram of Molerus and Kurtin Žisopropanol solution at 2% vrv.. given by a known log normal distribution and F Ž d e . is
calculated from Eq. Ž1.. This will occur for small bubble
diameter inside the column equals the diameter of the size dispersions Žsmall z ., because the bubbles rise at
bubble formed at the orifice. In Section 4.3, some results practically the same velocity. The range of z for the
based on the experimental conditions and on the conserva- present experimental data was 1.1–1.3, which allows for
tion of bubble size distribution functions are shown, which the above hypothesis.
also support this hypothesis under the above stated condi- By definition the Sauter mean diameter of the bubbles
tions. inside the column can be calculated from the bubble size
The bubbles are formed under the constant-flow regime distribution:
due to the distributor design. Of the several bubble forma- y1
` `
tion models existing for this regime, we chose two models dS s 3 2
H0 d F Ž d . d Ž d . H0 d F Ž d . d Ž d . Ž 16 .
e e e e e e
which usually give accurate predictions for the air–water
system in the considered range of gas superficial velocities The integrals in Eq. Ž16. were calculated numerically
w1x. The first one is the Davidson and Schuler’s
¨ w13x model, by an adaptive integrator with global error control w15x.
recommended for high gas velocity flows, which provides
the bubble volume at the end of the formation step in the
form
Q 06r5
V B s 1.378 Ž 12 .
g 3r5
The second is the model of Gaddis and Vogelpohl w14x,
recommended for low gas velocity flows, which provides
the bubble volume at the end of the formation step by the
numerical solution of Eq. Ž13.:
c1 c2
V B s c 0 q 1r3 q 2r3 Ž 13 .
VB VB
where
p d0 s Q0 Õ Q02
c0 s , c1 s 10.9 , c 2 s 1.16 Ž 14 .
rL g g g

4.3. Bubble size distributions

The accumulated fraction of the bubbles with diameter Fig. 4. Bubble size distribution Ž u G s 0.65 cmrs, isopropanol solution at
lower than a given value was obtained from the experi- 2% vrv..
 ´
P.L.C. Lage, R.O. Espositor Powder Technology 101 (1999) 142–150 147

UŽ d e ., the correlation of Karamanev w10x was used, slightly

modified as explained previously.
The integrals of Eq. Ž4. were calculated numerically
using the same integrator described previously w15x. The
error associated with this indirect measurement of the gas
hold up is estimated as being, at most, equal to the error
associated with the determination of the volume of the
bubble whose diameter is equal to the Sauter mean diame-
ter, that is, 30%. The error in the gas hold up is, probably,
even smaller, due to the cancellation of errors in the
integrals involving d e in the numerator and denominator
of Eq. Ž4..

6. Results and discussion

6.1. Mean bubble diameter

The values for the Sauter mean diameter obtained for

Fig. 5. Comparison between the bubble size distributions at column inlet each experimental point of the three isopropanol solutions
and inside the column.
are shown in Table 1. The bubble diameters calculated by
the models of Davidson and Schuler ¨ w13x and Gaddis and
w x
Vogelpohl 14 are also listed in Table 1. It should be
noted that, while the first bubble formation model has no
Since it is a mean value, one can estimate that the error in
functional dependence on the liquid surface tension, the
the Sauter mean diameter must be lower than that obtained
second one does. However, the variation in the diameters
for the diameter of each bubble. Thus, an error of 10% in
calculated by the Gaddis and Vogelpohl model w14x using
d S is estimated. For the largest gas flow rates, this error
the surface tension of each solution is very small Žabout
might be larger, because fewer bubbles Žabout 50. were
0.005 cm.. Thus, the arithmetic mean values were com-
used in the determination of the bubble size distribution.
puted and presented in Table 1 for the Gaddis and Vo-
gelpohl correlation w14x.
Analyzing the values for d S , one can verify that there is
5. Gas hold-up determination
little difference between the values obtained for the solu-
tions at a given value of u G , if u G - 5 cmrs. Larger
5.1. Correlations
discrepancies occur for the two largest gas flow rate
values, where the estimated error in the distribution deter-
The correlations of Akita and Yoshida w4x and Hikita et
mination is larger, and for two intermediary gas flow rate
al. w3x allow for estimating the global gas hold up based on
values for the isopropanol solution at 2% vrv. Some kind
the physical properties of the phases and the column
diameter. They are given respectively by:
1 1
´ gd c2 r L 8 gd c3 12 uG
Ž1y´ .
4
s 0.2
ž s / ž / ž( /
Õ L2 gd c
Ž 17 . Table 1
Comparison between the experimental data and predictions for the bubble
diameter
y0 .131
uG m L 0.578 m4L g uG d S Žcm. d S Žcm. d S Žcm. d DS d GV
´ s 0.672 ž s / ž /
rL s 3
Žcmrs.
0.65
Ž2%.
0.36
Ž1%.
–
Ž0.5%.
0.32
Žcm.
0.21
Žcm.
0.28
0.062 0.107 1.17 – 0.40 – 0.30 0.31
rL mG
1.68 0.38 0.39 0.36 0.33 0.33
= ž / ž /
rG mL
Ž 18 .
2.20
2.71
–
0.39
0.41
0.39
–
0.37
0.35
0.38
0.36
0.38
3.23 0.38 0.41 – 0.40 0.40
5.2. Bubble size distributions 3.75 0.52 0.43 0.43 0.42 0.42
4.16 0.54 – – 0.43 0.44
5.72 – 0.53 – 0.48 0.49
Under the conditions analyzed, it is possible to apply 6.16 0.42 0.49 0.43 0.50 0.50
Eqs. Ž4. and Ž5. to determine the global gas hold up inside 7.02 0.41 0.55 0.47 0.52 0.52
the column with Õ Ž d e . s UŽ d e .. For the computation of
148 ´
P.L.C. Lage, R.O. Espositor Powder Technology 101 (1999) 142–150

of nonsystematic experimental error must have originated

this latter discrepancy.
A possible explanation for the discrepancies in the d S
values obtained for the highest gas flow rates is the change
in the bubbling regime, when liquid circulation and bubble
coalescence and break up begin to occur. Some effects of
liquid circulation were visually observed for the two larger
values of the gas flow rate. This and the analysis using the
correlation of Molerus and Kurtin w2x corroborate this
hypothesis.
The predictions for bubble diameter from the bubble
formation models are reasonable, except for the Davidson
¨
and Schuler’s model w13x for small superficial gas veloci-
ties. Compared to the experimentally determined d S val-
ues, the maximum error for the diameter estimated by the
correlation of Gaddis and Vogelpohl w14x is about 25%.
However, in general, the estimates of bubble diameter with
this correlation show errors of 10–15%, that is, in the
order of the experimental error in the bubble diameter
determination. Fig. 7. Diagram of Molerus and Kurtin Žisopropanol solution at 0.5%
Figs. 3, 6 and 7 show the experimental points ´ = g for vrv..
isopropanol solutions at 2%, 1% and 0.5% vrv, respec-
tively, together with the curves obtained using the correla-
tion of Molerus and Kurtin w2x for various values of bubble
diameter. It is clear that the solutions with 2% and 1% vrv solution at 0.5% vrv, Fig. 7 shows results presenting
isopropanol concentrations, for u G - 5 cmrs, follow the similar behavior for the other two solutions, but with a
curves obtained from the correlation for diameters between larger dispersion of the experimental data. In this case, the
0.3 and 0.4 cm. For u G ) 5 cmrs, the diameters become surfactant concentration might be too low to equally affect
larger, between 0.5 and 0.8 cm. This is not in accordance the bubble coalescence phenomena for every gas flow rate.
with the d S values shown in Table 1. Adding this to the
visual observations, it can be concluded that there was a 6.2. Gas hold up
regime transition for u G ) 5 cmrs. For the isopropanol
The results of the direct measurement of the gas hold up
are given in Table 2 for the three isopropanol solutions.
The same table includes the ranges of gas hold up values
given by the correlations of Akita and Yoshida w4x and
Hikita et al. w3x, using the estimated values of surface
tension for the three isopropanol solutions. Only the value
ranges are shown, because the value of the surface tension
does not appreciably affect the gas hold up predictions.
Finally, from the bubble size distribution functions ob-
tained at each experimental point for each isopropanol
solution, the gas hold up was calculated from Eq. Ž4.,
using the estimated surface tension for each of the iso-
propanol solutions. Although the bubble size distribution
functions obtained for the different solutions, for a given
u G value, are different, the values obtained for the gas
hold up are very close to each other, being equal for
several u G values. In this way, only the ranges obtained
for the gas hold up for the different u G values are listed in
Table 2.
From Table 2, it can be verified that the correlations of
Akita and Yoshida w4x and Hikita et al. w3x provide gas hold
up predictions which are too much smaller than the experi-
mental data. On the other hand, the gas hold up values
Fig. 6. Diagram of Molerus and Kurtin Žisopropanol solution at 1% vrv.. calculated from the bubble size distributions agree very
 ´
P.L.C. Lage, R.O. Espositor Powder Technology 101 (1999) 142–150 149

Table 2
Comparison between the experimental data and predictions for the gas hold up
u G Žcmrs. ´exp Ž2%. ´exp Ž1%. ´exp Ž0.5%. ´distr ´AY ´H
0.65 0.024 " 0.01 – 0.002 " 0.01 0.032–0.033 0.021 0.038–0.039
1.17 – 0.054 " 0.009 – 0.056 0.035–0.036 0.053–0.054
1.68 0.083 " 0.01 0.083 " 0.009 0.065 " 0.009 0.080–0.082 0.048–0.049 0.065–0.067
2.20 – 0.12 " 0.02 – 0.10 0.060–0.061 0.076–0.078
2.71 0.16 " 0.01 0.15 " 0.02 0.13 " 0.02 0.13 0.070–0.071 0.086–0.088
3.23 0.18 " 0.07 0.17 " 0.07 – 0.15–0.16 0.080–0.081 0.095–0.097
3.75 0.23 " 0.06 0.21 " 0.07 0.29 " 0.05 0.17 0.089–0.091 0.103–0.106
4.16 0.25 " 0.1 – – 0.18 0.096–0.098 0.110–0.113
5.72 – 0.34 " 0.09 – 0.25 0.120–0.121 0.132–0.135
6.16 0.42 " 0.1 0.38 " 0.1 0.30 " 0.1 0.28–0.29 0.125–0.127 0.138–0.141
7.02 0.49 " 0.1 0.51 " 0.08 0.35 " 0.1 0.31–0.33 0.136–0.138 0.148–0.153

well with the experimental values. Considering the esti- L diameter of largest bubble Žcm.
mated experimental errors of ´exp , only 7 of the 23 experi- N number of bubbles in bubble column
mental points do not agree with the gas hold up values n number rate of bubbles Žsy1 .
calculated using the bubble size distributions. Q volumetric flow Žcm3rs.
Re Reynolds number
r0 parameter in Eq. Ž8. Žcm.
7. Conclusions U terminal rise velocity Žcmrs.
u superficial velocity Žcmrs.
The homogeneous regime in a bubble column was V volume Žcm3 .
achieved with aqueous isopropanol solutions for superfi- Õ absolute ascension velocity Žcmrs.
cial gas velocities below 5 cmrs. Under these conditions, t time Žs.
the correlation of Molerus and Kurtin w2x was shown to be z axial coordinate Žcm.
a quick way of estimating the mean bubble diameter. The Greek letters
correlation of Gaddis and Vogelpohl w14x was shown to be d parameter in Eq. Ž8. Žcm.
¨
a little better than the Davidson and Schuler’s model w13x, Dr difference between liquid and gas densities
with predictions for the mean bubble diameter which agree Žgrcm3 .
with the experimental data in 10–15% of the cases. The ´ gas hold up
correlations of Akita and Yoshida w4x and Hikita et al. w3x g dimensionless gas velocity
did not give reasonable predictions for the gas hold up, r density Žgrcm3 .
showing errors of up to 100%. On the other hand, the m dynamic viscosity Žpoise.
indirect measurement of gas hold up using the experimen- n kinematic viscosity Žcm2rs.
tally obtained bubble size distribution, was shown to be s surface tension Ždynrcm.
very accurate. The gas hold up values obtained from the z parameter of lognormal distribution
bubble size distribution appear to be more accurately Subscripts
determined than the distribution itself, by a fortuitous AY Akita and Yoshida correlation
cancellation of errors. B bubble
c column
distr from size distribution
8. Nomenclature DS Davidson and Schuler¨ model
e equivalent
CD drag coefficient exp experimental
ci parameters defined in Eq. Ž7. G gas phase
d diameter Žcm. GV Gaddis and Vogelpohl model
d 50 parameter of the log normal distribution Žcm. H Hikita et al. correlation
F size distribution function in bubble column Žcmy1 . h hydraulic
f bubble escape frequency Žsy1 . I injected bubbles
g gravity acceleration Žcmrs 2 . L liquid phase
H liquid–gas system height Žcm. o orifice
I size distribution function of injected bubbles S Sauter
Žcmy1 . T total Žliquid and gas phases.
150 ´
P.L.C. Lage, R.O. Espositor Powder Technology 101 (1999) 142–150
Acknowledgements
The authors are grateful for the support received from
CNPq, grant number 523268r96-3.
References
w1x A.D. Pinto, Bubble formation at orifices: initial evaporation and
bubbling processes Žin Portuguese., MSc Thesis, Programa de En-
´
genharia Quımica COPPErUFRJ, Chaps. 3 and 4, 1990.
w2x O. Molerus, M. Kurtin, Hydrodynamics of bubble columns in the
uniform bubble regime, Chem. Eng. Sci. 40 Ž1985. 647–652.
w3x H. Hikita, S. Asai, K. Tanigawa, K. Segawa, M. Kitao, Gas hold up
in bubble columns, Chem. Eng. J. 20 Ž1980. 59–67.
w4x K. Akita, F. Yoshida, Bubble interfacial area and liquid phase mass
transfer coefficient in bubble columns, Ind. Eng. Chem. Proc. Des.
Dev. 13 Ž1974. 84–91.
w5x W.D. Deckwer, A. Schumpe, Improved tools for bubble column
reactor design and scale-up, Chem. Eng. Sci. 48 Ž1993. 889–911.
w6x P.L.C. Lage, The non-uniform bubble size distribution in bubble
columns: the choice of bubble mean diameter for mass transfer Žin
Portuguese., Proc. of XXVI ENEMP ŽBrazilian Congress on Particu-
late Systems. II, 1996, 631–636.
w7x R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases and
Liquids, 4th edn., Mc Graw-Hill, New York, 1987, 648.
w8x K.J. Valentas, O. Bilous, N.R. Amundson, Analysis of breakage in
dispersed phase systems, Ind. Eng. Chem. Fundam. 5 Ž2. Ž1966.
271–279.
w9x K.J. Valentas, N.R. Amundson, Breakage and coalescence in dis-
persed phase systems, Ind. Eng. Chem. Fundam. 5 Ž4. Ž1966.
533–542.
w10x D.G. Karamanev, Rise of gas bubbles in quiescent liquids, AIChE J.
40 Ž1994. 1418–1421.
w11x T. Tadaki, S. Maeda, On the shape and velocity of single air bubbles
rising in various liquids, Kagaku Kogaku 25 Ž1961. 254–264.
w12x R. Clift, J.R. Grace, M.E. Weber, Bubbles, Drops and Particles,
Academic Press, New York, 1978, 181.
w13x ¨ Bubble formation at an orifice in an
J.F. Davidson, B.O.G. Schuler,
inviscid liquid, Trans. Inst. Chem. Engrs. 38 Ž1960. 335–442.
w14x E.S. Gaddis, A. Vogelpohl, Bubble formation in quiescent liquids
under constant flows conditions, Chem. Eng. Sci. 41 Ž1986. 97–105.
w15x P.L.C. Lage, R.H. Rangel, Total thermal radiation absorption by a
single spherical droplet, J. Thermophys. Heat Transfer 7 Ž1993.
101–109.