Chemical Engineering Science 62 (2007) 6263 6273

www.elsevier.com/locate/ces

Correction of the penetration theory based on mass-transfer data from bubble

columns operated in the homogeneous regime under high pressure
Stoyan Nedeltchev ,1 , Uwe Jordan, Adrian Schumpe
Institute of Technical Chemistry, Technical University of Braunschweig, Hans-Sommer-Strasse 10, 38106 Braunschweig, Germany
Received 20 November 2006; received in revised form 15 May 2007; accepted 17 July 2007
Available online 24 July 2007

Abstract
A new correction term was developed which allows the classical penetration theory to be applied successfully to kL a data obtained from oblate
ellipsoidal bubbles formed in bubble columns operated in the homogeneous regime at various pressures (0.1.4.0 MPa). The correction factor
is a function of both the Etvs number Eo and dimensionless gas density ratio. The new correlation was compared with literature kL a data
in 18 pure organic liquids, 14 adjusted liquid mixtures and tap water. In some of the liquids (tetralin, xylene and ethanol) not only air but also
other gases (nitrogen, helium and hydrogen) were used. In total, 263 experimental kL a points are tted with an average relative error of 10.4%.
In the theoretical approach for the kL a prediction, the gasliquid contact time (used in the penetration theory) is dened as the ratio of
bubble surface to the rate of surface formation. All further calculations are based on the geometrical characteristics (bubble length and height)
of an oblate ellipsoidal bubble. It was found that the new correction factor fc gradually reduces with the increase of both supercial gas
velocity uG and gas density G (operating pressure P ).
2007 Elsevier Ltd. All rights reserved.
Keywords: Bubble columns; Mass transfer; Absorption; Bubble; Organic liquids; Modelling

1. Introduction
Bubble columns are simple but very effective gasliquid
contacting systems. In these chemical reactors, the dispersion of
gas in the form of bubbles is achieved by passing the gas through
multiple orices into the liquid. The variables deserving special consideration are the size and shape of the bubbles, their
rise velocity, circulation, coalescence, effects of surface-active
substances and the rate of mass transfer. Bubble volume and
bubble shape determine the surface-to-volume ratio which is a
very important parameter in determining overall mass-transfer
rates. It is essential in the scale-up and design of bubble columns
to be able to predict mass-transfer rates reliably. The volumetric liquid-phase mass-transfer coefcient kL a is one of the
Part of this work was presented at the Seventh GermanJapanese Symposium on Bubble Columns held in Goslar, Germany, May 2023, 2006.
Corresponding author. Fax: +49 531 391 5357.
E-mail address: snn13@gmx.net (S. Nedeltchev).
1 Present address: Institute of Chemical Engineering, Bulgarian Academy
of Sciences, 1113 Soa, Bulgaria.

0009-2509/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2007.07.030

most important parameters for the economy of the processes in

bubble columns.
The mass-transfer properties of the bubbles in liquids determine the efciency and dimensions of bubble columns. There
are many practical situations where the bubbles undergoing
mass transfer are not spherical. So, it is of value to attempt
to predict the effect of bubble shape on mass transfer. Wellek
et al. (1966) indicated that the change in bubble shape affects
not only interfacial area but also the liquid-phase mass-transfer
coefcient kL .
Bubbles with small Reynolds numbers ReB are usually
spherical. At high ReB , the inertial forces tend to cause a
distortion from the spherical case. As ReB is increased, bubble oscillation (unsteady-state distortion in shape) will set in.
Ultimately bubble break-up will occur. Bubbles moving in
low-viscosity liquids are rst distorted to approximately an
oblate ellipsoidal shape. The bubble shape can be characterized by the bubble eccentricity. The addition of surface-active
agents causes a decrease in bubble distortion.
Wellek et al. (1966) reported that bubble deformation reduces
as the liquid viscosity increases. Calderbank (1967) found a

6264

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

decrease in the eccentricity (maximum width to maximum

height) with increasing viscosity, accompanied by the appearance of tails behind small bubbles and of spherical indentations in the rear surfaces.
Since the size, shape and velocity of a bubble may change appreciably as it rises and dissolves, its mass-transfer coefcient
may vary accordingly. Calderbank and Lochiel (1964) showed
that the bubble dissolves more quickly than it expands due to
the decrease in hydrostatic head above it. If the viscous or inertial forces do not act equally over the surface of a bubble, they
may cause it to deform and eventually break up.
Lochiel and Calderbank (1964) as well as Calderbank (1967)
have derived mass-transfer correlations for both solid and uid
spheres, uid oblate spheroids and uid spherical caps. All
of these correlations are based on the classical penetration
theory put forward by Higbie (1935). These mass-transfer
equations are of importance in gas absorption, liquidliquid
extraction, etc. Lochiel and Calderbank (1964) have shown that
steady-state mass-transfer conditions around a bubble become
established as soon as the bubble moves through a distance
approximately equal to its own equivalent spherical diameter. Miller (1974) demonstrated that mass transfer depends
on the mean bubble size. Bubble shape, motion and the frequency and extent of interface uctuations also affect the mass
transfer. Kawase et al. (1987), Garcia-Calvo (1989, 1992)
and Garcia-Ochoa and Gomez (2004) have derived also some
promising mass-transfer correlations which involve aqueous
solutions.
According to Higbies (1935) penetration model, the liquidside mass-transfer coefcient kL is dependent on both the
molecular diffusion coefcient DL and the gasliquid contact
time tc :

4DL
kL =
.
(1)
tc
This equation holds when the Reynolds number is very high
(bubbles in potential ow) (Calderbank, 1967). The exposure or
contact time tc that characterizes the residence time of microeddies at the interface is generally unknown but it can be
described by an adequate model (Garcia-Ochoa and Gomez,
2004). The parameter tc can be expressed as a ratio of bubble
diameter to bubble rise velocity (Timson and Dunn, 1960;
Deindoerfer and Humphrey, 1961) or bubble surface to the
rate of surface formation (Nedeltchev et al., 2006a,b). Alternatively, the contact time tc can be evaluated by means of
Kolmogorovs theory of isotropic turbulence, i.e., as a ratio
between eddy length and eddy uctuation velocity (Deckwer,
1980; Garcia-Ochoa and Gomez, 2004).
According to Vasconcelos et al. (2002) Eq. (1) applies when
the interface is mobile, the contact time with the liquid is short
and the penetration of the dissolved gas is slight. An important
factor determining bubble mobility is size. It is important to
note that Higbies (1935) equation has not hitherto been tested
with systems of gas bubbles for extreme variations of diffusion
coefcients and liquid viscosities. Eq. (1) was developed for
a uid in potential ow with a short interface contact time.

A good description of the contact time is also given by Miller

(1974). Values of kL reduce with a decrease in bubble rise
velocity which prolongs the exposure time of liquid elements at
the bubble surface. It is noteworthy that the effect of the bubble
wake on the mass transfer is not explicitly taken into account in
Eq. (1). According to Calderbank (1967) the penetration theory
throws little light on problems of mass transfer in bubble wakes.
For ellipsoidal and spherical-cap bubbles Eq. (1) yields
higher theoretical kL values than the experimental ones (Miller,
1974). Zieminski and Raymond (1968) argue that the bubble
distortion lowers the mass-transfer coefcients. The authors
cited some papers where constant shape factors were introduced in order to correct for deviation from sphericity.
Zieminski and Raymond (1968) reported that above a bubble
size of 3 103 m the kL value gradually drops. Therefore,
some correction factor (lower than unity) should be introduced
which will broaden the application of the penetration theory to
other bubble shapes that are frequently observed at industrial
conditions.
1.1. Summary of the available correction factors due to
bubble shape differences
Timson and Dunn (1960) corrected the penetration theory
due to the nonsphericity of the bubbles. Miller (1974) introduced the following correction factor:
fc = 683de1.376 .

(2)

The author argues that by means of a correction term (Eq. (2))

based on the mean equivalent bubble diameter Eq. (1) can be
made generally applicable for bubble sizes above 4 103 m.
This dimensional correction factor was derived only for masstransfer data obtained during the stripping of carbon dioxide
from aqueous solution. Nedeltchev (2003) has proposed a similar correction term for kL a prediction in both toluene and
gasoline. Eq. (2) shows that the larger the bubble, the higher
is the correction factor. The larger bubbles are characterized
with greater frequency and extent of surface deformations. With
larger ellipsoidal bubbles, more surface disturbances occur,
which along with the larger bubble wakes lead to intensication of mass transfer (Miller, 1974).
Another correction factor for oblate ellipsoidal bubbles has
been introduced by Lochiel and Calderbank (1964), Calderbank
(1967) and Calderbank et al. (1970):

2.96
fc = 1
,
(3)
ReB
where
ReB =

de uB L
.
L

(4)

This correction factor implies that larger ellipsoidal bubbles will

have higher ReB and thus for them Eq. (1) should be multiplied
by a factor closer to unity. The applicability of this correction
term is limited to the experimental conditions employed by the
authors. In all these correction methods, the gasliquid contact

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

time was expressed as a ratio of bubble diameter to bubble rise

velocity.
Cockx et al. (1995) corrected also the relation for the specic
interfacial area with a coefcient that accounts for the shape
of the bubble in the case of oblate ellipsoids. This correction
factor also exhibits a slight decrease as the bubble size shrinks.
The correction term was tested only for oxygen transfer into tap
water at ambient pressure. Vasconcelos et al. (2002) developed
also a method to calculate the ellipsoidal surface area from the
bubble equivalent diameter de based on a shape factor which
is a function of the bubble aspect ratio.
All previously derived correction factors were valid only
under the experimental conditions tested by the authors. A
great disadvantage is the fact that the correction terms have not
been tested at high pressure. To the best of our knowledge, no
attempt was made in the literature hitherto to derive a generally
applicable correction term.
2. Mathematical model (Nedeltchev et al., 2006a)
The kL a value can be predicted if one knows how to estimate
both the liquid-side mass-transfer coefcient kL and the interfacial area a. These parameters depend on the bubble diameter.
Both kL and a are closely related to the bubble bed hydrodynamics. kL incorporates the effects of the complex liquid ow
eld surrounding the rising gas bubbles. The interfacial area a
inherently reects the system bubble behavior.
Gas bubbles are nonspherical, except when interfacial tension and viscous forces are much more important than inertial
forces (Clift et al., 1978). Most of the bubble sizes in this work
fall in the range of 1.4.6 103 m. Such bubbles are no longer
spherical and follow a zigzag or helical upward path. Viscous
drag is augmented by vortex formation in the wake, and velocity
of rise remains fairly constant over the bubble size range
(Miller, 1974).
As a rough approximation, Higbie (1935) assumed that the
average time of exposure for mass transfer (called contact time
or exposure time for mass transfer) can be estimated as follows:
tc =

Bubble surface
.
Rate of surface formation

(5)

In this work all theoretical kL a calculations were based on this

denition of the gasliquid contact time tc . The latter characterizes the residence time of the micro-eddies (responsible for
mass transfer in the liquid lm) at the interface, i.e., at the bubble surface. It is practically impossible to measure directly the
contact time tc , so usually it is estimated from correlations for
the bubble surface and the rate of surface formation.
Higbie (1935) postulated that a gas bubble moving through
a liquid splits the liquid at its advancing tip. The penetration
theory assumes unsteady-state absorption of a gas by a uid
element adjacent to the surface. The element moves at a uniform
velocity from the front of the bubble to the rear as penetration
into it occurs. Timson and Dunn (1960) showed that the bubble
surface increases when the spherical bubble is deformed into
an ellipsoid. This leads to higher contact time and thus lower
kL value.

6265

In order to calculate the volumetric liquid-side mass-transfer

coefcient kL a one also needs to know how to calculate the
interfacial area a. The formula for its calculation depends on
the bubble shape. The specic interfacial area a is a function
of the number of bubbles NB , the bubble surface area SB and
the total dispersion volume Vtotal (Painmanakul et al., 2005):
a=

N B SB
,
Vtotal

(6)

where Vtotal = AH .
The theoretically calculated kL a values were obtained as a
product of both Eqs. (1) (using Eq. (5) for the contact time)
and (6). Experimentally, on the other hand, it is much easier
to measure the product of kL and a than the individual values. In the homogeneous ow regime there is a narrow bubble
size distribution and the researchers use frequently the Sautermean bubble diameter ds for their mass-transfer calculations.
The bubbles rise uniformly in nearly straight lines and have
roughly uniform bubble size (Krishna, 2000; Lucas et al., 2005).
Therefore, it is a reasonable simplication in the homogeneous
ow regime to use the mean bubble diameter and disregard the
bubble size distribution.
For the sake of theoretical prediction of the kL a values
by means of Eqs. (1), (5) and (6) one needs to calculate the
bubble surface area SB , the rate of surface formation Rsf
and the number of bubbles NB . The calculation of the bubble
surface area SB depends on the bubble shape (Painmanakul
et al., 2005). Two excellent diagrams for bubble shape determination are available in the books of Clift et al. (1978) and
Fan and Tsuchiya (1990) in the form of loglog plots of the
bubble Reynolds number ReB vs. the Etvs number Eo with
due consideration of the Morton number Mo. A comparison
among the experimental conditions used in this work and the
above-mentioned standard plots reveals that the formed bubbles are no longer spherical but oblate ellipsoidals that follow
a zigzag upward path as they rise. Vortex formation in the
wake of the bubbles is also observed.
An oblate ellipsoidal bubble is characterized by its length
l (major axis of the ellipsoid) and its height h (minor axis of
the ellipsoid). The surface area SB of an ellipsoidal bubble
was calculated as follows (Fan and Tsuchiya, 1990; Nedeltchev
et al., 2006a,b):


 2
1
h
(1 + e)
l2
1+
ln
,
(7a)
SB = 
2
l
2e (1 e)
where the eccentricity e is

 2
h
.
e= 1
l

(7b)

In order to calculate both ellipsoidal bubble length l and

height h, the correlations of Terasaka et al. (2004) (for
2 < T a < 6) were used (see Nedeltchev et al., 2006a,b). The
Tadaki numbers Ta fell always in the range specied above.
However, the correlations of Terasaka et al. (2004) imply that
one needs to know a priori the bubble equivalent diameter de .

6266

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

Very often in the literature it is assumed that de can be

approximated by the Sauter-mean bubble diameter ds . Bubble
shape, motion and any tendency for the interface to ripple, uctuate or otherwise deform are all related to the bubble size.
In turn, bubble size is determined by the physical characteristics of the system and the operating conditions. The ds value
was estimated by means of the correlation of Wilkinson et al.
(1994) which is one of the most frequently cited in the literature.
This equation implies that the bubble size decreases as supercial gas velocity uG or gas density G (operating pressure
P ) increases. The calculated ds values for all liquids examined
correspond to an ellipsoidal shape.
The bubble equivalent diameter de of an ellipsoidal bubble
was calculated by assuming a sphere of volume equal to the
volume of the ellipsoidal bubble:
 2
de3
4
l h
= 
(8a)
VB =
6
3 2 2
or
de = (l 2 h)1/3 .

(8b)

Estimating the characteristic length of the ellipsoidal bubbles

with the same surface-to-volume ratio (the same ds value as
calculated from Wilkinson et al.s, 1994, correlation) required
an iterative procedure but led to only insignicantly different
values than simply identifying the equivalent diameter de with
ds when applying the correlations for both l and h (see Terasaka
et al., 2004).
The equivalent bubble diameter de is needed also for the calculation of the bubble rise velocity uB . The latter was estimated
by means of the correlation of Mendelson (1967). This equation is particularly suitable for the case of ellipsoidal bubbles.
Mendelson (1967) correlation for uB along with Wilkinson
et al.s (1994) relationship for ds estimation was also used to
calculate the bubble Reynolds number ReB (Eq. (4)) needed
for the estimation of both l and h values.
The bubble rise velocity uB and both the bubble length l and
height h of an ellipsoidal bubble take part in the calculation
of the rate of surface formation Rsf (Higbie, 1935). For oblate
ellipsoidal bubbles (see Nedeltchev et al., 2006a,b),

(l h)2
l 2 + h2
Rsf = 

uB .
(9)
2
8
Eq. (9) is needed for the calculation of the contact time tc
(Eq. (5)) and thus the liquid-side mass transfer coefcient kL
(Eq. (1)).
The number of bubbles NB was deduced from the bubble
residence time (aerated liquid height H divided by the bubble rise velocity uB ) and the bubble formation frequency fB
(Painmanakul et al., 2005; Nedeltchev et al., 2006a,b). The latter was expressed as the volumetric ow rate QG divided by
the bubble volume VB (see Eq. (8a)).
The substitution of the relations for NB and Vtotal into
Eq. (6) yield
a=

fB SB
.
A uB

(10)

Therefore, the theoretical kL a values for ellipsoidal bubbles can

be calculated as a product of kL value (estimated by means of
Eqs. (1), (5), (7a) and (9)) and a value (estimated by means
of Eq. (10)). Eqs. (1) and (10) have a theoretical background,
whereas Eq. (9) is based on a semi-theoretical consideration.
The theoretical kL a values have to be multiplied with some
correction factor fc in order to match satisfactorily the experimental kL a values. Specically, the accurate kL a values
for ellipsoidal bubbles should be calculated by means of the
following formula:

4DL Rsf fB SB
kL a = fc
.
(11)
SB A uB
The introduction of the correction factor fc could be attributed
to the fact that even when Eq. (1) is modied for oblate ellipsoidal bubbles it does not take into account the effect of the
bubble wake and surface disturbances. Due to these supplementary effects on the mass-transfer rate some additional term
should be introduced.
Nedeltchev et al. (2006a) have illustrated that at pressures up
to 1 MPa a good prediction of the experimental kL a values is
achieved when the correction factor is expressed as a function
only of the Etvs number Eo:
fc = 0.185Eo0.737 ,

(12)

where
Eo =

g(L G )de2
.


(13)

The Etvs number Eo represents the gravitational forceto-surface tension force ratio. As mentioned above, the bubble
shape depends on the Eo value. Eqs. (12) and (13) show that
as G increases and especially as bubble size reduces, Eo and
thus fc become lower.
Herein, we shall apply the above-described theoretical
method to conduct a systematic comparison between predicted
and experimental kL a results over a wide range of physicochemical properties, which are characteristic of gasliquid
systems. Both pure organic liquids and liquid mixtures as well
as tap water will be considered. In such a way, we shall assess
the capability of the correction method to predict the numerous
experimental kL a data available in the literature. The main goal
of this contribution is to improve our previous correction factor
(see Nedeltchev et al., 2006a) in order to render it applicable to
kL a data obtained at high pressures. Following the introduction
of an additional correction term accounting for the gas density
effect, the predicted kL a values at high pressures are improved
by as many as 18% (in terms of the maximum relative error).
3. Experimental results considered
Volumetric liquid-phase mass-transfer coefcients kL a measured in 18 organic liquids, 14 liquid mixtures and tap water
at 293.2 K (Tables 1 and 2) have been considered. Results for
1-butanol, ethanol (96%), toluene, decalin and tap water were
reported by Jordan and Schumpe (2001) for pressures P of 0.1,

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

6267

Table 1
Properties of the organic liquids and tap water (293.2 K)
Liquid

L (kg m3 )

L (103 Pa s)

 (103 N m1 )

DL (O2 ) (109 m2 s1 )

Acetone
Anilin
Benzene
1-Butanol
Cyclohexane
Decalin
Ethanol (96%)
Ethanol (99%)
Ethyl acetate
Ethylbenzene
1,2-Dichloroethane
1,4-Dioxane
Ligroin A (b.p. 90110 C)
Ligroin B (b.p. 100140 C)
Methanol
Nitrobenzene
2-Propanol
Tap water
Tetralin
Toluene
Xylene

790
1022
879
809
778
884
793
791
900
867
1234
1033
714
729
790
1203
785
1000
968
866
863

0.327
4.4
0.653
2.94
0.977
2.66
1.24
1.19
0.461
0.669
0.82
1.303
0.470
0.538
0.586
2.02
2.42
1.01
2.18
0.58
0.63

23.1
43.5
28.7
24.6
24.8
32.5
22.1
22.1
23.5
28.6
29.7
32.2
20.4
21.4
22.2
38.1
21.1
72.7
34.9
28.5
28.4

5.85a
0.97b
3.46a
1.29b
3.29a
1.60b
2.3b
2.37b
3.46a
2.94a
2.67a
1.97a
3.55a
3.29a
3.81b
1.63b
1.44b
2.1b
1.58b
3.83b
3.63a

a Measured
bD

value (ztrk et al., 1987).

values for oxygen were taken or calculated from Schumpe and Lhring (1990).

Table 2
Properties of the liquid mixtures (293.2 K)
Liquid mixture

Key Fig. 2b

L (kg m3 )

L (103 Pa s)

 (103 N m1 )

DL (O2 ) (109 m2 s1 )

Benzene/cyclohexane 6.7%
Benzene/cyclohexane 13.4%
Benzene/cyclohexane 31.5%
Benzene/cyclohexane 54%
Benzene/cyclohexane 78.5%
Benzene/cyclohexane 90%
Toluene/ethanol 5%
Toluene/ethanol 13.6%
Toluene/ethanol 28%
Toluene/ethanol 55%
Toluene/ethanol 73.5%
Toluene/ethanol 88.5%
Toluene/ethanol 94.3%
Toluene/ethanol 97.2%

A
B
C
D
E
F
G
H
I
J
K
L
M
N

865
854
834
814
797
787
863
859
852
838
823
807
800
796

0.634
0.628
0.631
0.672
0.772
0.858
0.578
0.587
0.616
0.731
0.961
1.013
1.103
1.135

27.6
26.9
26.2
25.4
24.9
24.9
27.6
27.3
25.5
25.0
24.2
23.3
22.7
22.2

3.61a
3.63a
3.62a
3.47a
3.16a
2.95a
5.0b
5.13b
4.57b
4.3b
4.0b
2.64a
2.49a
2.44a

aD

L values for oxygen were taken or calculated

b Measured value (ztrk et al., 1987).

from Schumpe and Lhring (1990).

0.2, 0.5, 1, 2 and 4 MPa, respectively. The kL a data in the other

pure and organic liquids were measured at ambient pressure by
ztrk et al. (1987). These authors presented also some kL a
data in tap water. All kL a data in the present work refer to the
dispersion volume.
The bubble column used by Jordan and Schumpe (2001) had
an inner diameter of 0.1 m and a height of 2.4 m. Three different
gas distributors were used: perforated plate, 19 ]1 103 m
(D1), single hole, 1 ]4.3 103 m (D2), and single hole,
1 ]1 103 m (D3). The clear liquid height was set at 1.3 m.
Jordan and Schumpe (2001) measured their kL a values with

a dynamic oxygen desorption technique. The oxygen desorption from the liquid was traced with an optical probe (MOPS,
Comte, GmbH, Hannover, Germany) based on uorescence extinction by oxygen. By dissolving the uorophore in the liquid,
rather than xing it to the tip of the glass ber, an instantaneous
sensor response was achieved.
ztrk et al. (1987) carried out their kL a experiments in a
jacketed glass bubble column of 0.095 m ID. The clear liquid
height was set at 0.85 m. A single tube of ]3 103 m ID
was used as the gas distributor (D4). Air, nitrogen, hydrogen
or helium was employed as the gas phase.

6268

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

The kL a values were measured by dynamic oxygen absorption or desorption methods. The oxygen fugacity in the liquids
was measured with a polarographic oxygen electrode (WTWEO 90) inserted horizontally at half the dispersion height. The
electrode response time was 3 s in water and less in most organic liquids. For absorption runs, oxygen was desorbed by
sparging nitrogen. After disengagement of all nitrogen bubbles,
a preadjusted air ow was fed by switching two magnet valves,
and the increase in oxygen fugacity was recorded. For desorption runs, oxygen-free inert gas (nitrogen, hydrogen or helium)
was sparged into air-saturated liquid.
In the present work only kL a data obtained in the homogeneous ow regime of bubble column operation were analyzed.
The upper boundary (the so-called transitional gas velocity)
of this ow regime was determined by the formulas of Reilly
et al. (1994). According to the bubble shape diagrams presented
by Clift et al. (1978) and Fan and Tsuchiya (1990) the bubbles
formed under all operating conditions were oblate ellipsoids.
4. Results and discussion
By means of a nonlinear regressional routine applied to the
experimental kL a data the following new correction factor was
derived:

0.15
0.94 G
fc = 0.124Eo
ref
G

0.94
G
0.15
g(L G )de2
= 0.124
,
(14)

1.2
3 for air at
where ref
G is the reference gas density (1.2 kg m
ambient conditions: 293.2 K and 0.1 MPa). Two hundred and
sixty-three experimental kL a values are tted with an average
relative error of 10.4%. It is worth noting that this correction
factor fc combines the individual corrections of both kL and
a due to the ellipsoidal shape of bubbles. The average relative
error without the gas density correction term is 14.9%. Comparisons between the F -test (1.155) value and the tabulated one
as well as the two-sample t-test (0.39) value and the tabulated
one have shown that there is no signicant difference between
the sets of predicted and experimental kL a values. The residual standard deviation with the gas density correction term is
3.91 103 , whereas this statistical parameter without the gas
density correction term is 6.62 103 .
The dimensionless gas density ratio is the additional correction term that has not been reported hitherto. Its introduction is
needed because the correlation of Wilkinson et al. (1994) was
tested only up to 1.5 MPa. In our work we correlate kL a data
at pressures of 2 and 4 MPa that could not be tted without
this term. Moreover, the correlations of Terasaka et al. (2004)
have been derived under ambient pressure. To the best of our
knowledge, these equations have not been validated under high
pressure. It is worth noting that such a dimensionless gas density correction factor has been also used by Krishna (2000)
for correcting his correlations for large bubble rise velocity
and dense-phase gas holdup. The different correction factors

reported in our previous papers (Nedeltchev et al., 2006a,b)

are due to both lower number of liquids considered and lower
operating pressures (up to 1 MPa).
It is interesting to note that the new correction factor
(Eq. (14)) does not involve the effect of the liquid viscosity.
However, the liquid viscosity takes part in the correlation for
the bubble size ds proposed by Wilkinson et al. (1994). The
bubble size ds affects the estimation of the bubble rise velocity uB and the bubble geometrical characteristics (l and h).
In addition, the liquid viscosity participates in both the bubble Reynolds number ReB and Morton number Mo and thus
Tadaki number Ta which is crucial for the determination of
both l and h values. The latter determines the bubble surface
and the rate of surface formation and thus the gasliquid contact time which is needed for the calculation of the kL value.
Therefore, the effect of the liquid viscosity is well incorporated
into our mathematical model.
Figs. 1ac show the parity plots of kL a values measured
in four organic liquids aerated with three different distributors
(D1, D2 and D3) at pressures up to 4 MPa.
It is clear that by means of the new correction factor
(Eq. (14)) the kL a values can be predicted reasonably well
(within 20%). It is worth noting that the kL a values obtained
in tap water aerated with gas spargers D1 and D4 also can
be predicted satisfactorily by means of Eq. (14). One of the
merits of our work is the successful prediction of kL a data at
very high pressure (up to 4 MPa). The presented theoretical
approach predicts satisfactorily the kL a values measured by
making use of four different gas distributors (D1D4). It is
known that the sparger has no substantial inuence on both
the gas holdup and mass transfer, provided that the diameter
of the sparger holes is not too small (Wilkinson et al., 1994).
Fig. 2a exhibits that the correction method predicts satisfactorily the kL a values measured in 15 different organic liquids
aerated with gas sparger D4 at ambient pressure. In comparison with our previous paper (Nedeltchev et al., 2006a) the
new liquids are acetone, cyclohexane, ligroin B, ethanol (99%)
and methanol. In the case of xylene and tetralin not only air
(nitrogen) but also helium (He) and hydrogen were used.
Ethanol (96%) was also sparged with helium.
The new correction method applies not only to pure organic
liquids but also to liquid mixtures. Fig. 2b shows that the predicted kL a values are in reasonable agreement with the experimental results. The keys for the different organic mixtures
used are given in Table 2. All 14 organic mixtures were aerated
with a gas sparger D4. As in some other parity plots, high
kL a values tend to be overpredicted. This indicates the onset
of coalescence at high gas velocity (ow regime transition).
In Fig. 2b are included also predicted and experimental kL a
values obtained in ethanol (96%), toluene and 1-butanol aerated
with helium (He) at different pressures by using different gas
spargers (D2 and D3). These data demonstrate again that the
developed theoretical approach along with the new correction
factor (Eq. (14)) is valuable and capable of predicting numerous experimental kL a data.
Fig. 3 shows the product fc (G /1.2)0.15 (see Eq. (14))
as a function of Eo. This product increases with Eo for 13

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

6269

Fig. 1. Parity plot of kL a values measured in (a) 1-butanol and decalin at pressures up to 4 MPa (gas distributors: D1 and D2), (b) toluene and decalin at
pressures up to 4 MPa (gas distributors: D1 and D2) and (c) ethanol (96%) at pressures up to 4 MPa (gas distributors: D1, D2 and D3).

organic liquids at ambient pressure. The trend holds also

for different gases (air, nitrogen, helium and hydrogen). The
fc (G /1.2)0.15 values fall in the range of 0.31.1 which

corresponds to Eo values from 2 to 10. It is worth noting that the

fc values for the liquids and operating conditions examined in
this work are always lower than unity (0.30.8). In the case of

6270

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

Fig. 2. Parity plot of kL a values measured in (a) 15 different organic liquids at ambient pressure (gas distributor: D4) and (b) 14 different liquid mixtures
(P = 0.1 MPa, sparger D4) as well as in ethanol, toluene and 1-butanol aerated with helium (He) at different pressures and gas spargers (D2, D3). The legend
keys for the liquid mixtures are given in Table 2.

the liquid mixtures specied in Table 2 the fc (G /1.2)0.15 vs.

Eo relationship follows the same trendline as the one shown in
Fig. 3. For the sake of brevity this gure is omitted.
Nedeltchev et al. (2006a) have shown that the values of the
correction factor (Eq. (12)) fall in line with the predictions of
the correlation of Wellek et al. (1966) for the bubble aspect
ratio E:
E 1 = 0.163Eo0.757 ,

(15a)

where
E=

l
.
h

(15b)

This similarity holds also for Eq. (14). Fig. 3 exhibits that the
fc (G /1.2)0.15 trendline at Eo 5 is very close to the predictions of Eq. (15a). Beyond this Eo value the values of the

product fc (G /1.2)0.15 become systematically higher than

(E 1) values.
It was found that the correction factors fc gradually decrease with the increase of gas density G (operating pressure). As the supercial gas velocity uG increases, a gradual
fc reduction is also observed. Both trends are illustrated for
four organic liquids in Fig. 4. The decreasing fc trends are
explainable in terms of Eq. (14). The new correction factor
fc is primarily dependent on the Etvs number Eo which
involves the bubble equivalent diameter de . As uG or G increases, the bubble size decreases (according to the correlation
of Wilkinson et al., 1994) leading to lower Eo and thus lower
fc (this trend is stronger than the effect of the gas density
ratio).
Larger ellipsoidal bubbles are characterized with higher fc
values since larger wakes or vortices are formed behind them

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

6271

Fig. 3. Product fc (G /1.2)0.15 as a function of the Etvs number Eo for 13 different organic liquids at ambient pressure (gas distributor: D4 unless
otherwise mentioned).

Fig. 4. Effects of both supercial gas velocity uG and gas density G (operating pressure P ) on the correction factor fc . Gas distributor: D1perforated
plate, 19 ]1 103 m, triangular pitch.

demonstrated that liquid drops (ethyl acetate drops in water and

water drops in isobutanol) with larger diameter (higher ReB )
are characterized with higher correction factors.
The presented correction of the penetration theory is applicable in the homogeneous ow regime of bubble column
operation. The upper boundary (the so-called transitional gas
velocity) of this ow regime was determined by the formulas
of Reilly et al. (1994).
Only in the case of two other organic liquids, carbon tetrachloride and ethylene glycol (reported by ztrk et al., 1987),
the presented correction method was not capable of tting the
experimental kL a data satisfactorily. Table 1 shows that the
correction method is applicable to organic liquids with densities L in the range of 714.1234 kg m3 and viscosities L in
the range of 0.327.2.94 103 Pa s. Carbon tetrachloride has
a higher liquid density (L = 1593 kg m3 ), whereas ethylene
glycol has a higher liquid viscosity (L = 19.94 103 Pa s).
5. Conclusions

which enhance the mass-transfer coefcient (see Lochiel and

Calderbank, 1964; Miller, 1974) and thus the correction factor
fc becomes closer to unity. In addition, with larger ellipsoidal
bubbles more surface disturbances occur which also enhance
the mass-transfer characteristics. Fan and Tsuchiya (1990)
argue that larger ellipsoidal bubbles (larger ReB ) will have
larger shedding frequency of vortex pairs from a bubble which
will increase the mass-transfer coefcient across the base of
an ellipsoidal bubble and thus the overall liquid-phase masstransfer coefcient. It is worth noting that Eq. (1) is valid at
very high bubble Reynolds numbers ReB (Calderbank, 1967).
It means that in the same liquid larger ellipsoidal bubbles will
have higher ReB and thus their correction factors fc will be
higher (closer to unity). The correction factor given in Eq. (3)
supports this explanation. Lochiel and Calderbank (1964) have

Experimental kL a values (263 points) in 18 pure organic

liquids, 14 adjusted mixtures and tap water were predicted with
reasonable accuracy by means of a new correction factor fc
that was applied along with Higbies (1935) penetration theory.
This correction term was correlated succesfully to the Etvs
number Eo and a dimensionless gas density ratio. The new
correction factor was found applicable at pressures up to 4 MPa.
This allows a clear-cut calculation of kL a in the homogeneous
regime of bubble column operation based on the contact time
dened as the ratio of bubble surface to the rate of surface
formation.
As both supercial gas velocity uG and gas density G (operating pressure) increase, the correction factor fc gradually
reduces.

6272

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

Notation
a
A
de
ds
DL
e
E
Eo
fB
fc
g
h
H
kL
kL a
l
Mo
NB
P
QG
Rsf
ReB
SB
tc
Ta
uB
uG
VB
Vtotal

specic interfacial area (referred to dispersion

volume), m1
cross-sectional area of the reactor, m2
bubble equivalent diameter, m
Sauter-mean bubble diameter, m
molecular diffusion coefcient, m2 s1
bubble eccentricity, Eq. (7b), dimensionless
bubble aspect ratio, Eq. (15b), dimensionless
Etvs number, Eq. (14), dimensionless
bubble formation frequency, s1
correction factor, dimensionless
gravitational acceleration, m s2
height of an ellipsoidal bubble, m
aerated liquid height, m
liquid-side mass-transfer coefcient, m s1
volumetric liquid-side mass-transfer coefcient
(referred to dispersion volume), s1
length (width) of an ellipsoidal bubble, m
Morton number (g4L /L 3 ), dimensionless
number of bubbles in the dispersion, dimensionless
operating pressure, Pa
gas ow rate, m3 s1
rate of surface formation, m2 s1
bubble Reynolds number, Eq. (4), dimensionless
bubble surface, m2
gasliquid contacttime, s
Tadaki number (=ReB Mo0.23 ), dimensionless
bubble rise velocity, m s1
supercial gas velocity, m s1
bubble volume, m3
dispersion volume, m3

Greek letters
L
G
ref
G
L


liquid viscosity, Pa s
gas density, kg m3
reference gas density (air at ambient conditions),
kg m3
liquid density, kg m3
surface tension, N m1

Acknowledgment
Dr. Stoyan Nedeltchev gratefully acknowledges the postdoctoral fellowships provided to him by the Deutscher
Akademischer Austauschdienst (DAAD) and the Alexander
von Humboldt Foundation.
References
Calderbank, P.H., 1967. Gas absorption from bubbles. The Chemical Engineer
45, CE209CE233.
Calderbank, P.H., Lochiel, A.C., 1964. Mass transfer coefcients, velocities
and shapes of carbon dioxide bubbles in free rise through distilled water.
Chemical Engineering Science 19, 485503.

Calderbank, P.H., Johnson, D.S.L., Loudon, J., 1970. Mechanics and mass
transfer of single bubbles in free rise through some Newtonian and nonNewtonian liquids. Chemical Engineering Science 25, 235256.
Clift, R., Grace, J.R., Weber, M., 1978. Bubbles, Drops and Particles, rst
ed. Academic Press, New York, NY.
Cockx, A., Roustan, M., Line, A., Hbrard, G., 1995. Modelling of mass
transfer coefcient kL in bubble columns. Transactions of the Institution
of Chemical Engineers 73, 627631.
Deckwer, W.-D., 1980. On the mechanism of heat transfer in bubble column
reactors. Chemical Engineering Science 35, 13411346.
Deindoerfer, F.H., Humphrey, A.E., 1961. Mass transfer from individual gas
bubbles. Industrial and Engineering Chemistry 53, 755759.
Fan, L.-S., Tsuchiya, K., 1990. Bubble Wake Dynamics in Liquids and
LiquidSolid Suspensions, rst ed. Butterworth-Heinemann Series in
Chemical Engineering. Stoneham, USA.
Garcia-Calvo, E., 1989. A uid dynamic model for airlift reactors. Chemical
Engineering Science 44, 321323.
Garcia-Calvo, E., 1992. Fluid dynamics of airlift reactors: two-phase friction
factors. A.I.Ch.E. Journal 38, 16621666.
Garcia-Ochoa, F., Gomez, E., 2004. Theoretical prediction of gasliquid mass
transfer coefcient, specic area and hold-up in sparged stirred tanks.
Chemical Engineering Science 59, 24892501.
Higbie, R., 1935. The rate of absorption of a pure gas into a still liquid during
short periods of exposure. Transactions of the A.I.Ch.E. 31, 365389.
Jordan, U., Schumpe, A., 2001. The gas density effect on mass transfer in
bubble columns with organic liquids. Chemical Engineering Science 56,
62676272.
Kawase, Y., Halard, B., Moo-Young, M., 1987. Theoretical prediction of
volumetric mass transfer coefcients in bubble columns for Newtonian
and non-Newtonian uids. Chemical Engineering Science 42, 16091617.
Krishna, R., 2000. A scale-up strategy for a commercial scale bubble column
slurry reactor for FischerTropsch Synthesis. Oil and Gas Science and
TechnologyRevue de lIFP 55, 359393.
Lochiel, A.C., Calderbank, P.H., 1964. Mass transfer in the continuous phase
around axisymmetric bodies of revolution. Chemical Engineering Science
19, 471484.
Lucas, D., Prasser, H.-M., Manera, A., 2005. Inuence of the lift force
on the stability of a bubble column. Chemical Engineering Science 60,
36093619.
Mendelson, H.D., 1967. The prediction of bubble terminal velocities from
wave theory. A.I.Ch.E. Journal 13, 250253.
Miller, D.N., 1974. Scale-up of agitated vessels gasliquid mass transfer.
A.I.Ch.E. Journal 20, 445453.
Nedeltchev, S., 2003. Correction of the penetration theory applied for
prediction of mass transfer coefcients in a high-pressure bubble column
operated with gasoline and toluene. Journal of Chemical Engineering of
Japan 36, 630633.
Nedeltchev, S., Jordan, U., Schumpe, A., 2006a. Correction of the penetration
theory applied to the prediction of kL a in a bubble column with organic
liquids. Chemical Engineering and Technology 29, 11131117.
Nedeltchev, S., Jordan, U., Schumpe, A., 2006b. A new correction factor
for theoretical prediction of mass transfer coefcients in bubble columns.
Journal of Chemical Engineering of Japan 39, 12371242.
ztrk, S.S., Schumpe, A., Deckwer, W.-D., 1987. Organic liquids in a bubble
column: holdups and mass transfer coefcients. A.I.Ch.E. Journal 33,
14731480.
Painmanakul, P., Loubire, K., Hbrard, G., Mietton-Peuchot, M., Roustan,
M., 2005. Effect of surfactants on liquid-side mass transfer coefcients.
Chemical Engineering Science 60, 64806491.
Reilly, I.G., Scott, D.S., De Bruijn, T.J.W., MacIntyre, D., 1994. The role of
gas phase momentum in determining gas holdup and hydrodynamic ow
regimes in bubble column operations. The Canadian Journal of Chemical
Engineering 72, 312.
Schumpe, A., Lhring, P., 1990. Oxygen diffusivities in organic liquids at
293.2 K. Journal of Chemical and Engineering Data 35, 2425.
Terasaka, K., Inoue, Y., Kakizaki, M., Niwa, M., 2004. Simultaneous
measurement of 3-dimensional shape and behavior of single bubble in
liquid using laser sensors. Journal of Chemical Engineering of Japan 37,
921926.

S. Nedeltchev et al. / Chemical Engineering Science 62 (2007) 6263 6273

Timson, W.J., Dunn, C.G., 1960. Mechanism of gas absorption from bubbles
under shear. Industrial and Engineering Chemistry 52, 799802.
Vasconcelos, J.M.T., Orvalho, S.P., Alves, S.S., 2002. Gasliquid mass transfer
to single bubbles: effect of surface contamination. A.I.Ch.E. Journal 48,
11451154.
Wellek, R.M., Agrawal, A.K., Skelland, A.H.P., 1966. Shape of liquid drops
moving in liquid media. A.I.Ch.E. Journal 12, 854862.

6273

Wilkinson, P.M., Haringa, H., Van Dierendonck, L.L, 1994. Mass transfer
and bubble size in a bubble column under pressure. Chemical Engineering
Science 49, 14171427.
Zieminski, S.A., Raymond, D.R., 1968. Experimental study of the behavior
of single bubbles. Chemical Engineering Science 23, 1728.