You are on page 1of 15

Chemical Engineering and Processing 40 (2001) 97 111

www.elsevier.com/locate/cep

A comparative study of gas hold-up, bubble size, interfacial area


and mass transfer coefficients in stirred gasliquid reactors and
bubble columns
Mounir Bouaifi 1, Gilles Hebrard *, Dominique Bastoul, Michel Roustan
Laboratoire dIngenierie des Procedes de LEn6ironnement (L.I.P.E., E.A. 833), Departement Genie des Procedes Industriels,
Institit National des Sciences Appliquees, 31077 Toulouse Cedex 4, France
Received 13 May 1998; received in revised form 31 May 1999; accepted 31 May 2000

Abstract
The overall gas hold-up mg, the volumetric mass transfer coefficient k1a, the liquid side mass transfer coefficient k1, the
volumetric interfacial area a, the bubble size dbs and bubble distribution have been characterised in two bubble columns and a
non-standard vessel equipped with various dual-impeller combinations. In the bubble columns, the effects of gas flow rate, sparger
type and column diameter were investigated. In the dual-impeller stirred reactor, the effects of rotational speed, gas flow rate,
impeller type and diameter were studied. The performances of the two reactors are compared here, and some relationships are
proposed and compared with existing correlations. 2001 Elsevier Science B.V. All rights reserved.
Resume
La retention gazeuse globale mg, le coefficient de transfert de matie`re volumetrique kla, le coefficient de transfert de matie`re cote
liquide kl, laire interfaciale volumetrique a, le diame`tre dbs et la distribution de diame`tre des bulles ont ete caracterises
respectivement dans deux colonnes a` bulles et une cuve non standard equipee de diverses configurations mettant en jeu deux
mobiles dagitation. Les effets du type de distributeur de gaz et du diame`tre de la colonnes ont ete etudies pour le premier type
de reacteur. Les effets de la vitesse dagitation, du debit gazeux et du type dagitateur ont ete mis en evidence pour le second type
de reacteur. Les performances des deux reacteurs sont comparees entre elles et confrontees a` des correlations existantes dans la
litterature. 2001 Elsevier Science B.V. All rights reserved.
Keywords: Bubble size; Gas hold-up; Interfacial area; Mass transfer coefficients; Gas liquid contactors

1. Introduction
In gas liquid reactors, mass transfer from the gas
phase to the liquid phase is the most important goal of
the process. The volumetric mass transfer coefficient is
a key parameter in the characterisation and design of
industrial stirred or non-stirred gas liquid reactors.
However, very few data are to be found dealing sepa* Corresponding author.
E-mail addresses: mounir@cre.chalmers.se (M. Bouaifi), hebrard@insatlse.insa-tlse.fr (G. Hebrard).
1
Present address: Department of Chemical Reaction Engineering,
Chalmers University of Technology, Kemivagen 4, SE-412 96
Goteborg, Sweden. Tel.: +49 31 772 3034; fax: + 49 31 772 3035.

rately with mass transfer coefficient and interfacial area


in the bubble columns or multi-impeller stirred gas liquid reactors [13] which are currently used in various
chemical and biological processes. Most investigations
performed on these systems are limited to the determination of the volumetric mass transfer coefficient k1a.
Unfortunately, this parameter is global and not sufficient to provide an understanding of the mass transfer
mechanisms. The separation of the parameters k1 and a
should be considered for better comprehension of the
gas liquid mass transfer mechanisms. It also allows us
to identify which parameter (k1 or a) controls the mass
transfer. Few results concerning the separation of k1
and a have yet been obtained [4,5]
The objectives of the present work are to:

0255-2701/01/$ - see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 2 5 5 - 2 7 0 1 ( 0 0 ) 0 0 1 2 9 - X

98

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

Fig. 1. The experimental apparatus associated to the agitated reactor.

1. study the interfacial area, volumetric mass transfer


coefficient, and separation of kl and a in two types
of gas liquid reactors in order to understand the
mass transfer mechanisms,
2. compare their values for two gas liquid contactors,
3. propose some correlations for each type of
contactor.

cally. Four types of impellers were used: Lightning


axial flow impellers A-315 (Npu = 0.75) and A-310
(Npu = 0.35), the four 45 Pitched Blade Turbine
Pumping Down (PBTD) (Npu = 1.30) and the Rushton
Disk Turbine (RDT) (Npu = 5.10 5.40). The different
configurations tested are given in Table 1. The gas
phase was air and the liquid phase tap water. Compressed air was introduced via a ring sparger 0.165 m in

2. Material and methods

Table 1
The various dual impeller systems studied

2.1. Material
2.1.1. Dual-impeller gas liquid reactor
The experiments were performed in a non-standard,
dished-bottom cylindrical tank of internal diameter
T = 0.43 m and ratio H1/T = 2, equipped with various double impeller combinations (Fig. 1). Four wide
longitudinal baffles (0.1T) were mounted symmetri-

Configuration

Lower impeller

Upper impeller

D/T

A
B
C
D
E
F

A-315
A-315
A-315
RDT
RDT
RDT

A-315
A-310
P.B.T.D.
A-315
A-310
PBTD

0.33
0.330.44
0.330.44
0.33
0.33
0.33

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

99

Fig. 2. The experimental apparatus associated to the bubble columns.

diameter, with 90 holes 1 mm in diameter spaced 10


mm apart on two circles. Gas flow rates ranged from
0.54 10 3 to 2.62 10 3 m3/s, gas velocities in a
hole ranged from 7 to 37 m s 1 and rotational speed
from 1.66 to 11.67 s 1.

2.1.2. Bubble columns


Experiments were carried out on two vertical bubble
columns made of PVC (Fig. 2).
The columns 1 and 2 were 0.15 and 0.20 m in
diameter respectively and 2 m in height (liquid
height). All experiments were performed without
throughput of liquid. Superficial gas velocities
ranged from 0.2510 2 to 4 10 2 m s 1. The liquid
was tap water and the gas used far all experiments was
air. Three different spargers were used, all of which
covered the whole cross-section of the column. The first
sparger was a perforated plate, the second a
sintered glass porous plate, and the third a perforated
flexible membrane. The diameter of the holes was
adapted to the gas throughput. Further details on these
spargers are given in Table 2. The columns were
equipped with pressure taps, located 0.2 m (column 1)
and 0.5 m (column 2) apart, connected to water
manometers.

2.2. Methods
2.2.1. Flow patterns
The qualitative flow patterns observed with the
stirred systems are reported in Fig. 3. The combination
of two A-315 impellers gave a strong axial component
and a big loop tended to form (Fig. 3a). When the
A-310 was placed as the upper impeller, the flow patterns generated by this impeller were similar to those
for the A-315. For axial dual-impeller configurations A
and B, each impeller pumped the liquid axially. Video
recordings show that a radial liquid velocity appeared
near the lower impeller. In this case again, a big loop
tended to form (Fig. 3a). The flow patterns generated
by the RDT were radial. After impact with the wall, the
stream lines curved and an axial liquid velocity component appeared. The ascending axial flow generated by
the upper impeller countered the one generated by the
lower impeller. For configurations D and E (Fig. 3c),
the reactor operated in one loop. There was no creation
of compartments. The PBTD used as upper impeller
generated a mixed flow pattern. Independently of the
lower impeller type, two loops tended to form. Each
impeller established its own zone of influence and the
reactor operated in two compartments (Fig. 3b and 3d).

100

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

The dual-impeller systems C and F can be regarded as


two independent loops, each created by one impeller,
with reciprocal mass exchange. Systems A, B, D and E
can be regarded as single-loop systems. For a given
rotational speed, the mixing time obtained with configurations A, B, D and E was less than that for systems C
and F. This behavior was due to the compartmental
flow pattern created only by the mixed configurations C
and F.
Concerning the bubble columns, four gas flow states
were identified (Fig. 4), depending on the superficial gas
velocity, the column diameter and the type of gas
sparger [6].

2.2.2. Gas hold-up


In the dual-impeller stirred reactor, gas hold-up was
determined by measuring the aerated liquid height relative to the gas-free liquid level. Two pressure taps were
located near the surface. In the bubble column, an
optical probe (Optoflow by Photonetics), located at a
height of 1.25 m above the gas sparger, connected to an
Optoelectronic type Y module and linked to a computer, was used to determine the local gas hold-up and
to establish the radial gas hold-up profiles. Overall gas
hold-up values were obtained by averaging the local
values over the cross sectional area of the column.
2.2.3. Bubble diameter
The bubble size was measured using a photographic
method. Pictures were taken in different planes with a
CCD high definition camera (Sony DXC 930 P)
associated to a video recorder and a high definition
screen. The bubbles at the wall were taken as representative of those inside the reactor. For a given hydrodynamic condition (N and Ug), 150 200 bubbles were
selected in various locations at different times and their
diameters estimated. For the dual impeller systems, the
measurements were done in three horizontal planes: the
plane of the upper impeller, that of the lower impeller,
and a plane situated mid-way between the two impellers.
The bubble size in the bubble columns was measured in
uniform flow at a height of 1.25 m above the gas
distributor and a distance of 0.1 m from the wall. The
bubbles were generally ellipsoids. The local bubble diameter was calculated using the following relationship
dbi =(a%2b%)1/3,

(1)

where a% and b% are the diameter and the width of the


ellipsoid respectively. For a given hydrodynamic condition and a given position, 150 200 bubbles were selected to estimate the Sauter mean diameter, dbs,
calculated by relation 2:
%nid 3bi
,

dbs =

(2)

%nid 2bi
where ni is the number of bubbles with an individual
diameter dbi.

2.2.4. Mass transfer coefficient


In the dual-impeller agitated reactor, the volumetric
mass transfer coefficient kla was measured using the
unsteady state method with an ISIS oxygen probe,
placed mid-way between the two impellers. The oxygen
concentration was reduced to zero by adding 150
mg l 1 of sodium sulphite (Na2SO3) and 2 mg l 1 of
cobalt ions. The response time of the ISIS probe is
about 30 s and can be neglected in kla determination
because in our experimental conditions the response
time never exceeded 2.3/kla. The sulphite concentration
was low enough not to cause any changes to the
coalescing properties of tap water. Perfect mixing was
assumed for the liquid phase. When the oxygen concentration increased, the mass transfer rate was given by
the following equation:
dCl
= kla(C*
Cl).
l
dt

(3)

In batch bubble columns, the volumetric mass transfer coefficient kla was measured using the steady state
method. During stationary conditions, the mass balance
on the gas phase allowed the volumetric mass transfer
coefficient, kla to be calculated. The oxygen concentration in the liquid phase was reduced to zero (adding
sulphite and cobalt) and that in the outlet gas phase
reached a constant value. Oxygen gas concentration was
measured with a Servomex OA.1100 oxygen analyser.
The estimation of the mass transfer coefficient was done
at the same temperature, 20C.

2.2.5. Axial liquid dispersion coefficient Ezl


Axial mixing in bubble columns was described by the
axial dispersion model. A transient state method was

Table 2
The characteristics of the different gas spargers used in bubble columns
Gas sparger

Sq. pitch arrangement (mm)

Orifice diameter (mm)

Holes per cm2

Thickness (m)

Perforated plate
Porous plate
Membrane

5
Apparent porosity: 0.3
5

2.5

(initial 0.5)

20103
20103
2.5103

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

101

Fig. 3. The qualitative flow patterns for the various studied agitated systems.

used. A tracer pulse (NaCl, 250 g l 1) was injected at


the bottom of the column just above the gas distributor
and the time variation of tracer concentration was
detected 1.8 m above the gas sparger, using a conductance probe (Tacussel XE 100) connected to a conductimeter and a computer. The axial liquid dispersion
coefficient Ezl was calculated by fitting the Ohki and
Inoue [7] and Kafarov et al. [8] model to the experimental curve generated by the impulse response.

2.2.6. Power consumption


In the dual-impeller gas liquid reactor the power
consumption Pg/Vl was determined by measuring the
torque using strain gauges mounted on the impeller
shaft. In order to take the power consumption related
to the gas pressure drop into account we considered the
total specific power consumption to be:
Pg/Vl tot =Pg/Vl +z1gUg.

(4)

In the case of bubble column reactors in which


mixing is induced pneumatically, the total specific
power consumption can be related to the total gas
pressure drop according to the well-known equation
Pg/Vl tot =Q DP/Vl = Q(zlgH +DPs)/Vl
=(zlgH+DPs)Ug/H.

(5)

the energy generated by the gas expansion into account.


However, when the stirred systems were compared with
similar works in the literature, only the mechanical
energy was considered because, in most of these works,
the energy due to the gas expansion is neglected.

3. Results and discussion

3.1. Gas hold-up


The gas hold-up, which is one of the most important
parameters characterising the hydrodynamics of reactors, depends mainly on the gas velocity, physical properties of the liquid and type of gas sparger (Hebrard et
al. [6]).

3.1.1. Dual-impeller gasliquid reactor


Gas hold-up in dual-impeller reactors has been studied by Abrardi et al. [9], Chiampo et al. [10] and Pinelli
et al. [11]. It has been shown that the gas hold-up mg
increases when the gas is efficiently dispersed. In this
case, when the gas dispersion is carried out by the
impellers, the liquid flow patterns have a predominant
effect. The gas hold-up increases with the rotational

With a perforated plate or a porous plate, the specific


sparger pressure drop DPs can be neglected. The total
gas pressure drop is only affected by the liquid height.
With a membrane sparger the total gas pressure drop
corresponds to the specific sparger pressure drop, which
cannot be neglected, plus the liquid static pressure. The
specific sparger pressure drop increases with the superficial gas velocity. In order to avoid the gas sparger
effect, a specific power consumption (Pg/Vl)% related to
the dispersion into the gas liquid mixing, was determined for the three sparger using Eq. (6) expressed as
(Pg/Vl)%=zlgUg.

(6)

It should be noted that the comparison between the


stirred systems and bubble columns was made taking

Fig. 4. The gas flow regimes in bubble columns.

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

102

Fig. 5. The linear regression mg = C1 (Pg/Vl)0.24 U 0.06


obtained in the
g
case of dual-impeller configurations.

Fig. 6. The gas hold-up versus the total specific power consumption
for the columns 1 and 2 with the three types of gas spargers.

speed. The increase in gas hold-up with Pg (or N) can


be explained by the increase in circulation of the liquid
and the decrease in bubble size.
It was observed that, for configurations A to F with
a ratio D/T = 0.33, the gas hold-up was practically
unaffected by the upper impeller but was strongly affected by the lower impeller type.
The axial configurations A, B and C gave similar gas
hold-up results. The same behavior was observed with
the mixed configurations D, E and F. However, for a
given value of Pg/Vl and Ug, the mg corresponding to the
mixed configurations was 16% larger than with the
axial configurations.
The experimental data can be correlated in terms of
superficial gas velocity and specific power consumption for operating conditions situated beyond the
complete dispersion point (N E NCD, Fig. 5). It should
be noted that the point NCD corresponds to the situation of homogeneous gas dispersion in which bubbles
reach the bottom of the reactor (Bakker[12]). For axial
configurations A, B and C, the A-315 impeller had
sufficient pumping capacities for the bubbles to
arrive at the bottom of the reactor and the gas was well
dispersed. For mixed configurations D, E and F,
complete dispersion was achieved when bubbles arrived

The constants C1 were 22.4 for the axial configurations and 24.8 for the mixed configurations.

3.1.2. Bubble columns


For all the gas spargers used, mg increased when gas
velocity increased. The dependence of the gas hold-up
on the superficial gas velocity Ug can be written as:
mg = C2U xg ,

(Ug(cm s 1)).

(8)

[1]

In bubbly flow, the values of x ranged from 0.83 to


1.19, in good agreement with values found in the
literature [13]. Experimental values of C2 and x obtained in this work are presented in Table 3.
In bubbly flow, for a given superficial gas velocity
condition, with porous and membrane spargers, the gas
hold-up data were greater than those obtained with the
perforated plate. This can be attributed to the bubble
size (as shown later) which strongly depends on the
sparger used. The smaller the bubbles, the greater the
gas hold-up values.
It appears, however, in Fig. 6, that the total power
consumption required to reach a given value of gas
hold-up with a Membrane sparger is higher than that
necessary with a porous or perforated plate.
Moreover it can be seen in Fig. 6 that, for all the gas
spargers, the variation of gas hold-up versus total
power consumption can be written as

mg = C1(Pg/Vl)0.24U 0.65
g

mg = C3(Pg/Vl)0.85
tot .

at the bottom of the reactor after impact with the wall

(N ENCD,

mg in %,

Ug E0.018 m s 1,

(9)

Pg/Vl in W m ,
R 2 =0.91),

Experimental values of C3 obtained in this work are


presented in Table 4.

(7)

Table 3
The values of C2 and x for different types of gas spargers used in the bubble columns
Installation

Column 1
Column 2

Porous

Membrane

Perforated plate

C2

C2

C2

3.62
3.43

0.91
1.03

4.25
3.12

0.99
1.05

3.66
2.2

0.83
1.06

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111


Table 4
The values of C3 for different types of gas spargers used in the bubble
columns
Installation

Column 1
Column 2

Porous

Membrane

Perforated plate

C3

C3

C3

0.052
0.087

0.049
0.042

0.052
0.070

3.1.3. Comparison between the two types of reactors


In order to compare the gas hold-up provided by the
two types of reactor, the experimental data obtained at
Ug = 0.011 m s 1 with the bubble column and the
dual-impeller gas liquid reactor are reported (Fig. 7)
against total the specific power consumption using Eq.
(5) and Eq. (4) respectively. With bubble columns, the
power consumption was determined by considering the
total gas pressure drop in Eq. (5).
It can be seen in Fig. 7 that the power consumption
required to reach a given gas hold-up value with the
two types of reactors are about the same.
However, whatever the type of reactor used, it is
difficult to propose a general correlation to predict gas
hold-up, which depends on the stirrer configuration
(stirred gas liquid reactor) and gas sparger type (bubble column).
3.2. Bubble size and distribution
Bubble size is a very important parameter in gas liquid reactor design. The bubble size study can be very
useful for a better understanding of gas dispersion
mechanisms. It also has a direct influence on gas hold-

103

up and interfacial area, so it is a good criterion for


evaluating the efficiency of a gas liquid contactor.

3.2.1. Dual-impeller gasliquid reactor


Bubble size and distribution have been investigated
by Calderbank [14], Takahashi et al. [15], who found
that the bubble size strongly depends on the measurement position. Bubble sizes near the vessel wall at the
same level as the impeller or in the upper levels were
found to be more than four times those in the impeller
region. Greaves and Barigou [16] studied the local
bubble size in 22 positions using a submersible capillary
probe. It was shown that the bubble size distribution
varied widely from one region of the reactor to another,
deviating considerably from a normal distribution. At
the impeller region with low gas flow rate, the increase
of impeller speed caused a sharp drop in the mean
bubble size while no effect was observed for high gas
flow rates, owing to the increased role of bubble coalescence. Parthsarathy and Ahmed [17] carried out experiments with bubbles of known initial size. The effects of
the bubble size and power input on the gas hold-up
were investigated. They also found that the bubble size
could be determined by the energy dissipation rate in
the impeller zone.
For the present study, bubble size measurements
were made only for configurations B and C (D/T =
0.44) and A (D/T = 0.33).
The results obtained show that the bubble size distribution depends on the operating conditions and the
configuration type. In flooding regime (Fig. 8), the
bubble size distribution is very heterogeneous, reflecting
the non-homogeneous gas dispersion in the liquid
phase. It was also observed that the bubble distribution

Fig. 7. The gas hold-up versus total specific power for the dual impeller configurations and the bubble columns 1 and 2 at Ug =0.011 m s 1.

104

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

Fig. 8. The bubble size distribution for the dual impeller configuration A at Ug = 3.58 10 3 m s 1 and N =4.00 s 1 (NB NCD).

depended on the location where the measurements were


performed. Bubble size distribution in a given location
depended strongly on the associated hydrodynamic
regime. Beyond the complete dispersion point, the
lower impeller had sufficient pumping capacity and was
able to disperse the gas phase in the entire reactor. As
a consequence, the bubble distribution was more uniform (Fig. 9), between 50% and 60% of the bubbles
having a diameter between 1 and 3 mm. In this case,
the bubble size was controlled by the flow patterns
generated by the impellers. The mean Sauter diameter
obtained in the lower part of the vessel was greater than
that observed in the upper part (Fig. 9). The Sauter
diameter dbs estimated in the upper part of the reactor
was generally smaller than that estimated between the
impellers or in the lower part of the reactor.
The Sauter mean diameter is reported (Fig. 10)
against specific power consumption for different gas
flow rates (3.58 10 3 0Ug (m s 1) 010.82 10 3).
The results show that dbs is affected by the power
consumption but not by the gas flow rate. The Sauter
diameter decreases when the power consumption increases. This is the consequence of the simultaneous

Fig. 10. The Sauter diameter versus the specific power consumption
for the dual impeller configurations.

increase in shear characteristics and turbulence. For the


various configurations and gas flow rates studied, the
following relationship is proposed for predicting the
Sauter diameter:
dbs = 10.1 10 3(Pg/Vl) 0.20
(dbs in m,

Pg/Vl in W m3,

R 2 = 0.90).

(10)

The mean measured values obtained for Sauter diameter in this study can compared with two correlations from the literature:
Calderbank [14]:
dbs = 4.15

| 3/5
m 1/2 + 0.0009,
(Pg/Vl)2/5 g

3/5
l

(11)

Lee and Meyrick [18]:


dbs = 4.25

Fig. 9. The bubble size distribution for the dual impeller configuration A at Ug = 7.18 10 3 m s 1 and N =6.66 s 1 (N\ NCD).

| 3/5
m 1/2.
(Pg/Vl)2/5 g

3/5
l

(12)

It should be noted that correlations 11 and 12 were


obtained with single-impeller systems equipped with
Rushton disk turbines. For these systems, the specific
power consumption exponent obtained is about 0.40
while the one obtained in this work was about 0.20.
However, the literature investigators (4, 14) introduce
the gas hold-up into their correlations. As a consequence, the real value of the specific power exponent
would be less than 0.40. The comparison between
the size of bubbles measured and that predicted from
the literature shows rather good agreement. An average
error of about 26% is found with Calderbanks [14]
correlation (Fig. 11). The average error is about 15%
when our results are compared with those predicted
from Lee and Meyricks [18] correlation (Fig. 11).

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

105

Table 5
The experimental relationships relating the Sauter diameter to the
total specific power consumption for the different types of gas
spargers

Fig. 11. The comparison between the measured Sauter diameter and
the literature models in the case of dual impeller configurations.

3.2.2. Bubble columns


The variation of the bubble Sauter diameter with the
total specific power consumption is shown in Fig. 12
for each sparger. The variation of dbs with Pg/Vl tot,
depends on the gas sparger. With the perforated plate,
dbs decreases with increasing Pg/Vl tot. On the contrary,
dbs increases slightly with increasing Pg/Vl tot when the
porous plate or the flexible membrane are used.
The experimental correlations relating the bubble
diameter to the specific power consumption are reported in Table 5 for different types of gas spargers.
According to Hinzes [19] equation, bubble diameters
obtained in the bubble column equipped with the perforated plate decrease when the specific power dissipated
increases.

Fig. 12. The Sauter diameter versus the total specific power for the
bubble column with diameter DC = 0.2 m with different gas spargers.

Ranging bubble
diameter

Correlation

Perforated plate

4.57.5

Porous plate

3.75

Membrane

3.74.2

Dual-impeller
reactor
Hinze [19]

35.9

dbs =16103
(Pg/Vl)0.23
tot
dbs =2.15103
(Pg/Vl)0.16
tot
dbs =2.79.1 03
(Pg/Vl)0.08
tot
dbs =10.1103
(Pg/Vl)0.20
tot
dbs =3 8103
(Pg/Vl)0.40
tot

Fig. 13 shows the effect of the gas sparger on the


bubble distribution. The bubble size distribution obtained with the perforated plate is more widely spread
than that obtained with the other spargers. For the
porous plate, the average bubble diameter dbs ranges
from 2.5 to 4.5 mm. For the flexible membrane, the
bubble population is very homogeneous; about 80% of
the bubbles have diameters in the 3.5 to 4.5 mm range.

3.2.3. Comparison between the two types of reactors


In the dual-impeller gas liquid reactor, dbs seems to
be affected only by the power consumption and not by
the gas flow rate. The Sauter mean bubble diameter
decreases when the power consumption increases. In
the bubble column, the variation of dbs depends on
both the gas sparger and the total power consumption.
As in the dual-impeller reactor, the bubble diameter
provided by the perforated plate distributor decreases
when specific power increases. With this type of gas

Fig. 13. The bubble size distribution for all gas spargers for the
bubble column with diameter DC =0.2 m at Ug =0.006 m s 1.

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

106

3.3.1. Dual-impeller gasliquid reactor


In Fig. 14, the change of interfacial area with the
power consumption is reported for Ug = 0.0072 m s 1.
It appears that the interfacial area strongly depends on
the power consumption and seems independent of the
impeller configuration and its diameter. For example,
for a specific power Pg/Vl = 250 W m 3 and Ug =
0.0072 m s 1, the mean interfacial area is 58 m2 m 3.

Fig. 14. The volumetric interfacial area versus the specific power
consumption for the configurations A, B and C at Ug = 0.0072 m
s 1.

sparger, the bubble size only depends on the intensity


of the turbulence or liquid circulation created. On the
contrary, dbs increases with increasing Pg/Vl tot when the
porous plate or the flexible membranes are used. With
this type of gas sparger, the bubble size depends on the
sparger type, i.e. on the power consumption necessary
to provide bubbles.

3.3. Interfacial area


The interfacial area is one of the most important
parameters for gas liquid reactor design. Generally,
the interfacial area depends on the units geometrical
size, the operating parameters and the physical and
chemical properties of the liquid [13]. Knowing the
bubble size and gas hold-up, it is possible to evaluate
the specific interfacial area:
a=

6mg
.
dbs(1mg)

(13)

3.3.2. Bubble columns


The interfacial area a depends on the gas sparging
conditions, gas flow rate, gas hold-up and bubble size.
The variation of the interfacial area with the total
specific power, for each sparger is shown in Fig. 15.
The variation of a with Pg/Vl tot, depends on the gas
sparger. The specific interfacial area increases with increasing Pg/Vl tot for each operating condition.
3.3.3. Comparison between the two types of reactors
In the dual-impeller reactor as in the bubble column
(Fig. 15), the interfacial area is linked to the gas
hold-up and the bubble diameter, which depend (as
shown previously) on specific power consumption or
type of gas sparger.
In both types of reactors, the interfacial area ranges
between 20 and 133 m2 m 3 for a specific power varying between 50 and 1000 W m 3. For any total power
consumption used, it can be seen in Fig. 15 that the
interfacial areas created by the bubble columns are
about 30% higher than those created by the stirred
reactor equipped dual impellers. If we compare our
results to those obtained by Nagel et al. [20] and Wild
et al. [21] we note that, with a power dissipation
ranging between 100 and 1000 W m 3, the interfacial
areas provided by the bubble columns are quite similar.
However, the interfacial areas provided by the dual-impeller reactor for configurations A, B and C are lower
than those obtained with the simple stirred vessels
studied by Nagel et al. [20] and Wild et al. [21].

Fig. 15. The interfacial area versus the total specific power for the three spargers in bubble columns and for the configurations A,B and C at
Ug = 0.0072 m s 1.

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

107

Fig. 16. The linear regression k1a= 0 0218(Pg/Vl)0.50 U 0.60


for the configurations with a ratio D/T =0.33.
g

Considering the previous gas hold-up and bubble


diameter correlation, it is possible to calculate the
interfacial area generated by the two types of reactors.Eq. (7), Fig. 10 and Fig. 13 give for the dual-impeller gas liquid reactor:
a= 5.94C1(Pg/Vl)0.44U 0.65
g ,

(14)

Eqs. (9) and (13) and Table 5 give, for bubble column
2 equipped with the perforated plate:
a=0.26(Pg/Vl)1.08
tot .

(15)

The values of the constant in Eq. (14) strongly depend on the system used. The constant and the specific
power exponent in Eq. (15) are related to the type of
gas sparger used. The scatter of the data shows that is
difficult to predict the specific interfacial area without
considering the gas sparger effect and the configuration
of the impellers.

3.4. The 6olumetric mass transfer coefficient


In a gas liquid reactor, the mass transfer performance can be characterised by measuring the volumetric mass transfer coefficient kla, which is the product of
the liquid mass transfer kl and the interfacial area a.

mainly because of the difference in the system scale and


the measuring methods used.
The volumetric mass transfer coefficient values were
compared with predictions from the literature. The
correlations proposed by Roustan [1], Vant Riet [22],
Lopes et al. [23] and Moucha et al. [3] present good
agreement for the same power input and gas superficial
velocity. It should, however, be noted that Moucha et
al. [3] used the dynamic pressure-step method. Their kla
measurements were carried out in the bottom level and
the upper levels. Our results agree well with their kla
values obtained in the upper levels. The maximum
relative deviation obtained is about 13%. The kla values
obtained in this study are larger than those proposed in
the literature with single impeller systems. The
results show that the volumetric mass transfer coefficient is greater in the dual-impeller reactor compared
with the single-impeller ones. This confirms the results
of Roustan [1] and Bouaifi et al. [24], who observed
that the gas hold-up obtained in two and three- impeller systems was about 30% higher than that obtained
with a single-impeller for a given gas flow rate and
power input.

3.4.1. Dual-impeller gas liquid reactor


It was observed, for a given gas flow rate and rotational speed above the complete dispersion point (N E
NCD), that all the configurations studied with a ratio
D/T = 0.33 give similar kla values for the same power
input. For the configurations with a ratio D/T = 0.33,
the experimental data are well represented by the following correlation in terms of superficial gas velocity
and power consumption (Fig. 16):
kla = 0.0218 (Pg/Vl)0.50U 0.60
g
(N ENCD,

Ug 0 0.018 m s 1,

R 2 =0.91).

(16)

One of the serious problems of the experimental


studies is the discrepancy between the correlation

Fig. 17. The volumetric mass transfer coefficient k1a versus the
superficial gas velocity for bubble columns with different gas spargers.

108

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

Fig. 18. k1a values provided by the two types of reactor versus total power consumption.

3.4.2. bubble columns


In Fig. 17, the volumetric mass transfer coefficient
kla is reported versus the superficial gas velocities. The
kla increases with increasing Ug for all operating
conditions.
In column 2 the kla seems to be affected only by the
interfacial area a. The kla provided from the membrane
and the porous spargers are higher than those obtained
from the perforated plate, which creates the smallest
interfacial area.
In column 1 the volumetric mass transfer coefficient
kla is not only related to the interfacial area. In fact the
kla obtained in column 1 for the three spargers are
similar while their interfacial areas are different. This
means that, in column 1, the liquid-side mass transfer
coefficient kl varies with the operating conditions (gas
velocity and sparger type).
So the separation of kla into kl and a can be very
interesting for a better understanding of mass transfer
coefficient evolution.
In the dual-impeller reactor, the kla increases with
gas flow rate and rotational speed, but it is independent
of impeller configuration for the same power consumption. In the bubble column the kla variations may be
due to the interfacial area a and liquid-side mass transfer coefficient kl.
3.4.3. Comparison between the two types of reactors
Fig. 18 compares the volumetric mass transfer coefficient obtained in the two types of reactors versus the
total specific power consumption.
With both reactors and for all operating conditions,
kla increases with increasing total power consumption.
But throughout the range of specific power variation,
the kla values obtained with bubble columns are larger
than those provided by the stirred gas liquid reactor.
This difference can be explained by larger values of
interfacial area in bubble columns as shown previously.
It is interesting to underline that kla values obtained
in the stirred gas liquid reactor seem to be influenced

only by the superficial gas velocity and the specific


power consumption whereas with bubble column kla
values depend on total power consumption and sparger
type.

3.5. The liquid-side mass transfer coefficient kl


The kl values were estimated by the following
relationship:
dbs(1mg)
.
6mg

kl = kla

(17)

The mass transfer rate depended not only on the gas


hold-up and bubble size but also on the value of kl. The
liquid-side mass transfer coefficient kl depended on the
diffusivity coefficient and turbulence created in the
liquid phase. It varied with bubble size and flow pattern
around the bubbles. Small bubbles with a diameter of
up to about 1 mm in water behaved like rigid spheres in
contact with the liquid film adhering to the phase
interface. For larger bubbles, the kl values were considerably higher than those recorded for smaller bubbles.

3.5.1. Dual-impeller gasliquid reactor


The liquid side mass transfer coefficient kl depended
on the turbulence created in the liquid phase. Most of
the proposed relationships for predicting kl are obtained using isotropic turbulence models. Kawase and
Moo-young [5] and Lamont and Scott [25] propose the
following equations:Kawase and Moo-Young [5]:
kl = 0.301(Pmw)1/4Sc 1/2,

(18)

Lamont and Scott [25]:


kl = 0.40(Pmw)1/4Sc 1/2,

(19)

Bakker [12] showed that the turbulence in an stirred


reactor was far from isotropic. Application of the penetration theory of mass transfer (Higbie [26]), leads to
the equation:

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

kl = C4

'

DO2 1/6
P ,
dbs g

Fig. 19. The Liquid side mass transfer coefficient versus the specific power.

(20)

where DO2 corresponds to the diffusivity coefficient


of oxygen in water.
The results obtained in the dual-impeller reactor
show that kl only depends on the power consumption.
For a given power input, the gas flow rate and the
configuration type have no effect on the value of kl
(Fig. 19). That means that, if the gas flow rate is
increased for a given rotational speed, the power consumption decreases and, as a consequence, kl decreases.
However, experimental values of kla increase. This is
principally caused by the increase of the interfacial area
a. The experimental data lead to the following relation:
kl = 1.12(Pg/V1)0.22.

'

109

(21)

The specific power consumption exponent in Eq. (21)


is of the same order as the values proposed in theoretical relations 18, 19 and 20.
It would be interesting to make measurements with
other liquids in order to define the effects of the liquid
properties on the parameters kl and the limits of validity of the literature models. McFarlane [27] reported
that, in some alcoholic solutions, the mass transfer was
limited by a low value of the parameter kl although the
interfacial area and the gas hold-up could reach high
values.

3.5.2. Bubble columns


Fig. 20 shows the liquid-side mass transfer coefficients estimated under homogeneous flow using Eq.
(17), versus the specific power (Pg/Vl)% calculated using
relation 6. In order to avoid the gas sparger effect, the
specific power considered (Pg/Vl)% is that related to the
dispersion into the gas liquid mixture. The results are in
good agreement with the liquid side mass transfer coefficient obtained using Higbies [26] penetration theory, which assumes that the effective contact time is
given in bubble columns by the ratio dbs/G:

kl = 1.13

DO2 1/2
G ,
dbs

(22)

assuming that the slip velocity G is a constant in the


range of superficial gas velocities used.
The kl values, ranging between 310 4 and 6
10 4 m s 1, are in good agreement with those found in
the literature [27], [22]. Calderbank and Moo-Young
[28] found that the kl value was 4.2 10 4 m s 1 for
an air-water system at 20C. The results obtained show
that kl can be considered to be independent of (Pg/Vl)%.
It depends, however, on the type of gas sparger and
column diameter.
Fig. 21 shows that kl strongly depends on the axial
liquid dispersion coefficient. The liquid side mass transfer coefficient increases when the axial liquid dispersion
coefficient increases. The increase in the axial liquid
dispersion coefficient corresponds to an improvement in
the liquid circulation velocity. In this situation, the slip
velocity increases and the effective contact time decreases. In consequence, kl increases.

Fig. 20. The liquid side mass transfer coefficient versus the specific
power into bubble columns.

110

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

Fig. 21. The liquid side mass transfer coefficient versus axial liquid
dispersion coefficient in bubble column (homogeneous regime).

3.5.3. Comparison between the two types of reactors


The physical phenomena which control the liquid
side mass transfer coefficient in bubble columns and
dual-impeller reactors are different. In the dual-impeller
reactor, the kl values depend on specific power consumption, i.e. on liquid velocity and local turbulence.
As far as the bubble columns are concerned, kl depends
on the effective contact time, which is controlled by the
axial liquid dispersion coefficient and slip velocity.

4. Conclusions
The objectives of this work were to characterise the
mass transfer parameters in stirred multi- impeller gas
liquid reactors and bubble columns, and to propose
and compare some relationships related to the two
types of contactors.
In the dual-impeller reactor as in the bubble column,
the gas hold-up depends on the total specific power
consumption. As in the dual-impeller reactor, the bubble diameter provided by a bubble column equipped
with a perforated plate decreases when the specific
power increases. Conversely, dbs increases with increasing Pg/Vl tot when a porous plate or flexible membrane
is used in the bubble column. There are no important
differences between bubble diameters provided by the
two reactors.
In the dual-impeller reactor as in the bubble column,
the interfacial area is linked to gas hold-up and bubble
diameter, which in turn depend on the specific power
consumption (dual impeller combination) and gas
sparger type. In the two types of reactors, the interfacial area ranges between 20 and 133 m2 m 3, for specific powers varying from 50 to 1000 W m 3. For the
same total power consumption, the area created by the
bubble columns is about 30% higher than that created
by the stirred axial dual impeller systems A, B and C.
The volumetric mass transfer coefficient kla obtained
in the stirred gas liquid reactor seems to be influenced

only by the superficial gas velocity and the specific


power consumption whereas, in bubble columns, kla
depends on total power consumption and sparger type.
With both reactors and for all operating conditions,
kla values increase with increasing total power consumption. For the same total power consumption, the
kla values obtained with bubble columns are higher
than those provided by the stirred gas liquid reactor.
This difference is explained by the higher values of
interfacial area obtained in bubble columns.
In the dual-impeller reactor, the liquid-side mass
transfer coefficient kl is strongly related to the power
input and not affected by the impeller type or gas flow
rate. In the bubble columns, kl can be considered as
independent of the power dissipated. It depends on the
axial liquid dispersion coefficient, which is influenced
by the type of gas sparger used.

Appendix A. Nomenclature

a
C1
C*1
dbs
Dc
D
DO2
Ezl
f
G
Hl
k1a
kl
N
NCD
DP
Pg
Pm
Qg
Sc
T
Ug
Greek
mg
w
|
z

interfacial area (m1)


dissolved oxygen concentration (kg m3)
concentration in equilibrium with oxygen
partial pressure (kg m3)
sauter bubble diameter (m)
bubble column diameter (m)
impeller diameter (m)
diffusion coefficient (m2 s1)
axial liquid dispersion coefficient (m2 s1)
fraction of bubble diameter distribution (%)
slip velocity (m s1)
tank liquid height (m s1)
volumetric mass transfer coefficient (s1)
liquid-side mass transfer coefficient (m s1)
rotational speed (s1)
rotational speed at complete dispersion point
(s1)
gas pressure drop (Pa s)
power consumption in aerated liquid (W)
power consumption per unit mass of liquid
(W kg1)
gas flow rate (m3 s1)
w
schmidt number
tank diameter (m) DO2

 

superficial gas velocity (m s1)


symbols
overall gas hold-up (%)
kinematic viscosity (m2 s1)
surface tension (kg s2)
liquid density (kg m3)

Subscripts
g
gas

M. Bouaifi et al. / Chemical Engineering and Processing 40 (2001) 97111

l
C.D.

liquid
complete dispersion

[14]

[15]

References
[1] M. Roustan, Contribution a` lEtude des Phenome`nes dAgitation et de Transfert de Matie`re dans les Reacteurs Gaz-Liquide,
Doctoral thesis, INSA Toulouse, France, 1978.
[2] J.M. Charles, Contribution a` lEtude de lExtrapolation des
Reacteurs Agites Mecaniquement, Doctoral thesis, INSA Toulouse, France, 1978.
[3] T. Moucha, V. Linek, J. Sinkule, Measurement of kla in multiple-impeller vessels with significant axial dispersion in both
phases, Trans. I. ChemE. 73 (1995) 286.
[4] W. -D. Deckwer, I. Adler, A. Zaidi, A comprehensive study on
CO2 inter-phase mass transfer in vertical co-current and counter-current gas flow, Can. J. Chem. Eng. 56 (1978) 43 55.
[5] Y Kawase, M. Moo-Young, Mathematical models for design of
bioreactors: application of kolmogoroffs theory of isotropic
turbulence, The Chem. Eng. J. 43 (1990) B19.
[6] G Hebrard, D Bastoul, M Roustan, Influence of gas sparger on
the hydrodynamic behavior of bubble columns, Trans. I.
ChemE. 74 (1996) 406.
[7] Y. Ohki, H. Inoue, Longitudinal mixing of the liquid phase in
bubble columns, Chem. Eng. Sci. 25 (1970) 1.
[8] V. Kafarov, V. Vygon, V.A. Rudakov, G.A. Mikheeva, Method
for determination of the coefficient of longitudinal mixing for the
continuous phase under the conditions of a nonflow-through
system in high-rate extractors, Teor. Osn. Khimich.
Tekhnologii 7 (4) (1973) 550556.
[9] V. Abrardi, G. Rovero, S. Sicardi, G. Baldi, R. Conti, Hydrodynamics of gas liquid reactor stirred with multi-impeller system,
Trans. Inst. Chem. Eng. 68 (1990) 516.
[10] F. Chiampo, R. Guglielmetti, R. Manna, R. Conti, Gas liquid
mixing in a multi-impeller stirred vessel, in: Seventh European
Conference on Mixing, Brugge, Belgium, 1991 pp. 333341
[11] D. Pinelli, M. Nocentini, F. Magelli, Hold-up in low viscosity
gas liquid systems stirred with multiple impellers. Comparison
of different agitators types and sets, in: Proceeding of the
Eighth International Conference on Mixing, Cambridge, U.K.,
1994, pp. 81
[12] A. Bakker, Hydrodynamics of stirred gasliquid dispersions,
Ph.D. Thesis, Delft University, Netherlands, 1991.
[13] Y.T. Shah, W.D. Deckwer, B.G Kelkar, S.P. Godbole, Design

[16]

[17]

[18]
[19]
[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

111

parameters estimations for bubble column reactors, A.I.Ch.E.J.


28 (1982) 353.
P.H. Calderbank, The interfacial area in gas liquid contacting
with mechanical agitation, Trans. Inst. Chem. Eng. 36 (1958)
443.
K. Takahashi, A.W. Nienow, Bubble Sizes, Gas hold-up and
coalescence rates in aerated vessels agitated by a rushton turbine:
spacial variations, in: Proceeding of the Sixth European Conference on Mixing, Pavia, Italy, 1988, p. 285
M. Greaves, M. Barigou, The internal structure of gasliquid
dispersions in a stirred reactor, in: Proceeding of the Sixth
European Conference on Mixing, Pavia, Italy, BHRA, Springer,
Berlin, 1988, pp. 313 320
R. Parthsarathy, N. Ahmed, Gas Hold-up in Stirred Vessels:
Bubble size and power input effects, in: Proceeding of the Sixth
International Conference on Mixing, Pavia, Italy, 1988, p. 295
J.C. Lee, D.L. Meyrick, Gas liquid interfacial area in salt
solutions in an agitated tank, Trans. Inst. Eng. 48 (1970) T37.
J.O. Hinze, Fundamentals of the hydrodynamic mechanisms of
splitting dispersion processes, A.I.Ch.E.J 1 (1955) 289.
O. Nagel, B. Hegner, H. Kurten, Criteria for the selection and
design of gas liquid reactors, Int.Chem.Eng. 21 (1981) 161172
(translated from German).
G. Wild, N. Midoux, J.C. Charpentier, Energetique des reacteurs
gaz-liquide, in: P Le Goff (coordinateur) Energetique Industrielle, Technique et Documentation,Lavoisier, Paris, Vol.3,
1982, p. 485 515
K. Vant Riet, Review of measuring methods and results in
nonviscous gas liquid mass transfer in stirred vessels, Ind.&
Eng.Chem.Proc.Des.& Dev. 18 (1979) 357.
M.M. Lopes De Figueiredo, P.H. Calderbank, The scale-up of
aerated mixing vessels for specified oxygen dissolution rates,
Chem. Eng. Sci. 34 (1979) 1333.
M. Bouaifi, M. Roustan, R. Djebbar, Hydrodynamics of Multistage Agitated Gas Liquid reactors, in: Proceeding of the Ninth
European Conference on Mixing, Paris, France, 1997, 137
J.C. Lamont, D.S. Scott, An eddy cell model of mass transfer
into the surface of a turbulent liquid, A.I.Ch.E.J. 16 (1970)
(1970) 513.
R. Higbie, The rate of absorption of pure gas into still liquid
during short periods of exposure, Trans. Am. Inst. Eng. 31
(1935) 365.
C.M. McFarlane, Gas liquid mixing studies with a-315 and
prochem hydrofoil agitators, Ph. D. Thesis, Birmingham University, UK, 1991.
P.H Calderbank, M.B. Moo-Young, The continuous phase heat
and mass transfer properties of dispersions, Chem. Eng. Sci. 16
(1961) 39.

You might also like