You are on page 1of 11

Chemical Engineering Science 63 (2008) 3212 -- 3222

Contents lists available at ScienceDirect

Chemical Engineering Science


journal homepage: w w w . e l s e v i e r . c o m / l o c a t e / c e s

Bubbling process in stirred tank reactors I: Agitator effect on bubble size, formation and
rising
Mariano Martn , Francisco J. Montes, Miguel A. Galn
Departamento de Ingeniera Qumica y Textil, Universidad de Salamanca, Pza. de los Cados 1-5, 37008 Salamanca, Spain

A R T I C L E

I N F O

Article history:
Received 19 July 2006
Received in revised form 18 March 2008
Accepted 21 March 2008
Available online 27 March 2008
Keywords:
CFD
Bubbles
Stirred tanks
Bioreactors
Bubbling process
Impellers

A B S T R A C T

The study of the hydrodynamics generated by impellers and its effect on the generation of bubbles and
on their rising and dispersion is of key importance to improve the knowledge about the contact between
phases and the mass transfer rates, particularly in cases where it is the limiting step. CFD simulations and
high-speed video techniques are used to study the hydrodynamics developed by five different impellers,
each located at three different positions above the dispersion device. Furthermore, two dispersion devices
with one and two holes, respectively, are also used. The effect of the impellers on the characteristics
of the bubbles and of the dispersions generated has been analysed. Bubbles generated under stirring
are smaller than those generated in stagnant fluids. It is also shown that the initial bubble size at the
orifice determines the contribution of the impeller and the perforated plate to the Sauter mean diameter.
Although bubble formation is chaotic, the formation period is predictable based on three variables: the
location of the impeller, its rotational speed and the gas flow rate. Bubble mean diameter was correlated
to classical equations based on Kolmogorov's theory. Only when impellers are capable of breaking the
bubbles, Kolmogorov's theory is completely verified.
2008 Elsevier Ltd. All rights reserved.

1. Introduction
In order to improve the mass transfer rate between phases, many
times the gas phase is dispersed into the liquid phase not only using
dispersion devices but also mechanical mixing. This unit operation
is widely used in the chemical and biochemical industries. In fermentation processes and waste water treatment, the dispersion of
the gas phase is the most important operation.
The main factors affecting the mass transfer rate in gas--liquid
dispersions are the physico-chemical properties of the liquid, the
sparger design, the diameter of the orifice, the tank design, the type
of agitator and its relative dimensions, the power input, the gas
flow rate, the continuous phase, the presence of a chemical reaction,
the electrolyte concentration and the presence of catalysts (Sideman
et al., 1966). Mixing also defines the generation of bubbles inside
the tank (their frequency and initial size), their mean size and the
hydrodynamics of the whole system.
Several studies (Hassan and Robinson, 1977; Arjunwadkar et al.,
1998; Gogate et al., 2000; Veera et al., 2001; Bouaifi et al., 2001; Alves
et al., 2002) have been carried out in order to understand the effect of
the hydrodynamics on the mass transfer. The typical hydrodynamic

Corresponding author. Tel.: +34 923294479; fax: +34 923294574.


E-mail address: mariano.m3@usal.es (M. Martn).

0009-2509/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2008.03.028

parameters studied have been the bubble size distribution, the gas
hold up, G , the rising velocity of the bubbles and the effective power
input, which is very important for the scale up of the equipment.
However, the effect of the hydrodynamics developed by the impellers on the bubbling has barely been studied. It has been reported
by Marshall et al. (1993) that bubbles generated under flow fields
were smaller in comparison with those growing in stagnant liquids.
Only few models try to determine the effect of known liquid flows
on the generation of bubbles at submerged orifices (Marshall et al.,
1993; Nahra and Kamotani, 2003; Loubire et al., 2004).
Inside stirred tanks, the liquid flow under the impeller is characterised by turbulent eddies (McCabe et al., 1991). Bubble formation process and the dispersions generated will be affected by the
flow patterns developed by the impeller, which defines the flow
over the orifices as well as the stresses acting on the bubbles. According to Barabash and Belevitskaya (1995), bubble stability under
these stresses depends on its size. Furthermore, the impeller can also
guide the gas through preferential paths, so that the resident time
of bubbles is reduced (Parente et al., 2004).
In this paper, high-speed video techniques are used to study the
generation of bubbles under mixing conditions as well as to study
the mean size of the population of bubbles for five different agitators
(two pitched blade turbines, a modified blade, a Rushton turbine and
a propeller), two dispersion devices, and several gas flow rates and
rotational speeds. The position of the impellers was also modified

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

along the vertical axis. The final aim of this work is to study and
characterise the hydrodynamics in order to explain its effect on the
mass transfer rate. This explanation will be carried out in the second
part of the work.
As a first approach to the problem, the hydrodynamics of the
liquid will be simulated using computational fluid dynamics (CFD).
This approach is easy and saves time and money (Aubin et al., 2004;
Brucato et al., 1998; Montangne et al., 2001). It has already been
verified by visual methods such as Doppler velocimetry (LCD), or
particle image velocimetry (PIV), for typical geometries and agitators such as Rushton or pitched turbines (Ranade et al., 1992; Dong
et al., 1994; Hockey and Nouri, 1996; Nienow, 1997; Sahu et al.,
1999; Baldi and Yianneskis, 2004).
2. Theoretical considerations
Bubble initial volume is responsible for its stability in the fluid
flow and it will determine whether the bubbles are going to be
broken or not (Barabash and Belevitskaya, 1995). The breakage
of the bubbles is what provides contact area between phases.
Therefore, bubble dispersion depends strongly on the bubbling
process.
Although the models available in the literature for bubble growing
in non-stagnant fluids consider only the effect of known liquid flow
rates over the orifice (Marshall et al., 1993; Nahra and Kamotani,
2003; Loubire et al., 2004), they conclude that the flow reduces the
bubble initial volume since it drags the bubble from the orifices and
so it detaches sooner.
The theoretical study for the mean diameter of the bubbles in
an agitated gas--liquid dispersion has traditionally been carried out
according to Kolmogorov's theory of isotropic turbulence. The maximum stable diameter for a bubble or a drop is a function of the
Weber number of the system (Pacek et al., 1998).
The Sauter mean diameter, d32 , is widely considered to be proportional to the maximum stable diameter (Calderbank, 1958; Pacek
et al., 1998; Alves et al., 2002; Garca-Ochoa and Gmez, 2004), and
can be expressed as
d32 = C1

3/5
3/5 2/5

(1)

where  is assumed to be the input power per unit mass (Alves et al.,
2002). Calderbank (1958) also considered the effect of the gas hold
up on the mean diameter of the bubbles.
Although the Sauter mean diameter has traditionally been considered to be proportional to the maximum stable diameter, this
assumption is valid only if the bubble size is controlled by breakup processes. On the other hand, when coalescence processes play
an important role, d32 has been reported to be proportional to the
minimum diameter of the dispersion (Shinnar, 1961):
d32 = C2 0.25

(2)

However, since 1967, most authors have used the first consideration. Looking for simple equations for predicting bubble mean size
in stirred tanks and avoiding theoretical considerations, the typical
equation used is

d32 = kd


Pg 
V

(3)

The constants in the former equations depend on the dispersion


device as well as on the impeller type.
These equations can also be obtained as the solution of a population balance in which bubbles are broken due to turbulence whereas
their stability depends on the surface tension (Galindo et al., 2000).
 in Eq. (3) would depend on the balance between break-up and
coalescence processes, and thus it is smaller next to the impeller.

3213

Meanwhile, coalescence increases  (Bouaifi et al., 2001; Alves et al.,


2002).
3. Materials and methods
3.1. Experimental setup
The experimental setup is shown in Fig. 1. Bubbles are generated
inside a laser-sealed glass tank whose dimensions are 151515 cm.
In the middle of the tank, there is a gas chamber with dimensions
5 5 5 cm divided into two equal chambers. On the gas chamber,
a substructure, 2 2 2 cm, is placed in order to accommodate the
sieve plate on it. The liquid level reaches 8 cm of water over the
perforated plate. Deionised water was used as liquid media (20 C, =
998 kg/m3 ,  = 0.073 N/m, L = 1.037 103 Pa s). This configuration
allows an accurate visualisation of the bubbling process as well as
the study of the dispersions generated by different impellers and
dispersion devices.
Air bubbles are generated at two different perforated plates. The
first has only one orifice of 2 mm of diameter. The second has two
orifices of 2 mm each. In order to avoid coalescence between growing
bubbles, the separation between holes was 6 mm. However, bubble
deformation due to liquid flow makes coalescence easier (Martin
et al., 2008). Nevertheless, the big bubble resulting from coalescence
at the perforated plate would be easily broken by the impeller and
the reported cases of coalescence were of no statistical importance.
In the first case, only one of the two gas chambers was used while the
other was sealed. When the two-holed perforated plate was used,
each hole was the exit for each of the gas chambers. Due to the
configuration of the experimental setup, three gas flow rates of 0.6
106 , 1.4 106 and 2.8 106 m3 /s were used with the one-holed
dispersion device along with one gas chamber. Meanwhile, for the
two-holed dispersion device, lower flow rates were to be used, 0.3
106 , 0.6 106 and 1.4 106 m3 /s, since the air flow can break
the seal between chambers. A constant pressure regime has been
used.
The experimental conditions have been selected in order to generate normal flow patterns inside the tank. The impeller must be
located neither close to the dispersion device nor close to the free
surface. The mixing must be turbulent and it is desirable for the rotation velocity to be low in order to avoid flooding due to either
cavitation or central vortexes (Placek et al., 1986).
Therefore, the impeller was situated in the middle of the tank
at three different distances, h, from the dispersion device of 0.02,
0.035 and 0.05 m. Three rotational speeds, 180, 280 and 430 rpm,
were used and this experimental scheme was used for five different
impellers, a Rushton turbine, a propeller, a modified blade and two
different pitched blade turbines.
The dispersion generated as well as the individual bubbles were
recorded by means of a high-speed video camera able to record up
to 1000 frames per second. Motionscope software was used to edit
and study the images and to carry out the measurements of the
bubble sizes and formation time and frequency.
The Sauter mean diameter of the dispersion was calculated using Eq. (4), counting up to 200 bubbles. In case there was no such
number in a single frame, several frames, separated enough in time
not to study the same bubbles, were analysed in order to obtain a
representative value:

d32 = 

3
ni deqi
2
ni deqi

(4)

where ni is the number of bubbles with an equivalent diameter deqi .


It is determined considering the bubbles as ellipsoids, Fig. 2.

3214

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

Fig. 1. Experimental setup: 1. High-speed video camera; 2. optical table; 3. bubble column; 4. illumination source; 5. air bottles; 6. rotameter; 7. computer; 8. impeller.

Fig. 2. Equivalent diameter calculus.

3.2. Power number calculation


3.2.1. Theoretical model
The power number determines the real power input in the system. Due to the fact that in the experimental setup the dissipated
energy was kept low, so that bubbles could be recorded as individual entities, the experimental measurements were not reliable and
an alternative method was used. The hydrodynamics of the system
has been modelled using CFD.
Ansys CFX solves the momentum transport equations,
Navier--Stokes equations and continuity equation using the finite
element method. The turbulence method used is known as k.. The
value for the model constants are C = 0.09, C1 = 1.44, C2 = 1.92,
k = 1.0 and  = 1.3. The normal stress (2/3) k, is added to the
hydrostatic pressure. The equations solved can be found in Brucato
et al. (1998).
The boundary conditions to be imposed correspond to the tank
walls, the impeller geometry, both defined as smooth walls, and the
free surface that correspond to the atmosphere.
3.2.2. Geometries
Fig. 3 represents the geometries of the five different impellers
with their photographs. These 3-D geometries, as well as the tank,
were created using CFX . The dimensions of the geometries are on
the figures so that it is possible to clone the stirrers and the tank.
As it can be seen in the representation of the tank, a hole has to be
designed in order to make a way for the impeller. Its boundaries will
be coupled with the boundaries of each of the impellers.
4. Results
4.1. Power number verification
Table 1 shows the calculated power number for the different agitators at different positions using CFD. Since the results for the three
rotational speeds at each position were similar, a mean value was
calculated. The values in Table 1 are small. However, they are in
accordance with the fact that in case of regions in the tank with
relative low flow velocity, in spite of high turbulence near the im-

peller, the energy required to move the flow is small (McCabe et al.,
1991). The liquid velocity was much lower below and surrounding
the gas chamber than that in the impeller region, due to the location of the dispersion device, in order to record the generation of
the bubbles. This effect is similar to that shown for non-Newtonian
fluids for fluid regions far from the impeller (McCabe et al., 1991).
Furthermore, in the absence of baffles, the power number decreases
with the Reynolds number (McCabe et al., 1991).
Although Ranade et al. (1992) and Ranade (1997) have successfully compared experimental values of the power number resulting
from their simulations with calculated values, it is desirable to confirm the power numbers calculated due to the non-standard geometries used in this work.
To verify the values of the power number (P0 ), we tried to measure the power input. However, it turned out to be a very small
value and within the experimental error of the clamp AC/DC current
tester. Therefore, the simulated values of P0 were validated measuring mixing times.
The power number and the dissipated energy are related to the
mixing time in a tank, . Several empirical correlations have been
proposed to determine : Eq. (5) (Van't Riet and Tramper, 1991), and
Eq. (6) (Nienow, 1997):
 3
T
1/3
P0
N 1
D

=3

 1/3
T
1/3 D2/3
D

 = 5.9

(5)

(6)

To experimentally determine , the concentration of a colorant


(blue of methylene) will be monitored after its injection in the tank.
Five ml will be injected at a corner of the tank and every 5 s a sample
of 2.5 ml will be extracted. The volumes of the samples are negligible with respect to the total volume of the tank and the dilution also
allows the verification of Beer's law. The concentration will be measured using UV-V spectrophotometry at 590 nm, the wavelength at
which the maximum of absorption of the colorant is obtained, and
a HITACHI U-2000 Spectrophotometer. The colorant verifies Beer's
Law (Atkins, 1991). The mixing time is that for which the concentration after the injection differs less than 5% with that at time infinity.
The experimental conditions for which the mixing time was measured were h = 2 cm for each of the first five impellers at an impeller
speed of 180 rpm. Under these conditions, the experimental relative
error is the smallest, since the biggest mixing time is obtained and
the injection is performed without directing the colorant into the
impeller blades.
The measured mixing times were compared with those obtained
using the power number in empirical equations (5) and (6) (see
Fig. 4). The experimental results agree well with the predicted by
Eqs. (5) and (6), so that the simulated power numbers can be now
used in the study.

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

3215

Fig. 3. Impellers (a) Pitched blade turbine: T = 6 cm, Tj = 0.6 cm, Wa = 1.5 mm, Wb = 1 mm, Re = 5.1 cm, Rb = 6.5 cm, Di = 5 mm; (b) Modified blades: T = 4.8 cm, Ti = 0.35 cm,
Tj = 0.4 cm, R = 0.25 cm, Tk = 0.5 cm, D = 1 cm; (c) Rushton turbine: T = 5.2 cm, Td = 3.2 cm, Ti = 1.5 cm, Tj = 1 cm; (d) Pitched blade turbine: T = 5.6 cm, Tj = 0.7 cm, 45 ,
Di = 0.5 cm, R = 11.15 cm, W = 0.6 cm; (e) Propeller: T = 5 cm; Tlobulus = 1.7 cm, 30 .

Table 1
Simulated Power number
h (m)

Pitched blade (impeller 1)

Modified blade (impeller 2)

Rushton turbine (impeller 3)

Pitched blade (b) (impeller 4)

Propeller (impeller 5)

0.02
0.035
0.05

0.075
0.083
0.076

0.07
0.06
0.09

0.11
0.10
0.10

0.036
0.034
0.034

0.04
0.036
0.036

Table 2
Bubble formation without agitation
Qc (m3 /s)
6

tf (ms)

tt (ms)

deqi (mm)

1 orifice

0.6 10
1.4 106
2.8 106

31
30
30

170
88
47

7.0
7.9
8.7

2 orifices

0.3 106
0.6 106
1.4 106

33
33
32

186
145
99

4.8
5.3
5.1

4.2. Bubbling process

Fig. 4. CFD code verification.

The aeration number for all the impellers was kept low so that
the study of bubbles in dispersions generated is possible. Otherwise,
bubbles could not be visually identified as entities. According to
the empirical equations for the aeration number (Hughmark, 1980;
Walas, 1990), the aerated and the unaerated power can be considered
as equal under our experimental conditions.
Table 2 summarizes the results of the characteristics of the bubbles in the absence of stirring as a starting point for the comparison
of the effect of stirring on the bubbles.

3216

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

Although the generation of a single bubble under stirring is a


chaotic process, bubble formation time and size change from one
bubble to another under the same experimental conditions. In this
context, several patterns can be found.
Bubbles generated under stirring are affected by the flow developed in the tank and by turbulent eddies. In the first place, bubbles
on the dispersion device can be dragged due to the liquid flow developed by the impeller (see Fig. 5). Furthermore, it can also be seen
that for the same experimental conditions, bubble shape and formation time vary from one bubble to another, supporting the chaos behind bubbling. Second, the motion of bubbles depends on the main
driving force and the flow pattern generated by the impeller. For
the highest gas flow rate experimentally used, the initial bubble diameter is approximately 9 mm. In this case, bubbles are big and the

Fig. 5. Bubbles swept by the impeller (t, ms). Modified blade h = 2 cm; N = 180 rpm;
Qc = 0.6 106 m3 /s.

buoyancy force has an important contribution. However, at the same


time, the drag force increases because the cross area of the bubble
is also large. Therefore, bubble motion results from the combination
of both diving forces together with the flow lines generated by the
impeller, which depend on its geometry. Finally, the flow pattern
can lead the bubbles through preferential paths. The impeller location plays an important role in bubble suction from the orifice. The
higher the location of the impeller, the greater the pressure difference between the position of the impeller and that of the dispersion
device.
Based on the experimental results exposed before, the formation
period of the bubbles, tt , is almost predictable because it depends on
the gas flow rate and it is directly related to the flow pattern over the
holes of the dispersion devices as well as to the pressure difference
between the impeller and the dispersion device. Apart from the wellknown effect of the gas flow rate, which decreases tt , an increment
in power input also reduces tt for all the impellers. Bubbles are
sucked by the upwards vertical flow lines generated by the impeller.
The position of the impeller does not always modify tt ; however,
in general terms, the higher the impeller is located, the shorter the
period time, due to the suction effect of the central ascendant current
under the impellers and the pressure difference generated between
the impeller and the perforated plate. Fig. 6 shows the experimental
results for the one-holed dispersion device.
In contrast, bubble formation time, tf , is not so predictable. Nevertheless, tf usually decreases slightly with the gas flow rate and the
power input. Bubbles are sucked and dragged from the dispersion
device and their detachment is earlier compared with stagnant fluids.

Fig. 6. Period of bubble formation. 1 orifice dispersion device.

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

3217

Fig. 7. Initial bubble size. 1 orifice dispersion device.

As it can be seen in Fig. 7, in general, the power input reduces the


initial bubble size. Bubbles are swept and helped detach from the
dispersion device due to the impeller agitation. Since tf decreases
slightly with the impeller speed, so does the initial bubble size. The
gas flow rate across the orifice also has an important effect on the
mean bubble size. Since tf remains almost constant with the gas flow
rate, bigger bubbles are generated as the gas flow rate increases.
Initial bubble size is important to determine whether the bubbles
are going to be dragged by the flow or could be broken.
Both the raising of the bubbles and the path followed are influenced by the initial size of the bubbles along with the flow pattern
generated by the impeller. The movement of big bubbles is determined not only by the buoyancy force, but also by drag forces due to
the large cross-sectional areas, considering that the bubbles do not
break. Both contributions and the flow lines generated by the different impellers, as a result of the particular geometry of their blades,
can make the bubble rise in zig-zag (pitched blade turbines) or in
a line (Propeller and two modified blades). In the case of smaller
bubbles, both effects reduce their contribution to the movement
of the bubbles. As a result, bubbles can easily be dragged by the
flow. Fig. 8 represents the rising paths for the pitched blade turbine and the propeller. To the right, the gas flow rate increases from
0.6 106 to 2.8 106 m3 /s. Meanwhile, downwards the rotational speed increases from 180 to 430 rpm. The photographs are
taken at the highest distance (h = 0.05 m). The rising path generated
by both can be fitted to a Fourier transform. The particular flow pattern developed by the propeller makes the bubbles rise almost in a
line. For lower positions of the impeller, the paths are barely developed, since the bubbles are generated in the zone of influence of the
impeller.

From these results, the bubble mean size and the effect of the
position of the impeller on the mass transfer rate can be easily explained. Bubbles will be broken either by deformation due to the
flow or by direct impact with the blades. According to the particular rising path, the bubbles can even avoid the effect of the blades
in certain conditions; therefore, it can be predicted that as the impeller is placed higher, bubbles will be broken mainly as a result of
their deformation, and if it is located lower, the direct effect of the
impeller will be the main break-up mechanism.
4.3. Bubble dispersions
Mean bubble size depends on the capability of the impeller of
breaking the bubbles generated at the dispersion device. Meanwhile,
bubble stability in a tank depends on its size. Big bubbles are easily
deformed and broken while smaller ones are dragged by the flow
across the tank with little deformation. Fig. 9 represents the deformation of a bubble generated under a Rushton turbine. The regular
shapes shown in stagnant fluids are no longer seen under stirring.
The Sauter mean diameter for both dispersion devices, one-holed
and two-holed perforated plates, was determined considering the
bubbles as ellipsoids and calculating the equivalent diameter of each
bubble, deqi , of the dispersion according to Fig. 2. The definition of
the Sauter mean diameter was used, Eq. (4), so that the experimental
results could be fixed to Eq. (3). Table 3 summarizes the results.
The results presented in Table 3 can be explained based on the
observed hydrodynamics. Big bubbles generated from the one-holed
dispersion device can be easily deformed and broken. As a result, the
values of  are large, in absolute terms. Smaller bubbles are more
stable in the flow and this leads to smaller values of , in absolute

3218

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

Fig. 8. Bubbles rising under pitched blade and propeller. h = 0.05 m.

Fig. 9. Bubble deformation under a Rushton turbine (t, ms). h = 0.05 m; N = 430 rpm; Qc = 1.4 106 m3 /s.

Table 3
Sauter mean diameter versus power input
One-holed

Two-holed

kd 10

kd 103

Pitched blade
Modified blade
Rushton turbine
Pitched blade (b)
Propeller

5.2
3.4
4.6
4.0
4.2

0.26
0.35
0.38
0.34
0.25

5.0
4.8
3.7
3.4
4.5

0.30
0.22
0.26
0.23
0.04

Air--water system

4.4

0.25

3.8

0.16

The effect of the five impellers.

terms. As a result, the dispersion device has a higher contribution to


the mean size of the dispersion.
Considering all the impellers for each dispersion device, mean
values of kd or  were obtained. As it can be seen in Table 3, the
mean values of kd and  are similar to those reported by Bouaifi
et al. (2001) and Alves et al. (2002), for the air--water system.

Furthermore, the results obtained impeller by impeller are also


in accordance with Alves's results, which reveal that if break-up
mechanisms determine bubble size,  increases, in absolute terms.
Meanwhile, coalescence and lack of breakage decrease , in absolute
terms.
The location of the impeller along the vertical axis for h = 0.02
and 0.035 cm, the locations of the impeller where it was possible
to measure a complete dispersion in the experimental setup, has
low effect on the break-up efficiency of the power input for the
dispersions generated from the one-holed perforated plate. Bubbles are big enough to be easily broken. The only impeller whose
results were clearly affected by the position of the impeller was
the first pitched blade turbine, impeller 1. A higher position of
the impeller resulted in a loss of break-up efficiency due to the
fact that the bubbles could avoid the effect of the impeller. However, if the bubbles generated at the dispersion device are smaller,
such as those generated from the two-holed perforated plate, a
higher position of the impellers reduces the break-up efficiency
for almost all the impellers. Table 4 summarizes the experimental
results.

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222


Table 5
Sauter mean diameter versus power input

Table 4
Effect of the location of the impeller on the break-up efficiency

h (cm)

One-holed

2
3.5
5

0.23
0.27
0.30

0.35
0.27
0.35

Modified blade

2
3.5

0.33
0.33

0.29
0.12

Rushton turbine

2
3.5

0.37
0.38

0.32
0.20

Pitched blade (b)

2
3.5

0.32
0.35

0.27
0.18

Propeller

2
3.5

0.24
0.26

0.05
0.07

Pitched blade

3219

Two-holed

One-holed
kd

kd

Pitched blade
Modified blade
Rushton turbine
Pitched blade (b)
Propeller

0.62
0.46
0.67
0.52
0.54

0.26
0.32
0.34
0.35
0.24

0.78
0.54
0.71
0.57
0.82

0.29
0.20
0.20
0.20
0.03

Air--water system

0.57

0.25

0.65

0.15

The effect of the five impellers and the initial bubble size.

Since the initial bubble size determines the stability of the bubbles, and this size depends on the dispersion device and the gas flow
rate impeller, it is convenient to use the following equation to isolate
the effect of the impeller on the bubbles:
d32
= kd
db

Fig. 10. Bubble mean size one-holed dispersion device.

Fig. 11. Bubble mean size two-holed dispersion device.

Figs. 10 and 11 show the effect of the power input on the mean
bubble size of the dispersion for the two different dispersion devices
with all the impellers.  decreases as the generated bubbles at the
orifice are smaller since they are more stable in the flow.

Two-holed


Pg 
V

(7)

The empirical coefficients are presented in Table 5. The values of 


show almost no change for the one-holed perforated plate and a little
change for the two-holed perforated plate. However, kd increases
when the stability of the initial bubble increases. Meanwhile, the
values of kd in Table 3 remain stable.
It is possible to explain the coefficients of the correlation (Tables 3
and 5) based on the photographs of the bubbles (Fig. 12).
When the impeller is capable of breaking the bubbles, in general
for big initial bubbles, for example, those generated from the oneholed perforated plate,  was next to that predicted by Kolmogorov's
theory (0.4). For these cases the mean bubble diameter of the dispersion is defined by the break-up process.
However, if the impeller cannot break the bubbles, this coefficient
is lower, in absolute terms. There are four different reasons for obtaining a low value of . The first reason is if the bubble size is within
the stable range for the operating conditions. In general, those generated from the two-holed perforated plate are small bubbles which
follow the flow lines. The blades of the impeller can barely deform
them. The second reason is if the bubbles can avoid the blades. This
occurs when the rising path of the bubbles, resulting from the hydrodynamics generated by each impeller combined with the effect of
drag and buoyancy forces on the bubbles, allows them to bypass the
impeller blades. The third reason is in case the blades of the impeller
cannot break the bubbles. This occurs when either the geometry of
the blades does not allow developing a wake where the bubbles can
be retained or the bubbles are not hit by the impeller blades. The
propeller and the three-blade turbine named as impeller 1 show this
fact. The last reason can be the coalescence processes taking place
due to bubble collisions. However, they were rarely found in the
recording. As a result, the dependency of the Sauter mean diameter
on the power input is less important.
Using these principles, it is possible to justify the break-up efficiency of each impeller:
For the pitched blade turbine (impeller 1) the change in the
regime with the rotational speed can be seen in Fig. 12(a). At low
rotational speeds, bubbles cannot be broken and rise almost as if no
interference with the impeller exists. Bubbles can avoid the blades.
An increase in the rotational speed increases the break-up processes
and leads to the generation of a true dispersion. The sharp blades of
this impeller allow a small loss of break-up efficiency when the initial bubbles are small, such as those generated from the two-holed
perforated plate, since bubbles are deformed until they break up in
the discharge of the gas or cut.
The particular geometry of the modified blade impeller, impeller 2, determines its effect on the bubbles. The peaks generate a

3220

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

Fig. 12. (a) Pitched blade turbine: rotational speed effect h = 0.02 m, Qc = 0.6 106 m3 /s; (b) Modified blade: rotational speed effect---h = 0.02 m, Qc = 2.8 106 m3 /s; (c)
Gas dispersion generated by a Rushton turbine (d) Gas dispersion generated by a pitched blade turbine. Effect of gas flow rate---h = 0.02 m, N = 430 rpm. (e) Gas dispersion
generated by a propeller.

low-velocity region along which the bubbles rise. The geometry of


the blade determines the break-up mechanism, whether the bubbles are broken due to the deformation at the end of the impeller

or due to the direct cut with the blade. This impeller suffers from
a high loss of efficiency in bubble break-up when their initial size
is smaller, because they can either avoid the blades in their rising

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

movement, so that they are neither cut by the blades nor retained
at the end of the blades to be broken at the discharged, or cannot
be deformed (Fig. 12(b)).
In the case of the Rushton turbine, the most used impeller for fermentation processes, two results can be reported. In the first place,
the disk retains the bubbles up to their break-up. This fact increases
the gas hold up and also the fact that when bigger bubbles are generated, break-up processes are easier. Fig. 12(c) represents the dispersion of gas generated by a Rushton turbine. This impeller has almost
the same break-up efficiency no matter the initial size of the bubble,
the effect of the disk retaining the bubbles until their breakage in
the discharge of the gas phase is the key, maintaining the value of
 next to that predicted by Kolmogorov's theory for both dispersion
devices.
The second pitched bladed turbine breaks the bubbles due to
deformation of the gas at the concave blade. Fig. 12(d) shows the
generation of the dispersion of bubbles in this particular case. The
blades are more hydrodynamic than those of the first turbine.
The bubbles are not cut by the blades. The flow pattern developed,
able to deform the bubbles, and the accumulation of gas at the
end of the impeller allow a small loss of efficiency in the break-up
process with the initial size of the bubble.
The most peculiar dispersion generated by any of the impellers is
that generated by the propeller. This impeller, typical in ship propulsion, develops a flow pattern in which the bubbles are barely broken
but retained under the impeller. Tables 3 and 5 present the absence
of break-up processes for small bubbles, those generated at the twoholed perforated plate, where  is almost zero. The propeller can
only smack the bubbles; meanwhile, the geometry of the blades of
the turbine does not generate a wade wide enough so as to retain
the bubbles up to their break-up by deformation. The positive value
is due to the deformation of the bubbles in the flow. An explanation can be found in the flow pattern developed by the impeller. A
low-pressure region under the impeller is obtained and therefore,
bubbles remain there for a while.
Furthermore, the physical effect of the impellers on the gas phase,
shown in the equations, whether they are theoretical or empirical,
is responsible for their break-up. This is the reason why there are
differences between the empirical fitting coefficients obtained for
different geometries and impeller configurations.
5. Conclusions
Since mass transport processes depend on the hydrodynamics
of the system, its characterisation allows a better understanding of
the process. CFD has proved to be a powerful tool in order to study
difficult geometries in an easy and safe way.
The mixing in a tank is used to improve the contact between
phases by dispersion of the gas phase. The dispersion generated by
each impeller depends on its geometry, responsible for the dominant
break-up mechanism (bubble deformation or elongation, direct cut),
and it is conditioned by its effect on the bubbling, which, although
is a chaotic process, determine the initial bubble size. In general,
the gas flow rate increases the initial bubble size; at the same time
power input and a higher location of the impeller reduces it. Furthermore, although the effect of the position of the impeller on the
characteristics of the growing bubbles is not well defined, a higher
position allows the bubbles to avoid the direct effect of the impeller,
reducing its efficiency. This is a very important result regarding the
mass transfer rate.
The dispersion device determines the mean bubble size and the
generated dispersion if the power input is unable to break the bubbles. Small bubbles, such as those generated in the two-holed dispersion device, are difficult to be broken and move with the flow.
Bigger bubbles are easily deformed and broken.

3221

The performance of the impellers in the breakage of the bubbles can be compared with the results of Kolmogorov's theory
of turbulence. Coalescence processes as well as stable bubbles,
small bubbles unable to be broken, reduce the dependency of
the mean bubble diameter on the power input, and only when
the break-up of the bubbles domain the process of gas dispersion, the theory of Kolmogorov is verified. The use of Calderbank's
equation for the specific area (Calderbank, 1958) is limited to the
impellers able to break the bubbles, where coalescence processes
can be neglected, since it is somehow based on Kolmogorov's
theory.
The rising path of the bubbles depends on the flow pattern of
each of the impellers as well as on the bubble size, which determines
the contribution of the drag and buoyancy forces to the movement
of the bubble.
Notation
a
a
b
C1 , C2 , C
d32
db

specific superficial area, m1


horizontal bubble diameter, m
vertical bubble diameter, m
model constants
Sauter mean diameter, mm
initial bubble diameter, m

deqi
D0
DC
Di
Fr , F
FlG

equivalent diameter, m, deqi = (a 2 b )(1/3)


orifice diameter, m
tank diameter, m
shaft diameter, m
volume forces, N m3
aeration number FlG = Qc 3

h
H
k
L
ni
N
P
Pg
P0

impeller height above the dispersion device, m


blade height, m
kinetic turbulent energy J kg1
impeller characterists, m
number of bubbles of class i
rotational speed, s1
unaerated power, W
aerated power, W
Power number P0 = 5 P 3

Qc
R

gas flow rate, m3 s1


blades curvature, m

Re
T
Ti
tM
tf
tt
uG
V
w
W

Reynolds number, Re = 
impeller diameter, m
blades length, m
mixing time, s
formation time of a bubble, ms
formation period of bubbles, ms
gas superficial velocity, m s1
liquid volume, m3
blade velocity, m s1
blade width, m

We

Weber number, We =

NT

T N 

T 2N

N 2 T 3


Greek letters
, ,
, 

g
, eff , T
G



empirical coefficients
dissipated energy, W kg1
gas hold up
laminar, effective and turbulent viscosities, Pa s
gas density, kg m3
liquid density, kg m3 1
superficial tension, N m
stress, N m2
angular velocity, rad s1

3222

M. Martn et al. / Chemical Engineering Science 63 (2008) 3212 -- 3222

Acknowledgements
The support of the Ministerio de Educacin y Ciencia of Spain
providing an F.P.U. fellowship to M. Martn is greatly acknowledged.
The funds from the project reference CTQ 2005-01395/PPQ are also
appreciated. We thank Prof. J. Cuellar, Chemical Engineering Department, University of Salamanca, for lending us some of the impellers
used in this paper.
References
Alves, S.S., Maia, C.I., Vasconcelos, J.M.T., Serralheiro, A.J., 2002. Bubble size in aerated
stirred tanks. Chemical Engineering Journal 89, 109--117.
Arjunwadkar, S.J., Sarvanan, K., Pandit, A.B., Kulkarni, P.R., 1998. Optimizing the
impeller combination for maximum hold up with minimum power consumption.
Biochemical Engineering Journal 1, 25--30.
Atkins, P.W., 1991. Fisicoqumica, 3a Edicin. Addison Wesley Iberoamericana S.A.
Wilmington, Delaware, EE. UU.
Aubin, J., Fletcher, D.F., Xuereb, C., 2004. Modeling turbulent flow in stirred tanks
with CFD: the influence of the modelling approach, turbulence model and
numerical scheme. Experimental Thermal and Fluid Science 28, 431--445.
Baldi, A., Yianneskis, M., 2004. On the quantification of energy dissipation in
the impeller stream of a stirred vessel from fluctuating velocity gradient
measurements. Chemical Engineering Science 59, 2659--2671.
Barabash, V.M., Belevitskaya, M.A., 1995. Mass transfer from bubbles and crops
in mechanically agitated apparatuses. Theoretical Foundations of Chemical
Engineering 29 (4), 333--342.
Bouaifi, M., Hebrard, G., Bastoul, D., Roustan, M., 2001. A comparative study of gas
hold up, bubble size, interfacial area and mass transfer coefficients in stirred
gas liquid reactors and bubble columns. Chemical Engineering and Processing
40, 97--111.
Brucato, A., Ciofalo, A., Grisafi, F., Micale, G., 1998. Numerical prediction of flow fields
in baffled stirred vessels: a comparison of alternative modelling approaches.
Chemical Engineering Science 53, 3653--3684.
Calderbank, P.H., 1958. Physical rate processes in industrial fermentation, Part 1: the
interfacial area in gas liquid contacting with mechanical agitation. Transactions
of the Institute Chemical Engineering 36, 443--463.
Dong, L., Johansen, S.T., Engh, T.A., 1994. Flow induced by an impeller in an
unbaffled tank---II. Numerical modelling. Chemical Engineering Science 49 (20),
3511--3518.
Galindo, E., Pacek, A.W., Nienow, A.W., 2000. Study of drop and bubble sizes in
a simulated mycelial fermentation broth of up to four phases. Biotechnology
Bioengineering 69 (2), 213--227.
Garca-Ochoa, F., Gmez, E., 2004. Theoretical prediction of gas--liquid mass
transfer coefficient, specific area and hold up in sparger stirred tanks. Chemical
Engineering Science 59, 2489--2501.
Gogate, P.R., Beenackers, A.A.C.M., Pandit, A.B., 2000. Multiple-impeller systems with
a special emphasis on bioreactors: a critical review. Biochemical Engineering
Journal 6, 109--144.

Hassan, I.T.M., Robinson, C.W., 1977. Stirred tank mechanical power requirement
and gas hold up in aerated aqueous phases. A.I.Ch.E. Journal 23 (1), 48--56.
Hockey, R.M., Nouri, J.M., 1996. Turbulent flow in a baffled vessel stirred by a 60
pitched blade impeller. Chemical Engineering Science 51 (19), 4405--4421.
Hughmark, G.A., 1980. Power requirements and interfacial area in gas--liquid
turbine agitated systems. Industrial & Engineering Chemistry Process Design and
Development 19, 638--641.
Loubire, K., Castaignde, V., Hbrard, G., Roustan, M., 2004. Bubble formation at
al flexible orifice with liquid cross flow. Chemical Engineering Processing 43,
717--725.
Marshall, S.H., Chudacek, M.W., Bagster, D.F., 1993. A model for bubble formation
from an orifice with liquid cross flow. Chemical Engineering Science 48 (11),
2049--2059.
Martin, M., Garca, J.M., Montes, F.J., Galan, M.A., 2008. On the effect of the sieve
plate configuration on the coalescence of growing bubbles. Chemical Engineering
and Processing: Process Intensification, in press, doi:10.1016/j.cep.2007.10.002.
McCabe, W.L., Smith, J.C., Harriot, P., 1991. Operaciones unitarias en Ingeniera
qumica, 4o Ed. McGraw Hill, Barcelona (Chapter 9).
Montangne, G., Lee, K.C., Brucato, A., Yianneskis, M., 2001. Numerical simulations
of the dependency of flow patterns on impeller clearance in stirred vessels.
Chemical Engineering Science 56, 3751--3770.
Nahra, H.K., Kamotani, Y., 2003. Prediction on bubble diameter at detachment from
a wall orifice in liquid cross flow under reduced and normal gravity conditions.
Chemical Engineering Science 58, 55--69.
Nienow, A.W., 1997. On impeller circulation and mixing effectiveness in the turbulent
flow regime. Chemical Engineering Science 52 (15), 2557--2565.
Pacek, A.W., Man, C.C., Nienow, A.W., 1998. On the Sauter mean diameter and size
distributions in turbulent liquid/liquid dispersions in a stirred vessel. Chemical
Engineering Science 53 (11), 2005--2011.
Parente, E., Piraino, P., Fidaleo, M., Moresi, M., 2004. Overall volumetric oxygen
transfer coefficient in an aerated bench-top stirred fermenter in aqueous
dispersions of sodium alginate. Biotechnology and Applied Biochemistry 40,
133--143.
Placek, J., Tavlarides, L.L., Smith, G.W., Fort, I., 1986. Turbulent flow in stirred tanks:
Part II A two scale model of turbulence. A.I.Ch.E. Journal 32 (11), 1771--1786.
Ranade, V.V., 1997. An efficient computational model simulating flow in stirred
vessels: Rushton turbine. Chemical Engineering Science 52 (24), 4473--4484.
Ranade, V.V., Mishra, V.P., Saraph, V.S., Deshpande, G.B., Joshi, J.B., 1992. Comparison
of axial flow impeller using a laser Doppler anemometer. Industrial Engineering
Chemistry Research 31, 2370--2379.
Sahu, A.K., Kumar, P., Patwardhan, A.W., Joshi, J.B., 1999. CFD modelling and mixing
in stirred tanks. Chemical Engineering Science 54, 2285--2293.
Shinnar, R., 1961. On the behaviour of liquid dispersions in mixing vessel. Journal
Fluid Mechanics 10, 259--275.
Sideman, S., Hortacsu, O., Fulton, J.W., 1966. Mass transfer in gas--liquid contacting
systems. Industrial Engineering Chemistry 58 (7), 32--47.
Van't Riet, K., Tramper, J., 1991. Basic Bioreactor Design. Marcel Dekker, New York.
Veera, U.P., Patwardhan, A.W., Joshi, J.B., 2001. Measurement of gas hold-up profiles
in stirred tank reactors by gamma ray attenuation technique. Transactions of
the Institute of Chemical Engineering 79 (Part A), 684--688.
Walas, S.M., 1990. Chemical Process Equipment: Selection and Design. ButterworthHeinemann, Boston (Chapter 10).

You might also like