Chem. Eng. Technol.

10 (1987) 204-215

204

Gas/Liquid Interfacial Area per Unit Volume and Volumetric

Mass Transfer Coefficient in Stirred Slurry Reactors
Michael Schmitz, Artur Steiff and Paul-Michael Weinspach*
In three-phase systems, where the liquid constitutes the continuous phase, solid is the catalyst
and gas represents the dispersed phase, there are decisive criteria which have to be observed in
reactor design. These are e.g. the interfacial area per unit volume between gas and liquid, the
volumetric mass transfer coefficient and the mass transfer coefficient. The basic aim of the
present work was therefore the investigation of these parameters in relation to the main influencing parameters. Process parameters stirrer speed and superficial gas velocity were varied as well
as the physical properties such as liquid viscosity, solids concentration, particle diameter and a
geometrical parameter, i.e. reactor diameter. The sulphite method was employed for the determination of these values. The test results confirmed the known relationships of power consumption
and superficial gas velocity. An increase in the liquid viscosity leads to a decrease in all the tested
values. In most cases, suspended solid particles lead to a lowering of the test values. The influence
of the tank diameter on the plots of the test values against specific power consumption turned
out to be invariant so that a scale-up of geometrically similar systems can be carried out at constant
power consumption, superficial gas velocity and liquid viscosity.

1 Introduction
The investigation and optimization of multiphase systems is at
present a major activity in process engineering. Although several dimensioning results for single- and two-phase reactors are
available, such data material for three-phase reactors is still
lacking.
Apart from the scientific aim of determining the complex relationships and interactions in such systems and of describing
these processes by physical models, from the industrial point of
view, it is just as important to solve the problems of scale-up
from a laboratory or pilot plant to an industrial size.
The mentioned three-phase reactors are
- fixed bed reactors and
- slurry reactors.
The term fixed bed reactor embraces all reactors, where both
gas and liquid flow in co- or countercurrent through stationary
solid filling material which can be either reactive or inert. In
slurry reactors, i.e. three-phase bubble column, three-phase
fluidized-bed reactor and stirred slurry reactor, the solid is suspended either by mechanical or by fluid-induced agitation. An
extensive comparison of these reactor types, together with their
process engineering advantages or disadvantages, was given by
Shah [l] and Kiirten [2].
The stirred three-phase reactor is employed in industry for the
catalytic hydrogenation of unsaturated fats and fatty acids and

Dr.-Ing. M. Schmitz (present address: Fa. Enka AG, Werk Oberbruch,

Boos Fremery StraRe, D-5138 Heinsberg), Dr.-Ing. A. Steiff and Prof.
Dr.-Ing. P.-M. Weinspach, Lehrstuhl f i r Therrnische Verfahrenstechnik der Abteilung Chemietechnik der Universitit Dortmund, Postfach
500 500, D-4600 Dortmund 50.

0 VCH Verlagsgesellschaft mbH, D-6940 Weinheim, 1987

for the reaction between HCl and CH3OH in the presence of

ZnCl2 catalyst.
A common feature of all these processes is the gas component
which passes through the gadliquid interface. If mass transfer
is a diffusion-controlled process, the magnitude of the interfacial area per unit volume exerts a decisive influence. The main
objective of this work was to provide a contribution to the
dimensioning of stirred slurry reactors, based on the influence
of power consumption on interfacial area per unit volume. In
addition, the volumetric mass transfer coefficient and the mass
transfer coefficient were to be determined at identical fluid dynamic conditions.

2 Measurement Method
Since conventional physical test methods for the determination
of interfacial area per unit volume (photography, conductivity,
photoelectric probe method) are not suitable for three-phase
systems and would lead to incorrect results (reasons are discussed in [3]), a chemical method, i.e. the sulphite method, was
chosen for the determination of the test values.
The overall reaction equation is:
Na2S0,

O2

k2

co2+

Na2S0,

Absorption of oxygen in cobalt-catalyzed aqueous sodium sulphite solutions is a method frequently used both in laboratory
and in industrial plants to determine the interfacial area per unit
volume and the volumetric mass transfer coefficient.
For the basic principles of the sulphite method and derivations
of the equations as well as an extensive explanation of the calculation procedure, the reader is referred to [3,4].
0930-7516/87/0306-0204 $02.50/0

205

Chem. Eng. Technol. 10 (1987) 204-215

3 Material Systems
The employed mass transfer system was an aerated, aqueous sodium sulphite solution with suspended inert solid particles. The
test conditions were set as follows'):

4 Experimental Set-up

pH:
temperature:

8,
20 "C,

pressure:

atmospheric pressure,

sulphite concentration:

0.4 kmol/m3 5

0.8 kmo1/m3,
catalyst concentration in
determination of interfacial area:

4 x 1 0 - ~kmoi/m3 5 cC:+ 5

8 x
catalyst concentration in
determination of volumetric mass transfer
coefficient k,a,

cC:+

kmol/m3,

5 x

kmo1/m3.

The physical properties used for the calculations are described

in [3]. CMC (carboxymethyl cellulose) was added to pure sulphite solution in order to increase liquid viscosity. The calculated liquid viscosities, according to the concept of Metzner and
Otto [ 5 ] , were between 1.45 and 140 mPa s, depending on the
added CMC-proportions.
1) List of symbols at the end of the paper.

I Pressure

air

(Coolngwoferr

Inert glass beads were used as the solid. Their density was between 2460 and 2490 kg/m3; particle fractions with average diameters of 0.088 and 0.32 mm were used.

Fig. 1 shows a diagram of the equipment. The central part is

a stirred slurry reactor ( I ) , in which the liquid represents the
continuous phase, while the gas and solid are the dispersed
phases. Two agitated vessels made of glass (internal diameters
0.2 and 0.45 m) are used, with geometries corresponding to
those usually mentioned in the literature. A stainless steel vessel, 1.5 m in diameter, was used as well. The ratio of nonaerated liquid height to reactor diameter is unity. Four, symmetrically arranged baffles, utilized as cooling elements, have
diameters of 0.1 D.
Sparger rings were employed as gas distributors. Their arrangement allowed the same hole gas velocities at equal superficial
gas velocities. The turbine impellers have diameters of 1/3 D.

5 Results
5.1 Influence of Impeller Rotational Speed and Superficial Gas
Velocity
The influence of stirring intensity and superficial gas velocity
was to be described first for the two- and three-phase systems
I

I t
I

1 Reactor
2 Torque
pick-up
.
.
.
3 Direct current motor
4 Tacho-alternator
5 Pulse generator
6 Pressure regulator
7 Flowmeter
8 Adjusting valve
9 Pressure meter
10 Air filter
11 pH measurement
12 Gas analyzer for oxygen
13 sampling of liquid

Fig. 1.
Experimental plant.

Chem. Eng. Technol. 10 (1987) 204 - 215

206
with and without added CMC. Fig. 2 shows the interfacial area
per unit voIume as a function of stirrer speed for three different
superficial gas velocities. In the two-phase system, the interfacial area is independent of stirring speed in the range of low
speeds, i.e. from 0 to 200 l/min, because the stirrer is flooded.

As shown in Fig. 2 , the influence of superficial gas velocity

dominates in the flooded range. The line, drawn according to
a correlation of Wiedmann [6], describes the demarcation to the
non-flooded range. To the right of this line, there is a strong
linear increase in the interfacial area per unit volume with increasing stirring speed, produced by the setting-in dispersion effect of the agitator and the corresponding dispersion of gas bubbles. In this range, an increase of superficial gas velocity leads
only to a slight increase in the interfacial area.
For the described three-phase data ($s = 0.3; d , = 0.32 mm),
the steeper slope of the regression curve leads to the conclusion
that the interfacial area becomes more dependent on stirrer
speed. In the range of low stirrer speeds (for the used tank n
< 1200 1/min) where the solid is completely suspended according to the 1-s criterion, the solid reduces the interfacial area.
This is caused by the non-homogeneous suspension of the solid,
or rather the significant concentration profile over the reactor
height, as experimentally determined by Bohnet [7]. The solid
is mainly restricted to the bottom part of the reactor where it
restrains the formation of a bubble layer in the lower part of the
reactor and impairs the dispersion effect of the stirrer. At high
rotational speeds (n > 1200 l/min), whereby the solid attains
a nearly homogeneous suspension state, the interfacial areas of

the three-phase system tend to higher values than those of the

two-phase system. This can be explained by higher kinetic energy of the particles at high stirrer speeds; the particles deform
an ascending gas bubble and thus produce an enlargement of the
interfacial area. The statement, that a non-homogeneous solid
dispersion adversely affects the interfacial area, is supported by
experience, namely, in three-phase systems, in contrast to comparable two-phase systems, the interfacial areas are smallest at
high rotational speeds and the highest set superficial gas veloci~ 4.6 cm/s.
ty of U G =
The relationship between the interfacial area and stirring speed
(see Fig. 2 ) allows no further conclusions as to whether the addition of inert solid particles, in order to increase the interfacial
area, is of energetic advantage or not. For an energetic assessment, a plot of interfacial area versus power consumption is
more suitable (see Fig. 3). This plot also permits a comparison
of different reactor types. In lower power consumption ranges
( P t o t / V ~< 1 kW/m3), the curves for different superficial gas
velocities coincide. Within this range, it is of no importance
whether the dissipated power is provided by stirring or by aeration. In higher power consumption ranges (Ptot/VL > 1 kW/
m3), a fanning of the curves is obtained for the two-phase
system. At constant power consumption, higher superficial gas
velocities lead to significantly larger interfacial areas.
Therefore, a description of interfacial area per unit volume,
with Ptot/VL as the only parameter, e.g. as presented by Midoux
[8], cannot be recommended. At constant power consumption,
the values for the three-phase system are, as a rule, lower than
those for the two-phase system. Only at the lowest superficial
~ 0.34 cm/s, larger interfacial areas are
gas velocity of U G =
produced by the employed particles in the three-phase system,
at a high power consumption.
The results of Fig. 3 are presented in Fig. 4 for the volumetric
mass transfer coefficient. The shapes of the curves in the two
L,

I l l

I l l

I l l

rn

lo3

10

1o2

103 I/min

n -

Fig. 2. Interfacial area per unit volume as a function of stirrer speed and
superficial gas velocity, for stirred two- and three-phase systems ($CMC =
0).

10

lo-
901

VL

kW/m3

lo

Fig. 3. Interfacial area per unit volume as a function of total power consumption per unit volume in stirred two- and three-phase systems ($cMc = 0).

Chem. Eng. Technol. 10 (1987) 204-215

207

Fig. 4. Volumetric mass transfer coefficient as a function of total power consumption per unit volume in stirred two- and three-phase systems ($CMC
0).

diagrams are almost identical, which leads to the conclusion that

the kL-value varies only slightly.
The fanning of the curves in the two-phase system shifts
towards lower power consumptions, which can be attributed to
the stronger influence of the superficial gas velocity on the ~ L U L value or to a relationship between the kL-value and the superficial gas velocity.

Fig. 5. Interfacial area per unit volume as a function of total power consumption per unit volume in stirred two-phase systems ( ~ C M C = 0 and
0.5%).

The measurements for the results shown in Figs 5 and 6 were

also carried out in three-phase systems. These results are summarized in Figs 7 and 8. In analogy to the two-phase system,
increased liquid viscosity leads to an intensified effect of the superficial gas velocity on a,-and k,a,-values. Furthermore, viscosity increase causes a stronger reduction of the kLaL-value,
in comparison to the aL-value.

~ 0.34 cm/s, the kLaL-value Finally, the influence of the total power consumption and superAs shown by the curve for U G =
is more reduced by the suspended solid than the interfacial area ficial gas velocity on the mass transfer coefficient k, should be
per unit volume. The mass transfer coefficient kL is obviously discussed. Figs 9 and 10 present the mass transfer coefficient
as a function of the total power consumption per unit volume
influenced by the suspended solid.
for the two superficial gas velocities. Liquid viscosity and solids
The way in which the total power consumption and superficial
gas velocity influence the interfacial area in more viscous sodium sulphite solutions (vL = 25 to 60 mPa s) is shown graphically in Fig. 5. For comparison, the results for pure sulphite solution are presented as well. Apart from a few tests at high superficial gas velocities (uGo = 4.6 c d s ) and, simultaneously, a
low stirring intensity, the interfacial areas of the highly viscous
sulphite solution are lower than those of the pure sulphite
system. The reasons for this are explained later. The gentle
curve gradient of the highly viscous system indicates a reduced
influence of stirring intensity at increasing liquid viscosity. This
effect is particularly distinctive for the high superficial gas velocity of uGo = 4.6 cm/s. As shown by the relatively large
horizontal distance between the curves, the influence of superficial gas velocity in the highly viscous system is more significant
than that in a system of low viscosity.

Corresponding to Fig. 5 , the volumetric mass transfer coefficient is plotted for the total power consumption in Fig. 6 . Very
striking is the significantly higher percentage decrease of the
k,a,-values when viscosity increases. This decrease is more
pronounced than that of interfacial area, and is caused by a
simultaneous influence of viscosity increase on the kL-and
a,-values.

Fig. 6 . Volumetric mass transfer coefficient as a function of total power

consumption per unit volume in stirred two-phase systems ($CMC = 0 and
0.5%).

208

Chem. Eng. Technol. 10 (1987) 204-215

5

lo5

1
-

m
S

lo3

k,

aL

lo2

D.2Ocm.

100

10-1
10
10-1

100
ptot

1 VL

kWlm3

102

kW/m3

PIo, I V L
10

dp=0.32mm, P s = 2 L 9 O 2

10

Fig. 9. Mass transfer coefficient as a function of total power consumption

(uc0 = 4.6 cm/s).

Fig. 7. Interfacial area per unit volume as a function of power consumption

per unit volume in stirred threephase systems ($CMC = 0 and 0.5%).

1o5

m
S

kL

10
PtOl

10-l

10
hot

VL

kW/m3

10

Fig. 8. Volumetric mass transfer coefficient as a function of power consumption per unit volume in stirred three-phase systems ($CMC = 0 and
0.5%).

fraction were varied as well. Since the mass transfer coefficient

can be obtained by division of the k,a,-value by the a,-value,
i.e. from two different test runs, the k,-value obviously shows
a wider scatter range than the other test values.
All mass transfer coefficient measurements can be approximated by horizontal lines; no influence of power consumption on
k,-value is present. Similar results were obtained by Calderbank [9, 101, Reith [ l l ] , Juvekar [12] and Matheron [13]. Opposite effects may possibly be responsible for the fact that the
k,-value is independent of power consumption:
1. On account of enhanced dispersion effect, an increasing
power consumption leads to smaller bubble diameters. Since

VL

Fig. 10. Mass transfer Coefficient as a function of total power consumption

( u c 0 = 1.9 c ~ s ) .

small bubbles behave as rigid spheres, the liquid film, surrounding the bubble, increases in thickness which reduces
the k,-value.
2. At the same time, liquid turbulence increases with increasing
power consumption. A decreasing liquid film thickness and
an increasing k,-value are the result.
The two effects neutralize each other in the investigated aqueous
sodium sulphite solution, with or without added CMC.

5.2 Effect of Liquid Viscosity

The measurements under variation of liquid viscosity were carried out with CMC-concentrations of $CMC = 0 to 0.75 mass-%
and, in exceptional cases, $CMC = 1 mass-%.
Fig. 11 shows the interfacial area per unit volume, for different
stirring speeds, as a function of representative liquid viscosity.

209

Chem. Eng. Technol. I0 (1987) 204-215

100
1
S

10"

2
100

10'

102

'I,

mPas

lo3

I
1o-z
10.'

Fig. 11. Interfacial area as a function of representative liquid viscosity.

For all the tested stirrer speeds, an increasing liquid viscosity

leads to decreasing interfacial areas. In agreement with Hattori
[14], up to 20 mPa s, the decrease is relatively small but it becomes larger at higher viscosities. For the correlation of results,
this implies that the relationship between the interfacial area and
viscosity cannot be described by a simple power law.
Fig. 12 shows the interfacial area per unit volume as a function
of the total power consumption for the two-phase system at a
~ 1.9 cmls. Addition of CMC
superficial gas velocity of L J G =
leads to smaller interfacial areas as a result of increasing viscosity. Several theories were developed in literature in order to
explain this phenomenon. For example, Yagi [15] points out
that, at high viscosities, a disk-shaped gas pocket forms around
the stirrer, which restrains the radial pumping capacity and formation of small bubbles. The formation of stable gas trails behind the blades of the stirrer is regarded by Hocker [16] as
responsible for the altered bubble spectrum with a larger mean

1
-

kW/rn3

10'

Fig. 13. Volumetric mass transfer coefficient as a function of total power

consumption, with viscosity as parameter (two-phase system, D = 20 cm).

diameter. Own observations are restricted to a 2 cm thick layer

on the reactor wall, since the sulphite solution, containing
CMC, was rather turbid. At high viscosities, very small bubbles
were observed, which remain in the liquid for quite a while and
also individual large bubbles which possibly form by detachment from the gas trails behind stirrer blades or from the diskshaped gas pockets. The small bubbles depleted of oxygen, due
to their long residence time, do not contribute to mass transfer
and thus to the active interfacial area. Therefore, the mean bubble size is displaced to higher values.
Fig. 13 shows the measured kLaL-values at the same operating
conditions against the total power consumption. The kLaLvalues of the 0.75 % CMC-solution could not be included in the
diagram, on account of a much too slow decrease in sulphite
concentration, which could not be analyzed by the available
Io - ~

100
Ptd 1 VL

m
S

lo3

OL

kL
102

1 0"

hot

I Vl

Fig. 12. Interfacial area per unit volume as a function of total power consumption, with viscosity as parameter (two-phase system, D = 20 cm).

10"

10-'
Ptot / V L

kW/m3

10'

Fig. 14. Mass transfer coefficient as a function of the total power consumption, with viscosity as parameter (two-phase system, D = 20 cm).

210

Chem. Eng. Technol. I0 (1987) 204-215

lo[

1
-

1
-

Io3

OL

k~O L

lo-

102

lo
1o1

10
P,I

/VL

kW/m3

10

10

Fig. 15. Interfacial area per unit volume as a function of total power consumption, at different viscosities (three-phase system, D = 45 cm).

measuring techniques. In comparison to Fig. 12, the curve

slopes are nearly the same which underlines the independence
of the kL-value of power consumption. It is quite obvious, that
the lowering of the kLaL-value with increasing viscosity is more
pronounced than in the case of interfacial area per unit volume.
For Pt,t/ VL = 4 kW/m3, the interfacial area is reduced by 55 % ,
if addition of CMC is increased from 0 to 0.5%; on the other
hand, the kLaL-value is reduced by 75 % . This indicates that the
kLaL-value is on the one hand influenced by interfacial area, for
reasons which were already mentioned, and, on the other hand,
by the mass transfer coefficient.

lo

10
Ptot I V L

10

kW/m3

Fig. 16. Volumetric mass transfer coefficient as a function of total power

consumption, at different viscosities (three-phase system, D = 45 cm).

throughout the whole concentration range. This was to be expected since the stirrer speed and the superficial gas velocity
were adjusted to match the conditions at the point of intersection
between the two- and three-phase curves in Fig. 2. Assessment
of the interactions between gas bubbles and solid particles was
based on the equation derived by Lee [7]. For a collision between a gas bubble and a particle, Lee obtained a critical Weber
number using a balance of forces, under the consideration of inertial force of the particles and the interfacial force:
Wec,it =

Ps d*

U L

= 3 .

The effect of viscosity increase on the mass transfer coefficient

is shown in Fig. 14. The decrease of the kL-value with increasing viscosity can be explained by reduced turbulence in the liquid which results in a thicker laminar liquid film around the
gas bubbles. Lower mass transfer coefficients are obtained as
a result.
Figs 15 and 16 represent the influence of liquid viscosity on interfacial area and volumetric mass transfer coefficient in stirred
three-phase systems. As in the two-phase system, an increase
of liquid viscosity leads to decreasing test values.
5.3 Influence of Particle Concentration and Particle Diameter

If the Weber number is higher than 3, the particle is capable of

dividing or deforming the bubble, so that a larger interfacial
area is generated.
The disadvantage of using this equation lies in the fact that the
relative velocity between gas bubble and particle uIel in stirred

t
D.2Ocm

The measurements should provide answers to the following

questions:

*CMC

1. From what particle concentration will the tested parameters

be significantly affected?
2. In what way are the values influenced?

vu = L . 2 r n l s , n = 1 2 0 0 l l m l n

p,

= 0

2190kglm

3
0

10

15
JIS

In some test series, the particle concentration was varied beone of these test series,
tween 0 and 30 mass-%. Fig.
- 17 presents
.
in absence of CMC. There was no change in the interfacial area

20

25

30

LO

Fig. 17. Interfacial area per unit volume as a function of solids mass fraction
( ~ M C= 0; dp = 0.32 m, D = 20 cm).

21 1

Chem. Eng. Technol. 10 (1987) 204-215

2
1
-

rn

103

aL

10

15

20
$5

10

20

15

JI,

25

30

LO

50

Fig. 18. Interfacial area per unit volume as a function of solids mass fraction
($CMC = 0.5%; d, = 0.32 mm; D = 20 cm).

reactors is unknown and cannot be estimated by approximate

methods. In bubble columns, as a first approximation, Urel can
be set equal to the bubble rise velocity. Since, compared to the
two-phase system, there is no change in the interfacial area in
Fig. 17, i.e. the Weber number is around 3, urel can be calculated by using Eq. (2). It leads to a value of urel = 0.48 m/s.
A sulphite solution with 0.5% CMC was also used at the same
stirring speed and superficial gas velocity (see Fig. 18). Up to
a solids fraction of $s = lo%, the interfacial area per unit
volume remains constant; between 10 and 30%, there is a linear
reduction. The interfacial area is reduced by 30% of the initial
value. It is possible that the viscosity increase, which reduces
liquid turbulence as well as the settling velocity of the particles,
decreases the relative velocity between bubbles and particles
since both are more strongly affected by the superimposed liquid flow. Collisions between particles and bubbles do not lead
to a deformation or division of the bubbles and, therefore, no
enlargement of interfacial area is produced.
Fig. 19 shows the measured interfacial areas under comparable
operating conditions but with particles of diameter dp = 0.088
mm added to the pure sulphite solution. Up to 5 % solids content
makes no difference to the interfacial area. The increase of the
solids fraction to 30% produces a linear increase in the interfacial area by 70%. On account of particle diameter reduced to
d, = 0.088 mm, small particles require a markedly higher velocity relative to that of the bubbles, in order to exceed the critical Weber number Wecrit = 3, according to the theory of Lee
[ 7 ] . Using Eq. (2), a value of ure1 = 0.93 m/s is calculated for
relative velocity. With the present state of knowledge there is
no explanation as yet as to why small (dp = 0.088 mm) and not
large (dp = 0.32 mm) particles lead to an increase of interfacial
area at the set operating conditions of n = 1200 l/min and UG,,
= 1.9 cm/s.
The solids concentration was varied in the reactor with diameter
D = 0.45 m, at a constant tip velocity of uu = 4.2 m/s (see

25

30

LO

Fig. 19. Interfacial area per unit volume as a function of solids fraction
(&-Mc = 0; dp = 0.088 mm; D = 20 cm).

Fig. 20). As in the case of Fig. 18, the increasing solids concentration produces a decrease of interfacial area on using the highly viscous solution.
On the one hand, the last four diagrams have shown that suspended solids influence the interfacial area only from a concentration of $s = 5 to 10% upwards and, on the other hand, it
is very difficult to predict whether the interfacial area will be
enlarged or reduced. The interfacial areas plotted for the different particle diameters in Fig. 21 as a function of stirring speed,
can be transferred to Fig. 22, provided that certain assumptions
are made.
This is described in the following for large particles (d, = 0.32
mm), based on the assumptions:
1. The interfacial area remains constant up to a solids fraction
of $s = 7.5%.
2. The interfacial area varies linearly with solids concentration.

10

15

20
JIS

25

30

LO

Fig. 20. Interfacial area per unit volume as a function of solids fraction
($CMC = 0.5%; d p = 0.32 mm; D = 45 cm).

212

Chem. Eng. Technol. 10 (1987) 204-215

loL

1
m

QL

lo3

OL

'-lo-'

10'
PI01 /VL

n-

kW/m3

10'

Fig. 21. Interfacial area per unit volume as a function of stirring speed foi
different particle diameters.

Fig. 23. Interfacial area per unit volume as a function of total power con~ 4.6 cm/s).
sumption, for different particle diameters ( U G =

Both assumptions are confirmed by Figs 17 to 20. If a linear

relationship exists between interfacial area and stirring speed,
only four experiments are required to obtain the graphs shown
in Fig. 22, for example, at n = 700 l/min and n = 1800 l/min
in the two- and three-phase systems. The resultant curves exhibit a point of intersection at about 1200 l/min in which there
is no influence of solids throughout the whole concentration
range. The line n = 1200 l/min in Fig. 22, resulting from this
point of intersection, has no inversion point. The horizontal sections of the curves (0 5 GS I7.5%) in Fig.22 were obtained
by linear interpolation between two-phase values while the end
points (GS= 30%) of the ascending or descending curve sections were obtained by interpolation between three-phase values
of Fig. 21. Combination of the horizontal sections with the end

values for GS = 30% leads to the curves presented in Fig.22

and thus to a rough qualitative estimation of the influence of inert particles on interfacial area in the overall stirring speed and
particle concentration range.
Subsequent diagrams allow a discussion of the influence of particle diameter at a solids fraction of GS =30%. Fig. 23 shows
the interfacial area per unit volume as a function of the total
power consumption per unit volume for different particle di~ 4.6 cm/s. Both
ameters, at a superficial gas velocity of U G =
particle fractions reduce the interfacial areas per unit volume
practically in the whole power consumption range. This effect
is particularly pronounced for large particles (dp = 0.32 mm).
Small particles, with dp = 0.088 mm, cause a smaller reduction
of the interfacial area. At low power consumptions, at which
stirrer speed is 1.3 times higher than the suspension rotational
speed, the suspension state of the particles is of some importance. On account of their high settling speed, large particles
are not nearly as well suspended as the small ones and they restrict the formation of a bubble layer in the bottom part of the
reactor and considerably reduce the interfacial area.
As shown in Fig. 24, the same particle fractions were also used
in a sulphite solution with 0.25 % CMC. A general reduction of
the interfacial area can be observed on addition of solids since
liquid turbulence is lowered by the viscosity increase. This
results in a reduced relative velocity between gas bubbles and
particles and in a restricted ability of the particles to deform the
gas bubbles.

lo2
0

10

15
$5

20

25 % 30

Fig. 22. Interfacial area per unit volume as a function of solids fraction and
stirrer speed.

Finally, the relationship between the mass transfer coefficient

and particle diameter should be discussed. In Fig.25, the kLvalue for a given superficial gas velocity is plotted against the
total power consumption. The kL-value of the two-phase system
is compared with the value of the stirred three-phase system,
with the addition of equivalent particle fraction to pure sulphite
solution (VL= 1.45 mPa s).

213

Chem. Eng. Technol. 10 (1987) 204-215

L ,

1
m

lo3

aL

lo2

I
10'

10'
ptot

3
lo-'
PI01

v,

100

kW/m3

I
FL(o)=3.3x10-'
/

kW/m3

10'

Fig. 26. Interfacial area per unit volume as a function of total power consumption, for different reactor diameters (uc0 = 1.9 cmis; $CMC = 0).

10'

Fig. 24. Interfacial area per unit volume as a function of total power consumption, for different diameters (uc0 = 1.9 cm/s).
lo-3,

VL

model tests (Index M) to larger reactors (Index H) requires geometrical similarity, the same material system and the same physical properties for the main construction. These three conditions are often only approximately or partly fulfilled.
Own results confirmed the condition that for the interfacial area
per unit volume as well as for the volumetric mass transfer
coefficient, a scale-up should be performed at constant specific
power consumption of the stirrer.

kL

This statement was verified for stirred two- and three-phase systems of different viscosities. Figs 26 and 27 illustrate this assertion for the system of low viscosity at a superficial gas velocity
of uGo = 1.9 cm/s. Further results on scale-up in presence of
CMC are presented in [3].

10.'

5
lo-'

p,

kW/m3

loo

'

VL

10'

Fig. 25. Mass transfer coefficient as a function of total power consumption,

for different particle diameters (uG,,= 1.9 cm/s).

The plot confirms the assumption that the mass transfer coefficient is independent of the power consumption per unit volume.
The suspended solid leads to a 10% decrease of the mass transfer coefficient when using the small particles (dp = 0.088 mm)
and to a 20% decrease on using the large particles (dp = 0.32
mm). A possible explanation is provided by the average bubble
diameter, which is reduced, especially in the presence of larger
particles. Smaller rigid bubbles form a thicker liquid film
which, in turn, results in lower kL-values.

5.5 Correlation of Data

For a statistical analysis of the results, a regression analysis was
used to determine a model for chemical engineering processes.
100
1
S

1o4

5.4 Influence of Reactor Diameter

In mixing technology, scale-up methods exploit the test results

obtained in a small-scale system to produce equivalent process
results by the application of suitable scale-up rules. Since there
are many stirring tasks (homogenization, suspension, dispersion, heat transfer), it is impossible to solve all the scale-up
problems with one scale-up criterion. A complete scale-up of

10'

10-1
Ptot

'

VL

kW/m3

10'

Fig. 27. Volumetric mass transfer coefficient as a function of total power

consumption, for different reactor diameters (UC, = 1.9 c d s ; $CMC = 0).

Chem. Eng. Technol. 10 (1987) 204-215

214
A power law, extended by a viscosity term, was used as the
regression model:

fluence on the tested parameters, are based on the assumption

that, below a concentration of $, = 0.075, the influence of the
solid is negligibly small:

1.33 ($s ULZ

Both equations were first used for the two-phase system. Since
the relationships between the tested values and liquid viscosity
are linear only in parts (see Fig. 1l ) , a function, describing the
whole viscosity range, would be far too imprecise.
Therefore, the viscosity spectrum was divided into two ranges.
The low viscosity range covers viscosities between 1.45 and 20
mPa s while the high viscosity range extends to 75 mPa s. The
results of the correlation calculation are presented in Table 1 .

1.42 ($s - 0.075)

Ranges of validity:
0.075

300 w/m3

IP R / v L

0.34 x

GS

I 0.3,

dp = 0.32 mm,
p s = 2490 kg/m3,

300 w / m 3 5
The validity range of the correlations comprises the following
parameter ranges:

PR/vL

0.34 x

m/s

5 x 1 0 - ~ Pa s

< vL

5 10000 w/rn3,
I uGo 5

5 75 x

4.6 x lO-m/s,
Pas.

5 10000 w/m3,

m/s 5 uGo

I4.6

The values aLzand k,aLz for the two-phase system must be calculated by using Eqs (3) and (4).

m/s.

It should be mentioned that the correlations are only valid for

geometrically similar plants and for non-coalescing systems.
The results in Table 1 show that the exponents ai in Eqs (3) and
(4) are the same for equivalent viscosity niveaus, which confirms that the mass transfer coefficient k , is independent of the
power consumption per unit volume. In absolute terms, the exponent c1 is always higher in the correlations of the k,a,-values
than in those of the a,-value, since the influence of viscosity on
the k,-and a,-values is incorporated into the k,a,-value.
The correlations of three-phase data are restricted to liquids of
viscosities in excess of 5 mPa s since, at these viscosities, the
addition of solids always reduces the investigated values. At low
viscosities (pure sulphite solution) the tested parameters were
reduced and also increased (see Figs 3 and 4), so that a single
correlation would be too imprecise. Both the following correlation equations, which permit a rough estimation of the solids inTable 1. Correlation of Two-phase Values.
Objective A
value
aL

2.87

0.76 0.34

-0.18

0.926

1 . 4 5 1~0 - 5 q L
5 2 0 ~ 1 0 - ~

aL

3.17

0.39 0.48

-1.11

0.920 2 0 X 1 0 - 3 5 1 ) L
c 7 5 x lo-

kLaL

5 . 9 7 x 1 0 - 4 0.78 0.36

-0.27

0.922

kLaL

1.39X10-4 0.32 0.33

-1.51

0.903 2 0 ~ 1 0 - 5 3 ,
575x

1 . 4 5 1~0 - 5 q L
5 2 0 l~o -

In order to obtain a generally valid correlation which, apart

from the solids concentration, also reflects the solids density
and particle shape, it would be necessary to conduct a large
number of additional experiments. Unfortunately, the time allotted to this investigation was insufficient.

Received: May 20, 1986 [CET 171

Symbols used
interfacial area per unit volume
concentration of cobalt ions
concentration of sulphite ions
reactor diameter
impeller diameter
particle diameter
rate constant (2nd order reaction)
volumetric mass transfer coefficient
stirrer speed
total power consumption per unit volume
stirrer power consumption per unit volume
correlation coefficient
superficial gas velocity
relative velocity between gas bubbles and
particles
stirrer tip speed
representative liquid viscosity according to [5]
liquid viscosity
density of solid
surface tension
mass fraction of carboxymethyl cellulose
mass fraction of solid
critical Weber number

Chem. Eng. Technol. 10 (1987) 204-215

Indices
D
Z

215
[7] Bohnet, M., Niesmak, G., Chon.-Ing.-Tech. 51 (1979) pp.
314-315.

three-phase system
two-phase system

[S] Midoux, N., Charpentier, J.C., Third European Conference on Mixing, April 4th-6th, 1979, Paper F 4, pp. 337 -356.
[9] Calderbank, P.H., Trans. Znst. Chem. Eng. 37(1959) pp. 173 - 185.
[lo] Calderbank, P.H., Moo-Young, M.B., Chem. Eng Sci. I6(1961) pp.
39 - 54.

References
[ I ] Shah, Y.T., Gus-Liquid-SolidReuctor Design, McGraw-Hill, New
York 1979.
[2] Kiirten, H.. Zehner, P., Lecture, GVC-Annual Meeting, Aachen,
1978.
[3] Schmitz, M., Dissertation, Univ. Dortrnund 1983.
[4] Schmitz, M., Steiff, A,, Weinspach, P.-M., Chem.-Ing.-Tech.54
(1982) pp. 852 - 853.
[5] Metzner, A.B., Otto, R.E., AIChE J. 3 (1957) pp. 3- 10.
[6] Wiedrnann, J.A., Dissertation, Univ. Dortmund 1982.

[ I l l Reith, T., Thesis, T.H. De@ 1968.

[12] Juvekar, V.A., Sharma, M.M., Chem. Eng Sci. 28 (1973) pp.
976 - 978.
[I31 Matheron, E.R., Sandall, O.C., AZChEJ. 25 (1979) pp. 332-338.
[I41 Hattori, K., Yokoo, S., Imada, O., J. Ferment. Technol. 50 (1972)
pp. 737 - 741.
[15] Yagi, H., Yoshida, F., Ind. Eng Chem., ProcessDes. Dev. 14(1975)
pp. 488 - 493.
[16] Hocker, H., Dissenation, Univ. Dortmund 1979.
[ 171 Lee, J.C., Sherrard, A.J., Buckley , P.S., Proc. Int. Symp. Fluidization, Toulouse, 1. -5. Oct. 1973, pp. 407-416.