European AI~I=II~| I~"LJ~I

Journal of J'~IJI~ I~II~

European J. Appl. Microbiol. Microbiologyand

Biotechnol. 10, 3 1 - 4 0 (1980) Biotechnology
9 by Springer-Verlag 9 1980

Hydrodynamics and Mass Transfer in an Airlift Loop Fermentor *

Ulfert Onken and Peter Weiland

Lehrstuhl fiir Technische Chemie B, Universit~it Dortmund, Emil-Figge-Stral~e, D-4600 Dortmund 50,
Federal Republic of Germany

Summary. The hydrodynamics and mass transfer behaviour of an airlift fermen-

t o t with an external loop (height 10m) has been investigated by measuring gas
and liquid velocities, gas hold-up, liquid mixing and oxygen transfer coefficients.
Liquid phase properties, i.e., ionic strength, viscosity and surface tension have
been varied by altering the fermentation media. Results are compared with those
from bubble column experiments performed in the same unit. It is shown, that
more uniform two-phase flow in the airlift leads to advantages in scale-up and
operation.

Introduction

In aerobic fermentation processes, the principal task of the reactor is to gas the culture
medium with air so that the microbes are supplied with oxygen. The gas-liquid interface
required for the transfer of oxygen can be produced in different ways. For submerged
fermentation processes which are most common, various possibilities exist (Fig. 1).
One possibility is the stirred tank in which the air is dispersed by mechanical agita-
tion. Essentially different from this traditional method of aeration with its rather high
energy requirement are the other types of bioreactors. In these the air is dispersed pneu-
matically. The most simple device of this sort is the bubble column. Combination of a
bubble column with recycling, yields the two types of loop reactors: internal loop reac:
t o t and external loop reactor. The principle of effecting liquid flow in these loop reac-
tors is called airlift, hence the name airfift reactor.
Although external loop reactors are employed by several companies for the produc-
tion of single cell protein (Gow et al. 1975; Kanazawa et al. 1975), little information
has yet been published on the behaviour of this type of reactor. Mass transfer and mix-
ing behaviour has been investigated only in two laboratory units of 3.0 m and 1.25 m
height b y Lin et al. (1976). In order to obtain basic data for reactor design we have
investigated a pilot airlift fermentor (Weiland 1978). The first part of our work reported
in this paper was carried out without biomass.

* Presented at the First European Congress on Biotechnology, Interlaken, September 25-29, 1978

0171--1741/80/0010/0031/~ 02.00
32 U. Onken and P. Weiland

G G G G

T
~ o o l
o ~ o I

o o l
o o
~' o o I

0 Ol

0 o I

o o I

.LL&.3J
G

Loop Reactors Fig. 1. Reactors for aerobic

Stirred Vessel BubbleColumn Inner Loop Outer Loop fermentations

l gas I gas
4-

direction of flow 1/

J -:
gas
Fig. 2. Schematic diagram of the airlift
c --/ loop fermentor
Hydrodynamics and Mass Transfer in an Airlift Loop Fermentor 33

Materials and Methods

Our investigations were performed in a tubular glass fermentor (Fig. 2) with a 100 mm
diameter up-flow column (a) and a 50 mm diameter down-flow column (b). The total
height of the apparatus is over 10 m. Gas is fed in above the lower end of the riser. The
dispersed gas reduces the overall density in the riser compared to that in the ungassed
down-flow column. Thus liquid circulation is induced in the loop. By means of a control
valve (c), circulation of the liquid can be reduced. When this valve is totally closed, the
up-flow column is operated as a bubble column. Thus in both types of reactors the gas is
dispersed pneumatically and the air-lift and the bubble column can be compared in the
same unit.
Liquid velocity is measured by a magnetic flowmeter (d) in the down-flow column.
The active height of the gas-liquid dispersion is 8.5 m. In these experiments either a
stainless steel porous plate (mean pore diameter 150/~m) or a stainless steel perforated
plate (diameter of holes 0.5 m m ) were used as aerators.

Results

Liquid Velocity
Liquid velocity is the characteristic parameter which distinguishes air-lift and bubble
columns. Figure 3 shows the superficial liquid and gas velocities in the up-flow column
for liquids with different physical properties. Whereas continuous bubble columns are
generally operated at liquid throughputs corresponding to overall velocities of less than
5 cm/s, liquid velocities measured in the airlift reactor are up to one order of magnitude

0,5 [ I I
porous plote
o woter
~' 0,4 _

E sodium chloride t woter

[] 0,1m [=0,1
i
>-
9 1,0m I=1,0

___ 0,3 2-proponol / water

-S
z~ 1,65% o = 6 0 ' 1 0 -3 N/m
O
J
9 4.90% G = 5 0 1 0 -3 N/m /0 j
LU
socchorose / woter
//.,~5 q
>
0 3L,5 % ~1 = 4 mPo
D 0 51,8% ~ = 16 rnPa

0
0,2
_J

_.J
<
0
la

W
n

U) /

0,1
0,001 0,01 0,1 0,16

SUPERFICIAL GAS VELOCITY [m/sl

Fig. 3. Superficial liquid velocity as function of superficial gas velocity

34 U. Onken and P. Weiland

higher. The straight lines running parallel in the double logarithmic diagram show that
for all liquids investigated in this work, the velocity varies approximately with a power
of 0.4 of the superficial gas velocity.
A higher viscosity lowers the liquid velocity. Increasing the viscosity by a factor of
10 by dissolving saccharose in water decreases liquid velocity by about 10%. The liquid
velocity is not affected by the two types of gas distributors employed in our investiga-
tions. When the difference in hydrostatic pressure between the two columns is known,
the liquid velocity can be determined exactly from the pressure drop in the pipes.

Gas Hold-Up
Figure 4 shows how the gas hold-up varies with the liquid velocity, when the gas veloci-
ty is altered. It shows the strong influence of liquid velocity on two-phase flow; gas.
hold-up decreases markedly with liquid velocity. Maximum gas hold-up is obtained for
stagnant liquid, which means bubble column operation. The minimum gas hold-up was
about one third of the maximum values and is reached at full circulation of the liquid.
The low gas hold-up in the airlift column is caused by an acceleration of the rising
bubbles in the upflowing liquid. The resulting decrease in gas hold-up can be determined
by calculating the slip velocity between gas and liquid flow (Richardson 1954).
The uniform direction of gas and liquid flow in the airlift does no only diminish gas
hold-up, but also leads to a totally different effect of liquid phase properties and of the
type of gas distributor on gas hold-up, compared to bubble column operation. As can be
seen from Fig. 5, variations in viscosity, surface tension and ionic strength show little
influence on gas hold-up in the airlift column, in contrast to bubble column operation,
i.e., without circulation. In addition a regime of unstable hold-up appearing in the bub-
ble column with saccharose solutions does not occur in the airlift.

25-

porous pLote
2 0 84 0,1 m sodium cNoride
super ficiel gos velocity
[m/s]
O 0,02
O_
D 15 z~ 0,03
i v 0,04
s [3 0,05
6 O ~ O 0,06
I
O O

"v ~o--_ !"~ I

5 ~-~~ -

0 4 -
0 0,1 0.2 0,3 0.4
SUPERFICIAL LIQUID VELOCITY [m/s
Fig. 4. Gas hold-up as function of superficial liquid velocity
Hydrodynamics and Mass Transfer in an Airlift Loop Fermentor 35

30
25
20

15

10

~5
&
6
T
L/3
(.9

0,5 4 ~ - -
0,001 O~ 0.1 0,16

SUPERFICIAL GAS VELOCITY [m/s]

Fig. 5. Gas hold-up as function of superficial gas velocity

The difference in behaviour between the two types of reactors is caused mainly by the
high velocity of the liquid in the airlift column with its lower gas hold-up and corres-
pondingly higher average distances between bubbles. This results in a decrease in coales-
cence frequency. This reduction of coalescence is enhanced by a more uniform distribu-
tion of bubbles on column cross section in the airlift compared to bubble columns
(Malnes 1966); Serizawa 1975). Therefore the addition of compounds which reduce
coalescence such as alcohols or electrolytes only have a small effect on gas hold-up in
the airlift reactor whereas in bubble columns the addition of these compounds increases
gas hold-up more than two-fold. In addition, the decrease in coalescence by liquid flow
enlarges the region of stable bubble flow in the airlift. The beginning of large bubble
formation and "slugging" was observed at liquid velocities three times higher than
during bubble column operation. In concentrated saccharose solutions and therefore
higher viscosities, formation of spherical cap bubbles and slugs is promoted leading to a
considerable decrease in gas hold-up at low gas through-puts, corresponding to superfici-
al gas velocities of about 0.02 cm/s. At higher gas velocities, characteristic for the air-
lift reactor, lower gas hold-up and more uniform bubble distribution largely prevents
formation of bubble swarms. Because of coalescence restraining and high bubble veloci-
ties in the airlift reactor, the type of gas distributor has little influence on gas hold-up in
comparison to bubble columns.

Mass Transfer
The mass transfer is much more important than gas hold-up when reactor efficiency and
design are assessed. In order to obtain the necessary data we measured oxygen absorp-
tion from air. A dynamic method was used to determine the response of the system to
a step change of the feed gas from nitrogen to air. For this purpose, the oxygen con-
36 U. Onken and P. Weiland

I~LO,"102 - - I
Is -1] porous, plate
9 o water
sodium chloride 0.1 m I =0.1
9 sodium chloride 0.4 m I : 0.4
8 ~" 2-propanol 1,65 % a : 60,10"3NIm ___
~' 2 - p r o p a n o t /.,90 % : 50.10"3 NIm
I~ socch~rose ~.70 ~ ~ : 4 rr', Po,s
7 II socch~ros,e 51,80% xl : 16 mPa,s
i

i
L _ A / ~AI

I
Jq

(? J

Fig. 6. Volumetric mass transfer

0#5 0.I 0,15 coefficient ( ~ a ) for oxygen in water
SUPERFICIAL GAS VELOCITY [m/s] and in aqueous solutions

centration in the liquid phase was measured as a function of time by means of pola-
rographic electrodes. Because of the negiegible small dead times of the electrodes,
the experiments could be evaluated on the basis of the model of Aiba and Huang
(1969), taking into account the time lag in electrode response. By using a procedure
described in detail by Weiland (1978), the volumetric mass transfer coefficients ~L a,
defined by Eq. (1), were obtained from the experimental results.

n_.._= *
VR ~La(CL _CL ) (1)

Volumetric mass transfer coefficients which have been determined in this way are
plotted against the superficial gas velocity in Fig. 6. Curves in the diagram are for
water, sodium choride solutions, isopropanol and saccharose. The results for pure
water agree well with the volumetric mass transfer coefficients measured by L i n e t al.
(1976). The enhancement of mass transfer by sodium chloride and b y isopropanol
is caused b y coaiescence restraining. Increasing the electrolyte concentration above
0.4 mol/1 had no further effect. The higher viscosity of the saccharose solution causes
a decrease in the mass transfer.
Addition of compounds which reduce bubble coalescence, increase the volumetric
mass transfer coefficients by a factor of about two. In bubble columns, mass transfer
is enhanced b y compounds which reduce bubble coalescence about sixfold (Oels et
al. 1976) in comparison to water. The overall difference of volumetric mass transfer
coefficients between the airlift reactor and the bubble column is parallel to the lower gas
hold-up in the airlift. With regard to the influence of viscosity (see curves for saccharose
Hydrodynamics and Mass Transfer in an Airlift Loop Fermentor 37

solutions, Fig. 6), there is a considerable decrease in the volumetric mass transfer coef-
ficients with increasing liquid viscosity. It should be noted that this effect is distinctly
greater than the variation of the diffusion coefficient with viscosity.
The rather small values of the volumetric mass transfer coefficients are not disadvan-
tageous, however, because the airlift reactor will be operated at gas velocities at least
2 or 3 times higher than bubble columns. In contrast to bubble columns, it is there-
fore possible to supply sufficient oxygen to the liquid phase even for very high reactors.

110
Da •

[crn2/sf

100

90
porous plate

1.65 % 2- proponol

VSS=2.5 cm/s

80
0,1 0,2 0,3
SUPERFICIAL LIQUID VELOCITY [m/s]

Fig. 7. Effect of liquid velocity on axial dispersion coefficient

rt
Meosuring
Cell ~ '
Trclcer-
Input

C**

4 TIME

Fig. 8. Pulse propagation of tracer input

38 U. Onken and P. Weiland

Mixing Behaviour
A knowledge of longitudinal mixing of the liquid phase is required especially for conti-
nuous operation. Longitudinal mixing and the mixing times required to obtain a de-
finite degree of homogenisation were investigated.
A pulse technique was employed, using NaC1 as tracer and measuring electrical
conductivity. Figure 7 shows that high liquid velocities are of minor influence on the
axial dispersion coefficient. An increase of liquid velocity from 0.05 m/s to 0.3 m/s
diminishes the axial dispersion coefficient of the liquid b y about 10%. Therefore,
liquid phase longitudinal mixing in the airlift is comparable to that in bubble columns
(Deckwer et al. 1973). For high gas throughputs the axial dispersion coefficient in the
100 mm diameter column is in the range of 100 to 150 cm 2 Is, depending on the
physical properties of the liquid phase and on the gas velocity. Measurements with
different solutions have shown that at high gas velocities the axial dispersion coefficient
of the liquid decreases with diminishing surface tension and increasing liquid viscosity.
The second parameter which describes mixing behaviour is the mixing time. It is
used as a basis for comparing various reactor types as well as a parameter for scaling up.
It is defined as the time, which is required to homogenize the concentration impulse
to yield a defined degree of inhomogeneity. Figure 8 shows pulse propagation for a
definite location of the probe. The inhomogeneity (i) is defined as the relative de-
viation of the actual maximum concentration (6) from the mean value (coo) which
is reached after perfect mixing (Lehnert 1972):
c-coo
i = - - (2)
coo

The time elapsed from tracer input to a definite degree of inhomogeneity is the mixing
time (tM). The time for one circulation (circulation time, tc) is defined as the interval
between two adjacent concentration peaks. In Fig. 9 mixing times (t M) for two degrees

700
D 2-proponol t65 %
[] CMC RIT 5000 0,25 %
600 O wtater

t~4, t c
Is/

400

~oD-'M Irlhomogeneity 10%

200

[3 ~ O - tN
O- [/~homogeneity 50%

0
I tc
i

0 0.02 0,0/* 0,06

SUPERFICIAL GAS VELOCITY [m/s]

Fig. 9. Influence of gas throughput on mixing time and on circulation time

Hydrodynamics and Mass Transfer in an Airlift Loop Fermentor 39

of inhomogeneity (i) are shown together with circulation time (t c) as a function of

superficial gas velocity.
The results clearly demonstrate that there is no appreciable difference in mixing
time for liquids differing in physical properties within the ranges of variation used in
this work. This is not the case, however, for the sugar solutions of higher viscosity, not
shown in Fig. 9. The influence of gas velocity on mixing time is comparable with the
results reported b y Lin et al. (1976). On the whole it can be stated, that the mixing
time decreases considerably with increasing gas throughputs for gas velocities smaller
than 0.04 m/s, whereas for higher gas velocities this effect becomes unimportant. In
addition, mixing processes in the airlift reactor are rather slow. From a comparison
of mixing times and circulation times it can be seen that for gas velocities higher
than 0.03 m/s the degree of inhomogeneity is only dependent on the number of cycles.
After 5 to 6 cycles, inhomogeneity is less than 10%.

Conclusion

The airlift reactor is a very interesting type of bioreactor. It has lower volumetric mass
transfer coefficients than bubble columns. Higher rates of aeration are ~equired, how-
ever, therefore airlift fermentors should be designed with larger heights than bubble
columns, in order to obtain high oxygen utilization. Liquid velocities will, therefore,
be rather high and this leads to improved heat transfer.
Investigations of the airlift reactor with fermentation systems are in progress in a
pilot installation adapted for this purpose.

Acknowledgement. Thanks are due for the Fonds der Chemischen Industrie for financial support,
and to the Studienstiftung des deutschen Volkes for granting a scholarship (P.W.)b

Nomenclature
a Specific interfacial area per volume of dispersion (m 2/m 3 )
c Local concentration of tracer (kmol/m 3 )
coo Concentration of tracer at infinite time (kmol/m 3 )
CL Concentration of oxygen in the liquid bulk (kmol/m 3)
c; Concentration of oxygen in the interface (kmol/m 3 )
Dax Axial dispersion coefficient (cm 2/s)
I Ionic strength (kmol/m 3)
i Inhomogeneity [defined in Eq. (2)]
Rate of oxygen transfer (kmol/s)
tc Circulation time (s)
tM Mixing time (s)
V R Volume of gas-liquid dispersion (m 3 )
VSG Superficial gas velocity in up-flow column (m/s)
40 U. Onken and P. Weiland

Greek Letter Symbols

/3L Oxygen transfer coefficient (m/s)

r/ Dynamic viscosity (m Pa s)
o Surface tension (m N/m)

References

Aiba S, Huang SY (1969) Chem Eng Sci 2 4 : 1 1 4 9 - 1 1 5 9

Deckwer W-D, Graeser U, Langemann H (1973) Chem Eng Sci 1223-1225
Gow JS, Littlehailes JD, Smith SRL (1975) SCP-Production from Methanol. In:
Tannenbaum SR, Wang DIC (eds) Single-cell protein II. MIT-Press, Cambridge,
pp 370-384.
Kanazawa M (1975) The production of yeast from n-paraffins. In: Tannenbaum SR,
Wang DIC (eds) Single-cell protein II. MIT-Press, Cambridge, pp 438-451
Lehnert J (1972) Verfahrenstechnik 6 : 5 8 - 6 2
Lin CH, Fang B S, Wu C S, Fang HY, Kuo FF, Hu CY, (1976) Biotechnol Bioeng 18:
1557-1572
Malnes D (1966) Inst Atomenergy Kjeller Res Establishment, Rep 110
Oels U, Schiigerl K, Todt J (1976) Chem Ing Technol 48 : 73 and MS 316/76
Richardson JF, Zaki WN (1954) Trans Inst Chem Engr 32 : 35-53
Serizawa A, Kataoka I, Michiyoshi I (1975) Int J Multiphase Flow 2 : 235-246
Weiland P (1978) Doctoral Thesis, University Dortmund

Received November 23, 1979