Documente Academic
Documente Profesional
Documente Cultură
By Jos C. Merchuk*
Department of Chemical Engineering, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel
Fluid-dynamic Characterization
The largest amount of publications related to airlift reactors during
recent years, report on measurements of fluid dynamic characteristics
and mass transfer rates. Some of them are valuable confirmations or
extensions of prior knowledge, and some of them report experimental
results in modified airlift design, which usually cannot be represented by
prior correlations. In any case, the need for collecting and unifying all
those experimental results still exists. We suggested years ago a unified
format for airlift data collection and correlation (Har-Noy et al., 1997).
Al-Masri and Abasaeed (1998) did extensive measurements of liquid
velocity and riser holdup in a series of external loop airlift reactors, and
proposed empirical correlations that fit their data better than other
proposed correlations. Still there is here a lack of a massive bank of data
for validation of those correlations.
One of the most basic characteristics of gas/liquid dispersions is the
bubble size. Couvert et al. (1999) have presented a series of very careful
measurements of liquid velocity, riser and downcomer gas holdup and
Sauter-mean bubble diameter. The last measurements were related to
the gas pressure in the tubular membranes that acted as spargers, and
here. The interesting point in this work is the use of the graphs
displaying gas slip velocity versus overall fluid velocity in order
to make a quantitative determination of the transition from
homogeneous gas flow, where the slip velocity decreases with
the total flow, to heterogeneous gas flow where it increases
with the total flow. Figure 1 shows the transitions for three
different diameters of the downcomer, each giving a different
resistance to liquid circulation. Similar results were found by
partially closing the valve in the downcomer. The same group
attacked the problem of regime identification from a quite
different angle (Vial et al., 2001), proposing the use of
auto-correlation functions of wall-pressure fluctuations for
identification of flow regimes in airlift reactors. This method
seems to be experimentally simple and very promising for this
purpose.
Van Benthum et al. (1999c) studied the transitions between
gas recirculation regimes in airlift reactors that they define as
follows: Regime 1 (no gas recirculation), Regime 2 (stationary
bubble front in the downcomer) and Regime 3 (gas recirculation).
It is apparent that Regime 2 has little interest from an industrial
point of view. The gas that stays practically stationary in the
downcomer will not have much influence on mass transfer,
since it will fast become depleted of oxygen, or other
components of interest. The authors made a systematic study of
the effect of suspended solids and liquid throughflow, and
managed to define criteria for the transition from Regime 2 to
Regime 3. The transition depends on the suspended solids and
on the throughflow of liquid.
See et al. (1999) made a careful study of the effect of drag
and frictional losses on the hydrodynamics of airlift reactors, to
elucidate in which measure the use of available methods and
correlations are applicable to airlift reactors. They recommend a
careful evaluation of the frictional losses in all the elements of
the circuit, including straightening vanes, gas liquid separator,
and riser-wall friction. On the other hand they found that the
use of correlations for fully developed single-phase flow in the
downcomer (their external loop reactor has no gas recirculation)
does not introduce much error.
Models
Airlift reactors especially the fluid dynamics of airlift reactors
have a strong appeal and there are many researchers that
have tackled the task of describing the flow of liquid, gas
and solids in these bioreactors. In Table 1, some of the models
that have been published during the last six years are
presented. Twelve of them model gas/liquid (G/L) systems,
and eight attempt the representation of three-phase,
gas/liquid/solid (G/L/S) systems. Some overlapping exists, since
several models are presented in a general form that fits both
G/L and G/L/S systems.
The table shows clearly that the liquid velocity and the gas
holdup in the riser are seen as the main variables studied and
appear in all the proposed models. Indeed, those variables are
closely related, since the difference in gas holdup between riser
and downcomer is the only driving force for liquid circulation in
the system. This has been clearly stated by Heijnen et al. (1997),
which expressed the pressure difference per unit height
required as:
DP
= (e Gr - e Gd ) - (e Sr - e Sd )(r S - r L )
gH
(1)
Table 1. Selected models for airlift fluid dynamics and mass transfer
Source
Heijnen et al.(1997)
van Benthum et al. (1999a)
Garcia-Calvo et al. (1999)
Hwang and Lu (1997)
Freitas et al.(1999)
Mudde and Van der Akker (2001)
Camarasa et al.(2001a)
Camarasa et al.(2001b)
Mousseau et al. (1998)
2Phase
3Phase
Gas
recirculation
VL
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
eGr
eGd
x
x
x
x
x
x
x
x
x
x
x
x
x
Cockx et al.(1997)
Couvert et al.(2001)
Tobajas et al.(1999)
Camacho-Rubio et al.(2001)
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
(2)
CFD
Remarks
3 regimes
ALR and
extension
kLa
es constant
kLa,
ammonia
x
x
eG
x
x
eSd
eSr
x
Axial
profiles
x
kLa
O2 axial
profiles
x
x
x
b=
1
>1
k L at c
(3)
both riser and downcomer, and a flow map of the liquid that
includes radial profiles of velocity. Mudde and Van Den Akker
(2001) carried out two- and three-dimensional simulation of
an airlift, trying to minimize the use of ad-hoc closure
terms. Surprisingly, they do not find clear advantage in
the three-dimensional model, since both approaches coincide
with experimental measurements of gas holdup. The two
dimensional model predicts a strong influence of the
sparger geometry, and seems to be better at low gas superficial
velocities.
Camarasa et al. (2001a, and b) developed a model for
external-loop airlift reactors (no gas recirculation) based on a
careful momentum balance, using the drift-flux model (Zuber
and Findlay, 1965) for gas holdup prediction. The model fits
fairly well their experimental results of mean velocity and gas
holdup in the riser, as well as axial variations of the local gas
holdup along the riser. They combined this fluid dynamic
information with a cells-in-series model for mass transfer, which
also addresses the mixing behavior of the phases.
Couvert et al. (2001) present a simple model for the prediction
of liquid velocity and gas holdup in rectangular airlift reactors of
different scales. They do consider gas recirculation, and their
model predicts gas holdups in both riser and downcomer,
which are close to the measured values.
The fundamental and solid model presented a decade ago by
Young et al. (1991) for an external airlift reactor (1991) was
upgraded (Sez et al., 1998), adding to the initial model the
effect of gas buoyancy forces in the gas. Based only on the
physical properties of the gas and liquid phases, the reactor
dimensions and the gas input, the model predicts gas and
liquid velocity and gas holdup along the riser for bubbly flow.
Total disengagement of the gas at the top is assumed.
The seminal model by Ho et al. (1977) has been an inspiration
and a basis for much improved models that represent the airlift
reactor as a sequence of perfectly mixed cells (Camarasa et al.
2001b; Steiff et al., 1997; Orejas, 1999). Those models are
convenient for handling biological or chemical reactions. The
simplification of the mathematical treatment may sometimes
outweigh the unrealistic stepwise change of the variables
along the reactor that is obtained. Steiff et al. (1997) produced
a considerable amount of experimental data that can be used
to calibrate the model parameters. Of special interest, since
these data are scarce in the literature, are the data of gas
recirculation, measured following the technique proposed by
Siegel et al. (1986).
Shechter et al. (2002) presented a three-phase model for the
fluid dynamic description of an airlift unit, part of an overall
process (referred to below). The model is based on a
momentum balance over riser and downcomer, and a series of
macroscopic equations that describe the continuity of liquid
and solid as it passes from the riser to the downcomer. Figure 2
shows the influence of solid loading on the variables of the
system: liquid and solid velocity in the riser (the corresponding
values in the downcomer are not shown to avoid overcharging
the figure), gas and solid holdup in the riser and solids holdup
in the downcomer. The model predicts a small increase in both
gas holdup and solids velocity as the amount of solids increases
at constant gas superficial velocity. The solids holdup in both
the riser and the downcomer increases as expected, and the
increase in the downcomer is much sharper than in the riser.
The most affected variable is the liquid velocity, ULr, which
decreases almost to half due to the change in solids loading.
The Canadian Journal of Chemical Engineering, Volume 81, June-August 2003
Figure 2. Influence of solids loading on riser gas holdup (er ), riser and
downcomer solids holdups (eSr and eSd), liquid and solid velocity in the
riser (ULr and USr), in a rectangular section airlift reactor of a water
treatment system, as predicted by the model by Shechter et al. (2002).
The superficial gas velocity was 0.006 (m/s), the area ratio (Ar/Ad)=
2.85, the solid density 950 (m/s) and the liquid density 955 (m/s). The
line-identifiers are: = er (-) ; = eSr (-) ; = eSd (-) ; x = ULr (m/s) ; *
= USr (m/s).
Figure 3. Influence of particle density on riser gas holdup (er ), riser and
downcomer solids holdups (eSr and eSd), liquid and solid velocity in the
riser (ULr and USr), in a rectangular section airlift reactor of a water
treatment system, as predicted by the model by Shechter et al. (2002).
The superficial gas velocity was 0.006 (m/s), the area ratio (Ar/Ad)=
2.85, the solid density 950 (m/s) and the liquid density 955 (m/s). The
line-identifiers are: : = er (-) ; = eSr (-) ; = eSd (-) ; x = ULr (m/s) ; *
= USr (m/s).
from the gas to the culture. The solution that has been almost
universally adopted for this problem which attains to citric
and other organic acids, antibiotics, etc. is to find the
conditions under which the biomass takes the form of fungal
pellets. The advantage in gas-liquid transfer rate, because of the
decrease in viscosity, usually outweighs the added transfer
resistance stemming from the intraparticle diffusion of oxygen.
But the formation of pellets in optimal size and compactness is
a very complex matter. Metz and Kossen (1977) pointed out
the multiplicity of variables and the difficulties of a priori prediction of the operation conditions required. Not much has been
advanced in this matter since. Jin et al. (1999) found empirically
the optimal gas flow rate in their 4.5 L ALR. It is worthwhile to
note that they seem to have been able to scale up these
conditions to their pilot-plant 160 L ALR.
Lazarova et al. (1997) studied experimentally the fluid
dynamics and the performance for wastewater treatment of an
split-vessel airlift with a rectangular section. They studied
carefully the influence of suspended solids on gas holdup both
in the riser and the downcomer, as well as the influence of
the ratio of riser to downcomer cross sectional areas on
liquid velocity. They compare the experimentally measured
velocities for different reactor heights without proposing any
correlation. Their measurements of mass transfer rate do not
reveal important changes with respect to those obtained in
water once the biofilm is developed. The main aspect stressed
by the researchers is the capacity for nitrification observed in
various stages.
Many applications of ALR have been reported in processes
where the point of interest here is simply that the process,
which can take place in a conventional stirred tank, can be run
using an ALR as well, with the consequent savings in energy
requirements, etc. For example, the use of Aspergilius niger for
textile wastewater (biological decoloration) was reported by
Assadi and Jahangiri (2001). Campos et al. (2002) used an ALR
in a combined (microfiltration and biological) treatment of
oilfield wastewater treatment. They obtained satisfactory results
in TOC and COD reduction in a continuous process, using an
Design Modifications
We have commented already on some interesting airlift design
modifications, as the BASE shown in Figure 4, where an
extension is added to the reactor providing a region of fluid and
solids flow that is completely different from the usual airlift flow
regions, and still can be controlled, mainly by the headspace
pressure in the main reactor. The GLAD (gas lift advanced
dissolution), a completely different concept, is a method for
sequestration of low purity CO2 emitted by thermal power
plants (Kosugi et al., 2001). The dissolution tube is approximately
200 m long, and the drainpipe reaches more than 1000 m in
depth. In fact, this is not a gas lift reactor, since this is a
once-through system, and recirculation, a basic characteristic of
airlift reactors, does not exist. It is therefore, a gas-lift pump in
Figure 7. Extractive bioconversion in a four-phase external-loop airlift bioreactor. In the riser, gas and a solvent are injected. Both gas and solvent
are completely separated at the top and only the aqueous medium and the suspended alginate beads circulate. From Sajk and Vunjak-Novakovic
(2000), with permission.
Bioprocess Applications
While airlift reactors have become very popular in research
institutes, the stirred tank reactor remains as the undisputed
leader in the realm of industrial bioreactors. That is possibly the
reason many publications that compare airlift and stirred tank
reactors. Kim et al. (1997) compared the production of
b-glucosidase by Aspergillius niger in various bioreactors, and
reported that the best results correspond to an airlift reactor.
Figure 8. The Helical Flow Promoters: The case of HFP located at the
top of the downcomer. The grey arrows indicate the flow of the liquid.
top and do not return to the bulk of the liquid. They report to
have doubled the final concentration of the plant cells, and also
the amount of secreted proteins to the medium.
Visnovsky et al. (2003) reported a comparison of a concentric
airlift reactor and a stirred tank bioreactor, for the growth of
UFL-Ag-286 insect cells, and for the production of Anticarsia
gemmatalis multiple nucleopolyhedrovirus (AgMNPV). While no
relevant differences could be observed in the doubling time and
viability of the insect cells, important differences were found in
the kinetics of adsorption of AgMNPV-NOVs (non-occluded
virus) on the insect cells, and in the rate of production of
AgMNPV-OVs (occluded virus). The NOVs were more
quickly and efficiently adsorbed on the cells in the airlift reactor
than in the stirred tank reactor. The onset of OVs production
was earlier in the airlift reactor than in the stirred tank reactor
(Figure 10). In addition, the titers of OVs obtained in the airlift
reactor were slightly higher than those obtained in the stirred
tank reactor for both baculovirus progenies. These results
indicate clearly the influence of reactor fluid dynamics on the
performance of the process.
10
Figure 12. Cyclic light history of cells in the Air Lift Reactor, according
to the model by Wu and Merchuk (2002). Illuminance is taken a nil in
the riser, a mean value is taken for the separator, and in the riser it
varies with the radius of the column.
Figure 14. Simulation of the effect of draft tube height on cell growth.
The initial cell concentration was x0 = 8 106 (cel/ml), the gas superficial velocity was 0.00331 (m/s), and the diameter of the column 0.2
(m) and the light intensity I0 = 250 mEm2s1. Wu and Merchuk
(2003).
Figure 15. Simulation of the effect of the ratio of cross sectional areas
Ar/Ad on the cell growth. The initial cell concentration was x0 = 8 106
(cel/ml), the gas superficial velocity was 0.00331 (m/s), The height of
the column was 1 (m), the diameter of the column 0.2 (m) and the
light intensity I0 = 250 mEm2s1. Wu and Merchuk (2003).
11
Selection Strategy
A point that usually appears at the early stages of the development
of a system where an ALR is considered is the selection of the
type of bioreactor. Should an external circulation ALR be
adopted, or rather an internal circulation ALR?
In the case of bioreactors (except biological leaching of
minerals) fluidization is not a serious problem. The solids have
usually a density that is not far apart of that of water, and
fluidization is attained at relatively low liquid velocities.
Therefore, the higher liquid velocities that can be attained in
external loop ALR are not a decisive point. On the contrary,
excessive liquid velocity may translate into damages to the cells
due to shear stress. The main difference remaining is therefore,
the presence of gas in the downcomer. In processes where
the oxygen concentration in the liquid may be depleted in
the downcomer, recirculation of gas will alleviate the
problem. Obviously, a wise consideration of the geometric
design, in order to shorten the residence time in the
downcomer will also contribute.
Our present understanding of the dynamic behaviour of the
ALR system indicates that rather than maintaining the classic
classification of those reactors into internal and external
recirculation devices, attention should be given to the design of
the top of the reactor where gas separation takes place. Since
most of times the main element to take into consideration is gas
disengagement, the elements must be balanced will be: a) The
time that a bubble must spend in the gas separator before
entering the downcomer (function of the liquid velocity in the
gas separator and the length of path of the liquid element), and
b) The ascending velocity of the bubble in the liquid. A clear
example would be a split-vessel ALR, that may be seen as an
external recirculation reactor on one hand, but offers also a
short residence time in the gas separator which provides the
basic characteristics of an internal recirculation ALR.
It should be taken into account that this aspect constitutes a
serious scale-up problem. As the diameter, or equivalent
diameter of the reactor increases, the path of the liquid element
traveling from riser to downcomer becomes longer and the
chances of bubble disengagement increase too. Ingenuity in
the geometric design has an important role here.
Recapitulation
The distinctive characteristics of airlift reactors are conferred by
the fluid dynamics of the gas/liquid or gas/liquid/solid systems
circulating. These characteristics are usually expressed as gas
holdup, liquid and solid velocities and mass transfer rate. It is
important for the design engineer to recognize whether there is
a need to consider those variables separately for each of the
distinct zones of the reactor (structural model). The decision on
this can be made comparing the circulation time and the mass
transfer characteristic time (kLa1). In any case, only a correct
understanding of the behaviour and interconnection of riser,
separator, downcomer and bottom of the reactor will allow the
reliable scale up from the laboratory to pilot or industrial size.
Several models that allow the simulation of the fluid dynamics
of airlift reactors have been presented during the last years. It
12
Nomenclature
Ar
Ad
b
D
g
H
I
I(t)
k
kLa
r
P
t
tc
T
ur
uz
ULr
UGr
USr
uq
VGL
VLr
VLd
x1
x2
x3
x0
xf
Dx
z
Greek Symbols
eG
eGd
eGr
eS
eSd
eSr
q
rL
rS
Dr
Subscripts
d
r
s
downcomer
riser
separator
Abbreviations
ALR
BOD
CFD
COD
G/L
G/L/S
HFP
NOV
OV
PFD
PSF
STR
TOC
airlift reactor
biological oxygen demand
computational fluid dynamics
chemical oxygen demand
gas/liquid
gas/liquid/solid
helical flow promoter
non-occluded virus
occluded virus
photon flux density
photosynthetic factory
stirred tank reactor
total organic carbon
References
Al-Masri, W.A. and A.E. Abasaeed, On the Scale-up of External Loop
Airlift Reactors, Newtonian Systems, Chem Eng. Sci 53, 40854094
(1998).
Assadi, M.M. and M.R. Jahangiri, Textile Wastewater Treatment by
Aspergilius niger, Desalination 141, 16 (2001).
Bakker, W.A.M., P. Kers, H.H. Beefting, J. Tamper, and C. D. de Gooijer,
Nitrite Conversion by Immobilized Nitrobacter agilis in an Airlift
Loop Bioreactor Cascade: Effects of Combined Substrate and Product
Inhibition, J Ferm. Bioeng. 81, 390393 (1996).
Bendjaballah, N., H. Dhaouadi, S. Poncin, N. Midoux, J.M. Hornut, and
G. Wild, Hydrodynamics and Flow Regimes in External Loop Airlift
Reactors, Chem. Eng. Sci. 54, 52115221 (1999).
Camacho Rubio, F., J.L. Garcia, E. Molina, and Y. Chisti, (2001). Axial
Inhomogeneities in Steady-State Dissolved Oxygen in Airlift
Bioreactors, Predictive Models, Chem. Eng. J. 84, 4355 (2001).
Camarasa, E., E. Carvalho, L.A.C. Meleiro, R. Maciel Filho, A.
Domingues, G. Wild, S. Poncin, N. Midoux, and J. Bouillard, A
Hydrodynamic Model for Air-Lift Reactors, Chem. Eng. Proc. 40,
121128 (2001a).
Camarasa, E., E. Carvalho, L.A.C. Meleiro, R. Maciel Filho, A.
Domingues, G. Wild, S. Poncin, N. Midoux, and J. Bouillard,
Development of a Complete Model for an Air-Lift Reactor, Chem.
Eng. Sci. 56, 493502 (2001b).
Camarasa, E., E. Carvalho, L.A.C. Meleiro, R. Maciel Filho, A.
Domingues, G. Wild, S. Poncin, N. Midoux, and A. Bouillard,
Complete Model for Oxidation in Airlift Reactors, Computers &
Chem. Eng. 25, 577584 (2001c).
Campos, J.C., R.H.M. Borges, A.M. Olivera Filho, R. Nobrega, and
G.L. Sant Anna Jr., Oilfield Wastewater Treatment by Combined
Micro Filtration and Biological Precess, Water Research 36, 95104
(2002).
Chisti, Y., Airlift Bioreactors. Elsevier Applied Science, New York, NY
(1989).
Chisti, Y., Pneumatically Agitated Bioreactors in Industrial and
Environmental Bioprocessing: Hydrodynamics, Hydraulics and
Transport Phenomena, Appl. Mech. Rev. 51, 33112 (1998).
Cockx, A., A. Line, M. Roustan, Z. Do-Quang, and V. Lazarova,
Numerical Simulation and Physical Modeling of the Hydrodynamics
of an Airlift Internal Loop Reactor, Chem. Eng. Sci. 52, 37973793
(1997).
13
14