See discussions, stats, and author profiles for this publication at:

https://www.researchgate.net/publication/228033687

Bubble Columns

Chapter · June 2000

DOI: 10.1002/14356007.b04_275

CITATIONS READS

4 2,640

2 authors, including:

Matthias Kraume
Technische Universität Berlin
386 PUBLICATIONS 4,491 CITATIONS

SEE PROFILE

Available from: Matthias Kraume

Retrieved on: 02 June 2016
 Bubble Columns 1

Bubble Columns
For other industrial reactors and their applications, see → Stirred-Tanc and Loop Reactors, → Tubular
Reactors, → Fixed-Bed Reactors, → Fluidized-Bed Reactors, → Three-Phase Trickle-Bed Reactors,
→ Reaction Columns, → Thin-Film Reactors, → Metallurgical Furnaces, and → Biochemical Engineering.
Peter Zehner, BASF Aktiengesellschaft, Ludwigshafen, Federal Republic of Germany
Matthias Kraume, BASF Aktiengesellschaft, Ludwigshafen, Federal Republic of Germany

1. Introduction . . . . . . . . . . . . . . . . 2 2.14. Airlift Loop Reactors . . . . . . . . . . 16

2. Bubble Columns and Modifications 4 3. Downflow Bubble Columns . . . . . . 18
2.1. Design and Applications . . . . . . . . 4 3.1. Design and Applications . . . . . . . . 19
2.2. Gas Distribution . . . . . . . . . . . . . 5 3.2. Operating Conditions
2.3. Flow Regimes . . . . . . . . . . . . . . . 6 and Gas Holdup . . . . . . . . . . . . . 20
2.4. Fluid Dynamics . . . . . . . . . . . . . . 7 3.3. Mass Transfer . . . . . . . . . . . . . . . 21
2.5. Bubble Size . . . . . . . . . . . . . . . . 8 4. Jet Loop Reactors . . . . . . . . . . . . 22
2.6. Bubble Rise Velocity . . . . . . . . . . 9 4.1. Design and Applications . . . . . . . . 23
2.7. Dispersion of the Liquid Phase . . . . 9
4.2. Typical Dimensions . . . . . . . . . . . 26
2.8. Dispersion of the Gas Phase . . . . . . 10
4.3. Energy Balance . . . . . . . . . . . . . . 27
2.9. Gas Holdup . . . . . . . . . . . . . . . . 10
2.10. Specific Interfacial Area . . . . . . . . 12 4.4. Mixing Behavior and Fluid Dynamics 27
2.11. Volumetric Mass-Transfer 4.5. Gas Holdup . . . . . . . . . . . . . . . . 28
Coefficient . . . . . . . . . . . . . . . . . 13 4.6. Mass Transfer . . . . . . . . . . . . . . . 30
2.12. Heat Transfer . . . . . . . . . . . . . . . 14 4.7. Three-Phase Loop Reactor . . . . . . 31
2.13. Slurry Bubble Columns . . . . . . . . 14 5. References . . . . . . . . . . . . . . . . . 31

Symbols (see also → Principles of Chemi- fi fraction of cross-sectional area of draft

cal Reaction Engineering and → Model Reac- tube
tors and Their Design Equations) F cross-sectional area, m2
Fi cross-sectional area of draft tube, m2
FR cross-sectional area of reactor, m2
h height, m
Variables hR height of gas – liquid mixture, m
ht height of reactor, m
a specific interfacial area, m−1
JD dispersion flow
A interfacial area, m2
kL liquid-phase mass-transfer coefficient,
d diameter, m
m/s
dh diameter of holes, m
P power, W
di inner diameter of draft tube, m
r radial distance from column axis, m
dn nozzle diameter, m
t time, s
D diffusion or dispersion coefficient, m2 /s
u superficial velocity, m/s
DG, L diffusion coefficient of dissolved gas in
v velocity, m/s
liquid, m2 /s
vrG relative velocity of bubble swarm in liq-
eM energy dissipation rate per unit mass,
uid, m/s
W/kg
vrS relative velocity of particle swarm in liq-
en jet power per unit volume, W/m3
uid, m/s
eV energy dissipation rate per unit volume,
V volume, m3
W/m3
V̇ volumetric flow rate, m3 /s
f fraction of cross-sectional area
z axial coordinate, m

c 2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

10.1002/14356007.b04 275
2 Bubble Columns

Greek symbols nearly always proceeds in the liquid phase. De-

−1 −2 pending on the application, special measures to
α heat-transfer coefficient, W K m
intensify mass transfer between the two phases
ε volume fraction
may be useful, or the residence-time distribution
εG gas holdup
of one or both phases may be modified.
ζ drag coefficient of circulation flow
The liquid may also contain inert, cat-
η dynamic viscosity, kg m−1 s−1
alytically active, or reactive particles in sus-
ν kinematic viscosity, m2 /s
pension. Oxidation, hydrogenation, chlorina-
 density, kg/m3
tion, phosgenation, alkylation, and other pro-
∆ density difference between liquid and
cesses have long been performed in bubble-
gas, kg/m3
column reactors in the chemical industry. In
∆S density difference between liquid and
1978, more than 107 t/a of chemical products
solids, kg/m3
were made in bubble columns [1]. Since then,
σ surface tension, N/m
marked growth has occurred. Industrial reac-
ϕ mass concentration, kg/m3
tors for high-tonnage products have capacities
of 100 – 300 m3 . Larger bubble columns, with
Subscripts capacities up to 3000 m3 , are employed as fer-
menters for protein production from methanol.
a annular space
The largest units (20 000 m3 ) are those for waste-
b bubble
water treatment.
bS Sauter diameter
Scientific interest in bubble columns has in-
c, circ circulation
creased considerably in the past 10 – 15 years.
D downflow
Up to the mid-1970s, only 10 to 20 publications
G gas phase
appeared annually; by the mid- to late 1980s, the
h hole
number had increased to 80 per year. This led
i inside draft tube
to the development of many empirical correla-
L liquid
tions and theoretical models enabling the math-
max maximum value
ematical simulation of bubble- column reactors.
min minimum value
Some academic research groups and commer-
M per unit mass
cial software developers have offered simulation
n nozzle
programs.
p particle
The mixing of a liquid and a gas having only
r relative
partial mutual solubility is one of the unit op-
R upflow, reaction mixture
erations in chemical technology. As Figure 1
slip slip
shows, this operation takes one of three prin-
S solids
cipal forms. The simplest design is the bubble
t reactor
column (Fig. 1 A) in which gas is fed into the
V per unit volume
column at the bottom and rises in the liquid, es-
caping from it at the upper surface; the gas is
consumed to a greater or lesser extent (depend-
1. Introduction ing on the intensity of mass transfer and chem-
ical reaction). When the off-gas contains high
Bubble columns are devices in which gas, in concentrations of valuable reactants, part of it
the form of bubbles, comes in contact with liq- is recycled to the reactor. This recycle design,
uid. The purpose may be simply to mix the liq- however, lowers the concentration profile in the
uid phase. Far more often, however, substances bubble column and must be optimized from an
are transferred from one phase to the other, for economic standpoint. In a simple bubble column
example, when gaseous reactants are dissolved the liquid is led in either cocurrently or counter-
in a liquid or when liquid reaction products currently to the upward gas stream and has a long
are stripped. Both processes can take place si- residence time. The flow direction of the liquid
multaneously. A chemical or biological reaction phase has little effect on the gas-phase residence
time, which is comparatively short. Thus, in the
 Bubble Columns 3

Figure 1. Principal methods of gas – liquid mixing

A) Bubble column; B) Downflow bubble column; C) Jet loop reactor

simple column, the flow of gas is always from in the evacuation and compression of gases has
bottom to top, and the stream can be made up of also been reported [2].
both fresh and recycle gas. In both types of column energy must be sup-
Longer gas-phase residence times can be plied continuously to the two-phase system to
achieved with the downflow bubble column keep the liquid and gas mixed. Only in this way
shown in Figure 1 B. The liquid is pumped down can separation of the phases be counteracted or
through the column at a velocity of more than reversed. In the first case, the simple bubble col-
20 cm/s, so that gas let in at the top is entrained in umn, this energy is supplied by the gas. In the
the flow and can even be held in a suspension- downflow bubble column the energy is supplied
like state until it has reacted completely. Usu- by the downflowing liquid.
ally, however, unconsumed gas is removed with A different mechanism comes into play in
the liquid and separated. Special designs per- the jet loop reactor (Fig. 1 C). Here no net flow
mit phase separation inside the apparatus. The of gas or liquid occurs along the column; in-
downflow bubble column is used mainly when stead, an internal circulating flow is produced.
large liquid streams are to be contacted with One way to achieve this is with a propeller, but
small gas streams and a short liquid residence other approaches exist. In the most commonly
time is required. The necessary velocity cannot used type of loop reactor, the jet loop reactor,
always be obtained with the liquid inlet to the the flow is driven by a high-velocity liquid jet. As
reactor. Thus, like the gas in an ordinary bub- in the downflow bubble column, gas is let in at
ble column, the liquid in the downflow bubble the top and dispersed by the jet energy. Bubbles
column can be recycled. Typical applications can be distributed throughout the reactor volume
for downflow bubble columns are the ozonation only if the downward liquid flow velocity in the
of drinking water and the treatment of water in internal tube is greater than the slip velocity of
swimming pools. A special use of such devices the bubbles. Accordingly, a minimum power in-
put is required.
4 Bubble Columns

These three basic methods of dispersing gas and the nonuniform distribution of gas bubbles
in liquid are generally not used in their pure over the cross section can be reduced by the in-
forms. The variety of problems in chemical and stallation of trays (Fig. 2 B), packings (Fig. 2 C),
biotechnical processes has led to many differ- or shafts (Fig. 2 D). All these devices can oper-
ent contacting devices that combine these basic ate either cocurrently or countercurrently. To set
techniques. up the most homogeneous possible bubble flow,
static mixer elements can also be placed in the
ascending flow section (Fig. 2 E).

Figure 3. Hydroformylation of propene

a) Stripping zone; b) Reaction zone

Hydroformylation. The hydroformylation

of propene is carried out in simple bubble
columns. The reaction is homogeneously cat-
alyzed by rhodium complexes. Usually the
propene and the CO/H2 gas mixture are let in
at the bottom of the reactor. Incompletely re-
Figure 2. Types of bubble-column reactors
A) Simple bubble column; B) Cascade bubble column with acted gas, saturated with the reaction product,
sieve trays; C) Packed bubble column; D) Multishaft bubble exits the reactor. The hydroformylation product
column; E) Bubble column with static mixers is separated from the gas stream by condensation
and forwarded to downstream processing, while
the gas is recycled to the reactor. Because the
2. Bubble Columns and heat of reaction cannot be completely removed
Modifications by evaporative cooling using the enthalpy of va-
porization of the product, the bubble column is
2.1. Design and Applications also equipped with an external cooling loop.
One great advantage of the process is that the
Bubble columns are very adaptable gas – liquid product is recovered from the reaction mixture
contacting devices; possible designs are shown without additional separation operations which
in Figure 2. The simplest form of bubble col- would damage the expensive catalyst system.
umn (Fig. 2 A) consists of a vertical tube with The close coupling between the product and the
no internals. Gas is fed in at the bottom while recycle gas necessary to discharge it (i.e., a cer-
liquid is led through the apparatus cocurrently tain quantity of gas is required for product dis-
or countercurrently. This simple form is seldom charge for thermodynamic reasons), however,
used in practice; instead, a number of modifi- presents some problems. First, the gas flow rate
cations are employed. The back-mixing of gas causes a high gas holdup, which reduces the re-
and liquid phases in the simple bubble column action volume and thus decreases the productiv-
 Bubble Columns 5

Figure 4. Oxidation of montan waxes in cascade bubble columns

a) Cascade bubble-column reactors; b) Separators; c) Final purification of wax oxidate; d) Off-gas treatment

ity of the reactor. Second, large bubbles occur, the fourth (oxidative degradation of wax acids)
which limit the delivery of gaseous reactants to is not. The residence-time distribution in the re-
the liquid phase in the reactor. For these reasons, actor must be controlled so that the desired re-
recycle gas is admitted to the bubble column at actions go as far as possible without the unde-
two levels (Fig. 3) [3]. About half of the recy- sirable reaction occurring to any marked extent.
cle gas is fed via the bottom sparger to disperse Oxidation is performed in four cascaded bub-
reactants into the overlying reaction zone. The ble columns connected in series (Fig. 4). In the
remaining recycle gas is let in via the top sparger, first bubble column, the crude wax for bleach-
which lies slightly below the liquid surface, to ing is metered in along with half of the required
facilitate separation of the reaction product. Fi- amount of chromic acid. Air is supplied to en-
nally, the CO/H2 reactant stream is fed at various hance mixing of the reactants. The spent chromic
levels to supply CO that has been consumed by acid is separated from the wax downstream of
the reaction in the liquid phase. both the first and the second bubble columns.
Another 25 % of the total acid required is added
Oxidation of Montan Waxes. Bubble to the second and third columns. The reaction
columns are used in a cascade when a narrow preferably takes place at 100 – 125 ◦ C and 1 – 5
residence-time distribution is required, for ex- bar, with a residence time of 1 – 3 h for the en-
ample, to prevent or limit undesired consecutive tire cascade. The enthalpy of reaction is removed
reactions. Reducing back-mixing (i.e., a narrow by partial evaporation of the water contained in
residence-time distribution) is also useful when the chromic acid. After exiting the fourth bub-
reaction-engineering considerations dictate that ble column, the oxidized product, spent acid, and
the gas must be fed to various points in the reac- off-gas are separated in two separators.
tor or when a liquid reactant must be degraded
to the greatest extent possible.
Montan waxes from brown coal must be 2.2. Gas Distribution
deresinified, oxidatively bleached, and esterified Usually, the gas is dispersed to create small bub-
(optional) [4], [5]. Oxidation of the waxes con- bles and distribute them uniformly over the cross
sists of several consecutive reactions; the first section of the equipment to maximize the in-
three steps (oxidation of resins and dark- colored tensity of mass transfer. The formation of fine
substances, saponification of montan waxes, ox- bubbles is especially desirable in coalescence-
idation of wax alcohols) are desirable, whereas hindered systems and in the homogeneous flow
6 Bubble Columns

regime (Section 2.3). In principle, however, sig- jet to disperse gas in a zone of high energy-
nificant mass transfer can be obtained at the dissipation rate [11–13]. Figure 6 illustrates sev-
gas distributor through a high local energy- eral frequently used dynamic gas spargers. The
dissipation density [6], [7]. simple two-phase jet nozzle alone (Fig. 6 A) or
In most cases, gas bubbles are generated by with momentum-transfer tube (Fig. 6 B) is not
pores or holes or in the shear zone of a liquid able to simultaneously disperse gas and suck in
jet. Figure 5 shows typical forms of “static” the gas stream. This can be achieved, however,
gas spargers, in which bubble formation occurs with the ejector jet nozzle (Fig. 6 C), the ejec-
without any additional energy supplied from tor (Fig. 6 D), and the Venturi tube (Fig. 6 E). In
outside. The simplest of these devices, the dip nozzle selection the ratio of the gas – liquid vol-
tube (Fig. 5 A), only gives an acceptably uni- umetric flow rates must always be considered.
form gas distribution over the cross section at Common values lie between 0.5 and 2. However,
some distance above the sparger. Perforated much higher values can be achieved in special
plates (Fig. 5 B) and perforated ring spargers cases with momentum-transfer tubes [12].
(Fig. 5 C) are more effective. Both of these re-
quire a certain minimum gas flow rate to achieve
uniform distribution and prevent the liquid from 2.3. Flow Regimes
getting into the sparger [8–10]. Very fine bub-
bles can be generated by the use of porous plates The upward motion of bubbles gives rise to three
(Fig. 5 D), but their pores are susceptible to foul- distinct flow regimes. The crucial quantity for a
ing, and this type of sparger is seldom used in flow regime is the superficial gas velocity. The
full-scale equipment. homogeneous flow regime is marked by a nar-
row bubble-size distribution, and bubbles are
distributed relatively uniformly over the cross
section of the apparatus. This regime extends
to superficial gas velocities of 0.03 – 0.08 m/s,
depending on the gas – liquid system and gas
sparger type.
The uniform distribution of gas bubbles van-
ishes at higher gas rates, and a highly turbulent
flow structure appears. In this heterogeneous or
churn-turbulent flow regime, large bubbles or ag-
glomerates of bubbles form and travel upward at
high velocity (see Section 2.6), mainly in the axis
of the column. The circulating flow that results
may be so vigorous that bubbles of a size corre-
sponding to that in the homogeneous regime are
actually transported downward in the zone near
the column wall (see Section 2.4).
In the small-diameter columns often used as
laboratory equipment, slug flow occurs at high
gas flow rates. Large bubbles are stabilized by
the column wall and take on the characteristic
slug shape.
The relationship between superficial gas ve-
locity and reactor diameter is illustrated by the
flow map of Figure 7 [14]. The broad transition
Figure 5. Static gas spargers regions are due to the effects of the gas distrib-
A) Dip tube; B) Perforated plate; C) Perforated ring sparger; utor, the gas – liquid system, and the liquid rate.
D) Porous plate
A knowledge of the flow regime is particularly
Dynamic spargers offer an alternative to the important because it strongly affects the produc-
static types. They use the power of a liquid tivity of bubble- column reactors.
 Bubble Columns 7

Figure 6. Dynamic gas spargers

diameter, gas holdup, bubble diameter and rise

velocity, viscosity of the liquid, and dispersion
height. Published analyses deal with both lami-
nar liquid circulation, which is only of theoreti-
cal interest [16–18], and turbulent flow, to which
the following discussion is devoted. For exam-
ple, Miyauchi and coworkers use a force bal-
ance over an annular, axially symmetrical vol-
ume element to obtain the velocity profile shown
in Figure 8, [19]. Calculation of the velocities,
however, requires knowledge of the gas holdup
as a function of radial position.
Models of circulation velocity based on en-
Figure 7. Flow regimes in bubble columns ergy balances, in contrast, assume a cell struc-
ture in the bubble column similar to that shown
in Figure 9 [20], [21]. In slender bubble columns,
2.4. Fluid Dynamics both calculations and experimental results show
that the height of the circulation cells hc is equal
Rising gas bubbles entrain liquid in their wakes. to the apparatus diameter d t [20], [22]. Joshi
As a rule, this upward flow of liquid is much and Sharma take into account the energy input
greater than the net liquid flow rate. Because of due to gas compression and energy losses by
continuity, regions therefore exist in which the dissipation in the wakes of the rising bubbles,
liquid is predominantly moving downward. as well as liquid transport across the liquid sur-
Many theoretical and experimental studies face (hydraulic pump), thus obtaining a veloc-
have described the flow behavior of the liquid ity profile over the cross section. Hills [23] and
phase [15]. The circulation velocity is given as Kojima and coworkers [24] have determined ve-
a function of superficial gas velocity, column locity profiles experimentally in bubble columns
8 Bubble Columns

having diameters of 0.14 and 5.5 m, respectively depends chiefly on bubbles far from the gas dis-
(Fig. 10). For the mean circulation velocity ῡL, c tributor,the following discussion only concerns
in bubble columns with additional liquid rate uL , these.
Joshi and Sharma [20] give the expression

 
3 εG uL
v̄L,c = 1.4 gdt uG ± −εG vrG (2.1)
1−εG

where εG is the gas holdup (+ for countercur-

rent, − for cocurrent). Zehner, using a force
balance, arrives at a similar relation for the mean
circulation velocity [21]:


3 1 ∆
v̄L,c = · gdt uG (2.2)
2.5 L

The velocity profiles derived from the models

and, in particular, the mean velocities enable cal-
culation of the essential fluid-dynamic parame-
ters in bubble columns [20], [21], [25].

Figure 9. Cell structure in bubble columns

Two basic methods – photography and probe

techniques – exist for determining bubble size;
however, they do not lead to identical results.
Both methods are subject to certain limitations
in view of the marked bubble selection that
may occur (i.e., not all bubble sizes can be de-
tected) [26], [27]. In particular, any measure-
ment method only leads to realistic results if the
flow is homogeneous (i.e., a narrow bubble-size
distribution is found). As yet, no method can
be recommended for the measurement of large
Figure 8. Radial distribution of liquid velocity in a bubble bubbles in the heterogeneous flow regime.
column
If bubbles are generated in a region of high
turbulence (as with dynamic gas spargers), the
following formula [28] can be used to describe
2.5. Bubble Size the Sauter diameter d bS (mean bubble diame-
ter, calculated from the volume to surface ratio)
Analysis of bubble size in bubble columns must
[29], [30].
distinguish between bubble-size distribution just
after bubble formation at the sparger and size  0.6  0.25
2 σ ηG
distribution further away from the distributor. dbS = ε0.5
G (2.3)
e0.4
M L ηL
Because of breakup and coalescence of the rising
bubbles, the two distributions can differ signif- This formula is based on Kolmogorov’s theory
icantly. Since the efficiency of bubble columns of isotropic turbulence.
 Bubble Columns 9

When static gas spargers are used, the bubble 2.6. Bubble Rise Velocity
diameter is only weakly dependent on gas veloc-
ity. Descriptive correlations [31–34] are applica- In the homogeneous flow regime, bubbles of al-
ble only to the systems and sparger geometries most uniform size and shape rise in the form of
for which they were obtained; a generally valid a swarm distributed uniformly over the column
description of bubble size does not yet exist. The cross section. When the regime changes, larger
maximum bubble diameter d b, max can be used bubbles or agglomerates of bubbles form in ad-
for purposes of estimation [27], [35]. For low- dition to the bubbles already present [36], [37].
viscosity liquids, the maximum bubble diameter These aggregates rise at a markedly higher ve-
is given by locity than the small bubbles. Figure 11 shows
measured velocities for large and small bubbles

σ [36]. Large bubbles first appear at a superficial
db,max = 3 (2.4)
gL gas velocity of ca. 0.03 m/s. The formation of
large bubbles, however, depends strongly on the
where σ is the surface tension. For the water – air type of sparger used. With sintered plates, for ex-
system, d b, max = 8 mm. Larger bubbles have ample, larger bubbles do not appear at gas rates
a high probability of being unstable and thus lower than ca. 0.1 m/s. As shown in Figure 11,
breaking up. The Sauter diameter for real distri- large bubbles have a rise velocity that is four or
butions is between 40 and 60 % of the largest more times larger than small ones. Thus, most
stable bubble diameter. This estimate is not, of the gas transport in the heterogeneous flow
however, applicable to the heterogeneous flow regime is accomplished by large bubbles. In this
regime due to the binodal bubble-size distribu- regime, the quantity of gas transported by small
tion in this regime. bubbles remains constant, whereas the quantity
transported by large bubbles increases linearly
with gas velocity. This relationship applies to co-
alescing and coalescence-hindered gas – liquid
systems.

Figure 11. Velocities of rising bubbles for the system

water – air
Reactor: d t = 0.44 m, ht = 5 m; Gas distributor: perforated
plate (d h = 3 mm)

2.7. Dispersion of the Liquid Phase

Figure 10. Calculated radial profiles of liquid velocity in Because of the large-scale circulation flows,
bubble columns [20] back-mixing occurs in both phases. The result-
ing dispersion flow J D is usually governed by an
10 Bubble Columns

equation analogous to Fick’s first law for molec- 2.8. Dispersion of the Gas Phase
ular diffusion. For the one-dimensional case of
axial dispersion, which is generally sufficient for Due to the large-scale circulation flow both the
a description, follows liquid and gas phases are dispersed. Further-
more, the formation of large and small bub-
dc bles, coalescence, and breakup result in addi-
JD = −DL (2.5)
dz tional dispersion in the gas phase. Whereas the
The dispersion coefficient DL is essentially a gas phase in a bubble column with a smaller
function of the superficial gas velocity and the diameter flows with virtually no back-mixing,
column diameter (e.g., see [38]). Flow direc- large units behave more like stirred tanks. The
tion or liquid velocity does not show any ef- gas-phase dispersion coefficient depends more
fect, provided the superficial liquid velocity re- strongly on gas velocity and column diameter
mains within the range common in industry than does that of the liquid phase. For this rea-
(uL < 0.03 m/s). The dispersion coefficient can son, the degree of axial gas mixing is especially
be estimated fairly accurately on the basis of relevant for scale-up when the gas phase is ex-
fluid-dynamic models. For example, Joshi and pected to show strong concentration variations.
Sharma [20] and Zehner [21] give dispersion Many formulas in the literature describe the
coefficients derived from the mean circulation dispersion coefficient as a function of different
velocity. Each of these formulas gives a good de- independent variables. A particularly suitable
scription of the experimentally determined dis- formula is [40]:
persion coefficients known from the literature.  3
By way of example, Figure 12 compares experi- uG
DG = 5 × 10−4 d1.5
t (2.7)
mental results reported by various workers with εG
the theoretical relation derived by Zehner: This formula is not, however, dimensionally ho-

mogeneous (DG in cm2 /s, uG in cm/s, d t in cm),
dt v̄L,c dt 3 1 ∆ and the gas holdup must be known. By contrast,
DL = = · gdt uG (2.6)
2 2 2.5 L the equation
The equation emphasizes that DL strongly de- ∆
gdt uG
L
pends on column diameter. εG DG =0.2dt uG 3
(2.8)
vrG

derived by Zehner and Schuch is dimension-

ally correct [41]. However, more recent mea-
surements [42] have shown that this correla-
tion must be modified in the heterogeneous flow
regime (uG ≥ 5 cm/s) because the proportional-
ity is somewhat different:

εG DG ∼ (uG dt )1.65

2.9. Gas Holdup

Gas holdup is one of the most important operat-

ing parameters because it not only governs phase
fraction and gas-phase residence time but is also
Figure 12. Liquid-phase dispersion coefficient measured by crucial for mass transfer between liquid and gas.
various authors [39] Gas holdup depends chiefly on gas flow rate, but
also to a great extent on the gas – liquid system
involved. Accordingly, many correlations that
 Bubble Columns 11

have been published only apply to the systems The effects of physical properties on gas
investigated. holdup are exceedingly complex. Increasing the
Gas holdup is defined as the volume of the viscosity of the liquid phase leads to increased
gas phase divided by the total volume of the dis- bubble coalescence and thus a decrease in gas
persion: holdup. Above ca. 50 mPa · s, however, the gas
VG holdup remains constant [46]. Although surface
εG = (2.9) tension is not very important for the gas holdup,
VG +VL
The relationship between gas holdup and gas ve- a change in coalescence behavior may have last-
locity is generally described by the proportion- ing effects. When gas-phase residence times are
ality long and gas distribution is obtained with per-
forated or sintered plates, the presence of salts
εG ∼un
G or alcohols that counteract coalescence has lit-
In the homogeneous flow regime, n is close to tle effect [44]. In contrast, gas holdup increases
unity. When large bubbles are present, the expo- markedly in systems sparged by two-phase noz-
nent decreases, i.e., the gas holdup increases less zles when coalescence is hindered [47]. Such
than proportionally to the gas flow rate (Fig. 13). behavior can be attributed to a small-bubble gas
The higher the contribution of large bubbles to holdup higher than that in coalescing systems,
the total gas holdup, the smaller is the exponent whereas the content of large bubbles is identi-
n. In the fully developed heterogeneous flow cal [36]. Small bubbles formed under high shear
regime, n finally takes on values between 0.4 stresses in the region near the two-phase nozzle
and 0.7, depending on the gas – liquid system. cannot recombine so the gas holdup increases
significantly with this type of gas distributor.
The relation of Akita and Yoshida [48] is
suitable for estimating the gas holdup and is
based on the investigation of numerous systems:
εG
c1 =
(1−εG )4
 2 1  3  1  
gdt L 8 gdt 12 uG
√ (2.10)
σ νL gdt

For pure liquids and nonpolar solutions the con-

stant c1 is 0.2, for electrolyte solutions it is
0.25. However, reliable results cannot be ex-
pected for systems that have not been investi-
gated in this study. The effects of reactor pres-
sure on gas holdup have not been fully explained.
Although some authors find no effect between
Figure 13. Gas holdup and fraction of large bubbles (sys- 1 and 16 bar [49], others find that gas holdup
tem: water – air; gas distributor: perforated plate d h = 3 mm) increases with pressure in systems with small
The effect of low liquid velocities uL on gas sparging holes (d h ≤ 1 mm) or with sintered
content is generally negligible. At high flow plates [44], [50–52]. Transition from the homo-
rates the gas holdup decreases in cocurrent sys- geneous to the heterogeneous regime occurs at
tems because gas bubbles pass through the col- higher gas flow rates as pressure increases.
umn more quickly. In contrast, the gas holdup Gas holdup is generally a function of posi-
rises in countercurrent systems; this can lead to tion in the bubble column. Axial profiles of gas
extremely high gas holdup, especially in down- holdup show a zone near the gas distributor in
flow bubble columns [43]. which the holdup increases to the value that char-
Above 0.1 m, the reactor diameter is of sec- acterizes the following equilibrium zone. The
ondary importance for gas holdup, as measure- gas holdup at the top of the column, in the zone
ments on units having diameters between 0.1 and
5.50 m show [34], [36], [44], [45].
12 Bubble Columns

of bubble breakup, is markedly higher than the

equilibrium value [45]. A 6εG
a= = (2.11)
Gas holdup also depends on radial position. VR dbS
The profile shows gradients only near the wall in where V R is the volume of the reaction mixture
the homogeneous flow regime [23], [45], [53]. In and d bS is the mean bubble diameter (Sauter di-
contrast, a parabolic radial gas holdup distribu- ameter, Section 2.5). As Figure 15 shows, the
tion appears in the heterogeneous regime [23], interfacial area increases with increasing gas
[37], [45], as a consequence of the preferential flow rate. An exception occurs when a porous-
rising of large bubbles or agglomerates of bub- plate sparger is used; like gas holdup, interfacial
bles in the axis of the column. Figure 14 shows area decreases on transition to the heterogeneous
radial profiles of gas holdup at various gas flow flow regime and then approaches the same val-
rates in the water – air system. ues observed with perforated plates. The growth
in interfacial area with increasing gas velocity is
always greater in the homogeneous than in the
heterogeneous flow regime. The reason lies in
the formation of large bubbles in the heteroge-
neous regime: the interfacial area of large bub-
bles per unit volume is markedly lower than that
of smaller ones.

Figure 14. Radial profiles of local gas holdup Figure 15. Specific interfacial area as a function of super-
d t = 0.45 m; ht = 6.2 m; h = 3.03 m (at measurement point); ficial gas velocity
perforated plate d h = 1 mm a) d t = 0.102 m; b) d t = 0.29 m; c) d t = 0.14 m; d) d t = 0.1 m;–
– – Porous plate; —- Perforated plate

The specific interfacial areas attainable in

2.10. Specific Interfacial Area various gas – liquid reactors can be compared on
the basis of power input P per unit volume [29].
The area of the gas – liquid interface is one of Experimental values can be described by the re-
the most important process parameters. Espe- lation
cially at high reaction rates (e.g., when a bubble  m
column is employed as an absorber), the interfa- P
a=k εn
G (2.12)
cial area becomes a crucial factor in equipment VR
sizing. Like gas holdup, interfacial area depends The exponent m is between 0.4 and 1 [54]. The
on the geometry, operating conditions, and gas – plot in Figure 16 enables a direct comparison
liquid system. Gas holdup and interfacial area to be made between reactors with respect to the
per unit volume are related as energy required to produce a given interfacial
area.
 Bubble Columns 13

Figure 16. Specific interfacial area as a function of specific

power input [55]
a) Stirred tank; b) Bubble column with porous plate; c) Bub-
ble column; d) Bubble column with two-phase jet nozzle
( jet loop reactor); e) Packed column; f ) Bubble column
with injector nozzle

2.11. Volumetric Mass-Transfer

Coefficient
The mass transfer between the gas and the liq-
uid phase in a bubble column can be described in
most cases by the volumetric mass-transfer coef- Figure 17. Mass-transfer coefficients in bubble columns
ficient k L a, which is the liquid-phase masstrans-
fer coefficient k L multiplied by the specific in- According to experimental results, the col-
terfacial area. Gas-phase resistance can usually umn diameter above about 15 cm has no ef-
be neglected, so k L a gives an adequate descrip- fect on mass-transfer coefficient. Some correla-
tion. To determine the mass-transfer rate, how- tions nonetheless include reactor diameter [31],
ever, the driving concentration difference must [57], [60]. Akita and Joshida [31] state that
be known which in turn requires a knowledge of the value of the column diameter used for cal-
mixing behavior in the gas and the liquid phase. culation should not be increased beyond 0.6 m.
In industrial units (d t > 1 m), estimates can be Based on this premise, their correlation for k L a
based on the assumption of complete mixing in is
both liquid and gas phases.
Like gas holdup and interfacial area, k L a also kL ad2t
=
depends on the gas flow rate, type of sparger, DG,L
and gas – liquid system. The mass-transfer coef-  0.5  2 0.62  3 0.31
νL gdt L gdt
ficient and the gas rate are again proportional to 0.6 2
ε1.1
G
DG,L σ νL
one another: (2.13)

kL a∼un
G and has the best experimental support. The mass-
where n can be between 0.7 and 0.92 [31], [56– transfer coefficient increases in coalescence-
59]. Mass-transfer coefficients two- to threefold hindered systems [54], [61]. This increase de-
higher can be achieved in the homogeneous flow pends on the system and the concentration
regime if a porous plate is used as sparger instead of coalescence-hindering substance. The maxi-
of a perforated plate (Fig. 17). In the heteroge- mum gain in mass-transfer coefficient due to the
neous regime, however, the effect of the sparger presence of electrolytes, however, is only 30 %.
is negligible.
14 Bubble Columns

Figure 18. Examples of heat exchanger inserts

A) Parallel single tubes; B) Tube bundle in cross flow; C) Longitudinal tube bundle; D) Tube spiral or helix; E) Draft tube
with jacket

2.12. Heat Transfer

 1/3
In many cases, heat must be removed when oper- π
l=db
ating bubble columns. A particularly simple so- 6εG
lution is to utilize the latent heat of vaporization and ῡ L, c is calculated by Equation (2.2). On the
of the liquid phase for heat removal, although whole, these two approaches correlate well with
this is not always feasible. In addition, many pos- literature data. Heat transfer in bubble columns
sibilities exist for heat transfer through heated or with heat-exchange internals has not been in-
cooled surfaces, as shown in Figure 18. In this tensively studied [65–69]. For tube bundles ar-
way, up to ca. 30 m2 /m3 of heat-transfer area can ranged in an axial direction (Fig. 18 C), the heat-
be installed in a bubble column. transfer coefficient increases with increasing
The turbulent flow generated by rising gas tube pitch and decreases when the free cross-
bubbles increases heat transfer even at low gas sectional area increases [68], [69]. A similar re-
rates (Fig. 19). The increase in heat-transfer co- lationship is found for a tube bundle arranged in
efficient α, with gas throughput is markedly cross flow (Fig. 18 B), but here a marked effect
greater in the homogeneous than in the hetero- of liquid throughput occurs [67].
geneous regime. The installation of tube bundles leads to an
The heat-transfer coefficient does not depend overall change in fluid dynamics and thus in
on the column diameter, type of sparger, or coa- mixing behavior. For example, tubes installed
lescence behavior of the gas – liquid system. in cross flow hinder flow in the longitudinal di-
Two distinct concepts are used to describe rection and thus reduce dispersion in the liquid
the heat-transfer coefficient at the wall. Whereas phase [70]. In contrast, the arrangement of heat-
Kast [62] and Deckwer [63] consider ra- transfer surfaces in the flow direction leads to
dial flow and heat transported by it, Joshi and more intense mixing of the liquid phase by in-
coworkers [64] and Zehner [21] use circulation tensifying circulation [71], [72].
velocities derived by them for physical model-
ing. Here with the following relation for the heat-
transfer coefficient α can be derived [21]:
2.13. Slurry Bubble Columns
 2
1/2
v̄L,c
α= 0.18 (1−εG ) λ2L L cpL · (2.14) Solid particles are present in bubble columns in
l ·νL
a wide variety of processes; they must be held
where
 Bubble Columns 15

in suspension by the rapid liquid circulation al- centration (< 10 wt %) and low settling veloc-
ready discussed. The presence of the solid phase ity of the particles the gas holdup is nearly un-
in a slurry bubble column means that all pro- changed [74]. In contrast, the gas holdup de-
cess parameters behave differently, and in a more creases at higher settling velocities with increas-
complicated way, than in a two-phase bubble ing solids concentration [75–77]. The strength
column. of this effect differs from one flow regime to an-
other. The decrease is particularly marked when
an increase in solids content leads to a change
from the homogeneous to the heterogeneous
regime. On the other hand, in the heterogeneous
regime the reduction in gas holdup is only slight
with increasing solids content. Yasunishi and
coworkers [78] verify and recommend the gas
holdup relation of Koide and coworkers [79]
εG
= (2.16)
(1−εG )4
 −0.252
4
0.918 g ηL
0.277 uGσηL L σ 3
0.748 0.881 −0.168
ϕ̄S S −L d t u G L
1 + 4.35 S L ηL

for a wide range of solids concentrations and liq-

uid properties. For coalescence-hindered aque-
ous electrolyte solutions a coefficient 0.364 must
be used instead of 0.277 in Equation (2.16).
The mixing behavior of the liquid and the
gas phase is very similar to that in the two-
phase bubble column. The axial solids distri-
bution can be described by a one-dimensional
sedimentation – dispersion model. The solids
Figure 19. Heat-transfer coefficient at reactor wall dispersion coefficient is generally lower than the
d t = 0.196 m; ht = 6.20 m; liquid velocity uL = 1.2 cm/s corresponding liquid dispersion coefficient [80].
The minimum gas velocity necessary to hold The difference between the two values increases
the solids in suspension increases as the con- rapidly with increasing settling velocity of par-
centration and density of the particles increase. ticles. For small solid particles (vrS < 0.01 m/s),
The increment depends, however, on the phys- the effect of superficial gas velocity on the axial
ical properties of the solid and liquid phases. solids concentration profile is negligible above
Many empirical equations for the critical gas ve- the minimum gas velocity for suspension [81].
locity for complete suspension show a marked The specific interfacial area declines contin-
increase with increasing single-particle settling uously with increasing solids content [76], [82].
velocity. Equation (2.15) [73] can be used for This phenomenon can be explained by the for-
design purposes: mation of larger gas bubbles, due to the pres-
uG,min ence of solid particles that lead to the observed
vrS
decrease in gas holdup. If, however, very fine
    √ 0.24 particles are used in aqueous electrolyte solu-
S −L 0.6 ϕ̄S 0.146 gdt
= 0.801 tions (i.e., systems with hindered coalescence),
L S vrS
  0.578 
the interfacial areas produced do not differ from
gηL4 those in the two-phase system [83].
1 + 807  (2.15)
L σ 3 The effect of solids on the volumetric mass-
transfer coefficient depends largely on particle
where vrS is the relative settling velocity of the properties, solids content, and physical proper-
particle swarm in the liquid. At low solids con- ties of the liquid. At low solids concentration (up
16 Bubble Columns

to ca. 3 – 5 wt %), the mass-transfer coefficient the catalyst. In the external cooling loop, the high
matches that of the two-phase bubble column. heat of reaction is removed and suspension of the
A higher solids content, like increasing parti- solid catalyst is assisted by the circulating liq-
cle size, leads to a drop in k L a relative to the uid stream. Complete conversion of benzene is
solid-free condition [78], [79], [84], [85]. For accomplished in a fixed-bed reactor (b) installed
very fine particles (d P < 36 µm), however, k L a downstream on the gas side.
also decreases with decreasing particle diameter
[86]. Overall, the behavior of k L a is governed
by the interfacial area per unit volume because
the change in k L is generally small.
The mass-transfer coefficient between liquid
and solid increases roughly as the fourth root of
the gas flow rate, decreases with increasing liq-
uid viscosity and particle diameter, and becomes
partly independent of these factors at high gas
rates. Two distinct models can be used for the
mathematical description, but they lead to simi-
lar values if the solids are completely suspended.
Values reported by various authors are compared
in Figure 20 [87].
Figure 21. Hydrogenation of benzene to cyclohexane
a) Main reactor; b) Secondary reactor; c) Steam
drumht = 10 m; d t = 2.5 m; T = 195 ◦ C; p = 22 bar; Gas ve-
locity: 7.5 cm/s; Liquid residence time ∼ 3 h

The usual reaction conditions in the bubble-

column reactor are 200 ◦ C and 22 bar. Typical
gas velocities are ca. 0.08 m/s; liquid-phase res-
idence time is ca. 3 h.

2.14. Airlift Loop Reactors

In contrast to bubble columns, airlift loop re-
Figure 20. Liquid – solid mass-transfer coefficient esti-
mated with five different correlations [87]
actors are characterized by a well-defined liq-
ν L = 10−6 m2 /s; L = 1000 kg/m3 ; DG, L = 10−9 m2 /s; uid circulation, which is achieved by dividing
—- ∆S /L = 1.5;– – – ∆S /L = 0.3 the reactor into sections with and without gas
sparging. The difference in gas holdup between
these two zones drives the liquid circulation. In
Hydrogenation of Benzene. In the IFP pro- principle, two types of airlift loops can be identi-
cess, benzene is hydrogenated to cyclohexane in fied (Fig. 22). In the first (airlift reactors with in-
a slurry bubble column [88] (→ Cyclohexane, ternal loop), either a concentric tube (Fig. 22 A,
Chap. 4.1.). This process was used to produce B) or a plane partition (Fig. 22 C) divides the
1.8 × 106 t of cyclohexane in 23 plants world- column into riser and downcomer sections. In
wide in 1991. the second (airlift reactors with external loop,
Figure 21 shows the slurry bubble column (a) Fig. 22 D), two separate tubes form the upflow
in which benzene is hydrogenated on suspended and downflow zones; the tubes are joined by two
Raney nickel. The hydrogen-rich gas that is let in horizontal sections at top and bottom.
at the bottom of the main reactor provides hydro- The dependence of liquid circulation veloc-
gen for the reaction and also strips the product ity on superficial gas velocity is described by the
cyclohexane out of the reactor. Thus, the pro- purely empirical relation
cess can be operated without the need for ex-
pensive equipment to separate the product from v̄L,c =C1 uC2
(2.17)
G
 Bubble Columns 17

The value of the constant C 1 is determined by re- and viscosity have much less effect on airlift
actor geometry and the physical properties of the loop reactors than on bubble columns because
system; C 2 depends on both flow regime and re- the interactions between bubbles are far weaker
actor geometry [89–91]. A physical model based as a result of the high circulation velocity. For
on an energy balance [39], [92] leads to the re- the same reason, the homogeneous flow regime
lation in airlift loop devices extends to much higher
 gas rates than in bubble columns [91], [93].
3
v̄L,c =c· ghR uG (2.18)

An exponent of 0.33 for gas flow rate is of the

order of usual experimental results.

Figure 23. Gas holdup in airlift reactors with external loop

(system: 0.1 mol/L NaCl solution – air)d t = 0.1 m;
ht = 8.5 m; Porous plate = 150 µm

The gas holdups that occur in airlift reactors

with internal loop are only slightly lower than
those in bubble columns. The decrease in gas
holdup in the riser is partly offset by an increase
in the downcomer [94], [95].
As in the bubble column, the volumetric
mass-transfer coefficient increases with increas-
ing gas flow rate. Because the liquid-phase mass-
transfer coefficients k L are the same in bubble
columns and airlift reactors [96], [97], the dif-
ference in k L a results from differences in inter-
Figure 22. Types of airlift loop reactors facial area. Airlift reactors with external loop al-
A) Concentric draft tube with external recycle; B) Concen-
tric draft tube with internal recycle; C) Deep shaft reactor
ways have lower mass-transfer coefficients than
(ICI); D) External loop bubble columns because the lower gas holdup
implies a smaller area for mass transfer. The
Airlift reactors with external loop (Fig. 22 D)
k L a values for airlift reactors with internal loop,
are usually run at much higher gas and liquid
on the other hand, are similar to the values for
flow rates than conventional bubble columns.
bubble columns (Fig. 24) because here the gas
The high circulation velocities significantly
holdups differ only slightly [93], [98–100].
change the nature of the two-phase flow, namely,
In contrast to bubble columns, the residence-
the gas holdup declines with increasing circu-
time distribution of airlift reactors is influenced
lation velocity (see Fig. 23). The highest gas
not just by longitudinal mixing but also by back-
holdup occurs in the bubble column (uL = 0),
mixing due to the circulation flow. Reactors with
where the absolute velocity of the rising bubbles
an external loop exhibit axial dispersion coeffi-
is lowest because of the zero liquid velocity. Pa-
cients whose values are up to 20 % lower than
rameters such as surface tension, coalescence,
18 Bubble Columns

those of bubble columns, depending on the cir- Biological Wastewater Treatment. Airlift
culation velocity [93]. The consequences of this loops are employed to provide well-defined
back-mixing for the reactor yield of airlift loops, back-mixing of the liquid phase. This is desir-
however, are far less serious because the high liq- able, for ex-ample, when uniform temperature
uid velocities lead to far larger Bodenstein num- and concentration distributions must be main-
bers (for a definition of the Bodenstein number, tained in the reaction medium to equalize feed
see → Mathematical Modeling). The axial dis- variations as quickly as possible or to prevent
persion coefficients of airlift reactors with inter- settling of solids from the mixture.
nal loop are much lower. Measurements of axial Airlift reactors are used in biological waste-
mixing of the gas and liquid phases show a de- water treatment [5], [103], [104]. These units are
crease in the dispersion coefficients by roughly closed vessels ca. 15 – 25 m tall and 10 – 45 m in
a factor of three [101]. diameter; they have small space requirements,
very good oxygen utilization, and greatly re-
duced off-gas and noise emissions. The contents
of the reactor circulate through one or more draft
tubes (Fig. 25); sparging occurs outside the draft
tubes. During operation, the growing microor-
ganisms must constantly be provided with suf-
ficient oxygen and substrate, and adequate mix-
ing of the wastewater – activated-sludge mix-
ture must be insured. If these conditions are
not satisfied, solids will settle and anaerobic fer-
mentation processes may occur. The usual con-
ditions are as follows: superficial gas velocity
1 – 3 mm/s, gas holdup between 1 and 3 %, and
roughly 25 circulations of reactor contents per
hour. The utilization of atmospheric oxygen is
Figure 24. Comparison of mass-transfer coefficients for air- more than 50 %. The wastewater has a residence
lift reactors and bubble columns time between 6 and 15 h.
(system: salt solutions; gas distribution: small bubbles)

As in bubble columns, mixing times decrease

with increasing superficial gas velocity because 3. Downflow Bubble Columns
the circulation velocity becomes greater. Since
the circulation time t c and the mixing time t M Chapter 2 described ordinary bubble columns
are directly proportional [92] in which gas flows from bottom to top and has
 0.5 a short residence time (gas sparging method A
tM FD
=γ (2.19) in Fig. 1). In downflow bubble columns, by con-
tc FR
trast, the gas and liquid phases are transported to-
(where γ = 3.5 for internal circulation and 5.2 for gether from top to bottom (Fig. 1 B). This regime
external circulation), the relationship between demands liquid velocities vL greater than the rel-
mixing time and circulation velocity can be ex- ative velocity vrG between the two phases. De-
pressed as pending on the liquid velocity chosen, very low
−1
tM ∼v̄L,c gas velocities can be achieved
The mixing time is also directly proportional to vG = vL − vrG (3.1)
the distance traveled. Five to six passes are re-
quired for a degree of mixing of 90 %. The liquid or long residence times. This is an advantage es-
circulation in airlift reactors, with their high cir- pecially when a large-volume liquid stream must
culation velocities, leads to higher heat-transfer come in contact with a small-volume gas stream.
coefficients than in bubble columns [102]. As in In the extreme case, a virtually suspended state
bubble columns, the heat-transfer coefficient in
airlift loop reactors increases with gas flow rate.
 Bubble Columns 19

Figure 25. Biohoch reactor (Hoechst)

a) Settling zone; b) Aeration chamber; c) Draft tube

of the bubbles can be realized, with an arbitrarily Adding a liquid recycle creates diverse pro-
long residence time. Under certain conditions, cess design options. The back-mixing involved,
this permits complete conversion of the gas. Nor- which is usually undesirable, can often be ac-
mally, however, part of the gas must be assumed ceptable. With a liquid recycle the downflow
to exit the reactor without reacting, mainly when bubble column can be operated on small feed-
the gas contains components that do not react. In streams. The recycle loop also provides a sim-
such cases, the cocurrent motion of the phases ple way of adding or removing heat, so that the
is a disadvantage because only one theoretical temperature profile in the reactor becomes more
transfer unit can be realized. uniform.
From the standpoint of process engineering,
the downflow bubble column with integrated
3.1. Design and Applications separator (Fig. 26 B) differs little from the sim-
ple downflow bubble column. The integrated
As with bubble columns, a variety of designs ex- separator is well suited when larger quantities of
ist for downflow bubble columns. These differ off-gas must be removed. A typical application
mainly in the way the gas is let in, the bubbles is in the ozone treatment of water air or oxygen
are generated, and the unreacted gas is removed. with a low ozone content is fed to the reactor and
Figure 26 shows some examples. the quantity of exit gas is almost the same as the
The simple downflow bubble column quantity of inlet gas. The design of the reactor
(Fig. 26 A) is particularly suitable for gases that with integrated separator is simple. The shoul-
are soluble in the liquid phase and/or fast re- dered form is not suitable for high pressure.
actions. Unreacted gases cannot be separated The downflow – upflowbubble column
in the column, so an extra separator may be (Fig. 26 C) combines a downflow bubble col-
required. The simple downflow bubble column umn and an ordinary bubble column. Particu-
is often employed at high pressure (> 100 bar). larly long gas residence times are possible. The
A slender geometry makes it possible to reduce liquid routing shown in Figure 26 C gives a fre-
the wall thickness of the cylindrical reactor. To quently desirable residence-time distribution:
improve mass transfer between gas and liquid the downflow section features mixing similar to
phases, the vessel can be packed with particles, a stirred tank by virtue of the pump stream. This
which also reduce both the required liquid rate is advantageous with a high heat of reaction,
and the axial mixing of liquid and gas. Usually, which can be removed with the pump stream.
however, packings are used as catalyst supports. The bubble column in the outer annular space
The classical application of this type of device is merely carries the outflow, and back-mixing in
the hydrogenation of a wide range of substances. this zone can be suppressed by internals (pack-
20 Bubble Columns

ing, static mixers, sieve trays). High conversion gases with low inerts content can be completely
are achieved in higher-order reactions. converted under pressure with this method, also
called dip-tube sparging.
The lower part of the downflow bubble col-
umn serves as the separator. Only small gas bub-
bles are carried out of the reactor, which still
have some reactivity. The pump is therefore pro-
tected against excessive contents of gas in the
liquid, even in coalescence-hindered systems.

3.2. Operating Conditions and Gas

Holdup

Gas is fed in at the top of the column and dis-

tributed as uniformly as possible over the cross
section. Large gas holdups (εG ≈ 0.3 – 0.35)
can be obtained even in coalescing systems
such as water – air [105–110]. When the sys-
tem is coalescence-hindered, for example, if car-
boxymethyl cellulose or ethanol has been added
to the water – air system, values up to εG ≈ 0.45
can be achieved [108].
The gas holdup can be estimated as
uG
εG = (3.2)
vL − vrG

The relative velocity vrG for coalescing aqueous

systems takes on values in the range 0.2 m/s ≤
vrG ≤ 0.3 m/s. The liquid velocity vL can be
calculated from the liquid rate uL and the gas
holdup εG :
uL
vL = (3.3)
1−εG

Hence the gas holdup can be expressed as

uG
εG = uL (3.4)
1−εG
−vrG

If gas and liquid rates are given, Equation (3.4)

Figure 26. Designs of downflow bubble columns
A) With external gas separator; B) With integrated gas sepa- can be solved for the gas holdup:
rator; C) Combined downflow – upflow with bubble column 

in annulus; D) Dip-tube gas sparging with internal gas re-  2
B 2 uG
cycle εG = + +1−1
2 B vrG
Downflow – Upflow Bubble Column
(Fig. 26 D). Another design combines downflow where
and ordinary bubble columns. However, the top
of the downflow section is in the gas space of the uL +uG
B= −1 (3.5)
reactor. Fresh gas together with recycle gas (that vrG
has escaped from the liquid surface) is drawn in Finally, Equation (3.4) can be written so as to
here and dispersed in the liquid. Pure gases or yield the liquid rate:
 Bubble Columns 21

flow rate and rising pressure has been reported

  [108].
uG
uL = (1−εG ) +vrG (3.6) If the gas holdup is known, the specific inter-
εG
facial area can be estimated as
A flow chart can then be derived for the
water – air system (vrG = 0.23 m/s). In Figure 27, 6εG
a=
the gas velocity is plotted versus the superficial dbS
liquid velocity, with the gas holdup εG as a pa- which gives maximum values up to
rameter. For εG > 0.35, flow is in the hetero- a = 450 – 700 m−1 . Figure 28 compares these
geneous regime. In this regime a highly turbu- values to those measured in simple bubble
lent two-phase flow develops, resulting in strong columns and packed columns [108] (referred
mixing of liquid and gas. In the extreme case, not to the total volume but to the liquid vol-
gas accumulates at the top and can propagate ume). This figure illustrates the advantages of
throughout the reactor from there. Kulkarni re- downflow bubble columns, which have higher
ports somewhat different results [111], possibly aL values at low gas rates.
because of less uniform gas feeding. A liquid-phase mass-transfer coefficient
k L ≈ 3.7 × 10−4 m/s has been calculated by
plotting the volumetric mass-transfer coefficient
k L a versus the volumetric gas holdup [110] and
the bubble size stated above. Somewhat lower
estimates, up to 3 ×10−4 m/s, have been re-
ported in [109], [112]. However, the mass-trans-
fer coefficient k L cannot be measured directly,
and large uncertainties are to be expected. The
volumetric mass-transfer coefficients k L a mea-
sured in bubble columns at equal gas holdups are
roughly the same. Because of the very different
volumetric gas flow rates in upflow and down-
flow bubble columns, different gas conversions
are achieved; Figure 29 compares these figures
for sulfite oxidation with air [110].
Data on the axial back-mixing of the
gas phase have been reported [111]. At gas
Figure 27. Flow chart for downflow bubble columns
rates of uG = 0.001 – 0.01 m/s, the dispersion
Gas velocity is plotted as a function of liquid velocity, coefficient takes on constant high values:
with gas holdup as parameter, for the water – air system DG ≈ 0.2 m2 /s (d t = 0.025 m, uL = 0.334 m/s,
(vrG = 0.23 m/s) calculated with Equation (3.6). εG = 0.025 – 0.09).
For systems with hindered coalescence, de- Back-mixing of the liquid phase under com-
termination of these values and other process parable conditions is also markedly less than in
parameters normally requires experimentation normal bubble columns [113].
[105].
Examples. Steiner and Herbrechtsmeier
studied the oxidation of sulfite solutions with
air in the downflow bubble column, finding a
3.3. Mass Transfer twelvefold higher depletion than with a simple
bubble column (Fig. 29) [110].
The following statements are applicable only to
A process for the absorption of gases con-
coalescing systems. In the homogeneous flow
taining nitrogen oxides has been reported [114].
regime, all bubbles in the downflow bubble col-
Virtually complete conversion can be obtained
umn are almost equal in size. Diameters of
through the use of a downflow bubble column
3 mm ≤ d b ≤ 4 mm are observed [112]. A slight
3 m high.
decrease in bubble size with increasing liquid
22 Bubble Columns

Figure 28. Comparison of mass-transfer area per unit liquid volume for various sparged devices as a function of gas velocity,
with liquid velocity as parameter, according to [108]

Figure 29. Comparison of oxygen depletion in upflow and downflow bubble columns for sulfite oxidation with air [110]
T = 22 ◦ C; [SO2−
3 ] = 0.4 – 0.8 mol/L; [Co
3+ ] = 7×10−6 mol/L; pH = 8.0

The degradation of organic contaminants by 4. Jet Loop Reactors

ozonolysis is a well-known method of water
treatment. For economic and safety reasons, vir- Jet loop reactors are among the most versatile
tually complete depletion of the ozone is desir- gas – liquid contactors. The momentum of the
able, which can be achieved in downflow bubble liquid jet issuing from the nozzle enhances in-
columns as reactors [115]. No danger of fouling ternal circulation and opposes demixing of the
exists in these devices, and high liquid through- phases (distribution method C in Fig. 1). The liq-
puts can be handled. Figure 30 is a flow sheet uid jet can be utilized to suck in, compress, and
of the entire process. Only a single theoreti- disperse fresh or recycle gas (Figs. 31, 32, 33,
cal mass-transfer unit can be realized, but this 34). The liquid-jet data are important process
drawback can be overcome by the proposed use parameters. The liquid volumetric flow rate, ve-
of a reactor cascade [115]. The phases are led locity, and power can be varied over wide ranges.
through the cascade countercurrently. Finally, the size of the draft tube and the upper
 Bubble Columns 23

Figure 30. Plant with downflow bubble column for ozone treatment of water
a) Downflow reactor; b) Pressurizing pump; c) Ozone generator; d) Compressor; e) Deozonizer

flow-reversal zone strongly affect fluid dynam- A combination of all three distribution meth-
ics and gas separation. Other possible variations ods (bubble column, downflow bubble column,
are offered by the nozzle configuration (Figs. 32, and jet loop) is shown in Figure 31 D. The pro-
33, 34). cess characteristics of one sparging type will pre-
The four examples in Figure 31 illustrate op- dominate, depending on the selected gas and liq-
tions for the direction flow pattern phases. In all uid flow rates. At high gas flow rates, for exam-
cases the gas is incorporated into the liquid via ple, the liquid surface level rises above the nozzle
the nozzle located in the gas space. The liquid orifice. The liquid jet then no longer entrains gas
jet entrains gas bubbles until the nozzle orifice bubbles, serving only to drive the circulation and
is closed by the rising liquid surface. The in- disperse the bubbles. This versatile type of dis-
corporation of more gas submerges the nozzle tribution can be further refined through variation
and blocks the gas inlet, the surface level then of the nozzle position and the use of self-priming
drops again; this self-regulating mechanism en- ejectors (Figs. 32, 33, 34). To evaluate a design,
ables the gas holdup to be controlled. the essential process parameters must be esti-
A jet loop reactor (also called a gas- mated, which is not always feasible because of
circulation reactor) [116] without net gas or liq- the many possible variations. For the basic forms
uid throughput is shown in Figure 31 A. Both shown in Figure 31, however, some information
phases are let in at the top and discharged at the can be derived from a power balance (Section
top (the gas phase is consumed to a higher or 4.2).
lower degree). This corresponds to dispersion
method C of Figure 1.
If the gas is under pressure, it can also be let 4.1. Design and Applications
in at the bottom to intensify circulation. The re-
sult is a net gas rate, as in the bubble column Loop Reactors with Downflow Liquid Jets.
(Fig. 31 B); here, distribution methods A and C Figure 32 illustrates several loop reactors with
of Figure 1 are combined. downward-pointing nozzles. These reactors are
Figure 31 C shows a combination of sparging even suitable for suspension catalysis. If the
methods B and C in Figure 1. The liquid is fed pump or feedstream is cut off or fails, the noz-
at the top and discharged at the bottom. This jet zle drains clear, and the danger of the nozzle
loop reactor has an additional net flow of liquid, being plugged by the suspended catalyst is thus
as in a downflow bubble column. reduced. Another advantage is the long gas res-
24 Bubble Columns

idence time. From the inlet at the top, the gas fers much higher gas holdups and better mass-
circulates through the loop at least once. More- transfer performance (Sections 4.5 and 4.6). The
over, the devices are designed so that the gas jet only has to supply the recycle gas. Another
is internally recycled. This is important for the marked increase (up to a factor of two) in the gas
complete conversion of gases containing little or holdup is achieved by installing a momentum-
no inerts. transfer tube in the reactor. This can be sub-
merged (Fig. 33 B) or can extend above the liq-
uid surface (Fig. 33 C).

Figure 31. Options for phase routing in the “gas-

circulation” type of jet loop reactor
A, C) With surface gas sparging; B, D) With pressure sparg-
ing (with gas throughput); C, D) With bottom outlet for
liquid, inlet at top (with liquid throughput)
In the gas- circulation reactor of Figure 32 A Figure 32. Types of jet loop reactors with downward liquid
jet
(see also Fig. 31) [116], the gas can also be let A) Gas-circulation reactor with pressure sparging; B) Gas-
in at the bottom of the reactor, independently circulation reactor without connected gas space; C) Gas-
of the nozzle, if the gas is available at reactor circulation reactor with submerged nozzle; D) “Compact”
pressure. For a given jet power, this design of- reactor with annular space for liquid injection
 Bubble Columns 25

To prevent separation of the gas phase, the In the jet loop reactor proposed by Räbiger
nozzle can be built directly into the reactor top and coworkers, the liquid is fed in via an annu-
(Fig. 32 B) [117]. Gas bubbles separating in this lar nozzle (Fig. 32 D) [119], [120]. Gas can be
zone are immediately entrained by the liquid jet drawn in via the center tube (ejector fashion) or
and redispersed into the circulating flow. supplied under pressure (injector fashion). For
application of this reactor type in wastewater
treatment see [121], [122].

Figure 33. Options for spontaneous gas sparging from top

Surface sparging in gas-circulation reactor without (A) and
with (B) momentum-transfer tube; dip-tube sparging with-
out (C) and with (D) self-priming ejector
The submerged nozzle in Figure 32 C can ei-
ther accept pressurized gas from outside the sys-
tem [118] or suck the gas in. Internal gas recycle
is also possible. The suction of the nozzle can
be enhanced by applying a swirl to break up the
Figure 34. Design options for loop reactors with upward
liquid jet, provided the liquid nozzle orifice is liquid jet
set back somewhat to the rim of the nozzle. The A) With expanded head for gas separation; B) With gas re-
ejectors and ejector nozzles discussed in Chap- cycle via a self-priming ejector; C) With gas recycle via an
ter 2 are suitable for deeper submergence (see ejector jet nozzle and perforated plates; D) With external
recycle tube
also Fig. 34 B and C).
26 Bubble Columns

Submerged nozzles supply the gas in the time distribution of a stirred tank. Back-mixing
loop flow. The penetration depth of the jet dif- is suppressed in the second section (perforated
fers. The reactor is thus easier to start up and plates). For higher-order reactions, the conver-
the circulation flow can be build up in a sim- sion of both liquid and gaseous components can
pler way. This is particularly important in batch be increased in this way. With pure gases, re-
processes. In principle, submerged nozzles rep- cycle ensures adequate distribution because the
resent an intermediate stage between surface gas flow rate is sufficient for an even gas load of
sparging (Fig. 32 A and B) and pressurized gas the perforated plate.
sparging through nozzles at the bottom of the Figure 34 D shows a reactor (analogous to the
reactor (Fig. 34). airlift loop of Fig. 22 D) with external recircu-
In the gas- circulation reactor of Figure 33, lation. This device has an operational behavior
gas sparging takes place through the free surface. comparable to that of reactors with internal cir-
If a momentum-transfer tube is used (Fig. 33 B), culation. Its advantages include better gas sepa-
gas enters the circulating flow at a greater depth. ration and simpler heating or cooling facilities.
This arrangement functions even when the liq- The accessible heat-transfer area is larger, and
uid level drops and the momentum-transfer tube a conventional heat exchanger can be integrated
extends into the gas space (Fig. 33 C). In Fig- directly into the loop.
ure 33 D, the tube is led outside the reactor, so The most important characteristics of jet
that gas can be delivered directly from outside loop reactors with upward-pointing nozzles can
without any mixing with recycle gas. The last be summed up as follows:
two types of gas sparging (Fig. 33 C and D) are
1) They are particularly suitable for higher gas
also referred to as dip-tube sparging designs.
throughputs or when the gas has a high con-
Jet Loop Reactors with Upflow Liquid Jet tent of inerts
(Fig. 34). In the second major variant of the jet 2) They have larger interfacial areas than bub-
loop reactor, the nozzle points upward. This de- ble columns, especially with system with
sign is closely related to the airlift loop reactor hindered coalescence
(Section 2.14). The liquid jet mainly produces 3) They offer more intensive back-mixing than
smaller gas bubbles so that conversion of the bubble columns
gas phase can be improved. On the other hand, 4) They are less suitable for suspension cataly-
the circulation velocity is increased. At least a sis, because the nozzle can become plugged
portion of the gas is thereby driven through the
reactor faster than in airlift reactors, with the
possible result of lower conversion; this danger 4.2. Typical Dimensions
exists particularly in the reactor of Figure 34 A,
which uses a jet nozzle. The widened separation In jet loop reactors, height-to-diameter ratios
zone at the top can reduce gas recirculation or ht /d t of 5 to 20 are common. When several
cut it off altogether; the result is improved mix- nozzles are used along with internal tubes, ar-
ing of the liquid phase because the circulation bitrarily small height-to-diameter ratios can be
velocity increases with increasing gas holdup achieved. Values larger than 20 are also seen in
in the downcomer. Gas recycle is recommended high-pressure operation and in pilot plant reac-
whenever part of the gas leaves the column un- tors.
converted; it can be accomplished without a me- The optimal diameter d i of the internal tube
chanical recycle-gas compressor if an ejector or is dictated by the direction of flow. If flow in
an ejector jet nozzle is employed (Fig. 34 B and the tube is downward, tubes with a diameter ra-
C). The ejector jet nozzle offers a higher com- tio d i /d t of 0.2 – 0.5 are suggested [116], [119].
pression efficiency than the ejector and produces The narrower internal tubes have the principal
larger interfacial areas [12]. From the chemi- advantages of higher gas holdup and better mass-
cal reaction engineering standpoint the combi- transfer at low energy dissipation rate (up to
nation of a loop reactor with a series of perfo- 1 kW/m3 ). As jet power increases, the wider
rated plates is very interesting (Fig. 34 C). The tubes become better in these respects [119],
lower part (loop reactor) exhibits the residence- [123].
 Bubble Columns 27

For upward flow, Blenke and coworkers de- 2) Power dissipation per unit volume due to
termined the optimal internal tube diameter as friction between phases (slip power):
d i /d t = 0.59 [124]; this result applies to single-
and two-phase reactors. eslip = −∆gεG (1−εG ) vrG (4.6)
A crucial parameter of the jet loop reactor
is the nozzle diameter d n . Common values are A combination of these terms gives the bal-
in the range of d n /d t = 0.02 – 0.1. For a given ance equation:
jet power, large nozzles are more efficient than
ethrust +eG +eL +eslip +ecirc = 0 (4.7)
smaller ones for coalescing systems. In systems
with hindered coalescence, in contrast, this re- In general, the quantities vL, i (i.e., ecirc ) and εG
lationship can be reversed. (i.e., eslip ) are unknown, so direct evaluation is
impossible.

4.3. Energy Balance

4.4. Mixing Behavior and Fluid
In the general case of Figure 31 D (combina-
Dynamics
tion of bubble column, downflow bubble col-
umn, and jet loop reactor), a total of five me-
The flow processes in jet loop reactors are par-
chanical power terms can be identified [125].
ticularly crucial for the mixing and residence-
These can be referred to the reaction volume
time distribution of both phases. The investiga-
π 2 tion of the relationship between mixing time and
VR = d hR (4.1)
4 t liquid circulation time showed that complete ho-
Three types of power are delivered to the reac- mogenization of the liquid phase requires ca. ten
tor: passes [126].
The residence-time distribution of the liquid
1) Jet thrust power ethrust per unit volume: in the jet loop reactor has been investigated both
 2
experimentally and theoretically [127], [128].
vn2 vL,i dn Values of dispersion coefficients in the liquid are
ethrust =fi L (4.2)
hR dt presented in [118], [128]. The crucial parameter
for mixing is the internal circulation flow gen-
where f i is the fraction of cross-sectional
erated by the liquid let in through the nozzle.
area of the inner tube and vn is the nozzle
In certain cases, this quantity can be estimated
velocity
from the energy balance (Eq. 4.7).
2) Power input per volume due to liquid flow
The residence-time distribution of the dis-
eL (as in a downflow bubble column) :
persed gas phase can be found in [129]. The
dispersion coefficients are not substantially dif-
eL = ∆gεG uL (4.3)
ferent from those of the liquid phase.
3) Power input per unit volume due to gas flow
eG (as in a bubble column; holds approxi- Fluid Dynamics of Single-Phase Flow. The
mately only for small pressure changes): simplest formulation is that of single-phase flow.
For eG = 0, eslip = 0, eL = 0, the following holds
eG ≈ ∆g (1−εG ) uG (4.4) for liquid circulation flow


This power is transformed to heat by two 2
mechanisms: vL,i dt = vn dn (4.8)
ζfi

1) Power dissipation per unit volume due to cir- The drag coefficient of circulation flow ζ can
culation flow: be found in [118], [130–132]. Typical values
3
for bottom gas feed are ζ = 4; for top gas feed,
vL,i
ecirc = −ζ fi ¯ (4.5) ζ ≈ 0.25 – 2.
2hR
 28 Bubble Columns

Fluid Dynamics of Two-Phase Flow. In circulation velocity is almost independent of jet

two-phase flow the momentum balance can be velocity. Only at larger values does vL, c increase
solved for the circulation flow vL, i only if the linearly with vn as in single-phase flow [133–
gas holdup εG is known, which is not normally 135]. Investigations on radial velocity profiles
the case. Only in the gas- circulation reactor in loop reactors are found in [135].
(Fig. 32 A, B) can the gas holdup be definitely
specified within a certain range. Under the as-
sumption of low gas holdup εG , for example,
the following implicit relation is reported [123]:

(vn dn )2 =
 
ζ fi 2 ∆ εG (vrG −uL ) −uG
vL,i + ghR d2t (4.9)
2 L vL,i

The resistance coefficients ζ obtained for single-

phase flow can be used to a good approximation.
Figure 35 shows a logarithmic plot of Equation
(4.9) for uL = 0 and uG = 0. The liquid velocities
in the two-phase regime are always lower than
in single-phase flow. If the velocity goes below a
minimum value, the flow becomes unstable and
stops, as shown by the nonlinear behavior of the
curves at low velocities. Gas sparging then takes Figure 35. Circulation flow rate in gas-circulation reactor
place only in the upper part of the internal tube. as a function of jet conditions and gas holdup, calculated
These features are illustrated in Figure 35. The with Equation (4.9)
Assumptions: ζ = 2.5; vrG = 0.23 m/s; Geometric dimen-
√minimum flow velocities vL, i, min are a factor of sions: d t = 0.3 m; d h = 0.015 m; ht = 2.0 m; f i = 0.25
3 smaller than the single-phase velocities. This
means that all possible velocities in two-phase
flow are in the range
1 vL,i dt
√ ≤ ≤1
3 2
vn dn
ζfi

The minimum flow velocity can also be ob-

tained from Equation (4.9):


3 ∆/L
vL,i,min = ghR [εG (vrG −uL ) −uG ] (4.10)
ζfi

At high gas velocities uG or when the nozzle is at

the bottom of the reactor (Fig. 34 A), gas holdup
cannot be freely selected. Instead, it adjusts it-
self as a function of fluid-dynamic conditions.
At present, flow velocities cannot be calculated
in advance. Figure 36. Measured circulation flow in pressure-sparged
Bohner’s measurements for the jet loop re- jet loop reactor as a function of jet conditions, with gas ve-
locity as parameter [133]
actor are plotted in Figure 36 in the form d t = 0.29 m; ht = 2.0 m; f i = 0.35

v̄L,c dt =f (vn dn ,uG )

4.5. Gas Holdup
[133], [134]. Two regions can be identified. At
low jet velocities vn , the jet loop behaves like In the gas-circulation reactor (Figs. 31 and 32 A
an airlift loop reactor (Section 2.14). The liquid and B), the gas holdup can be set arbitrarily
 Bubble Columns 29

within certain limits. For a given jet velocity vn , tiny amounts of methanol reduce the tendency
the maximum values εG, max can be calculated to coalesce. The primary gas bubbles generated
from the power input by the liquid jet by the jet retain approximately the same size
L 3 π
over the whole apparatus because they do not
Pn = v Fn where Fn = d2n (4.11a) coalesce. This reduces their slip velocity vrG so
2 n 4
that much higher gas holdups and interfacial ar-
or the jet power per unit volume eas can be obtained.
 2
Pn L vn3 dn
en = = · (4.11b)
VR 2 hR dt

The equations reported by Tebel and Zehner

are presented here in simplified form [123].
The maximum specific energy dissipation rate
caused by the slip between the two phases
(Eq. 4.6) is given by


4 2 dn
eslip,max = en (4.12)
31.5 ζfi dt

which can be solved as follows for the maximum Figure 37. Comparison of measured and calculated maxi-
gas holdup: mum gas holdup as a function of specific jet power in gas-

circulation reactor
2 2 2en ddn System: water – air with added methanol; measured values
t
εG,max = · after [123]; calculations with Equation (4.13)
31.5 ζfi ∆gvrG

d t = 0.14 m; ht = 1.32 m; d i = 0.055 m; d n = 4.9 mm
2 2 2Pn ddn
t
= · (4.13) More complex relationships apply if pres-
31.5 ζfi ∆gvrG VR
surized gas is let in at the bottom of the gas-
Equation (4.13) is compared with measured val- circulation reactor. The gas holdup then sets it-
ues in Figure 37. The lowest gas holdups are ob- self, analogously to behavior in a bubble column,
tained in the coalescing system water – air. Even an airlift loop reactor (Sections 2.9 and 2.14),

Figure 38. Gas holdup measured in the pressure-sparged jet loop reactor [133] versus jet conditions at various gas velocities
ht = 2 m; d t = 0.29 m; f i = 0.33
30 Bubble Columns

or the ordinary loop reactor (Fig. 34 A). Typical processes. In the easily coalescing water – air
measured values for this type of gas sparging are system, the bubble diameter corresponds essen-
presented in Figure 38. As the gas rate increases, tially to the local energy dissipation rate, a quan-
gas holdup increases rapidly. The jet velocity, by tity that is distributed very unevenly over the vol-
contrast, has relatively little effect. ume in jet loop reactors. Zones of particularly
Zehner and Thelen obtained the expres- high energy-dissipation rate include the imme-
sion diate action region of the liquid jet and, to a lesser
  degree, the regions of loop flow reversal. This is
eslip eG the reason why the smallest bubbles are observed
=f (4.14)
2en ddn 2en ddn near the jet in the compact reactor (see Fig. 32 D)
t t
[120]. As Figure 40 shows, bubble size decreases
for jet loop reactors with pressurized gas sparg-
from ca. 3 mm to almost 2 mm with increasing
ing [125]. Figure 39 shows the gas holdup for
energy dissipation rate Pn /V R . In the other much
different reactor types and sparging types, based
larger regions, the bubbles quickly coalesce to
on this relation. To within measurement er-
bigger (3 – 4 mm) units. Jet loop reactors accord-
ror, the same values are obtained for the gas-
ingly do not feature smaller air-in-water bubbles,
circulation reactor with pressurized gas sparging
on average, than bubble columns or downflow
(Fig. 31 B) and for the jet loop reactor (Fig. 34).
bubble columns operated in the homogeneous
These values are fitted well by the correlation
flow regime. For a given gas holdup, compara-
 0.8 ble mass-transfer should therefore be expected.
eslip eG
= 1.5 (4.13a) The relationships among the mean bubble diam-
2en ddn 2en ddn
t t eter (Sauter diameter) d bS , the gas holdup εG ,
which can be solved directly for the gas holdup: the specific interfacial area a, and the volumetric
mass-transfer coefficient k L a have been pointed
 0.2  0.8
2Pn ddn uG out in Section 3.3. These considerations lead to
t
εG = 1.5 (4.13b) the following guideline values for the water – air
∆gvrG VR vrG
system:

Figure 39. Relationship between slip power slip , jet power

n , and gas compression power G for determination of gas
holdup Figure 40. Bubble sizes in the compact reactor (system:
a) Surface sparging in gas-circulation reactor; b) Transition water – air)
between surface and pressure sparging; c) Pressure sparging d t = 0.1 m; ht /d t = 5.9; d n = 5 mm; d i /d t = 0.6
in gas-circulation reactor and normal jet loop reactor

Usual specific energy dissipation rates:

4.6. Mass Transfer Pn /V R ≈ 1 – 10 kW/m3
Mean bubble diameter (Sauter diameter):
The mean bubble diameter in a sparged reactor is d bS ≈ 3.5 mm
always the result of distribution and coalescence Maximum gas holdup: εG ≈ 6 – 30 %
 Bubble Columns 31

Special interfacial area: a ≈ 100 – 600 m−1 the maximum possible gas holdup in a gas-
Volumetric mass-transfer coefficient: circulation reactor with gas sparging at the sur-
k L a ≈ 0.04 – 0.2 s−1 face [76]. As Figure 41 shows, the solids content
εS has a strong effect that cannot be accounted
If liquid mixtures and ionic or detergent so- for merely by the additional slip power dissipa-
lutions exhibit a noncoalescing behavior, much tion of fine particles. Instead, the slurry must be
higher gas holdups and smaller bubble diameters assumed to yield larger gas bubbles because of
(significantly less than 1 mm) can be achieved. its higher apparent viscosity. The higher slip ve-
The volumetric mass-transfer coefficients and locity of these larger bubbles might then explain
specific interfacial areas behave in a similar way; the marked dependence on solids concentration.
in systems with hindered coalescence they may Räbiger [138] and Wachsmann [139] also
be a factor of 5 to 10 higher than the water – air studied the effect of solids. At constant energy
values. A compilation of volumetric mass-trans- dissipation rate and constant volumetric gas flow
fer coefficients in various types of sparged ap- rate, the dependences are roughly similar to
paratus can be found in [136]. those found in [76]. Only for low solids and gas
holdups does a slight increase in gas holdup oc-
cur relative to two-phase systems.
This has been explained by bubble breakup
by large solid particles [142]. The effect should
not occur below the critical Weber number
S dS vb2
W e= =3
σ
However, this statement partially contradicts
Räbiger’s results [138]. Technically, these dis-
crepancies are insignificant. Up to a solids con-
centration of 10 vol %, no major differences exist
Figure 41. Effect of solids on gas holdup in the gas-
circulation reactor in gas holdup and mass-transfer between normal
System: 1 wt % NaCl – air – glass spheres; d p = 0.075 – and three-phase jet loop reactors. Higher volu-
0.15 mm metric particle concentrations do not, however,
normally occur in jet loop reactors.
4.7. Three-Phase Loop Reactor
When solids are suspended in sparged loop re- 5. References
actors the same engineering considerations are
necessary as in slurry bubble columns (Section 1. H. Gerstenberg, Chem.-Ing.-Tech. 47 (1975)
no. 5, 209.
2.13). Many workers have studied the effect of
2. M. Bobik, Chem.-Ing.-Tech. 53 (1981) MS
solid particles on fluid dynamics and mass-trans-
922.
fer performance [76], [137–139]. 3. BASF, EP 254 180, 1991 (P. Zehner et al.).
The fluid-dynamic principles of solid – liquid 4. Hoechst, DE-OS 2 855 263, 1978 (A. Riedel et
systems are comparable to those of gas – liquid al.).
systems (Sections 4.3 and 4.4) [132], [140]. 5. R. Steiner, Chem. Eng. Process. 21 (1987)
These considerations have been extended to 1 – 8.
threephase systems so that theoretical models 6. W.-D. Deckwer, J. Hallensleben, M. Popovic,
are available for complex multiphase flow [120], Can. J. Chem. Eng. 58 (1980) 190.
[123], [141]. In principle, the solid phase is ac- 7. H. Buchholz, Dissertation, Universität
counted for by a supplemental energy term for Hannover 1979.
the slip power dissipation of the particle swarm 8. K. Ruff, T. Pilhofer, A. Mersmann,
(Section 4.3). How interactions between phases Chem.-Ing.-Tech. 48 (1976) no. 9, 759 – 764.
influence individual slip velocities is not clear. 9. G. Neubauer, T. Pilhofer, Chem.-Ing.-Tech. 50
For example, Kürten and coworkers examined (1978) no. 2, 115 – 116.
32 Bubble Columns

10. Mersmann, A., Ger. Chem. Eng. (Engl. 36. H. Wezorke, Dissertation, Universität
Transl.) 1 (1978) 1. Dortmund 1986.
11. P. Krötsch, Fortschr. Verfahrenstech. Abt. D 37. R. Beinhauer, Dissertation, TU Berlin 1971.
16 (1978) 389. 38. Y. T. Shah, G. J. Stiegel, M. M. Sharma, AIChE
12. P. Zehner, Chem.-Ing.-Tech. 47 (1975) no. 5, J. 24 (1978) 369 – 400.
209. 39. M. Kraume, P. Zehner, Chem.-Ing.-Tech. 61
13. H. Gerstenberg, Chem.-Ing.-Tech. 51 (1979) (1989) no. 4, 332 – 333.
no. 3, 208 – 216. 40. K.-H. Mangartz, T. Pilhofer, VT,
14. Y. T. Shah, B. G. Kelkar, S. P. Godbole, W. D. Verfahrenstech. 14 (1980) no. 1, 40 – 44.
Deckwer, AIChE J. 28 (1982) 353 – 379. 41. P. Zehner, G. Schuch, Ger. Chem. Eng. (Engl.
15. J. B. Joshi, Y. T. Shah, Chem. Eng. Commun. Transl.) 8 (1985) 282 – 289.
11 (1981) 165 – 199. 42. G. Blasey, Dissertation, Universität Dortmund
16. K. Rietema, S. P. P. Ottengraph, Trans. Inst. 1987.
Chem. Eng. 48 (1970) T 54. 43. P. Herbrechtsmeier, R. Steiner,
17. J. R. Crabtree, J. Bridgwater, Chem. Eng. Sci. Chem.-Ing.-Tech. 52 (1980) 468.
24 (1969) 1755. 44. W. Hikita et al., Chem. Eng. J. 20 (1980)
18. S. M. Bhavraju, T. W. F. Russel, H. W. Blanch, 59 – 67.
AIChE J. 24 (1978) 454. 45. R. Nottenkämper, A. Steiff, P. M. Weinspach,
19. K. Ueyama, T. Miyauchi, AIChE J. 25 (1979) Ger. Chem. Eng. 6 (1983) 147 – 155.
258 – 266. 46. H. F. Bach, T. Pilhofer, Ger. Chem. Eng. (Engl.
20. J. B. Joshi, M. M. Sharma, Trans. Inst. Chem. Transl.) 1 (1978) 270. H. Buchholz,
Eng. 57 (1979) 244. R. Buchholz, H. Niebeschütz, K. Schügert, Eur.
21. P. Zehner, Ist. Chem. Eng. 26 (1986) no. 1, J. App. Microbiol. Biotechnol. 6 (1978) 115.
22 – 35. 47. K. Schügerl, J. Lücke, U. Oels, Adv. Biochem.
22. A. Lübbert: “Mass Transfer with Chemical Eng. 7 (1977) 1. G. Keitel, U. Onken,
Reactions in Multiphase Systems,” NATO Chem.-Ing.-Tech. 54 (1982) no. 3, 262 – 263.
Advanced Study Institute, Cesme/Izmir, Aug. 48. K. Akida, F. Yoshida, Ind. Eng. Chem. Proc.
10 – 11, 1981. Des. Dev. 13 (1974) 84 – 91.
23. J. H. Hills, Trans. Inst. Chem. Eng. 52 (1974) 49. H. Kölbel, E. Borchers, H. Langemann,
1 – 9. Chem.-Ing.-Tech. 33 (1961) no. 10, 668 – 675.
24. E. Kojima et al., J. Chem. Eng. Jpn. 13 (1980) W.-D. Deckwer, Y. Luisi, A. Zaidi, M. Ralek,
16. Ind. Eng. Chem. Proc. Des. Dev. 10 (1980)
25. M. Kraume, P. Zehner, German/Japanese 699.
Symposium Bubble Columns, Schwerte 1988, 50. M. H. Oyevaar, K. R. Westerterp, Chem. Eng.
preprints. Process. 25 (1989) 85 – 98.
26. R. Buchholz, K. Schügerl, Eur. J. Appl. 51. M. H. Oyevaar, R. Bos, K. R. Westerterp,
Microbiol. Biotech. 6 (1979) 301 – 315. Chem. Eng. Sci. 46 (1991) 1217 – 1231.
27. E. Blaß, Chem.-Ing.-Tech. 60 (1988) no. 12, 52. P. M. Wilkinson, L. v. Dierendonck, Chem.
935 – 947. Eng. Sci. 45 (1990) 2309 – 2315.
28. P. H. Calderbank, Chem. Eng. 45 (1976) CE 53. A. Serizawa, I. Kataoka, I. Michiyoshi, Int. J.
209. Multiphase Flow 2 (1975) 235 – 246.
29. O. Nagel, B. Hegner, H. Kürten, 54. W.-D. Deckwer, Reaktionstechnik in
Chem.-Ing.-Tech. 50 (1978) no. 12, 934 – 944. Blasensäulen, Salle und Sauerländer, Aarau
30. H. Unno, I. Inoue, Chem. Eng. Sci. 35 (1980)
1985.
1571.
55. U. Oels, J. Lücke, R. Buchholz, K. Schügerl,
31. K. Akita, F. Yoshida, Ind. Eng. Chem. Proc.
Ger. Chem. Eng. (Engl. Transl.) 1 (1978) 115.
Des. Dev. 13 (1974) 84 – 91.
56. W.-D. Deckwer, R. Burchhart, G. Zoll, Chem.
32. T. Miyahara, Y. Matsuba, T. Takahashi, Int.
Eng. Sci. 29 (1974) 2177.
Chem. Eng. 23 (1983) 517 – 523.
57. Y. Kawase, B. Halard, M. Moo-Young, Chem.
33. T. Miyahara, N. Naga, T. Takahashi, Int.
Eng. Sci. 42 (1987) 1609 – 1617.
Chem. Eng. 23 (1983) 524 – 531.
58. H. Hikita et al., Chem. Eng. J. 22 (1981)
34. K. Koide et al., Chem. Eng. Jpn. 12 (1979)
61 – 69.
98 – 104.
59. J. S. Cho, N. Wakao, J. Chem. Eng. Jpn. 21
35. A. Mersmann, Chem.-Ing.-Tech. 61 (1989),
(1980) 576 – 581.
no. 2, 97 – 104.
 Bubble Columns 33

60. S. P. Godbole, A. Schumpe, Y. T. Shah, N. L. 86. J. Kratochvil, K. Winkler, J. Zahradnik: Coll.

Carr, AIChE J. 30 (1984) 213 – 220. Czech. Chem. Commun. 50 (1985) 48 – 60.
61. K. Akita, Int. Chem. Eng. 29 (1989) 127 – 135. 87. K. Muroyama, German/Japanese Symposium
62. W. Kast, Int. J. Heat Mass Transfer 5 (1962) Bubble Columns, Schwerte 1988, preprints.
329 –336. 88. “Cyclohexane,” Hydrocarbon Processing
63. W.-D. Deckwer, Chem. Eng. Sci. 35 (1980) 1977, no. 11, 144.
1341 –1346. 89. U. Onken, P. Weiland, Adv. Biotechnol.
64. J. B. Joshi et al., Chem. Eng. Commun. 6 Processes 1 (1983) 67 – 95.
(1980) 257 –271. 90. M. Y. Chisti, Airlift Bioreactors, Elsevier
65. A. N. Khoze et al., J. Eng. Phys. 20 (1971) Applied Science, London 1989.
759 –763. 91. P. Verlaan, J. C. Vos, K. van’t Riet: J.
66. L. S. Aksel’rod, N. I. Vorotnikova, A. A. Chem. Technl. Biotechnol. 45 (1989)
Kozlov, Heat Transfer-Sov. Res. 8 (1976) 109 – 121.
25 – 33. 92. R. A. Bello, C. W. Robinson, M. Moo-Young,
67. R. Wendt, Dissertation, Universität Dortmund Can. J. Chem. Eng. 40 (1984) 573 – 577.
1983. 93. P. Weiland, U. Onken, Chem.-Ing.-Tech. 52
68. H.-J. Korte, Dissertation, Universität (1980) 986 – 987.
Dortmund 1987. 94. P. Weiland, Ger. Chem. Eng. (Engl. Transl.) 7
69. H.-J. Korte, A. Steiff, P.-M. Weinspach, (1984) 374 – 385.
German/Japanese Symposium Bubble 95. Y. Kawase, M. Moo-Young, Chem. Eng.
Columns, Schwerte 1988, preprints. Commun. 40 (1986) 67 – 83.
70. R. Wendt, A. Steiff, P.-M. Weinspach, 96. T. Reith, Ph. D. Thesis, Universität Delft 1968.
Chem.-Ing.-Tech. 55 (1983) no. 10, 796 – 797. 97. W. Hirner, H. Blenke, VT, Verfahrenstech. 11
71. Y. T. Shah, C. A. Ratway, H. G. McIlvried, (1977) no. 5, 297 – 303.
Trans. Inst. Chem. Eng. 56 (1978) 107 – 112. 98. K. Koide, H. Sato, S. Iwamoto, J. Chem. Eng.
72. K. Bernemann, Dissertation, Universität Jpn. 16 (1983) 407 – 413.
Dortmund 1989. 99. K. Koide et al., J. Chem. Eng. Jpn. 16 (1983)
73. K. Koide, T. Yasuda, S. Iwamoto, E. Fukuda, 413 – 419.
J. Chem. Eng. Jpn. 16 (1983) 7 – 12. 100. K. Nakao et al., German/Japanese Symposium
74. S. Kara et al., Ind. Eng. Chem. Proc. Des. Dev. Bubble Columns, Schwerte 1988, preprints.
21 (1982) 584 – 594. 101. K. Schügerl, Bioreaktionstechnik, vol. 2, Salle
75. Y. Kato et al., Ist. Chem. Eng. 13 (1979) und Sauerländer, Aarau 1991.
562 – 567. 102. P. K. Ouyoung, M. Y. Chisti, M. Moo-Young,
76. H. Kürten, P. Zehner, Ger. Chem. Eng. (Engl. Chem. Eng. Res. Des. 67 (1989) 451 – 456.
Transl.) 2 (1979) 220 – 227. 103. G. Müller, G. Sell, DECHEMA-Monogr. 86
77. T. Sauer, Dissertation, Universität Paderborn (1980) 589.
1986. 104. S. J. Heijnen et al., Chem. Eng. Technol. 13
78. A. Yasunishi, M. Fukuma, K. Muroyama, J.
(1990) 202 – 208.
Chem. Eng. Jpn. 19 (1986) 444 – 449.
105. Y. Bando et al., J. Chem. Eng. Jpn. 21 (1988)
79. K. Koide, A. Takazawa, M. Komura, H.
no. 6, 607 – 612.
Matsunaga, J. Chem. Eng. Jpn. 17 (1984)
106. A. Ohkawa, Y. Shiokawa, N. Sakai, J. Chem.
459 – 466.
Eng. Jpn. 18 (1985) no. 6, 172 – 174.
80. T. Matsumoto, N. Hidaka, S. Morooka, AIChE
107. A. Ohkawa, Y. Shiokawa, N. Sakai, N. Imai, J.
J. 35 (1989) 1701 – 1709.
Chem. Eng. Jpn. 18 (1985) no. 6, 466 – 469.
81. H. Mürtz, Dissertation, Universität Dortmund
108. P. Herbrechtsmeier, H. Schäfer, R. Steiner,
1990.
Chem.-Ing.-Tech. 56 (1984) MS 1221.
82. M. Fukuma, K. Muroyama, A. Yasunishi, J.
109. P. Herbrechtsmeier, H. Schäfer, R. Steiner,
Chem. Eng. Jpn. 20 (1987) 28.
Chem.-Ing.-Tech. 53 (1981) MS 884.
83. E. Sada, H. Kumazawa, C.-H. Lee, H.
110. P. Herbrechtsmeier, R. Steiner,
Nakamura, Ind. Eng. Chem. Proc. Des. Dev.
Chem.-Ing.-Tech. 50 (1978) no. 12, 944 – 947.
26 (1987) 112.
111. A. Kulkarni, Y. T. Shah, Chem. Eng. Commun.
84. K. Nguen-Tien, H. Patwari, A. Schumpe,
28 (1984) 311 – 326.
W.-D. Deckwer, AIChE J. 31 (1985) 194.
85. H.-G. Springmann, A. Steiff, P.-M. Weinspach, 112. A. Ohkawa et al., J. Chem. Eng. Jpn. 20
Chem.-Ing.-Tech. 63 (1991) 162 – 163. (1987) no. 1, 99 – 101.
34 Bubble Columns

113. W. J. Albrecht, G. Schulze-Pillot, 129. H.-J. Warnecke, D. Vaupel, I. Prüss, H.

Chem.-Ing.-Tech. 50 (1987) no. 22. Langemann, Chem.-Ing.-Tech. 61 (1989
114. R. Steiner, Chem. Eng. Process. 21 (1987) no. 10, 815 – 817.
1 – 8. 130. H. Blenke, K. Bohner, S. Schuster,
115. R. Steiner, DECHEMA-Monogr. 94 (1983) Chem.-Ing.-Tech. 37 (1965) no. 3, 289 – 294.
164 – 197. 131. H. Blenke, K. Bohner, W. Pfeifer,
116. H. Kürten, P. Zehner, Annual Conference of Chem.-Ing.-Tech. 43 (1971) no. 1 – 2, 10 – 17.
Chemical Engineering, Straßburg, Oct. 1980. 132. P. Zehner, Habilitationsschrift, Universität
117. BASF, EP 0 087 670, 1983 (H. H. Schuster, H. Kaiserslautern 1988; Fortschr.-Ber. VDI Z.
Dreher, J. Hambrecht). Reihe 3, 1988, no. 160.
118. H. Blenke in: Loop Reactors, Advances in
133. K. Bohner, Dissertation, Universität Stuttgart
Biochemical Engineering 13, Springer-Verlag,
1971.
Berlin 1979.
134. K. Bohner, H. Blenke, Verfahrenstechnik
119. N. Räbiger, A. Vogelpohl, Chem.-Ing.-Tech.
(Mainz) 6 (1972) no. 2, 50 – 57.
53 (1983) MS 1111.
120. N. Räbiger, Habilitationsschrift, Technische 135. D. K. Jain et al., Can. J. Chem. Eng. 68 (1990)
Universität, Clausthal 1988; Praxiswissen 1047 –1051.
thermische Verfahrenstechnik, Verlag TÜV 136. H.-J. Warnecke, M. Geisendörfer, D.
Rheinland, Köln 1988. Ch. Hempel, Chem. Eng. Technol. 11 (1988)
121. H. Loosen, ZFL 40 (1989) no. 9, 494 – 496. 306 – 311.
122. B. Lohrengel, A. Vogelpohl, Chem.-Ing.-Tech. 137. L.-S. Fan, S.-J. Hwang, A. Masuura, Chem.
62 (1990) MS 1856. Eng. Sci. 39 (1984) no. 12, 1677 – 1688.
123. K. H. Tebel, P. Zehner, Chem. Eng. Technol. 138. N. Räbiger, Chem.-Ing.-Tech. 57 (1985) no. 3,
12 (1989) 274 – 280. 248 –249.
124. H. Blenke, K. Bohner, W. Hirner, 139. U. Wachsmann, N. Räbiger, A. Vogelpohl,
Verfahrenstechnik (Mainz) 3 (1969) no. 10, VDI-Ber. 607 (1986) 1537 – 1566. U.
444 – 452. Wachsmann, N. Räbiger, A. Vogelpohl, Ger.
125. P. Zehner, H.-G. Thelen: Chem. Eng. (Engl. Transl.) 8 (1985) 411 – 418.
“Mehrphasenströmung,” Vortrag auf dem 140. P. Zehner, Chem.-Ing.-Tech. 62 (1980) MS
GVC-Fachausschuß, Bremen 1991. 850.
126. A. Jaramillo, I. Adler, A. Fiechter, Third 141. K. H. Tebel, P. Zehner, Annual Meeting of the
European Congress on Biotechnology, Process Engineers, Straßburg, Sept 17 – 19,
Munich, Sept. 10 – 14, 1984. 1986.
127. J. Lehnert, Verfahrenstechnik (Mainz) 6 142. J. C. Lee, 4th Proc. European Symp., Chem.
(1972) 58. React. Eng., 1971, p. 27.
128. H.-J. Warnecke, M. Weidenbach, J. Prüss, H.
Langemann, Chem.-Ing.-Tech. 59 (1987)
no. 6, 496 –499.

Bubble Memories → Information Storage Materials

Buffing → Abrasives
Building Bricks → Construction Ceramics