Biochemical Engineering Journal 27 (2006) 323–330

Mixing time in bioreactors under aerated conditions

Dimiter Hadjiev ∗ , Nour Eddine Sabiri, Adel Zanati
Laboratoire d’Etudes Thermiques, Energétiques et Environnement, Université de Bretagne Sud, site de Pontivy,
Allée des Pommiers, 56300 Pontivy, France

Received 10 December 2004; received in revised form 24 May 2005; accepted 2 August 2005

Abstract

The major difficulty needing a solution in agitated and aerated bioreactors is closely connected to the optimum hydrodynamics regime in
the stirred tank. It is evident that even a simple stirred reactor is geometrically complex and it seems that the mixing time can be one useful
criterion for estimation of mixing intensity and therefore the estimation of oxygen profiles and mass transfer coefficients. The mixing time
measures are relevant to the length of time for which detectable inhomogeneities last in the extremities of the vessel and can identify the
extent of possible concentration gradients in a reactor that should be used as a gradientless unit for bio-kinetic studies. The parameter depends
mainly on the mixing system characteristics, the physicochemical properties, the aeration and the rate of cell multiplication. To quantify the
influence of some of these factors experiments have been performed using a wide spread laboratory scale stirred dual-impeller bioreactor
configuration marketed by “Bioflow”. Experimental data have been obtained for various internal geometries of the reactor under aerated and
non-aerated conditions administering real wastewater. A multi-regression analysis method was used to find mathematical correlation relating
the influence of the rotation speed and the gas flowrate with the reactor’s capacity to mix.
© 2005 Elsevier B.V. All rights reserved.

Keywords: Bioreactor; Mixing time; Aeration

1. Introduction determine the effects of aeration on the power input [1–3] and
the types of interactions between the bubbles and the agitator
Mixing is indispensable in biotechnology, as it defines [4], while much less work has been conducted on the effects
the environment during the cultivation in bioreactors. The of gassing in multiple-impeller systems. Some studies [5]
development of the biomass in a reactor leads to continuous indicated that the mixing time in gassed systems is evermore
modifications of the medium. In these conditions, one of the longer than that in a single liquid phase, while Pedersen et al.
most important problems is to establish an optimum hydro- claimed the opposite [6]. Other results [7–9] demonstrated
dynamic regime. Mixing time contains useful information that the mixing time increases and decreases, according to
about flow and mixing within the vessel and can be useful the aeration rate and the turbine agitation speed. At higher
for biosynthesis processes scale-up. The mixing time, Tm , rotational speeds (loading and complete dispersion regime),
denotes the time required for the tank composition to achieve the mixing is mainly due to the impeller action, which, in
a specified level of homogeneity following addition of a tracer turn, is affected by the gas flow rate, G. As a result, Tm in this
pulse at a single point in the vessel. regime is larger than in unaerated liquids and increases with
The study of two-phase gas–liquid systems is of great an increase in the value of G. At loading–flooding transition
interest in biochemical engineering, since many fermenta- of the lowermost impeller, according to Abardi et al. [10],
tion processes employ a gas phase. Numerous studies have the pumping efficiency of the stirrer (especially the bottom
taken place with laboratory scale and large-scale reactors to stirrer) is quite low and a circulation loop promoted by the
bubbling of the gas probably prevails. For the calculation of
∗ Corresponding author. Tel.: +33 2 9727 6770; fax: +33 2 9727 8153. the mixing time, several equations have been proposed but
E-mail address: [email protected] (D. Hadjiev). their accuracy is quite limited. Furthermore, most of them

1369-703X/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.bej.2005.08.009
324 D. Hadjiev et al. / Biochemical Engineering Journal 27 (2006) 323–330

This analysis shows the lack of information about homog-

Nomenclature enization rates in aerated stirred laboratory-scale reactors and
the need to develop some comprehension of mixing under
AF agitation factor (according to Guillard and variable geometry conditions. The objective of this paper is
Trägardh) to study the effect of geometry, agitation speed and gas flow
a, b, c, e, f, h, i, j, k, m, n, p, q, r numerical coefficients rate on the mixing time in a wide spread lab-scale stirred
in Eq. (3) and Eq. (4) dual-impeller bioreactor marketed by “Bioflow” uncovering
B baffle width its specific mixing dynamics. The mixing time measures are
B clearance between the baffle and the wall of important because they are relevant to the length of time for
the reactor which detectable inhomogeneities last in the extremities of
D tank diameter the vessel and can identify the extent of possible concentra-
d impeller diameter tion gradients in a reactor that should be used as a gradientless
d impeller spacing unit for bio-kinetic studies. Further the dimensional analysis
d top clearance methodology was used to correlate the experimental data.
G volumetric gas flow rate To check the validity and generalize the proposed correlation
g gravitational acceleration some estimated mixing times are compared to the experimen-
H liquid height tal values obtained with two-phase industrial bioreactors.
K, K1 constants in Eq. (3) respect to Eq. (4)
l impeller blade length
MF mixing factor NTm /NTmo 2. Experimental
N impeller rotation speed
nb number of blades 2.1. Laboratory scale reactor and operating conditions
nimp number of impellers
nbaf number of baffles The experiments were carried out in a Bioflow 3000
P power, required to drive the agitator batch dual turbine-stirred reactor with a capacity of 3 L, with
Tm mixing time computer-controlled and recorded parameters (see Fig. 1).
V volume of the reactor The main geometrical characteristics are given in Table 1.
w impeller blade width The experiments were carried out using distilled water and
X variable defined by Eq. (6) genuine wastewater. The operating conditions are reported in
Y bottom clearance

Greek letters
εT local energy dissipation rate
η liquid kinematic viscosity
ρ liquid density

can predict this parameter for Re > 10,000, while transient

regime can also be observed in large-scale and laboratory-
scale reactors. Recently, Guillard and Trägardh [11] proposed
a new analogy mixing–time model (AMTM) for estimating
mixing times in industrial turbine-agitated aerated reactors.
It is based on the hypothesis of the existence of an analogy
between the mixing time in a single liquid phase and that
in a gassed liquid phase when the same hydrodynamic con-
ditions are applied. They found that the internal geometry
of the reactor does not affect the model. It should be men-
tioned, however, that all validations of this model have been
performed for industrial reactors with preliminary optimized
geometry. This geometry was not modified during experi-
ments while, in the same time, numerous studies [9] point
at the fact that the flow patterns are a strong function of the
clearance from the bottom, spacing between impellers, clear-
ance above the upper impeller and the gas sparging pattern,
etc. Fig. 1. Scheme of the bioreactor.
 D. Hadjiev et al. / Biochemical Engineering Journal 27 (2006) 323–330 325

Table 1
Characteristics of the bioreactor and the turbine impeller
d (mm) 64
d/D 0.36
H/D 1.15
N of baffles 4
N of blades 6
w/d 0.2
l/d 0.27
B/d 0.24
B /d 0.01
N of impellers 2

Table 2. The gas flowrate was measured using a Brooks “Read

Out and Control Electronics 0152” flowmeter and coupled
with a Brooks “Smart Mass Flow” regulator.
The values of the mixing time have been determined
by means of top injections of 2 M NaOH used as a tracer
(ρ = 1080 kg m−3 ). Top injection means that the tracer is
added at the top of the reactor 2 cm under the liquid surface Fig. 2. Example concentration variations registered by the technique
near the edge of a baffle and on the opposite to the pH elec- employed (N = 0.83 s−1 , d /d = 1.7, Y/d = 1, Q = 50 L h−1 ).
trode. All injections were performed with a syringe (injection
time of 1 s). trodes, but the temporal variations of concentration inside
Two pH electrodes were used as follows: one probe each zone are not identical. A rapid initial axial transport can
mounted at 20 mm from the vessel bottom; another probe be observed for both zones followed by a progressive slow
was mounted in the middle of the reactor. The first one was increase in pH for the middle zone, while an almost linear
connected to the bioreactor computer-recording system, the decrease can be detected in the bottom of the reactor. The
other one to a Sartorius Professional pH Meter PP 15. Each results given in Fig. 2 show that for some geometrical con-
experiment has been carried out several times, at identical figurations the compartments in a compartment model can
conditions, the average value for the mixing time being used. have different volumes and therefore different mixing dynam-
The maximum experimental error was of 3–5%. ics. Understanding the influence of the geometry factor is
From the mixing kinetics obtained, mixing times at 95% therefore of interest in biochemical engineering allowing the
were determined. These times are defined as the time required prediction of favorable operating conditions and improving
for the system to mix to a prescribed final state of mixture the bioreactor design.
represented by the degree of homogeneity, I. Therefore, the
following homogeneity criterion was applied:
pH∞ ± 0.05 pH 3. Results and discussion
I= × 100 = 95%
pH∞
In order to collect data for mixing time in different com-
with pH = pH∞ − pH0 (pH0 and pH∞ are the initial and partment volumes, both single and two phase flow cases were
the final pH values). studied. In all experiments, the dimensionless mixing time
NTm was determined with and without aeration.
2.2. Description of concentration temporal variation
3.1. Mixing in single liquid phase
In Fig. 2, curves recorded with both electrodes are shown.
The results show a similar delay of 7–10 s for both pH elec- The dimensionless mixing times have been determined
for various turbine rotation speeds and for different geom-
Table 2
Experimental conditions
etry. Two series of experiments have been carried out. In
the first one, the lower stirrer was placed on the shaft at a
Operating conditions Range
distance varying between 32 and 109 mm (0.5d and 1.72d)
Turbine Re number (Re) 1365–13653 from the upper one, while the clearance between the latter
Froude number (Fr) 7.23E−4–7.23E−2
d /d 0.5–1.7
and the liquid interface remained constant (32 mm—0.5d).
Y/d 1–2.2 The results given in Fig. 3 show the influence of the Re
d /d 0.5–1.7 number on the mixing number. The latter always increases
Aeration number (F = GN−1 d−3 ) 0.063–0.32 with Re and no significant influence of the geometry (d /d
Modified aeration number 0.34E−2–1.67E−2 and Y/d) can be detected. As the Re number determines the
(AF = GVr −1 N−1 nimp −1 )
flow pattern in the tank, it is obvious that in this case the
326 D. Hadjiev et al. / Biochemical Engineering Journal 27 (2006) 323–330

regime. Further, in transient regime, the dimensionless mix-

ing number was found to increase with an increase in the
impeller spacing, which may be attributed to the fact that
the flow fields of the impellers become more independent of
each other [12]. In turbulent flow, however, there is a quite
pronounced role of the distance to the interface, which can be
attributed to the location of the injection point (near the inter-
face). Fluid–fluid mixing is usually modelled by a structure
of mixed compartments. Each compartment has two circula-
tion loops, above and below the impeller and each loop can
be considered as an ideally mixed tank. Significant axial mix-
ing exists between the adjacent circulation loops. As the top
clearance increases, the volumes of the ideally mixed tanks
change. The mixing dynamics is therefore influenced by the
disproportion between the circulation loops. Furthermore, it
can be noticed that the results obtained with d /d = 1.5 show
a completely different comportment and the mixing number
steadily increases with N for N > 1.67 (100 rpm, Re > 5000).
This can be attributed to the quite equilibrated compartment
Fig. 3. Influence of the rotation speed on the mixing number at constant
volumes, i.e. circulation loops (d = 2d = 1.17Y). This is the
distance between the upper stirrer and the liquid interface.
reason for this geometry to be recommended in reactors with
parameters “d “and Y do not influence the hydrodynamic multiple impellers.
conditions at transient and low turbulent regimes. Further, In conclusion, the existing approaches for estimating mix-
the influence of the distance to the interface was studied. The ing times in non-aerated stirred reactors cannot be used for
upper stirrer was placed at a distance varying between 32 reactors equipped with multiple impellers. As stated above,
and 109 mm (0.5d and 1.72d) from the lower one, while the the mixing time is dependent on the turbulence-dependent
bottom clearance was maintained constant. The results are mass exchange between the compartments formed by the
presented in Fig. 4. No general behaviour can be observed impellers. More complex mixing models have to be used.
but two zones can be clearly outlined. At low impeller rota- These can be compartment models or networks-of-zones
tion speed (N < 2, Re < 10,000), corresponding to a transient models but they require complex calculations. An empirical
regime, the mixing numbers are always lower than those in model based on dimensional analysis can be a good approach
turbulent flow. The mixing number is therefore inversely pro- to the problem combining fairly good accuracy and rapid-
portional to the turbulent diffusion respective to the operating ity. All the elements concerning this empirical model will be
given in Section 4.

3.2. Mixing time in two phase systems

The influence of the operating regime on the mixing

number is presented in Fig. 5 for four gas flow rates. The
experiments were carried out at constant geometry (64 mm
bottom clearance, 109 mm clearance between the impellers
and 32 mm top clearance). As it can be seen, the mixing num-
ber is steadily increasing with the Re number. Furthermore,
it seems that the types of interactions between the bubbles
and the impellers and between the bubbles and the tank are
not modified despite a five times increase in the modified
aeration number (AF increases from 0.003 to 0.015). In fact,
according to Shukla et al. [13], low aeration numbers always
correspond to dispersion/loading operational regimes, i.e. the
impellers are effectively dispersing the gas.
In Fig. 6, the experimental mixing numbers are plotted
versus the gas flow rate for both transient and low turbu-
lent regimes. All experiments have been carried out at con-
stant geometry (78 mm bottom clearance, 96 mm between
Fig. 4. Influence of the rotation sped on the mixing number for various the impellers and 32 mm top clearance). As it can be seen,
distances between the upper stirrer and the liquid interface. in transient regime the increase in the gas flow rate does not
 D. Hadjiev et al. / Biochemical Engineering Journal 27 (2006) 323–330 327

Fig. 7. Influence of the gas flow rate on the mixing number for different
Fig. 5. Influence of the rotation speed on the mixing number for various gas distances between the upper stirrer and the liquid interface.
flow rates and constant geometry.
109 mm (0.5d and 1.7d) from the lower one. As it can be
influence significantly the mixing number. The influence is seen, aeration does not always affect the mixing number in
more important in turbulent regime (when Re > 10 000). In the same way and no general behaviour can be outlined. The
the same time the relationships NTm = f(G) for Re = constant mixing number always decreases for d /d = 1 and for d /d = 1.7
are complex. According to Mann [14] these results can be (top clearance, respectively, 78 and 32 mm), while it remains
related to the gas–liquid flow structure. When the variables independent for a top clearance of 57 mm. It is evident that
G and N are changed, transition from “bubble column shape” increasing the top clearance, the volumes of the ideally mixed
around the shaft of the turbine to complete dispersion via compartments change. The axial mixing between the adjacent
the formation of circulation zones can be observed. In this circulation loops in the upper section differs from the axial
case the latter can be detected for 45 m3 h−1 < G < 55 m3 h−1 mixing between the loops in the lower compartment. The
(0.007 < AF < 0.009). Thus, the effect of gassing on the mix- interstage liquid exchange flow rates are therefore modified.
ing time is dependent on the operating regime of the impeller. These results clearly support the role of the reactor’s geome-
At lower impeller speeds, a gas controlled regime of the tank try on the liquid axial mixing as already shown in Fig. 4. It is
is observed and the mixing time increases marginally or stays evident, that the situation becomes much more complex when
constant as shown in Fig. 6. gas is introduced and the rates depend also on the superficial
Fig. 7 contains the results for the influence of the impeller gas velocity.
spacing at a constant bottom clearance of 64 mm. The upper
impeller was placed on the shaft at distances from 32 to 3.3. Main theoretical approaches

3.3.1. The analogy mixing time model

According to Guillard and Trägardh [11] it is possible
to simplify the mathematical formulation of mixing time
predictions under aerated conditions assuming an analogy
between mixing in a single and a two phase flow. The anal-
ogy mixing–time model is therefore based on the analogy
between the mixing time in single liquid phase (without aer-
ation) and that in two gas–liquid phases. A mixing factor MF ,
is defined as the ratio between NTm with and without aeration.
For a top injection point this factor depends on the modified
aeration number as follows:
MF = 10−2.97 (AF )−0.96 (1)
The gas phase thus influences the mixing process such
that mixing times under aerated conditions are generally
Fig. 6. Influence of the gas flow rate on the mixing number for various shorter than mixing times in a single liquid phase. Accord-
rotation speeds and at constant geometry. ing to Manikowski et al. [15] the flow in the bulk exhibits
328 D. Hadjiev et al. / Biochemical Engineering Journal 27 (2006) 323–330

are calculated for a constant top clearance (d /d = 0.5) and

normalized to the number of impellers. As it can be seen the
clearance between the impellers influences considerably the
results because of its influence on the convective exchange
between the zones.
In summary, the two mixing number correlations tested
here cannot qualitatively estimate the influence of the reac-
tor’s geometry on the dimensionless mixing number NTm .
Predict the configurations giving shorter mixing times can
improve bioreactor design and would be of interest in bio-
chemical engineering. Such configurations cannot be deter-
mined using the above-mentioned models and therefore,
another approach is needed.

4. Correlations for mixing time

Fig. 8. Influence of the modified agitation factor AF on the mixing factor Most of the existing mixing models concern single

MF (‘* data normalised to the number of impellers).
impeller non-aerated systems in turbulent flow conditions
and consider generally one geometrical factor, which is the
zigzag motion, with secondary recirculation zones above and d/D ratio. Only the influence of the off-bottom clearance and
below the Rushton impeller, while separated regions between the spacing between impellers on the power consumption
each impeller can be clearly observed in single liquid phase. was studied. A more complete model given by Norwood and
These zigzag motions of the flow can explain the increase and Metzner [19] takes account of the influence of the Re and
the decrease of the mixing time in two phase systems com- Fr numbers and shows clearly the existence of three zones
pared to that obtained in non-aerated conditions. The patterns corresponding to the laminar, transitory and turbulent regime
of the bubbles induce the formation of convective oscilla- flow. In the latter, the mixing time number is independent on
tory motion in the liquid phase. According to Guillard and the Re number.
Trägardh the volumes around each impeller are like “open As it can be seen, dimensional analysis is frequently used
structures” which communicate with each other. Changing to assist in the interpretation of complex phenomena involv-
the distance between the impellers and the top clearance, the ing many variables. In the present instance it can be assumed
volumes of these open structures change and therefore the that the mixing time in an aerated and agitated vessel is gov-
convective exchange between the zones is limited. The results erned by a multitude of geometrical (H, D, d, d , d , Y, B, l,
presented in Fig. 8 and compared to Eq. (1) do not demon- w, nb , nimp , nbaf ) and technological (N, G, ρ, η) factors. The
strate a very good agreement especially for data obtained in analysis leads to the following dimensionless grouping and
transient regime. The slope of the line is modified but the variables.
critical value of AF for these geometric configurations (the
a




value of AF at which MF = 1) corresponds to that obtained by H d  b d  c Y e d f h
NTm = K nb
the authors for top and surface injections AF ≈ 0.8 × 10−3 . D d d d D

k
m  
j B l w n P q r
3.3.2. The turbulence model × niim nbaf Re Fr Fg (3)
Recently, Nienow [16] used the turbulence model outlined d d d
by Ruszkowski [17] and Grenville et al. [18] to describe many
results obtained independently using impellers with very dif- with NTm the dimensionless mixing number; Re = (Nd2 )/η
ferent d/D ratios and flow patterns over a very wide range of the impeller Reynolds number; Fr = (N2 d)/g the impeller
mean energy dissipation rates. He proposed Froude number; Fg = G/(Nd3 nim ) the modified aeration num-

−1/3 ber related to a single-impeller system; (H/D), (d /d), (d /d),
−1/3 d (Y/d), (d/D), (B/d), (l/d), (w/d) dimensionless geometric
Tm = AD (ε̄T )
2/3
(2)
D groups.
The use of a specific reactor requires maintaining some of
where (ε̄T ) is the local energy dissipation rate. He gives
the dimensionless geometric groups constant. Therefore, the
A = 5.9 and suggests that Eq. (2) suitably modified, i.e. using
general expression becomes:
(Pg ) or (ε̄T )g , respectively, may be used for aerated conditions
too, provided the impeller is not flooded. The comparison
b
c
e
d d  Y
between some experimental data obtained in turbulent flow NTm = K1 Rep Fr q Fgr (4)
with Eq. (2) is also presented in Fig. 8. The mixing times d d d
 D. Hadjiev et al. / Biochemical Engineering Journal 27 (2006) 323–330 329

Table 3
Experimental and estimated values
Operating conditions Reactor A
Tank diameter (m) 2.09
Type of injection Surface
Number of impellers 4
Impeller diameter, da 0.33
Impeller spacing, d /d 2.12
Bottom clearance, Y/d 1.52
Liquid height (m) 7.32
NTm exp b 63–38
NTm calc b 56–41
a Normalised with tank diameter.
b Normalised to the number of impellers.

constant top clearance K1 = 1.75, while it is equal to 1, for a

constant bottom clearance. If these values of K1 are used the
Fig. 9. Influence of X on the mixing time number. model offer a quite good agreement with the experimental
data, the average deviation being of ±13.5%.
After evaluating the coefficients b, c, e, p, q and r, by trial The proposed Eq. (5) was used to estimate the mix-
and error the final form of the equation has been determined: ing times in industrial reactors. Some experimental results

 2
   obtained by Guillard and Trägardh are given in Table 3 and
d d compared with Eq. (5) which can be seen to be in very good
NTm = −13.981 + 27.972 − 4.1327
d d agreement. It can be concluded that the proposed equation

−0.54 is not restricted only to laboratory scale agitated and aerated
d
× Re0.275 Fr 0.275 Fg−0.04 (5) units.
d
Fig. 9 shows the plot of the experimental mixing time num-
bers on: 5. Conclusions

 2
  
d d The influence of various parameters on the mixing time
X = −13.981 + 27.972 − 4.1327
d d in an aerated laboratory Rushton turbine-agitated reactor has

−0.54 been studied. The important role of the geometry has been
d demonstrated and the results have been related to the flow
× Re0.275 Fr 0.275 Fg−0.04 (6)
d structure and the compartmentalisation of the reactor. It was
found that some configurations allow shortening consider-
for a great number of experiments. ably the mixing time in the case of top injection, while
The obtained correlation does not show a significant influ- for other the latter is always found to be longer than under
ence of the modified aeration number Fg on the dimensionless non-aerated conditions. A new correlation for estimating the
mixing time. This result seems in agreement with the data mixing times in aerated stirred reactors has been proposed. It
obtained by Arjunwadkar et al. [20] concerning the influence takes into account the influence of the top and bottom clear-
of the gas flow number on the relative power input PG /P0 in ances and the distance between the impellers in the case of
multi-impeller systems. Apparently, in dispersion regime, the multiple-impeller systems. The fact that the operating condi-
parameter does not affect to an appreciable extent the relative tions cover a broad range of Re numbers (from transitory to
power input and, therefore, the mean energy dissipation rate turbulent flow), Fr numbers and Fg modified aeration num-
(ε̄T ) determined as P/ρV. Taking this analysis further, one bers, as well as, the comparison with results obtained in
can say that in a particular reactor geometry (d/D = constant, industrial reactors show that the derived correlation has a
d /d = constant) according to Eq. (5): relatively good general validity and can be used to scale-up
NTm ∝ d 1.5 the reactors geometry in the case of multi-impeller systems,
when the impellers are not flooded, i.e. are effectively dis-
The implications for improved mixing are similar to those persing the gas.
for the turbulence approach given by the generalized equation
for mixing time recommended by Nienow [16].
At the same time, Eq. (5) shows clearly that the change References
in the size of the “open structures” influences considerably
the slope of the straight line and therefore the coefficient K1 . [1] A.W. Nienow, D.J. Visdom, Flow over disc turbine blades, Chem.
When the distance between the impellers is modified at a Eng. Sci. 29 (1974) 1994.
330 D. Hadjiev et al. / Biochemical Engineering Journal 27 (2006) 323–330

[2] K. Van’t Riet, J.M. Boom, J.M. Smith, Power consumption, impeller [11] F. Guillard, Ch. Trägardh, Mixing in industrial Rushton turbine-
coalescence and recirculation in aerated vessels, Trans. Inst. Chem. agitated reactors under aerated conditions, Chem. Eng. Process. 42
Eng. 54 (1976) 124. (2003) 373.
[3] F.G. Bader, Mass transfer in a multiturbine fermentor. I. Gassed and [12] V. Hudcova, V. Manchon, A.W. Nienow, Gas–liquid dispersion with
ungassed power inputs, in: Proc. Int. Conf. Bioreactor Fluid Dyn., dual Rushton turbine impellers, Biotech. Bioeng. 34 (1989) 617.
Cambridge, England, 1986, p. 269. [13] V.B. Shukla, U.P. Veera, P.R. Kulkarni, A.B. Pandit, Scale-up of
[4] J.M. Smith, M.N.C.G. Warmoeskerken, The dispersion of gases in biotransformation process in stirred tank reactor using dual impeller
liquids with turbines, in: Proc. Fifth European Conf. Mix., Würzburg, bioreactor, Biochem. Eng. J. 8 (2001) 19.
Germany, 1985. [14] R. Mann, Gas–liquid stirred vessel mixers: towards a unified theory
[5] L.K. Ju, G.G. Chase, Improved scale-up strategies of bioreactors, based on networks-of-zones, Chem. Eng. Res. Des. 64 (1986) 23.
Bioprocess Eng. 8 (1992) 49. [15] M. Manikowski, S. Bodemeier, A. Lübert, W. Bujalski, A.W.
[6] A.G. Pedersen, J. Nielsen, J. Villadsen, Characterization of biore- Nienow, Measurement of gas and liquid flows in stirred tank reactors
actors using isotope tracer techniques, in: Proc. Sixth European with multiple agitators, Can. J. Chem. Eng. 72 (1994) 769.
Congress on Biotechnol., Elsevier Science BV, Amsterdam, 1994, [16] A.W. Nienow, On impeller circulation and mixing effectiveness in
p. 931. the turbulent flow regime, Chem. Eng. Sci. 52 (1997) 2557.
[7] J.M.T. Vasconcelos, S.S. Alves, A.W. Nienow, W. Bujalski, Scale- [17] S. Ruszkowski, A rational method for measuring blending perfor-
up of mixing in gassed multi-turbine agitated vessels, Can. J. Chem. mance and comparison of different impeller types, in: Proc. Eighth
Eng. 76 (1998) 398. European Mix. Conf., Inst. Chem. Eng., Rugby, UK, 1994, pp.
[8] P. Urabel, R.G.J.M. Vander Lans, Y.Q. Cui, K.C.A.M. Luyben, Com- 283–291.
partment model approach: mixing in large scale aerated reactors with [18] R.K. Grenville, S. Ruszkowski, E. Garred, Blending of miscible liq-
multiple-impellers, Chem. Eng. Res. Des. 77 (1999) 291. uids in the turbulent and transitional regimes, in: 15th NAMF Mixing
[9] P.R. Gogate, A.A.C.M. Beenackers, A.B. Pandit, Multiple-impeller Conference 1995, Banff, Canada, 1995.
systems with special emphasis on bioreactors: a critical review, [19] A.B. Metzner, K.W. Norwood, Flow patterns and mixing rates in
Biochem. Eng. J. 6 (2000) 109. agitated vessels, AIChE J. 6 (1960) 432.
[10] V. Abadi, G. Rovero, S. Sicardi, G. Baldi, R. Conti, Hydrodynamics [20] S.J. Arjunwadkar, K. Saravanan, A.B. Pandit, P.R. Kulkarni, Opti-
of a gas–liquid reactor stirred with multi-impeller systems, Trans. mizing the impeller combination for maximum hold-up with mini-
Inst. Chem. Eng. 68 (Part A) (1990) 516. mum power consumption, Biochem. Eng. J. 1 (1998) 25.