Sie sind auf Seite 1von 9

Biochemical Engineering Journal 8 (2001) 111–119

Volumetric oxygen transfer coefficients (kL a) in batch

cultivations involving non-Newtonian broths
A.C. Badino Jr. a , M.C.R. Facciotti b , W. Schmidell b,∗
a Department of Chemical Engineering, Universidade Federal de São Carlos, P.O. Box 676, SP 13565-905 São Carlos, Brazil
b Department of Chemical Engineering, Escola Politécnica da Universidade de São Paulo,

P.O. Box 61548, SP 05424-970 São Paulo, Brazil

Received 21 June 1999; accepted 9 January 2001

The oxygen transfer in non-Newtonian fermentation broths of Aspergillus awamori, during batch cultivations in conventional 10 l
bioreactor has been investigated. Values of the volumetric oxygen transfer coefficient (kL a), obtained at various impeller speeds, air flow
rates and at distinct initial substrate concentrations were correlated with operational variables, geometric parameters of the system and
physical properties of the broths utilising rigorous techniques in order to obtain a set of reliable and accurate data. An experimental device
was constructed for on-line rheological measurements, and the apparent dynamic viscosity was determined from the broth rheograms. In
order to measure power requirements, a torque meter was developed and non-Newtonian fluids with rheological characteristics similar to
the Aspergillus fermented broths were utilised to obtain reference curves and correlations in the fermentor where the cultivations took place.
Gas balancing method and a modified dynamic method were utilised simultaneously to determine kL a values. The rigorous methods thus
developed allowed adequate evaluation of the oxygen transfer in the cultivations and also permitted good fits of four different traditional
correlations for kL a to the experimental data. © 2001 Elsevier Science B.V. All rights reserved.
Keywords: Volumetric oxygen transfer coefficient; Power consumption; Broth rheology

1. Introduction taken when applying these correlations to non-Newtonian

systems containing electrolytes such as fermentation broths
Fermentation broths containing mycelial cells frequently [3].
exhibit a pseudoplastic non-Newtonian rheological be- Two types of correlations have been proposed for the vol-
haviour, which can be described by the power-law model. umetric oxygen transfer coefficient (kL a). The first does not
This behaviour exerts a profound effect on the bioreactor make use of any dimensional criterion. In these correlations,
performance, affecting mixing pattern, power requirement, kL a is related to the gassed power consumption per unit vol-
heat and mass transfer processes [1]. The increase in the ume of broth (Pg /V) and the superficial gas velocity (v s ), as
broth apparent viscosity (µap ) during aerobic fermentations originally proposed by Cooper et al. [4]:
can be partially compensated by increments in the operating  a1
conditions (N and Q), in order to maintain adequate kL a kL a α (vs )b1 (1)
values. Nevertheless, high impeller speeds (N) lead to the V
formation of high shear zones close to the impellers, with where the values of the constants a1 and b1 may vary con-
consequent physical damage to the cells and a reduction in siderably, depending on the system geometry, the range
the process productivity [2]. Due to the importance of the of variables covered and the experimental methodology
volumetric oxygen transfer coefficient (kL a) in the perfor- used. Although initially developed for fluids very distinct
mance and scale-up of conventional bioreactors, the litera- from fermentation broths, this type of correlation has been
ture describing various correlations for kL a in Newtonian widely used in fermentation systems [3,5–7]. In a more
fluids is rather extensive. However, particular care should be recent work, Montes et al. [8] determined values of kL a in
yeast broths (Trigonopsis variabilis) over wide ranges of
∗ Corresponding author. Tel.: +55-11-818-2284; fax: +55-11-211-3020. both impeller speeds and superficial gas velocities in three
E-mail addresses: (A.C. Badino Jr.), different mechanically-stirred, sparger-aerated and baffled (W. Schmidell). bioreactors (2, 5 and 15 l) in order to consider the effect

1369-703X/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S 1 3 6 9 - 7 0 3 X ( 0 1 ) 0 0 0 9 2 - 4
112 A.C. Badino Jr. et al. / Biochemical Engineering Journal 8 (2001) 111–119

Nomenclature YCO
carbon dioxide molar fraction
a1 , a 2 , a 3 constants of Eqs. (1), (2) and (8) (–) in outlet gas (–)
b1 , b2 , b3 constants of Eqs. (1), (2) and (8) (–) Greek letters
c2 , c3 constants of Eqs. (2) and (8) (–) α constant of proportionality
C dissolved oxygen concentration in the φ air specific air flow rate (vvm: air volume
broth in steady-state (mmol O2 l−1 ) per broth volume per minute) (min−1 )
C∗ dissolved oxygen saturation γ av average shear rate (s−1 )
concentration in the broth (mmol O2 l−1 ) µap apparent dynamic viscosity (Pa s)
d3 constant of Eq. (8) (–) µg air viscosity (Pa s)
Di impeller diameter (m) µw water viscosity (Pa s)
Dt tank diameter (m) ν ap apparent kinematic viscosity (m2 s−1 )
Fr (=Di N2 /g) Froude number (–) ρ density (kg m−3 )
GLA glucoamylase activity (U l−1 ) σ surface tension of filtrate (N m−1 )
k constant of Eq. (10) (–) σ∗ (= σ/(ρ(νap 4 g)1/3 )) Eq. (8) (–)
ke electrode sensitivity (h−1 or s−1 ) τ av average shear stress (Pa)
kL a volumetric oxygen transfer coefficient
(h−1 or s−1 )
K consistency index (Pa sn ) of the fermentor scale-up on kL a. Experimental data were
n flow behaviour index (–) fitted using the correlation proposed by Cooper et al. [4]
ṅin inlet dry molar gas flow rate and the values for the parameters a1 , b1 and proportionality
(mol O2 h−1 ) constant were 0.35, 0.41 and 3.2 × 10−3 , respectively. Due
ṅout outlet dry molar gas flow rate to the fact that most of the yeast broths behave as slightly
(mol O2 h−1 ) non-coalescent fluid, according to the authors, the corre-
N impeller speed (rpm) lation improved the prediction of kL a values with respect
NP power number (–) to other generic correlations usually developed for strong
P0 ungassed power consumption (W) coalescent and non-coalescent fluids. Extended forms of
Pg gassed power consumption (W) this first type of relationship have been proposed in the
Pgi gassed power consumption per literature, incorporating terms like impeller speed (N) and
impeller (W) the fluid apparent viscosity (µap ). For example, Ryu and
(Pgi /Q)∗ (= (Pgi /Q)/(ρ(gvap )2/3 )) Humphrey [9] observed the influence of the broth appar-
dimensionless group of Eq. (8) (–) ent viscosity (µap ) on kL a in penicillin fermentation and
Q air flow rate (m3 s−1 ) proposed the following correlation:
QO2 specific oxygen uptake rate  a2
(mmol O2 g−1 h−1 ) Pg
kL a α (vs )b2 (µap )c2 (2)
O2 maximum specific oxygen uptake V
rate (mmol O2 g−1 h−1 )
where again, the constants a2 , b2 , c2 as well as the
QO2 X global oxygen uptake rate
proportionality constant, depend on the geometry of the sys-
(mmol O2 g−1 h−1 )
tem. More recently, Garcı́a-Ochoa and Gómez [10] utilised
R2 correlation coefficient (–)
this correlation to fit experimental data obtained in a stirred
Rem modified Reynolds number (–)
tank reactor of 20 l of working volume. The rheology of the
S substrate concentration (g l−1 ) xanthan gum solutions was described both in terms of the
S0 initial substrate concentration (g l−1 ) power-law and the Casson models. Experiments were per-
Sc (= µap /(ρDO2 )) Schmidt number (–) formed considering the effect of the number and type of stir-
Sh∗ kL aD2i /DO2 , modified Sherwood rers (paddle or turbine), the number of blades of the stirrers
number (–) and the sparger type (ring and disk) in the volumetric oxygen
Si∗ dimensionless group in Eq. (7) (–) transfer coefficient (kL a). In the same way, Gavrilescu et al.
vs (= 4Q/(π Dt2 )) superficial gas [1] have assessed the problem of oxygen transfer in antibiotic
velocity (m s−1 ) broths produced by nine different microorganisms belonging
V broth volume (l or m3 ) to the actinomycetes and the fungi. The following equation
Vi broth volume per impeller (m3 ) for kL a was established from a wide variety of experimental
X cell concentration (g l−1 ) data obtained in 20 and 100 m3 stirred tank bioreactors:
YOin2 oxygen molar fraction in inlet gas (–)  0.4  0.65
Pg 0.5 µap
YOout oxygen molar fraction in outlet gas (–) kL a = 0.025 (vs ) (N )
V µg
A.C. Badino Jr. et al. / Biochemical Engineering Journal 8 (2001) 111–119 113

Additionally, Shin et al. [11] verified that in high cell others since the variable Pgi represents the gassed power
density cultures of fast-growing aerobes such as recombi- consumption per impeller, this equation can be applied to
nant E. coli where the cell mass increases up to more than systems with different numbers of impellers and, therefore,
70 g/l, the oxygen availability can be the rate-limiting of with different geometries. This possibility was verified by
the fermentation process. In that way, the following corre- Jurecic et al. [3], investigating submerged cultures of Bacil-
lation for kL a incorporating the effect of cell density (X) in lus licheniforms for bacitracin production in bioreactors
oxygen transfer has been proposed: of 100 l and 67.5 m3 with two and four turbine impellers,
 0.55 respectively.
Pg Although several correlations have been developed for
kL a = 0.0192 (vs )0.64
V different systems, most of them are not specific to fermenta-
× (1 + 2.12X + 0.20X 2 )−0.25 (4) tion broths or, when developed with such a purpose, they do
not determine all the variables throughout the cultivation.
The second type of correlation for kL a is based on dimen- Surface tension of the liquid (σ ), for instance, is considered
sional analysis. This approach presents certain advantages to be constant throughout the whole cultivation and, there-
because the correlations obtained for a known system can fore, it is not considered in the correlations. Furthermore, in
be used to estimate kL a in other systems with different the development of kL a correlations in cultivations involv-
dimensions. Yagi and Yoshida [12] proposed the following ing filamentous microorganisms, particular care should be
correlation, valid for both Newtonian and non-Newtonian taken when measuring some variables, like the rheological
fluids, for the standard six-bladed turbine in standard con- properties of the broth and the kL a itself, when traditional
figuration in a vessel of 0.25 m diameter: equipment and methodologies are utilised. In recent works,
Badino Jr. et al. [15], as well as Svihla et al. [16], verified
Sh∗ = 0.060(Rem )1.50 (Sc)0.50 (Fr)0.19 that the use of conventional bench rheometers, such as the
 µ v 0.60  ND 0.32 concentric cylinder, to determine rheological properties of
ap s i
× (5) viscous mycelial suspensions, is often unsatisfactory. The
σ vs
main problems are caused by pellet’s size, generally of the
This correlation fitted to the experimental data in a same order of magnitude as the annulus, as well as the ten-
reasonable extent, with maximum deviations of 30 and dency of the suspension to become heterogeneous due to
80%, between experimental and calculated Sh∗ values, settling and particle interaction. Thus, they propose the use
for Newtonian and non-Newtonian fluids, respectively. In of rotational rheometer provided with a helical impeller to
order to determine the combined effects of the variables, obtain reliable and accurate rheological data for filamentous
including operational conditions, liquid and gas properties fermentation broths. In addition, Olsvik and Kristiansen
and geometry of the tank, Garcı́a-Ochoa and Gómez [10] [17] concluded that on-line continuous measurement of
employed a similar dimensionless equation to fit the set broth rheology enables the uniform treatment of the samples
of experimental data of kL a obtained in the same system and statistically more “correct” measurements.
under operational conditions above specified. Li et al. [13] The aim of this work is to investigate the influence of
proposed a relationship among kL a, the impeller speed (N) operating conditions on the rheological properties of the
and the apparent viscosity (µap ), for fermentation broths fermented broth and on the oxygen transfer during batch
of Aureobasidium pullulans and Xanthomonas campestris cultivations of A. awamori in a bench scale fermentor.
in 100 l fermentor, which correlated quite well the great Methodologies for on-line rheological measurements and
variety of experimental data, as follows, respectively: for determination of both power consumption and kL a have
  been developed in order to obtain a set of reliable and accu-
µap −0.33
Sh∗ = 5.40 × 104 (Fr)0.45 (6) rate data. Four classical correlations mentioned previously
µw (Eqs. (1), (2), (5) and (8)) for the volumetric oxygen transfer
  coefficient (kL a), including the operating conditions, geo-
µap −0.58
Sh∗ = 3.32 × 103 (Fr)0.23 (7) metrical parameters of the system and physical properties
µw of the broth (apparent viscosity and also surface tension of
the filtrated broth), were fitted to the experimental values.
Zlokarnik [14] suggested, also through dimensional analysis,
a general dependence of kL a on physico-chemical properties
and operating conditions, as follows: 2. Materials and methods
Q Pgi ∗ a3
kL a α (Sc)b3 (σ ∗ )c3 (Si∗ )d3 (8) 2.1. Microorganism and culture media
Vi Q
where the dimensionless variable Si∗ , not exactly defined, A spore suspension (2×107 spores per millilitre) of A.
is related to the coalescence behaviour of solutions. How- awamori NRRL 3112, stored in cryotubes (glycerol 20%
ever, this correlation presents a feature not found in the v/v) at −20◦ C, was used throughout this work.
114 A.C. Badino Jr. et al. / Biochemical Engineering Journal 8 (2001) 111–119

Growth medium was composed of (in g/l): total reducing Dissolved oxygen concentration was measured by a
sugars (TRS) from cassava flour and assayed as glucose, sterilisable galvanic electrode (Mettler-Toledo InPro6000
20.0; yeast extract, 0.2; MgSO4 ·7H2 O, 1.0; (NH4 )2 SO4 , Series), bearing a 0.0254 mm thick FEP (fluorinated ethy-
10.0; Na2 HPO4 ·12H2 O, 7.56; KH2 PO4 , 7.0, pH 5.0. Two lene propylene) membrane. Readings of the oxygen (YO2 )
fermentation media with different compositions were and carbon dioxide (YCO2 ) molar fractions in the outlet
utilised. The first had the same composition as the growth gas were furnished by gas analysers produced by Beckman
medium and the second had initial TRS concentration (S0 ) Industrial (model 755) and Fuji Electric (model IR-730),
of 40.0 g/l, with addition of silicone antifoam (0.08 ml/l) to respectively. The experimental apparatus is shown in Fig. 1.
2.3. Analytical determinations and
2.2. Fermentation equipment and cultivation procedure surface tension measurement

Nine batch cultivations were conducted at 35◦ C and pH Cell concentration (X) was evaluated as dry matter
5.0 in a 10 l Microferm MF-14 fermentor (New Brunswick obtained after vacuum filtration and drying at 85◦ C for 6 h.
Sci. Co. Inc., USA) provided with four baffles in order to Substrate concentration (S) was determined as equivalent
favour turbulence and to prevent vortex formation. For each glucose utilising Glucose Autoanalyser II (Bran-Luebbe),
fermentation, spores were initially germinated in five 1 l after the polysaccharide in the sample had previously been
Erlenmeyer flasks, each holding 200 ml of growth medium. converted to glucose through an enzymatic hydrolysis [18].
Flasks were inoculated with 1 ml of the spore suspension Liquid surface tension of the filtrated broth (σ ) was mea-
from the cryotubes and inoculated in a rotary shaker (New sured with a Fisher Scientific Co. ring tensionmeter.
Brunswick Sci.) for 24 h at 35◦ C and 300 rpm. The con-
tents of the flasks were transferred to the fermentor with 2.4. Rheological properties
9 l of fermentation medium, performing a total of 10 l of
fermentation broth. The cultivations were carried out at dif- Continuous on-line rheological measurements were
ferent operational conditions. The ranges of impeller speed recorded every 2 h of cultivation, utilising the experimen-
(300 < N < 700 rpm) and specific air flow rate (0.2 < tal device constructed and calibrated by Badino Jr. et al.
φ air < 1.0 vvm) are frequently utilised during filamentous [15] based upon the original reports of Kemblowski and
fungal cultivations in bench-scale bioreactors, in order to Kristiansen [19]. In order to determine flow curves of fer-
provide adequate mixing and oxygen supply to the cells. mentation broths, a rotational helical ribbon impeller was
The assays were designated as follows: E20-7-3, E20-5-6, connected to Brookfield rheometers, models LV-DVIII and
E20-7-9, E40-5-2, E40-5-6, E40-5-10, E40-7-2, E40-7-6 RV-DVIII (Brookfield Eng. Lab. Inc., USA), which mea-
and E40-7-10. The label E20-7-3, for instance, specifies sured the impeller torque (T) at different rotational impeller
the following experimental conditions: S0 = 20 g/l, N = speeds (N). The average shear rate, γ av , and the average
700 rpm and φ air = 0.3 vvm, respectively. Samples were shear stress, τ av , around the impeller were calculated from
withdrawn each 3 h for analytical determinations. the calibration equations, γ av = 28.91N and τ av = 10530T .

Fig. 1. Experimental apparatus.

A.C. Badino Jr. et al. / Biochemical Engineering Journal 8 (2001) 111–119 115

For these continuous on-line rheological measurements, The inlet molar gas flow rate (ṅin ) was measured utilis-
the fermentation broth was pumped out of the bioreactor ing a mass flowmeter model 5811-N (Brooks Instrument
through silicone rubber tubing and into the impeller rheome- Division), and the outlet molar gas flow rate (ṅout ) was
ter by a peristaltic pump. The outlet of the impeller shaft calculated from the nitrogen mass balance in the gas phase
in the rheometer was provided with a rotating labyrinth by the following equation:
seal. Aseptic conditions were ensured by introducing sterile
air through the glass tubing next to the head of the ves- ṅout = (13)
sel. The average residence time in the loop was fixed at 1 − YOout
90 s, in order to allow good measurements, without causing
The volumetric oxygen transfer coefficient (kL a) could be
particle settling or even disturbing the performance of the
calculated for the steady-state condition, from the mass
n ) was fitted to on-line balance for dissolved oxygen as follows:
The power law model (τav = Kγav
experimental data and the rheological properties, con- QO2 X
sistency index (K) and flow behaviour index (n), were kL a = (14)
C∗ − C
where C is the dissolved oxygen concentration in the
2.5. Power consumption broth at the steady-state given by the galvanic electrode
and C∗ is the dissolved oxygen saturation concentration in
Ungassed and gassed power consumptions were esti- the broth estimated by the method proposed by Schumpe
mated under the previously mentioned different operating et al. [23].
conditions (N and Q), every hour during cultivation, by us- The volumetric oxygen transfer coefficient (kL a) was
ing traditional reference curves and correlations to interpret also evaluated through an alternative method, based on the
data from a purpose-built torque meter based on flat strain traditional dynamic method taking into consideration the
gauges of 120 # resistance. The new device was connected electrode sensitivity (ke ) [24]. The electrode sensitivity (ke )
to the stirrer shaft inside the fermentor, to avoid frictional is defined as the inverse of the time taken for the electrode
losses in the mechanical seal. In order to obtain the reference to reach 63.2% of its final value when exposed to a step
curves and correlations, non-Newtonian fluids of similar change in oxygen concentration. After each cultivation the
rheological characteristics to the fermentation broths were electrode sensitivity (ke ) was determined in triplicate as-
used [20]. Ungassed power consumption (P0 ) was obtained says utilising xanthan gum solution 0.4% w/v at 35◦ C to
through a graph of power number (NP ) as a function of simulate the fermentation broth.
modified Reynolds number (Rem ), as proposed by Metzner
et al. [21] for non-Newtonian pseudoplastic fluids.
3. Results and discussion
NP = (9)
ρN 3 Di5 3.1. Rheological behaviour of fermentation broths
ρN 2−n Di2
Rem = (10) With the average shear rate (γ av ) ranging from 0.96 to
K(k n−1 ) 72.27 s−1 , all fermentation broths at different cultivation
where k = 11.4, under the conditions considered. times, behaved as non-Newtonian pseudoplastic fluids.
Gassed power consumption (Pg ) was estimated through a Good fits between the power law model and the experimen-
traditional correlation, proposed by Michel and Miller [22], tal data were achieved, showing that on-line rheological
obtained for the experimental system used in cultivations: measurement is a consistent and reliable technique for
 0.44 mycelial cultivations. Fig. 2 illustrates typical flow curves
P02 ND3i (rheograms) obtained at different times for the run E40-7-6.
Pg = 0.832 (11)
Q0.56 Significant changes in the rheological parameters, fluid
consistency index (K) and flow behaviour index (n) as
2.6. Oxygen transfer depicted the Fig. 3 for the run E40-7-6, were observed as
the fermentations proceeded: K increased rapidly while n
Global oxygen uptake rate (QO2 X) as well as the volu- decreased to values between 0.3 and 0.5. Interestingly, an
metric oxygen transfer coefficient (kL a) were obtained by added feature of the on-line technique was its ability to ac-
both steady-state and dynamic methods. By the gas balanc- curately anticipate the instant at which the culture medium
ing method, QO2 X was determined from an oxygen mass is depleted of carbon source, especially when measure-
balance in the gas phase by the following expression: ments are taken frequently. As can be observed in Fig. 3,
immediately after the carbon source is exhausted (∼27 h),
ṅin YOint2 − ṅout YOout the K value decreases suddenly, probably as a consequence
QO2 X = 2
(12) of both hyphae fragmentation and lysis.
116 A.C. Badino Jr. et al. / Biochemical Engineering Journal 8 (2001) 111–119

and Kargi [26], both cultivating Aspergillus niger strains,

utilising glucose as the main carbon source, found Qmax O2
values in the range 3–4 mmol O2 g−1 h−1 , which agrees very
well with the present results.
With respect to kL a determinations, the simultaneous use
of the two measuring methods allowed reliable determina-
tions of kL a over a wide range (60.1–768 h−1 ) showing that
the gas balancing and the alternative dynamic method taking
into account the electrode response time, should be used in
a complementary manner: while the gas balancing method
provides accurate measurements mainly after the cell con-
centration reaches higher values, the alternative dynamic
method is particularly useful to evaluate kL a at the begin-
ning of the fermentation, when the differences in oxygen
Fig. 2. Typical flow curves obtained at different times of the run E40-7-6 composition between inlet and outlet gas are quite small and
(10 h: τav = 0.019γav0.60 , 16 h: τ = 0.261γ 0.43 , 22 h: τ = 0.837γ 0.47
av av av av
and 30 h: τav = 1.554γav 0.39 ). the levels of dissolved oxygen concentration are very high,
a situation at which the gas balancing and other methodolo-
gies are often inadequate. Nevertheless, dynamic methods
for determinations of kL a are useful only when the dissolved
oxygen concentration is fairly above the critical value (Ccrit );
for lower values, the gas balancing method would be more
appropriate [24].
It is known that besides the agitation (N) and aeration (Q)
conditions, a variable that negatively affects kL a is the broth
apparent viscosity (µap ), because it offers resistance to the
oxygen transfer from the gaseous to the liquid phase. Val-
ues of apparent viscosity (µap ) varied between 0.013 and
0.257 Pa s and the estimated maximum value of ungassed
power consumption (P0 ) was 187 W. The strong influence
Fig. 3. Variation in substrate concentration (S) and rheological parameters of the broth apparent viscosity (µap ) on the volumetric oxy-
K and n during run E40-7-6. gen transfer coefficient (kL a) during experiment E40-5-10 is
shown in Fig. 5. It can be observed that kL a decreased from
3.2. Characteristics of oxygen uptake and oxygen transfer 381 to 124 h−1 when µap increased from 0.013 to 0.257 Pa s
in the time interval from 0 to 37 h of cultivation. In addi-
Typical time course profiles of specific oxygen uptake rate tion, in Fig. 5 the time course of the relative surface ten-
(QO2 ) and of dissolved oxygen concentration (C) obtained sion of the filtrated broth (σ /σ w ) is depicted, where σ w =
during an experiment with oxygen limitation (E40-5-2) is 7.04.10−2 N m−1 is the surface tension of water at 35◦ C
shown in Fig. 4. Considering all runs, the maximum values [27]. The relative surface tension of the filtrated broth pre-
of specific oxygen uptake rate (Qmax sented an ascendant behaviour, varying from 0.897 to 0.943
O2 ) ranged from 2.8 to
−1 −1 for the same time period considered before. Such variation is
4.5 mmol O2 g h . Metwally et al. [25], as well as Shuler
relatively modest and it is probably due to the consumption

Fig. 4. Time profiles of specific oxygen uptake rate (QO2 ) and of dissolved Fig. 5. Influence of apparent viscosity (µap ) and relative surface tension
oxygen concentration (C) for the experiment E40-5-2. of filtrated broth (σ /σ w ) on the kL a during run E40-5-10 (γ av = 95 s−1 ).
A.C. Badino Jr. et al. / Biochemical Engineering Journal 8 (2001) 111–119 117

Table 1
Correlations for the volumetric oxygen transfer coefficient (kL a)
Correlation Reference Regression Experimental
coefficient (R2 ) values
kL a = 41.44(Pg /V )0.47 (vs )0.39 Cooper et al. [4] 0.90 286
kL a = 19.89(Pg /V )0.53 (vs )0.43 (µap )−0.12 Ryu and Humphrey [9] 0.92 305
 µ v −0.02  ND −0.25
ap s
Sh∗ = 1.45(Rem )0.83 (Sc)0.43 (Fr)0.74
Yagi and Yoshida [12] 0.94 316
σ vs
Q Pgi ∗ 0.53
kL a = 1.45 (Sc)−0.20 (σ ∗ )−0.40 Zlokarnik [14] 0.96 318
Vi Q

of the dissolved nutrients from the original culture medium

as the fermentation proceeds.

3.3. Correlations for kL a

As mentioned before, four classical correlations have been

utilised to correlate the volumetric oxygen transfer coef-
ficient (kL a) with the operation variables (N, Q, and Pg ),
geometric parameters of the system (V and Di ) and phys-
ical properties of the broths (µap and σ ), as described by
Cooper et al. [4], Ryu and Humphrey [9], Yagi and Yoshida
[12] and Zlokarnik [14]. The correlations were fitted to
the experimental values and the parameters were estimated Fig. 7. Comparison of the experimental data with those calculated using
through Marquardt’s procedure that utilises the least squares the correlation proposed by Ryu and Humphrey [9] (DM: dynamic method,
non-linear regression [28]. The criterion for the best fit and GB: gas balancing for kL a determination).
parameter optimisation was the sum of squares of residuals
(SSR). No convergence problems were found in non-linear
regression and all confidence intervals, computed for the The data in Table 1, as well as the figures, show that good
95% of probability, were below 25% of the estimated values. fits were obtained for all the correlations, indicating that any
The fitted correlations, together with the respective correla- of the adjusted models can be used to estimate, with good
tion coefficients (R2 ) and numbers of experimental values precision, the volumetric oxygen transfer coefficient (kL a)
used in the regressions, are presented in Table 1. Figs. 6–9 in these cultivations.
illustrate the quality of the fits. However, it can also be observed that the more complex
The simultaneous use of the two methods, gas balancing is the correlation, also the better is the quality of the fit
(GB) and dynamic method (DM), allowed a great number of (higher R2 values). The correlation proposed by Cooper
kL a values to be obtained as previously discussed, with the et al. [4], for instance, being less complex because it only
obtention of similar values in the same operating conditions. considers the effects of the gassed power consumption per

Fig. 6. Comparison of the experimental data with those calculated using Fig. 8. Comparison of the experimental data with those calculated using
the correlation proposed by Cooper et al. [4] (DM: dynamic method, GB: the correlation proposed by Yagi and Yoshida [12] (DM: dynamic method,
gas balancing for kL a determination). GB: gas balancing for kL a determination).
118 A.C. Badino Jr. et al. / Biochemical Engineering Journal 8 (2001) 111–119

parameters obtained from the correlation of Zlokarnik [14],

presented values similar to those obtained by Jurecic et al.
[3], employing non-Newtonian fermentation broths of B.
licheniformis in pilot and industrial scales (a3 = 0.35, b3 =
−0.30 and c3 = −0.50). Also, it should be emphasised that
although a lot of correlations for kL a include dimensionless
terms, in the literature investigated no works have been
found that present measurements of parameters such as the
surface tension of the filtrate (σ ) during cultivation.

4. Conclusions

Fig. 9. Comparison of the experimental data with those calculated using The main objective of this work was to investigate the
the correlation proposed by Zlokarnik [14] (DM: dynamic method, GB: influence of operating conditions on the rheological prop-
gas balancing for kL a determination). erties of the broth and on the oxygen transfer during batch
cultivations of A. awamori in a bench scale fermentor. Four
unit volume (Pg /V) and the superficial gas velocity (v s ) on classical correlations found in the literature, which correlate
kL a, fitted the data less well than those correlations that the volumetric oxygen transfer coefficient (kL a) with the
consider the effects of the physical properties of the broth process operational variables (N, Q and Pg ) and the geomet-
on kL a. Despite its simplicity, the correlation proposed by ric parameters of the system has been utilised. In particular,
Ryu and Humphrey [9] gave also a good fit, practically as in some of these correlations, kL a is also correlated with
good as the more complex correlations proposed by Yagi the physical properties of the broth (apparent viscosity and
and Yoshida [12] and by Zlokarnik [14] that need the knowl- liquid surface tension), which are time dependent during
edge of additional variables of the broth, such as apparent the fermentation process.
viscosity (µap ) and surface tension (σ ), for convenient ap- Good fits were obtained for all the correlations tested,
plication. However, if we intend to use a kL a correlation in considering experimental data obtained with Aspergillus cul-
the design or scaling-up of conventional bioreactors, it is tivation in quite distinct operational conditions, observing
more convenient to use more complex correlations, such as that the fit was even better for the more complex correla-
those proposed by Yagi and Yoshida [12] and by Zlokarnik tions, which take into account the physical properties of the
[14], which relate dimensionless variables. broth. It is obvious, however, that the utilisation of such cor-
In addition, it should be observed that the parameters relations means the necessity of more complex analytical
estimated by the regressions are very close to the values determinations, which is not always possible.
found in the literature. The parameters of the correlation Finally, it should be pointed out that the good fits obtained
proposed by Cooper et al. [4], a1 = 0.47 and b1 = 0.39, are a consequence of the quality and accuracy of the gen-
respectively, are within the ranges of values mentioned by erated experimental values, which are a result of carefully
Kawase and Moo-Young [29] for non-Newtonian systems, elaborated methodologies implemented in the cultivations.
which are 0.37 < a1 < 0.80 and 0.20 < b1 < 0.84. The
value obtained for the constant c2 of the correlation pro-
posed by Ryu and Humphrey [9] (c2 = −0.12) is very References
different from that obtained in the original work (c2 =
−0.86). However, it should be pointed out that the broths [1] M. Gavrilescu, R.V. Roman, V. Efimov, Acta Biotechnol. 13 (1993)
studied had different physicochemical properties and the [2] J.J. Smith, M.D. Lilly, R.I. Fox, Biotechnol. Bioeng. 35 (1990) 1011.
discrepancy may be due to the fact that different equations [3] R. Jurecic, M. Berovic, W. Steiner, T. Koloini, The Can. J. Chem.
have been used to calculate the apparent viscosities (µap ) Eng. 62 (1984) 334.
of the fermentation media: while the power law model [4] C.M. Cooper, G.A. Fernstrom, S.A. Miller, Ind. Eng. Chem. 36
has been used in this work to describe the non-Newtonian (1944) 504.
[5] M. Griot, V. Saner, E. Heinzle, I.J. Dunn, J.R. Bourne, in: Proceedings
behaviour of Aspergillus fermentation broths, Ryu and
of the 2nd International Conference on Bioreactor Fluid Dynamics,
Humphrey [9] employed the Bingham plastic model. How- Cambridge, 1988, p. 17.
ever, as expected, both of the values for the constant c2 [6] V. Linek, J. Sinkule, Biotechnol. Bioeng. 35 (1990) 1034.
were coherently negative, indicating an inverse proportion- [7] A.C. Badino Jr., M. Barboza, C.O. Hokka, Advances in Bioprocess
ality between kL a and µap . The values obtained for the Engineering, Kluwer Academic Publishers, The Netherlands, 1994,
p. 157.
parameters from the correlation of Yagi and Yoshida [12]
[8] F.J. Montes, J. Catalan, M.A. Galan, Proc. Biochem. 34 (1999) 549.
could not be compared with values found in the literature, [9] D.Y. Ryu, A.E. Humphrey, J. Ferment. Technol. 50 (1972) 424.
because no published work has been found that applies [10] F. Garcı́a-Ochoa, E. Goméz, Biochem. Eng. J. 1 (1998) 1.
this correlation to non-Newtonian fermentation broths. The [11] C.S. Shin, M.S. Hong, J. Lee, Biotechnol. Technol. 10 (1996) 679.
A.C. Badino Jr. et al. / Biochemical Engineering Journal 8 (2001) 111–119 119

[12] H. Yagi, F. Yoshida, Ind. Eng. Chem. Proc. Des. Dev. 14 (1975) 488. [22] B.J. Michel, S.A. Miller, AIChE J. 8 (1962) 262.
[13] G.-Q. Li, H.-W. Qiu, Z.-M. Zheng, Z.-L. Cai, S.-Z. Yang, J. Chem. [23] A. Schumpe, G. Quicker, W.D. Deckwer, Adv. Biochem. Eng. 24
Tech. Biotechnol. 62 (1995) 385. (1982) 1.
[14] M. Zlokarnik, Adv. Biochem. Eng. 8 (1978) 133. [24] A.C. Badino Jr., M.C.R. Facciotti, W. Schmidell, J. Chem. Technol.
[15] A.C. Badino Jr., M.C.R. Facciotti, W. Schmidell, Braz. J. Chem. Biotechnol. 75 (2000) 469.
Eng. 14 (1997) 359. [25] M. Metwally, M. El Sayed, M. Osman, P.P.F. Hanegraaf, A.H.
[16] C.K. Svihla, S.N. Dronawat, J.A. Donnelly, T.C. Rieth, T.R. Hanley, Stouthamer, H.W. Van Verseveld, Antonie van Leeuwenhoek 59
Appl. Biochem. Biotech. 63 (1997) 375. (1991) 35.
[17] E.S. Olsvik, B. Kristiansen, Biotechnol. Adv. 12 (1994) 1. [26] M.L. Shuler, F. Kargi, Bioprocess Engineering: Basic Concepts,
[18] W. Schmidell, J.R.G. Menezes, Rev. Microbiol. 17 (1986) 194. Prentice-Hall, New Jersey, 1992.
[19] Z. Kemblowski, B. Kristiansen, Biotechnol. Bioeng. 28 (1986) 1474. [27] R.H. Perry, C.H. Chilton, Chemical Engineers’ Handbook, 5th
[20] P.R. Vilaça, A.C. Badino Jr., M.C.R. Facciotti, W. Schmidell, Edition, McGraw-Hill Kogakusha, Tokyo, 1973.
Bioprocess Eng. 22 (2000) 261. [28] D.W. Marquardt, J. Soc. Ind. Appl. Math. 11 (1963) 431.
[21] A.B. Metzner, R.H. Feehs, L.H. Ramos, R.E. Otto, J.B. Tuthill, [29] Y. Kawase, M. Moo-Young, Chem. Eng. Res. Des. 66 (1988)
AIChE J. 7 (1961) 3. 284.