Communication to the Editor
Determination of the Volumetric Mass
Transfer Coefficient (k,a) Using the Dynamic
Gas Out-Gas In Method: Analysis of
Errors Caused by Dissolved Oxygen Probes
L. A. Tribe, C. L. Briens, and A. Margaritis
Department of Chemical and Biochemical Engineering, University of
Western Ontario, London, Ontario, Canada N6A 5B9
Received June 28, 1994lAccepted December 13, 1994
There are many dynamic methods for measuring the vol-
umetric mass transfer coefficient. The gas out-gas in
method can directly determine the volumetric m a s s
transfer coefficient in a bioreactor system and provide
estimates of the volumetric microbial oxygen uptake rate
and the average oxygen saturation concentration at the
gas-liquid interface. The errors on these parameters are
large if the dissolved oxygen probe response time is not
considered. For reliable measurements, deconvolution
of the oxygen probe measurements must be made. o
1995 John Wiley & Sons, Inc.
Key words: volumetric mass transfer coefficient oxygen
uptake rate probe response time dynamic gas out-gas
in method.
INTRODUCTION
In aerobic bioreactors, sufficient levels of oxygen must be
supplied to the liquid broth to establish optimum conditions
for microbial growth. Often, the oxygen transfer rate is the
limiting factor of a bioreactor system and an important pa-
rameter for scale-up. The oxygen transfer rate is dependent
upon the volumetric mass transfer coefficient, k,a, and the
driving force, C; - C, (where C, is the dissolved oxygen
concentration and C: is the oxygen saturation concentration
in the liquid phase at the gas-liquid interface).*18
The oxygen transfer rate can be measured by chemical or
physical techniques. The application of a chemical tech-
nique requires accurate reaction kinetics which can be dif-
ficult to determine.20 Moreover, the chemicals (e.g., so-
dium sulfite) cannot be used in bioreactor systems.
A physical technique usually involves performing an up-
step or downstep in the oxygen concentration of the biore-
actor inlet gas and measuring the dissolved oxygen concen-
tration. In simulated broths without viable microorganisms,
the volumetric mass transfer coefficient can be determined
by either absorption or desorption. Steady state is first es-
tablished with either air or nitrogen sparging and either
oxygen or nitrogen is then substituted for the air or nitro-
gen .4,5,7,10,16,18,20 As pointed out by Kim and Chang,
* To whom all correspondence should be addressed
Biotechnology and Bioengineering, Vol. 46, Pp. 388-392 (1995)
0 1995 John Wiley & Sons, Inc.
performing an upstep with oxygen-enriched air is preferable
to a downstep with nitrogen since it eliminates the risk of
going below the critical dissolved oxygen concentration in
actual microbial broths. A variant of this technique is the
dynamic pressure method, where the pressure inside the
bioreactor is increased causing an instantaneous increase in
the interfacial oxygen concentration. 2-14
Although these physical methods can be applied to biore-
actor systems, they require either an independent measure-
ment of the volumetric oxygen uptake rate by the microor-
ganisms in the bioreactor4 or an accurate determination of
the interfacial oxygen concentration, Ct2 The oxygen up-
take rate measurement requires taking a sample and per-
forming an experiment in a separate laboratory unit. The
interfacial oxygen concentration, C;, cannot be easily de-
termined since it depends on the evolving broth composi-
tion, the local pressure, and the oxygen concentration in the
gas, which varies as the gas bubbles rise through the broth.
The physical gas out-gas in method avoids these prob-
lems by determining directly the volumetric mass transfer
coefficient in the actual microbial broth.3 The dissolved
oxygen concentration is monitored during a short period of
non-aeration (gas out) and subsequent re-aeration (gas
in). The gas-out period must be short and the dissolved
oxygen concentration, C,, kept above the critical oxygen
concentration, Ccrit,to ensure that the volumetric oxygen
uptake rate is approximately constant during the gas out-gas
in experiment; the respiration rate coefficient is constant at
dissolved oxygen concentrations above the critical value
and the biomass concentration does not significantly change
over short periods of time. The gas out-gas in method can
also provide estimates of both the microbial oxygen uptake
rate and the average oxygen interfacial concentration.
For all the physical methods the determination of the
oxygen mass transfer coefficient becomes inaccurate when
the response time constant, T ,of the dissolved oxygen probe
is large. The response time constant is defined as the time at
which the probe reaches 63.2% of its final value when
exposed to a step change in concentration. For reasonably
accurate values, a criterion of T + (l/k,a) is usually rec-
ommended. 19-21 The commercially available sterilizable
CCC 0006-35921951040388-05
dissolved oxygen probes which are used for measurements
in bioreactor systems have large response time constants of
lCL100 s.21Several models of the probe response have been
developed,'1318but most modem sterilizable probes can be
approximated by a first-order hence the general
use of the response time constant, 7 , to characterize dis-
solved oxygen probes. This article quantifies the errors
caused by neglecting the probe response time when inter-
preting the results of gas out-gas in experiments. As shown
below, previous attempts have made assumptions that do
not correspond to realistic conditions.
METHOD
Figure 1 shows an example of the results from a typical gas
out-gas in experiment. For the gas-out period, the change in
dissolved oxygen concentration equals the volumetric mi-
crobial oxygen uptake rate, given by the equation.
which yields, by integration,
The volumetric microbial oxygen uptake rate, Q,,X, can
thus be obtained from the variation with time of the dis-
solved oxygen concentration, c,, during the gas-out period.
For the gas-in period, the rate at which the dissolved oxygen
concentration changes is given by the
0 20 40 60 80 100 120
Time (s)
Figure 1. Difference between the actual, C,, and the measured probe
signal, y, for dissolved oxygen concentration changes in a bioreactor dur-
ing a gas out-gas in experiment. Assumptions: C,,,, = 0.1 mg OJL, k,a
= 0.07 s I , and Po, X = 0.25 mg O,/L/s. (-)
~ Actual concentration.
(---) Probe signal.
Equation (3) can be rearranged to 3317
Therefore, k,a can be determined from the slope of C12
versus dC,/dt. Using the value of the microbial oxygen
uptake rate obtained from the gas-out results, the oxygen
interfacial concentration can be obtained from the intercept,
which is (Cf - Qo2X/k,a). Alternately, since initially, un-
der steady-state conditions,'
Equation (3) can be integrated from the time, t,, at which
the air flow is re-started to any subsequent time, t , to
give'X2'
Thus, k,a can be determined from the slope of ln(C,, -
C,) versus t. If required, the oxygen interfacial concentra-
tion can then be obtained from Equation ( 5 ) and the value of
the volumetric microbial oxygen uptake rate obtained from
the gas-out results.
Data from Bambot et a1.2 for the response of a sterilizable
Clark-type electrode from Ingold Electronics to a step in
dissolved oxygen concentration verified the applicability of
the first-order system to the response of such a probe. This
type of probe is commonly used to measure dissolved ox-
(3) ygen concentrations in bioprocesses. Their data gave a
probe response time of approximately 3 1 s. Ingold lists 98%
response in 60 s , which,
~ for a first-order system, corre-
sponds to a response time constant of 15 s based on 63.2%
response. The discrepancy may be due to the aging of the
membrane. A first-order system with a response time of 15
s was assumed to minimize the errors caused by the probe
response. Other probes may have different response char-
acteristics, but errors will be of the same order of magnitude
as predicted with a first-order probe.
The probe signal, at time t, was obtained by convolution
of the concentration history from the start of the experi-
yt2
....... ment until time t with the probe response characteristics.
.......
For t < t,,
y(r) = CLO- Qo,X [f - T (1 - e-"")] (7)
and for t > ts,
cs
I
140
Determined values for the oxygen uptake rate and the vol-
umetric mass transfer coefficient may vary depending on
COMMUNICATION TO THE EDITOR 389
the experimental data selected for the calculations. There-
fore the following boundaries were set (see Fig. 1):
The aeration should be re-started when the dissolved
oxygen concentration reaches 1.5 times the critical con-
centration. That is, C, = 1.5Cc~t. The volumetric mi-
crobial oxygen uptake rate should remain approximately
constant over a gas out-gas in experiment.
The initial probe signal, yo, corresponds to the actual
initial dissolved oxygen concentration in the bioreactor,
Co (i.e., yo = Co).
Experimental data selected for analysis for the gas-out
and gas-in periods are within the range defined by ytl
and yt2 (see Fig. 1):
yfl = yo - 5 Ayl where Ayl = 0.01~~
yt2 = ymini + 5 Ayz where Ay2 = 0.Olymini
and yminiis the minimum probe signal value obtained from
a gas out-gas in experiment. These boundaries were fol-
lowed in evaluating the errors caused by neglecting the
probe response time.
In practice, the probe response time is often neglected
and the volumetric microbial oxygen uptake rate and the
volumetric mass transfer coefficient are obtained directly
from the probe signal. Linear regression of the decreasing
probe signal versus time is performed between ytl and yf2.
The negative of the slope yields an estimate of the volu-
metric microbial oxygen uptake rate, (Q,, me,
as indicated
by Equation (9):
(9)
There are two methods to obtain estimates of the volumetric
mass transfer coefficient, &a),.
Method 1 . For the increasing probe signal,
The estimated volumetric mass transfer coefficient is the
negative of the inverse of the slope of y versus dyldt eval-
uated between the boundaries ytl and yf2.
Method 2. For the increasing probe signal,
The estimated mass transfer coefficient is the negative of
the slope of lnb, - y) versus t evaluated between the
boundaries ytl and yr2.
The percent error on the volumetric microbial oxygen-
uptake rate and the volumetric mass transfer coefficient
were evaluated by
390 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 46, NO. 4, MAY 20, 1995
Percent error on Qo,X =
\ /
and
Table I lists the range of values of the oxygen uptake rate,
the critical dissolved oxygen concentration, and the volu-
metric mass transfer coefficient used in this article. This
range of values was derived from Atkinson and Mavituna'
and represents typical aerobic fermentation conditions.
RESULTS AND DISCUSSION
Only the error from neglecting the probe response is eval-
uated here. All other sources of error are not included. For
example, this article does not consider the assumption of
rapid bubble disengagement during the gas-out period,
which is not verified for tall vessels. Other authors have
proposed solutions to this problem.2o321However, the de-
termined value of the volumetric mass transfer coefficient
will still be inaccurate unless the probe response time is
properly considered.
Figure 1 illustrates the large difference between the actual
and measured concentration profiles in a bioreactor during a
typical gas out-gas in experiment. The boundaries used for
the evaluation of the oxygen uptake rate and the volumetric
mass transfer coefficient from the probe response are indi-
cated. Figure 1 shows that the probe indicated a dissolved
oxygen concentration of over 2.6 mg O,/L when the actual
concentration was 0.15 mg 02/L. Neglecting the probe re-
sponse time may therefore lead to oxygen starvation.
Figure 2 shows that the oxygen uptake rate may be
greatly underestimated by neglecting the probe response
time. The error increased when either the oxygen uptake
rate or the volumetric mass transfer coefficient was in-
creased. Such errors would lead to a defective scale-up.
Neglecting the probe response time caused large errors on
the volumetric mass transfer coefficient. A comparison of
Figures 3 and 4 show that, for most cases, larger errors
resulted using method 1. With either method, the error was
almost independent of the volumetric oxygen uptake rate
(Figs. 3 and 4). A reliable estimate of the error on the mass
transfer coefficient can thus be made even when the oxygen
Table I. Range of values of volumetric microbial oxygen uptake rate,
critical oxygen concentration, and volumetric oxygen mass transfer coef-
ficient.
Parameter Range Unit
Volumetric oxygen uptake rate 0.05-1.0 rng O,/L/s
Critical oxygen concentration 0-2 rng O,/L
Volumetric mass transfer coefficient 0.014.15 S-'
 0 0

-1 0 a 4,
p -20
2 -20
W
Y
m
c
b
Q -40 m,
3 Y -30
C
W c
0

:
0 -60 2b -40
W c
5 c
c
0
-80 a
i -50

b
c
c -60

i
a -100
-70

-120 I I I 7 -80 1 I I 1 - 7
00 02 04 06 08 10 00 02 0.4 06 08 10
Oxygen Uptake Rate, Qo2X (rng0,lLls) Oxygen Uptake Rate, Qo2X (rng0,lLls)

Figure 2. Error on the volumetric microbial oxygen uptake rate caused
by neglecting the probe response time. Assumptions: C,,, = 0.1 mg 0,IL
and the valuesof k,aare 0.03 s - ' (a), 0.05 s - ' (b), 0.07 s - ' (c), 0.09 s - '
(d), 0.11 s-' (e), 0.13 s C ' (9, and 0.15 s C 1 (8).
Figure 4. Error on the volumetric mass transfer coefficient (method 2)
caused by neglecting the probe response time. Assumptions: C,,,, = 0.1
mg0,iLand thevaluesofkLaare0.03 s - ' (a), 0.05 s C ' (b), 0.07 s C 1(c),
0.09 s - ' (d), 0.11 s - l (e), 0.13 s C 1(f). and 0.15
s (8).
C '

uptake rate is not known. This relative error increased as the

mass transfer coefficient increased. the oxygen concentration of the bioreactor inlet gas is per-
According to van't Riet,20 the error on the mass transfer formed. Figure 4 shows that for a mass transfer coefficient
coefficient caused by neglecting the probe response time is less than 0.07 s-' the error in bioreactors could be as high
less than 6% when the mass transfer coefficient is smaller as 37%. Linek et a1.'' also assume that the probe has equil-
than the inverse of the probe response time constant (k,a C ibrated with the bioreactor broth when the gas is turned back
1 /=~ 0.07 s-'). This criterion assumes that the probe has on at time ts, as shown in Figure 1. This approximation can
equilibrated with the bioreactor broth before an upstep in only be valid if the critical dissolved oxygen concentration
and the microbial volumetric oxygen uptake rate are very
0
small. Such conditions are rarely encountered in practice.
Neglecting the probe response time and then underestimat-
-10
ing the required volumetric mass transfer coefficient would
lead to a defective scale-up.
-20
Figure 5 shows that the error on the volumetric mass
transfer coefficient caused by neglecting the probe response
3 -30
time is practically independent of the critical oxygen con-
c centration. In practice, however, the effect of probe cali-
0
L

-40
bration errors may become significant as the critical oxygen
b
c
concentration becomes larger.
c

a
i -50
If the interfacial oxygen concentration, C,*, is known, the
volumetric mass transfer coefficient, k,a, could theoretical-
-60
ly be obtained from Equation (5) and the value of the ox-
ygen uptake rate estimated from the gas-out results. How-
-70
ever, this would lead, in most cases, to larger errors than
obtained by using the gas-in results.
-80 I I I I I

0.0 0.2 0.4 0.6 0.8 10 CONCLUSIONS

Oxygen Uptake Rate, Qo2X (rng0,ILls)
Figure 3. Error on the volumetric mass transfer coefficient (method 1)
caused by neglecting the probe response time. Assumptions: C,,, = 0.1
mg0,/LandthevaluesofkLaare0.03sC' (a),0.05s-'(b),0.07sC'(c),
0.09 s C ' (d), 0.11 s-' (e), 0.13 s-' (0, and 0.15 s - ' (g).
Neglecting the probe response time causes large errors on
the oxygen uptake rate and on the volumetric mass transfer
coefficient even when the standard criterion that the probe
response time constant be much smaller than the inverse of
the volumetric mass transfer coefficient is verified. The

COMMUNICATION TO THE EDITOR 391


-I0
-20
1i a --..
\.\ , A,,
00 0.2 0.4 0.6
Oxygen Uptake Rate, QO2X(mgO,/L/s)
Figure 5. Comparison of the error on the volumetric mass transfer co-
efficient, caused by neglecting the probe response time for various values
'
of kLa and C,,,,. The values of kLa are 0.03 s - (a), 0.05 s C ' (b), 0.07 s -
( ~ ) , 0 . 0 9 s - ~ ( d ) , O s. l- I' ( e ) , 0 . 1 3 s C ' ( f ) , a n d 0 . 1 5 s ~ ' ( g )Valuesof
C,,,,: (-.-) 0 mg 0,iL; (--) 0.5 mg 0,iL; (--) 1 mg 0,iL;
(. . . . .) 1.5 mg 0,iL; (-) 2 mg 0,iL.
calculation method which uses the derivative of the mea-
sured probe response to determine the mass transfer coef-
ficient should be avoided. The alternate (method 2) yields
less inaccurate estimates. For reliable measurements and
scale-up, deconvolution of the oxygen probe measurements
should be performed.
The authors thank the Natural Sciences and Engineering Re-
search Council (NSERC) of Canada for their financial support.
NOMENCLATURE
dissolved oxygen concentration (mg O,/L)
dissolved oxygen saturation concentration in liquid phase at
gas-liquid interface (mg O,/L)
estimated dissolved oxygen saturation concentration in liquid
phase at gas-liquid interface (mg 02/L)
critical dissolved oxygen concentration (mg 0,iL)
initial dissolved oxygen concentration (mg 0,iL)
dissolved oxygen concentration at t = t, (mg 0 /L
volumetric oxygen mass transfer coefficient (s ?) )
estimated volumetric oxygen mass transfer coefficient (s I)
respiration rate coefficient (mg 02/g dry biomass weighus)
volumetric microbial oxygen uptake rate (mg O,/L/s)
estimated volumetric microbial oxygen uptake rate (mg 0,i
L/s)
time (s)
time corresponding to minimum probe signal value (s)
time at which gas flow is re-stafied (s)
characteristic probe response time constant (s)
biomass concentration (g dry weight/L)
initial probe signal (mg O,/L)
probe signal at time t (mg O,/L)
392 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 46, NO. 4, M A Y 20, 1995
Y11 upper boundary for analysis (mg 0,iL)
YO lower boundary for analysis (mg O,/L)
y,,,, minimum probe signal value (mg 0,iL)
A,,
References
parameter used to calculate y r , (mg O,/L)
parameter used to calculate yr2 (mg 02/L)
I . Atkinson, B., Mavituna, F. 1991. Biochemical engineering and bio-
technology handbook, 2nd edition. MacMillan, London.
2. Bambot, S. B., Holavanahali, R., Lakowicz, J. R., Carter, G . M.,
Rao, G . 1994. Phase fluorometric sterilizable optical oxygen sensor.
Biotechnol. Bioeng. 43: 1139-1 145.
3. Bandyopadhyay, B., Humphrey, A. E., Taguchi, H. 1967. Dynamic
measurement of the volumetric oxygen transfer coefficient in fermen-
tation systems. Biotechnol. Bioeng. 9: 533-544.
4. Bartholomew, W. H., Karow, E. O., Sfat, M. R., Wilhelm, R. H.
1950. Oxygen transfer and agitation in submerged fermentations.
Mass transfer of oxygen in submerged fermentation of Streptomyces
griseus. Ind. Eng. Chem. 42: 1801-1809.
08 10 5. Chang, H. N., Halard, B., Moo-Young, M. 1989. Measurement of
KLa by a gassing-in method with oxygen-enriched air. Biotechnol.
Bioeng. 34:1147-1157.
6. Cole-Parmer Instrument Company Catalogue. 1993-94. Niles, IL.
7. Dang, N. D. P., Karrer, D. A , , Dunn, I. J. 1977. Oxygen transfer
coefficients by dynamic model moment analysis. Biotechnol. Bioeng.
. 19: 853-865.
8. Dunn, I. J., Einsele, A. 1975. Oxygen transfer coefficients by the
dynamic method. J. Appl. Chem. Biotechnol. 25: 707-720.
9. Gauthier, L., Thibault, J., LeDuy, A. 1991. Measuring k,a with
randomly pulsed dynamic method. Biotechnol. Bioeng. 37:889-893.
10. Kim, D.-J., Chang, H.-N. 1989. Dynamic measurement of KLa with
oxygen-enriched air during fermentation. J. Chem. Tech. Biotechnol.
45: 39-14.
11. Linek, V., Sinkule, J . 1990. Comments on validity of dynamic mea-
suring methods of oxygen diffusion coefficients in fermentation media
with polarographic oxygen electrodes. Biotechnol. Bioeng. 35:
1034-1041.
12. Linek, V., Benes, P., Vacek, V. 1989. Dynamic pressure method for
kLa measurement in large-scale bioreactors. Biotechnol. Bioeng. 33:
1 4 0 6 1 4 12.
13. Linek, V . , Moucha, T . , Dousova, M., Sinkule, J. 1994. Measure-
ment of kLa by dynamic pressure method in pilot-plant fermentor.
Biotechnol. Bioeng. 43: 4 7 7 4 8 2 .
14. Linek, V., Sinkule, J., Benes, P. 1991. Critical assessment of gas-
sing-in methods for measuring k,a in fermentors. Biotechnol. Bioeng.
38: 323-330.
15. Linek, V., Sobotka, M., Prokop, A. 1973. Measurement of aeration
capacity of fermentors by rapidly responding oxygen probes. Biotech-
nol. Bioeng. Symp. 4: 429453.
16. Nakanoh, M . , Yoshida, F. 1980. Gas absorption by Newtonian and
non-Newtonian liquids in a bubble column. Ind. Eng. Chem. Process
Des. Dev. 19: 19C-195.
17. Rane, K. D., Sims, K . A. 1994. Oxygen uptake and citric acid pro-
duction by Candida lipolytica Y 1095. Biotechnol. Bioeng. 43:
I3 1-1 37.
18. Robinson, C. W., Wilke, C. R. 1973. Oxygen absorption in stirred
tanks: A correlation for ionic strength effects. Biotechnol. Bioeng. 15:
755-782.
19. Ruchti, G . , Dunn, I. J., Bourne, J. R. 1981. Comparison of dynamic
oxygen electrode methods for the measurement of KLa. Biotechnol.
Bioeng. 23: 277-290.
20. van't Riet, K . 1979. Review of measuring methods and results in
nonviscous gas-liquid mass transfer in stirred vessels. Ind. Eng.
Chem. Process Des. Dev. 18(3): 357-364.
21. van't Riet, K . , Tramper, J. 1991. Basic bioreactor design. Marcel
Dekker, New York.