BAB 9 Rajiv Dutta

Sterilization 207
5 X 10 12 m-3, which needs to be reduced to such an extent that only one

organism can survive during two months of continuous operation.
The heat-resistant bacterial spores in the medium can be
characterized by an Arrhemus coefficient (k do ) of 5.7 x 10 39 hr-1 and an
activation energy (Ed) of 2.834 x 105 kJ /kmol (Deindoerfer and
Humphrey, 1959). The sterilizer will be constructed with the pipe
with an inner diameter of 0.102 m. Steam at 600 kPa (gage pressure) is
available to bring the sterilizer to an operating temperature of 125°C.
The physical properties of this medium at 125°C are c = 4.187 kJ /kg K,
P = 1000 kg/m 3, and J1 = 4 kim hr.
a. What length should the pipe be in the sterilizer if you assume
ideal plug flow?
b. What length should the pipe be in the sterilizer if the effect of
axial dispersion is considered?
Solution:
a. The design criterion can be calculated from Eq. (8.5) as
v = ln~ = In (5 x 1012m-3)(2m3/hr)(24hr/day)(60days) = 37.2

n
Since the temperature at the holding section is constant,
Eg. (8.5) is simplified to
V = kdThold
From the given data, kd can be calculated by using Eq. (8.4) to
yield
Therefore,
r - V - 37.2 = 0.0983 hr
hold - k - 387.6
d
The velocity of nledium is
2 m 3 /hr
u = - - - - = 245 m/hr
JE-O.I02 2 m 2
4
The length of the sterilizer is
L = liThoId = 24.1m
b. The Reynolds number for the medium flow is
N = 0.102 (245)(1000) = 6.24 x 103
Re
4
D
From Figure 8.2 0.8 for N Re = 6.24 X 103
iidt
208 Fundamentals of Biochemical Engineering
Therefore,
o ~ 0.8il dt = 20 m 2 jh
Now, substituting all the values given and calculated in this
problem to Eq. (8.25) will result in an equation with only one
unknown, L, which can be solved by using any non-linear
equation solver:
L = 26.8
Therefore, the holding section should be 26.8 m, which is 2.7 m
longer than the result from the assumption of ideal plug flow.
Cooling Section: For the cooling section, a quench cooler with

adequate heat removal capacity is effective. Another technique is to
inject the hot medium through an expansion valve into a vacuum
chamber, which is known as flash cooling. Both of these take a very
short time; therefore, the sterilization during the cooling period can
be assumed to be negligible.
A shell-and-tube or a plate-and-frame heat exchanger can also be
employed for cooling. The temperature versus residence time
relationship for cooling using an isothermal heat sink is
T H =Tc-(Tc-TH)exp(_UArcOOl) (8.27)
2 1 cW
For cooling using a countercurrent heat sink of equal flow rate and
heat capacity,
(8.28)
8.6 AIR STERILIZATION

For aerobic fermentations, air needs to be supplied continuously.
Typical aeration rates for aerobic fermentation are 0.5 - 1.0 vvm (air
volume per liquid volume per minute). This requires an enormous
amount of air. Therefore, not only the medium but also the air must
be free of microbial contaminants. All of the sterilization techniques
discussed for medium can also be employed for air. However,
sterilization of air by means of heat is economically impractical and is
also ineffective due to the low heat-transfer efficiency of air
compared with those of liquids. The most effective technique for air
sterilization is filtration using fibrous or membrane filters.
The cotton plug, routinely used as a closure for tubes or flasks of
sterile solution, is a good example of removal of microorganisms
from air by a fibrous filter. A simple air filter can be made by packing
cotton into a column. However, with cotton filters the pressure drop
is high and wetting can be a breeding ground for the contamination.
Sterilization 209
Therefore, glass fibers are favorable as filter medium because they

give a lower pressure drop and are less liable to wetting or
combustion. Modern fibrous filter systems are cylinders made from
bonded borosilicate microfibers sheathed in reinforcing
polypropylene mesh in which the layers increase in fineness and
density from the center outward. This type of design can deliver over
3 m 3 /s of sterile air at 0.1 bar of pressure drop (Quesnel, 1987).
With fibrous filters, airborne particles are collected by the
mechanisms of impact on, interception, and diffusion.
Impaction: When an air stream containing particles flows around

a cylindrical collector, the particle will follow the streamlines until
they diverge around the collector. The particles because of their mass
will have sufficient momentum to continue to move toward the
cylinder and break through the streamlines, as shown in Figure 8.3.
The collection efficiency by this inertial impaction mechanism is the
function of the Stokes and the Reynolds number as:
.
TJlmp.
= f (N N ) =
SV Re
(CfPPJ; Vo DeVop)
18,uDc' f.1
(8.29)
where Cf is known as Cunningham correction factor. The value of Cf

can be estimated from the empirical correlation developed by Davies
(Strauss, 1975),
Cf = 1 + ~~ [ 1.257 + 0.400exp( -1.l0~~)] (8.30)
where A is the mean free path of gas molecules based on the

Chapman-Enskog equation,
A= (O-.-£9P )I::~ (8.31)
Fig. 8.3 Flow pattern around cylindrical fiber, showing the path of
particles collected by inertial impaction.
The efficiency lJimp is defined as the fraction of particles

approaching the collector which impact. Various correlations are
available in the literature. An empirical correlation for the efficiency
developed by Thorn is (Strauss, 1975):
3
N St
l1imp = 3
N St
2
+ 0.77NSt + 0.22
for N Re c =10 (8.32)
Another correlation propsed by Friedlander (1967) is

0.075 N\~
l1i01P = (8.33)
The efficiency increases with increasing particle diameter or air flow
velocity.
Interception: The inertial impaction model assumed particles had

mass, and hence inertia, but no size. An interception mechanism is
considered where the particle has size, but no mass, and so they can
follow the streamlines of the air around the collector. If a streamline
which they are following passes close enough to the surface of the
fiber, the particles will contact the fiber and be removed (Figure 8.4).
The interception efficiency depends on the ratio of the particle
diameter to the cylindrical collector diameter (l( = dp / Dc):
'1int=
1
2.002- 1nN
Rec
[(1+K)ln(l+I\)- ~i~:;n (8.34)
which was developed by using Langmuir's viscous flow equation

(Strauss, 1975). The ratio l( is known as interception parameter. The
collection efficiency by interce'ption increases with the increase of the
particle size.
Fig. 8.4 Flow pattern around cylindrical fiber, showing the

interception collection mechanism.
Sterilization 211
Diffusion: Particles smaller than about 1 micron in diameter

exhibit a Brownian motion which is sufficiently intense to produce
diffusion. If a streamline containing these particles is sufficiently
close to the collector, the particles may hit the collector and be
removed. Contrary to the previous two mechanisms, the collection
efficiency by diffusion increases with decreasing particle size or air
velocity. The typical size of particles collected by this mechanism is
less than about 0.5 f.1m. The efficiency of collection by diffusion can be
estimated by an equation analogous to Langmuir'S equation,
Eq. (8.34), as (Strauss, 1975):
-
17dif -
1
2.002 _ In N
[(1
+
Z) 1 (1
n +
2) 2(2 + Z)]
- 2(1 + Z)
(8.35 )
Rec
where 2 is the diffusion parameter defined as
1
2 = [224(2002 -lnNRe ) D Br
c vD
c
]3 (8.36)
Friedlander (1967) suggested the following correlation

17dif = 1.3N p;2/3) + 0.7J(2 (8.37)
where N pe is Peclet number, an important dimensionless parameter
in the theory of convective diffusion. It is defined as
vcD c
N pe =- - = NReNSc (8.38)
D Br
where N sc is the Schmidt number which is defined as
f.1
N sc = [ ) (8.39)
P Br
The diffusivity due to Brownian motion for submicron size particles
can be estimated from
D _ CfkT
(8.40)
Br - 3npdp
where k is Boltzmann's constant (1.38054 x 10-9 11K).
Combined Mechanisms: The total collection efficiency of a

fibrous filter is obtained from the combined effect of the preceding
three mechanisms. One straightforward way to combine the
collection efficiencies of the different mechanisms is to add them
together, but this implies that a particle can be collected more than
once, which does not make sense. A better approach is to use the
following correlation,
1]c =1 - (1 - 1]imp) 1- 1]int) (1 - 1]dif) (8.41)
212 Fundamentals of Biochc111ical Engineering
which allows only the particles not collected by one mechanism to be

collected by the others. Substitution of Eqs. (8.32), (8.34), and (8.35)
into Eq. (8.41) will result in the correlation for the collection efficiency
by the combined mechanisms. Pasceri and Friedlander (1960)
correlated the combined collection efficiency as
6 2 05
11c = 2/3 0.5 + 31C NR~c (8.42)
N sc N Re c
As mentioned earlier, with an increase of the superficial air velocity
(vo)~ lJimp and 11int increase whereas lJdif decreases. Therefore, the
combined collection efficiency normally decreases to reach a
minimum point and then increases with increasing superficial air
velocity.
Effect of Multiple Layers and Packing: All correlations for the
collection efficiency discussed so far are based on the ideal case of a
single cylindrical collector. Now, let's examine a filter unit consisting
of randomly oriented multiple layers. Consider an area (A) of filter at
a right angle to the gas flow and with a depth dh. If the packing
density a. is defined as the volume of fiber per unit volume of filter
bed, the velocity within the filter void space is equal to
V= ~ (8.43)
(1- a)
v
Cn - _O-A(1-a) - -.....
1-(1
dh
Fig. 8.5 Shell balances around a differential element of a filter.

A mass balance on the particles for the control volume (Figure 8.5)
results in
Input - ()utput = Collected by the filter (8.44)
) d( ~~v~)
C ~A(1-a)- Cll~+ dh A(1-a)
11 1- a 1- a dh
__Vo __ c
= I-a 11
(Adh)17 0 L
c C
(8.45)
Sterilization 213
where L is the length of cylindrical fiber per unit volume of filter bed,
which is related to the packing density a. and the average collector
diameter Dc as
rcD 2 L
a= _ c _ (8.46)
4
Simplifying Eq. (8.45) and substituting Eq. (8.46) for L gives
_ den = 4a1Jc dh (4.47)

Cn nD c(l-a)
which, on integration, results
In Cn =
Cna
-~(~)11c
n Dc 1- a
(8.48)
where B is the filter depth. Therefore, the collection efficiency for the
filter bed can be estimated as
11/ = 1 - -Cn
Cno
= 1- exp[ - -4B- ( -a-)
n Dc 1- a
11c] (8.49)
When fibers are packed together in a filter bed, the velocity will be
increased and the flow pattern will be changed, which increases the
collection efficiency from impaction and interception. Chen (1955)
has determined fiber interference effects experimentally and
suggests
11 a = 11f (1 + 4.5a) (8.50)

which is applicable for a < 0.1 and 11f< 1/(1 + 4.5 a)
In summary, in order to estimate the collection efficiency of a filter
bed, you have to calculate: 11c by using either Eq. (8.41) or Eq. (8.42),
11/ using Eq. (8.49), and 11a using Eq. (8.50). However, it should be
noted that the predictions of the collection efficiency from various
correlations vary widely due to empiricism or the oversimplification
in developing the models represent a complex collection mechanism.
Furthermore, since the effect of the gas velocity on the collection
efficiency is large, the collection efficiency can decrease significantly
by increasing or decreasing the gas velocity. To insure sterility in a
fermenter system, a conservative approach needs to be taken which
considers the minimal efficiency conditions due to possible velocity
fluctuations and prediction error from various correlations. In
assessing the filtration job to be accomplished, Humphrey (1960)
recommended that the design should permit only a one-m-a-
thousand chance of a single contaminant penetrating the filter during
the entire course of the fermentation.
Example 8.3
A filter bed of glass fibers (Dc = 15 /.lm, the bed depth B = 10 em, and
packing density a = 0.03) is being used to sterilize air (20°C, 1 atm)
with an undisturbed upstream velocity, vo, of 10 cm/s. The air stream
contains 5,000 bacteria per cubic meter (d p = 1 /.lm and Pp = 1 g/ em ).
a. Estimate the single fiber collection efficiency by mertial
impaction, by interception, and by diffusion.
b. Estimate the single fiber collection efficiency based on
combined mechanisms by using Eq. (8.41) and Eq. (8. 42) and
compare the results.
c. Estimate the collection efficiency (1J a ) of the filter bed.
d. Show how the superficial velocity V o affects the various single
fiber collection efficiencies.
Solution:
a. The velocity within the filter void space is from Eq. (8.43)
v= ~ = _1_0_ = 10.3 cm/s

(I-a) 1-0.03
To estimaterJimp by using Eq. (8.32) or Eq. (8.33), we need to

calculate the Reynolds and the Stokes number from the given
condition and the physical properties of air at 20°C and 1 atm
(p = 1.20 X 10-3 g/ cm 3 , /.l = 1.8 X 10-4 g/ em s)
N = Dcvp = (1.5 x 10-3 )(10.3)(1.20 x 10-3 ) = 0.103

Re c /1, 1.80 X 10-4
where all units are in the cgs system. The mean free path A and
the Cunningham correction factor can be estimated from
Eq. (8.31) and Eq. (8.30) as:
A= [ 1.80 X 10-4] n(29) = 6.50 x 10-6 cm

0.499(1.20 x 10-3 ) 8(8.314 X 10 7 )(293)
Cf = 1 + ~~ [1.257 + OAooexr(-1.10 ~~)] = 1.16

The Stokes number is
1.16(1)(1 x 10-4)2(10.3) = 0.0247

3
18(1.80 x 10-4)(1.5 x 10- )
Sterilization 215
Therefore, from Eq. (8.33), the single fiber collection efficiency

by inertial impaction is
11imp = 0.075N§t2 = 0.075(0.0247)1.2 = 8.82 X 10--4

The single fiber collection efficiency by interception is from
Eq. (8.34) for 1\= d,IO( = 0.0667 and N Re ( = 0.103,
hint = 1
2.002-1nN Re ,
[0 + /() In(l + /()- /(_~!~]
2(1 + 1(-)
= 0.96 x 10-4
The diffusivity dup to the Brownian n1otion can be estimated

from Eq. (8.40),
_ CfkT _ 1.16(1.38 x 10-6 )293

DB - - - - - - - - - - - : - - - - 4-
r 3nJ.1d p 3n(1.80 x 10-4)(1 x 10- )
2
= 2. 8y a Q2 77 x 10-5 cm2 / s
Ox au
Therefore, the Peclet number is
Dcv 1.5 x 10-3 (10.3) 4
N pe = - = 5 = 5.57 x 10
D Br 2.77 x 10-
The single fiber efficiency by diffusion can be estimated
from Eq. (8.37),
l1dif = 1.3Np~2/3) + 0.71(2 = 4.00 x 10- 3

Therefore, the single fiber collection efficiency by inertial
impaction, interception, and diffusion is 8.82 x 10-4 , 9.96 X 10-4 ,
and 4.00 x 10-3 , respectively.
b. The combined collection efficiency can be estimated from
Eq. (8.41) as:
11c = 1 - (l - 8.82 x 10-4)(1- 9.96 x 10-4)(1- 4.00 x Hy-J) = 0.0059
Instead, if we use Eq. (8.42),
1.80 x 10-4
Ns = _/1_ = = 5.41 X 105
C pOBr (1.20 x 10-3 )(2.77 x 10--7 )
Therefore,
6
rye = (5.41 x 10 5 )2/3(0.103)°.5 + 3(0.0667)2(0.103)°.5 = 0.0071
In this example, diffusion is predominant over impaction and

interception. If we use Eq. (8.32) and Eq. (8.35) for the
estimation of 1Jimp and 1Jdlf' respectively, 1Jinp = 6.82 X 10-3 and

1Jdif = 6.97 X 10- , which are significantly lower than those
4
predicted from Eq. (8.33) and Eq. (8.37). Strauss (1975)

compared the predicted values from various correlations and
found that the diffusion collection efficiency 1Jdif from Eq. (8.35)
tends to predict the lower value than the experimental values.
c. If we choose to use 1Je = 0.0059 which was predicted from
Eqs. (8.33), (8.34), (8.37), and (8.41), the collection efficiency for
the filter bed can be estimated from Eq. (8.49) as
1J = 1- ex [- 4(10) (0.03 )0.0059] = 0.79

J P n(l.5 x 10-3 ) 1- 0.03
and the collection efficiency including the interference effects is
from Eq. (8.50),
1J a = 0.79 [1 + 4.5 (0.03)]
d. Similarly, we can calculate various collection efficiencies at
various values of vo' As shown in Table 8.1, with an increase of
vo, 1Jimp and 1Jint increased whereas 1Jdif decreased. Therefore, the
combined collection efficiency 1Je decreased initially, reached a
minimum value, and then increased with increasing YO. The
minimum collection efficiency was reached when V o was about
5 cm/s.
Table 8.1 Various Collection Efficiencies for 1 tIm Particle
3 3 3 3 3
Vo 10
1Jirnp X 10
1Jint X 10
1Jdif X 1Je x 10 1Je x 10
cm/s Eq. (8.33) Eq. (8.34) Eq. (8.37) Eq. (8.41) Eq. (8.42)
0.1 0.004 0.479 22.3 22.8 28.6
1 0.056 0.647 7.25 7.941 0.3
5 0.0384 0.857 4.53 5.76 7.01
10 0.883 0.996 4.00 5.87 7.10
15 1.44 1.10 3.79 6.32 7.54
20 2.03 1.19 3.67 6.88 8.05
50 6.09 1.60 3.42 11.1 10.8
8.7 NOMENCLATURE
A surface area across which heat transfer occurs during
sterilization, m 2
B filter bed depth, m
Cf Cunningham correction factor, dimensionless
Sterilization 217
en cell number density, number of cells/m3

c specific heat of medium, J/kg K
Dc collector diameter, m
dp particle diameter, m
dt pipe diameter, m
D axial dispersion coefficient, m 2 / S
DBr diffusivity due to the Brownian motion, m 2 / s
Ed activation energy for thermal cell destruction in Arrhenius
equation, J/kmol
H enthalpy of steam relative to raw medium temperature, J/kg
In flux of microorganisms due to axial dispersion, m-2s-1
k Boltzmann's constant, 1.3805 x 10-23 J/K or 1.3805 x 1-16 erg/K
kd specific death rate, S-l or kg/m 3s
L length of holding section, m
M initial mass of medium in batch sterilizer, kg
Mw molecular weight of gas molecules, kg/kmol
ms steam mass flow rate, kg/ s
me coolant mass flow rate, kg/s
Npe Peclet number (/line uL/D or vODe/D Br), dimensionless
NRe Reynolds number (dtUP/JlL ), dimensionless
NRec collector Reynolds number (DevOp/Jl), dimensionless
Nsc Schmidt number (Jl/pD Br ), dimensionless
N St Stokes number (CfPpd'i v o/18 jlDc), dimensionless
n number of cells in a system
q rate of heat transfer, JI s
R gas constant, 8.314 x 103 JIkmol K or 8.314 x 107 erg/mol K
S cross-sectional area of a pipe, m-2
T absolute temperature, OK
To initial absolute temperature of medium, OK
Tc absolute temperature of heat sink, OK
TCo initial absolute temperature of heat sink, OK
TH absolute temperature of heat source, OK
t time, s
td doubling time, s
U overall heat-transfer coefficient, J/ s m 2 K.
U velocity, m/ s
u velocity, m/s
Urn maximum velocity, m/ s

v fluid velocity within the filter void space, m/ s
Vo undisturbed upstream fluid velocity, m/s
W mass of medium in a sterilizer, kg
x x-directional distance, m
Z the diffusion parameter defined in Eq. (8.36), dimensionless
a packing density defined as the volume of fiber per unit
volume of filter bed, dimensionless
1] collection efficiency, dimensionless
l( the ratio of the particle and the collector diameter (d p / Dc)'
dimensionless
A the mean free path of gas molecules, m
Il fluid viscosity, kg/m s
ilL liquid viscosity, kg/m s
P density, kg/m 3
Pp density of particles, kg/m 3
r residence time, s
r average residence time, s
V design criterion for sterilization, dimensionless
8.8 PROBLEMS
8.1 A fermenter containing 10 m 3 of medium (25°C) is going to be
sterilized by passing saturated steam (500 kPa, gage pressure)
through the coil in the ferrnenter. The typical bacterial count of
the medium is about 3 x 10 12 m-3, which needs to be reduced to
such an extent that the chance for a contaminant surviving the
sterilization is 1 in 100. The fermenter will be heated until the
medium reaches 115°C. During the holding time, the heat loss
through the vessel is assumed to be negligible. After the proper
holding time, the fermenter will be cooled by passing 20 m 3 /hr
of 25°C water through the coil in the fermenter until the
medium reaches 40°C. The coils have a heat-transfer area of
40 m 2 and for this operation the average overall heat-transfer
coefficient (U) for heating and cooling are 5,500 and 2,500 kJ /hr
m 2 K, respectively. The heat resistant bacterial spores in the
medium can be characterized by an Arrhenius coefficient (k d )
of 5.7 x 1039 hr-1 and an activation energy (Ed) of 2.834 x 105 kJ/kmgl
(Demdoerfer and Humphrey, 1959). The heat capacity and
density of the medium are 4.187 kJ /kgK and 1,000 kg/m 3,
respectively. Estimate the required holding time.
Sterilization 219
8.2 A continuous sterilizer with a steam injector and a flash cooler

will be employed to sterilize medium continuously. The time
for heating and cooling is negligible with this type of sterilizer.
The typical bacterial count of the medium is about 5 x 10 12 m-3,
which needs to be reduced to such an extent that only one
organism can survive during the three months of continuous
operation. The heat resistant bacterial spores in the medium can
be characterized by an Arrhenius coefficient (k do ) of 5.7 x 1039 hr-1
and an activation energy (Ed) of 2.834 x 10 5 kJ /kmol
(Deindoerfer and Humphrey, 1959). The holding section of the
sterilizer will be constructed with 20 m-long pipe with an inner
diameter of 0.078 m. Steam at 600 kPa (gage pressure) is
available to bring the sterilizer to an operating temperature of
125°C. The physical properties of this medium at 125°C are
c = 4.187 kJ/kgK and p =1,000 kg/m 3, and J1 = 4 kg/m hr.
a. How much medium can be sterilized per hour if you
assume ideal plug flow?
b. How much medium can be sterilized per hour if the effect
of axial dispersion is considered?
8.3 You need to design a filter for a 10,000-gallon fermenter that
will be aerated at a rate of 535 ft 3 /min (at 20°C and 1 atm). The
bacterial count in the air is 80 per ft 3 . Average size of the
bacteria is 1 J1m with density of 1.08 g/ cm3 . You are going to use
glass fibers (Dc = 15 j1,m) with packing density a = 0.03. The
cross-sectional area of the filter will be designed to give a
superficial air velocity Vo of 5 ft/ s.
a. What depth of the filter would you recommend to prevent
contamination?
b. How is the answer in (a) changed if Vo is decreased to 1 ft/s?
Explain the results.
8.4 Based on the combined mechanisms of impaction, interception,
and diffusion, a minimum efficiency should result for spheres
depositing on cylindrical collectors.
a. Develop an expression for the particle size corresponding
to this minimum efficiency. As an approximation, ignore
the effect of particle size on the Cunningham correction for
slip.
b. What is the particle size corresponding to the minimum
efficiency for the filter bed of glass fibers described in
Example 8.3?
8.9 REFERENCES
Aiba, S., A. E. Humphrey and N. F. Millis, Biochemical Engineering (2nd ed.),
pp. 242 - 246. Tokyo, Japan: University of Tokyo Press, 1973.
Chen, C. Y., "Filtration of Aerosols by Fibrous Media," Chem. Rev. 55 (1955):
595 - 623.
Deindoerfer, F. H. and A. E. Humphrey, "Analytical Method for Calculating
Heat Sterilization Time," AppI. Micro. 7 (1959a): 256 - 264.
Deindoerfer, F. H. and A. E. Humphrey, "Principles in the Design of
Continuous Sterilizers," Appl. Micro. 7 (1959b): 264 - 270.
Felder, R. M. and R. W. Rousseau, Elementary Principles of Chemical Processes
(2nd ed.) pp. 630 - 635. New York, NY: John Wiley & Sons, 1986.
Friedlander, S. K., "Aerosol Filtration by Fibrous Filters," in Biochemical and
Biological Engineering Science, vol 1., ed. N. Blakebrough. London,
England: Academic Press, Inc., 1967, pp. 49 - 67.
Humphrey, A. E., "Air Sterilization," Adv. Appi Micro. 2 (1960): 301- 311.
Levenspiel, 0., "Longitudinal Mixing of Fluids Flowing in Circular
Pipes,"Ind. Eng. Chem. 50 (1958): 343 - 346.
Levenspiel, 0., Chemical Reaction Engineering (2nd ed.), p. 272. New York, NY:
John Wiley & Sons, 1972.
McCabe, W. L., J. C. Smith, and P. Harriott, Unit Operations of Chemical
Engineering (4th ed.), pp. 76 - 90. New York, NY: McGraw-Hill Book Co.,
1985.
Pasceri, R. E. and S. K. Friedlander, "The Efficiency of Fibrous Aerosol
Filters," Can. J. Chem. Eng., 38 (1960): 212 - 213.
Pelczar, M. J. and R. D. Reid, Microbiology, pp. 441 - 461. New York, NY:
McGraw-Hill Book Co., 1972.
Quesnel, L. B., "Sterilization and Sterility," in Basic Biotechnology, eds. J.
Bu'lock and B. Kristiansen. London, England: Academic Press, 1987, pp.
197 - 215.
Strauss, W., Industrial Gas Cleaning (2nd ed.), pp. 182, 278 - 297. Oxford,
England: Pergamon Press Ltd., 1975.
Wehner, J. F. and R, H. Wilhelm, "Boundary Conditions of Flow
Reactor," Chern. Eng. Sci. 6 (1956): 89 - 93.
9
Agitation and Aeration
9.1 INTRODUCTION
One of the most important factors to consider in designing a
fermenter is the provision for adequate mixing of its contents. The
main objectives of mixing in fermentation are to disperse the air
bubbles, to suspend the microorganisms (or animal and plant
tissues), and to enhance heat and mass transfer in the medium.
Since most nutrients are highly soluble in water, very little mixing
is required during fermentation just to mix the medium as
microorganisms consume nutrients. However, dissolved oxygen in
the medium is an exception because its solubility in a fermentation
medium is very low, while its demand for the growth of aerobic
microorganisms is high.
For example, when the oxygen is provided from air, the typical
maximum concentration of oxygen in aqueous solution is on the
order of 6 to 8 mg/L. Oxygen requirement of cells is, although it can
vary widely depending on microorganisms, on the order of 1 giL h.
Even though a fermentation medium is fully saturated with oxygen,
the dissolved oxygen will be consumed in less than one minute by
organispls if not provided continuously. Adequate oxygen supply to
cells is often critical in aerobic fermentation. Even temporary
depletion of oxygen can damage cells irreversibly. Therefore, gaseous
oxygen must be supplied continuously to meet the requirements for
high oxygen needs of microorganisms, and the oxygen transfer can
be a major limiting step for cell growth and metabolism.
Mixing provided by a laboratory shaker apparatus is adequate to
cultivate microorganisms in flasks or test tubes. Rotary or
reciprocating action of a shaker is effective to provide gentle mixing
and surface aeratiOll. For bench-, pilot-, and production-scale
fermenters, the mixing is usually provided by mechanical agitation
with or without aeration. The most widely used arrangement is the
radial-flow impeller with six flat blades mounted on a disk (Figure 9.1),
which is called flat-blade disk turbine or Rushton turbine.
Radial-flow impellers (paddles and turbines) produce flow
radially from the turbine blades toward the side of the vessel, where
the flow splits into two directions: one part goes upward along the
side, back to the center along the liquid surface, and down to the
impeller region along the agitating shaft; and the other goes
downward along the side and bottom, then back to the impeller
region. On the other hand, the axial flow impellers (propellers and
pitched blade paddles) generate flow downward to the tank bottom,
then up the side and back down the center to the impeller region.
Therefore, the flat-blade disk turbine has the advantage of limiting
the short-circuiting of gas along the drive shaft by forcing the gas,
introduced from below, along the path into the discharge jet.
Db
....--.-.
[J!OW
Top view
Fig. 9.1 Typical dimensions of flat-blade disk turbine

d,: Db : Dw : Dd = 1 : 0.25 : 0.2 : 0.67.
Mass-Transfer Path: The path of gaseous substrate from a gas

bubble to an organelle in a microorganism can be divided into several
steps (Figure 9.2) as follows:
1. Transfer from bulk gas in a bubble to a relatively unmixed gas
layer
2. Diffusion through the relatively unmixed gas layer
3. Diffusion through the relatively unmixed liquid layer
surrounding the bubble
4. Transfer from the relatively unmixed liquid layer to the bulk
liquid
5. Transfer from the bulk liquid to the relatively unmixed liquid
layer surrounding a microorganism
6. Diffusion through the relatively unmixed liquid layer
7. Diffusion from the surface of a microorganism to an organelle
in which oxygen is consumed
Agitation and Aeration 223
Microorganism
Gas bubble
Fig. 9.2 Schematic diagram of the path of a gaseous

substrate to an organelle in a cell.
Steps 3 and 5, the diffusion through the relatively unmixed liquid

layers of the bubble and the microorganism, are the slowest among
those outlined previously and, as a result, control the overall mass-
transfer rate. Agitation and aeration enhance the rate of mass transfer
in these steps and increase the mterfacial area of both gas and liquid.
In this chapter, we study various correlations for gas-liquid mass
transfer, interfacial area, bubble size, gas hold-up, agitation power
consumption, and volumetric mass-transfer coefficient, which are
vital tools for the design and operation of fermenter systems. Criteria
for the scale-up and shear sensitive mixing are also presented. First of
all, let's review basic mass-transfer concepts important in
understanding gas-liquid mass transfer in a fermentation system.
9.2 BASIC MASS-TRANSFER CONCEPTS

9.2. 1 Molecular Diffusion in Liquids
When the concentration of a component varies from one point to
another, the component has a tendency to flow in the direction that
will reduce the local differences in concentration.
Molar flux of a component A relative to the average molal velocity
of all constituent JA is proportional to the concentration gradient
dCA/dz as
. dC A
fA = - D AB - (9.1)
dz
which is Fick's first law written for the z-direction. The DAB in Eq. (9.1)
is the diffusivity of component A through B, which is a measure of its
diffusive mobility.
Molar flux relative to stationary coordinate NA is equal to
CA dC A
N A = C(N A +NB)-DAB~ (9.2)
where C is total concentration of components A and Band N B is the

molar flux of B relative to stationary coordinate. The first term of the
right hand side of Eq. (9.2) is the flux due to bulk flow, and the second
term is due to the diffusion. For dilute solution of A,
NA ::::JA (9.3)
Diffusivity: The kinetic theory of liquids is much less advanced

than that of gases. Therefore, the correlation for diffusivities in
liquids is not as reliable as that for gases. Among several correlations
reported, the Wilke-Chang correlation (Wilke and Chang, 1955) is the
most widely used for dilute solutions of nonelectrolytes,
o 1.117 x 10-16(~MB)0.5T
D AB = V;O.6 (9.4)
J1 bA
When the solvent is water, 5kelland (1974) recommends the use of the
correlation developed by Othmer and Thakar (1953).
o 1.112 x 10-13
D AB = (9.5)
)11.1V~A6
The preceding two correlations are not dimensionally consistent;
therefore, the equations are for use with the units of each term as 51
unit as follows:
D~B diffusivity of A in B, in a very dilute solution, m 2 / s
M B molecular weight of component B, kg/kmol
T temperature, OK
J1 solution viscosity, kg/m 3s
V bA solute molecular volume at normal boiling point, m 3 /kmol
0.0256 m 3 /kmol for oxygen [See Perry and Chilton (p.3 -
233, 1973) for extensive table]
~ association factor for the solvent: 2.26 for water, 1.9 for
methanol, 1.5 for ethanol, 1.0 for unassociated solvents, such
as benzene and ethyl ether.
Example 9.1
Estimate the diffusivity for oxygen in water at 25°C. Compare the
predictions from the Wilke-Chang and Othmer-Thakar correlations
with the experimental value of 2.5 x 10-9 m 2 / s (Perry and Chilton,
p. 3 - 225, 1973). Convert the experimental value to that corresponding
to a temperature of 40°C.
Solution:
Oxygen is designated as component A, and water, component B. The
molecular volume of oxygen \1BA is 0.0256 m 3 /kmol. The association
factor for water ~ is 2.26. The viscosity of water at 25°C is 8.904 x 10-4
kg/m s (CRC Handbook of Chemistry and Physics, p. F-38, 1983). In
Eq. (9.4)
1.173 x 10-16 [2.26(18)]°·5 209
DO =
AB
= 2.25 x 10-9 m 2/ s
(8.904 X 10-4)1.1(0.0256)°·6
InEq. (9.5)
1.112 X 10-13
D~B = = 2.27 X 10-9 m 2 /s
(8.904 x 10-4 )1.1(0.0256)°.6
If we define the error between these predictions and the
experimental value as
0/ (DAB)predicted - (DAB)experimental 100
/0 error = x
(DAB) experimental
The resulting errors are -9.6 percent and -9.2 percent for Eqs. (9.4)
and (9.5), respectively. Since the estimated possible error for the
experimental value is ± 20 percent (Perry and Chilton, p. 3 - 225, 1973),
the estimated values from both equations are satisfactory.
Eq. (9.4) suggests that the quantity DAB Ji!f is constant for a given
liquid system. Though this is an approximation, we may use it here to
estimate the diffusivity at 40°C. Since the viscosity of water at 40°C is
6.529 X 10--4 kg/m s from the handbook,
D° at 40°C = 2.5 X 10-9 (8.904 X 10-4)( 313) = 3.58 x 10-9 m 2 / s

AB 6.529 X 10-4 298
If we use Eq. (9.5), D4B J.111 is constant,
D~B at 40°C = 2.5 X 10-9 (8.904 x 1O~)1.1 = 3.52 X 10-9 m 2 /s

6.529 x 10
9.2.2 Mass-Transfer Coefficient

The mass flux, the rate of mass transfer qc per unit area, is
proportional to a concentration difference. If a solute transfers from
the gas to the liquid phase, its mass flux from the gas phase to the
interface N c is
Nc = q~ = kC<C c - CC1 ) (9.6)
where C c and CCi, is the gas-side concentration at the bulk and the
interface, respectively, as shown in Figure 9.3. KG is the individual
mass-transfer coefficient for the gas phase and A is the mterfacial
area.
226 Fundall1e1ltals of Biochernical Engineering
Similarly, the liquid-side phase mass flux Nt is
(9.7)
where kL is the individual mass-transfer coefficient for the liquid

phase.
Gas LiqUid
CL CLi
Fig, 9.3 Concentration profile near a gas-liquid interface and an

equilibrium curve.
Since the amount of solute transferred from the gas phase to the
interface must equal that from the interface to the liquid phase,
NG =N L (9.8)
Substitution of Eq. (9.6) and Eq. (9.7) into Eq. (9.8) gives
CG -CGi kL
-- = (9.9)
CL -C L. I
kG
which is equal to the slope of the curve connecting the (C L , C G ) and
(C LI , CGI ), as shown in Figure 9.3.
It is hard to determine the mass-transfer coefficient according to
Eq. (9.6) or Eq. (9.7) because we cannot measure the interfacial
concentrations, CLI' or C G1 . Therefore, it is convenient to define the
overall mass-transfer coefticient as follows:
(9.10)
where C~ is the gas-side concentration which would be in
equilibrium with the existing liquid phase concentration. Similarly,
C L is the liquid-side concentration which would be in equilibrium
with the existing gas-phase concentration. These can be easily read
from the equilibrium curve as shown in Figure 9.4. The newly
defined KG and KL are overall mass-transfer coefficients for the gas
and liquid sides, respectively.
CL CLi Ci
Fig. 9.4 The equilibrium curve explaining the meaning of C~ and C~
Example 9.2
Derive the relationship between the overall mass-transfer coefficient
for liquid phase KL and the individual mass-transfer coefficients, kL
and kG. How can this relationship be simplified for sparingly soluble
gases?
Solution:
According to Eqs. (9.7) and (9.10),
kL(C Li - CL) = KL(C! - C1) (9.11)
Therefore, by rearranging Eq. (9.11)
1 C~ - CL
kL CL. - CL
1
=_
1 (Ct. -CL)+(C~ -C L .)
1 1
(9.12)
kL CL. - CL
1
1 l(C~-CL.)
=_+_ I
kL k L Ct. - CL I
Since
kL(C L.- CL) = kG(Cc-C G.) (9.13)
1 1
By substituting Eq. (9.13) to Eq. (9.12), we obtain

*
1 1 CL - CL.
- 1 = _+_ 1
1
=_ 1
+_ (9.14)
KL kL kG CG - CG. I
kL kGM
which is the relationship between K Lf kLf and kG. M is the slope of the
line connecting (C LI , CGI ) and (C~, Cc ) as shown ill Figure 9.4.
For sparingly soluble gases, the slope of the equilibrium curve is

very steep; therefore, M is much greater than 1 and from Eq. (9.14)
KL ~ kL (9.15)
Similarly, for the gas-phase mass-transfer coefficient,
KG ~ kG (9.16)
9.2.3 Mechanism of Mass Transfer

Several different mechanisms have been proposed to provide a basis
for a theory of interphase mass transfer. The three best known are the
two-film theory, the penetration theory, and the surface renewal
theory.
The two1ilm theory supposes that the entire resistance to transfer is
contained in two fictitious films on either side of the interface, in
which transfer occurs by molecular diffusion. This model leads to the
conclusion that the mass-transfer coefficient kL is proportional to the
diffusivity D AB and inversely proportional to the film thickness zf as
kL = DAB (9.17)
zf
Penetration theory (Higbie, 1935)assumes that turbulent eddies
travel from the bulk of the phase to the interface where they remain
for a constant exposure time teo The solute is assumed to penetrate
into a given eddy during its stay at the interface by a process of
unsteady-state molecular diffusion. This model predicts that the
mass-transfer coefficient is directly proportional to the square root of
molecular diffusivity
1/2
kL = 2 DAB (9.18)
( 1C t )
e
Surface renewal theory (Danckwerts, 1951) proposes that there is an
infinite range of ages for elements of the surface and the surface age
distribution function l/J(t) can be expressed as
l/J (t) = sest (9.19)
where s is the fractional rate of surface renewal. This theory predicts
that again the mass-transfer coefficient is proportional to the square
root of the molecular diffusivity
kL = (sDAB)1/2 (9.20)
All these theories require knowledge of one unknown parameter,
the effective film thickness zf the exposure time 4, or the fractional
rate of surface renewal s. Little is known about these properties, so as
theories, all three are incomplete. However, these theories help us to
visualize the mechanism of mass transfer at the interface and also to

know the exponential dependency of molecular diffusivity on the
mass-transfer coefficient.
9.3 CORRELATION FOR MASS-TRANSFER COEFFICIENT

Mass-transfer coefficient is a function of physical properties and
vessel geometry. Because of the complexity of hydrodynamics in
multiphase mixing, it is difficult, if not impossible, to derive a useful
correlation based on a purely theoretical basis. It is common to obtain
an empirical correlation for the mass-transfer coefficient by fitting
experimental data. The correlations are usually expressed by
dimensionless groups since they are dimensionally consistent and
also useful for scale-up processes. The dimensionless group
important for correlations can be derived by using Buckingham Pi
theory as shown in the following example.
Example 9.3
The mass transfer coefficient kL of oxygen transfer in fermenters is a
function of Sauter mean diameter D32, diffusivity DABf and density p(
viscosity J.1c of continuous phase (liquid phase). Sauter-mea n
diameter D32 can be calculated from measured drop-size distribution
from the following relationship,
n
£..J n·D~
"" I I
i=l
D 32 = - n - - - (9.21 )
£..J n·D~
"" I I
i=l
Determine appropriate dimensionless parameters that can relate the
mass transfer coefficient by applying the Buckingham-Pi theorem.
Solution:
The first step of Buckingham-Pi theorem is to count the total number
of parameters. In this case, there are five parameters: kLf D32 , DAB' PC'
and J.1c' all of which can be expressed with three principle units: mass
M, length L, and time T. Therefore,
Number of parameter: n =5
Number of principle dimension: r =3
In developing the dimensionless groups, every dimensionless
group will contain r = 3 of repeating parameters and the total number
of dimensioness group will be n - m as
Number of repeating parameter: m = r =3
Number of dimensionless group: n -m = 5 - 3 = 2
230 Fundarnentals of Biochelllical Engineering
Therefore, we need to choose three repeating parameters. Since we

want to develop a correlation for kL , we want it to be only one
dimensionless group and therefore, cannot be a repeating parameter.
You can choose any three out of D32 , DAB' p, and f.1, although you may
end up different set of dimensionless groups depending on how you
select them. If we choose D32 , DAB' and p as repeating parameters,
two dimensionless groups can be constructed as
n1 = a;2D~B p~ kL (9.22)
n2 = Da32 DifAB PLc'JLc (9.23)
Now, the exponents of the above equations can be determined so that
both groups can be dimensionless by substituting each parameter
with their dimensions as
ni = L a(L 2 /T)b(M/L 3 )C(L/T) (9.24)
Collecting all exponents forM, T, and L unit,
M: c=O
T: -6-1 = 0 b =- 1
I.,: a + 2b - 3c = 0 a =1
Therefore, the fi rst d i Inensionless group is
o -1 0 kL D32 (9.25)
II) = D32DABPckr= - -
DAB
\vhich is known dS Sherwood number N Sh . Similarly,
= D132 D-AB P -0c JiL= Pc

1
1_·1 2 (9.26)
DABpc
which is known as Schnlidt llumber, N Sc
Earlier studies in 1l1aSS transfer between the gas-liquid phase

reported the volumetric mass-transfer coefficient kLa. Since kLa is the
combination of two experiInetltal parameters, mass-transfer
coefficient and mterfacial area, it is difficult to identify which
parameter is responsible for the change of kLll when we change the
operating condition of a fernlenter. Calderbank and Moo-Young
(1961) separated kl.ll by 1l1easuring interfacial area and correlated
mass-transfer coefficients in gas-liquid dispersions in Inixing vessels,
and sieve and sintered plate colulnn, as follo\vs:
1. For small bubbles less than 2.5 1l1Ill in diameter,
k = 0.31 N -2/ J [t1Pj1~.g]1/3 (9.27)

r sc
Pc
N
or N Sit =.
0 31NSc l(1 1/3
LI' . (9.28)
where N Gr is known as Grashof number and defined as

3
t1
Grashof number, NGr = D 32 pcg2 ·,p (9.29)
J1c
The more general forms which can be applied for both small
rigid sphere bubble and suspended solid particle are
ill
kL = 2D.AB +O.31Nsc--2/3(I1PJ1;_g (9.30)
D32 PI J
or
1/1 i1
N Sh =2.0 + 0.31 N!)C NC / (9.31)
Eqs. (9.30) and (9.31) were confirmed by Calderbank and Jonp~

(1961), for mass transfer to and from dispersions of low-density
solid particles in agitated liquids which were designed to
simulate mass transfer to microorganisms in fermenters.
2. For bubbles larger than 2.5 mm in diameter,
1/3
k = 0 42 N -1/2 ~PJ1cg (lJ.12)
L · Sc ( 2
Pc ]
or
N Sh 0 42 N Sc 1/2 N Gr 1/'\
=. (9.33)
Based on the three theories reviewed in the previous section,

the exponential dependency of moll'cular diffllsivity on the
mass-transfer coefficient is expectt'd .to' bl' soml' value between
0.5 and 1. It is interesting tOl1ott' thdt the exponential
dependency of molecular diffusivityin the preceding
correlations is 2/3 or '1/2, v\'hirh is v"ithin the rC1nge predicted
by the theories.
Example 9.4
Estimate the mass-transfer cOt:'fficient for the oxygen dissolution in
water 25°C in a lnixing vesseleqllippl'dvvith flat-hlade disk turbine
and sparger by lIsing ('C1ldl'rbank (lnd Moo- Young's rorrl'lations.
Solution:
The diffusivity of thl' oxygl'n in vvatpr 25°C is 2.5 x lOll m 2 Is
(Example .1). The viscosity and density of water at 25°C is 8.904 x Hl 4
kg/rn s (CR\ Handbook (~r Cht'lllistry lllld Physics, p. F-38, 1983) and
997.08 kg/rn' (Perry and Chilton, p. 3 - 71, 1973},respectively. The
density of air can be calculated from the ideal gas law,
Pau RT 8.314 x 10 (298) 3

. = PM = 1.01325 x 10 (29) = 1.186 k 1m3
g
Therefore the Schmidt number,
sc = _J1_ =
N 8.904 x 10-4 9 = 357.2
pO AB 997.08(2.5 x 10 )
Substituting in Eq. (9.27) for small bubbles,
k = 0.31(357.2)-2/3 [(997.08 -1.186)(8.904 x 10-4)(9.81)]113

L (997.08)2
= 1.27 X 10-4 mls

Substituting in Eq. (9.32) for large bubbles,
k = 0.42(357.2)-1/2 [(997.08 -1.186)(8.904 x 10-4)(9.81)]113

L (997.08)2
= 4.58 X 10-4 mls

Therefore, for the air-water system, Eqs. (9.27) and (9.32) predict
that the mass-transfer coefficients for small and large bubbles are
1.27 x 10-4 and 4.58 x 10-4 mis, respectively, which are independent of
power consumption and gas-flow rate.
Eq. (9.32) predicts that the the mass-transfer coefficient for the
oxygen dissolution in water 25°C in a mixing vessel is 4.58 x 10-4m / s,
regardless of the power consumption and gas-flow rate as illustrated
in the previous example problem. Lopes De Figueiredo and
Calderbank (1978) reported later that the value of kL varies from
7.3 x 10-4 to 3.4 X 10-3 mis, depending on the power dissipation by
impeller per unit volume (Pm/v) as
k L oc (S1!L f33 (9.34)
This dependence of kL on Pm/v was also reported by Prasher and

Wills (1973) based on the absorption of CO2 into water in an agitated
vessel, as follows:
0.25
kL = 0.592 D 1£2 ( Pm
)
(9.35)
VJ.1c
which is dimensionally consistent. However, this equation is limited
in its use because the correlation is based on only one gas-liquid
system.
Akita and Yoshida (1974) evaluated the liquid-phase mass-transfer

coefficient based on the oxygen absorption into several liquids of
different physical properties using bubble columns without
mechanical agitation. Their correlation for k L is
k L 0 32 = 0.5(~) 1/2( gD3i3)1/4( gD 322Pc )3/8 (9.36)

DAB DAB VC (J
where v c and (j are the kinematic viscosity of the continuous phase

and interfacial tension, respectively. Eq. (9.36) is applicable for
column diameters of up to 0.6 m, superficial gas velocities up to
25 m/ s, and gas hold-ups to 30 percent.
9.4 MEASUREMENT OF INTERFACIAL AREA

To calculate the gas absorption rate qL for Eq. (9.7), we need to know
the gas-liquid interfacial area, which can be measured employing
several techniques such as photography, light transmission, and laser
optics.
The interfacial area per unit volume can be calculated from the
Sauter-mean diameter 0 32 and the volume fraction of gas-phase H, as
follows:
6H
a= - (9.37)
0 32
The Sauter-mean diameter, a surface-volume mean, can be
calculated by measuring drop sizes directly from photographs of a
dispersion according to Eq. (9.21).
Photographic measurement of drop sizes is the most
straightforward method among many techniques because it does not
require calibration. However, taking a clear picture may be difficult,
and reading the picture is tedious and time consuming. Pictures can
be taken through the base or the sidewall of a mixing vessel. To
eliminate the distortion due to the curved surface of a vessel wall, the
vessel can be immersed in a rectangular tank, or a water pocket can
be installed on the wall (Skelland and Lee, 1981). One limitation of
this approach is tilat the measurement of drop size is limited to the
regions near the wall, which may not represent the overall dispersion
in a fermenter. Another method is to take pictures by immersing the
extension tube with a objective lense in the tank as described by
Hong and Lee (1983).
Drop size distribution can be indirectly measured by using the
light-transmission technique. When a beam of light is passed through
a gas-liquid dispersion, light is scattered by the gas bubbles. It was
found that a plot of the extinction ratio (reciprocal of light

transmittance, 1 IT) against interfacial area per unit volume of
dispersion a, gave a straight line, as follows (Vermeulen et aI., 1955;
Calderbank, 1958):
1
T = m1+m2a (9.38)
In theory, m1 is unity and m2 is a- constant independent of drop-size

distribution as long as all the bubbles are approximately spherical.
The light transmission technique is most frequently used for the
determination of average bubble size in gas-liquid dispersion. It has
the advantages of quick measurement and on-line operation. The
probes are usually made of mirror-treated glass rods (Vermeulen
et al., 1955), internally blackened tubes with mirrors (Calderbank,
1958), or fiber optic light guide (Hong and Lee, 1983).
9.5 CORRELATIONS FOR A AND D32

9.5.1 Gas Sparging with No Mechanical Agitation
Leaving the vicinity of a sparger, the bubbles may break up or
coalesce with others until an equilibrium size distribution is reached.
A stable size is achieved when turbulent fluctuations and surface
tension forces are in balance (Calderbank, 1959).
Akita and Yoshida (1974) determined the bubble-size distribution
in bubble columns using a photographic technique. The gas was
sparged through perforated plates and single orifices, while various
liquids were used. The following correlation was proposed for the
Sauter-mean diameter:
032 = 26(g D~pc )~05(g~.~ )-0012( Vs )-0.12

(9.39)
Dc (J Vc ~gDc
1($Y!PL r.5( g~~ fl

and for the interfacial area
aD c = H1.13 (9.40)
where Dc is bubble column diameter and V s is superfical gas velocity,

which is gas flow rate Q divided by the tank cross-sectional area.
Eqs. (9.39) and (9.40) are based on data in columns of up to 0.3 m in
diameter and up to superficial gas velocities of about 0.07 In/s.
9.5.2 Gas Sparging with Mechanical Agitation

Calderbank (1958) correlated the mterfacial areas for the gas-liquid
dispersion agitated by a flat-blade disk turbine as follo\\'s:
1. For V s < 0.02 mis,
ao =I.44 [
(Pm/ V )O.4 Pc 0.2
06
](Vs )1/2
- (9.41)
(1. ~
ND )0.3
for NJiei (· VI < 20,000
s
where N Re is the impeller Reynolds number defined as
2
N = D[Npc (9.42)
Rei J1c
The interfacial area for N~: (ND[ I VS )O.3 > 20,000 can be calculated
from the interfacial area ao ~btained from Eq. (9.41) by using the
following relationship.
IOg10 (~~) = (1.95XI0-5)N~~(N~I f3 (9.43)
1. For V 5 > 0.02 m/ s, Miller (1974) modified Eq. (9.41) by

replacing the aerated power input by mechanical agitation Pm
with the effective power input P e' and the terminal velocity V t
with the sum of the superficial gas velocity and the terminal
velocity V s + Vt. The effective power input Pe combines both gas
sparging and mechanical agitation energy contributions. The
modified equation is
ao=1.44[(PeIV)0.4PcO.2]( Vs )1/2 (9.44)

<10.6 Vt + Vs
Calderbank (1958) also correlated the Sauter-mean diameter for
the gas-liquid dispersion agitated by a flat-blade disk turbine
impeller as follows:
1. For dispersion of air in pure water,
<1 0.6 ] 0.5 -4

D32 =4.15 [ 04 02 H + 9.0 x 10 (9.45)
(Pmlv) . Pc .
2. In electrolyte solutions (NaCI, Na2S041 and Na3P04),
D - 2 25 [ (10.6 ] H O.6S ( J1d )0.25

32 (9.46)
-. (P /v)0.4 pc O.2 J1c
m
3. In alcoholic solution (aliphatic alcohols),
032 = 1.90[ (P /v)0.4 Pc(10.6

0.2
]HO.65 (Jld )0.25
J1c
(9.47)
nl
where all constants are dimensionless except 9.0 x 10-4m in Eq.

(9.45). Again the preceding three equations can be modified for high
gas flow rate (V s > 0.02 m/s) by replacing Pm by Pe, as suggested by
Miller (1974).
9.6 GAS HOLD-UP

Gas hold-up is one of the most important parameters characterizing
the hydrodynamics in a fermenter. Gas hold-up depends mainly on
the superficial gas velocity and the power consumption, and often is
very sensitive to the physical properties of the liquid. Gas hold-up
can be determined easily by measuring level of the aerated liquid
during operation ZF and that of clear liquid ZL. Thus, the average
fractional gas hold-up H is given as
H = Zr - ZL (9.48)
ZF
Gas Sparging with No Mechanical Agitation: In a two-phase
system where the continuous phase remains in place, the hold-up is
related to superficial gas velocity V s and bubble rise velocity V t
(Sridhar and Potter, 1980):
H= Vs (9.49)
Vs + Vt
Akita and Yoshida (1973) correlated the gas hold-up for the
absorption of oxygen in various aqueous solutions in bubble
columns, as follovvs:
H = 0.20(gD~PC )1/8(gD~ )1/12(~) (9.50)

(1- H)4 (1 v~ ~gDc
Gas Sparging with Mechanical Agitation: Calderbank (1958)

correlated gas hold-up for the gas-liquid dispersion agitated by a
flat-blade disk turbine impeller as
H= (_V~~r/2 + (216 x 10-4{ (Pm/~:::Pc 0.2]( ~ r/ 2

(9.51)
where 2.16 x 10-4 has a unit (m) and V t = 0.265 mls when the
bubble size is in the range of 2-5 mm diameter. The precedi11g
equation can be obtained by combining Eqs. (9.41) and Eq. (9.45) by
means of Eq. (9.37).
For high superficial gas velocities (V s > 0.02 m/s), replace Pm and
V t of Eq. (9.51) with effective power input Pf and V t + V s, respectively
(Miller, 1974).
Side view Top view

Fig. 9.5 A torque table to measure the power consumption
for mechanical agitation.
9.7 POWER CONSUMPTION

The power consumption for mechanical agitation can be measured
using a torque table as shown in Figure 9.5. The torque table is
constructed by placing a thrust bearing between a base and a circular
plate, and the force required to prevent rotation of the turntable
during agitation F is measured. The power consumption P can be
calculated by the following formula
P = 2ltrNF (9.52)
where N is the agitation speed, and r is the distance from the axis to
the point of the force measurement.
Power consumption by agitation is a function of physical
properties, operating condition, and vessel and impeller geometry.
Dimensional analysis provides the following relationship:
2 2
p3 5
= f(PN0 1 N 0[
" ,
°T !!- , Ow)
, ...
(9.53)
pN D[ Jl g O[ O[ O[
The dimensionless group in the left-hand side of Eq. (9.53) is
known as power number N p , which is the ratio of drag force on
impeller to inertial force. The first term of the right-hand side of Eq.
(9.53) is the impeller Reynolds number N Rei which is the ratio of
inertial force to viscous force, and the second term is the Froude
number NFr which takes into account gravity forces. The gravity force
affects the power consumption due to the formation of the vortex in
an agitating vessel. The vortex formation can be prevented by
installing baffles.
For fully baffled geometrically similar systems, the effect of the
Froude number on the power consumption is negligible and all the
length ratios in Eq. (9.53) are constant. Therefore, Eq. (9.53) is
simplified to
100 ,..-----.--------r------r----~---__.__--__,
10 -I:--~lIE:""""_~r-------+----+----_+_---_+_--____I
Np Flat six-bla e turbine
Propeller
O. 1 +--..J--l................&....L.l.Llf---~......L..L-IU-Ut--L-....L-..L..L-L..L.Uf----L....L....I.....L.L.L...u+-~..L..J....L..L.L.I.If----L-.a.-a...u...u..Lf
1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
NRei
Fig. 9.6 Power number-Reynolds number correlation in an agitator
with four baffles (each O. 1Dr). (Rushton, et aI., 1950)
N p = a (NRe ).8 (9.54)

Figure 9.6 shows Power number-Reynolds number correlation in
an agitator with four baffles (Rushton et al., 1950) for three different
types of impellers. The power number decreases with an increase of
the Reynolds number and reaches a constant value when the
Reynolds number is larger than 10,000. At this point, the power
number is independent of the Reynolds number. For the normal
operating condition of gas-liquid contact, the Reynolds number is
usually larger than 10,000. For example, for a 3-inch impeller with an
agitation speed of 150 rpm, the impeller Reynolds number is 16,225
when the liquid is water. Therefore, Eq. (9.54) is simplified to
N p = constant for N Re > 10,000 (9.55)
The power required by an impeller in a gas sparged system Pm is
usually less than the power required by the impeller operating at the
same speed in a gas-free liquids Pmo. The Pm for the fIat-blade disk
turbine can be calculated from Pmo (Nagata, 1975), as follows:
loglO = Pm =-192 D[
4.38( D[ 2N )0.115 ( D[N 2)1.96(gI J( ~ )
(9.56)
( Dr )
T
Pmo v 9 ND]3
Example 9.5
A cylindrical tank (1.22 m diameter) is filled with water to an
operating level equal to the tank diameter. The tank is equipped with
four equally spaced baffles whose width is one tenth of the tank
diameter. The tank is agitated with a 0.36 m diameter flat six-blade
disk turbine. The impeller rotational speed is 2.8 rps. The air enters
through an open-ended tube situated below the impeller and its
volumetric flow rate is 0.00416 m 3Is at 1.08 atm and 25°C.
Calculate the following properties and compare the calculated
values with those experimental data reported by Chandrasekharan
and Calderbank (1981): Pm = 697 W; H = 0.02; kLa = 0.0217 S-l
a. Power requirement
b. Gas hold-up
c. Sauter-mean diameter
d. Interfacial area
e. Volumetric mass-transfer coefficient
Solution:
a. Power requirement: The viscosity and density of water at 25°C
is 8.904 x 10-4 kg/m 3s (CRC Handbook of Chemistry and
Physics, p. F-38, 1983) and 997.08 kg/m3 (Perry and Chilton,
p. 3-71, 1973), respectively. Therefore, the Reynolds number is
N = pNDf = (997.08)(2.8)(0.36)2 = 406,357

Re J1 8.904 X 10-4
which is much larger than 10,000, above which the power number
is constant at 6. Thus,
P mo = 6p N3Df = 6(997.08)(2.8)3(0.36)5 = 794 W
The power required in the gas-sparged system is irom Eq. (9.56)
_Pm 192(0.36')4.38( 0.36 2(2.8) )0.115

IOglO- - - -- -7
Pmo 1.22 8.93xl0
(0.36)2.8 2)1.96( ~:~~)( 0.00416 )

( 9.81 (2.8)0.36 3
Therefore, Pm = 687 W
b. Gas hold-up: The interfacial tension for the air-water interface
is 0.07197 kgl S2 (CRC Halldbook of Chemistry and Physics,
p. F-33, 1983). The volume of the the dispersion is
v = .~ 1.222 (1. 22) = 1.43m3
4
The superficial gas velocity is
v = ~ = 4(0.00416) = 0.00356 rn/ s

nDT 2 n1.22 2
Substituting these values into Eq. (9.51) gives
H = (0.00356H)1I2 + 2.16 x 10--4[(687/1.43)0.4997.080.2] (0.00356)1/2

0.265 0.07197°·6 0.265
The solution of the preceding equation for H gives
H = 0.023
c. Sauter-mean diameter: In Eq. (9.45)
0.07197°·6 ] 0.5 -4
D 32 = 4.15 [ 04 02 0.023 + 9.0 x 10
(687/1.43) . 997.08 .
= 0.00366m = 3.9 mm!

d. Interfacial area a: In Eq. (9.37)
a = _6H = 6 (0.023) = 37.6 m-1

D32 0.00366
e. Volumetric mass-transfer coefficient: Since the average size of
bubbles is 4 mm, we should use Eq. (9.32) . Then, from
Example. 4
Therefore,
kLa = 4.58 x 10-4 (37.7) = 0.0175-1
The preceding estimated values compare well with those
experimental values. The percent errors as defined in Example .1 are
-1.4 percent for the power consumption, 15 percent for the gas hold-
up, and -21.7 percent for the volumetric mass-transfer coefficient.
9.8 DETERMINATION OF OXYGEN-ABSORPTION RATE

To estimate the design parameters for oxygen uptake in a fermenter,
you can use the correlations presented in the previous sections,
which can be applicable to a wide range of gas-liquid systems in
addition to the air-water system. However, the calculation procedure
is lengthy and the predicted value from those correlations can vary
widely. Sometimes, you may be unable to find suitable correlations
which will be applicable to your type and size of ferment ers. In such
cases, you can measure the oxygen-transfer rate yourself or use
correlations based on those experiments.
The oxygen absorption rate per unit volume q/ v can be estimated
by
qa = K La(C1. - C L) = kLa(Cl - Cr) (9.57)

v
Since the oxygen is sparingly soluble gas, the overall mass-transfer
coefficient K L is equal to tIle individttal mass-transfer coefficient K L .
Our objective in fermenter design is to maximize the oxygen transfer
rate with the minimum power consumption necessary to agitate the
fluid, and also minimum air flow rate. To maximize the oxygen
absorption rate, we have to maximize K L, a, C 1J - Ct - However, the
concentration difference is quite limited for us to control because the
value of Cl is limited by its very low maximum solubility. Therefore,
the main parameters of interest in design are the mass-transfer
coefficient and the mterfacial area.
Table 9.1 lists the solubility of oxygen at 1 atm in water at various
temperatures. The value is the maximum concentration of oxygen in
water when it is in equilibrium with pure oxygen. This solubility
decreases with the addition of acid or salt as shown in Table 9.2.
Normally, we use air to supply the oxygen demand of fermenters.
The maximum concentration of oxygen in water which is in
equilibrium with air C*L at atmospheric pressure is ab0ut one fifth
of the solubility listed, according to the Henry's law,
Table 9.1 Solubility of Oxygen in Water at 1 atm. a
Temperature Solubility
°C mmol 02/L mg 02/L
o 2.18 69.8
10 1.70 54.5
15 1.54 49.3
20 1.38 44.2
25 1.26 40.3
30 1.16 37.1
35 1.09 34.9
40 1.03 33.0
a Data fronl [nternational Critical Tables, Vol. Ill.

New York: Mc(]raw-l-Iill Book Co .. 192R. p, ~7 t.
P0 2
Cl = H. (T) (9.58)
°2
where P0 2 is the partial pressure of oxygen and H02 (T) is Henry's
law constant of oxygen at a.temperature, T. The value of Henry's law
constant can be obtained from the solubilities listed in Table 9.2. For
example, at 25°C, Cl is 1.26 mmol/L and P0 2 is 1 atm because it is
242 Fundamentals of Biochernical Engineering
pure oxygen. By substituting these values into Eq. (9.58), we obtain

H 02 (25°C) is 0.793 atm L/mmol. Therefore, the equilibrium
concentration of oxygen for the air-water contact at 25°C will be
0.209atm
Cl = 0.264 mInoL = O 2 IL - 8.43 mg/L
0.793 atm L/mmol
Ideally, oxygen-transfer rates should be measured in a fermenter
which contains the nutrient broth and microorganisms during the
actual fermentation process. However, it is difficult to carry out such
a task due to the complicated nature of the medium and the ever
changing rheology during cell growth. A common strategy is to use a
synthetic system which approxin1ates fermentation conditions.
Table 9.2 Solubility of Oxygen in Solution of Salt or Acid at 25°C. a
Cone. Solubility, mmol O2 / L

mol/L HCI H2 SO 4 NaCI
0.0 1.26 1.26 1.26
0.5 1.21 1.21 1.07
1.0 1.16 1.12 0.89
2.0 1.12 1.02 0.71
a Data from F. Todt, Electrochemishe Sauer-stotfmess-

ungen, Berlin: W. de Guy & Co., 1958
9.8.1. Sodium Sulfite Oxidation Method

The sodium sulfite oxidation method (Cooper et al., 1944) is based on
the oxidation of sodium sulfite to sodium sulfate in the presence of
catalyst (Cu++ or C~++) as
1a
N a2 SO3 +"2 Cu orCo++ N a2 SO4 (9.59)
2 -------.+)
This reaction has following characteristics to be qualified for the

measurement of the oxygen-transfer rate:
1. The rate of this reaction is independent of the concentration of
sodium sulfite within the range of 0.04 to 1 N.
2. The rate of reaction is much faster than the oxygen transfer rate;
therefore, the rate of oxidation is controlled by the rate of mass
transfer alone.
To measure the oxygen-transfer rate in a fermenter, fill the
fermenter with a 1 N sodium sulfite solution containing at least 0.003
M Cu++ ion. Turn on the air and start a timer when the first bubbles of
air emerge from the sparger. Allow the oxidation to continue for
4 to 20 minutes, after which, stop the air stream, agitator, and timer at
the same instant, and take a sample. Mix each sample with an excess
of freshly pipetted standard iodine reagent. Titrate with standard
sodium thiosulfate solution (Na 2S20 3) to a starch indicator end point.
Once the oxygen uptake is measured, the kL a may be calculated by
using Eq. (9.57) where CL is zero and Cl is the oxygen equilibrium
concentration.
The sodium sulfite oxidation technique has its limitation in the fact
that the solution cannot approximate the physical and chemical
properties of a fermentation broth. An additional problem is that this
technique requires high ionic concentrations (1 to 2 mol/L), the
presence of which can affect the interfacial area and, in a lesser
degree, the mass-transfer coefficient (Van't Riet, 1979). However, this
technique is helpful in comparing the performance of fermenters and
studying the effect of scale-up and operating conditions.
Example 9.6
To measure kL a, a fermenter was filled with 10 L of 0.5 M sodium
sulfite solution containing 0.003 M Cu++ ion and the air sparger was
turned on. After exactly 10 minutes, the air flow Was stopped and a
10 mL sample was taken and titrated. The concentration of the
sodium sulfite in the sample was found to be 0.21 mol/L. The
experiment was carried out at 25°C and 1 atm. Calculate the oxygen
uptake and kLa.
Solution:
The amount of sodium sulfite reacted for 10 minutes is
0.5 - 0.21 = 0.29 mol/L
According to the stoichiometric relation, Eq. (9.59) the amount of
oxygen required to react 0.29 mol/L is
0.29 12 = 0.145mol/L
Therefore, the oxygen uptake is
g
(0.145 mole 02/ L ) (32 02/
600s
mol
~
) = 7.73 10-3 g/Ls
The solubility of oxygen in equilibrium with air can be estimated

by Eq. (9.58) as
C* L-
-
P02 = (1 atm)(O.209 mol 02/ mol air) = 8.43 x lO-4g / L
H o 2 (T) (793 atm L/mol)(1 moll32g)
Therefore, the value of kLa is, according to the Eq. (9.57),

3
ka- qa/ v _ 7.73x10- g/Ls 0.917s-1
L - C~ -C/o - (8.43x10-3 g/L-0)
9.8.2 Dynamic Gassing-out Technique

This technique (Van't Riet, 1979) monitors the change of the oxygen
concentration while an oxygen-rich liquid is deoxygenated by
passing nitrogen through it. Polarographic electrode is usually used
to measure the concentration. The mass balance in a vessel gives
(9.60)
Integration of the preceding equation between t1 and t2 results in
In[C~ - Cd t1)]
CL - CL (t 2 )
k La = --.;;:~---...;;;. (9.61)
t 2 - t1
from which kLa can be calculated based on the measured values of
CL(f 1) and CL(f 2 )
9.8.3 Direct Measurement

In this technique, we directly measure the oxygen content of the gas
stream entering and leaving the fermenter by using gaseous oxygen
analyzer. The oxygen uptake can be calculated as
qa = (QinCo2/in - Qout C02/out) (9.62)

where Q is the gas flow rate.
Once the oxygen uptake is measured, the kL a can be calculated by
using Eq. (9.57), where CL, is the oxygen concentration of the liquid in
a fermenter and Ct is the concentration of the oxygen which would
be in equilibrium with the gas stream. The oxygen concentration of
the liquid in a fermenter can be measured by an on-line oxygen
sensor. If the size of the fermenter is rather small (less than 50 L), the
variation of the Ct - CL in the fermenter is fairly small. However, if
the size of a fermenter is very large, the variation can be significant.
In this case, the log-mean value of C*L - CL of the inlet and outlet of
the gas stream can be used as
air off
deL
dt
Fig. 9.7 Dynamic technique for the determination of kLB.
* *
* C) - (C L - CL)in - (C L - CL)out
(c L - L L M - * * (9.63)
In[(CL - CL)inl(C L - CL)out]
9.8.4 Dynamic Technique

By using the dynamic technique (Taguchi and Humphrey, 1966), we
can estimate the kLa value for the oxygen transfer during an actual
fermentation process with real culture medium and microorganisms.
This technique is based on the oxygen material balance in an aerated
batch fermenter while microorganisms are actively growing as
dCL = kLa(Ct - Cd - '0 Cx (9.64)

dt 2
where '02 is cell respiration rate [g 02 / g cell h].

While the dissolved oxygen level of the fermenter is steady, if you
suddenly turn off the air supply, the oxygen concentration will be
decreased (Figure 9.7) with the following rate
dC L
- = ro Cx (9.65)
dt 2
since kLa in Eq. (9.64) is equal to zero. Therefore, by measuring the

slope of the CL vs. t curve, we can estimate r 02CX . If you turn on the
airflow again, the dissolved oxygen concentration will be increased
according to Eq. (9.64), which can be rearranged to result in a linear
relationship as
C= C1 - _l_(dC
L
kLa dt
L +'0 Cx)
2
(9.66)
The plot of Ct versus dCLldt + r02 will result in a straight line which
has the slope of -II (kLa) and the y-intercept of C1 .
9.9 CORRELAliON FOR KLa
9.9.1 Bubble Column

Akita and Yoshida (1973) correlated the volumetric mass-transfer
coefficient kLa for the absorption of oxygen in various aqueous
solutions in bubble columns, as follows:
-0.62
kL a = 0 . 6 D AB 0.5 V c -0.12.!!.-. D T0.17 gO.93 H1.1 (9.67)
( )
Pc
which applies to columns with less effective spargers.
In bubble columns, for 0 < V s < 0.15 mls and 100 < Pglv < 1100
W 1m 3, Botton et al. (1980) correlated the kLa as
p I )0.75
_ ~Lq_ = ~ (9.68)
0.08 ( 800
where Pg is the gas power input, which can be calculated from
Pg = 800( Vs )0.75 (9.69)

v 0.1
9.9.2 Mechanically Agitated Vessel

For aerated mixing vessels in an aqueous solution, the mass-transfer
coefficient is proportional to the power consumption (Lopes De
Figueiredo and Calderbank, 1978) as
Pm )0.33
kL DC
( - (9.70)
V
The interfacial area for the aerated mixing vessel is a function of

agitation conditions. Therefore, according to Eq. (9.41)
a oc (P; f4 v2.5 (9.71)
Therefore, by combining the above two equations, kLa will be

p )0.77
kLa oc;
( Vs°.5 (9.72)
Numerous studies for the correlations of kLa have been reported

and their results have the general form as
kLa = bl ( ;
p )b2
V: 3 (9.73)
Agil"ation and Aeration 247
where bl , bl , and bl vary considerably depending on the geometry of

the system, the range of variables covered, and the experimental
method used. The values of b2 and b3 are generally between 0 to 1 and
0.43 to 0.95, respectively, as tabulated by Sideman et al. (1966).
Van't Riet (1979) reviewed the data obtained by various
investigators and correlated them as follows:
1. For"coalescing" air-water dispersion,
kLa = 0.026( P~l f4 V2 54

(9.74)
1. For "noncoalescmg" air-electrolyte solution dispersions,
kLa = 0.002 (P; f7 VsO. 2 (9.75)
both of which are applicable for the volume up to 2.6 m:); for a wide
variety of a~itator types, sizes, and D/D r ratios; and 500 < P mil' <
10,000 W 1m . These correlations are accurate \vithin approximately
20 percent to 40 percent.
Example 9.7
Estimate the volumetric mass-transfer coefficient kLa for the gas-
liquid contactor described in Example .4 by using the correlation for
kLa in this section.
Solution:
From Example .4, the reactor volume v is 1.43 m 3, the superficial gas
velocity V s is 0.00356 mis, and power consumption Pill is 687 W. By
substituting these values into Eq. (9.74),
kLa = 0.026 (687 )0.4 0.003560.5 = 0.018 S-1

1.43
9.10 SCALE-UP
9.10.1 Similitude
For the optimum design of a production-scale fermentation system
(prototype), we must translate the data on a small scale (model) to the
large scale. The fundamental requirement for scale-up is that the
model and prototype should be similar to each other.
Two kinds of conditions must be satisfied to insure similarity
between model and prototype. They are:
1. Geometric similarity of the physical boundaries: The model

and the prototype must be the same shape, and all linear
dimensions of the model must be related to the corresponding
dimensions of the prototype by a constant scale factor.
2. Dynamic similarity of the flow fields: The ratio of flow
velocities of corresponding fluid particles is the same in model
and prototype as well as the ratio of all forces acting on
corresponding fluid particles. When dynamic similarity of two
flow fields with geometrically similar boundaries is achieved,
the flow fields exhibit geometrically similar flow patterns.
The first requirement is obvious and easy to accomplish, but the
second is difficult to understand and also to accomplish and needs
explanation. For example, if forces that may act on a fluid element in
a fermenter during agitation are the viscosity force F v , drag force on
impeller F 0' and gravity force Fe, each can be expressed with
characteristic quantities associated with the agitating system.
According to Newton's equation of viscosity, viscosity force is
Fv = ~(~;)A (9.76)
where du/dy is velocity gradient and A is the area on which the

viscosity force acts. For the agitating system, the fluid dynamics
involved are too complex to calculate a wide range of velocity
gradients present. However, it can be assulned that the average
velocity rradient is proportional to agitation speed N and the area
A is to D I' which results.
Fv = DC )1 NDl (9.77)
The drag force F0 can be characterized in an agitating system as
F'/) ex:
Pmo
-- (9.78)
DIN
Since gravity force Fe is equal to mass m times gravity constant g,
Fc oc pD.!, (cs (9.79)
The summation of aJl forces is equal to the inertial force F1 as,
~F = Fv + Fv + F(; = PI oc pOl N 2 (9.80)
Then dynamic similarity between a model (111) and a prototype (p)
is achieved if
(9.81 )
or in dimensionless forms:
(;~ \ = (~ ) (:~) = (;~) (:~) = (:~)

m p m p m
(9.82)
The ratio of inertial force to viscosity force is

F] - pD] 4N 2 - p]
D 2N = N .
(9.83)
Fv - JiND]2 - Ji Rel
which is the Reynolds number. Similarly,

!L = pD[4N
2
= pN 3 D/ = 1 (9.84)
Fv Pmo/D]N Pn10 Np
2 2
F[ = pD/N = D[N =N rr (9.85)
Fe pD]3 g g
Dynamic similarity is achieved when the values of the

nondimensionalparameters are the same at geometrically similar
locations.
(NRei)P = (NRei)m
(Np)p = (NP)m (9.86)
(NFr)p = (NFr)m
Therefore, using dimensionless parameters for the correlation of
data has advantages not only for the consistency of units, but also for
the scale-up purposes.
However, it is difficult, if not impossible, to satisfy the dynamic
similarity when nlore than 011e dimensionless group is involved in a
system, which creates the needs of scale-up criteria. The following
example addresses this problem.
Example 9.8
The power consumption by an agitator in an unbaffled vessel can be
expressed as
p = f(PN~/ ,!,!2gD ])
pN 3 D]5
Can you determine the power consumption and impeller speed of

a I,OOO-gallon fermenter based on the findings of the optimum
condition from a geometrically similar one-gallon vessel? If you
cannot, can you scale up by using a different fluid system?
Solution:
Since Vp/V m = 1,000, the scale ratio is,
(D1)p = 1,0001/3= 10 (9.87)

(D] )m
To achieve dynamic similarity, the three dimensionless numbers
for the prototype and the model must be equal, as follows:
(9.88)
PND]2) = [PND;)
[ /l p /l m
(9.89)
[
N D1 ) = [_~~DI) (9.90)
g P g m
If you use the same fluid for the model and the prototype, Pp = Pm
and J.1 p = J.1m. Canceling out the same physical properties and
substituting Eq. (9.87) to Eq. (9.88) yields
3
5 Np
(P mo)p = 10 (Pmo)m [
- ) (9.91)
N nz
The equality of the Reynolds number requires

N p = 0.01 N m (9.92)
On the otller hand, the equality of the Froude number requires
1
N p = Jl6 N m (9.93)
which is conflicting with the previous requirement for the equality of

th.e Reynolds number. Therefore, it is impossible to satisfy the
requiremerlt of the dynamic similarity unless you use different fluid
systems.
If Pp # Pm and /lp# /lnZ' to satisfy Eqs. (9.89) and (9.90), the following
relationship must hold.
[~ t = 3~.6[~)p (9.94)
Therefore, if the kinematic viscosity of the prototype is similar to

that of water, the kinematic viscosity of the fluid, which needs to be
employed for the model, should be 1/31.6 of the kinematic viscosity

of water. It is impossible to find the fluid whose kinematic viscosity is
that small. As a conclusion, if all three dimensionless groups are
important, it is impossible to satisfy the dynamic similarity.
The previous example problem illustrates the difficulties involved
in the scale-up of the findings of small-scale results. Therefore, we
need to reduce the number of dimensionless parameters involved to
as few as possible, and we also need to determine which is the most
important parameter, so that we may set this parameter constant.
However, even though only one dimensionless parameter may be
involved, we may need to define the scale-up criteria.
As an example, for a fully baffled vessel when N Re > 10,000, the
power number is constant according to Eq. (9.55). For a geometrically
similar vessel, the dynamic similarity will be satisfied by
(pNP30 r5 (P3
rno mo
) ) (9.95)
p = pN 0/ m
If the fluid employed for the prototype and the model remains the
r
same, the power consumption in the prototype is
(Pmo)p = (Pmo)m [~: r[ ~~~~: (9.96)
where (D/)p/(D/)m is equal to the scale ratio. With a known scale ratio
and known operating conditions of a model, we are still unable to
predict the operating conditions of a prototype because there are two
unknown variables, P ma and N. Therefore, we need to have a certain
criteria which can be used as a basis.
9.10.2 Criteria of Scale-Up

Most often, power consumption per unit volume P ma/v is employed
as a criterion for scale-up. In this case, to satisfy the equality of power
numbers of a model and a prototype,
Pmo Pmo N pI
( 0/ ) P = ( 0/ ) m[ N m
(D)
p )3[ ]2 (9.97)
(Or)m
Note that P maiO?) represents the power per volume because the
liquid volume is proportional to DI for the geometrically similar
vessels. For tIle constant P mjDt,
3 [(D1)m ]2
[NNp
m) = (OJ; (9.98)
As a result, if we consider scale-up from a 20-gallon to a 2,500-

gallon agitated vessel, the scale ratio is equal to 5, and the impeller
speed of the prototype will be
N = [(D1 )m
p
]2/3 N m
= 0.34 N m (9.99)
(D1)p
which shows that the impeller speed in a prototype vessel is about
one third of that in a model. For constant P mol v, the Reynolds number
and the impeller tip speed cannot be the same. For the scale ratio of 5,
(NRe.)p = 8.5 (NRe.)m (9.100)
1 1
(ND/)p = 1.7 (ND/)m (9.101)

Table 9.3 shows the values of properties for a prototype (2,500-
gallon) when those for a model (20-gallon) are arbitrarily set as 1.0
(Oldshue, 1966). The parameter values of the prototype depend on
the criteria used for the scale-up. The third column shows the
parameter values of the prototype, when P mjv is set constant. The
values in the third column seem to be more reasonable than those in
the fourth, fifth, and sixth columns, which are calculated based on the
constant value oj· Qlv, ND l , and NDlpl/1, respectively. For example
when the Reynolds number is set constant for the two scales, the
PnlOlv reduces to 0.16 percent of the model and actual power
consumption Pmo also reduces to 20 percent of the model, which is
totally unreasonable.
As a conclusion, there is no one scale-up rule that applies to many
different kinds of mixing operations. Theoretically we can scale up
based on geometrical and dynamic similarities, but it has been shown
that it is possible for only a few limited cases. However, some
principles for the scale-up are as follows (Oldshue, 1985):
1. [t is important to identify which properties are important for
the optimum operation of a mixing system. This can be mass
transfer, pumping capacity, shear rate, or others. Once the
important properties are identified, the system can be scaled up
so that those properties can be maintained, which may result in
the variation of the less important variables including the
geometrical similarity.
2. The major differences between a big tank and a small tank are
that the big tank has a longer blend time, a higher maximum
impeller shear rate, and a lower average impeller shear rate.
3. For homogeneous chemical reactions, the power per volume
can be used as a scale-up criterion. As a rule of thumb, the
intensity of agitation can be classified based on the power input
per 1,000 gallon as shown in Table 9.4
Table 9.3 Properties of Agitator on Scale-Upo
Property Model Prototype

20 gal 2,500 gal
P mo 1.0 125. 3126 25 0.2
Pmd V ' 1.0 1.0 25 0.2 0.0016
N 1.0 0.34 1.0 0.2 0.04
01 1.0 5.0 5.0 5.0 5.0
Q 1.0 42.5 125 25 5.0
QIV 1.0 0.34 1.0 0.2 0.04
N0 1 1.0 1.7 5.0 1.0 0.2
NRe 1.0 8.5 25 5.0 1.0
a Reprinted with permission from 1. Y. Olclshue, "Mixing Scale-Up Techniques" Biotech.
Bioeng. 8 (1966):3-24. ©1966 by John Wile} & Sons. Inc.. New York. NY.
4. For the scale-up of the gas-liquid contactor, the volumetric

mass-transfer coefficient kLa can be used as a scale-up criterion.
In general, the volumetric.mass-transfer coefficient is
approximately correlated to the power per volume. Therefore,
constant power per volume can mean a constant kLa.
5. Typical impeller-to-tank diameter ratio D]/T for fermenters is
0.33 to 0.44. By using a large impeller, adequate mixing can be
provided at an agitation speed which does not damage living
organisms. Fermenters are not usually operated for an
optimum gas-liquid mass transfer because of the shear
sensitivity of cells, which is discussed in the next section.
Table 9.4 Criteria of Agitation Intensity
Horsepower per 1000 gal Agitation Intensity

0.5 - 1 Mild
2-3 Vigorous
4 - 10 Intense
9.11 SHEAR-SENSITIVE MIXING

One of the most versatile fermenter systems used industrially is the
mechanically agitated fermenter. This type of system is effective in
the mixing of fermenter contents, the suspension of cells, the breakup
of air bubbles for enhanced oxygenation, and the prevention of
forming large cell aggregates. However, the shear generated by the
agitator can disrupt the cell membral1e and eventually kill some
microorganisms (Midler and Fin, 1966), animal cells (Croughan et al.,
1987), and plant cells (Hooker et al., 1988). Shear also is responsible
for the deactivation of enzymes (Charm and Wong, 1970). As a result,
254 Fundalnentals oj~ Biochemical Engineering
for optimum operation of an agitated fermentation system, we need

to understand the hydrodynamics involved in shear sensitive
mixing.
For the laminar flow region of Newtonian fluid, shear stress r is
equal to the viscosity f.1 times the velocity gradient du / dy as
du
r = -f.1- (9.102)
dy
which is known as Newton's equation of viscosity. The velocity
gradient is also known as shear rate.
For the turbulent flow,
du
r= -1} - (9.103)
dy
where 1} is eddy viscosity, which is not only dependent upon the
physical properties of the fluid, but also the operating conditions.
Therefore, to describe the intensity of shear in a turbulent system
such as an agitated fermenter, it is easier to estimate shear rate du/ dy
instead of shear stress. I Even though we use the shear rate as a
measure of the shear intensity, we should remember that it is the
shear stress that ultimately affects the living cells or enzyme.
Depe11ding upon the magnitude of viscosity and also whether the
flow is laminar or turbulent, there is a wide range of shear stress
generated for the same shear rate. However, the shear rate is a good
measure for the intensity of the agitation.
In agitated systems, it is difficult, if not impossible, to determine
the shear rate, because of the complicated nature of the fluid
dynamics generated by impellers. A fluid element in an agitated
vessel will go through a wide range of shear rates during agitation:
maximum shear when it passes through the impeller region and
minimum shear when it passes near the corner of the vessel. Metzner
and Otto (1957) developed a general correlation for the average shear
rate generated by a flat-blade disk turbine, based on power
measurements on non-Newtonian liquids, as:
[- dy
dU)
av
= 13 N for N Re . < 20
1
(9.104)
9.12 NOMENCLATURE
A interfacial area, m 2
a gas-liquid interfacial area per unit volume of
dispersion for low impeller Reynolds numbers, m-I
1 For a non-Newtonian fluid, the viscosity is not constant even for the laminar flow. There-
fore, shear rate is easier to estimate than shear stress.
gas-liquid interfacial area per unit volume of

dispersion, m-1
b constant
C concentration, kmol/m 3
CD friction factor and drag coefficient, dimensionless
D diameter of bubble or solid particle, m
diffusivity of component A through B, m 2 Is
diffusivity of component A in a very dilute solution
of B, m 2 Is
Sauter-mean diameter, m
impeller diameter, m
tank diameter, m
impeller width, m
force,N
molar flux of component A relative to the average
molal velocity of all constituents, kmoll m 2s
K overall mass-transfer coefficient, ml s
k individual mass-transfer coefficient, ml s
g acceleration due to gravity, ml S2
H volume fraction of gas phase in dispersion,
dimensionless
impeller speed, S-1
mass flux of A and B relative to stationary coordinate,
kmol/m2 s
Froude number (D j N 2 I g), dimensionless
mass flux from gas to liquid phase and from liquid to
gas phase, respectively, kmol/m 2s
NCr Grashof number (D~2 Pc~pglJ1c2), dimensionless
No number of sparger orifices or sites
Np power number (P molpD~N3), dimensionless
N Reb bubble Reynolds number (D BM V tPcl J1c), dimensionless
N Re . impeller Reynolds number (D~ NP clJ1c), dimensionless
1
N SC Schmidt number (J1 clPc DAB), dimensionless
NSh Sherwood number (k L D32/D AB), dimensionless
P total pressure, N I m 2
Pe effective power input by both gas sparging and
mechanical agitation, W
power dissipated by sparged gas, W
power dissipated by impeller in aerated liquid
dispersion, W
power dissipated by impeller without aeration, W
q the rate of mass transfer, kmoll s

Q volumetric gas-flow rate, m 3 Is
R gas constant
length of radial arm for dynamometer system, m
fractional rate of surface renewal, S-l
exposure time for the penetration theory, s
velocity, m/ s
sparger hole velocity, m/ s
superficial gas velocity, gas-flow rate divided by tank
cross sectional area, m/ s
terminal gas-bubble velocity in free rise, ml s
volume of liquid, m 3
level of the aerated liquid during operation, and that
of clear liquid, m
Zf film thickness in the two-film theory, m
1] 0.06, fraction of jet energy transmitted to bulk liquid
L1" difference in density between dispersed (gas) and
continuous (liquid) phases, kg/m 3
y shear rate, S-l
J1 viscosity, kgl ms
v kinematic viscosity, m 2 I s
(J) angular velocity, S-1
lC absolute pressure, N I m 2
pressure at sparger, N/m 2
P density, kg/m3
l/J(t) surface age distribution function
a- surface tension, kg/s 2
r shear stress, N / m 2
SUBSCRIPT
c continuous phase or liquid phase
d dispersed phase or gas phase
G gas phase
g gas phase
L liquid phase
9.13 PROBLEMS
9.1 Derive the relationship between the overall mass transfer
coefficient for gas phase KG and the individual mass-transfer
coefficients, kL and kG' How.can this relationship be simplified

for sparingly soluble gases?
9.2 Prove that Eq. (9.27) is the same with Eq. (9.28), and Eq. (9.30) is
the same with Eq. (9.31).
9.3 The power consumption by impeller P in geometrically similar
fermenters is a function of the diameter 01 and speed N of
impeller, density p and viscosity )1. of liquid, and acceleration
due to gravity g.Determine appropriate dimensionless
parameters that can relate the power consumption by applying
dimensional analysis using the Buckingham-Pi theorem.
9.4 A cylindrical tank (1.22 m diameter) is filled with water to an
operating level equal to the tank diameter. The tank is equipped
with four equally spaced baffles, the width of which is one
tenth of the tank diameter. The tank is agitated with a 0.36 m
diameter, flat-blade disk turbine. The impeller rotational speed
is 4.43 rps. The air enters through an open-ended tube situated
below the impeller and its volumetric flow rate is 0.0217 m 3 /s
at 1.08 atm and 25°C. Calculate:
a. power requirement
b. gas hold-up
c. Sauter.. mean diameter
d. mterfacial area
e. volumetric mass-transfer coefficient
Compare the preceding calculated results with those
experimental values reported by Chandrasekharan and
Calderbank (1981): Pm = 2282 W; H= 0.086; kLa = 0.0823 s-l.
9.5 Estimate the volumetric mass-transfer coefficient kL a for the
gas-liquid contactor described in Problem .4 by using a
correlation for k L a and compare the result with the
experimental value.
9.6 The power consumption by an agitator in an unbaffled vessel
can be expressed as
2
P3mo = f(pND r )
pN D/ 5 Jl
Can you determine the power consumption and impeller speed
of a l,OOO-gallon fermenter based on findings of the optimum
condition from a one-gallon vessel by using the same fluid
system? Is your conclusion reasonable? Why or why not?
9.7 The optimum agitation speed for the cultivation of plant cells in
a 3-L fermenter equipped with four baffles was found to be 150
rpm.
258 FlIlldll1llClltl1/~ (~f BiocJzcfllical Enxillcerins
a. What should. be the impeller speed of a geometrically

similar 1/000 L fermenter if you scale up based on the same
power consumption per unit volume.
b. When the impeller speed suggested by part (a) was
employed for the cultivation of a 1/000-L fermenter, the
cells do not seem to grow well due to the high shear
generated by the impeller even though the impeller speed
is lower than 150 rpIn of the model system. This may be
due to the higher impeller tip speed which is proportional
to NO I . Is this true? Justify your answer with the ratio of the
impeller tip speed of the prototype to the model fermenter.
c. If you use the impeller tip speed as the criteria for the scale-
up, what "viiI be the impeller speed of the prototype
fermenter?
9.8 The typical oxygen demand for yeast cells growing on
hydrocarbon is about 3 g per g of dry cell. Design a 10-L stirred
fermenter (the diameter and height of fermenter, the type and
diameter of impeller) and determine its operating condition
(impeller speed and aeration rate) in order to meet the oxygen
demand during the peak growth period with the growth rate of
0.5 g dry cell per liter per hour. You can assume that the
physical properties of the medium is the same as pure ,vater.
You are free to make additional assumptions in order to design
the required fermenter.
9.14 REFERENCES
Akita, K. and F. Yoshida, IJGasHoldup and Volumetric Mass Transfer

Coefficient in Bubble Colum.n," I&EC Proc. Des. Dev. 12 (1973): 76-80.
Akita, K. and F. Yoshida, "Bubble Size, Interfacial Area, and Liquid-Phase
Mass Transfer Coefficient in Bubble Columns," I&EC Proc. Des. Dev. 13
(1974):84-91.
Botton,R., D. Cosserat, and J. C. Charpentier, 1J()perating Zone and Scale Up
of Mechanically Stirred Gas-Liquid Reactors," Chenl. Eng. Sci. 35
(1980):82-89.
Bowen, R. L., "Unraveling the Mysteries of Shear-sensitive Mixing Systems,"
Chern. Eng. 9 (1986):55-63.
Calderbank, P. H., "Physical Rate Processes in Industrial Fermentation, Part
I: Interfacial Area in Gas-Liquid Contacting with Mechanical Agitation,"
Trans. Instn. Chern. Engrs. 36 (1958):443-459.
Calderbank, P. H., "Physical Rate Processes in Industrial Fermentation, Part
II: Mass Transfer Coefficients in Gas-Liquid Contacting with and without
Mechanical Agitation," Trans. Instn. Chern. Engrs. 37 (1959):173-185.
Calderbank, P. H. and S. J. R. Jones, "Physical Rate Properties in Industrial
Fermentation - Part III. Mass Transfer from Fluids to Solid Particles
Suspended in Mixing Vessels," Trans. Instn. Chenl. Engrs. 39 (1961):363-
368.
Calderbank, P. H. and M. B. Moo-Young, "The Continuous Phase Heat and
Mass-Transfer Properties of Dispersions," Chern. Eng. Sci. 16 (1961):39-54.
Chandrasekharan, K. and P. H. Calderbank, "Further Observations on the
Scale-up of Aerated Mixing Vessels," Chern. Eng. Sci. 36 (1981):819- 823.
Charm, S. E. and B. L. Wong, "Enzyme Inactivation with Shearing,"
Biotechnol. Bioeng. 12 (1970): 1103-1109.
Cooper, C. M., G. A. Fernstrom, and S. A. Miller, "Performance of Agitated
Gas-Liquid Contactors,"Ind. Eng. Chern. 36 (1944):504-509.
Croughan, M. S., J.-F. Hamel, and D. I. C. Wang, "Hydrodynamic Effects on
Animal Cells Grown in Microcarrier Cultures," Biotech. Bioeng. 29
(1987):130-141.
CRC Handbook of Chemistry and Physics. Cleveland, OH: CRC Press, 1983.
Danckwerts, P. V., "Significance of Liquid-Film Coefficients in Gas
Absorption," Ind. Eng. Chern. 43 (1951): 1460-1467.
Higbie, R. "The rate of absorption of a pure gas into a still liquid during short
periods of exposure," Tram. AIChE 31 (1935):365-389.
Hong, P. O. and J. M. Lee, "Unsteady- State Liquid-Liquid Dispersions in
Agitated Vessels," I&EC Proc. Des. Dev. 22 (1983): 130-135.
Hooker, B. S., J. M. Lee, and G. An, "The Response of Plant Tissue Culture to
a High Shear Environment," Enzyme Microb. Technol. 11 (1989):484-490.
Lopes De Figueiredo, M. M. and P. H. Calderbank, "The Scale-Up of Aerated
Mixing Vessels for Specified Oxygen Dissolution Rates," Chern. Eng. Sci.
34 (1979): 1333-1338.
Metzner, A. B. and R. E. Otto, "Agitation of Non-Newtonian Fluids," AIChE J.
3(1957):3-10.
Midler, Jr., M. and R. K. Finn, "A Model System for Evaluating Shear in the
Design of Stirred Fermentors," Biotech. Bioeng. 8 (1966):71-84.
Miller, D. N., "Scale-Up of Agitated Vessels Gas-Liquid Mass Transfer,"
AIChE. J. 20 (1974):442-453.
Nagata, S., Mixing: Principles and Application, pp. 59-62. New York, NY: John
Wiley & Sons, 1975.
Oldshue, J. Y., "Fermentation Mixing Scale-Up Techniques," Biotech. Bioeng. 8
(1966):3-24.
Oldshue, J. Y., "Current Trends in Mixer Scale-up Techniques," in Mixing of
Liquids by Mechanical Agitation, ed. J. J. Ulbrecht and G. K. Patterson.
New York: Gordon and Breach Science Publishers, 1985, pp. 309-342.
Othmer, D. F. and M. S. Thakar, "Correlating Diffusion Coefficients in
Liquids," Ind. Eng. Chern. 45 (1953):589-593.
Perry, R. H. and C. H. Chilton, Chemical Engineers' Handbook, (5th ed.). New
York, NY: McGraw-Hill Book Co., 1973.
Prasher, B. D. and G. B. Wills, "Mass Transfer in an Agitated Vessel," I&EC
Proc. Des. Dev. 12 (1973):35 1-354.
Rushton, J. H., E. W. Costich, and H. J. Everett, "Power Characteristics of
260 Fundanlentals of Biochemical Engineering
Mixing Impellers, Part II," Chern. Eng. Prog. 46 (1950):467-476.

Sideman, S., O. Hortacsu, and J. W. Fulton, "Mass Transfer in Gas-Liquid
Contacting System," Ind Eng. Chern. 58 (1966):32-47.
Skelland, A. H. P., DiffnlionalMass Transfer, pp. 56-58. New York, NY:
JohnWiley& Sons, 1974.
Skelland, A. H. P. and J. M. Lee, "Drop Size and Continuous Phase Mass
Transfer in Agitated Vessels," AIChE J. 27 (1981):99-111.
Sridhar, T. and O. E. Potter, "Gas Holdup and Bubble Diameters in
Pressurized Gas-Liquid Stirred Vessels," / & E. C. Fundarn. 19 (1980):21-
26.
Taguchi, H. and A. E. Humphrey, "Dynamic Measurement of Volumetric
Oxygen Transfer Coefficient in Fermentation Systems," J. Ferment. 7ec/
z.(Jap.) 44 (1966):881-889.
Van't Riet, K., "Review of Measuring Methods and Results in Nonviscous
Gas-Liquid Mass Transfer in Stirred Vessels," I&EC Proc. Des. Dev.
18(1979):357-364.
Vermeulen, T., G. M. Williams, and G. E. Langlois, "Interfacial Area in
Liquid-Liquid and Gas-Liquid Agitation," Chern. Eng. Prog. 51
(1955):85F-94F.
Wilke, C. R. and P. Chang, "Correlation of Diffusion Coefficients in Dilute
Solutions," AIChE J. 1 (1955):264-270.
10
Downstream Processing
10.1 INTRODUCTION
After successful fermentation or enzyme reactions, desired products
must be separated and purified. This final step is commonly known
as downstream processing or bioseparation, which can account for
up to 60 percent of the total production costs, excluding the cost of
the purchased raw materials (Cliffe, 1988).
The fermentation products can be the cells themselves (biomass),
components within the fermentation broth (extracellular), or those
trapped in cells (intracellular), examples of which are listed in
Table 10.1. As shown in Figure 10.1, if the product of our interest is
the cell, cells are separated from the fermentation -broth and then
washed and dried. In the case of extracellular products, after the cells
are separated, products in the dilute aqueous medium need to be
recovered and purified. The intracellular products can be released by
rupturing the cells and then they can be recovered and purified. The
downstream processing for enzyme reactions will be similar to the
process for extracellular products.
Table 10.1 Examples of Bioprocessing Products
and Their Typical Concentrations
Types Products Concentration
Cell itself baker's yeast, single cell protein 30 gIL
Extracellular alcohols, organic acids, amino acids 100 gIL
Extracellular enzymes, antibiotics 20 gIL
Intracellular recombinant DNA proteins 10 gIL
Bioseparation processes make use of many separation techniques

commonly used in the chemical process industries. However,
bioseparations have distinct characteristics that are not common in
the traditional separations of chemical processes. Some of the unique
characteristics of bioseparation products can be listed as follows:
1. The products are in dilute concentration in an aqueous
medium.
2. The products are usually temperature sensitive.
. . I Extracellular
I
Supernatant ---+ Recovery ---+ I I Punficatlon ---+ products
.. . I
Punflcation ---+
Intracellular
products
Cells ---+
~
Cell products Cell debris
Fig. 10.1 Major process steps in downstream processing.
3. There is a great variety of products to be separated.

4. The products can be intracellular, often as insoluble inclusion
bodies.
5. The physical and chemical properties of products are
similar to contaminants.
6. Extremely high purity and homogeneity may be needed for
human health care.
These characteristics of bioseparations products limit the use of
many traditional separation technologies and also require the
development of new methods.
10.2 SOLID-LIQUID SEPARATION

The first step in downstream processing is the separation of
msolubles from the fermentation broth. The selection of a separation
technique depends on the characteristics of solids and the liquid
medium.
The solid particles to be separated are mainly cellular mass with
the specific gravity of about 1.05 to 1.1, which is not much greater
than that of the broth. Shapes of the particles may be spheres,
ellipsoids, rods, filaments, or flocculents. Typical sizes for various
cells vary widely such as,
bacterial cells: 0.5 - 1 ~m
yeast cells: 1 - 7 7 Jlffi
fungal hyphae: 5-15 ~m in diameter, 50 - 500 llm in length
suspension animal cells (lymphocytes): 10 - 20 llffi suspension
plant cells: 20 - 40 llm
The separation of solid particles from the fermentation broth can
be accomplished by filtration or centrifugation.
10.2.1 Filtration
Filtration separates particles by forcing the fluid through a filtering
medium on \vhich solids are deposited. Filtration can be divided into
Downstream Processing 263
several categories depending on the filtering medium used, the range

of particle sizes removed, the pressure differences, and the principles
of the filtration, such as conventional filtration, microfiltration,
ultrafiltration, and reverse osmosis. In this section we limit our
discussion on the conventional filtration which involves large
particles (d p > 10 f.lm). This technique is effective for dilute
suspension of large and rigid particles. The other filtration
techniques are discussed in the purification section at the end of this
chapter.
A wide variety of filters are available for the cell recovery. There
are generally two major types of filters: pressure and vacuum filters.
The detailed descriptions of those filter units can be found in
Chemical Engineers Handbook (Perry and Chilton, 1973). The two types
of filters most· used for cell recovery are the filter press and rotary
drum filters. A filter press is often employed for the small-scale
separation of bacteria and fungi from broths. For large-scale
filtration, rotary drum filters are usually used. A common filter
medium is the cloth filter made of canvas, wool, synthetic fabrics,
metal, or glass fiber.
Assuming the laminar flow across the filter, the rate of filtration
(dVfl dt) can be expressed as a function of pressure drop ~p by the
modified D'Arcy's equation as (Belter et al., p. 22, 1988),
.l dVf = ilp
(10.1)
A dt J1(LI K)
where A is the area of filtering surface, k is D'Arcy's filter cake
permeability, and L is the thickness of the filter cake. Eq. (10.1) states
that the filtration rate is proportional to the pressure drop and
inversely proportional to the filtration resistance Llk, which is the
sum of the resistance by the filter medium RM and that by the cake Rc,
kL = RM + Rc (10.2)
The value of RM is relatively small compared to R CI which is

proportional to the filtrate volume as
L apcVj
k ::" Rc = -A- (10.3)
where a is the specific cake resistance and Pc is the mass of cake solids
per unit volume of filtrate.
Substitution of Eq. (10.3) into Eq. (10.1) and integration of the
resultant equation yields
L apcVj
k ::"R c = - A - (10.4)
which can be used to relate the pressure drop to time when the
filtration rate is constant.
Eq. (10.4) shows that the higher the viscosity of a solution, the
longer it takes to filter a given amount of solution. The increase of the
cake compressibility increases the filtration resistance a, therefore,
increases the difficulty of the filtration. Fermentation beers and other
biological solutions show non-Newtonian behaviors with high
viscosity and form highly compressible filter cakes. This is especially
true with mycelial microorganisms. Therefore, biological feeds may
require pretreatments such as:
1. Heating to denature proteins
2. Addition of electrolytes to promote coagulation and
flocculation
3. Addition of filter aids (diatomaceous earths or perlites) to
increase the porosity and to reduce the compressibility of cakes
Although the filtration theory reviewed helps us to understand the
relationship between the operating parameters and physical
conditions, it is rarely used as a sole basis for design of a filter system
because the filtering characteristics must always be determined on
the actual slurry in questions (Perry and Chilton, p. 19-57, 1973). The
sizing and scale-up of production-scale filter are usually done by
performingfiltration leaf test procedures (Dorr-Oliver, 1972). From the
test data, the filtration capacity can be expressed either in dry pounds
of filter cakes per square foot of filter area per hour (lb / ft 2 hr) or in
gallons of filtrate per square foot per hour (gal/ft2 hr), depending on
whether the valuable product is the cake or filtrate, respectively.
Since the filtration leaf test is perfortned under ideal conditions, it is
common to apply a safety factor to allow the production variation of
operating conditions. Safety factor commonly used is 0.65 (Dorr-
Oliver, 1972). The following example shows the use of the filtration
leaf data for the sizing of production-scale filter unit.
Example 10.1
Filtration leaf test results indicate that the filtration rate of a protein
product is 50 dry lbs/ (ft 2 hr). What size production filter would be
required to obtain 100 dry Ibs of filter cake per hour?
Solution:
Let's apply safety factor of 0.65, then
50 ~
2
x 0.65 =32.5 ~
2
ft hr ft hr
The size of the filter required is,
1001bs/hr = 3.08 ff
32.51bs (ft 2hr)
10.2.2 Centrifugation
Centrifugation is an alternative method when the filtration is
ineffective, such as in the case of small particles. Centrifugation
requires more expensive equipment than filtration and typically
cannot be scaled to the same capacity as filtration equipment.
Two basic types of large-scale centrifuges are the tubular and the
disk centrifuge as shown schematically in Figure 10.2. The tubular
centrifuge consists of a hollow cylindrical rotating element in a
stationary casing. The suspension is usually fed through the bottom
and clarified liquid is removed from the top leaving the solid deposit
on the bowl's wall. The accumulated solids are recovered manually
from the bowl. A typical tubular centrifuge has a bowl of 2 to 5 in. in
diameter and 9 to 30 in. in height with maximum rotating speed of
15,000 to 50,000 rpm (Ambler, 1979).
The disk centrifuge is the type of centrifuge used most often for
bioseparations. It has the advantage of continuous operation. It
consists of a short, wide bowl 8 to 20 in. in diameter that turns on a
vertical axis (Figure 10.2b). The closely spaced cone-shaped discs in
the bowl decrease the distance that a suspended particle has to be
moved to be captured on the surface and increases the collection
efficiencies. In operation, feed liquid enters the bowl at the bottom,
flows into the channels and upward past the disks. Solid particles are
thrown outward and the clear liquid flows toward the center of the
bowl and is discharged through an annular slit. The collected solids
can be removed intermittently or continuously.
i
(a) (b)
Fig. 10.2 Two basic types of centrifuges: (a) tubular and (b) disk.
266 Fundanzentals of Biochemical Engineering
When a suspension is allowed to stand, the particles will settle

slowly under the influence of gravity due to the density difference
between the solid and surrounding fluid, a process known as
sedimentatiqn. The velocity of a particle increases as it falls and
reaches a constant velocity (known as terminal velocity) at which
Weight force - Buoyancy force = Drag force
The expression for the terminal velocity can be derived from the
balance of the forces acting on a particle as,
d: (Ps - p)a
vt = (10.5)
18/l
which is applicable when the Reynolds number is less than 1
(dpvp/u < 1) which is always the case for biological solutes. In the
case of a sedimentation process, the acceleration term in Eq. (10.5) is
equal to the acceleration due to gravity. Due to the small difference in
density between the cells and the broth, simple settling can take a
long time unless cells are large or the cells form a large aggregate.
Under the centrifugal force, the acceleration term in Eq. (10.5)
becomes
a = oir (10.6)
where OJ is the angular velocity and r is the radial distance from the
center of a centrifuge to a particle. Therefore, the increase in
acceleration by the centrifugal force speeds up the settling process.
10.3 CELL RUPTURE@

Once the cellular materials are separated, those with intracellular
proteins need to be ruptured to release their products. Disruption of
cellular materials is usually difficult because of the strength of the
cell walls and the high osmotic pressure inside. The cell rupture
techniques have to be very powerful, but they must be mild enough
so that desired components are not damaged. Cells can be ruptured
by physical, chemical, or biological methods.
Physical Methods Physical methods include mechanical

disruption by milling, homogenization, or ultrasonication. Typical
high-speed bead mills are composed of a grinding chamber filled with
glass or steel beads which are agitated with disks or impellers
mounted on a motor-driven shaft. The efficiency of cell disruption in
a bead mill depends on the concentration of the cells, the amount and
size of beads, and the type and rotation speed of the agitator. The
optimum wet solid content for the cell suspension for a bead mill is
typically somewhere between 30 percent to 60 percent by volume.
'The amount of beads in the chamber is 70 percent to 90 percent by
volume (Keshavarz et al., 1987). Small beads are generally more

efficient, but the smaller the bead, the harder it is to separate them
from ground solids. Cell disruption by bead mills is inexpensive and
can be operated on a large scale.
A high-pressure homogenizer is a positive displacement pump with
an adjustable orifice valve. It is one of the most widely used methods
for large-scale cell disruption. The pump pressurizes the cell
suspension (about 50 percent wet cell concentration) to
approximately 400 to 550 bar and then rapidly releases it through a
special discharge valve, creating very high shear rates. Refrigerated
cooling to 4 or 5°C is necessary to compensate for the heat geIlerated
during the adiabatic compression and the homogenization steps
(Cliffe, 1988).
An ultrasonicator generates sound waves above 16 kHz, which
causes pressure fluctuations to form oscillating bubbles that implode
violently generating shock waves. Cell disruption by an
ultrasonicator is effective with most cell suspensions and is widely
used in the laboratory. However, it is impractical to be used on a large
scale due to its high operating cost.
Chemical Methods Chemical methods of cell rupture include the
treatment of cells with detergents (surfactants), alkalis, organic
solvents, or by osmotic shock. The use of chemical methods requires
that the product be insensitive to the harsh environment created by
the chemicals. After cell disruption, the chemicals must be easily
separable or they must be compatible with the products.
Surfactants disrupt the cell wall by solubilizing the lipids in the
wall. Sodium dodecylsulfate (SDS), sodium sulfonate, Triton X-IOO,
and sodium taurocholate are examples of the surfactants often
employed in the laboratory. Alkali treatment disrupts the cell walls in
a number of ways including the saponification of lipids. Alkali
treatment is inexpensive and effective, but it is so harsh that it may
denature the protein products. Organic solvents such as toluene can
also rupture the cell wall by penetrating the cell wall lipids, swelling
the wall. When red blood cells or a number of other animal cells are
dumped into pure water, the cells can swell and burst due to the
osmotic flow of water into the cells.
Biological Methods Enzymatic digestion of the cell wall is a good
example of biological cell disruption. It is an effective method that is
also very selective and gentle, but its high cost makes it impractical to
be used for large-scale operations.
10.4 RECOVERY
After solid and liquid are separated (and cells are disrupted in the
case of mtracellular products), we obtain a dilute aqueous solution,
from which products have to be recovered (or concentrated) and

purified. Recovery and purification cannot be divided clearly
because some techniques are employed by both. However, among the
many separation processes used for the recovery step, extraction and
adsorption can be exclusively categorized as recovery and are
explained in this section.
10.4.1 Extraction
Extraction is the process of separating the constituents (solutes) of a
liquid solution (feed) by contact with another insoluble liquid
(solvent). During the liquid-liquid contact, the solutes will be
distributed differently between the two liquid phases. By choosing a
suitable solvent, you can selectively extract the desired products out
of the feed solution into the solvent phase. After the extraction is
completed, the solvent-rich phase is called the extract and the
residual liquid from which solute has been removed is called the
raffinate.
The effectiveness of a solvent can be measured by the distri-
bution coefficient K,
K =L (10.7)
x
where y* is the mass fraction of the solute in the extract phase at
equilibrium and x is that in the raffinate phase. A system with a large
K value is desirable, since it requires less solvent and produces a
more concentrated extract phase. The K value can be increased by
selecting the optimum pH. For example, the K value for penicillin F
between the liquid phases composed of water and amyl acetate is 32
at pH 4.0, however, it drops to 0.06 at pH 6.0 (Belter et al., p. 102,
1988). The addition of countenons such as acetate and butyrate can
also increase the K value dramatically.
Another measure for the effectiveness of solvent is the selectivity f3,
f3 = y ~ / xa (10.8)
y'b/Xb
which is the ratio of the distribution coefficient of solute a and that of
solute b. For all useful extraction operations, the selectivity must be
larger than 1. Other requirements for a good solvent include the
mutual insolubility of the two liquid systems, easy recoverability, a
large density difference between the two phases, nontoxicity, and low
cost.
The most widely used extractor is the mixer-settler that isa
cylindrical vessel with one or several agitators. The vessel is usually
equipped with four equally spaced baffles to prevent the vortex
formation. A radial-flow type agitator such as a flat-bladed turbine is

better for the extraction process than an axial-flow type sU.ch as a
propeller because the relatively low-density difference between the
two liquid systems does not require a strong downward flow toward
the bottom of the vessel.
The mixer and settler can be combined in one vessel or separated.
In the combined case, the agitator is turned off after extraction so that
the two phases can separate. For continuous operation, the mixer and
settler are separated. A settler is nothing but a large tank.
Single-stage Extraction
Extraction can be carried out as a single-stage operation either in
batch or continuous mode. Figure 10.3 shows the flow diagram for a
single-stage mixer-settler. For single-stage extraction design, we need
to estimate the concentration of solute in the extract and in the
raffinate with a given input condition. The overall material balance
for the mixer-settler yields
F, XF R,x ...
,...... -,..
5, Ys E,y
,..... .....
Fig 10.3 Flow diagram for the single-stage mixer settler.
F+5=R- E (10.9)
and the material balance for a solute gives
FX F +SYs = Rx + Ey (10.10)
If we assume the mixer-settler has reached an equilibrium,
y = Kx (10.11)
If inlet conditions (F, 5, xF' and Ys) are known, we have four
unknown variables (R, E, x, and y). However, since we have only
three equations, we need additional information to be able to solve
for the unknown variables, which are the equilibrium data of the
ternary system: solute, solvent, and diluent, which are usually
described graphically in triangular coordinates (Treybal, 1980).
In bioseparations, the solute concentration in the feed is usually
low, therefore, the changes of the extract and the raffinate streams are
negligible. We can assume that F = Rand 5 = E. In that case, we have
only two unknown variables, x and y, so we can solve it to obtain,
270 Fundanlentals of Biochemical Engineering
RXF + Eys (10.12)

Y= RIK+E
If pure solvent is used, Is = 0 and Eq. (10.12) can be simplified to
KXF
Y= l+K(EjR) (10.13)
which shows that the solute concentration in the extract depends on
the distribution coefficient K and the ratio of the extract and the
rafmate EIR.
Another parameter that is important in extraction is the
recoverability of solute. Even though Y is high, if the recoverability of
solute is low, the extractor cannot be regarded as an efficient unit. The
recoverability y can be defined as the fraction of the solute recovered
by an extractor as
Ey
y=- (10.14)
RXF
Substitution of Eq. (10.13) into Eq. (10.14) gives

K(EIR)
(10.15)
Y= l+K(EjR)
With a given system of constant K, a decrease of EIR increases y,

but decreases y. Therefore, an optimum operation condition has to be
determined based on the various factors affecting the economy of the
separation processes, such as the value of products, equipment costs,
and operating costs. It is interesting to note that Y depends on the
ratio EI R, but not on the values of E and R. Can we increase E and R
indefinitely to maintain the same y as long as E/R is constant for a
continuous extractor? The answer is "no." We should remember that
Eq. (10.13) is based on the assumption that the extractor is in
equilibrium. Therefore, the increase of E and R will shorten the
residence time; as a result, the extr-actor cannot be operated in
equilibrium and y will decrease.
Multistage Extraction
The optimum recoverability of solute by a single-stage extractor is
determined based on the K value and the E/R ratio. To increase the
recoverability further, several extractors can be connected
crosscurrently or countercurrently.
Multistage Crosscurrent Extraction In multistage crosscurrent

extraction, the raffinate is successively contacted with fresh solvent
which can be done continuously or in batch (Figure 10.4). With the
given flow rates of the feed and the solvent streams, the
concentrations of the extracts and the raffmates can be estimated in

the same manner as a single-stage extractor.
Multistage Countercurrent Extraction The two phases can be

contacted countercurrently as shown in Figure 10.5. The
countercurrent contact is more efficient than the crosscurrent contact
due to the good distributions of the concentration driving force for
the mass transfer between stages. The disadvantage of this scheme is
that it cannot be operated in a batch IIlode. The overall material
balance for the countercurrent extractor with N p stages gives
Ys Ys Ys
F
2 3
x
Fig. 10.4 Flow diagram for the crosscurrent multistage mixer-settler.
~ S
n
F
Fig 10.5 Flow diagram for the countercurrent multistage mixer-settler.
F+5 = RN p + E1 (10.16)
The material balance for the solute yields
FXF + SYs = RN PxNP + E1Yl (10.17)
If the flow rates of the raffmate and extract streams are constant,
F = RN p = Rand 5 = E1 =E. Therefore, Eq. (10.17) can be rearranged to
Yl= ~(XF-XNp)+YS (10.18)

which can be used to estimate YI with the known feed conditions
(F = R, S = E, x F, Ys) and the separation required (x Np ). The
concentrations of the intermediate streams can be estimated from the
solute balance for the first n stages as
R
Yn+l = Yl - E(XF - xn ) (10.19)
which is operating line for this extractor. For a linear equilibrium

relationship Yn = Kx n,
Fig. 10.6 The graphical estimation of the number of ideal

stages for multistage countercurrent extraction.
Yn + 1 = Yl - ~ ( xF - ~ ) (10.20)
which can be used to calculate Yn + 1 with a known value of Yn starting

from Yl' The total number of ideal stages required to accomplish a
certain job (Np ) can be estimated by continuing this calculation for
Yn + 1 until Yn + I ~ Ys' Another convenient equation that can be used to
calculate N p is (Treybal, 1980),
In[ ::p -_Y::/~ (l-fE)+fE]

(10.21)
Np = In (KE)
-
R
The number of ideal stages and intermediate concentrations of the
extract and the raffmate phases can be estimated graphically as
shown in Figure 10. 6. The equilibrium line (y* =Kx) and the operating
line, Eq. (10.19), are plotted. The operating line passes a point (x F, Yl)
and has the slope of R/ E. Then, steps are drawn as shown in Figure
10.6 until the final step passes a point (x N p Ys).
Example 10.2
Penicillin F is to be extracted from the clarified fermentation beer by
using pure amyl acetate as solvent at pH 4.0. The distribution
coefficient K of the system was found to be 32. The initial
concentration of penicillin in the feed is 400 mg/L. The flow rates of

the feed and the solvent streams are 500 L/hr and 30 L/hr,
respectively.
a. How many ideal stages (countercurrent contact) are required to
recover 97 percent of penicillin in the feed?
b. If you use three countercurrent stages, what will be the percent
recovery?
c. If you use three crosscurrent stages with equal solvent flow rate
(10 L/hr each), what will be the percent recovery?
Solution:
a. The mass fraction of the solute in the feed can be calculated by
assuming the density of the feed is the same as water,
400 . -4
XF =- = 4.0 x 10
10 6
Since pure solvent was used, Ys = O. For 97 percent recovery of
the solute,
XN
p
= XI' (1 - 0.97) = 1.2 x 10-5
and R = 500 = 0.52

KE 32(30)
From Eq. (10.21)
4
4.0 .....__.
In --_.__ X 10--
. (1- 0.52) + 0.52 ]
[ 1.2 x 10-5
N = = 4.3
p In(1/0.52)
The total number of ideal stages is 4.3.
b. Since N p = 3, from Eq. (10.21),
4.0 x 10--4 ]
In -~- (1 - 0.52) + 0.52
[
3=
In(1/0.52)
Solving the preceding equation for X3 yields
x3 = 2.9 X 10-5
Therefore, the percent recovery of the solute is
Y= xF - x3 X 100 = 930/0
xF
c. The concentrations of the extract and the raffinate leaving the

first extractor can be calculated from Eq. (10.13) and the
equilibrium relationship as
KXF = 32(40 x 10-4) =7.8 X 10-3

y= l+K(E/R) 1+32(10/500)
Xl = h
= 2.4 X 10-4
K
Those streams leaving the second and third stages can be
calculated from Eq. (10.13) by replacing XF with Xl and X2'
respectively.
Y2 = 4.8 X 10- , X2 = 1.5 X 10-4
3
Y3 = 2.9 X 10-3, x2 = 9.1 X 10-4

(Ts Therefore, the percent recovery of the solute is
Y= xF - x3 X 100 = 770/0
xl'
which is lower than that of the countercurrent contact.
10.4.2 Adsorption
A specific substance in solution can be adsorbed selectively by
cer~ain solids as a result of either the physical or the chemical
interactions between the molecules of the solid and substance
adsorbed. Since the adsorption is very selective while the solute
loading on the solid surface is limited, adsorption is an effective
method for separation of very dilutely dissolved substances.
Adsorption can be classified into three categories: conventional
adsorption, ion exchange, and affinity adsorption.
Conventional Adsorption Conventional adsorption is a reversible

process as the result of intermolecular forces of attraction (van der
Waals forces) between the molecules of the solid and the substance
adsorbed. Among many adsorbents available in industry, activated
carbons are most often used in bioseparations. Activated carbons are
made by mixing organic matter (such as fruit pits and sawdust) with
inorganic substances (such as calcium chloride), and the carbonizing
and activating with hot air or steam. They are often used to eliminate
the trace quantities of impurities from potable water or processing
liquids. However, they are also being used for the isolation of
valuable .products from the fermentation broth by adsorption and
then recovery by elution.
lon-exchange An ion-exchange resin is composed of three basic

components: a polymeric network (such as styrene-divmyl-benzene,
acrylate, methacrylate, plyamine, cellulose, or dextran), ionic
functional groups which are permanently attached to this network
(which may be amons or cations), and counterions. For example,
strong-acid cation-exchange resins contain fixed charges like -50 3,
which are prepared by sulfonating the benzene rings in the polymer.
Fixed ionic sites in the resin are balanced by a like number of charged
ions of the opposite charge (countenons) to maintain electrical
neutrality.
When a solution containing positively charged ions contacts the
cation-exchange resins, the positively charged ions will replace the
counter ions and be separated from the solution which is the
principle of the separation by ion-exchange. To illustrate the
principle of ion-exchange, consider the reaction of the hydrogen form
of a cation-exchange resin (H+R-) with sodium hydroxide
H+R- + Na+OH- ~ Na+R- + HOH (10.22)
In this reaction, the positively charged ions in the solution (Na+)
replaced the counterion (H+). In the case of a hydroxide-form anion-
exchange resin (R+OH-) with hydrochloric acid,
(10.23)
where the negatively charged ions in the solution (CI-) replaced the
counter ion (OH-).
Fig. 10.7 Principle of affinity adsorption. The specific ligand (L),

such as enzyme inhibitor and antigen, is immobilized
on a water-insoluble carrier (C). A solute (8) can react
selectively with the affinity ligand.
Affinity Adsorption Affinity adsorption is based on the chemical

interaction between a solute and a ligand which is attached to the
surface of the carrier particle by covalent or ionic bonds. The
principle is illustrated in Figure 10.7.
Examples of ligand/solute pairs are (Schmidt-Kastner and G6lker,

1987):
enzyme inhibitors - enzymes
antigens or haptens - antibodies
glycoprotems or polysacchandes - lectins
complementary base sequences - nucleic acids
receptors - hormons
carrier proteins - vitamins
Affinity adsorption offers high selectivity in many bioseparations.
However, the high cost of the resin is a major disadvantage and limits
its industrial use.
Adsorption Isotherm
For separation by adsorption, adsorption capacity is often the most
important parameter because it determines how much adsorbent is
required to ac~omplish a certain task. For the adsorption of a variety
of antibiotics, steroids, and hormons, the adsorption isotherm
relating the amount of solute bound to solid and.that in solvent can
be described by the empirical Freundlich equation.
y* = bXc (10.24)
where y* is the equilibrium value of the nlass of solute adsorbed per
mass of adsorbent and X is the mass fraction of solute in the diluent
phase in solute-free basis. 1 The constants band c are determined
experimentally by plotting log y* versus log X
Fig. 10.8 Flow diagram for the single-stage contact filtration unit.
Another correlation often employed to correlate adsorption data

for proteins is the Langmuir isotherm,
1 Since the concentration in the adsorbent phase Y is expressed as a solute-free basis,

the concentration in the diluent phase X is also expressed on a solute-free basis for
uniformity. However, for dilute solutions, the difference between X (mass fraction on a
solute-free basis) and x (mass fraction in solution including solute) is negligible.
y* = YmaxX (10.25)
K L +X
where Ymax is the maximum amount of solute adsorbed per mass of
adsorbent, and K L is a constant.
Adsorption Operation
Adsorption can be carried out by stagewise or continuous-contacting
methods. The stagewise operation of adsorption is called contact
filtration because the liquid and the solid are contacted in a mixer
and then the solid is separated from the solution by filtration.
Single-stage Adsorption Contact filtration can be carried out as a
single-stage operation (Figure 10.8) either in a batch or a continuous
mode. If we assume that the amount of liquid retained with the solid
is negligible, the material balance for a solute gives
(10.26)
Xo 2
Fig. 10.9 Two-stage crosscurrent adsorption
or ~ = X o -Xl (10.27)
Ls YI - Yo
where 5 s andL s are the mass of adsorbent and of diluent,
respectively. If we also assume that the contact filtration unit is
operated at equilibrium and that the equilibrium relationship can be
approximated by the Freundlich or Langmuir equation, the
adsorbent-solution ratio (5 s/ L s) can be estimated from Eq. (10.27)
after substituting in the equilibrium relationship for Y1 .
Multistage Crosscurrent Adsorption The amount of adsorbent

required for the separation of a given amount of solute can be
decreased by employing multistage crosscurrent contact, which is
usually operated in batch mode, although continuous operation is
also possible. The required adsorbent is further decreased by
increasing the number of stages. However, it is seldom economical to
use more than two stages due to the increased operating costs of the
additional stages.
If the equilibrium isotherm can be expressed by the Freundlich
equation and fresh adsorbent is used in each stage (Yo = 0), the total
amount of adsorbent used for a two-stage crosscurrent adsorption
unit (Figure 10.9) is
SSt +SS2 = Xo -Xl + Xl -X 2

(10.28)
Ls bXf bX 2
For the minimum total adsorbent, f
~(S51 +S52 ) =0 (10.29)

dX 1 Ls
Xn-1 XNp
::~ ~
Ls
n Np
.. Yn Yn+ 1 XNp+1
5s
Fig. 10.10 Multistage countercurrent adsorption

Therefore, from Eqs. (10.28)and (10.29), the relationship for the
intermediate concentrations for the minimum total adsorbent can be
obtained as
(10.30)
Multistage Countercurrent Adsorption The economy of the

absorbent can be even further improved by multistage
countercurrent ,adsorption (Figure 10.10). Like extraction, the
disadvantage of this method is that it cannot be operated in a batch
mode. Since the equilibrium line is not usually linear for adsorption,
it is more convenient to use a graphical method to determine the
number of ideal stages required or the intermediate concentrations
rather than using an analytical method. The solute balance about the
n stages is
LsXo + SSYn + 1 = LSXn + SSY1 (10.31)
which is the operating line.
The number of ideal stages can be estimated by drawing steps
between the equilibrium curve and the operating line as shown in
Figure 10.11. The intermediate concentrations can be also read from
the figure.
Example 10.3
We are planning to isolate an antibiotic from a fermentation broth
(10 L) by using activated carbon. The concentration of the antibiotic is
1.1 x 10-6 g per g water. Ninety-five percent of the antibiotic in
solution needs to be recovered. Absorption studies at the operating
condition gave the following result.
x x 106 (g solutelg water) 0.1 0.3 0.6 0.9 1.2

3
Y* x 10 (g solute/g carbon) 1.3 1.7 2.3 2.4 2.6
Fig. 10.11 The graphical estimation of the number of ideal stages

for countercurrent multistage adsorption.
a. Which isotherm (Freundlich, Langmuir, or linear) fits the data
best? Determine the experimental constants for the isotherm.
b. If we use single-stage contact filtration, how much absorbent is
needed?
c. If we use a two-stage crosscurrent unit, what is the minimum
total amount of absorbent? How much absorbent is introduced
to each stage?
d. If we use a two-stage countercurrent unit, how much total
absorbent is needed?
Solution:
a. In order to test which isotherm fits the data best, linear
regression analysis was carried out for the adsorption data. The
results are as follows:The mass fraction of the solute in the feed
can be calculated by assuming the density of the feed is the
same as water,
280 Fundamentals oj" Biochemical Engineering
Isotherm Plot Slope Intercept Carr. Cae".

Freundlich In Xvs. In Y* 0.29 -0.87 0.99
5
Langmuir 1/Xvs. 1/Y* 4.1 x 10- 381 0.97
Linear Xvs. Y* 1160 0.0013 0.95
The results show that any isotherm can fit the data very well. If we
choose the the Freundlich equation, which fits the adsorption data
the best, the slope and the intercept of the 10gX versus logY* plot
yields the equation,
y* = 0.13XO· 29
b. The concentration of the inlet stream was given as X o= 1.1 x
10-6. Since 95 percent of the solute in the input liquid stream is
recovered, the concentration of the outlet stream will be
X 2 = 0.05Xo = 5.5 x 10-8
From Eq. (10.27) and the Freundlich equation, Eq. (10.24), the
amount of the absorbent required is
5 - L Xo - X 2 = 9.6 g
5 - 5 bX
2
c. The minimum usage of absorbent is realized when Eq. (10.30) is
satisfied,
X 8)0.29 _0.29(1.1 X10-6) = 1 _ 0.29

( 5.5 x 10- Xl
The solution of the preceding equation yields Xl = 3.3 X 10-7.
Therefore, the alnount of adsorbent introduced into the first
and the second stages
5~ - L ~o - Xl = 4.2 g
51 - 5 bXf
5 - L Xl - X 2 = 2.5 g
52 - S bX
2
The total amount of adsorbent is 6.7g which is significantly less
than what was calculated for single-stage contact filtration. It
should be noted. that the minimum amount of adsorbent usage
is realized when a larger amount of the adsorbent is fed to. the
first stage followed by the smaller amount to the second stage.
d. For two-stage countercurrent adsorption, a solute balance of
both stages yields
5 - L X o -X 2
5 - 5 bXf
DO'lDl1stream Processing 281
Xo
Fig. 10.12 Breakthrough curve for fixed-bed adsorption.

A solute balance for the second stage only yields
S - L Xl - X 2
S - s bX
2
Equating the two equations results in
X o -X 2 _ Xl -X 2
bX1 bX 2
Inserting the given values and solving will yield Xl = 5.8 X 10-7
The amount of adsorbent required is
S =L Xl - X2 = 4.9
S S bXf g
Therefore, further reduction of adsorbent usage can be realized
with the countercurrent system.
Fixed-Bed Adsorption The adsorbent can be packed in a

cylindrical column (or tube in small scale) and the diluent passed
through for the selective adsorption of the solute of interest. It is one
of the most common methods for adsorption in both laboratory or
industrial scale separations.
One simple way to analyze the performance of a fixed-bed
adsorber is to prepare a breakthrough curve (Figure 10.12) by
measuring the solute concentration of the effluent as a function of
time. As the solution enters the column, most of the solute will be
adsorbed in the uppermost layer of solid. The adsorption front will
move downward as the adsorption progresses. The solute
concentration of the effluent will be virtually free of solute until the
adsorption front reaches the bottom of the bed, and then the
concentration will start to rise sharply. At this point (t b in Figure
10.12), known as the break point, the whole adsorbent is saturated
with solute except the adsorption layer at the very bottom. If the
solution continues to flow, the solute concentration will continue to
increase until it is the same as the inlet concentration (Xo). Normally,
the adsorption is stopped at the break point and the adsorbed
material is eluted by washing the bed with solvent at conditions
suitable for desorption. The amount of solute lost with the diluent
can be estimated from the graphical integration f)f the breakthrough
curve.
10.5 PURIFICATION
After a product is recovered or isolated, it may need to be purified
further. The purification can be accomplished by numerous methods
such as precipitation, chromatography, electrophoresis, and
ultrafiltration.
10.5.1 Precipitation
Precipitation is widely used for the recovery of proteins or
antibiotics. It can be induced by the addition of salts, organic
solvents, or heat.
The addition of salt precipitates proteins because the protein
solubility is reduced markedly by the increase of salt concentration in
solution. Precipitation is effective and relatively inexpensive. It
causes little denaturation.. Ammonium sulphate is the most
commonly employed salt. The disadvantage of ammonium sulphate
is that it is difficult to remove from the precipitated protein. Sodium
sulfate is an alternative but it has to be used at 35-40°C for adequate
solubility.
The use of organic solvents at lower temperature (less than ~5°C)
precipitates proteins by decreasing the dielectric constant of the
solution. The organic solvents should be miscible in water to be
effective. Acetone, ethanol, methanol, and isopropanol are commonly
employed organic solvents.
Heating also can promote the precipitation of proteins by
denaturing them. It is often used to eliminate unwanted proteins in a
solution. However, the selective denaturation without harming the
desired protein products can be difficult and often risky.
10.5.2 Chromatography
Chromatographic processes always involve a mobile phase and a
stationary phase. The mobile phase is the solution containing solutes
to be separated and the eluent that carries the solution through the
stationary phase. The stationary phase .can be adsorbent,
ion-exchange resin, porous solid, or gel, which are usually packed in

a cylindrical column.
A solution composed of several solutes is injected at one end of the
column and the eluent carries the solution through the stationary
phase to the other end of the column. Each solute in the original
solution moves at a rate proportional to its relative affinity for the
stationary phase and comes out at the end of the column as a
separated band. Depending on the type of adsorbent or the nature of
the solute-adsorbent interaction, they are called adsorption, ion-
exchange, affinity, or gel filtration chromatography.
The basic principles of adsorption, ion-exchange, and affinity
resins have been explained in the previous section on adsorption.
Chromatography is similar to adsorption because both involve the
interaction between solute and solid matrix. However, they are
different in a sense that chromatography is based on the different rate
of movement of the solute in the column, while adsorption is based
on the separation of one solute from other constituencies by being
captured on the adsorbent.
Gel filtration chromatography uses gel such as cross-linked
dextrans, polyacrylamide, and agarose as the stationary phase. The
gel contains pores of defined sizes. The smaller molecules penetrate
the pore structure to a greater extent and therefore have a longer
retention time than the larger molecules.
Chromatography has been used to purify proteins and peptides
from comple,\ liquid solutions extensively on the laboratory scale.
When chromatographic processes can be scaled up from laboratory
scale to a larger scale, the diameter rather than the height of a column
should be increased. In this way only the flow rate is increased
without changing the retention time and other operating parameters.
For large-scale chromatography, a stationary phase has to be selected
which has the required mechanical strength and chemical stability
properties.
For designing and operating a chromatograph, the yield and
purity of the separated products are the two important parameters to
be controlled, which can be estimated as follows: The yield of a
solute i collected between two times t 1 and t2 during chromatographic
separation can be calculated as (Belter et aI., p.188, 1988)
t2
. ld amount of solute i eluted Jt] Yi Lsdt
Yle =
total amount of solute
=
J; Yi Lsdt (10.32)
where Yi is the concentration of a solute i and Ls is the solvent flow

rate. The change of the solute concentration can be approximated by
the Gaussian form,
(tit _1)2]
Y1. = y.Imax exp [ _ max
20- 2
(10.33)
where Yi max is the maximum concentration, t max is the time for the
maximum, and 0- is the standard deviation of the peak.
Substitution of Eq. (10.33) into Eq. (10.32) and integration results in
yield = ~{erf[- t1ltmax -1] _ erf[- t2 /t max

2 Jier Jier
-I]} (10.34)
The purity of a solute i collected between two times t 1 and t 2

during chromatographic separation can be calculated as
amount of solute i eluted f~2 Yi Lsdt

purity = = ----- (10.35)
amount of impurity eluted L f~2 Y j Lsdt
]
10.5.3 Electrophoresis
When a mixture of solutes is placed in an electrical field, the
positively charged species are attracted to the anode and the
negatively charged ones to the cathode. The separation of charged
species based on their specific migration rates in an electrical field is
termed electrophoresis.
It is one of the most effective methods of protein separation and
characterization. The chief advantages of this method are that it can
be performed under very mild conditions and it has high resolving
power, resulting in the clear separation of similarly charged protein
molecules. However, in order to use this separation technique, the
components of a mixture must have an ionic form, and each
component must possess a different net charge.
When a charged particle q moves with a steady velocity UE through
a fluid under an electric field Ef , the electrostatic force on the particle
is counter-balanced by the fluid drag on the particle. For globular
proteins, the drag force can be approximated from Stokes law.
Therefore, the balance is
qEf = 31!Jld pu£ (10.36)
so that
qE f
(10.37)
UE = 2npdp
where the steady velocity of particle u£ is known as electrophoretic

mobility, which is proportional to the electrical field strength and
electric charge, but is inversely proportional to the viscosity of the

liquid medium and the particle size.
However, the increase of the electric field strength increases not
only the electrophoretic mobility, but also the mixing caused by the
gas release from the electrodes and the free convection caused by
electrical heating, which are two major problems in designing
electrophoresis setup. To minimize these problems, electrodes are
usually located in separate compartments and the electrophoresis is
performed in a gel, known as gel electrophoresis. Among various gels
available, polyacrylamide gels are most commonly used. They are
thermostablel, transparent, durable, relatively inert chemically,
nomonic, and easily prepared. The disadvantage of gel
chromatography is that setting up the gel is tedious and time
consuming and large-scale operation is not possible:
10.5.4 Membrane Separation

As we discussed earlier for the solid-liquid separation technique,
filtration separates particles by forcing the fluid through a filtering
medium on which solids are deposited. The conventional filtration
involves the separation of large particles (d p > 10 )lm) by using
canvas, synth~tic fabrics, or glass fiber as filter medium.
We can use the same filtration principle for the separation of small
particles down to small size of the molecular level by using
polymeric membranes. Depending upon the size range of the
particles separated, membrane separation processes can be classified
into three categories: microfiltration, ultrafiltration, and reverse
osmosis, the major differences of which are sumnla.rized in
Table 10.2.
Table 10.2 Pressure Driven Membrane Separation Processes
(Lonsdale, 1982)
ProcessSize Cutoff Molecular Pressure Material

Wt. Cutoff Drop (psi) Retained
Micro- 0.02 - 10,um 10 Suspended material
filtration including
microorganisms
Ultra-
filtration 10 - 200 A 3 00 - 300k 10- 100 Biologicals, colloids,
macromoleculres
Reverse
Osmosis 1 - 10A < 300 100 - 800 All suspended and
dissolved material
Microfiltration refers to the separation of suspended material such

as bacteria by using a membrane with pore sizes of 0.02 to 10 /lm. If
the pore size is further decreased so that the separation can be
achieved at the molecular level, it is called ultrafiltration. The typical
molecular weight cutoff for ultrafiltration is in the range of 300 to
300,000, which is in the size range of about 10 to 200A. Reverse osmosis
is a process to separate virtually all suspended and dissolved
material from their solution by applying high pressure (100 to 800
psi) to reverse the osmotic flow of water across a semipermeable
membrane. The molecular weight cutoff is' less than 300 (1 to lOA).
Protein products are in the range of the molecular cutoff for
ultrafiltration.
Membrane filters are usually made by casting a polymer solution
on a surface and then gelling the liquid film slowly by exposing it to
humid air. The size of the pores in the membranes can be varied by
altering the composition of the casting solution or the gelation
condition. Another common technique is to irradiate a thin polymeric
film in a field of a-particles and then chemically etch the film to
produce well-defined pores.
The solvent flux across the membrane J is proportional to the
applied force, which is equal to the applied pressure reduced by the
osmotic pressure ~n as (Belter et al., p. 255, 1988),
J = Lp(~p - (cr~n) (10.38)
where L p is the membrane permeability and cr is the reflection
coefficient (a = 1 if the membrane rejects all solute and cr = 0 if the
solute passes through membrane freely). The solvent flux is the
volume of solvent per unit membrane area per time, which is the
same as the solvent velocity. If the solution is dilute the osmotic
pressure is
~n = R' TC* (10.39)
where C* is the solute concentration at the membrane surface.
One of the major problems with the membrane separation
technique is concentration polarization, which is the accumulation of
solute molecules or particles on the membrane surface.
Concentration polarization reduces the flow through the membrane
and may cause membrane fouling. It can be minimized by using
cross-flow type filter and maintaining a high liquid flow parallel to
the membrane surface by recirculating the liquid through thin
channels. There are four basic membrane modules commonly
employed, plate-and-frame, shell-and-tube, spiral-wound, and
hollow-fiber membrane. All of these modules are cross-flow type.
Concentration polarization can be further reduced by prefiltering the
solution, by reducing the flow rate per unit membrane surface area,
or by backwashmg periodically.
The flux of solute from the bulk of the solution to the membrane
surface is equal to the solute concentration times solvent flux C/. At
steady state, it will be countered by the molecular .diffusion of th.e
solute away from the membrane surface as
CJ = _D de (10.40)
dz
which can be solved with the boundary conditions,
at z = 0 (10.41)
at z =8
to yield
D C*
J= - In- (10.42)
8 Cb
where 8 is the thickness of the laminar sublayer near the
membrane surface within which concentration polarization is
assumed to be confined. The concentration ratio, C* / Cb is known as
the polarization modulus.
Example 10.4
Figure 10.13 shows a typical batch ultrafiltration setup. As the
solution is pumped through the filter unit, the permeate is colJected
and the retentate is recycled. The volume of the solution reduces with
time and the solute concentration increases. Develop a correlation for
the time required to reduce the solution volume from Va to V Assume
that the concentration polarization is negligible. Also assume that the
membrane totally rejects the solute.
__..__._-L~·-·-·-··-··--Re-t-en-ta-t-e ------
Permeate
Fig. 10.13 Typical batch ultrafiltration setup.
Solution:
The decrease of the solution volume is equal to the membrane area A
times the solvent flux across the membrane 1.
dV = -AI (10.43)
dt
Substitution of Eq. (10.38) into Eq. (10.43) yields
dV
at = ALp (!lp - aMI) (10.44)
Since the solute 'is completely rejected by the membrane, (J = 1.

Therefore, substituting Eq. (10.39) into Eq. (10.4~' and rearranging
gives
dV = -AL ~
dt p p
(1- R'TC*)
!lp
(10.45)
If the concentration polarization is negligible, the solute

concentration on the membrane surface C* is equal to the bulk
concentration Cb Since the membrane rejects all solute, the total
solute (1Jl / = CbV) is constant. Therefore,
dV =-AL ~
dt p p
(1- R'TC*)
!lp
(10.46)
The integration of the Eq. (10.46) with the initial condition, V ~ Va

at t = 0, and-the rearrangement for t yields
t=. 1 [(Vo_V)+(R'Tm')ln(Vo-RITmlli1P)] (10.47)

AL pi1p i1p V - R'Tm'li1p
10.6 NOMENCLATURE
A area of filtering surface, m 2

a acceleration, m/ S2
C solute concentration, kg/m 3
Cb solute concentration at the bulk solution, kg/m 3
C* solute concentration at the membrane surface, kg/m 3
o molecular diffusivity, m 2 / s
dp particle diameter, m
E mass (or mass flow rate) of extract phase, kg (or kg/s)
EF electric field strength, 'NC- 1
F mass (or mass flow rate) of feed stream, kg (or kg/s)
J solvent flux, the volume of solvent per unit
membrane area per time, m/ s
k D'Arcy's filter cake permeability, m 2

~p pressure drop across the filter and the cake or
pressure drop across the membrane, Pa
K distribution coefficient, dimensionless
kL constant for the Langmuir isotherm, dimensionless
Lp permeability, m/Ns
Ls mass (or mass flow rate) of diluent or solvent, kg (or
kg/s)
m, n constants for the Freundlich equation,
m' total amount of solute, kg
q electric charge of a particle, C
Rc resistance coefficient for filter cake, 1 1m
RM resistance coefficient for filter medium, 11m
R mass (or mass flow rate) of raffinate stream, kg (or
kg/s)
R' gas constant, gas constant, JIkmol K
r radial distance from the center of a centrifuge to a
particle, m
5 mass (or mass flow rate) of solvent, kg (or kg/s)
S s m a s s (or mass flow rate) of adsorbent, kg (or kg/s)
s compressibility coefficient, dimensionless
T temperature
t time, s
uE electrophoretic mobility, m/s
V solution volume,m3
VI volume of filtrate collected, m 3
vt terminal velocity, ml s
X mass fraction of solute in solution in solute-free
basis, dimensionless
x mass fraction of solute in solution (in raffinate
phase for extraction), dimensionless
Y mass of solute per mass of adsorbent, dimensionless
y mass fraction of solute in extract phase (for
extraction) or in adsorbent (for adsorption),
dimensionless
a specific cake resistance, m/kg
aD specific cake resistance for incompressible cake, m/kg
290 Fundamentals ofBiochemical Engineering
f3 selectivity of extraction, dimensionless

r recoverabilityof solute, dimensionless
8 thickness of the laminar sublayer, m
/l fluid viscosity, kg/m s
~II osmotic pressure, N/m 2
P density, kg/m3
Pc mass of cake solids per unit volume of filtrate, kg/m3
Ps density of solid particle, kg/m3
a reflection coefficient, dimensionless
m angular velocity, s-1
10.7 PROBLEMS
10.1 An antibiotic, cycloheximide, is to be extracted from the
clarified fermentation beer by using methylene chloride as
solvent. The distribution coefficient K is 23. The initial
con~entration of cycloheximide in the feed is 150 mg/L. The
recovered solvent containin~ 5 mg/L of cycloheximide is being
used with the flow rate 1 m /hr. The required recovery of the
antibiotic is 98 percent.
a. If you use fOUf countercurrent stages, how much feed can
you process per hour (F)?
b. If you use fOUf crosscurrent stages with equal solvent flow
rate (0.25 m 3 /hr), how much feed can you process per hour
(F)?
10.2 Aspartic acid needs to be isolated from a fermentation broth.
The initial concentration of aspartic acid in the solution is
1.0 x 10-3 g/mL. We need to recover 98 percent of the aspartic
acid. The amount of aspartic solution is 1 m 3• The isotherm for
the adsorption of aspartic acid into an anion exchanger (Duolite
A162) is given as follows (Cowan et al., 1987)
x X 103 (g solute/g water) 0.02 0.03 0.05 0.07 0.17 0.55 1.1
y* (g solute/g adsorbent) 0.05 0.07 0.12 0.11 0.19 0.21 0.20 0.23 0.24
a. Which isotherm, Freundlich or Langmuir, fits the data
better? Determine the experimental constants for the
isotherm.
b. If we use single-stage contact filtration, how much Duolite
A162 is needed?
c. If we use two-stage crosscurrent filtration with an equal
amount of adsorbent for each stage, what is the total
amount of absorbent?
d. If we use two-stage countercurrent filtration, how much

total absorbent is needed?
10.3 Monoclonal antibodies (MoAb) will be separated from the
supernatant of a mammalian culture by using DEAE-Sephacel
as the adsorbent. The supernatant contains 100 ,ug/mL of
MoAb. Adsorption follows the Langmuir isotherm with Y'max
=116 mg of MoAb per mL of adsorbent (settled volume) and K'L
= 0.5 X 10-3 mg of MoAb per mL of supernatant (Desai et aI,
1987). The amount of supernatant is 1 L. If you use 2 mL of
DEAE-Sephacel for a single-stage, what percent of the MoAb
will be recovered?
10.8 REFERENCES
Ambler, C. M., "Centrifugation," in Handbook of Separation Techniques for
Chemical Engineers, ed. P. A. Schweitzer.
New York, NY: McGraw-Hill Book Co., 1979, pp. 4.55-4.84.
Belter, P. A., E. L. Cussler, and W. Hu, Bioseparations: Downstream Processing
for Biotechnology. New York, NY: John Wiley & Sons, 1988.
Cain, C. W., "Filtration Theory," in Handbook of Separation Techniques for
Chemical Engineers, ed. P. A. Schweitzer. New York, NY: McGraw-Hill
Book Co., 1979, pp. 4.3--4.8.
Cliffe, K., "Downstream Processing," in Biotechnology for Engineers, ed. A.
Scragg. Chichester, England: Ellis Horwood Ltd., 1988.
pp.302-321.
Cowan, G. H., I. S. Gosling, and vv. P. Sweetenham, "Modelling for Scale-Up
and Optimisation of Packed-bed Columns in Adsorption and
Chromatography," in Separations for Biotechnology., eds. M. S. Verrall and
M. J. Hudson. Chichester, England: Ellis Horwood Ltd., 1987. pp. 152-
175.
Desai, M. A., J. G. Huddleston, A. Lyddiatt, J. Rudge, and A. B. Stevens,
"Biochemical and Physical Chax~cterisation of a Composite Solid Phase
Developed for Large Scale Biochemical Adsorption," in Separations for
Biotechnology, eds. M. S. Verrall and M. J. Hudson. Chichester, England:
Ellis Horwood Ltd., 1987, pp. 200-209.
Dorr-Oliver Inc., Filtration Leaf Test Procedures. Stamford, CT: Dorr-Oliver
Inc., 1972.
Keshavarz, E., M. Hoare, and P. Dunnill, "Biochemical Engineering Aspects
of Cell Disruption," in Separations for Biotechnology, eds. M. S. Verrall and
M. J. Hudson. Chichester, England: Ellis Horwood Ltd., 1987, pp. 62-79.
Lonsdale, H. K., liThe Growth of Membrane Technology," J. Membrane Sci. 10
(1982):81-181.
Perry, R. H. and C. H. Chilton, Chemical Engineers' Handbook, (5th ed.), pp. 19-
57-19-85. New York, NY: McGraw-Hill Book Co., 1973.
Schmidt-Kastner, G. and C. F. Go/ker, "Downstream Processing in

Biotechnology," in Basic Biotechnology, eds. J. Bu'lock and B. Kristiansen.
London, England: Academic Press, 1987, pp.173-196.
Treybal, R. E., Mass-Transfer Operations (3rd ed.), p. 128, pp. 479-488. New
York, NY: McGraw-Hill Book Co., 1980.
SUGGESTED READING
Belter, P. A., E. L. Cussler, and W. Hu, Biosepamtions: Downstream Processing
for Biotechnology. New York, NY: John Wiley & Sons, 1988.
Verrall, M. S. and M. J. Hudson, Eds., Separations for Biotechnology. Chichester,
England: Ellis Horwood Ltd., 1987.

BAB 9 Rajiv Dutta

Uploaded by

Copyright:

Available Formats

BAB 9 Rajiv Dutta

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

BAB 9 Rajiv Dutta

Uploaded by

Copyright:

Available Formats

Sterilization 207

5 X 10 12 m-3, which needs to be reduced to such an extent that only one

v = ln~ = In (5 x 1012m-3)(2m3/hr)(24hr/day)(60days) = 37.2

Cooling Section: For the cooling section, a quench cooler with

8.6 AIR STERILIZATION

Therefore, glass fibers are favorable as filter medium because they

Impaction: When an air stream containing particles flows around

where Cf is known as Cunningham correction factor. The value of Cf

Cf = 1 + ~~ [ 1.257 + 0.400exp( -1.l0~~)] (8.30)

where A is the mean free path of gas molecules based on the

A= (O-.-£9P )I::~ (8.31)

The efficiency lJimp is defined as the fraction of particles

Another correlation propsed by Friedlander (1967) is

Interception: The inertial impaction model assumed particles had

which was developed by using Langmuir's viscous flow equation

Fig. 8.4 Flow pattern around cylindrical fiber, showing the

Diffusion: Particles smaller than about 1 micron in diameter

Friedlander (1967) suggested the following correlation

Combined Mechanisms: The total collection efficiency of a

which allows only the particles not collected by one mechanism to be

Fig. 8.5 Shell balances around a differential element of a filter.

_ den = 4a1Jc dh (4.47)

11 a = 11f (1 + 4.5a) (8.50)

v= ~ = _1_0_ = 10.3 cm/s

To estimaterJimp by using Eq. (8.32) or Eq. (8.33), we need to

N = Dcvp = (1.5 x 10-3 )(10.3)(1.20 x 10-3 ) = 0.103

A= [ 1.80 X 10-4] n(29) = 6.50 x 10-6 cm

Cf = 1 + ~~ [1.257 + OAooexr(-1.10 ~~)] = 1.16

1.16(1)(1 x 10-4)2(10.3) = 0.0247

Therefore, from Eq. (8.33), the single fiber collection efficiency

11imp = 0.075N§t2 = 0.075(0.0247)1.2 = 8.82 X 10--4

The diffusivity dup to the Brownian n1otion can be estimated

_ CfkT _ 1.16(1.38 x 10-6 )293

l1dif = 1.3Np~2/3) + 0.71(2 = 4.00 x 10- 3

In this example, diffusion is predominant over impaction and

estimation of 1Jimp and 1Jdlf' respectively, 1Jinp = 6.82 X 10-3 and

predicted from Eq. (8.33) and Eq. (8.37). Strauss (1975)

1J = 1- ex [- 4(10) (0.03 )0.0059] = 0.79

en cell number density, number of cells/m3

Urn maximum velocity, m/ s

8.2 A continuous sterilizer with a steam injector and a flash cooler

Fig. 9.1 Typical dimensions of flat-blade disk turbine

Mass-Transfer Path: The path of gaseous substrate from a gas

Fig. 9.2 Schematic diagram of the path of a gaseous

Steps 3 and 5, the diffusion through the relatively unmixed liquid

9.2 BASIC MASS-TRANSFER CONCEPTS

where C is total concentration of components A and Band N B is the

Diffusivity: The kinetic theory of liquids is much less advanced

D° at 40°C = 2.5 X 10-9 (8.904 X 10-4)( 313) = 3.58 x 10-9 m 2 / s

D~B at 40°C = 2.5 X 10-9 (8.904 x 1O~)1.1 = 3.52 X 10-9 m 2 /s

9.2.2 Mass-Transfer Coefficient

Nc = q~ = kC<C c - CC1 ) (9.6)

Similarly, the liquid-side phase mass flux Nt is

where kL is the individual mass-transfer coefficient for the liquid

Fig, 9.3 Concentration profile near a gas-liquid interface and an

Fig. 9.4 The equilibrium curve explaining the meaning of C~ and C~

By substituting Eq. (9.13) to Eq. (9.12), we obtain

For sparingly soluble gases, the slope of the equilibrium curve is

9.2.3 Mechanism of Mass Transfer

visualize the mechanism of mass transfer at the interface and also to

Cf = 1 + [ 1.257 + 0.400exp( -1.l0)] (8.30)

Cf = 1 + [1.257 + OAooexr(-1.10 )] = 1.16

IOg10 () = (1.95XI0-5)N(N~I f3 (9.43)