-

From Biotechnology: The Science and the Business, Moses, V. and Cape, R. E., editors,
Harwood Academic Publishers (New York), 1991.
Chapter 13
FermentationTechnology,Bioprocessing,
Scale-up and Manufacture ,
Yusuf Chisri and Murray Moo-Young
Thecommercialization of biotechnology-based processes for
the improyement ofhuman life would be impossible without
the supporting engineering disciplines. Chemical or process
engineering taking into accOunt the specificities ofbiological
systems has developed ;nto biochemical engineering which is
already a rapidly growing branch ofknowledge. Biochemical
leering has a crucial role in economically transforming
tfi1: labornory disco\'eries in biotechnology into large scale
manufacturing.Processdevelopmenttime - thetimebetween
initial process conception and full scale manufacture of
p-0duct for sale - may be considerably shortened by the
earliest invoh-ement of engineers in biotechnology research.
The type and the quality of research data, for example, can
lead tosignificantsavings in resources in subsequentstages of
biorr"c.. development.
This chapter is an overview of biochemical engineering
baseofbiotechnology.Theprocessingconsiderationscommon
to many biological systems are examined. The bioreactors
used in the production of biochemicals and biocatalysts
(enzymes,microorganisms,cellcultures)and thefundamentals
ofdesignofthesereactorsandsupportingsrstemsaretreated.
1"' 'iuct harvesting, purification and other downstream pro-
b.lOg operations are discussed with emphasis on newer
p'elcpments. The associated process control technologies
a:" considered. Finally, an overall process dimension is pro-
vJed by a complete bioprocess.
BIOPROCESS: GENERAL ASPECTS
Any large-scale operation involving the transformation of
some raw material (biological or non-biological) into some
product by means of microorganisms, animal or plant cell
cultures, or by materials (e.g. enzymes, organelles) deri\Oed
from them, may be termed a "bioprocess". The"product" of
suchprocesses may be saleable (e.g. insulin, penicillin, SCP,
enzymes) or they may have little commercial value (waste
tre:ltment). A bioprocess is typically made up of three steps
shown i.n Figure 1. The raw material or feedstock (see
Chapter 14) must be converted to 3 form which is suitable
for This is done in a pretreatment step which
may Involve one or more of the operations shown in
Figure1.Frequently,thewellestablishedchemic31engineering
operations suffice for the pretreatment stage and these will
not be discussed further.
The pretreatment step is followed by one or more bio-
reaction stages where the
PROCESS STAGES
-
Raw Material
I
PRETREATMENT
l
I
I
I
l
BIOREACTION
I
I
I
DOWNSTREAM
PROCESSING
I
I
8
It
desired biotransformation takes
I
OPERATIONS
Sorting
Sieving
Comminution
Hydrolysis
Medium formulation
$Ierilization
Biomass production
Metabolite biosynthesis
Immobilized enzyme and
cell catalysis
Biotransformations
Filiration
Centrifugation
Sedimentation
Flocculation
Cell Disruption
Extraction
UltrafiItration
Preci pitotion
Crystalization
Chromatography
Evaporation
Drying
Packaging
Figure 1. Bioprocess stages and the commonly used operations in
them.
I
Bingham
Casson
C\J
E
TO
Oilotanl
Z
n > I
Newtonian
n = I
l-
Pseudo-
plastic
n <: I ./ I
;' I
..--./ I
slope =fLo p I
I . I
Y
o )'1 I t
0
0
y (5-
1
)
Figure 2. Shear strcss \"s. shcar ratc plots for varIOUS flow
behaviours.
may be used. For mrcdial broths the power law model is
oftCl1 s:llisfactoty. The parameters K and narc dependcnt on
solids concentration; Il generally declines \\ith the concen-
! tration of solids, i.e., the suspension bccomes increasingly
: __ shear thinning, while K increases quite strongly with the
. solids content of the mycelial slurry. At present, however, no
satisfactory quantitati\'e relationships exist to describe the
influence of mycelial solids on K and n of these fluids. For
example, for broths of Aspergil!lls niger the dependence of
K on solids concentration (X) according to two different
correlations is shown in Figure 3. The agreement between
these equations is not at all satishctory. As a result, eYen
though a flow model may fit the experimental d:lta, the
parameters obtained ma;' ha\'c little yalue for bioreactor design,
Furthermore, all the foregoing flow models and the parameters
obtained on their basis apply strictly to laminar flow. Could
such K and n have any real value for process design purposes
under conditions of turbulent flow? This question is particu-
larly important in man;' biological slurries in which the
parameters K and n are not really constant but depend on the
shear rate range used in their determination. Questions such
as this remain to be ans\\'ered. Alternative approaches which
relate the design propertics such as oxygen transfer, mixing
and heat transfer directl;' to biomass content of a slurry are
useful as discussed later.
Permenter technology
10..-------------,---1
AspergillUS niger
K = 0.03 X2.48
based on Reuss
ela!. (1982)
c
(J)
0
1.0
0...

K= 4.3x IO-4X
3
.
3
based on
Allen (1987)
0.1
I 2
Figure 3. Dcpendence of consistency index (K) on solids concen-
tration (X) in broths of Aspergillus niger.
PRETREATMENT OPERATIONS
Media Sterilization
The successful operation of most productive fermentations
depends on the maintenance in culture of a single microbial
species with \\'hich the fermenter was initially inoculated. To
prevent contamination by other organisms the gaseous and
liquid feeds to the fermenter have to be sterilized, Either
phrsical elimination, for example by filtration, or destruction
of microorganisms mar be used to achieve sterile feed streams.
\,\'hile ionizing radiation and chemical sterilents rna:' be
used, sterilization by filtration and heat treatment are b;' far
the most common techniques. Liquids such as water and salt
solutions can be inexpensively sterilized by passing them
through absolute filters the pore size of which are smaller
than the dimensions of any contaminating particles. This
method is practicable only when the liquid is free of other
suspended and its initial microbial contamination is
4 6 8 10
X (gL-
1
)
Biotechnology/The the Business
pbcl.? The tr:lI1sform:uion may involve the conversion of a
substrate to biomass or biomass and some biochemical or
enzyme. Alternatively, the conversion may usc dead whole
cells (immobilized or in suspension) or enzymes as the bio-
catalytic agl.?ncy. Bioreactorsform thecoreofthe bioreaction
stq.1. The material produced in the bioreactors must usually
beprocessed further in thedownstreamsectionoftheprocess
to convert it to a useful form. Downstream processing con-
sists of predominantly physical operations which arc aimed
atconcentration and purificationoftheproduct.Thepurified
product may have to be in different physical forms (liquid,
" slurry, powder, crystalline) for different applications.
The properties of biological materials impose significant
unique constraints on the bioreaction and dO""nstream pro-
cessingstages. Thesestages arc treated more thoroughly later
on in this chapter after a consideration ofthe characteristics
ofthe biomaterials themselves.

PROPERTIES OFBIOLOGICAL
MATERIALS
Thedesign ofa bioprocess and theengineeringoftheprocess
equipment requires a careful consideration of the physical
and chemical properties of the material being handled and a
delimitationofthemaximumprocessingstresses(temperature,
pH. shear forces, contamination, pressure) that the material
safelytobate.Typicallyabioprocessmustoperatewithin
the physiological ranges of pH and temperature (pH - 7,
tempaature ::; 37C), the specific conditions being very
process-dependent. The pressures which would normally be
encountered (::; 2 MPa)in bioreactorsdo notseem todamage
microorganisms or enzymes but carbon dioxide associ:lted
toxicity effects due to increJsed solubility of CO
2
at higher
pressures may become important.
The sensitivity to sheJring forces mJy \"::Ir}" widely. In
general, bacteria and yeasts which grow JS small individual
cells arc quite shear tolerant. GeneticJlly engineered species
and wall-less mutJnts arc frequently susceptible to shear
damage due to their weaker cell wall/membrane structures.
Filamentous bJcteriJ and mycelial fungi which have larger
particle dimensions do show signs of mechanical damage in
high shear fields. Similarly, plant and mammJliJn cells are
more sensitive to shear. Enzymes, with the exception of
multienzyme complexes and membrane associated enzymes,
Jre. not damaged by shear in the Jbsence of gas-liquid

nearlyall processingoperationsmusthandleliquids
and slurries a more in-depth tre:1tment of biofluids follo",s.
Biological Fluids
Biofluids and slurries hll into two categories: (i) Ne",toniJn
fluids such as water, honey Jnd most bacterial and rcast
fermentJtion broths; and (ii) non-Newtonian media such as
polysaccharidefermentations and thebrothsofStreptomyces,
Aspergilli and Penicillia.
Newtonian fluids
AtconstanttemperatureJndpressure, Newtonian fluids have
J constJnt viscosity irrespective of the shear rate. for these
fluids shear stress (t) and shear rate arc linearly related. In
laminar flow
t = Y (1)
Plots of T vs. yare straight lines of slope equal to the
viscosity,
Non-Newtonian fluids
Viscosity of these fluids is dependent on the rJte of sheJr.
This dependence is commonly described by the power law "-
model:
t = KyO
":-.,
where K and n Jrethe consistencyindex and the flow beHav-
iourindex, respectively, ofthe fluid. By anJlogy with eq. (1)
In Jpparent viscosity CJn be defined for the
Newtoni:lI1 fluids:
(3)
for n = 1, eq. (3) reduces to constantviscosity form and the
fl uid is Newtonian. forn > 1, thefluid becomes increasingly
viscouswithshearanditis termeddilatantorshearthickening..
When n < 1 the fluid is shear thinning or pse/fdoplastic. .
Many biological media display pseudoplastic behaviour.
The shear stress vs. shear rate plots for the various flo"'-
behaviours are shown in Figure2. Theslopeofa line joining
anypointon theseplotstotheoriginis theJppJrentviscosity.
Clearly,theJpparentviscosityincreases(dilatJnt)ordcc; '5
(pseudoplastic) with shear rate for non-Newtonian fluiq J
Certain fluids do not flow until the applied shear s' A's
exceeds a minimum VJlue (To) known as the "yield
This t)Tpe of behaviour may be described bv the Bing/;'s .
, . ,vn
plastic flow model or by the Casson equJtion:
T = To + K )T (Bingham plastic) (of} C
y'T = y:r;, + K
c
yy (Casson model) (5)
where K
c
is a constJnt kno",-n as CJsson viscosity.
Estimation ofviscosity
Forsuspensions ofyeJsts Jnd bacteria gro""ing JS individuJI
cells suspended in a wJter-like medium, equationsofEinstein
= (1 + 2.5 ES)
and of Vand
(7)
168

sterilization is calcubtcd by gnphical
integration of k
J
VS.
time profiles
-In - J k
d
dt + kd thold +
(13)
No TOlal
(
N)
To
In continuous sterilization very rapid heating and cooling
is obtained and the contribu tions of heating and cooling
periods to sterilization is generally disregarded.
Heating Hold I Cooling'
w
n::
T
s
---------
I I
::>
I I
r- I I
<t
n::
I
I
I
I
w
Q.
I
I
I
I

I
I
uJ
r-
To
I ...
t heat
I
I
I
.,.....
t hold
I
I t cool

i
-'1
to
TIME
Figure 5. Time-temperature profile.
Continuous sterilization
Ad\'antages associated ""ith HTST sterilization, rapid heating
and cooling and precise control of holding time makes con-
tinuous sterilization the preferred method whenever it can be
employed. The kinetics of sterilization are identical to the
treatment gi\'en in the previous section. The raw medium is
'eated to the sterilization temperature either by continuous
:- injection or by a high efficiency heat exchanger (plate
or spiral exchangers). This is followed by a holJing coil
where the sterilization temperature is maintained for a time
equalling the residence time of the coil. Either flash cooling
or indirect heat exchange returns the feed to the fermentation
temperature. The steam injection and indirect heating schemes
are illustrated in Figure 6.
For a continuous fermentation feed flow of Q m\-l,
N may be easily calculated for a given level of contamination,
e.g., one contamination during operation time top, as
N = lIQ top' (14)
Hence, the holding time necessary for a given NINo, and the
length of holding coil rna)' be determined.
The design of the holding coil requires careful attention,
The velocity of flow is not uniform across the cross section
of a pipe and the flow is always faster at the axis of a straight
pipe. As a result, the residence times of different clements of
fluid in the pipe can be diffaent. A most conservative estimate
of the residence time is obtained from the equation
Fermenter technology
L
(15)
t
r
=
Un1 .\X
""here L is the length of the holding coil and U m .1X the
centreline velocity of Ho"" in the pipe. The maximum velocity
U
m
.
1X
depends on the Reynolds number of flow
PL U d
Re = :......::--- (16)

""here U is the average flo",' velocity through the pipe. \\' ell-
de\'e!oped turbulent flo"" (Re 2300) is desired in th<' pipe
to minimize the difference between U and UOl.l
x
' Values of
UfUm." as a function of Reynolds number are gi\'en in
Fioure 7 for straioht, circular pipes. In helical coils laminar
.
flo"" persists to significantly higher values of Reynolds
number than in straight pipes.
Air Sterilization
Aerobic fermentalions require a continuous supply of large
quantities of air or oxygen. The gas entering the fermentc'r
must be free of contaminants such as bacteria, spores, bac-
teriophages and other microorganisms. Similarly, the exhaust
gas from a fermenter must be treated to remo\'e micro-
organisms and spray particles which can be potentially harmful
to plant personnel and the environment.
Sterilization of fermentation inlet and exhaust gas is achiewd
predominantly by filtration on which depends the fail ure or
success of a fermentation operation. The smallest particles
which need to be remO\'ed from the air are bacteria and
viruses. The smallest bacteria are - 0.1 f.lm in diameter;
viruses are typically less than 0.3 and may be as small as
0.0-1 Either depth filtration or absolute filtration may be
used to free the air of umvanted particulates. Depth filtration
depends on passing the gas through a bed of packing such as
compressed glass wool or other fibrous material. The spaces
between fibres are larger than the dimensions of the particles
to be remowd and particles penetrate the filter bed to various
depths. The total filter depth is such that the required reman]
is achie\'ed. Absolute filters, on the other hand, have openings
which are smaller in size than the dimensions of the smallest
particle to be retained. Particle remonl is by a sieving action.
Porous polymer membranes, ceramic and metal membranes
are used as absolute filters. Polymer membranes in the very
narrow pore size distribution can be produced by subjecting
non-porous polymer films to bombardment by high
nuclear panicles. Howe\'er, absolute membrane filters are
easily fouled and produce relatively high pressure drops.
Prefilters are used prior to the air filters to remove gross
contamination such a dirt, oil and water droplets and foam to
extend the operational life of the filter.
Depth filtration
Sewral different mechanisms of particle retention operate in
a compacted bed of fibrous material: direct interception of
1
7
1
Science and the Business
ow. For some liquids such as blood serum filter sterilization
nay be the only viable technique if denaturation of its highly
abile constituents is to be prevented. On the other hand,
:oncentrated solutions of sugars and highly contaminated
nedia (e.g. molasses, cornsteep liquor) are commonly heat
terilized. Thermal denaturation of contaminating organisms,
\hen properly carried out, is among the most effective
nethods of sterilization.
:-feat Sterilization
fhermal denaturation of one or more enzymes in the con-
aminating microorganisms is used to render them non-viable.
fhe feed is brought to a sufficiently high temperature and
leld there for a certain time to ensure the destruction of the
nost resistant contaminant. The process may be conducted
)atchwise or continuously.
rJatch sterilization
fhe rate of destruction of microorganisms follows the first-
)rder kinetics:
(8)
iVhere kd. is the specific death rate and N the concentration of
:he contaminating organisms at any time. The time (t) required
:0 reduce the concentration of contaminants from an initial
value No to some value N is obtained from the integrated
:orm of eq. (8):
(9)
[ypically, No is 10
5
-10
9
/mL. The final concentration N
iepends on the degree of acceptable fermentation failure
;ince at some point the cost of further reducing the risk of
:ontamination would exceed the expense of a lost fermen-
:ation. For a 250 m
3
fermentation with an acceptable failure
'ate of one in fifty fermentations the final level of contami-
lation would be one microorganism in fifty fermenter volumes
N = 1/(50 X 250) = 1/12500 m
3
== 8 X 10-
11
per mL
[he thermal death rate constant k
d
is a function of temperature
k
d
= A e-t>E/RT (10)
t..E is the activation energy for the destruction of a
Jarticular microorganism, A is its Arrhenius parameter and T
s the absolute temperature. Bacterial spores such as those of
fJacillus stearothermophilus and Clostridium botulinum are
;ome of the most heat resistant, and sterilization processes
lre designed to be effective against them. The activation
:nergy for destruction of these spores is 2.5 - 2. 9 X lOs ]
,mol-
I
and A is - 1.6 X 10
36
S-I. Because k
d
increases with
]0
temperature, the higher the temperature the shoner the treat-
ment time needed to achieve a given bd (N/N
o
) of destruc-
tion as illustrated in Figure 4. However, the processes ,,hich
lead to microbial inactivation also cause destruction of heat
labile essential nutrients in the feed. Denaturation of nutrients
(k
n
) is also temperature dependent
k
n
= An e-t>En/RT (11)
o
Z
"-
Z
100C
IISOC
A = 1.6 X 10
36
5
b.E =2.9 X 10
8
J kmol-
I
0.1
0.001 =- =-'::" ----'I -'---_-..J
a 0.5 1.0 1.5
HOURS
Figure 4. Time needed to obtain a given level of destruction (N/
No) at various sterilization temperatures.
From equations (10) and (11) it follo,,'s that
k
d
= e(t>En - 6E)/RT
(12)
k
n
An
The activation energy for the thermal deactivation of most
nutrients (t..E
n
) is substantially lower than t..E for microbial
deactivation. For example t..E
n
for thiamine hydrochloride
(vitamin B
6
) is only 9.2 X 10
7
] kmol-
I
Hence, in order to
maximize microbial destruction relative to nutrient loss (i.e.
achieve higher kd/k
n
), sterilization at high temperature for a
shorter time is indicated. This is the basis for HTST or high-
temperature-short-time sterilization. Because of the short
exposure times (of the order of seconds), HTST is best
implemented in a continuous flow mode.
Either direct heat (steam injection) or indirect heat (coils,
jacket) may be used to heat the fluid to the sterilization
temperature where it is held for some time (holding time)
followed by cooling. The time-temperature profile of such a
process is shown in Figure 5.
1':-
In batch sterilization heating and cooling times are relatively'
long and some sterilization occurs during these periods.
During heating and cooling k
d
varies with time and the total
Fermenter technology
ofall viable organisms. In practice, each depth filter thickness
may haw a DOr penetration of 0.001% or less and since
se\'eral 1:1\'ers are often used in series, the actual penetration
is much I;,,er. Hence, depth filters are an effective meansfor
air sterilization. In some applications a depth filtration step
mav be followed by an absolute filter to gain additional
on airpurity. A typical packed filter arrangement
in fermentation application is shown in Figure 8.
BIOREACTION
Fermentation Process, Growth and Production
A sterilized batch fermenter containinga properly developed
fermentation medium (C and N" sources, micronutrients),
sufficientaeration,suppliesofpHcontrolchemicals,antifoams
and associated systems must be inoculated with the desired
microbial species to initiate the fermentation. The inoculum
consistsofamicrobialsuspensionin rapid exponentialgrowth
added at a concentration of 5-10% by volume to the
fermenter. The slower growing the organism, the larger the
volume ofinoculum used to avoid having long fermentation
times (costs) in the production vessel. Because industrial
fermenters tend to be quite large, inoculum preparation
from agar slants often requires several fermentation steps:
shake flasks, seed fermenter and secondary (or tertiary) seed
fermentation. In some instances quantities of spores for
inoculation produced in a seed stage arc blown directly into
the larger SC3le vessel with the ingoing air.
Following inoculation the growth of the microorganism
follows the typical pattern illustrated in Figure9. Inoculation
of cells into the fermenter often results in a period where
there is no increase in cell number: this period is known as
the "lag phase". The length ofthe bgphase is related to the
growth histOry of the inoculum, the composition of the
medium and the size of the inoculum. The composition of
the media used in seed and production vessel should be
identical to avoid or eliminate excessive lag. Additionally, as
pointed out earlier, rapidly growing cells (late exponential
growthphase)should be used for inoculationand the volume
ofthe inoculum should be such that possible osmotic st:.9-ck
effeels on dilution in the brger vessel are minimaV The
existence ofa lag phase shouId not be taken to mean lack of
metabolic activity in the cells; in fact, the lag phase is pre-
paratory to rapid exponential growth and is essentially an
ad3ptationperiod.Ne\'enheless, thelagphaseis unproductive
with respect to fermentation time in the production vessel
and fennentation optimization aims at reducingor bypassing
the lag. The bg phase is followed by exponential growth
duringwhichthecell number(ormass)increasesexponentially
with time.Thecell mass and cell number grO\\'th r3tes arenot
necessarily equal.
I Log phase
l/)
l/)
1:Exponential : Stationary Death
I growth
I _..:.p_h_a_s_e_--..;.._ phase
o
E
Q.l
U
Q.l
.0
o
>
I
I
I
I
I
J
I
I
I
o
FERMENTATION TIME
Figure-9. Typical progression ofmicrobial growth.
Increase in cell mass (X) during exponential growth often
follows the equation
dX
(18)
dt
where is the specific growth rate and k
d
the specific death
rate. Duringexponential growth k
d
and eq. (18) reduces
to
dX
=
(19)
For a cell mass concentration X
o
at the beginning of expo-
nential growth (X
o
usually equals inoculum concentration in
the fermenter) and taking the time at which exponential
growth commences as zero, eq. (19) can be integrated to
yield.
X
In - = (20)
X
o
Hence, a biomass doubling time td can be shown to be
In 2
td
(21)
11
Similarly, a cell number doubling time or mean generauon
time t
g
is gi yen by
In 2
t
g
=--
(22)
flN
where flN is the specific cell number growth r3te.
In bacteria, where cell division leads to two identical cells,
flc-; and fl "'illbe the same. Foryeasts, moulds, plantcells and
otherorganisms and fl are not always equal. The specific
growth rate is characteristic of the microorganism and is a
function of growth environment including temperature, pH,
173
-- --
- ---
Biotechnology/The Science and the Business
To vaccum
Holding
coil
Heater
Holding
Flash
evaporator
coil
sterile
Steam
medium
Raw
medium
injection
Sterile
medium
(a) Direct steam injection (b) Indirect heating
Figure 6. Stcrilization processes: (J) direct steam injcction; (b) indirect heating.
Raw
medium
..
E1.0..----------------------,
~
::J
o 0.8
~
a:
~ 0.6
u
o
-l
W 0. 4L----L........l-.L..J...LUJL.I.----L---l...-L...l...LU.l..l..-----J'---'--Ll-Ll..LJ.J
>
Id Id 10
4
10
5
REYNOLDS NUMBER. Re (-l
igure7. U/U
max
vs. Reynolds numberin straight, circular pipes.
articlesbyfilterfibres;inertial impaction onfibres ofparticles
I ith sufficiently high momentum that they are incapable of
following the gas stream as it flows around to avoid thefibre
~ n t,hey impact on the fibre; small particles mm'e around in
rhe gas by diffusion and Brownian motion, they eventuall)'
~ o l l i andareretained onfilterfibre. Othermodesofparticle
emoval mechanisms may occur to various degrees. Thl:
emoval of particles with filter depth follows the equation
No k
In N = rz (17)
7
2
where No and N are, respectively, the initial and final levels
of contamination in air, z is the filter depth to achieve N
contamination level and k
r
is the filter constant. k
r
depends
on bed void fraction, fibre diamcter, velocity of flow
and temperature. formation of condensate affects filter
performance and should be prevented.
The air filter is itself sterilized by direct or indirect steam
heating. Chemical sterilization is also practiced. ~
Acommontestoffilterperformanceis based onpenetration
ofdioctylpathalate (DOrtest) particles mainly of :::; 0.3 ~ l m
size. A penetration of0.003% orless corresponds to removal
PrefiIler
( Humidify
control)
Air
1------,
_steam
Filter
one-way
valve
steam
Condensate
drain
..
" .
.. .
Fermenter
Figure 8. Air filtration using apacked filter.
-
Fermenter technology
z
0
".
-
,-
I-
,/
/
<I:
/
/
cr:::
I-
/
/
Z / Product
w
/
/
(.)
Z
0
---
(a) (.)
TIME
/
/
/
".
....
/
/
/
/
/
/
/
/
Product
/
(b)
TIME
Figure 11. Relationship between growth and product formation: (a) growth associated product formation;
(b) non-growth associated production.
ofa purercarbon sourcesuch as glucose compared,say, with
molasses may reduce purification problems and simplify
pollution control and waste treatment.
The medium must provide sufficient carbon, nitrogen,
minerals and other nutrients to yield the required amount of
cell mass and product. Minimum requirements arc estimated
from the stoichiometryofgrowth and productformation. In
general,
C-source +N-source +minerals + specific nutrients +O
2
(vitamins, hormones)
-> cell mass + product + CO
2
+ H
2
0 (24)
Most nutrients arc supplied at levels well abo\'e the minimal
needs. Other considerations relating to fermentation feed-
stocks have been examined in Chapter 14.
BIOREACTORS
Fundamentals ofMass and Heat Transfer
Transport of mass and heat are encountered not only in
bioreactors but also in most other processing operations.
Heat sterilization of fermenters and temperature control
during a ferment:1tion are both dependent on heat transfer
phenomena. Similarly, the transfer of oxygen from a gas
phase into a liquid and within the liquid to the biocatalytic
particle are problems of mass transfer.
Gas-Liquid Mass Transfer
The transfer of oxygen from the gas-phase to the micro-
organism suspended in the gas-liquid dispersion takes place
along a certain pathway.Themost general transport route is
12)G':'S-lI0UID
INTERFACE
OXYGEN
141 BULK LIQUID
{3l LIQUID
FILM
AIR BUBBLE 1
(6)CELL- LIQUID
INTERFACE
(8) SITE OF
BIOCHEMICAL
REACTION
Fipure 12.. Oxygen transport path from the gas bubble to the
mlcroorgantsm.
depicted in Figure 12 which shows that eight resistances to
oxygen transfer can exist:
1. in a gas-film inside the bubble;
2. at the gas-liquid interface;
3. in a liquid film at the gas-liquid interface;
4. in the bulk liquid;
5. in a liquid film surrounding the cell;
6. at the cell-liquid interface;
7. the internal cell resistance; and
8. the resistancc at the sites of biochemical reaction.
Not aU these resistances are significant. Thus, in practice
bioreactors operate at such levels of turbulence in the
fluids that convective transportdominates in the ~ d y ofthe
Fennenter technology
:.. .. ~ ~ ~ r Tn other words, the mass transfer IS liquid
Since the transfer rate and the flux are related by
175
Biotechnology/The Science and the Business
medium compositIOn and dissolved oxygen levels. Some
typical doubling times for different classes ofmicroorganisms
are given in Table 1. Exponential growth is followed by a
stationaryphase duringwhich the growth and death rates are
equal.Exhaustionofagrowth-limitingnutrientin themedium
oraccumulation oftoxic material are possible causes ofonset
ofthe stationaryphase. Eventually theculture enters adeath
phase in which cell lysis orsome othermechanism ofloss of
cell viability overtakes growth.
Table 1
Typical Doubling Times of Some Industrially Important
Classes of Microorganisms
td (minutes)
Bacteria 45
Yeasts 90
Moulds 160
Protozoa 360
Mammalian hybridoma 630-1260
Plant cells 3600-6600
Effect of temperature on growth
Depending on the optimum temperature for growth, micro-
organisms are classified as psychrophiles, mesophiles and
thennophiles. Typical optimum growthtemperaturefor these
is giYen in Table 2. Actually, there is a range oftemperatures
ncar the values given in Table 2 over which these classes
of organism grow; the exact optimum growth temperature
dependsonthemicrobialspeciesandothergrowthconditions.
The efficiency of conversion of the carbon source to cell
massis temperaturedependentanddeclineswithtemperature.
Maximum growth yield is obtained at temperatures lower
thanthosefor maximumgrowth rate. Furthermore, tempera-
ture optima for growth and product formation are not
necessarily the same.
Table 2
Typical Optimum Temperature for Growth
Psychrophiles -15
Mesophiles - 37
Thermophiles -55
Substrate concentration effects on growth
The effect on growth of the concentration of a growth-
limiting substrate, such as acarbon sourcefor example, often
followsthebehayiourshowninFigure10.Thespecificgrowth
rate !l increases with substrate concentration until it is no
longer growth limiting. The curve in Figure 10 is described
by the Monod equation
S
(23)
j-Lt


2

Figure 10. Effect of substrate on specific growth
rate. K
s
is the substrate concentration at half llm
o
where is the maximum specific growth rate and K
s
is the
saturation constant. Numerically, K
s
is the substrateconcen-
tration corresponding to !lm/2. Thus, growth on a given
substrate may be described by two constants: and K
s

However, at high substrate concentrations inhibition of

growth due to substrate may be encountered. High substrate
concentrations may adversely affect product formation in
many fermentations.
Consideration ofthe relationship between cell growth and
product formation is essential to the successful conduct ofa .,----'
fermentation. Two simple, extreme possibilities are growth-
associated product formation (Figure 11(a)), in which case
the product formation results from primary energy metab-
olism, and nongrowth associated production where product
concentration is proportional to the quantity of biomass but
not to growth rate (Figure 11(b)).
Growth Media
Careful formulation of growth and production medium is a
prerequisite ofsuccessful fermentation. Microbial nutritional
and environmental needs have to be met as well as several
technico-economicconstraints. Mediumdevelopment aims to
maximize product yield and cDncentration at minimum
medium cost. Although traditional emphasis is on the fer-
mentation step, the choice of medium affects downstream
and upstream (pretreatment) activities and medium design
should be carried out in an overall process context. The usc
174
r
It = K ;,n-l (53)
,If' I
while the average shear rate (y) in the stirred vessel is often
approximated by the equation
;/
,
= k
,
K (54)
Substitution of equations (53) and (54) in equation (52) leads
to a modified Reynolds number:
-N'-n D'
PL - ,-
(55)
Re, = K kn-1
I
The Powernumber-Re,'nolds number behaviour for non-
Newtonian media is to that described for the
Newtonian fluids. The constant k, in eg, (54) is dependent on
the system geometry and to some extent on the rheology of
non-Newtonian media. The dependence of k
i
on the flow
index (n) of non-Newtonian fluids has been reported to be
4n
Fermenter technology
sterile air. These bioreactors employ multiple impellers, usually
Rushwn turbines. \X'ith some modifications the design
methods de,doped for the standard reacwr can be usefully
employed for other geometric configurations. 1\lultiple
impellers are located on a single shaft with a minimum distance
of one impeller diameter between impellers. Two diiferent
types of impellers may be used on the same shaft and are in
fact beneficial in some cases. For example, a combination of
radial flow Rushton turbines and axial flow propellers is
sometimes used to provide improved mixing in vessels \\'ith
HId
T
> 1. Recent investigations inw other impeller designs
ha,'e sho\\'n that impellers such as Prochem hydrofoils and
InrerMIG provide higher mass transfer at lower power con-
sumption than Rushton disc turbines. Some of these newer
impeller types are depicted in Figure 18. In many cases
existing stirred tank bioreactors are known w have been
upgraded by changing the Rushton turbine in favour of the
newer impellers. The lower shear characteristics of the latter
k= K'
I
(
3n + 1
added advantage.
/
where K' is impeller-tank geometry dependent. Some typical
values of kj are given in Table 5.
Table 5
Values of k
j
(eq. 54) for Various Impellers
k
1
Impeller (-)
6-Bladed turbine 11-13
Paddles 10-13
Propeller
-10
Helical ribbon -30
Quite distinct from an average shear rate in the tank, a
maximum shear rate on the impeller also exists, Considerations
of maximum shear to which a fermentation culture may be
exposed without harm are particularly important for fragile
biocatalysts \\hich also tend to have large particle sizes (animal
cells, plant cells, fihmentous organisms). The maximum shear
rate on a Rushton turbine in Newtonian media has recently
been expressed as
y = 3.3 Nl.
5
(D
j
2
PL )[1 (57)

which applies for Rei = 100 to 29,000.
Other Stirred Tank Configurations
The "standard" configuration of stirred tanks is not the
commonly used geometry in bioreactor applications; instead
tanks with height-to-diameter ratios of 3: 1 and 4: 1 are more
common because they permit better utilization of the expensive
Pneumatically Agitated Bioreactors
Although traditionally in extensive use, the mechanic"ally
stirred bioreactors have several significant limitations \\'hen
compared to pneumatic bioreacrors which are agitated b}' gas
injection. A comparison of pneumatic bioreactors (airlifts
and bubble columns) with stirred vessels is provided in Table
6. Numerous advantages of the gas-agitated reactors have led
to a clear preference for them particularly in the newer bio-
technology applications im'olving fragile material. The
following sections examine some of the design considerations
for bubble column and airlift bioreactors.
Bubble Columns and Modified Bubble Columns
Bubble columns are among the simplest of gas-slurry bio-
reactors. A gas-sparged pool of Ii uid with heioht-to-' er
ratio well a ove unlt\- constitutes a bubble column
19). Several modifications to the basic design are
possible, however, and Figure 19 illustrates some possible
configurations.
The energ" input to the fluid in the bubble column arises
redominantly from isothermal expansion of the injected gas
and epen s on t e super cia gas ve ocity
Pc
V = PL g UG
(58)
L
The main reactor performance characteristics gas holdup
(f), specific gas-liquid interfacial area (ad, overall volumetric
mass transfer coefficient (kLad, mixing, axial dispersion (Ed
and heat transfer - are controlled by gas flow and hence also
by the energy input. A wide range of gas velocities may be
used; howe,'er, the maximum velocity should be less than the
181
Biotechnology/The Science and the Business
liquid and hence the associated resistance (i.e. resistance 4 in
Figure 12) can be ignored. Similarly, for single cells or dis-
persed mycelia the resistance due to the liquid film on the
surface of the cell (i.e. resistance 5 in Figure 12) may be
neglected. This is so in spite of the fact that the density
differences between the microbial cells and the
fluid areverysmalland,consequently,theremaybeastagnant
liquid film surroundingthe microbe. Theminute dimensions
of a microbial cell and its large surface area are the reasons
for a negligible cell-liquid film resistance. The cell-liquid
interfacial resistance and the resistance at the reaction sites
(resistances 6 and 8, respectively, in Figure 12) can both be
disregarded because of the active oxygen transport through
the cell membrane as well as the rapid rates ofbiochemical
reactions. An intracellular resistance (i.e. resistance 7 in
Figure 12) can also be discounted since the enzymes for
terminal respiration are located in the cell membranes rather
than in theprotoplasm, and again becauseofthesmall sizeof
microbial cells. This efu!!inatesa,ll the transport resistances
.the interface. The oxygen
reduced to thatofthe
"Interfacial mass transfer. .
The mass transfer models
The region in the vicinity ofthe gas-liquid interface may be
visualized as consisting ofadjacent, stagnant, gas and liquid
films of some finite thickness as depicted in Figure 13.
According to this two-film model, the resistance to transfer
in each phase is localized in the thin films close to the
interface.Theinterfaceitselfis assumed to offer no resistance
to mass transfer; the interfacial concentrations are therefore
determined by the equilibrium relationship. Mass transfer
throughthestagnantfilmsisassumedtobesolelybymolecular
diffusion andthusat steadystatelinearconcentrationprofiles
exist in the films (Figure 13). Forthis situation the mass flux
(Jo,) of the diffusing species is related to the concentration
gradient in the film and to the film thickness (6) in
accordance with Fick's first law:
D
Jo
z
= "6
(25)
where D is the molecular diffusivity of oxygen in the film.
Theratio D/6 is known as the"mass transfercoefficient", k.
Equation (25) may be written for each of the two films.
Jo
z
= kG (C
G
- C
Gi
) (26)
= k
L
(C
Li
- Cd (27)
where kG and k
L
are the gas and the liquid mass transfer
coefficients, respectively. Since the interfacial concentrations
are in equilibrium, the flux may be expressed in terms ofthe
overall concentration driving force as follows:
(28)
DIRECTION OF DIFFUSION
..
INTERFACE
BULK BULK
GAS
GAS LIQUID LIQUID
FILM FILM
':J
!
C
G
I "STAGNANT"
I
FILMS
I
I
I
I
I
I T
I
I
I
I
C
L
I
I
I
1--8
G
8
L
--J
Figure 13. Oxygen concentration profile in the gas-liquid inter-
facial region.
where K
L
is the O\"erall mass transfer coefficient based on
liquid film. C:' is the equilibrium concentration in the liquid,
which, for a sparingly soluble gas such as oxygen, is related
to C
G
by the equilibrium relationship known as Henry's
law:
C
G
= He:' (29)
where H is the Henry's law constant.
From equations (26), (27), and (28), and the knowledge '-..-/
that CLi = H C
Gi
, it can be shown that
1 1 1
- = - + (30)
K
L
k
L
Hk
G
Forsparingly soluble gases, such as oxygen, H is very much
larger than unity. Moreoycr, kG is typically considerably
larger than k
L
because the gas phase diffusivities are vastly
greater than those in the liquids (d. Doxrgon/air = 10"'
at 20C), and at the same time, the
film thicknesses are smaller than those of the liquid films.
Underthesecircumstances thesecond term on the right hand
side of equation (30) becomes negligible and the equation
reduces to
1 1
(31)
K = k
L L
This implies that essentially all the resistance to
mass transferofa sparinglysoluble gas lies in the liquid film
"blo'v out" (spray formation) condition. The maximum liquid
velocities that may be used tend to be quite low because of
the long residence times typical of bioreactors.
The hydrodynamic regime of operation influences column
performance. At low gas velocities (:s 0.05 ms-
I
) in water-
like fluids, bubbles h;l\"e spheroidal shapes and rise uniformly
with no interaction. This is the bubble flo,,"' regime (Figure
20) which is also kno""n as "homogeneous" or "unhindered
bubble flow". As the gas velocity increases the bubble motion
becomes unstable and chaotic, and the column becomes more
turbulent. Larger bubbles with little definition to their shape
coexist with many small bubbles. This is the churn-turbulent
regime (Figure 20). The transition from bubbly to churn-
turbulent flow is gradual and it occurs oYer a range of gas
flow depending on the properties of the fluids and on reactor
geometry. Further increase in gas flow rates leads progress-
ively to increased bubble coalescence, slugging and annular
film flow; however, these other flow regimes are gener3lly
)t encountered in biore3ctor applications.
BUBBLE FLOW CHURN - TURBULENT
Figure 20. Bubble and churn flow regimes.
holdup.
t'he volume fraction of gas in dispersion, or "gas holdup" (E),
is an important characteristic of bubble columns and other
gas liquid reactors. By definition,
V
G
E = ---'=---
(59)
V
G
+ V
L
The reactor must be able to accommodate the gas holdup
produced under various conditions. Furthermore, the resi-
time of gas in liquid and hence the efficiency of util-
IZatIOn of gas depend on gas holdup
h
L
E
. Fermenter technology
The specific gas-liquid interfacial are3 for mass transfer is
controlled by a combination of gas holdup and me3n bubble
di3meter
6E
aL = ---- (61 )
dB (I - E)
where dB is the di3meter of a sphere h3ving the same surface-
to-volume ratio as the gas bubble, In practice a distribu'tion
of bubble sizes is encountered and dB is calculated using
N
I nj d/
dB = __
(62)
2
L nj d
i
i=l
",here nj is the frequency of occurrence of bubbles with a
diameter d
i
. Bubble size and frequency distribution may be
measured, fOf example, by electrical resistivity and other
similar probes. Alternatively, aL may be determined by one
of the direct measurement techniques such as sulfite oxidation.
In bubble columns gas holdup shows the following general )
dependence on gas velocity (
E = a UGb (63) )
The parameters a and b have been found to e 2.47 and 0.97,
and 0.49 an 0.46, respectIvely, for bubble (U
c
< 0.05 ms I)
and churn-flow (U
G
> 0.05 ms 1 regimes i, ai - for
broad ranges 0 bub Ie column geometries. Addition of in-
organic salts to water enhances gas holdup by a few percent
up to an ionic strength corresponding to - 0.15 M NaCI.
This effect is due to coalescence inhibition ",hich results
from electrical repulsion between like ions at the surfaces of
bubbles. For any given ionic strength, the type of inorganic
salt does not influence gas holdup.
Numerous other gas holdup correlations are aV3ibble in
the voluminous literature on bubble columns (see Reading
List). One example is the correlation
7
;
(64)
where c is either 0.20 (non-electrolytes) or 0.25 (electrolytes).
Equation (64) covers column diameters de = 0.152 - 0.6001
and U
G
= 0.004 - 0.33 ms-'. It applies to Newtonian media
such as water, glycerol and methanol.
In homogeneous non-Newtonian systems the following
equation may be employed
'-..-/
Biotechnology/The Science and the Business
mass transfer consumption
term term
where qo, and X are the specific oxygen consumption rate
(02/kg cells s) and cell concentration (kg celli m
3
)respectiyely.
Interruption of the air supply to the reactor eliminates the
mass transfer term (eq. 38) and the dissolyed oxygen con-
centration declines linearly with time (Figure 14) due to
oxygen consumption by the biomass, The slope of the C
L
YS.
t plot yields qo, X (Figure 14). The air supply is turned on
before the dissolved oxygen concentration has dropped to
the critical dissolyed oxygen leyel for the microbial species so
that the fermentation is not damaged. The oyerall volumetric
mass transfer coefficient is determined using the C
L
\'S. t
plot beyond the point of resumption of oxygen supply. Thus
eq. (38) is rearranged to
Air off

--'- _ C
crit
L-
TIME
Figure 14. The dissolved oxygen concentration (Cd vs. time.
C
L
= C,: - _1_ (qO' X + d CL)
(39)
k
L
aL dt
and kLaL is obtained from the slope of a plot of C
L
YS.
(qO' X + d (Figure 15). Rapid-response dissoh'ed
oxygen probes should be used to minimize the effect of
electrode delay on the measurements.
Gas-phase oxygen balance technique depends on measure-
ments of mass flow of aeration gas into and out of the
fermenter. The mass fraction of oxygen in the inlet and outlet
gas ,streams must also be determined (mass spectrometer,
paramagnetic oxygen analyzer) as well as the steady state
dissolved oxygen concentration (dissolved oxygen electrode).
The haL is obtained from the oxygen balance
M (x
o
- Xi) = V
L
h aL (C:' - Cd = V
L
qo, X (W)
where V
L
is the broth yolume, M the mass flow of gas and x
is the mass fraction of oxygen in gas (0 = outlet, i = inlet).
17
8
Figure 15. Calculation of kLaL'
Equation (40) assumes no eYaporation and it does not correct
for carbon dioxide production; howeYer, the necessary cor-
rections can be easily incorporated. With the steady-state
method any possibility of affecting the fermentation by inter-
ruption of air supply is circumvented.
i,;.
Heat Transfer
Most fermentations require careful temperature control. Heat
generated by agitation and aeration power input and that
generated by the fermentation itself needs to be estimated for
design of sufficient cooling capacity. Sterilization operations
also require knowledge of heat transfer and necessitate the
provision of sufficient heating capability.
Typical fermentation heat generation for bacterial, fungal
and yeast fermentations is of the order of 3-15 kW m-3.
The exact amount depends on the nature of the substrate and
its rate of oxidation. A highly reduced substrate such as a
hydrocarbon would release more heat per mole substrate on
complete oxidation than a relatively less reduced )
Methods for estimating the heat evolution have been discussed
in Bailey and Ollis.
2
Between 1 and 15 kW m-
3
of heat input occurs due to
agitation in stirred tank fermenters. In bubble columns and
airlifts the contribution of heat due to agitation is usually less
than 5 kW m-
3
Once the heat transfer rate ("'hich equals the
heat evolution rate pItts the heat generation due to agitation
at steady state) is established, the heat transfer area needed to
obtain this rate is calculated from:
(41)
where A
H
is the transfer area and /).T is the mean temperature
difference dri,-ing force. U
H
, the overall heat transfer coef-
ficient, is the sum of the resistances to heat transfer due to
the fluid films on either side of the heating or cooling surface,
fouling (corrosion, protein burn on) resistances on either
side, and the resistance due to the metal wall through which
the heat must pass. Hence,


velocity usin" most of the J\":tibbk correbtions is limited
onl)" to specific reactor geometries 0\"([ narrow r:tnges of
scale.Onl)"recentlydidamoregeneral:tirliftdesignprocedure
become a\"aibble as discussed bter in this section.
Theinducedliquidcircubtionisanimportantdistinguishing
characteristic ofairlift reactors. In other typcs ofbioreactars,
such as the bubble columns and the stirred tanks, the general
requirement of long residence times se\-crcly limits the
maximum linear flow "c1ocit)" through the reactors unless
recycle flow is en;ployed. In airlifts, high linear liquid velo-
cities arc attainablewithoutrecycleand theselead taimprowd
turbulence and good mixing, heat and mass transfer. The
liquid circubtion in an airlift reactor originates from the
difference in the bulk densities of the fluid in the riser and
the downcomer. The fluid circulates along a well defined
path: upflo\\" in the riser,i0\\'nflow in the downcomer. A
mean circulation velocity (ULd is defined as
(74)
where Xc IS the circulation path length and t
c
is the average
t
Air
External-loop
oi rlift
. .
. ,
o 0
Q 0 .. \)
, 0
c
Internal-loop
split-cylinder
air lift
Figure 22. External- and interaI-loop airlift reactors.
Fermenter technology
time for one complete circulation. The circulatory flo\'.' is
clearly revealed by'injection ofa tracer such as an acid pulse
into the downcomer (or riser) and follo\\'ing the tracer flow
at some downstream location. The characteristic deca:"ing
sinusoidal tracer response depicted in Figure 23 is obsern:d:
the time difference bct\\'een adjaceIH peaks is the circulation
time.
Unlikeano\'eral1, awrage,circulationvelocity(ULd,v ~ l u s
of a superficial velocity measured either in the dO\\'ncomer
(U
Ld
) orthe riser (U
L
,) are more me:lIlingful. The continuity
criterionleads to thefollo\\'ing relationship betweentheliquid
velocities in the riser and the downcomer:
(75)
TRACER
INLET
DETECTOR
Figure 23. Tracer response in an airlift reactor.
Thesuperficial\'elocitymustbedistinguished from the"linear
liquid velocity", also known as the "interstitial velocity",
because in reality the liq uid flo\'.' occupies only a pan of the
flow channel, th.e rest being taken up by the gas. The inter-
stitial velocity (Vd and the superficial velocity are related as
follows:
V
Lr
~
(76)
1 -
f,
and
V
Ld
~
(77)
1 -
Ed
The velocity of liquid circulation, \\hile itself controlled by
the gas holdups in the riser and the downcomer, in turn
l
'---/
Biotechnology/The Science and the Business
. The mechanical power requirements of ungassed stirred
tanks maybeestimated usingPowernumber(Po)vs. impeller
Reynolds (Re;) number plots, examples ofwhich are shown
in Figure 17. The Power number and the impeller Reynolds
number are defined, respectively, as
(-+6)
and
(-+7)
10
Flat-blade turbine
0
Turbulent Paddle
0-
Prochem impeller
1.0
Marine impeller
0./
I 10 10
2
10
3
10
4
10
5
Re (-)
I
Figure17. Powernumbervs.impellerReynoldsnumberforvarious
impellers.
Theexact nature ofPo - Rei plots is dependent on impeller
type and on the presence or absence of baffles. The power
absorption by liquids in unbaffled tanks in turbulent flow
(Rei > 10
2
) is significantly less than in baffled tanks. In
laminar flow (Rei ::; 10) the power number is inversely
dependent on Rei:
Po :x Rej-t
(-+8)
with the constant of proportionality dependent to some
degree on the type of impeller. Under developed turbulent
flow conditions (Rei > 10
4
) the power number becomes
independent of the impeller Reynolds number, but depends
on the impeller type. Because most applications are likely to
im'oh-e highly turbulent reactors, a compilation ofconstant
PO""er numbers for yarious geometries is provided in
Table 3. The values of the proportionality constant (eq. -+8)
are:given in Table 4.
Introductionofgas into themixing\"esse! always leads toa
reduction in the power absorption relative to the ungasscd
situation.Onceanestimateofungassedpower(P)is available,
the power input in the presence of gas may be calculated
using the Michel-\1iller equation
180
Table 3
Turbulent Power Numbers in Stirred Vessels
r
Geometry Po
(Baffled tank) (-)
Propeller (square pitch, 3-blades) 0.32
Turbine (6-bladed) 6.30
Turbine (6-curved blades) 4.80
Flat paddle (2-blades) 1.70
Prochem impeller (5-blades, D
i
=d
T
/2) 1.0
P
2
N D.3)O.45
PG = 0.72
(
Q0.56 I (49)
This equationprovides a good approximation in many appli-
cations but it should not be used for extreme values of gas
volumeflow (Q). Otherdesign parameterssuch as theoverall
gas holdup (e) and the \'olumetric mass transfer coefficient
(kLadcan becalculated on the basis ofavailable correlations
which have been summarized by Mann.
6
Some useful
correlations are
3
j
10 = 0.52 (N (PL D f65 1.4 (50)
and
(51)
Numerous other equations are available which may be more
suitable for specific situations.
Table 4
The Values ofc in Po = c Rej-I for Various Impellers
c
Impeller (-)
6-Bladed standard turbine (unbaffled -100
tank)
Helical ribbon (unbaffled tank) -380
Propeller (3-bladed, square pitch, baffled -40
tank (4-baffles))
Non-Newtonian media
Fornon-Ne",tonian media the impeller Reynolds number is
based on the apparent viscosity of the fluid:
Rei = PL ND? (52)

For the often observed power law behayiour the apparent
viscosity is giYen by
liquid flows in a YCrtica! pipe using either the' fluid of interest
or a reasonJble' simuhtion of the fluid. The experimentation
is yery simple and straightfonnrd. For example, for air-
water the following applies
E =0 Uc, (85)
r 0.24 + 1.35 (UCr + U Lry:'9J
when U
Lr
> 0.3 ms-
1

Knowledge of the riser and the downcomer holdups enables
the calculation of holdup (c:):
A
r
E
r
+ Ad Ed
E = (86)
A
r
+ Ad
and hence the height of gas-liq uid dispersion:
(87)
In equation (87) h
L
is the unaerated liquid height.
Design at the hydrodynamic and mass transfer level would
ilvolye the prediction of U Ln , En Ed and h
D
for any giYen
operating conditions (thn fluids) and gi\'en reactor geometry
(An Ad, A
b
and hd. A design flow chart for internal-loop
airlifts has been published.
12
Gas-liquid mass transfer
The measurement of the oycr:l1l yolumetric mass transfer
coefficient and gas holdup in a gi\'en fluid in a small bubble
column enables the calculation of the ratio kL/d
B
:
k
L
_ (1 - ) k a
"'--_---'------'L=---=L = "4'
(88)
dB - 6E
This ratio has been found to be constant (ljJ) for any specific
fluid oyer broad ranges of gas flow rates. Thus k
L
/ dB which
has been experimentally determined in a bench-top model
'eactor may be used for estimation of kLaL in larger production
vessels. An estimate of the gas holdup in the reactor is first
obtained using the procedure described earlier; kLaL is calcu-
lated as follows:
'\j.'6
haL (89) =0
(1 - )
For air-water the parameter '\j! is S-I. For fluids
made up of filamentous or fibrous solids suspended in a
water-like medium "4!(S-I) depends on the concentration C
s
(dry wt./vol.%) of solids:
(90)
where PL, flL, D
L
and a refer to the properties of the
suspending fluid.
Other considerations
Substrate injection. The problem of location of substrate feed
in an airlift vessel becomes p:lfticularly significant in
COntmuous and fed batch operations. For rapidly utilized
Fermenter technology
substrates, the concentration of which must be kept low for
reasons such as substrate toxicin' or substrate inhibition. the
microorganisms in a tall airlift m'ay be stan-cd of the
only a shon distance dmvnstream of the point of substrate
injection. Thus, multiple substrate feed points may be neces-
sary axially up a reactor if product yield reduction due to
substrate stan-ation is to be avoided.
The substrate balance for a differential volume of the riser
may be written as
(91 )
where S is the substrate concentration at an,' vertical position
z, VLr is the interstitial riser liquid w!ocity,'E
Lr
the riser axial
dispersion coefficient of the liquid phase and .Rs the rate of
substrate consumption. When the substrate concentration
must not fall below a critical minimum value Smin and it
should not exceed a maximum of Sma", because of inhibition
considerations, then eq. (91) may be solved with appropriate
reaction kinetic expression to determine the axial distance at
which fresh substrate addition becomes necessary.
Gas sparger
Perforated plate gas spargers are often used in airlifts and, in
keeping with the practice in bubble columns, these plates are
located at the base of the riser in the airlift. However, this
type of sparger positioning is inappropriate in airlift devices
because the recirculating flow from the downcomer leads to a
maldistributi'on of gas (Figure 24(a)). The use of perforated
pipe ladder type gas spargers located just above the point
where the flow from the downcomer meets the riser leads to
imprO\'ed gas/liquid flow (Figure 24(b)). Perforated pipes are
recommended for bioreactor applications.
Immobilized Enzyme Reactors
Immobilized enzyme (and immobilized whole cell) catalysts
(see Chapters 17 and 18) can be employed in a variety of
reactor configurations. Catalyst panicles may be used in sus-
pension as in stirred tank and fluidized bed reactors (Figure
25) or they may be held in place in fixed or packed bed
devices. Hollow fibre reactors containing catalyst immobilized
either throughout the thickness of the fibre wall or confined
to one side of it (e.g. perfusion systems) are possible. Fbt
polymer membranes containing immobilized catalyst ha\'e
been used in spirally wound configurations. Immobilized
particulate biocatalysts can, of course, also be used in airl;[t
and bubble column reactors so long as the solids loading and
density are not excessi\'e. In such reactors a compressed gas
prO\'ides the necessary agitation in the fluid and gas-liquid
mass transfer is not the main consideration.
Reactor efficiency is measured by the quantity of substrate
tr:lI1sformed per unit time per unit mass of immobilized
Biotechnology/The Science andthe Business
Table 6
A Comparison ofMechanically and Pneumatically
Agitated Bioreactors
Mechanical Agitation Pneumatic Agitation
(Stirred Tank) (Airlifts, Bubble Columns)
1. Mechanically complex Mechanically simple and
(stirrers, shaft, seals, robust
bearings)
2. Often high shear Gentle, low shear levels
(suitable for tissue culture,
plant cells, fragile genetically
(a ) engineered microorganisms)
Prochem Maxflo hydrofoil
3. Gas throughput limited High gas throughputs
by impeller flooding possible (particularly in
airlift devices)
---------------------- '---:-
4. Difficult to clean due to Easy to clean. Extended
mechanical complexity; asceptic operation possible
greater possibility of (useful in continuous
contamination over operation)
extended operation
5. Turbulence confined to More uniform distribution
impeller zone in viscous of turbulence
non-Newtonian media.
Gas channels through
the impeller zone while
the rest of reactOr
remains stagnant
6. Operationally flexible Limited operational
(controlled by impeller flexibility. Require more
speed and by gas flow careful design
rate)
(b)
Hydrofoil impeller
Sieve
plates
fI ~ ..
Vertical
baff les
- . .
.- -: "
.
-.
. ..
-
~
( c )
Air Air
Air
Figure 18. Some newer impellers.
Figure 19. Bubble columns.
Intermig
--
Fermenter technology
..

.. .
0
,
D
D 0
o "
o
D' 0
Cot01 yst
"
0
.
t>
"


" particles
o D
..

0
..
.. " 0
..
"
0 ..
0
.
0
..
00"
0
a) Stirred tank
b) Fluidized bed
..-
L
-
--
-<:::I":
W
\..I;
)1:


-y
+
Feed-.----"-----{
Feed
c) Packed bed
d) Hollow fibre
(e) Spiral wound membrane
system
module
Figure 25. DeploymerH of immobilized cat.11;'st: free suspension in stirred tank (a) or fluidized bed (b);
fixed catalyst in packed bed (c), hollow fibre (d) orspira! wound membrane (e),
Fibre wall
(permeable)
Membrane
supporting
cataIyst
Product
form \vhich is indicati\'(: of equal performance of the two
reactor systems in this regime. However, ""hen 5 K
s
the
reaction rate is first order in substrate (eq. 92) and the plug
flow system gives a b,>tter performance than the continuous
stirred tank. In the lattcr, all the catalyst \\'ould bc exposed
to a low substrate concentration and this can bc utilizcd
ad\'anrageously in continuous stirred tanks ,,'hen the re:lction
is inhibited by substrate.
The theor:tical efficiency of other types of reactors is
between thc two extremes of thepacked bed and continuous
Stirred tank flow geomctries.
Mass Transfer Effects
Heterogeneous catalysis has its associated mass con-
siderations. Mass transfer resistances at the interface of solid
supportand the bulk liquid and within the solid matrix often
reducetheeffectiveness of theimmobilized form. Adnntages
of immobilization should be weighed against possible disad-
Yantages in the process of choosing a particular form of
biocatalyst.
An:llysis of the interfacial and mass trJl1Sfer
and cat:dystperformanccis illustrated for a spherical catJ1:.st
GAS
Biotechnology/The Science and the Business
The shear rate (1') expression commonly employed for the
calculation of apparent yiscosity of fluids in bubble columns
(sec eq. (3 is
l' = 5000 U
G
(66)
which is due to Nishikawa and coworkers.
9
This expression
[eq. (66)] is used for the calculation of in eq. (65). Howeyer,
there is a considerable degree of uncertainty on the mean
shear rate in bubble columns.
According to some recent work, to the simple holdup
equation (eq. 63) should apply to non-Newtonian media
also. The parameters a and b now depend on the properties
of the fluid as well as on the flow regime. The parameter b
has been empirically correlated with the flow index according
to
b = 0.564 n-0.354
(67)
Equation (67) disregards any flow regime effects, but It IS
based on data on a variety of fluids including fermentation
broths of fungi Chaetomium cellulolyticum and Neurospora
sitophila. Other gas holdup data on slurries which simulate
fungal media is available elsewhere.
ll
,12
Gas-liquid mass transfer
T\\o of the correlations for the overall yolumetric mass transfcr
coefficicnt in N cwtonian fluids arc:
and
(69)
These equations were developed by Fair
13
and Akita and
Yoshida
7
, respectively. For air-water, a simple cquation is
kLaL = 2.39 X 10-
4
(P
G
lVd
S6
(70)
which has been shown to apply up to a height-to-diameter
ratio - 24. Notice (cqs. (68)-(70 that the oyerall yolumctric
mass transfer coefficient may be based either on the liquid
yolumc (kLad or on the volume of gas-liquid dispersion
(kLaD). These two arc related as follows:
kLao = kLaL (1 - E) (71)
The: mass transfer ,york on non-Newtonian media in bubble
columns is less extensive. Some equations which may be
useful in estimation of mass transfer performance are
kLao = 8.35 X 10-
4
U 0044
G
II -1.Q1
.tp
(72)
due to Godbole et al.
s
and
kLao = 3.15 X 10-
3
U
G
0.59 '[ -0.S4
,.lp (73)
J
due to Deckwer et al.
14
For additional information on non-
Newtonian systems the work of Schumpe and Deckwer
l5
r
should be consulted. Gas-slurry systems haw been treated 11.16
elsewhere.
A Yast amount of literature on bubble columns is available;
some of the main sources are listed in the Reading List.
Airlift Bioreactors
Airlift bioreactors consist of a liquid pool divided into t\\o
distinct zones only one of which is usually sparged by gas.
The different gas holdup in the gassed and ungassed zones
results in different bulk densities of the fluid in these regions
which causes circulation of fluid in the reactor by a gas-lift
action. The part of the reactor containing the gas-liquid upflow
is the "riser" and the region containing the downflowing
fluid is known as the "downcomer". Figure 21 shows the
schematic of an airlift reactor.
.
'-..t->=t-- DOWNCOMER (DOWNFLOW)
SPARGED RISERS
(UP FLOW j
SPARGER
GAS ....
.

Figure 21. Schematic of an airlift reactor.
Airlift reactors have been successfully applied to almost
evcry type of fermentation. Many examples haye been cited
17
in other works.
5
Recent applications include hybridoma
cell culture for monoclonal antibody production on a
commercial scale.
Airlift reactors are ayailable in two basic forms: (i) the
internal-loop airlifts in which the riser and the downcomer
are contained in the same reactor shell, and (ii) the external-
or outer-loop reactors where the riser and the downcomer
are two quite separate tubes which are linked ncar the top
and the bottom. The external- and internal-loop configurations
are shown in Figure 22. Modifications to the basic airlift
dcsign ha\'e been used to produce othcr sub-types of airlift
reactors, some of which have been discusscd by Chisti and
Moo-Young.
5
Estimation of such essential airlift reactor design parameters
as the overall gas holdup (E), volumetric mass transfer coef-
ficient (kLad and the magnitude of induced liquid circulation
Bioreactor Scale-up
Laboratory scale bioprocess denlopment identifies the opti-
mal fermentation conditions for the process. Oxygentransfer
requirements, maximum tolerable le\'els of shear, pH and
temperaturecontrolneedsshouldbecomeknownatthispoint.
Theobject ofscale-up is to reproduce on pilot orproduction
scale the successful fermentation results achiend in the lab-
oratory.The results are often specified as production rateper
unit fermenter volume.
In practice scale-up is quite complex. It is not generally
possible to reproduce exactly on the production scale all the
various parameters from laboratory or pilot scale units. For
example, at equal specific power inputs twO geometrically
similarstirred reactorsdo not have identical mixing times. As
a result scale-up is based on the strategy ofholdingconstant
only oneortwooftheseveralpossibleparametersatdifferent
fermenter scales. The parameter(s) held constant are those
which are considered to han the greatest impact on the
fermentation; furthermore, the criterion of geometric simi-
larity (i.e. keeping the ratios ofcorresponding lengths equal
on production and pilot-scale units) is not always rigidly
adhered to so thatsmall geometric variations may be utilized
to advantage as long as they do mot result in unpredictable
behaviour.
Thescale-upmethodswhichhanbeenmostoftenproposed
are as follows:
1. scale-up based on equal power input;
2. scale-up based on equal mixing times;
3. scale-up based on equal oxygen transfer (kLad;
4. scale-up based on equal shear rates (or impeller tlp
speed).
Thelist is notexhaustive. Forhighly aerobic fermentations
based on maintaining a constantoxygen transferrate
IS areasonableapproachbutinotherfermentations,limitations
such as those on shear rate may be equally important.
The following comments on scale-up apply to stirred tank
trpeoffermenters. Considerations for scale-up ofpneumatic
reactors, particularly theairlifts, "-'ere examined earlier in this
chapter.
Scale-up basedon equalPCIV
L
ratio
Th.e criterion of equal PG/V
L
on pilot-plant and production
unltS has been employed for certain antibiotic fermentations.
The. available evidence indicates that the necessary power
reqUIrements decrease with increasino fermenter volume ap-
. b
prOXimately as
P
G
<X V -0.37
L ( 112)
V
L
Consequently, keeping PG/V
L
constant in scale-up may not
Fermenter technology
be :In energy efficient approach. Furthermore, it may not be '
a satisfactory strategy for shear sensitiYC fermentations since
the impeller tip speed and Rernolds number scale-up by
factors of> 1when PdV
L
is held constantfor geometrically
similarnsscls.Table8showssomeoftheeffectsofgeometri-
call:' similarscale-up ofa 20 L reactor to 2.5 r:1
3
plant\"esse!.
Effects ofkeeping PG/VLconstant(i.e. PG/VL == 1 arbitrary
unit for both reactors) on impeller rpm (N), tip speed (:\0,)
and Reynolds number(Rej) is shown in Table 8. These par-
ametersscale-up by the respective ratios of0.34, 1.7and 8.S.
Table 8 also shows the of maintaining constant rpm
(N),constanttipspeed(NOJ) andconstantimpellerReynolds
number (Re;).
Table 8
Effects ofScale-upBasedonConstantPdV
L
(orConstant
N, NOJ, Rej) on Other Parameters
Laboratory
reactor Plant reactor
Parameter 20 L 2.5 m
J
Ii
/ V,

1 25 0.2 0.CJ16
N 0.34 1 0.2 0.C4
N 0i 1.7 5 1 0.2
Rei 8.5 25 5 1
Scale-up based on equalshear
The maximum shear rate is related to the impeller tip speed
which is held constant on scale-up. However, the shear rates
in the fluid which are governed by fluid turbulence or
Reynolds numbers do not scale-up proportionately because
the impeller Reynolds number does not remain constant.
(Table 8).
Scale-up basedon equaloxygen transfer
Maintenance of an equal o\-erall volumetric oxygen transfer
coefficient(kLadisoftentakentoensureequaloxygentransfer
on scale-up. Ihis is only when the oxygen transfer
dri\-ing force alSo remains unchanged on scale-up.
- For stirred tanks
(113)
where k] is dependent on geometry and k
2
and k
J
are scale-
dependent.
More complex scale-up methods rely on estimation of
kLaL as well as the spatial oxygen concentration profiles in
thereactorstoyieldavalueofoxygentransferrate. Operating
and scale-upparametersareadjusteduntil thedesired transfer
rates are obtained for realistic operating conditions.
(
I
\f
Biotechnology/The Science and the Business
affects these holdups by either enhancing or reducing the
\'elocity of bubble rise.
Airlift reactor design
Probably the first question faced by the designer of airlift
bioreactorsforaparticularapplicationwouldbeoneofchoice
ofconfiguration: external-loop or the internal-loop. Table 7
compares the performances of these two distinct geometric
types of airlifts; such a comparison could form the basis of
a preliminary choice. Generally internal-loop reactors have
better mass transfer characteristics. On the other hand, in
fluids in which \'Cry high viscosities necessitate greater tur-
bulenceand for adequate mixing and.mass transfer, the
external-loops may be preferable.
An energy balance over the circulating airlift loop can be
used to obtain the following equation for the superficial
liquid \elocity in the riser:
18
(78)
__T::---= + K r
(1 - E
r
)2 B Ad (1
= [ K 2g h U
Lr
D
Equation (78) is for low viscosity water-like fluids and it
ignores wall friction losses in the riser and the downcomer.
Theequation applies to external- and internal-loop configur-
ations ofairlift reactors.
TheparametersK
T
andK
B
arethefrictional losscoefficients
for the headspace and the bottom of the airlift reactors. For
typical internal-loop airlift the term containing K
T
(eq. 78)
can be ignored, while K
B
is dependent on the bottom
geometry:18
(
AAd)O.8
K
B
=ll.4 (79)
b
Equation (79) applies over an A
d
/ A
b
range of0.2 -1.8;A
b
is'
the free area for flow between the riser and thedowncomer.
In external-loop reactors K
B
= K
T
and a K
B
of 5 may be
Table 7
Relative Performance of External- and Internal-Loop Airlift Bioreactors
Reactor
Parameter
External-loops Internal-loops
Mass transfer (kLad
Overall holdup (E)
Riser holdup (lOr)
Downcomer holdup (fd)
Liquid velocity (U
Lr
)
Circulation time (tJ
Liquid Reynolds Nos. (shear)
Heat transfer
r86
lower
lower
lower
lower
higher
lower
higher
probably higher
used fordesign purposes for the following approximate geo-
metric ranges: Ab/A
d
= 1-2,Ab/A
r
= 0.25-1 and Lcp/d,p
= 2-7.
Recent research
19
has shown that for non-Newtonian,
pseudoplastic fluids, for which eq. (78) is unsuitable, the
following may be used for U
Lr
calculation:
ULrA
r
+ - PL g hD ULrAr (lOr - Ed)
1 3 [K
T
(A
r
)2 1 ]
+ 2PL U
Lr
A
r
(1 _ E)2 +K
B
Ad (1 _ Ed)2 = 0
r
(80)
The and in this equationarethefrictional pressure
drops in the riser and the downcomer, respectively. The in-
depthprocedureforthedeterminationofthesepressuredrops
is described elsewhere.19 Equations (78) and (80) assume that
the riser and the downcomer gas holdups are known; these
parameters are interrelated:
(81)
for internal-loops without a gas-liquid separator per se, and"-.-/
Ed =0.79 lOr - 0.057 (82)
.Ed = 0.46 lOr - 0.024 (83)
for external-loop airlifts. Equation (82) iSjSuitable for water-
like fluids while eq. (83) is more apprdpriate for slurries
encountered in fermentations ofsuch fungi as Penicillia and
Aspergilli andinthecultivationof microorganisms
like Streptomyces. .
The calculation of riser holdup for design and
scale-uprequiressomeexperimental particularly
when new applications are involved and fluids arc rheol-
ogically complex. Equations of the type -
need to be established by independent of gas and
higher
higher
higher
'.
higher
lower
probablY lGlwer "
r
J.

i
I
I
I
OverflOW
IJ
Solids
outlet

Tubular bowl
Fermenter technology
aeometl
T
. The v:tlue of L represents the are:t of a graYity
;ettling which is capable of the same sep:tr:tting ability
for continuous flmY oper:ttion as the centrifuge. For ap-
ropri:tte equipment selection the separation requirements
h:tye to be defined. It is uneconomical to specify equipment
for more stringent st?p:tration duty th:lI1 is rt?ally nect?ss:try.
Cle:trh', from eq. (1 H) the particle diameter and the density
differ;nce between the particle and the suspending fluid are
important factors affect separation. . .
Selection of a centrIfuge for any ne"'" appbcatlon ""'ould
almost always in\'oh"e expensi\"C pilot scale e\aluations. A
few simple laboratory tests can, hO""'e\'er, prO\'ide an indication
of whether or not the pilot run is e\"Cn worth pursuing, Ifa
sample of the slurry does not settle on standing under the
influence of gra\'ity o\"Cr seyeral days, it is unlikely that a
separator can achie\"C yery Howe.ve{', change in
characteristics such as particle sIze ""'hICh may be achle\'ed,
'for example, by the addition of flocculating agents or by
alteration of pH may implO\'e the likelihood of separating
difficult to settle solids, The flocs should be strong enough to
withstand the accelerational forces which are experienced as
the fluid enters the centrifuge and comes up to the same
rotation:tl speed as the bulk of fluid. Ifthe flocs formed are
easily disintegrated there is little ad\"Jntage to adding floccu-
lation chemicals.
A slight shaking of a bottle of gravity-settled solids can
provide an indication of how light the particles are. Other
solids properties such as the particle size and density need
also to be known. Here the density refers to solids as they
are in suspension and not dry. This difference is of particular
importance for biological solids which contain a high pro-
portion of ",,'ater and s""'ell when more is added. The solid s
may be fibrous (fungal mycelia) or slimy, or may occur as
pellets. These properties impact upon the choice of centrifuge.
example, some fibrous materials settle as mats under the
high centrifugal force and may cause de-sludging problems in
certain automatic solid-discharging centrifuges. Solid packing
characteristics and ease of settling can be easily judged by
spinning a test tube sample in a laboratory centrifuge at
- 3000 rpm for 3 to 5 minutes,
Equipment
SeW!"J! types of centrifuges arc a\'ailable; the more common
ones are:
(i) tubular bowl;
(ii) multichamber bO""'I;
(iii) disc-stack;
(iv) scroll discharge decanter centrifuge.
The particle size ranaes for which these configurations are
t>
suItable are shown in Figure 28.
Tub/dar bowl centrifllge. Shown schematically in Figure 29,
the tubular bowl is the simplest centrifuge configuration.
I

I

I

I

basket
.
ba
I I
!I Scroll discharge ,
I
I
Ultra
II
I
1 I
I
I
Tubular bawl
I Batch disc
I
NOlzle disc
I
Valve disc
Opening bawl
I
d Imperforate
I
I
I
I
I I
I
10-
2
10-
1
/0 10
2
10
3
PARTICLE DIAMETER (fLm)
Figure 28. Panicle size range for different types of centrifuges.
--., ,-Liquid
- C


1 solids
SOlids
""'" ( .. ,
Retained
solids
Decanler bowl
Figure 29. Centrifuges.
High 'g' -forces do permit good solids dewatering but the
operation is batch with respect to solids. The solids-handling
capacity is limited and solids recovery is labour intensive.
only slurries with low solids concentrations
can be economically de""':ttered.
flultichmnber bo'wl centrzfuge. This configuration (Figure
29) is basicall:-' a tubular bowl centrifuge with increased solids
capacity. Efficiency is maintained up to complete
filling of the chambers. Other oper:ttional char:tetcristics arc
nearly the same as for the tubular bo""'] machines.
Disc-st"ck centrzfuges. Disc-stack centrifuges come in sewral
types depending on whether the solids are retained or dis-
Mullichamber bowl
Disk slack
I
i
Biotechnology/The Science and the Business
DOWNCOMER
t
GAS
(a )
t
GAS
( b)
Figure 24. Positioning of gas spargers in airlift reactors: (a) poor
gas distribution; (b) irnpro\'ed gas injection.
catalyst for specified initialconcentration of substrate (50)
and its desired conversion X = (50 - 5)/5
0
, The conversion
characteristicsofdifferentreactorconfigurationscanbecalcu-
lated from a knowledge ofthe kinetics ofreaction. Thus, for
a reaction which obeys lvlichaelis-Menten kinetics
d5 k
r
e 5
(92)
dt - K
s
+ S
(e enzyme concentration, K
s
= Michaelis constant, 50
substrate concentration), we have for various reactors:
Bateb stirred tank
Change in quantity of Rate of substrate
substrate in the reactor consumption by reaction (93)
or
. d5 k
r
E 5
(94)
- V
L
dt = K + 5
s
whereE is thetotalamountofcatalystin thereactor.Equation
(94) may be integrated for 5 = 50 at t = 0 and 5 = 5 at t = t,>
to
k
r
E t = 5 _ 5 _ K In i
(95)
s
V
L
0 50
which can be rewritten in terms ofconversion Xas
k E t - -
= X 50 - K
s
In (1 - X) (96)
Continuous stirred tank ::-
The substrate concentration in the inlet stream is 50 and
because the reactor is well-mixed the substrate concentration ..
in the exit stream (5) is the same as in the volume of the '---/
reactor. A steadystate substrate balance in thereactorcan be
written as
5ubstrate flow into reactor = substrate flow out of reactor
+ substrate consumption due
to reaction (97)
or
Q
5
o
= Q 5 + K
r
E 5
K
s
+ 5
(98)
which can be rearranged and written in terms ofX:
krE
Q
= 5
0
X+ KsX
1 - X
(99)
Packed bed
Following the procedures outlined in the earlier examples,
the appropriate equation for a packed bed system with feed
flow rate Q is
krE - I
Q = 50 X - K
s
n (1 - X)
(100)
Because Q is the volumeprocessed in time t in a continuous
flow reactor and V
L
is thecorresponding volume for a batch
reactor, comparison of eqs. (96) and (100) shows that the
performance of batch stirred tank and plug flow systems is
identical. This is a general conclusion, irrespectinofreaction
kinetics. Howe\'er, kinetics alone do not determine reactor
choice and operational considerations are important. For
example, control of pH is operationally easier in stirred
reactors.
for reactions which display Michac1is-i\1enten kinetics,
continuous stirred tank and packed bed reactors operated
such that 5 K
s
, eqs. (99) and (100) reduce to an identical
188
JIil"'"''III'----------
1 dV
L
,0.P
(11S)
= hRF
The flow resistance R
r
is the sum of the resistallce due to
nlter medium (r
n
,) and that due to accul11ubted biomass:
. \\'
R
r
= 0: - + r (119)
A m
r
where c\: is the mean specific resistance of the biomass cake
and wis thedryweight ofaccumubted biomass. Atconstant
pressure (,0.P) plots of tlVL vs. VL for incompressible filter
cakes are linear, with slope and intercept dependent on con-
ditions of operation. However, biological materials usually
produce compressible cakes and experimental determination
of filtration volume vs. time relationship is necessary.
Microfiltration and ultrafiltration
Microfiltrationand ultra-filtration rely on porous membrane
filter media. Thebasic difference between the twooperations
is the "particle" size range handled, Microfiltration mem-
branes retain suspended solids down to BJCteria,
fungi and tissue cells are readily removed while pro-
tell1s and enzymes pass through the filter membrane at high
flux. Ult,rafiltration membranes hawmuch finer pores (1- 20
nm) which allow retention of proteins, enzymes and car-
bohydratesofnriousmolecularweiohtcutoffs.Thefollowin
o
discussioll places emphasis on the of bacterial
cells. Note that industrial ultranltration and microfiltration
srstems are physicallJ' and operationally simibr; the theore-
tICal of these two operations are equinlent. A
recent on ultrafiltration should be consulted for
details'. Inpracticalprocessingoperations microfiltration
IS usually employed in a cross-flow mode. The fluid to be
filtered flows parallel to the filter surface (Figure 31). The
:ross-flow of f:ed with respect to the filtrate flux generates
shearforces whichhelpto sweep thefilter surfaceofexcessive
solids build.up. However, inmost cases the buildup ofa thin
layerofsollds (concentration polarization)cannot beentirelv
prevented. '
The filtrate flux (]) through the membrane depends on the
pressure ,0.Pn,i> theviscosity of the suspend-
mg llquld and the hydraulic resistance of the membrane
Filtrate flux
Feed
_
. . I....
:'.:::.- +.<.,,0, _
!(....
72"3:=<;<3:<3<::I
Retenfate (concentrated
slurry of particles1
Filter membrane
Membrane support
Filtrate or permeate
Figure 3I. Principle of cross-flow filrer.
Fermenter technology
(R:-r) and the deposited solids (Rc):
] = ,0.P
n1
(120)
(R
c
+ R:-
1
)
The resistance of the solids layer depends on its thickness
(6
5
), voidage (5) and the particle size:
180 (1 - (5)2 6
5
R =
d
p
5
3
c
2
(121)
Biological solids from compressible cakes: the porosity (5)
of the solids layer decreases with increasing transmembrane
pressure and the flux does not increase linearly with ,0.P
n1
,
A typical rebtionship between filtrate flux and the trans-
membrane pressure is depicted in Figure 32, While for pure
liquids the flux varies linearly with ,0.P
n
l> for slurries in-
creasing ,0.P
n1
beyond acertainpointproduces no additional
benefit. As shown in Figure 32, for fixed ,0.P
TM
and other
operating parameters, an increase in cross-flow velocit\ en-
hances thepermeateflux duetoreduced solids byerthickness
at higher flow rate. Reliable theoretical prediction of the
a solid deposit is not possible and experimental
enluatlon IS necessary to determine suitable operatin
o
con-
ditions. The flux usuallyincreases with temperature (r:duced
viscosity) and with increasing flow rate parallel to the mem-
brane. In biological applications the upper temperature limit
would be determined by consideration of product stability
(usually < 40C). The choice of cross-flow rate would be a
balance between the pumping costs and the higher filtration
rate. The flux depends also on the concentration of solids in
the bulk fluid since it affects notonly the rate oftransportof
deposited solids back into the bulk flow (hence os) but also
the turbulence intensity on the retentate side. The flux de-
creases with solids concentration in the feed.
Theprocessequipmentconsistsofpolymermembranefilters
mountedin variousways. Ceramicmembranesare no\\' a\'ail-
/
/
/
I.
I
/ Pure liquid flux
/
/
/

/
/
t
increasing
/
cross-flow
/
_------ u, velocity
TRANSMEMBRANE PRESSURE (6P
TM
)
Figure 32. Typical relarionship between filtrate flux and trans-
membrane pressure.
195
Biotechnology/The Science and the Business
particle. For a substrate (e.g. oxygen or glucose) diffusing
into the catalyst particle the general substrate balance on the
particle is
Rate ofdiffusion into particle = rate of consumption (101)
since at steadystate there is no accumulation. Solution ofthe
appropriate balance equations leads to
(102)
where 5is the substrate concentration at any radius r in the
spherical particle of effectiYe diffusivity Dc; R
v
is the \'olu-
metric reaction rate. Equation (102) may be solved, either
analytically or numerically, using the applicable kinetic ex-
pression (RJ to obtain the positional yariation of substrate
concentration inside the particle. For example, for a first
order reaction with rate constant k
n
the expression for con-
centration at any radius r in the catalyst bead is
5 R
p
sinh [(k/Dc)O.5 r]
S; = -;sinh [(k/Dc)o.s R ] (103)
p
Equation (103) applies in the absence of interfacial mass
transfer limitation, i.e. when concentration in the liquid at
the solid/liquid interface (Sj) is the same as the bulk liquid
concentration.Thetotalrateofreactionintheparticle(RvV
p
) is
obtained by equating the total rate to the total diffusiYe flux
at the surface of the particle:
_ 2 D dS I (104)
RvV
p
- -4rrR
p
c dr R
p
dS/ drat the surface beingcalculated by differentiation ofeq.
(103) followed by replacement of r with R
p
The diffusional
resistance in the particle giyes rise to a concentration profile
within the particle, the intraparticle concentration being less
than in thebulkfluid. Hence,theaveragevolumetric reaction
rate is lower in the catalyst compared with a homogeneous
bulk liquid reaction. The ratio of the reaction rate obserwd
in the catalyst (presence ofdiffusional resistance) to the hy-
pothetical rate for the same reaction in the absence of
diffusional mass transferresistance (i.e. liquidphase reaction)
is the catalyst effectiveness factor YJ:
Rate of reaction with mass transfer limitation
YJ = (iOj)
Reaction rate in the absence of mass transfer
limitation
For the example of a first order reaction, the maximum
possible reaction rate (no mass transfer limitation) is
(lC6)
Generalized plots of the effectiYeness factor (11) for any nth
orderreactioncan be found in the literature. Figure26shows
the effectiYeness factor plotted against a Thiele modulus <;>
19
for any particle shape (sphere, cylinder, slab) is
I
<P = V
p
St- )112 (107)
A 2 Dc
p
when n > -1. Figure 26 is for a spherical particle and first
order reaction (n = 1). Clearly, for spherical geometry and
first order reaction the mass transfer limitation is negligible
(i.e., 11 = 1) for 0.3 :s <P (Figure 26).
The effect of mass transfer external to the particle (solid-
liquid interfacial mass transfer) on its effectiwness needs also
to be evaluated. Use of the boundary conditions
dS k
s
at r = R - = - - (5 - 5) (lOS)
p' dr Dc I I
dS
at r = 0-=0 (109)
, dr I
I
,
--
in the mass balance equation (eq. 106) leads to an equation I
which includes the external mass transfer. The mass transfer l
coefficient k
s
is calculated from the well-known correlations,
applicabletoaparticularhydrodynamicregime. For
for stagnant fluid around a spherical particle (i.e. negligible I
difference between the density ofthe particle and that ofthe I
suspending fluid), we haye
1.0 i
0.8
0.6
7](-) 0.4
0.2
,
0.2 2 4 6 8 10
_____Ji
Figure 26. Effectiveness factor (11) ys. Thiele modulus (<I for
I
I
spherical particles and first order reaction.
ksdp
Sh = D = 2.0 (110)
L
whereSh is the Sherwoodnumberandfor forced eonyection,
the Fr6ssling equation
kd .
33
Sh = p = 2.0 + 0.552 Re
o
.
s
Se
O
. (111)
L.
where Re and Scat;i-'.-the Reynolds and theSchmidtnumbers,
respectively. .'/
Seyeral other' equations for k
s
are a\'ailable for particular
applications such as particles in fluidized beds and slurries in
stirred tanks.
)
Retention sieve
......--Coolont
'---
r-++--- Retention sieve
C:::=:=J-./H----- Agitator discs
t
Feed
Figure 34. Bcad mill (vertical chamber).
r
Rotaling disc
Aglta.ors
,Feed
/;
Bearing
seals
...
I
I /
,"--------.- -----.J
I I
L J Coolant
Figure 35. Bead mill (horizontal chamber).
machines tends to fluidize the grinding beads to somedegree
thereby reducing grinding efhciency.
TIle kinetics ofcell disruption in bead mills depend on the
construction of the mill. First-order disruption kinetics haw
beenfound in mJchineswithpredominJntplugflow, whereas
in machines in which the rotor design (Figure 36) permits
signihcant backmixing the disruption de\iJtes from the first
order behJviour. For hrst order bJtch disruption the rate of
protein release by cell rupture is directly proportional to the
amOUnt of unrelc:ased protein:
dF.. . .
dt = kD (R,n - R)
(122)
where Ris the weight ofprotein released per unit weight of
cells, and R
m
is the maximum measurJble release ofprotein.
Equation (122) may be integrated for batch time (t) to give
Fermenter technology
Figure 36. Some rotors for bead mills.
In (. F..
m
.)= In Dj = kDt (123)
R
m
- R
whereD
i
is thereciprocalofthefractionofunreleasedprotein.
For continuous disruption in mills in which the flow may
be described in terms ofthe continuous stirred tankin series
(CSTR) model the disruption kinetics follow the equation
- R
m
- (k ;)iJ
(124) D
i
- . . - [1 + Dt
r
)
R
m
- R
where t
r
is the mean residence time and j is the number of
CSTRsin series.The\"alueofjmaybeobtainedexperimentally
from residence time distribution studies, whereas t
r
is given
by
V
c
t r = Q
(125)
The disruption rate constant, k
o
, is a function of seYeral
parameters:temperature,impellerrotationalspeed,beadload-
ina
0'
bead size and cell concentration. In addition, thedensit),
of bead material is expected to affect k
D
although little has
been written on this subject.
\Vithin limits, the disruption rate constant, k
D
, Increases
with the agitJtOr tip speed, U
T
:
k
D
= KyU, (126)
the practical upper limit on the impeller tip speed being
-15- 16 ms -1. Thepo"erconsumption increases with agita-
tor speed according to
p C NY diS (127)
where Nandd
i
are the rotation speed (rps) and the agitatOr
dise diameter, respectively. The constant c is a function of
the agitator design, suspension viscosity and the density of
suspension. Heat production and the associated cooling re-
quirements increJse with increasing agitJtion as alsodoes the
weJron the beads, agitator and the chJmberWJlls. A certJin
amount of weJr is unavoidable and bead material must be
cJrefully selected in cell disruption applications. Beads Jre
JYJibble in vJrious types of gbss, cerJmics and steels; note
thJt some materiJ]s (leJded gbss, for example) will not be
acceptJblewithproductsintendedforphJrmJceuticJIorfood
applications. The kinetic energy of the beads is an important
Jctor in disruption; it mJy be calculated as
Biotechnology/The Scienceand the Business
DO\VNSTREAM PROCESSING
OPERATIONS
General Considerations
The output of a bioreactor usually undergoes one or more
downstreamseparation,purification,stabilizationandpackag-
ing operations to produce a saleable product. This section
examinessomeofthemorecommonunitoperationsemployed
in downstream bioprocessing.
a.-erallprocessdesigninyolYesconsiderationofinteractions
between various downstream and upstream stages and the
bioreactor. In general, the smaller the number ofprocessing
steps necessary the more attractiye the process.
Solid-Liquid Separations
All bioprocesses involve one or more solid-liquid separation
steps. Following fermentation the biomass may need to be
separated from the broth, or the cell debris may have to be
remoyed following cell disruption. Purification of enzymes
and biochemicals by precipitation and crystallization also
employs solid-liquid separation processes.
The well established separation technologies
suchas sedimentation and filter presses are nottreated in this
section even though they find numerous applications in the
biotechnologyinduStry. Excellenttreatmentsofthesesubjects
f can be found in the chemical engineering literature. Here
\ only the operationsofparticular,interestin bioprocessingare
\ examined.
Centrifugation
Adyances in structural steels have made possible the use of
high-speed, corrosion resistantcentrifuges for large scale bio-
processes. Attention to ease of cleaning, containment and
aerosol suppression, and the availability of sterilizable ma-
chines have made centrifugation a yery important separation
operation in biotechnology industry.
Operational principle
A centrifuge is basically a sedimentation tank with enhanced
graYitational force to increase the rate of sedimentation. A
particle enters the cylindrical bowl of the centrifuge (Figure
27) the feed flowing at some constant yclocity. The
centrifugal force driyes the particle outwards towards the
walls ofthe bowl. A particle initially at radius ri (Figure 27)
would be atsome position r aftertimet. Ifthe residence time
of the liquid in the centrifuge is such that the r 2: rB, the
particle will reach the wall of the centrifuge and it will
sediment. Forparticles in Stokes' law regime (Re 1),the
terminal settling Yclocity at r3dius r is P
!?xiS of rotation
Zs
I
I
LIQUID

I
\
,
\ Path of
'J
, I particle
\
\
\
\

Lrs-!
t
Feed
Bowl wall
Figure 27. Particle motion in the bowl of a centrifuge.
{J}r (Ps - pdd/
U
l
= ( 114)
18
where w is the angular YClocity ofthe centrifuge. Since U
l
==
dr/dt,equation(114) mayberewrittenandintegratedbetween
the limits r = ri at t = 0 and r = rB at t = t
r
(the residence
!
time) to give
l
(115)
I

Note that ri cannot be zero and the feed must enter the"'--- 1
separating zone some distance from the axis of rotation.
The acceptable yolume flow rate through the centrifuge
which will allow the particle of diameter 2: d
p
to separate
may be calculated as
\
T' d'??
= -.!: = w-(Ps - pd p-;r (rB- - rj-) ZB
Q
t
r
18 In (rB/r
j
) (116)
The design of the centrifuge and strength of construction
materials limit the maximum rotational speed.
Performances ofdifferent centrifuges are compared on the
basis ofthe sigma concept which is also the commonly used
basis ofscale-up. The parameter L is defined as

(117)
2 U
l
Because U
l
(eq. 114) is dependent on the geometry of the
centrifuge (yi3 ZB, rn, rj), L (m
2
) is dependent on centrifuge
19
2
piZ0'Z
disruption rate constant is sensitin' to the design of the valve
seat and some designs are shown in Figure 38. While most of
the available disruption data applies to nh'e scats type (a) in
Fi"ure 38, recent work has shown that vain.' scat (c) is
'"
significantly bener than other designs in cell disruption
applic:nions.
From equation (131) it can be seen that the operating
pressure is the major influence on disruption rate. Typically
the operating pressure does not exceed 50- 60 MPa but higher
pressure (- 130 1\lPa) equipment is anilable. An optimal
choice of operating pressure is important because the power
consumption during disruption is a linear function of the
operating pressure, corresponding to about 3.5 kW per 100
MPa of operating pressure. The operating pressure also affects
heat generation: at operating temperatures higher than 40C,
protein denaturation during disruption may occur. Since the
temperature rise across a homogenizer due to adiabatic com-
pression is about 2"e per 10 MPa, inadequate precooling, or
failure to cool between multiple passes, can result in tempera-
tures above 40C and consequent denaturation. Although the
degree of disruption is increased by the number of passes
through the homogenizer, a minimum number is desirable
in practice. Multiple passes not only reduce the machine
throughput but also cause further disintegration of already
broken cell debris \\'hich leads to separation problems further
downstream.
A "microfluidizer" high pressure homogenizer which relics
on complex interactions between multiple liquid jets, cavitation
and impingement has recently become a\ailable. For bacteria
such as E. coli and B"o'Uus Sltbtilis this device gives a perform-
ance similar to the more traditional homogenizers but 95%
breabge of yeast (S. cerevisiae) required 30 passes in one
case compared with a residence time of only 3.3 minutes in a
bead mill.
Other considerations
A unit operation cannot be considered in isolation from the
rest of the process and the overall process must always be
kept in mind. For example, the cell disruption operation
afkcts the physical properties of cell slurry such as viscosity,
density, panicle size and settlability of suspension which in
turn affect the subsequent processing.
(a)
( bl (c) (d)
Flal type Knife edged Cane Iype Groo>ed
Figure 38. Valve seats for high pressure homogenizer.
Fermenter technology
Precipitation
Purification of enzymes and other biologically active proteins
is often a multistep process. A single purification may in\'olyc
cell disruption, debris removal, fractional precipitation, ion-
exchange chromatography, gel filtration, affinity chromato-
graphy and crystallization. The smallest possible number of
separation steps is desirable for reasons of economy. Generally,
no more than seven are used; many industrial processes make
use of relatively crude preparations of enzymes and one or
t\\o purification stages may be sufficient. The purification
scheme may be configured to produce several products from
a gi\'en mixture so reducing the unit COSt of purification.
Selective precipitation of proteins from a solution of se\'eral
proteins is among the oldest of purification and concentration
techniques. Purification factors of 3-10-fold are reiatiYCly
modest compared with chromatographic methods_ Howeycr,
precipitation methods can deal with large quantities of material
which may be quite crude (cell debris, suspended solids,
contaminants). Furthermore, continuous operation can be
cffecti\'e!y employed. Precipitation is encountered frequently
in protein purification schemes, predominantly as one of
the early purification stages. Additional steps may be used'
downstream to polish further the product obtained at the
preCipitation stage.
The process of precipitation converts the soluble protein to
an insoluble form by altering the solute-solvent interactions.
Protein molecules carry positive and negative charges, the net
charge being dependent on the solution pH, At the isoe!ectric
pH the protein molecule carries zero net charge and is least
soluble in a pobr soh'ent such as water. The isoelectric pH of
different proteins is usually different. Hence, by stepwise
variation of pH of a protein solution different protein fractions
may be precipitated and collected but exposure to extreme
pH values may denature proteins and cause the loss of their
biological activities. Protein precipitation by pH variatiOn is
employed in the food industry (e.g. in milk coagulation).
Unwanted, heat-labile protein may be coagulated out of
solution of relatively thermally satable components by heating.
\'\'ater is a strongly polar solvent (with high dielectric
constant = 80 at 20"C) which interacts with the charged
protein molecules to keep them in solution. The addition of
less polar soh-ents such as methanol, ethanol and acetone
(Kj = 21 at 20C) reduces the dielectric constant of the
solution and protein solubility accordingly declines. Fraction-
ation by organic soh'ents is sensitive to temperature, pH,
ionic strength and the presence of other metal ions. Manipu-
lation of these parameters provides flexibility in the selection
of separation conditions. Howeycr, org:1l1ic solYCnts haw a
tendency to denature proteins and temperatures as low as
-lOoe may have to be used to reduce denaturation in a less
polar environment.
Probably the most widely used protein precipitation tech-
nique is salting-out. Addition of salt (either as crystalline
I99
Biotechnology/The Science and the Business
charged and on the mechanism of discharge of solids. All
disc-stack machines (Figure 29) contain a set of conical plates
or discs separated by flow channels. The thin flow channels
mean a reduced "depth" for solids to settle and hence bener
performance." The feed is introduced to the rotating bowl by
a stationary inlet pipe and passes into the zone between the
discs where separation takes place. Under the influence of
centrifugal force, the particles traYeI radially outwards until
they strike one of the conical discs. The particles then slide
down the under side of the disc and are thrown into the
space outside the disc stack. The solids may either accumulate
here (solids-retaining disc centrifuge), or discharge from rhe
bowl continuously (nozzle discharge) or intermittently (solids
ejecting with peripheral or axial discharge).
For further details of these de\ices the literature mentioned
in the bibliography should be consulted.
Decanter centrifuge. The decanter bowl discharge centrifuge
(Figure 29) is suitable for slurries with high solids contents.
Solids are continuously discharged by the helical screw mech-
anism. Only reiatiYely low centrifugal forces are feasible.
Filtration
Rotary vacuum (or pressure) filters
Rotary vacuum drum filters appear to be the most commonly
employed type of filters in biochemical industry, particularly
in antibiotics manufacture and the production of such chem-
icals as citric acid by Aspergillus niger fermentations. Rotary
filters are available either for suction (or vacuum) operation
as in a Buchner funnel or for pressure operation. In the laner
the liquid is forced through the filter by the application of
pressure on the liquid surface. These filters have the advantage
of continuous operation and are useful when sterility and
containment requirements are not stringent.
A rotary vacuum filter consists of a drum frame covered
with filter cloth (canYas, nylon, Dacron, metal or glass fibre).
The internal volume of the drum is divided into radial
chambers (Figure 30) to which vacuum may be applied. The
. drum rotates (0.1-2 rpm) partly submerged in an agitated
trough of the slurry to be filtered. Application of vacuum
(250-500 mm Hg) to the submerged chambers (-30% of
filter area) of the drum results in the slurry being drawn
through the filter cloth; the initial layer of solids deposit acts
as the filter medium. Continuation of suction as the solids-
coated drum surface emerges from the slurry bath leads to
de"'a.tering ",hich may be followed by spray ",ash. Before
the drum re-enters the slurry the solids cake is taken off the
filter surface by a knife scraper (doctor blade). Other solids-
discharge mechanisms may be used such as strings ",hich can
be lifted off the filter surface. In some cases the filter cloth
itself is passed over small diameter rollers to crack the solids
cake which then drops off before the cloth returns to the
....
heads
Cake
recovery
Solids
cake
Scraper
,
I
Figure 30.
1-
drum surface. The type of solids discharge used depends on
the properties of the solids. In the cake remonl stage the
Rotary vacuum filter.
suction is discontinued; in some designs the suction chamber
may be under positive air pressure to assist cake remon!.
When high filtering capacity and no washing are desired,
filter drums with 60 to 70% submerged filter area may be
l
used. Filtration of fine or gelatinous solids which form imper-
meable cakes cannot be carried out effectively with a bare
filter cloth which is readily plugged. For such cases a precoat
filtration scheme employing filter aids can be used. Filtration
of Streptomyces broths often requires filter aids. In this type.
of operation a slurry of a filter aid such as diatomaceous
earth or cellulose fibres is filtered through the filter cloth to
form a porous cake of the filter aid. Subseqciently the broth
is filtered through this cake and as a layer of solids deposits
on the cake it is scraped off together with a thin layer of the
filter aid thereby exposing fresh filtration surface. Suction is
maintained throughout the entire cycle to keep the bulk
the filter aid material firmly attached to the filter drum.
Precoat filtration is usually limited to cases in which the
solid is not the desired product. Filter aid can be difficult to
recover and it may have to be discarded together with the
filtered solids. Many biological products are adsorbed to the
filter aid material and this loss can be significant particularly
when the product is expensive. Laboratory trials are indis-
pensible to satisfactory filtration performance for any new
application. The broth pretreatment conditions can radically
alter its filtration characteristics; changes in pH and tempera-
ture, for example, and the length of holding time at these
conditions lead to large changes in filtration properties of
Streptomyces broths. conditions have to
{be. found experimentally. he flux of filtrate, i.e. the volume
, .
o trate (VL co per unit time (t) per unit filter area
(A
r
), is related to the pressure drop driving force (L'\,P), the
Iviscosity 9f the continuous phase and the flow resistance
\.' (R
F
) by the Hagen-Poiseuille type equation:
I
194
which is usu31ly in aqueous solution (fcrment3tion broth
or liquor), when contacted with an immiscible
(cxtractJ.nt), distributes or pJ.rtitions itself between the nvo
phases, The extent of p3rtitioning is determined br the par-
tition coefficient, k", defined as
(13-t)
'11' b '
For organic acids and bases such as peniCl 111S, enzolc
acid, citric acid, erythromycin, k
p
is strongI)' pH depen
d
ent:
the salt forms of these compounds (e,g, sodium benzoate)
han' preferential solubility in aqueous media, while the acid
(e,g. benzoic acid) or base (erythromycine) forms show higher
solubilities in less polar, organic soh-ents. Variation of pH is
' . f
thus used to alter k
p
for t he desire d extraction. Add Itlon 0
. I f 'd d b d
inorg311lc sa ts, atty an s, etergents, etc., can e use to
manipulate k
p
which is also affected by temperarure. Organic
phases composed of mixtures of two or more organic solvents
have been employed and in these cases k
p
can be altered by
changing the soh-ent composition,
The ratio of the total amount of solute in the two phases is
knO'lvn as the "degree of separation" (G) and depends on the
"olumes of the soh'em and the broth:
G == C,oken{Vsohen, == k Vsohenr (135)
C
brorh
' Vbro,h p VbrOIl,
Equ:ltions (13-t) and (135) are eq uilibri urn rebtionships.
Various types of extraction equipment arc a\'ailable and
haw been discussed in chemical engineering literature, In
the an tibiotic industry centrifugal extractors such as the
Podbidniak extractor arc commonly used for very r3pid ex-
traction, Excessi"e exposure of the antibiotic or other product
to extraction conditions which may be dcliterious (e.g, low
pH in penicillin extraction) is a\'oidcd and because of the
high centrifugal fields the form3tion of stable emulsions is
reduced. Either whole broth or clear liquor may be extracted
but the extraction of \I,hole broth can lead to problems with
blockages due to solids. The extractors consist of se\'er31
perforated concentric shells attached to a central sh3ft which
acts also as the inlet and outlet for the fluid streams. The
sh3it and the shells rotate; the dense liquid is fed to the
innermost shell through the shaft and moves radially outward
under the action of centrifugal force, A light liquid fed at the
peripherr of the outermost shell moves inward counter current
to the flow of heav)' liquid, The continuous circular interface
(major interface) bet\l'een the two liquids lies somewhere
bet,,'ecn the inner and outermost shells and its position can
be controlled by Drying the operating conditions. Selection
of optimal operating conciitions for any new application re-
quires appropriate hboratory trials.
Proteins and other biopolymers which show reason3ble
solubility only in aqueous solutions, or are denatured in
organic soh'ems, rna)' be purified by liq uid-liquid extraction
Fermenter technology
bct""een (\\'0 aqueous pluses. Immiscible aqueous ph3ses for
such applications arc produced by dissoh'ing high concen-
tr3tions of twO different polymers or a polrmer-s3lt combi-
nJ.tion in ""ater. The reading list should be consulted for
additional information on this technique,
Chromatography
Industrial application of chromatography as a separation tech-
nique is a recent phenomenon, but already large scale pro-
duction chromatography h3s pro\'en itself particularly \'3luable
v
for protein purifications. The production of insulin, hormones
and medicinal enzymes makes use of this technique \Ihile
blood plasma fractionation by chromatography is also rapidly
developing, Ver)' complex mixtures can be separated or re-

soh-ed into their components.
_--"
Operating principles
All chromatographic sep3r3tions depend on physicochemical
interactions between the dissoh-ed components of a mixture
and a stationary phase. The latter is usually a solid (or a
liquid supported on a solid) contained in a packed column.
The mixture is applied to the column 3S' a small volume of
solution. The column is then washed or eluted with a soh-ent
which constitutes the mobile phase. Different components of
thl? mixture move down the column at different rates depending
on how strongly a particular component interacts with the
stationary phase. The components which associate strongly
\I,ith the stationary phase flow down the column at a slower
rate than those which do not bind strongly to the solid
packing. As a result of their different velocities the different
components of the mixture separate as they mo\'e down the
column and can be collected as separate components. The
phenomenon is analogous to what happens on a racing track:
aU contestants arc at the starting position in a single line: part
\I'a:-' down the track, however, they have separated and arc at
different distances from the starting posi tion and from each
other because of their different a\'erage speeds.
In industrial biological separations the mobile phase is
alw3)'s a liquid, usually an aqueous solution. Howe,'er, the
nature of the interactions between the mixture components
and the stationary phase can be \'ery different leading to
different types of chromatography: they include affinit), ion
exchange and hydrophobic chromatography as well as gel
filtration. \Vhile all these arc useful in biotechnological separ-
ations, ion exchange is widely used for protein purification
and affinitr chromatography is an especially powerful tech-
nique for biologically active substances. Beginning from the
time 3 mixture is applied to a chromatographic column anc
the flow of eluting soh'ent is surted, a concentration VS, timI
plot of the components emerging from the column can b
plotted as in Figure 41. The peaks correspond to the tWI
components the mixture, A and B. The extent of separatio
20
'--'"
Biotechnology/The Science and the Business
(128)
which indicates high density beads and high agitator tip
speeds (for higher bead velocity, Ub) for good disruption.
The optimum density of the bead material is dependent on
the apparent viscosity of the cell slurry inside the mip because
the viscous drag tends to reduce the bead velocity. The
optimum density of bead material may be estimated using the
equation.
Ps = 1016 + 183 (129)
for 1:s; :s; 20 Pa s. Bead diameter and bead loading are
other important considerations in a disruption operation.
Generally the disruption increases with increasing bead loads
and so does the po"er consumption and the production of
heat. Experience shows that bead loading should be 80 to
95% of the void volume of the disruption cbamber. Lower
loads lead to poor disruption efficiency.
In general, more rapid disruption is achieved with smaller
beads, the optimal bead size being dependent on the size of
the microbial cells being disintegrated: the smaller the cell
size the smaller should be the bead diameter. For fungal
hyphae, for example, bead sizes> 1 mm may be satisfactory.
For animal and plant cells and for yeasts the bead size should
be < 1 mm. The lower practical limit on bead size is about
0.3 mm and for small bacteria the cell disruption on a single
passage through the mill may not produce satisfactory disrup-
tion performance. Typically, two or more passes may be
needed to cause sufficient disruption of bacterial slurries
whereas a single pass may be enough for the larger yeast
cells.
Other factors such as the concentration of cells, the location
of the desired enzyme within a cell and the strength of the
cell wall affect product release by disruption. The cell wall
strength depends on the growth environment and the growth
stage at which the cells are harvested for disruption. Some of
these considerations have been reviewed elsewhere.:!l
The high pressure homogenizer
The high-pressure Manton-Gaulin APV type homogenizer is
among the most widely used liquid shear disruption devices.
The high pressure homogenizer consists of a positive dis-
placement piston pump with one or more plungers. The cell
suspension is drawn through a check valve into the pump
cylinder and, on the pressure stroke, is forced through an
adjust?ble discharge valve (Figure 37) with a restricted orifice.
As the cell slurry passes between the valve and the seat its
velocity increases rapidly to approximately 290 ms-
J
with a
corresponding decrease in pressure so that cavitation bubbles
form. The product velocity decreases again as the suspension
leaves the valve seat area causing the bubble to implode. The
shock energy released together with the associated turbulence
cause the disintegration of cell wall. The impingement of cell
Disrupted
cells
Feed __
Handwheel for valve
positioning
Figure 37. High pressure homogenizer valve assembly. See
reference 21.
slurry on the impact ring (Figure 37) possibly contributes to
disruption. Because both cavitation and impingement pro-
cesses are velocity associated, the pressure drop across the
vahe (i.e. the difference between the operating and the
atmospheric pressure) influences both of them and affects the
rate of disruption. The protein released by disruption depends
on the pressure difference (tlP) and the number of passes
(N ) through the valve as follows
p
dR ..
-d= k
D
tlp (R
m
- R) (130)
N
p
where (R,n R) is the amount of protein remaining to be
released. Equation (130) can be integrated for R = 0 at N
p
=
oand R= Rat N
p
= N
p
, to
In [Rm/(R
m
- R)] = k
D
VJ
p
tl p. (131)
The exponent a in the pressure term in equation (131) depends
both on the microbial cell being disrupted and on its growth
history; cells grown on simple synthetic media are generally
less robust. For Saccharomyces cerevisiae and Escherichia coli,
a values of 2.9 and 2.2, respectively, have been found. Within
limits, the disruption process is independent of the concen-
tration of cells in suspension. The optimum slurry viscosity
and solids concentration ranges tend to be narro""er for the
high-pressure homogenizers than for the bead mills. The
maximum slurry viscosity should normally not exceed 1Pas
for the homogenizer although more viscous material may be
processed in some circumstances. Similarly the maximum
acceptable particle size is about 20 a lower size (- 2
is preferable. The homogenizer is not suited for fungal broths
such those of Aspergilli or for clumps of plant cells. The
I _.
)
!I'ij1 illirmlilii _
In a \Yell packed column the channelling effects (parameter
C) arc independent of flow and arc minimized by using small
spherical panicles with a narrow size distribution as the
packing material. Although theoretical reasoning indicates
that a bed of small panicles would give better separation
efficiency, the operational requirement of acceptable pressure
drop through the bed places lower limits on the size of bed
material. The flow rate through a packed bed of rigid particles
for a giwn pressure drop is calculated using the Kozeny-
Carman equation
_ k, d/ __
UL - . 7
(142)
L fll (1 - (5)-
which applies for the laminar flow which normally occurs in
chromatography columns. This equation is useful because
it indicates the effects of bed midage (s), pressure drop
bed height (L), eluent ,oiscosity (fIr.) and the particle
size (d
p
) on the superficial velocity of the eluting solvent.
The packing particle size (d
p
), the eluent velocity (Ur.) and
the molecular diffusivity (Dr.) of the solute in the eluent can
be combined into a dimensionless reduced ,-clocity (Vr) given
by
v = Uld
p
(143)
r
l
D
For good separation performance Uland d
p
should be selected
to give a reduced velocity in the range 3-10. This criterion,
howeycr, yields quite low values of Uland in commercial
practice the separation performance is often sacrificed to
attain reasonable process throughputs.
Packing material
Selection of the column packing is the most important step in
chromatographic separations. Properties of the components
of the mixture and the operational requirements determine
the most appropriate column packing for the given separa-
tion needs. lvlany of the chromatographic packing materials
used in bioseparations are porous, hydrophilic substances
(cross-linked pol:;acrylamide, agarose, cross-linked dextran,
cellulose) which give rise to deformable particles. This restricts
both the maximum pressure drop th;lt may be used across the
column and the column height to relatively low values to
avoid the bed compressibility problems_ Although packing
materials with improYCd mechanical properties are becoming
available the practical solution to avoiding bed compressibilitY
while maintaining satisfactory separation performance is the
use of stacked columns. These consiSt of several packed sec-
tions, typically no more than 30 cm thick, arranged in series
in a stack to give the necessary column length (Figure 43);
each section of the stack is a self-comained packed bed.
Because the loading capacity (amount of mixture sample that
may be handled in a single batch operation) is dependent on
the amount of packing material, and because of the restrictions
on height of any packed section, the commercial columns
Fermenter technology
Feed

Staged column
pipes
Product
Figure 43. Stacked chromatography column.
tend to have large diameters_ FIO\y distribution, collection
and redistribution of flow between packed sections require
careful specialist design so that separation achieved is not lost
br remixing of components during these transferring oper-
ations. Column chromatography as currently practiced is a
batch separation technique: a batch of mixture is applied to
the column followed by elution "I\-ith a continuous soh-em
flow. HO"'ever, column designs which allow continuous ap-
plication of mixtures are being deYeloped. ,The literature in
the reading list should be consulted for details.
Drying
},!any products of the biochemical industry such as Yaccines,
enzymes, pharmaceuticals, etc, ha,'e to be dehydrated for
preservation. Dry products keep well and are easy to package
and transport. Several types of drying operations are em-
ployed; spray and freeze drying are particularly important
for thermolabile, biologically active products.
23
Biotechnology/The Science and the Business
Spray Drying
Spra)' drying is a method for rapid, continuous, drying of
solutions, emulsions and slurries. Pressure or centrifugal atom-
izers or gas-liquid jets are used to generate a fine spray of
solution droplets which are brought into continuous contact
with hot air in a large chamber (Figure 44). Large droplet
surface area and small droplet size ensure high evaporation
rates so that drying times are but a few seconds. The flow of
air is usually cyclonic. The dimensions of the drier must be
such that the droplets do not reach the walls until sufficiently
dry to prevent sticking and burn-on. A drying chamber tends
to be quite Lirge: 1-10 m in diameter being common. The
dry powder settles to the bottom from where it is removed
either pneumatically or mechanically, or by a combination of
these methods.
Advantages of spray driers are: continuous operation,
powdered product requiring no further size reduction and
rapid drying which leads to good product quality particularly
for heat-labile materials but relatively low thermal efficiency
is a limitation.
Aseptic spray drying equipment is available. All the air
used is filter sterilized and the drying and solids-handling
chambers operate under slight positive pressure. The instal-
lations can be operated leak-tight and are sterilizable. Anti-
biotics such as streptomycine sulfate for direct injection can
be spray dried. Highly heat-labile products like some enzymes
and blood sera can be successfully spray dried. Microorganisms
may be spray-dried for presen'ation and use as SCPo
Air
Drying--.
oir
Cyclone
'-----------''-----.. Product
Figure 44. Schematic of a spray drier.
Freeze Drying
freeze drying is the most gentle of the drying methods. The
material to be dehydrated is frozen and the ice crystals sub-
24
limed by slight warming without thawing. The process may
be carried out at atmospheric pressure or under vacuum and
sublimation assisted with infrared or microwave radiation, or
by contact heating. About 2800 kJ of heat needs to be supplied"
for each kg of ice remo\"ed. Freeze-dried products are easy to
reconstitute (rehydrate) for use; thus, vaccines, blood plasma,
hormones and enzymes are often freeze-dried but a disad-
vantage of the technique is the long processing time.
BIOPROCESS CONTROL
In order to ensure optimal functioning of a bioprocessing
plant se\'eral processing parameters need to be monitored and
controlled. Temperature, pH, product and substrate concen-
trations, dissoh-ed oxygen, and material flows are a few of
those which may have to be followed over time and manipu-
lated in some predetermined way so as to obtain the desired
product yields at minimal cost. ,
Computer-based control systems are increasingly encoun-"-,,, :
tered in biochemical processing plants and operations such as '
in-place cleaning, filling, sterilization sequences are often fully
automated. Control of the biochemical reactor or fermenter
is generally limited to control of pH, temperature and dissoh-ed
oxygen. A typically instrumented fermenter is shown in
figure 45. More extensiw control of fermentation processes
is desirable but it is restricted by two main factors: (i) the
availability of online sensors to measure the biological and
physicochemical parameters needed to follow the progress of
fermentation remains limited; and (ii) our limited ability to
interpret the available information in the context of the bio-
logical system so that the information obtained can be used
as a basis for control. Substantial research effort is underway
in overcoming these limitations.
Sophisticated control of fermentation systems presuppose'
the existence of a mathematical description - or model - 0).--
Inoculum
Anli! oom --+i:>,<J---{
Acid I
Alkc Ii -tJ----H
I
,
indicatorl eonltollel
r-'----l>'<r-_ Ex hou sf
r
lR
rIo." indicalor I
recordu

I
Cooling woterl

processin9
Figure 45. Typi':'ll fcrmenter instrument.ltion.
steam
f'

the process.Thekinetics ofgro,;\th and product in

combination with the physical system characterlstlcs form a
set of equations which constitute the process model. One of
the aims ofcontrol is optimisation ofthe process. III batch or
fed-batch cultures, profiles ofprogressive changes ill environ-
ment:l! conditions (temperature, pH, dissolwd oxygen, sub-
st;ate concentration) arc determined for maximization of
productyield. In continuous op.timisati?n is used
to select an environmental regIme for maxImum blOmass or
product formation.
A series of chaptc'rs in Comprehensh'e Biotechnology (see
further Reading) prm'ide in-depth insight into bioprocess
control and control instrumentation.
THE OVERALL PROCESS
6-Aminopenicillanic Acid
1
Penicillinsarc amongthe mostwidelyused antibiotics.About
I
10,000tonnesofbenzylpenicillin (Penicillin-G) arcproduced
per annum by fermentation with Penicillium chrysogenum.
I
Penicillin-G (PEN-G)itselfis oflimited use and mostofit is
comerted to 6-aminopenicillanic acid (6-APA) which is a
!
,
precursor for the manufacture of semi-synthetic penicillins
!
(Figure 46). Although completechemical synthesis of6-APA
I
is possible, it is economically not feasible on an indus trial
scale and an enzymetic method is commercially used for the
com'ersion of PEN-G to 6-APA. The enzyme responsible
for the transformation is penicillin acylase (Ee 3.5.1.11)
produced by the bacterium Escherichia coli. Several process
varients may be used: (1) E. coli cells grown separately are
killed and brought into contact with a solution ofPEN-G at
37C in a batch slurry reactor. After one use, however, the
cells lose most oftheirenzyme activity due to lysis and have
to be replaced. This is obviously a drawback. (2) Li\'e,
immobilized E. coli cells may be brought in contact with a
continuous flow of PEN-G solution. Either packed bed or
suspension reactor modes may be used. Side reactions and
contamination by substances required to sustain the cells are
some possible problems. (3) The enzyme penicillin acylase
may be extracted from E. coli, immobilized and used in
a continuous reactor. Scheme (l) has certainly been used
commercially and the other two routes probably have been,
too. The immobilized enzyme route is examined further in
the following section.
The overall process schematic flow sheet
23
is depicted in
Figure 47. A high :'ielding strain of E. coli grown in a batch
Stirred tank is used to obtain the enz)'me. The cell slurry is
concentrated by centrifugation and the cells disrupted in a
pressure homogenizer. Theenzyme penicillin acylase is
tsolated by a t'>\'o stage salt precipitation in a low-shear
helical ribbon impeller vessel. The first step precipit:ltes the
nucleicacids which are removed bycentrifugationalong with
Fermenter technology
... .. .. _.. _,
PenicIllin- G
6.Aminopenicillanic acid
.!-R .

// C0
2
H
o
Semisyntheticpenicillins
Figure 46. Semi-synthetic penicillins from benzylpenicillin
(Penicillin-G).
nE. COLI CELLS
rFROM FERMENTER
r-------,
NUCLEiC
CELL
ACID !---.....'CENTRI FUGATleN
DISRUPTION
PRECIPITATION

PEN -G
ENZYME
PURIFICATION a
IMM08111ZATION
I-c--""'-<
ALKALI
BUFFER
------w6 - APA TO RECOVERY
Figure 47. Schematic process flow sheet for production of 6-
aminopenicillanic acid.
thecelldebris.Theenzymeprecipitateundergoessomefurther
separation to a relatively pure form which is immobilized on
amberlite beads. The immobilized catalyst has a half-life of
the order of several months under the normal reactor COI1-
25
Biotechnology/The Science and the Business
ditions of 37"C and pH 6-8, long enough for industrial
purposes.
The kinetics ofthe deacylation reaction are such that it is
strongly inhibited by 6-APA and is also susceptible to some
substrate inhibition. Detailed calculations indicate a packed
bed plug flow reactor for the conversion. However, as the
deacylationreactionproceeds,thepHofthereaction medium
drops and good pHcontrol is necessary to avoid damage to
theproduct, substrate and enzymecatalyst. ControlofpHis
difficult in packed columns and in this reaction is achieved
by the addition of alkali to a buffered reaction medium.
Rapidanduniformmixingis necessarythroughoutthereactor
in order to avoid high local pH values which significantly
reduce the half-life ofthe enzyme. To satisfy the operational
requirement ofgood reactor mixing and at the same time to
approximatethereactionsystemtoa plug-flowtype,a battery
offour continuous stirred tank reactors (CSTRs) in series is
usedforthereaction.TheuseoffourCSTRsisaboutoptimum
and a PEN-G conversion ofmore than 95% is theoretically
possible. Addition of a fifth reaction stage in series would
increase theconversiononlyslightlysince theproductivityof
the catalyst in this stage would be low because it would be
subject to inhibition by the reaction products which reach
their maximum concentration in this last stage. A fifth stage
would, of course, increase the capital and operation costs.
The process plant for even a relatively simple process such
as the production of 6-APA can be surprisingly complex
whenincontinuousoperation.Therequirementsforautomatic
cleaning and catalyst replacement as well as the need to
accommodate the associated flow schemes leads to extensive
.pipework,valves,pumpsandotherancilliarieswhilethewhole
plant has, ofcourse, to be built to hygienic standards.
CONCLUSIONS
This overview of bioprocessing and bioprocess engineering
has naturally been limited to basic considerationsand a
tion only ofprocessingoperations has been presented. While
very many factors combineto influence the final outcome, it
is quiteclearthat the design ofbothequipmentand operating
protocols demands an integrated approach to the process
under consideration.
NQMENCLATURE
.Roman Symbols
A Arrhenius parameter(S-I); parameter ineq. (139)
l
(m
2
s- )
Ab Free area for flow between riser and downcomer
(m
2
)
Ad Cross sectional area ofdowncomer (m
2
)
B
b
C
6,C
C':'
Cbro,h
C
crit
C
G
C
Gi
C
i
C
L
CLi
C
p
C
s
CsoIvcnt
Ct'
C'X
c
D
De
Dr
Dj
D
L
d
dB
de
d
ep
dj
d
T
E
6,E
E
Lr
e
G
g
H
,0 ..
i
;
-I
Filter area (m
2
)
I
Heat transfer area (m
2
)
1
Arrhenius parameter for nutrient denaturation
(S-I)
Surface area ofpanicle (m
2
)
Cross sectional area of riser (m
2
)
Parameter in eq. 63 m-Is)b) .
Gas-liquid interfacial area per unit dispersion
volume (m-
I
)
Gas-liquid interfacial area per unit liquid volume

Parameter in eq. (139) (s)
Parameter in eqs 63 and 67 (-)
Parameter in eq. 139 (m)
Concentration difference (kg m-J)
Saturation or equilibrium concentration (kg m-J)
Concentration in broth (kg m-J)
Critical oxygen concentration (kg m-J)
Concentration in bulk gas (kg m-J)
Interfacial concentration on gas side (kg m-J)
Concentration of species i (kg m-J)
Concentration in bulk liquid (kg m-J)
Interfacial concentration on liquid side (kg m-J)
Specific heat capacity (] kg-
I
OC-
I
)
Concentration of solids (dry wt./vol.%
(g/lOO mL))
Concentration in solvent (kg m-
J
)
Tracer concentration (kg m-
J
)
Equilibrium tracer concentration (kg m-3)
Constant (eq. 64, 127) (as appropriate)
l
Diffusivity (m
2
s- )
Effective diffusivity in catalyst (m
2
s-I)
Reciprocal offraction of unreleased protein (-)
Impeller diameter (m)
l
Diffusivity in liquid (m
2
s- ) 0'--../
Diameter (m)
Sauter mean bubble diameter (m)
Column diameter (m)
Connecting pipe diameter (m)
Diameter of the ith bubble (m)
Tank diameter (m)
Total amount of enzyme in reactor (kg)
Activation energy (] kmol-
I
)
Liquid phase axial dispersion coefficient in riser
(m
2
s-
l
)
Actintion energy for nutrient denaturation
(] kmol-
I
)
Enzyme concentration (kg m-J)
Degree of separation (eq. 135) (-)
Gra\Oitational acceleration (ms-
2
)
Height equivalent of a theoretical plate (HETP)
(m)
Henis law constant (-)
H for component A (m)
206

'1
Jvlinimum HETP (m)
Hnl
Dispersion height (m)
h
D
Heat transfer coefficient (inside film)
h,
(J S-l
m
-2C-
1
)
Heat transfer coefficient (inside fouling) 'J
h
" (J s-
l
m-
2
C-
I
)
Unaerated liquid height (m)
h
L
Heat transfer coefficient (metal wall)
hill
. (] S-l
m
-
2
C-
I
)
Heat transfer coefficient (outside film)
h
o
(] s-
l
m-
2
C-
1
)
Heat transfer coefficient (outside fouling)
hot
(] S-l
m
-
2
C-
I
)
Ionic strength (-)
I
FI (k -., -I 3 -., -I)
ux 'g m -s or m m -s
J
2
O;..:ygen flux (kg m- s-
l
)
J0
2
Number of CSTRs in series (-)
J
K
Consistency index (Pa.s
n
)
K'
Constant (eq. 56) (-)
Frictional loss coefficient (bottom) (-)
K
B
Kc
Casson viscosity (Pa.s)
K
L
O\'erall mass transfer coefficient based on liquid
film (ms-
I
)
Saturation constant, Michaelis constant (kg m-3)
K
s
Salting-out constant (eq, 133) (kg m-
3
)
K
SL
K
T
Frictional loss coefficient (top) (-)
K
r
Constant (eq. 126) (m-
I
)

Dielectric constant (-)
k Mass transfer coefficient (ms-
I
)
k
D
Disruption rate constant (S-l)
k
J
Specific death rate (S-I)
k
t
Filter constant (m-
I
)
kG Gas film mass transfer coefficient (ms-
I
)
k Constant (eq. 54) (-)
k
I
L
Liquid film mass transfer coefficient (ms-
I
)
k
Specific nutrient denaturation rate (S-I)
ll
k
p Partition coefficient (eq. 134) (-)
k
r Rate constant (as appropriate)
k
s Solid-liquid mass transfer coefficient (ms-
I
)
k
T
Thermal conducti\ity (\'\I m-IOC-
I
)
k,
Constant (eq. 142) (-)
k, -
3 Constants (eq. 113) (as appropriate)
L
Length (m)
L(p
Connecting pipe length (m) . S'!
M
Mass flow rate of gas (kg m-3).......
m
Mass of bead (kg)
N
Microbial concentration (m-3 or kg m-3)
N
Impeller speed (rps)
N
A Number of theoretical plates for component A
(-)
N
0 Initial microbial concentration (m-3)
N
p Number of passes (-)
n
Flow behaviour index (-)
n,
Number of bubbles with diameter d
i
(-)
P



P
G
Po

Q
QH
R
R
R(
Re
Rei
R
f

R
m
R
p
R
s
R...
R"
r
rB
ri
So
Sp
T

To
T,
t
tA.B
t
c
t(ool
td
t
c
tG
t
g
th.:-.u
tholJ
to
top
t
r
U
Pow"
Pressure drop (Pa) --
Frictional pressure drop (downcomer) (Pa)
Frictional pressure drop (riser) (Pa)
Gassed power (W Ee.'"'I (,,)-1 jYJ)
Power number (-r- \"""vJ' <I
Transmembrane pressure (Pa) .
l
Volume flow rate (m
3
s- )
Heat transfer rate (J S-I)
Specific oxygen consumption rate (kg 02/kg
cellsecond)
Gas constant (J K-
I
kmol-
I
)
Protein concentration (kg m-3)
Resistance due to solids cake (m-
I
)
Reynolds number (-)
Impeller Reynolds number (-)
Total flow resistance (m-
I
)
Resistance of membrane (m-1)
Maximum releasable protein (kg m-3)
Particle radius (m)
Rate ofsubstrate consumption (kg S-I) .
Volumetric reaction rate in catalyst (kg S-lm-3)
Resolution (-)
Radial distance (m)
Radius of bowl (m)
Initial radial position (m)
Resistance of filter medium (m-
I
)
Rate of surface renewal (S-I)
Substrate concentration (kg m-3)
Schmidt number (-)
Sherwood number (eq. 110) (-)
Concentration in bulk liquid orconcentration in
liquid phase at solid/liquid interface (kg m-
3
)
Initial substrate concentration (kg m-3)
Protein solubility (kg m-3)
Absolute temperature (K)
Temperature difference (OC)
Initial temperature (oq
Sterilization temperature (OC)
Time (s)
Retention times of peaks A and B (s)
Circulation time (s)
Cooling time (s)
Mean doubling time (s)
Exposure time (s)
Gas phase residence time in liquid (s)
Mean generation time (s)
Heating time (5)
Holding time (s)
Zero time (s)
Operation time (s)
Residence time (s)
Average flow velocity (ms-
t
)
Bead velocity (ms-
I
)
2C
I
Biotechnology/The Science andthe Business
U
H
!lL
U
Lc
U
Ld
U
Lm
U
Lr
U
m
.,X
U
G
U
Gr
U
T
u,
V
broth
V
c
V
G
"h
YL
YLd
V
Lr
V
p
V
r
/
V.oJ-m,
w
\VIA, W
B
X
X
X
o
X
Xj
X
o
Xc
Z
ZB
Zj
Overall heat transfer coefficient (W m-
2o
C-
I
)
Fluid viscosity at wall temperature (Pa.s)
Liquid velocity (ms-
I
)
:Maximum specific growth rate (S-I)
Mean liquid circulation velocity (ms-
I
)
!J.N Specific cell number growth rate (S-I)
Superficial liquid velocity in downcomer (ms-
I
)
U
L
at H
m
(ms-
I
)

PL
Suspension viscosity (Pa.s)
Liquid density (kg m-
3
)
Superficial liquid velocit) in riser (ms-
I
)
Ps Density of solids (kg m-3)
Maximum velocity (centreline velocity) (ms-
I
)
L Sigma factor (eq. 117) (m
2
); summation
Superficial gas velocity (ms-
1
)
o Interfacial tension (N m-I)
Superficial gas velocity based on riser (ms-
I
) L Shear stress (Pa)
Tip speed (ms-
I
)
Terminal settling velocity (ms-
I
)
Volume of broth (m
3
)
Chamber volume (m
3
)
Volume of gas in dispersion (m
3
)
Liquid "olume (m
3
) .
I.nterstitialliquid velocity (ms-
I
)
YL in downcomer (ms-
I
)
V
L
in riser (ms-
I
)
Volume ofparticle (m
3
)
Reduced velocity (eq. 143) (ms-
I
)
Volume of soh-ent (m
3
)
Dryweight of accumulated biomass (kg)
Basal width ofpeaks A and B (s)
Solids concentration (dry) (kg m-
3
)
Conversion (-)
Initial concentration ofsolids (dry) (kg m-
Mass fraction of oxygen (-)
Mass fraction ofoxygen (inlet) (-)
Mass fraction of oxygen (outlet) (-)
Circulation path length (m)
Filter depth, axial distance (m)
Bowl length (m)
Charge on species i (-)
Greek Symbols
ex Parameter (eq. 84) (ms-
I
)
ex Mean specific cake resistance (m kg-I)
Parameter (eq. 84) (-); constant (eq. 133)
(kg m-
3
)
y Shear rate (S-I)
o Film thickness (m)
0G Gas film thickness (m)
0L Liquid film thickness (m)
Os Thickness of solids layer (m)
E O\wall gas holdup (-)
Ed Downcomer gas holdup (-)
E
r
Riser gas holdup (-)
ES Void fraction of solids (-)
11 Effectiveness factor (-)
Specific gro'wth rate (S-I)
Apparent viscosity (Pa.s)
!J.L Liquid viscosity (Pa.s)
La Yield stress (Pa)
<p Thiele modulus (-)
1jJ Constant kL/d
B
(eq. 88) (S-I)
W Angular velocity
Abbreviations
6-APA 6-Aminopenicillanic acid
CMC Carboxymethyl cellulose
CSTR Continuous stirred tank reactor
DOP Dioctylpathalate
HETP Height equivalent of a theoretical plate

I
,
,
!
.I
I
I
,
I
I
j
J
..
HTST High temperature and short time sterilization
PEN-G Penicillin-G (benzylpenicillin)
SCP Single cell proteins
3
)
REFERENCES
1. Coulson, J.M. and Richardson, J.F. (1977). Chemical Engin-
eering, vol. 1 (3rd cd.) Oxford: Pergamon Press.
2. Bailey, J.E. and Ollis, D.F. (1977). Biochemical Engineering
Fundamentals, New York: McGraw Hill.
3. Fair, J.R., Lambright, A.J. and Andersen, J.W. (1962). Heat
transfer and gas holdup in a sparged contactor. Ind. Eng.
Chern. Process Des. Develop., 1,33-36.
t
f
4. Deckwer, W.-D. (1985). Bubble column reactors. In Biotech
nology, vol. 2, p. 445. Ed. by H.-J. Rehm and G.
Weinheim: VCH. 1
5. Chisti, M.Y. and Moo-Young, M. (1987). Airlift reactors: t
Characteristics, applications and design considerations. Chem. t
Eng. Commun., 60, 195-242. ,
6. Mann, R. (1983). Gas-Liquid Contacting in Mixing Vessels, f
Rugby: Institution of Chemical Engineers. t
Akita, K. and Yoshida, F. (1973). Gas holdup and volumetric f
mass transfer coefficient in bubble columns. Ind. Eng. Chern. t
Process. Des. De\elop., 12, 76-80. l
8. Godbole,S.P., Schumpe, A., Shah,Y.T. andCarr,N.L(1984). }
Hydrodynamics and mass transfer in non-Ne'1iJtonian solutions !.
in a bubble column. AIChEJ., 30, 213-220.
9. Nishikawa, Kato, M. and Hashimoto, K. (1977). Heat
transfer in aerated tower filled with non-Newtonian liquid.
Ind. Eng. Chem. Process Des. Develop., 16, 133-137.
Chisti, M.Y. and Moo-Young, M. (1988). Gas holdup in
pneumatic reuctors. Chem. Eng. J., 38, 149-152.
Chisti, M.Y. and 1100-Young, M. (1988). Hydrodynamics
andoxygentrunsferinpneumaticbioreactordevices.Biotechnol. ,-<
Bioeng., 31, 487-494.
12. Chisri, M.Y. (1989). Airhft Bioreuctors, London: Elsevier
Applied Science.
208
_
/
hir,1.R. (1967). Dcsiglling g,L'-;?:/rze,{ re,letors. Chcl11.
--J3.
7-1 3), p. 67.
Dcckwer,\V.-D., h.,Schull1pc,A.andScTpCI11Cn,
14. Y. (1932). Ox)'gcn 1I1.15S tr,'''s}<.'r into "cr"ted CMC so/utiom in
a babble column. Biorcchnol. Bioeng., 2-1, 461-481.
15. Schul11!'e, A. and Dcck\\'cl, \\.-D. (1987). Viscous media in
to:.:..'cr biorclzC{ors: hydrod)l1.l?n:c chdhictcristics and 'rh155 (r...tl1S-
fer properties. Bioprocess 2, 79-94.
16. 1\loo-Young,;\1. and KJ\\'ase, l". (1987). Biorcc,etor d('sign for
s!ctrryfcnllcnt,ztion sptcms. IIIHorizonsoil3io.:hcl11iLcll
eering. p. 281, Ed. b
1
S. AibJ. Tokyo: Uniwrsiry of Tokyo
Press.
17. 1\kr.:huk,].C and Siegel, i\I.H. (1988). Air-/,ft /C<letors in
cbemic,d and biological tec!n;olog1'.]. Chcm.Tech. Biorechno!.,
-II. 105-J20.
IS. Chisri, }.I.Y., Habrd, B. ;lnd i\!oo-l"oung, }.!. (1988). Liquid
circulation in airhft reactors. Chem. Eng. SLi., 43, 451-457:
19.
Chisri, Y. and l\loo-Young. 1\1. (1988). Pri,dwion of liquid
circulation velocity in airlift re,lcors ",,it!; biological medi". ].
Chem. Tech. Biorec'hnol., 42. 211--2J9.
10. F!.lschel,E.,\Vandrey,C and hula,l\!.-R.(1983). Ultr,,!,ltr,'tion
for tbe separation of bioutalysts. Ad,. Biochemical
26, 73-142.
21. Chisri, Y. and 1\1. (1936). Dismption of microbi,z/
cells for intr,'cell"I'/r products. Enzyme \linob. Techno!., 8,
J94- 204.
I3dl, OJ.,Ho,lfc, },[ and Dunnill. P. (19S3). Tb" foml,ttion oj
protein precipit,lIL'S and their ce':tnfugal Ad\'. Bio-
chemical Engineering, 26, 1-72.
23. Chisri, M.Y. (1932). A riC:';' plUcess for the prod:{("tion of 6-
all1inopcnicilldnic acid fmm PO/icillill-G. The Polytechnic
Ib:1d,ln Journal, 1(1), SS92.
FURTHER READING
I. Reierences 2, 4-6, 12 and 20-22.
2. B:liley,].L (19S0). BlOchemic,z/ ""lction engmecrina and bio-
, . I h
C!}emlC<1 reClclors. C cm. Eng. Sci., 35, 1854-1886.
0
3. Blursrrom, LE. (19S5). BlOteefmoIogy: Ferment,Uion and
do:.;mst:mm processing. Chem. Eng. (IS february), pp.
126-b3.
4. Brown, D.E. and P.R. (1987). Cross-jlo:.;' septlriltion
Fermenter technology
of cells. Process Biochemistry (August), pp. 96-101.
5. Erickson, R.A. (I%4). Disk cemrifuge) in bio[cc!Jlz()iogv.
Chemid Engineering Progress, 8: (December), pp. 51- 54.
6. hechrcr, A. (Ediror, 19S2). A,h-. in Bioebcmicill Engil!et'ring,
25.Berlin: Springer-Verlag.
7. Karel, S.F., Libicki, S.B. and Roberrson, C.R. (1935). Thc
immobilzz,ltio'l of :';'hole ails: engincering principles.
Eng. Sci., -10, 132J-1354. .
8. Kesha"araz, E., Hoare, j\1. and Dunnill, P. (1937). Biocbeniml
engineering a;pects of cell dis rup tioII. In Sep'lrarions ior Bio-
rechnology Verrall and 111.]. Hudson, Eds.). Chichesrer:
Ellis Horwood. pp. 62-79.
9. Konecny,]. (1977). T!Jeoretic,zI alld pr,'etic,z/ "speCiS of im-
mobilized enz/,mes. In Surn!\' of Progress in ChemisnT, 8
(A.F. Scorr, Ed.). Ne'" Press: pp. 195-251.
Ie. Kroner, K.H., Nissinen, V. and Ziegler, H. (19S7). Impro:'e'!
dynamic fib,aion of miaobiu! suspel15iol1s. Bio/Technology, 5
(Seprember), pp. 921- 926.
11. Mackay, D. and Salusbury, T. (1988). Choosing bt'to:xel1 cen-
trifug'1tioll and aossf!o:.;' microfiltrdtiorl. The Chemical Engin-
eer, April. pp, 45-50.
12. Moo-Young,]I,I. (EditOr, 1935). Comprehensive Biotec!m%gy,
vo!. 2. Oxford: Pergamon Press.
13. Moo-Young,}.l. and Chisri, Y. (19S9). Considerations for
designing bioreactors for she.1r-sensiti';e wlture. Bioi
Technolog)', 6 (No"ember), pp. 1291-1296.
H. OIJshuc,].Y. (1939). Fluid mixing in 1989. Chcm. Prog-
ress, 85 (l\lay), pp. 33-42.
15. Ripperger, S. (J93S). Engineering aspects and applications of
aossf!o:.; miaofiltration. Chem. TCLhno!., 11, 17-2j.
16. K., Lucke, ]. and Oe!s, U. (1977). Buf,blc cu!:"nn
biore'ICiors. Ad,. Biochemical 7, J -84.
17. SiegelI, ].H., Dupre, G.D. and Pirkle, Jr., J.e. (19S6).
Chromatographic sep,/mtions in a cross-f!o':.;' MSB. Chcm.
Progrcss, 82 (Noycmber), pp. 57-61.
1S. Turunjian, R.S. (1985). Scale-Ifp consider.1tions for memf,r,mc
processes. Bio/Technology, 3 (jld:), pp. 615- 626.
19. Webb, e., Black, G..l\l. and Arkinson, B. (Edirors, 1936).
Process Engineering Aspats of Immobilized Cell Systems,
Rugby: Instirurion of Chemical Engineers.
20. Whirl', ]l,I.D. and Marcus, D. (1938). Disintegration of miao-
organisms. In Adnnces in Biorechnological Processes
(A. Mizrahi, Ed.). New York: Alan R. Liss, Inc. pp. 51-96.
2
I
.
.) [- C) J
/2
-fb
-'i - q!
? - 4
--, J/