h1 PDF

Downstream Processing
in Biotechnology
J.A. Wesselingh and
J. Krijgsman
in Biotechnology
in Biotechnology
J.A. Wesselingh
Emeritus, Department of Chemical Engineering,
University of Groningen
J. Krijgsman
Former Principal Scientist
Gist-Brocades, Delft (now DSM)
@Delft Academic Press / VSSD

First edition 2013
Published by Delft Academic Press / VSSD

Leeghwaterstraat 32, 2628 CA Delft, The Netherlands
tel. +31 15 27 82124, e-mail: hlf@vssd.nl
Publishers website: http://www.vssd.nl/hlf
About this book: http://www.vssd.nl/hlf/d030.htm
All rights reserved. No part of this publication may be reproduced, stored
in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise, without
the prior written permission of the publisher.
Printed version:
ISBN 978-90-6562-318-8
Electronic version:
ISBN 978-90-6562-319-5
NUR 952
http://www.vssd.nl/hlf/d030.htm
Key words: biotechnology
Contents
Front
Contents
Preface
iii
Symbols
Ferment
Release
Clarify
19
Concentrate (1)
35
Concentrate (2)
47
Extract
59
Crystallize
71
Distill
81
Purify (1)
95
10
Purify (2)
107
11
Process
121
Derivations
129
Answers
137
Index
143
Lessons
Back
Preface
The exciting part of biotechnology is thought to be bio. That is the place
for novel ideas, funding, research, patents However technology comes
behind that - without technology, no idea will make it to the mar-ket. An
important part of technology is downstream processing: the sepa-ration
and purification of the bio-product. This is usually the most expen-sive
part of a bio-project, and it can require much ingenuity and a huge effort to
develop a process that is clean and economic.
The development of a downstream process starts in the lab, but ends with a
plant. The plant may have to produce a million times the amounts made in
the lab. The challenge is to end with a good plant. To do this two groups of
people have to work together. These are the lab people micro-biologists,
biochemists and the process engineers. We have tried to write a book
that introduces the subject to both of you. In eleven lessons it describes all
the common steps used in downstream processing, starting in the lab and
ending in the plant. The idea is a course of perhaps one or two weeks,
depending on your starting knowledge. The level is under-graduate
engineering, perhaps graduate biochemistry. Engineers will find the
modeling a bit simple, biochemists the description of experi-ments rather
basic.
At the end of each lesson you will find a small number of exercises. Use
these to find out whether you have absorbed the lesson material. Brief
answers are provided at the end of the book. You can download fully
worked out exercises from www.vssd.nl/hlf/d030.htm in the form of pdf
files. If you have access to the program Mathcad Prime 2.0 (or later) you
can download the originals and play with them.
In the first version of a technical book it is difficult to avoid errors.
(Believe us, we have tried!). We would be pleased to hear mistakes and
points that are not clear via www.vssd.nl/hlf/.
This is the place to start thanking people, starting with those of you who
are going to react. Then the people who have helped us to build up our
understanding of the subject: the many students that we have had on
Downstream Processing and the colleagues of John at Gist Brocades and
DSM. It is now time to begin. On to lesson 1
Hans (J.A.) Wesselingh
John Krijgsman
LLL
Symbols
A
c
area
concentration
m2
kg m-3
C
CR
concentration
cake ratio
specific heat
kg m-3
J kg-1 oC-1
d
D
diameter (small, particle)

diameter (large)
distribution coefficient
m
m
-
E
f
g
slope of equilibrium line

extraction factor
fraction
gravitational acceleration
m s-2
h
H
j
k
K
l
m
mass enthalpy
molar enthalpy
(volume) flux
empirical constant
mole fraction distribution coefficient
length
mass
J kg-1
J mol-1
m s-1
depends
m
kg
m
M
mass flow
kg s-1
molar mass
(molar) amount
kg mol-1
mol
molar flow
mol s-1
pressure
heat flow
Pa
W
n
n
p
radius
R
RM
rC
S
release, removal, retention

medium resistance
cake resistance per height
settling ratio
T
t
temperature
time
velocity across, particle velocity
m s-1
unknown flow
volume
volume flow
m-2
C
s
m3
m3 s-1
Y
velocity along,
m s-1
concentration velocity
work flow (power)
m s-1
W
mole fraction in liquid

mole fraction in vapor
height, small
height, large
W
x
y
z
Z
Greek Symbols
H
I
concentration factor
change
void fraction in particle
void fraction
K
U
viscosity
density
Pa s
kg m-3
D
'
m2
s-1
6 sigma of a centrifuge
Z angular velocity
Subscripts
0
0,1
0,1,2..
initial
inlet, outlet of a single stage
A,B
stage number
components
cake, concentrate, cycle
E
F
G
L
solvent (extractant)
feed
gel
liquid
maximum
permeate
solid, saturation
water
M
P
Superscript
'
other phase
YL
1 Ferment
You have managed to get micro-organisms to make the right product in a
200 mL laboratory flask. So the job is finished or is it? No! You still
have to get the product out of the cells, to remove cell debris and other
contaminants, to concentrate, purify and test the product. Then you have
to scale up to production. Production has to be safe and economic; also it
should not produce too much waste. Finally, you will need to package the
product and to get it sold. You will have to keep in touch with your customers, and improve the product. There is far more work ahead, most of it
outside the laboratory...
Develop a Product
To see where we are, and what this course is about, let us stand back for a
moment. The subject is part of the development of a product (here a bioproduct) that is to help a user or customer. Such a development is a huge
effort, involving many people and large amounts of resources. The steps
in product development are shown in Figure 1-1.
Figure 1-1 Product development cycle
Product development is usually cyclic you cannot really say where it

begins. Even so, it is commonly taken to begin with an idea. This might
come from the market (market pull) or from the developer (market
push). The first steps are exploring or defining briefly looking at, and
evaluating many possibilities. For the kind of products we consider here
(chemicals, food, pharmaceuticals) this leads to experimentation in the
lab (small scale). A project then needs to become more focused, as lesspromising routes are discarded before the strict develop stage. This is
where the number of people involved becomes large as do the costs.
This step usually requires a pilot plant a small scale version of the large
plant. At some point, we will start to organise large scale production: to
find existing plant or to construct new plant. We begin to produce. While
we are trying out production, we are already starting to develop a market
a difficult part in many projects. Hopefully sales will take off and we
will be able to get back our investment. Even at this point, we cannot sit

'RZQVWUHDP3URFHVVLQJLQ%LRWHFKQRORJ\
back. We will have to keep on improving the product, the manufacturing
process and our marketing. The cycle should keep on turning with new
projects
Penicillin and insulin are products where the development has gone
through a large number of cycles, during several decades. These cycles
often involve the introduction of better micro-organisms that improve the
yield or quality of the product. These are first tested in the lab, then in the
pilot plant, and finally in production.
This book does not cover the whole of product development. It focuses on
the develop and produce blocks in our scheme (Figure 1-2). In each
block you will see the sub-blocks ferment, separate and formulate.
The fermenter contains the micro-organisms making the product. It is the
core of the bio-process. Even so, we only discuss it in so far as it interacts
with the separate step. Here we separate the product from the many contaminants that it what this book is about. This is downstream from the
fermenter hence the name. The formulate step is to bring the product
in the form required by the customer as a liquid, powder, granulate or
whatever, with its packaging. We do not discuss formulation in this book.
Figure 1-2 Subdivision of develop and produce
Develop has connections with idea formation and full scale production.
Forming ideas is often the world of laboratory people such as microbiologists and biochemists full scale production that of process operators and
engineers. The book has to bridge these worlds. The microbiologist may
find our treatment of microbiology very simple and that of engineering
daunting. On the other hand, some process engineers may find our treatment of engineering simple (Just in case you wonder: your authors are
process engineers, but they have extensive experience in biotechnology.)
Ferment (using Cells)

Bio-products are made inside the cells of micro-organisms. These are usually kept as a suspension (broth) in a fermenter or bio-reactor (Figure 13). It is quite difficult to process the products from a broth. The reasons
are:
1. The product may be excreted by the cells to give an extra-cellular
product. This is the case for a bio-product such as citric acid, antibiotics and enzymes. As we shall see, this is a desirable situation.
However, not all products are excreted for example many proteins
)HUPHQW
Figure 1-3 Contents of a fermenter
2.
3.
4.
5.
are retained in the cell. These intracellular products can be very valuable, but they have to be released from the cells before further processing.
The broth contains a lot of water, both inside and outside the cells.
This might be 80% in the production of a bulk chemical such as ethanol, or up to 95% for pharmaceutical proteins. All this water usually
has to be removed.
The broth will contain a large number of components. Some of these
may closely resemble the product. The product may be less than one
per cent of the broth. Even so, you may need a high purification up to
99.9999% for some applications in pharmaceuticals!
The broth can be viscous, especially if cell and product concentrations
are high. This makes for difficult handling in the first stages of the separation process.
Finally, we are working with live material: organisms that are continuously adapting and evolving. So fermentations are variable we have
Process Steps
The characteristics above determine the usual sequence of downstream
process steps (Figure 1-4):
Figure 1-4 Sequence of process steps
x
x
x
x
Intracellular products are released from the cells (lesson 2)

Cells and debris are separated to clarify the broth (lesson 3).
The product is concentrated by removing water (lessons 4 and 5).
Further separations then follow to purify the product (lessons 6 to 10).
Bio processes form large flows of waste water. You can reduce these
waste flows and also product losses by a good choice of the separations and their sequence. You will often have to recycle solvent
streams. We discuss recycling and the whole of sequencing in lesson 11.
The development of each step usually starts with lab experiments by lab
people, not by process people. These experiments form the base of all development work, and the scaling up to and construction of large installations. So it is important that also lab people have an idea of how their experiments will be used by process engineers and that the process engineers have some idea of experimentation.
Each separation step requires equipment. We will show the important
pieces, say a bit about their sizing and operation, and discuss their use of
energy and chemicals. We often need many separation steps. There are
losses in each step, and these accumulate rapidly. Suppose that each step
gives a loss of five percent (which is not much). The yield of the process
is then:
0.951 = 0.95 after one step
0.952 = 0.90 after two steps
0.953 = 0.86 after three steps
0.9510 = 0.60 after ten steps.

So we need to design and operate carefully to get a reasonable yield.
A Warning
We end this introduction with a warning. This book uses models (and
simple equations) to help you understand how different pieces of equipment work. You might get the impression that these would allow you to
design downstream processes from scratch. This would be a mistake: the
models help to interpret and extend experiments, but on their own they are
seldom sufficiently accurate. You cannot design a downstream process for
a new product without experimenting.
Further Reading
J.A. Wesselingh, Sren Kiil and Martin E. Vigild Design & Development
of Biological, Chemical, Food and Pharmaceutical Products, John Wiley
& Sons 2007.
Pauline M. Doran Bioprocess Engineering Principles, 2nd ed., Academic
Press 2012
)HUPHQW
Exercises
After each lesson you will find a few exercises. Most of these are small
and meant to get you to think about what you have just read.
1-1 Fermenter Contents
Yeast cells are almost spherical. They have a diameter d = 10 Pm. The
number of cells in a suspension is N = 5108 mL-1. The volume fraction of
water in the cell is HW = 70% and the densities of water and the solids in
the cell are UW = 1000 kg m-3 and US = 1400 kg m-3.
Calculate the volume fraction occupied by the cells and the fraction of
solids (dry mass fraction) in the suspension.
Hint: calculate everything for one cubic metre.
1-2 Yield and Profit
A bio process has the following steps with their yields on the most important raw material:
fermentation
0.85
product release 0.95
filtration
0.95
concentration
0.95
purification (1) 0.95
purification (2) 0.95
The initial production after losses is 1000 ton per year; this is sold for 10
million Euros per year. The cost of the process is 8 million Euros per year
giving a profit of 2 million Euros per year.
(a) Calculate the overall yield of the process.
(b) Assume that you can shave off 0.01 of each of the losses (so that the
yields become 0.86, 0.96, 0.96 and so on.) Calculate the new process yield.
(c) By how much will your production increase, assuming that you process the same amount of raw material?
(d) If you can sell the increase for the same price (a big if!), by how
much will your profit increase? By which percentage?
2 Release
Bio-products are formed inside cells. They are often released naturally to
the surrounding liquid. If this does not happen, geneticists may be able to
modify the cell so that it does release the product. Separation of whole
cells from a suspension is relatively easy and the whole cells can sometimes be put to use. So extracellular products are preferred. However,
many products are retained in the cell; especially those obtained by genetical manipulation, for example in E.coli. These intracellular products have
to be released for further processing. Before considering the different
methods for doing that, we first look briefly at the structure of cells.
Micro-organisms
Two important groups of micro-organisms used in biotechnology are the
bacteria and the yeasts (Figure 2-1).
2-1 Micro-organisms and their walls
Bacteria have dimensions of the order of a micrometre. Their walls are

thin about ten nanometres thick and often fragile. The two main
groups are those of the Gram-positive and the Gram-negative bacteria.
(The name Gram is from the Danish scientist who made this classification, it has nothing to do with the mass unit gram.) In the Gram-positive
bacteria, the cell wall consists of two layers: the inner one consisting
mainly of phospholipids, with a thin outer coating of protein or polysaccharide. The walls of the Gram-negative bacteria have two phospholipid
layers. The oily liquid films formed by phospholipids are hydrophobic, so
they form a barrier for water. They can be disrupted by surfactants.
Yeasts are larger than bacteria: they might have diameters up to ten micrometres. The walls are much thicker typically about a hundred nanometres and contain layers that mainly consist of the polysaccharides
glucan and mannan. As a result the walls are stronger than those of bacteria and more difficult to disrupt. Polysaccharide and protein layers can be
weakened or disrupted by enzymes.
Most intracellular products are present in the cytoplasm. However, some
interesting products are contained in organelles such as ribosomes or the
mitochondria that are found inside larger micro-organisms such as yeasts.
Ultrasonic Release
In the lab many techniques are used to release products. Most of them are
not used in production, so we do not discuss them. One exception is ultrasonic cell disruption. The equipment for this popular technique looks like
a small kitchen mixer. You stick it into a beaker of broth, turn it on, and
can feel the beaker heating up by the dissipation of the 20 kHz vibration.
The dissipation is of the order of 500 kW m-3: if you were to do the
same on a production scale, you might need a fair part of the electricity of
a full scale power station. Not all lab techniques can be scaled up!
Enzymatic Release
A second lab techniqueone that is used in productionis enzymatic
release. We add a solution of enzymes (often together with a surfactant)
to the cell suspension and mix. Then we wait, as the enzymes slowly
make the cell walls permeable and release the product. A few of the cell
wall degrading enzymes used for this purpose are shown in the table below.
organisms
component
enzyme
bacteria
murein
lysozyme
phospholipids
surfactant
glucan
glucanase
mannan
mannanase
protein
protease
yeasts
Enzymes do not simply dissolve the cell wall. The cell is a little capsule
with a wall that is (slightly) permeable to water, but much less to other
components in the cell. Originally, the concentration of these components
is higher in the cell than outside. So the concentration of water is lower
inside the cell. Water is drawn into the cell until the pressure is high
enough to counter water transport. This osmotic pressure is quite high:
in the order of megapascals. The enzymes are thought to form little holes
in the cell wall (Figure 2-2). This releases the pressure difference, so water diffuses inwards through the intact part of the wall. This water pumps
the cell contents out through the holes. All components in the cell liquid
Figure 2-2 Enzymatic release
5HOHDVH
are removed at the same rate. The concentration in the cell goes down,
and so does the water permeation. So the release rate decreases in time.
The fraction of the product released according to this model is given approximately by:
c
R
| 1 exp k t
cM
This has two parameters: the maximum attainable product concentration
cM and a rate constant k. Both have to be determined from an experiment,
where the concentration released is followed as a function of time. The
value of cM is proportional to the concentration of cells. The rate constant
k depends on the micro-organism, the enzymes used, their concentrations,
and the temperature. The inverse of the rate constant has the dimension of
a time. This is a measure for the time required for release; if it has a value
of 100 s you can expect everything to be released within a few hundred
seconds. In practical systems k-1 has a value between minutes and an hour.
The release as a function of time is shown in Figure 2-3.
Figure 2-3 Enzymatic release curve
You can determine both constants in the equation from a plot of the concentration versus time. The concentration cM is that of the plateau; the
slope near the origin has the value kcM (Figure 2-4).
Figure 2-4 Finding parameters in a plot
A disadvantage of enzymatic release is the high cost of the enzymes.

Scaling up is not a problem: you get the same results on a large scale as in
the lab (if you use the same conditions!).
Mechanical Release
Products contained in organelles often cannot be released using enzymes.
Here we must use a mechanical cell disruption method. Two commonly
used pieces of equipment are the bead mill and the homogenizer.
The Bead Mill
A bead mill (Figure 2-5) is a cylinder in which impellers turn with a tip
speed of about 10 m s-1. They are filled with little wear-resistant beads
with a diameter of about one millimetre; these are intensely agitated.
About half of the volume is occupied by the beads. Cells are disrupted
between them.
Figure 2-5 A bead mill
Bead mills can be operated in two very different ways (Figure 2-6). The
traditional way is to produce batches of homogenate. The space between
the beads is filled with broth and the mill turned on. After a certain time
typically 5 to 30 minutes the mill is turned off and the homogenate removed. For larger throughputs it is handier to use continuous operation.
Here the feed flows steadily into the machine, and homogenate flows out.
In batch operation, the cell disruption increases in time; in continuous operation the disruption increases along the machine, but is steady at any
fixed position.
If you assume that the release of product is proportional to the number of
intact cells around at any moment, then the release in batch operation
should follow an exponential function:
R 1 exp k t
Here k is again a rate constant that has to be determined experimentally;

its inverse might have a value of ten minutes. As time goes up, the negative exponential decreases to zero; the release then approaches unity. The
value of k depends on cell properties, on the geometry of the mill, how far
it is filled with beads and on the power input per volume of mill. (This is
in turn determined by the size and rotational speed of the discs.) The assumption behind the equation above is only roughly true. Weak cells are
disrupted first, so the release falls more rapidly than predicted. So experiments to determine k have to be done under conditions close to those that
are to be used in practice.

5HOHDVH
Figure 2-6 Batch and continuous operation
For continuous operation, the time should be replaced by the residence

time of the liquid in the mill. This is the ratio of the volume of liquid in
the mill to the volume flow through the mill:
V
V
This only applies if liquid is neatly pushed through the mill (if there is
plug flow). In practice this is not the case, and the performance of a continuous bead mill is not as good as the above procedure predicts.
Bead mills have to be cooled via the cylindrical part of the wall. With an
increase in scale, the wall area increases with the square of the dimensions, but the volume and throughput with the cube. So it becomes increasingly difficult to cool the mill. A single mill can handle up to a cubic
metre per hour.
Scaling up of bead mills has its uncertainties. It starts with batch experiments in the lab or pilot plant using a small mill. This should have
the same chamber geometry as the large machine, use the same cell suspension, the same beads and degree of filling, and have the rotational
speed chosen to give the same power input per volume of mill. From
these experiments we determine the rate constant and translate this into
the parameters of the large machine (batch or continuous).
The Homogenizer
The homogenizer consists of a high pressure pump, followed by a valve
(Figure 2-7). The piston pump has to deliver a constant liquid flow at
pressures between 50 and 150 MPa (a few hundred times the pressure of a
home water system). This requires a sturdy, expensive, multi-cylinder design, with pressure dampers. Such pumps are built for capacities that can
exceed 10 m3 hr-1. The valve is adjustable. In principle, the valve is no
different from that used to tap water. However, it has to be much sturdier,
and to be built using wear-resistant materials.
Cell disruption occurs just behind the slit of the valve. Liquid leaves the
slit with a velocity of hundreds of metres per second. Eddies draw local
spots of vacuum, where bubbles form (cavitation). These implode when

Figure 2-7 A homogenizer
the pressure increases locally, causing sharp shock waves of a high intensity. These in turn disrupt the cells.
The release by a homogenizer increases rapidly with the pressure difference. This is often described by an empirical equation of the type:
R
'p 1..3
1 exp k
'p0
The exponent in the pressure ratio is determined from experiments with

varying pressure differences. If you do not have such data, use the value
2. This equation tells that the release goes up to one with a high pressure
difference. However, we get a temperature rise in the homogenizer that
also increases with the pressure difference. This causes heat sensitive proteins to deactivate and we may find that the release goes through a maximum. The deactivation can be reduced by cooling the flows in and out of
the homogenizer.
To increase the yield we can also use a second pass (Figure 2-8). After a
single pass the fraction of intact cells left is (1 - R); after two passes it is
(1 - R)2. We seldom use more than two passes.
A homogenizer is a simple and robust piece of equipment. It can be used
for large scale production. However, it has its problems:
x a homogenizer is noisy, very noisy,
x the throughput is too high for use in the lab,
x the homogenizer valve wears rapidly and
x homogenizers are expensive.
You can do experiments in the lab by replacing the pump with a French
press. You can regard this as a cylinder which makes a single stroke.
Figure 2-8 Disruption of cells

5HOHDVH
Using Energy
Mechanical cell disruption is our first encounter with flows and the use of
energy. It is a good place to introduce a few tools for handling process
calculations.
You can consider both a bead mill and a homogenizer (Figure 2-9) as systems. A system is something on itself, around which you can draw a
boundary. We do this with dashed lines. You can draw boundaries in
many ways for example, the homogenizer can be regarded as having
two sub-systems: the pump and the valve. Just split in a way that is handy
for your problem.
Figure 2-9 Systems and sub-systems
To set up calculations on systems we use balances. A simple (but important) one is the mass balance:
(change of mass) = (sum of mass flows)
Mass flows in are positive; flows out are negative. In the examples
here the flow in will have the subscript 0; the flow out the subscript 1.
In symbols the mass balance becomes:
dm
m0 m1
dt
For a batch bead mill, this is trivial: there is no change of mass, and there
are no flows in or out (except when filling and emptying). The result is:
0 00
In the continuous homogenizer, there is again no change of mass (except
when starting or stopping) and it will be clear that the flow out is the
negative of the flow in:
0 m0 m1 o m1 m0
The second balance that we use is the energy balance. The energy balance
is usually written in terms of a property of fluids known as enthalpy. For
liquids and solids (but not for gases) this is equal to the energy per mass.
If you have studied thermodynamics you may know why we use enthalpies instead of energies: we will not explain that here. Enthalpy depends on pressure and temperature. The flows in our homogenizers are
mostly water; for this liquid the enthalpy is given approximately by:
p
h
T
U
Here Uis the density of water and its specific heat. (This is commonly
given by the symbol cP, but we need that symbol for other purposes.) We
take these to be constant with values of 1000 kg m-3 and 4000 J kg-1 oC-1.

The energy balance is more complicated than the mass balance. Energy
can enter via mass flows, but also as work flows (power) and as heat
flows.
change of enthalpy work heat

energy flows
flows flows
Also here, flows out are negative. In symbols the equation reads:
dE
m0 h0 m1h1 W Q
dt
For liquids and solids we can approximate the energy by the enthalpy,
giving:
d (mh)
dt
m0h0 m1h1 W Q
(The overbar denotes an average.) In our batch mill there are no flows
and if we want to avoid heating up we get:
0 0 0 W Q o Q
W
So the cooling must remove a heat flow that is equal to the power. We
have worked out the mass and energy balances for the batch mill, the homogenizer and the homogenizer pump in Figure 2-10. You see that you
need a constraint to solve the balances. For the batch mill, this is the design decision that the mill is not to heat up.
For the homogenizer we have decided that there will be no cooling: the
heat flow is zero. The balances then lead to an equation for the temperature increase in the homogenizer:
W
m
T1 T0
As we will see in the exercises, the temperature rise can be substantial. In

the pump we have assumed that there is no rise in temperature. We then
get an equation for the pressure rise in the pump:
UW W
m V
Note that the mass flow divided by the density is equal to the volume
flow. By assuming that there is no rise in temperature we have taken the
pump to be ideal, to have no losses. In a real piston pump perhaps ten percent of the power is lost as heat and the pressure rise is then ten percent
lower. Check that you can follow the derivations.
p1 p0
Combining the two equations above gives the temperature rise as a function of the pressure difference:
T1 T0
p1 p0
U

5HOHDVH
Figure 2-10 Balances for three systems
Further Reading
J.A. Asenjo ed., Separation Processes in Biotechnology, Marcel Dekker,
1990
Exercises
2-1 Enzymatic Release
Enzymatic release form yeast cells has been studied by Hunter and Asenjo (1990). We have brought their results into the form:
R 1 exp k c t
The rate constant is a function of composition:

k c
1
c
4
1 with W 7.5 10 s and cA
W cA
8 103 g L-1
The equation is valid in the range of enzyme concentrations 0 g L-1 < c <
0.3 g L-1. Note that there is some release even when the concentration of
enzyme is zero. This is known as autolysis.
(a) Plot the release over a period of 3 hours (about 10 000 s). Use enzyme concentrations of 0 and 0.3 g L-1.
(b) How long does it take to get a release of 0.95 (95%) of the enzyme?

2-2 The Homogeniser
At a pressure difference of 50 MPa (500 atmospheres) a homogenizer releases a fraction of 0.4 of a protein in a feed. The feed has a temperature
of 20 oC. The feed density is U = 1000 kg m-3 and the its specific heat =
4000 J kg-1 oC-1. Assume that the release as a function of the pressure difference follows:
'p 2
R 1 exp k
'p0
The homogenate leaves the homogenizer at a higher temperature. It is

then cooled in about thirty seconds in a heat exchanger. In these 30 seconds part of the released material can be deactivated thermally. The figure below shows an estimate of the fraction J remaining for different values of the outlet temperature. For this product deactivation becomes serious if the outlet temperature rises above 50 oC. We look at the background of this figure in lesson 4.
(a) Find the constant k in the homogenizer release equation.
(b) Estimate the release and the temperature rise for pressure differences of 60, 80, 100, 120 and 140 MPa.
(c) Multiply your release values with the fraction J remaining. This
tells you which fraction of the original product is left in the homogenate. What do you see?
2-3 Exercise Your Balancing

Balances are important in processing: this is the first of a number of exercises to help you to use them. The two balances that we use are the mass
balance and the energy balance:
d mh
dm
m0 m1
m0h0 m1h1 W Q
dt
dt
d
When using these balances, the first thing to do is to decide which terms
you need and which ones you can scrap. We will practice this using two
examples: that of a real (non-ideal) pump and that of a batch bead mill.

5HOHDVH
A Real Pump
The shaft work into the pump is provided by the drive, which is not part
of our system. The liquid has a density U = 1000 kg m-3 and a specific
heat = 4000 J kg-1 oC-1. Flows in are shown in the figure.
(a) The pump is a continuous system. How large is the mass flow
out?
(b) Is any heat flow shown? Could we scrap the heat flow?
(c) The pressure change is given. Write out the energy balance and
solve it to get the temperature change.
Batch Bead Mill

Our second exercise is on a batch bead mill. This is to be operated at ambient temperature, but we have forgotten to put the cooling on. The mill
contains metal (S) and liquid (L). The masses and specific heats of these
are given in the figure, as are the initial temperature and power input. Because there is no cooling, the temperature of the mill will increase in time.
You may assume that, at any moment, the whole mill has the same temperature. (If you know about heat transfer, you will realize that this is a
reasonable assumption.) This makes it easy to calculate the energy of the
system:
E
(a)
(b)
(c)
(d)
(e)
mS mL

p
US UL
mS S mLL T
Is there a change in pressure?

Are there mass flows in or out?
Are there heat flows in or out?
Does the energy content of the mill change in time?
The energy balance becomes
dE
dt
W or
mS S m L L
dT
dt
Solve this equation to get the temperature as a function of time. How long
does it take for the mill to heat up to 40 oC?

3 Clarify
For most products we have to remove cells (or cell debris) from the broth
before further processing. The three important techniques are:
x centrifuging,
x filtration and
x microfiltration.
Microfiltration is a membrane process that is similar to ultrafiltration,
which we consider in the next lesson. We will not discuss it here.
We can remove large particles by settling. In biotechnology this technique
is only used for precipitates, where fine particles have been agglomerated
to form larger ones. Even so, we first consider settling because the theory
gives a good starting point for that of centrifugation.
Settle
There are three forces working on a particle in a liquid:
x gravity, which pulls the particle downwards,
x a pressure gradient, which forces it upwards and
x the friction force between the moving particle and the liquid.
(If there are large velocity gradients, acceleration forces can also play a
role, but in settlers these can be neglected.)
Single Particle
For a small isolated spherical particle the three forces are:
S 3
S
d gU S d 3 gU L 3SKud
6
6
For steady settling, the sum of these forces is zero. After rearranging we
find the settling (or rising) velocity:
u
g U S U L d 2
18 K
The symbols denote:

g
acceleration by gravity
US density of the solid particle
UL density of the liquid
K
viscosity of the liquid
d
diameter of the particle
We see that the settling velocity is proportional to the square of the particle diameter. Most particles in bioprocesses are small:
precipitates
10-4 < d < 10-3 m
cells
10-6 < d < 10-5 m
cell debris and viruses
10-7 < d < 10-6 m
So settling velocities are small and it is difficult to clarify liquids.
Figure 3-1 gives an impression of the settling velocities that you may expect. This is for water as the liquid and for the range of densities that you

Figure 3-1 Settling velocities
may expect in biotechnology. Note the logarithmic scale the velocities

vary over a factor of ten thousand.
Measuring Cylinder (Lab)
Real particles are seldom spherical. Also they are seldom on their own.
Settling velocities in swarms can be much lower than those of single particles and cannot be easily predicted. However, it is not difficult to measure the settling velocity in a measuring cylinder (Figure 3-2). Single particles tend to catch up with the interface between the suspension and the
clear liquid. So this interface is often quite sharp and its position can easily be followed. If you dont see a clear layer within a few minutes, you
can better forget about settling.
Figure 3-2 Measuring the velocity
Continuous Settler
For large throughputs we use continuous settlers (Figure 3-3). The flow
in settlers is turbulent. Fortunately, much of the behaviour can be understood without considering details of the flow. It is important to realise that
the space for continuous settling is chosen to be long and shallow. (The

&ODULI\
Figure 3-3 Settler dimensions and flow
diagrams in the next section are misleading in this respect). Two simple
models of such tanks are the model with plug flow (no mixing) and that
with mixing of the liquid in the vertical direction.
Figure 3-4 Plug flow model
Plug Flow Settler

We first consider the model with plug flow (Figure 3-4). Starting point is
a rectangular settling tank with width W, length l and height Z. Liquid has
a volume flow from left to right through the tank; we assume that it has
y has the value:
the same velocity everywhere. The velocity
V
WZ
The time that the liquid stays in the tank (the residence time) becomes:
v
l
v
lWZ
WZ
V
Consider a particle entering at the top of the tank (this is the worst position). It sinks with a velocity u. The particle will reach the bottom of the
tank if the time required for settling is less than the residence time:
WZ
Z lWZ
d
V
u
All particles are removed up to a flow rate of:
VM
ulW
uA
Here A is the surface area of the tank. The capacity of a settler does not
depend on the height of the tank, but only on its surface area. A larger
height yields a longer residence time, but the particles also have to move
further. Further on, we will compare models using a settling ratio:
ul
S
vZ
It will be clear that we get complete removal when S > 1. When S < 1,
only those particles entering below a height z (Figure 3-5) will be re-

Figure 3-5 Plug flow (2)
moved. The fraction removed or removal becomes:

R 1 if S t 1; R S if S 1
The plug flow model predicts that a batch settler and a continuous settler
will show the same removal if the batch time is equal to the residence
time.
Figure 3-6 Vertically mixed settler
Vertically Mixed Settler

Also in this model (Figure 3-6), liquid flows from left to right, with the
same horizontal velocity everywhere. Here, however, we assume that the
liquid is turbulent, and that eddies cause rapid mixing in the vertical direction. A second assumption is that the length over height ratio is large,
so that mixing in the horizontal direction can be neglected.
All liquid enters with a concentration c0 of particles. As before, the particles have a settling velocity u. However, they do not fall steadily, as they
are moved up and down by turbulence. Particles that fall on the bottom
are assumed to stay there. In this model the removal becomes:
R 1 exp S
Figure 3-7 Removal in settlers

&ODULI\
Our two models are compared in Figure 3-7. The results are broadly similar, but there are two differences worth noting:
x The vertically mixed model does not have a sharp cut off point it
never gives a complete removal, except for a very large settling ratio.
x The removal from the vertically mixed model is less than that of a
batch settler with the same residence time.
Both observations are borne out in practice.
Centrifuge
Separating particles by gravity takes long if the particles are small. We
can then use centrifuges in which the separation forces are much larger.
Swing Centrifuge
A simple kind of centrifuge still used in the laboratory is the swing
centrifuge (Figure 3-8). The liquid is distributed between two test tubes
that are rapidly swung around. Particles in the test tubes are flung outwards with an acceleration that depends on their radial position r:
Z2 r
The acceleration is also proportional to the square of the angular velocity
Z, which is related to the rotational speed n:
Z 2Sn
This acceleration can easily be hundreds of times larger than that of gravity. If we get a good separation in a swing centrifuge at 1000 g within ten
minutes, we may be able to use centrifugation on a large scale.
(Residence times in a large centrifuge are much shorter, but so are the settling distances.)
Figure 3-8 Swing centrifuge
For scaling up we need the settling velocity ug under gravity. We can

estimate this from the value u obtained with the swing centrifuge using
ug
u
g
Z2 r
Here r is the average radius of the swing centrifuge.

We will look at two kinds of production centrifuge:
x the tubular centrifuge, with a small capacity and
x the disc centrifuge for larger throughputs.
Both work with continuous flow through the equipment.
Figure 3-9 Tubular centrifuge
Tubular Centrifuge
The tubular centrifuge (Figure 3-9) is hardly used nowadays. We look at
it because it shows the principle of centrifugation more clearly than modern machines. The cylinder has a radius r and a length L, giving a settling
area:
A 2SrL
The ratio of the centrifugal acceleration to that of gravity becomes:
Z2 r
g
So the centrifuge behaves like a settling tank with a surface area:

6
Z2 r
2SrL
If we know this sigma value (it is often supplied by centrifuge manufacturers) we can estimate the maximum capacity from a settling experiment
in the lab:
ug 6
VM
The equipment is continuous, so that the vertically mixed model of the

settler is better.
Disc-stack Centrifuge
Disc-stack centrifuges (Figure 3-10) contain a stack of conical discs.
They are the type of centrifuge used most in biotechnology. The settling
distance between the discs is small of the order of a millimetre. So disc
stack centrifuges can have a very large sigma value (up to 0.25 km2!).
The feed enters under the discs through a hollow axis. It first flows outwards before going inwards between the discs. Particles move outwards

&ODULI\
Figure 3-10 Disc-stack centrifuge
and are removed through the rim. (There are several ways of doing this.)
The liquid is removed through a slit along the axis. The geometry of the
settling area (Figure 3-11) leads to a more complicated equation for the
sigma value:
2S Z2
N r03 r13 tan D
3 g
Here N is the number of discs and D the disc angle. Again, you can use
this together with a lab experiment to estimate the maximum capacity of
the centrifuge. The derivation of this equation is given separately in the
appendix behind this book.
Figure 3-11 Stack geometry

Filter
Filtration (Figure 3-12) is the other important method for removing particles from a suspension. The particles are retained by a porous filter
medium usually a specially woven cloth. The particles form a cake
on the medium. Surprisingly, the pores in the medium are usually chosen
larger than those of the particles to be removed, perhaps three times larger than the largest particles. These form bridges and then the cake stops
the other particles. Liquid is forced through the cake and the medium by
a pressure difference. Clear filtrate is collected behind the medium.
Figure 3-12 Principle of filtration
Funnel Filter (Lab)

Figure 3-13 shows a funnel filter that is often used in the lab. It can also
be used to gather data for large installations, as we discuss below. The
medium covers the coarse pores of a support. The feed is poured into the
upper part of the funnel and pulled by vacuum into the receiving flask below. Initially the flow is only limited by the resistance of the medium.
However, the flow goes down rapidly as a cake forms especially when
the feed contains fine particles.
Figure 3-13 Funnel filter
Volume Flux
The volume flux j (flux for short) is the volume flow divided by the area
of the filter (Figure 3-14). It has the dimension of a velocity and is also
known as the superficial velocity. Filtration is only feasible when the
flux is not too low:
j ! 0.5 104 ms1 180 Lm2 hr 1
Cake Ratio
This is the ratio of the cake volume to the filtrate volume (Figure 3-15).
The filtrate volume used is determined at the moment that the top of the

&ODULI\
Figure 3-14 Volume flux
Figure 3-15 Cake ratio
cake falls dry: it does not include liquid inside the cake. If the cake ratio is
large we get no filtrate. We can then still separate by washing or displacing the liquid in the cake, as we discuss further on.
Scaling Up
If you want to use the results of the lab experiment to estimate the size of
large filters to be used in production, you should do the following:
x use the same feed,
x use the same filter medium,
x use the same pressure drop and
x use the same filtration time.
You then get the same flux on the two scales and scale up by increasing
the filter area. The feed should contain the same micro-organisms, with
the same concentrations. In the lab it is common to use filter paper as a
medium but you cannot use that if you need the data for scaling up. The
pressure drop is important because filter cakes of biological material are
compressible. So the cake resistance increases with increasing pressure
drop, sometimes greatly.
To get a thick cake or a short time cycle you may have to extrapolate your
measurements using techniques from the next paragraph.
Filter Resistances
The flux is determined by the pressure drop 'p across the filter, the viscosity K of the liquid and the combined resistances of medium and cake.
Initially, only the medium resistance RM is important (Figure 3-16). However, as the cake height h increases, the cake resistance rCh tends to domi-
Figure 3-16 Medium and cake resistance

nate. As you will find, the cake resistance rC per unit height has a somewhat surprising unit, and a very high value:
1012 m2 rC 1015 m2
Measuring Resistances
Both the medium and the cake resistance can be obtained from the measurement of the volume of filtrate versus time. If you plot (t/V) versus V
(Figure 3-17) the two values can be found from the intercept and the
slope:
RM
A 'p
intercept
K
rC
2 A2 'p
slope
KCR
We derive these equations separately in the appendix Derivations.
Figure 3-17 Filter plot
Filtration Remarks
The medium resistance is not that of the clean mediumit is much higher. It includes the resistance of the particle bridges over the holes. We can
only give an order of magnitude:
RM
105..106
d
For a cake consisting of equal spherical particles (a strongly idealised

cake!) you can estimate the resistance per height using:
rC
180
1 I
3
1
d2
Here d is the particle diameter and I the void fraction of the cake. For
large spherical particles I 0.4, but cakes of smaller particles tend to be
more open. The equation shows that the resistance increases rapidly as
particles get smaller. This is why cell debris cannot be filtered on its own.
Cell debris can be filtered if a filter aid is used (Figure 3-18).
Figure 3-18 Filter aid

&ODULI\
Filter aid consists of particles in the range from 20..50 Pm. These have to
be large enough to form a permeable cake, but small enough to trap the
debris. The volume of filter aid has to be substantially larger than that of
the particles trapped; the type and amount have to be determined by experiment.
Filter aid is usually added in two portions:
x a small part (say 10%) as precoat and
x the rest mixed with the feed (body feed).
Filter aid gives rise to two problems:
x it is difficult to get rid of the cake and
x it can cause much loss of product.
The cake resistance depends strongly on the void fraction I. The void
fraction can often be increased by adding flocculants to the feed, and by a
good choice of pH and ionic strength.
Washing and Blowing
At the end of the filtration, the top of the cake has just fallen dry. The
cake still contains liquidand the product in the liquid. The best way to
retrieve this is to displace the liquid with clean water (Figure 3-19). This
washing is more effective than blowing to remove product. In the ideal
case, the volume of clean liquid is equal to that of the liquid in the cake.
In practice you need more: two or three times as much. Blowing is often
applied after washing, as a first step in the drying of the cake.
Figure 3-19 Washing and blowing
Plate Filters
The two kinds of equipment that are most used in production are:
x plate filters, which are operated batchwise and
x drum filters, which run continuously.
Plate filters (Figure 3-20) consist of a large number of hollow plates.
These are parallel; the distance between them is a few centimetres. The
medium is drawn over both sides of the plates. Liquid flows from the outside to the inside, through the medium. The particles are retained by the
medium; after some time the space between the plates will be filled with
cake. The filter is then dismantled to remove the cake. (This can be done
automatically.)

Figure 3-20 Plate filter
Plate filters are built for pressure drops between 0.05 and 1 MPa. The
maximum pressure drop is often limited by the compressibility of the filter cake. The time required to filter a volume V is:
aV 2 bV with a
KrCCR
and b
2 A2 'p
KRM
A'p
This assumes that there is enough space for the cake between the plates.
Drum Filters
A drum filter or rotary vacuum filter (Figure 3-21) consists of a large
drum that rotates around a horizontal axis. The medium is drawn around
the cylindrical part of the drum. The pressure inside the drum is subatmospheric. The drum turns slowly though a trough with feed. Liquid is
pulled into the drum and a cake forms on the medium. Only the lower
part of the drum (perhaps one third) is in the trough; the rest of the drum
is used to wash and dry the cake. The cake is scraped off just before the
drum returns into the liquid.
The cycle time tC of a drum filter is of the order of half a minute. The volume flow is given by:
1 b b2 4at
t
2a
with a
KrCCR
2
2 fA 'p
, b
KRM
and t
fA 'p
ftC
Here f is the fraction of the drum used for filtration and A the cylindrical
area of the drum. The derivation closely follows that of the filter plot, as
given in the appendix.
Figure 3-21 Drum filter

&ODULI\
Further Reading
Martin Rhodes Introduction to Particle Technology, 7th ed. John Wiley &
Sons 1998, Chapter 2
Don W. Green and Robert H. Perry Perrys Chemical Engineers Handbook, 8th ed. McGraw-Hill 2008, Chapter 18
Exercises
3-1 A Simple Settler
To partly clarify a waste water stream, you consider using a settler. In a measuring cylinder you find a fairly
sharp interface between turbid and clear liquid. The
height as a function of time is given in the table. Settling
has finished after 10.5 minutes and we then have a cake
thickness of 4 cm.
The solids concentration in the feed is 50 kg m-3 and the
liquid and solid particles have densities of 1000 and
1400 kg m-3.
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(min)
(cm)
25
21
17
13
10
12
14
4
Find the effective settling velocity u.
16
5
What is the volume of solids in the feed? (Hint:
volume fraction questions are best answered by
considering one cubic metre of mixture. Think of a measuring
cube one metre high, one metre wide and one metre long.)
What is the volume fraction of cake?
What is the volume fraction of solids in the cake?
Consider a plug flow settler with an area of one square metre. What
is the maximum flow that it can handle?
We need to handle one hundred cubic metres per hour. How much
plug flow area would we need?
If we want a 99% removal, how much would we need using the
vertically mixed model?

3-2 Filling Up of a Settler
This is the next exercise on the use of balances. We will analyze the filling up of a settler. The flow is steady and the (uniform) particles are neatly distributed in the feed. We also assume plug flow. The particles settle
with a velocity u and form a cake layer with a uniform thickness until
the last particle has settled (see figure). During filling up the velocity of
the liquid above the cake increases, but the effect on settling is counteracted by the shorter settling distance. When the settler is operated at its
maximum capacity, the cake will just extend to the end of the settler.
To find how long filling up takes, we consider the cake as our system.
This has a variable upper boundary, as the height increases in time:
The solid and liquid have constant densities, so we can replace mass balances by volume balances. There are two such balances: one for the solid
and one for the liquid. There are only flows into the cake; no flows out.
The balance for the solids becomes:
dVS
u HS A
dt
The change of the volume of the cake becomes:
dVC
dt
For the height we get:

a)
b)
dz
dt
A dz
HcS dt
HS
HcS
Solve this equation to get the height as a function of time.

The settler has to emptied when the cake height reaches the top of
the vessel. Use the following parameters to find out how long this
takes:
u 3 104 ms-1 A 10 m2 Z 0.5 m
V
c)
1 dVS
HcS dt
3 -1
s HS
uA 0.003m
0
0.001 HcS
0.3
What happens in this model if the flow is less than the maximum
flow?

&ODULI\
d)
What happens if the feed contains a small fraction of particles that

settle much more quickly than the rest? Focus on what happens
near the inlet of the settler.
3-3 Centrifuge Capacity

You want to find out whether an existing disc-stack centrifuge can handle
a certain feed. In the lab you do an experiment with a swing centrifuge
with the following parameters:
average radius
r = 10 cm
acceleration at average radius
a = Z2r = 2103 m s-2
ratio against gravity
Rg = a/g = 161
settling velocity
z = 12 cm
We find that clarification takes 4 minutes. The disc-stack centrifuge has a
sigma value of 6 = 1200 m2.
Estimate the volume flow that can be handled by the centrifuge.
3-4 Sigma Value
Centrifuges are expensive pieces of equipment, but they can have a long
life. So you may well encounter second-hand machines, of which the data
have sometimes been lost. It is then handy if you can make an estimate of
the capacity.
Estimate the sigma value of a disc-stack centrifuge with the following parameters:
rotational speed
n = 9200 min-1
number of discs
N = 72
inlet radius
r0 = 8.0 cm
outlet radius
r1 = 1.8 cm
disc angle
D = 38o
3-5 Plate Filter (1)
In your plant you need to filter a bacterial broth. You get the following
results on a small pressure filter in the lab:
area
A = S/4(0.1 m)2
pressure difference
'p = 500 kPa
viscosity
K = 210-3 Pa s
filtration time
t = 2860 s
filtrate volume
V = 600 mL
cake height
h = 5.1 cm
(a)
(b)
(c)
(d)
(e)
Calculate the cake ratio.

We design a plate filter using the experimental cake thickness of 5.1
cm. How large will the flux be?
The batch has a volume of 10 m3. How much filtrate will this give?
Calculate the required area to do this in one go.
What will the area and filtration time be if we use three cycles?

3-6 Plate Filter (2)
The complete set of data of the previous experiment is given in the table to the right.
area
A = (S/4)(0.1 m)2
pressure difference
'p = 500 kPa
viscosity
K = 210-3 Pa s
filtrate volume
V = 600 mL
cake height
h = 5.1 cm
cake ratio
(a)
(b)
CR
hA
600 mL
0.668
(s)
(mL)
88
100
332
200
731
300
1286
400
1996
500
2862
600
Determine the medium resistance and the cake resistance-per-height

from the intercept and slope of a plot of t/V against t.
Calculate the constants a and b in the flux equation, then check that
this gives the same result as the last experimental point.
3-7 Drum Filter

We want to see whether an existing drum filter can be used to filter a suspension of yeast. Below are the relevant data from a lab experiment with
a funnel filter.
filter area
A = (S/4)(0.1 m)2
pressure difference
'p = 5104 Pa
viscosity
K = 210-3 Pa s
filtration time
tE = 119 s
cake height
h = 1.5 cm
The dimensions of the drum of the production filter are:
diameter
D=2m
length
L=4m
pressure difference
'p = 5104 Pa
filtration section angle 'J = (2S / 3) = 120o
fraction of filter used f = 'J / (2S) = 0.333
drum area
AD = SDL = 25.1 m2
rotational speed
n = 2 min-1
cycle time
tC = 1/n = 30 s
(a)
(b)
(c)
(d)
(e)
Calculate the cake ratio.

Estimate the flux during the filtration time.
Calculate the area used by the filtration and the volume flow processed by the filter.
Plot the values of t/V against V and determine the medium resistance and resistance-per-height of the cake from the intercept and
the slope.
Use these values to calculate the volume flow and check that you
get the same value as above.

4 Concentrate (1)
The clarified broth still contains a large amount of water. It is usually best
to remove this before further purification. This concentration step usually
provides some purification itself: for certain products this may be sufficient. We have divided this subject tin two parts
x
concentration methods that use heat and
x
other concentration methods.
We being with evaporation.
Evaporate
This is the oldest and simplest method to remove water (or other solvents)
and to concentrate a solution. Bio products are often unstable, so we need
to evaporate at low temperatures and pressures, and to keep the time in
the equipment as low as possible.
Figure 4-1 Rotary vacuum evaporator
In the Lab
Evaporation begins in the lab, for example in a rotating vacuum evaporator (Figure 4-1). The liquid to be vaporized is contained in a rotating flask
with a tilted outlet. The flask is partly submerged in a heating liquid. The
outlet of the flask is connected to a condenser and a vacuum pump. The
rotation of the flask mixes the liquid and increases the rate of heat transfer. During the experiment we obtain:
x
the temperature of the bath and product,
x
the amount of condensate versus time and
x
samples to measure liquid properties.
The experiment yields the following data:
x
the heat flow into the flask,
x
an indication of how far we can concentrate,
x
how the liquid viscosity changes,
x
whether the liquid foams on boiling and
x
whether we lose product.
If we concentrate too far, the liquid becomes viscous and the heat flux
goes down. At some point the liquid may no longer flow. If the liquid
foams this may give problems in a plant, where the volume of foam will
be much larger than in the lab. Finally, if the product is unstable we may
lose it when it de-activates or decomposes in the vaporizer.

In the Plant
Large evaporators often use bundles of vertical tubes (heat exchangers),
with the product flowing as a film down the inside of the tube (Figure 42). Such a film can be heated more quickly than in our rotating vacuum
evaporator, so we may have less deactivation, or may be able to use a
higher temperature. The heat of vaporization is provided by steam that
condenses on the outside of the tubes. The pressure on the outside is higher than on the inside, so that the temperature there is also higher.
Figure 4-2 A heat exchanger
After the exchanger we separate the vapor and the liquid concentrate
(Figure 4-3). The vapor is cooled and liquefied in a condenser. The pressure in the evaporator is determined by the pressure in this condenser.
When steam condenses it provides an amount of heat of 2.2 MJ kg-1. This
can only evaporate one kilogram of water (or a few kilograms of solvent).
Steam is a major cost in evaporation, so we try to use the steam several
times. Figure 4-4 shows how this is done in a two-stage evaporator. Here
the vapor from the first evaporator is used as heating steam in the second
one. The temperature (and so also the pressure) of the second vaporizer
Figure 4-3 Single stage evaporator

&RQFHQWUDWH
Figure 4-4 Two-stage co-current evaporator
has to be lower than that of the first. We can further reduce the steam consumption with more stages, but at the cost of more equipment.
With large numbers of stages, the steam consumption approaches a lower
limit. This can be determined using thermodynamics, but that is outside
the scope of this booklet. For this lower limit, think in terms of ten to
twenty times less steam than for a single stage. It is also possible to re-use
the steam via a compressor (Figure 4-5). This brings the steam up to a
higher pressure, where it condenses at a higher temperature and we can
recycle the heat into the vaporizer. Here we only need a single stage to
make good use of the energy input, but at the cost of the compressor and
the energy required to run it. Instead of the compressor, we can use a
steam ejector. This is cheaper and more reliable, but less efficient.
As we have found in the lab, we cannot vaporize all the liquid. The concentration of solids cannot be increased to more than 60% of the volume
(and often less). Near the maximum concentration, heat transfer is slow
and we need more and more equipment.
Figure 4-5 Vapor-compression evaporator

Some Properties
To design an industrial evaporator we need data. The first is the vapor
pressure of the solution against temperature and composition. Figure 4-6
shows the vapor pressure of pure water against the temperature. (Note
that the scale is logarithmic.) We need to reduce the pressure to less than
10 kPa (0.1 atmosphere) to evaporate water at 40 oC. The vapor pressure
of solutions of non-volatile substances such as sugars and salt is proportional to the mole fraction of water in the solution. The dashed line shows
this for a mole fraction of water of 0.9. For most substances this is a concentrated solution, as the molar mass of water is small. For a first understanding of evaporation you can neglect this effect.
Figure 4-6 Vapor pressure of water
To set up energy balances we need the enthalpies of our solutions. For

liquid water we have already seen in lesson 2 that this is a function of
pressure and temperature. This function is shown in Figure 4-7 in the lower left corner; the drawn line is for low pressures, the dotted line for the
high pressure of 50 MPa. We see that the effect of pressure is small. (It is
only important in pumps, where temperature effects almost disappear.)
Here we neglect it. With the units omitted, we get for the enthalpy of water:
h 4000 T
h in J kg 1 ; T in oC
Water boils at 100 oC. Steam is formed with an enthalpy that is 2.2 MJ
kg-1 higher: this is the enthalpy of vaporization. If we superheat the steam
above the boiling point, the enthalpy increases further, roughly as:
h
h in J kg 1; T in oC
2.4 106 2000 T

&RQFHQWUDWH
Figure 4-7 Enthalpies of water and steam
Energy Balance
There are two compartments inside the heat exchanger: the part outside
the tubes, where steam is fed, and the part inside the tubes, where the
product feed is partly evaporated. These form the two subsystems of a
model of the heat exchanger (Figure 4-8). We assume that steam condenses on the outside of the tubes at atmospheric pressure (100 kPa) and at a
temperature of 100 oC. The feed enters at 20 oC. On the inside of the tubes
the feed is boiling at 40 oC; the pressure is about 10 kPa. The flows are
both 1 kg s-1. We want to determine the heat flow Q-dot through the tube
walls and the steam fraction f of the mixture of feed (water) and steam
leaving the inside compartment.
Figure 4-8 Model and energy balances

The energy balances of the two compartments are worked out in the figure. The balance of the outside gives the heat flow of 2.2 MW; the balance of the inside then tells us that the fraction of the liquid that is vaporized is 0.85. The rest has been used to heat the feed at the inlet from
20 oC to 40 oC.
Exchange Area
The heat flow Q-dot (in W) through the wall of the evaporator is proportional to the area A and roughly proportional to the temperature difference
between bath and product:
DA 'T
The proportionality factor D is known as the heat transfer coefficient. It

usually has a value between 3000 and 700 W m-2 oC-1. This value is high
when the feed has a low viscosity, but decreases as the viscosity goes up.
For our exchanger we assume a heat transfer coefficient D of 1000 W m-2
o -1
C . The temperature difference 'T is 40 oC. The exchange area required
is then
Q
D 'T
2.2 106
1000 40
55m2
This will require a large number of tubes.

Product Losses
We may lose product in the evaporator by de-activation or decomposition. The rate r usually follows an Arrhenius equation:
r
'E
kc exp
RGT
This contains an activation energy 'E (in J mol-1), the gas constant RG of
8.314 J mol-1 K-1 and the temperature in K. The equation is difficult to
visualize, but for small changes in temperature the behavior is simple:
equal increments in temperature increase the rate by an equal factor. Figure 4-9 shows temperature increments that double the rate. So only small
temperature rises may be allowed especially when we are processing
enzymes or other heat sensitive products.
The residence time in an evaporator might be 100 seconds. Figure 4-10
shows how the activity then varies with temperature (using an activation
energy of 300 kJ mol-1). This is for three values of the constant k in the
Figure 4-9 Temperature increments that double the rate

&RQFHQWUDWH
Figure 4-10 Activity after 100 seconds (see text)
Arrhenius equation. The drawn curve is for a product that is stable at ambient temperature. For small temperature rises, the activity remains high,
but above a certain point it falls rapidly. At a slightly higher temperature
you will have lost all activity in 100 seconds! For less stable products
this happens at lower temperatures. The figure has been obtained by integrating the rate equation.
Exchange HeatSimply
Our evaporator is an example of heat exchange - an important technique
in process engineering. Heat exchange is nowadays a specialism, requiring the use of large incomprehensible computer models. However, there
are simple situations that are both easy to understand and useful. If you
would like to get the idea, read on.
Constant Temperature on Both Sides
You have already seen examples of this in the rotary evaporator in the lab
and in our film evaporator. In the rotary evaporator both the flask and the
bath are mixed, so on each side the temperature is the same everywhere
(except next to the wall of the flask).
Figure 4-11 Temperatures in a simple evaporator
In our film evaporators (Figure 4-11) we have steam condensing on one

side and water (mostly) boiling on the other side. A pure fluid boils or
condenses at a single temperature, so also here each side has a single temperature. (Here, that in the tubes is denoted by t; elsewhere t stands for
time.) For this situation the heat flow is given by
DA T1 t1
Countercurrent, Equal Heat Capacities

In many heat exchangers the two flows move in opposite directions: the
exchanger works in countercurrent mode (Figure 4-12). The equations of

Figure 4-12 A counter-current heat exchanger
countercurrent operations can become complicated, but there is one simple situation: when the heat capacities of the two flows are equal:
m
mcc
Now any change in temperature of temperature in one flow causes the opposite change in the other flow. As a result the temperature difference remains constant along the exchanger. We can again use
Q
DA 'T
Countercurrent, General
The temperatures will usually vary differently along both sides of the exchanger in a way such as shown in Figure 4-13.
Figure 4-13 Temperature profiles
The two temperature differences at the ends of the heat exchanger are
known as the GTD (greater temperature difference) and the LTD (lesser
temperature difference). The temperature difference to use in heat transfer
calculations is the logarithmic mean of these two or LMTD:
'T
GTD LTD
ln GTD LTD
If the GTD and LTD do not differ much this is close to the average tempe
-rature difference.
In some heat exchangers fluids are both heated or cooled and evaporated
or condensed. You can best understand these by thinking of them as split
into different sections.

&RQFHQWUDWH
Further Reading
Don W. Green and Robert H. Perry Perrys Chemical Engineers Handbook, 8th ed. McGraw-Hill 2008, Chapter 11
J.M. Coulson, J.F. Richardson Chemical Engineering, 5th ed. Pergamon
Press 2002, chapter 10.
J.P Holman Heat Transfer, 9th ed. McGraw-Hill 2002, Chapter 10
Exercises
4-1 Mass Balance of an Evaporator
The feed of an evaporator contains:
UW = 1000 kg m-3
water
mW = 950 kg
product
m0S = 50 kg
US = 1400 kg m-3
We want to concentrate this to get a volume fraction of product IS = 0.6.
This will be for the spray dryer in the next exercise.
(a)
(b)
How much water do we need to vaporize?

Which fraction of the water will we have removed?
4-2 A Spray Dryer

A spray dryer is a piece of equipment for making powders. It consists of a
large (mostly empty) tower with a conical bottom. The feed consists of a
solution or fine suspension of a solid in water. It is fed through a sprayer,
which rotates with a tip speed of around a hundred metres per second.
Drops are formed with a diameter of a few tenths of a millimetre. These
are shot into hot air, so that they dry rapidly and are solid before they
reach the wall. The solid powder is withdrawn from the bottom of the
cone.

There is a large circulation flow in the dryer. As a result, the temperature
is almost constant (except close to the hot air inlet). We take the dryer to
consist of two subsystems: the dryer (tower) and the air heater. The system is continuous, and the flows in and out are shown in the figure. Note
that are allowed to neglect the flow of solids in the energy balance. We
need the following parameters:
hL(T) = WT
enthalpy of water
W = 4000 J kg-1 oC-1
enthalpy of steam
'h = 2.2106 J kg-1
hV(T) = WT + 'h
hA(T) = AT
enthalpy of air
A = 1000 J kg-1 oC-1
T1 = 120 oC T2 = 60 oC
temperatures
T0 = 20 oC
-1
The water flow to be evaporated is mW 1kgs
(a)
(b)
(c)
Set up an energy balance of the dryer. Use this to calculate the air
flow.
Use the energy balance of the air heater to get the heat flow.
If the ratio of solids to liquid is 50/50, what is the powder production? And if the ratio is 60/40? Why would we want to concentrate
the feed as far as possible before using a spray dryer?
4-3 De-activation in an Evaporator

In this exercise we estimate the thermal de-activation of a bio-product in
an evaporator. We assume that the product goes through the evaporator in
plug flow. It has a certain residence time and the temperature is the same
everywhere. The rate of de-activation is
r
dc

dt
kc
For a given temperature k is constant. The concentration or activity then

decreases form its initial value c0 as:
c
exp kt
c0
The rate constant follows the Arrhenius equation with the parameters:
k T
k ' 1047 s-1 'E
'E
k ' exp
RGT
300 103 J mol-1 K-1 RG
8.314 J mol-1 K-1
Use a residence time of 100 s and calculate the remaining activity for
temperatures between 20 and 60 oC (293 .. 333 K).

&RQFHQWUDWH
4-4 Heat Exchange - Simply
In this exercise you are to calculate the heat flows and exchange areas required for the a number of simple heat exchangers. Use the same parameters in all of them:
W = 4000 J kg-1 oC-1 'h = 2.2106 J kg-1
D = 1000 W m-2 oC-1
(a)
Consider the counter-current exchanger below. What is the temperature difference along the exchanger for the upper stream? And for
the lower one? Use the energy balance of either stream to calculate
the required heat flow. Then calculate the required area.
(b)
In the following exchanger we heat a stream using steam. Calculate

the LMTD and compare this with the average temperature difference (GTD - LTD) / 2.
(c)
Do the same for the cooler below. Cooling is obtained by evaporating a solvent such as liquid propane or liquid ammonia.
(d)
As a last example we take an exchanger with unequal flows. Use an

energy balance of the upper flow to obtain the heat flow. Then calculate the LMTD and use this to obtain the required area of the exchanger.

5 Concentrate (2)
In this second lesson on how to concentrate we look at two processes that
work at ambient (or lower) temperatures:
x
precipitation and
x
membrane separations.
Precipitate
Proteins are precipitated by decreasing their solubility with a precipitant. The precipitate can be much more concentrated than the feed.
Protein Solubility
The solubility of proteins cannot be predicted accurately, although the
trends are known and understood. Proteins contain both weak basic and
weak acid groups of molecules. When the protein dissolves in water,
these can ionize, giving the protein a charge. At a high pH, the acid
groups are ionized, giving the protein a negative charge. Proteins have a
positive charge at a low pH. The pH where the charge is zero is known as
the iso-electric point or pI (Figure 5-1). Far away from the pI proteins repel each other, but near the pI this is much less so, and they can precipitate, usually in the form of flocs. Their solubility goes through a minimum at the pI.
Figure 5-1 Charge and solubility of proteins
Small ions form an electrical double layer around proteins. At a low concentration of the ions, the double layer is relatively thick, so the protein
molecules repel each other. When more ions are added, the layer thins
and the protein may precipitate. The solubility cS often follows:
log cS
k1 k2 I
with I
1
ci zi2
2
(See Figure 5-2.) The constants k1 and k2 depend on the protein / salt combination. k1 also depends on the pH and somewhat on temperature.

Figure 5-2 Protein solubility versus ionic strength
The variable I is the ionic strength. It contains the molar concentrations

ci of the different ions and their charge numbers zi.
Proteins have a low solubility in organic solvents. This is due to the low
dielectric constant of the solvents, which inhibits ionization. All this leads
to the three common ways of reducing solubility:
x
by adding a salt (usually ammonium sulphate)
x
with a solvent (usually acetone or an alcohol).
x
by changing the pH with acid or base
Salts (but not chlorides) have advantages:
x
they cause little denaturation,
x
they can often be used at ambient temperature,
x
they can give a fairly pure product.
The disadvantages are:
x
we have to process large amounts of salt water,
x
salts have to be removed from the product and
x
salts are corrosive.
You need up to 60 % of solvent for precipitation, and this may cause denaturation. So also here we may have to work at low temperatures. However, solvents can be recovered and re-used more easily than salts.
Acids and bases are cheap, and we do not need large amounts of them.
However, near the iso-electric point, proteins are often unstable. So also
here we may need to work at low temperatures.
Stirred Vessel (Mixer)
We usually precipitate in a stirred vessel or mixer (Figure 5-3). This is
the point to say something about such vessels. They come in great variety,
but most are cylindrical, with the stirrer in the axis. They have baffles to
avoid rotation of the liquid. The liquid height is about the same as the vessel diameter.
Of the many types of stirrer used, we only look at two:
x
the downward pumping propeller and
x
the turbine mixer.

&RQFHQWUDWH
Figure 5-3 Stirred vessel
The two types have rather different characteristics. The propeller gives a
large axial circulation that sweeps the bottom of the vessel; the turbine
gives radial flow with intense turbulence near the mixer blades. The propeller is good at suspending solids; the turbine at breaking up particles.
The stirrers are placed one diameter or less above the bottom of the vessel,
so that they can be used when the vessel is partly filled.
The flow in stirred vessels is highly turbulent. (The exception: small vessels in the lab containing viscous liquids you should avoid these.) Circulation velocities are a few tenths of the tip speed of the stirrer, which is
typically a few metres per second. Even in large vessels the time required
to circulate and mix the contents is less than a minute. Local mixing is
much faster again in the turbulence behind the stirrer blades.
The power input of the stirrer per kilogram of liquid (the dissipation) is a
good measure of the degree of agitation. The power consumed depends on
the rotational speed n, the diameter d of the stirrer, the density U of the
liquid and a constant Po known as the power number:
W
Po U n3d 5
For a propeller the power number has a value of about 0.4; for the turbine
a value of 5. So, with the same speed, the dissipation of the turbine is an
order of magnitude higher than that of the propeller. To get the average
dissipation e we divide the power by the mass of liquid in the vessel. This
gives:
3
d
e | 0.7 Po n3d 2
D
Here D is the vessel diameter. The dissipation varies strongly throughout
the vessel. A large part of the dissipation occurs in a small volume downstream of the stirrer blades. This volume is of the order of d 3; the dissipation near the stirrer is of the order:
eM
Po n 3d 2
So to avoid a high local dissipation, we need a large stirrer with a low

speed; for a high local dissipation we need the opposite (Figure 5-4).

Figure 5-4 Low shear and high shear
The behavior of bubbles, drops and particles in a stirred vessel is governed by the turbulence around them. To break down bubbles and drops
we need high shear rates. That is why you will often see small turbines in
fermenters and the extractors that we discuss in lesson 6. To form large
flocs in a precipitator we need low shear rates, but we have to keep the
flocs suspended. Here a large propeller is better.
Process Steps
We start the precipitation with strong agitation to rapidly mix in the precipitant (Figure 5-5). This liquid is brought in just downstream of the stirrer where the mixing is most intense. The flocs that are formed here are
small; they are not easily broken down by the stirrer. To form larger flocs,
we then reduce the agitation. If this is not sufficient to let the flocs grow,
you may add a suspension of small inert particles (for example gypsum).
The protein flocs adhere to these particles.
From Lab to Plant
When scaling up, several things have to be kept the same in the small and
large scales:
x
the feed composition,
x
the chemicals used
x
the geometry (shape) of the stirred vessel
x
the dissipation or intensity of agitation,
Figure 5-5 Process steps

&RQFHQWUDWH
x
the time for addition of the precipitant and
x
the times for the different mixing steps.
Note that the rotational speed of the stirrer is higher in the smaller vessel
than in the large one.
Use a Membrane
We can also concentrate using membranes. These can be regarded as filter media with very small pores. The common membrane processes are
shown in Figure 5-6:
(1) Reverse osmosis (RO) membranes have pores in the sub-nanometre
range. They retain most species except water. They do need a high
pressure difference (in the range of megapascals).
(2) Nanofiltration membranes have pores of around a nanometre. They
pass water and small ions such as Na+ and Cl-, but retain larger
molecules.
(3) Ultrafiltration (UF) membranes have pores of up to 10 nanometres.
They retain proteins, viruses and cells. We regard them in more detail further on.
(4) Microfiltration membranes have pores with a diameter of 0.2..0.5 Pm.
They retain cells, like ordinary filters.
In addition to these pressure-driven processes you may encounter pervaporation (PV) that we discuss in lesson 8, and electrodialysis (ED).
Electrodialysis is a fairly complicated process that removes ions from solution using an electrical field.
Figure 5-6 Pressure-driven membrane processes
Membrane processes have a low investment, they make good use of energy and they work at ambient temperature. On the other hand, they require
maintenance (cleaning and membrane replacement). Also, they do not
give sharp separations.
UF Membranes
The membrane processes have much in common, and we will only discuss one of them in any detail. Ultrafiltration modules are mostly in the
form of porous tubes with a diameter of a few millimetres to a centimetre

Figure 5-7 A UF module
(Figure 5-7). The membrane is a layer on the inner surface of the tube. It
may consist of a porous, spongy polymer; of porous carbon; of a porous
ceramic layer; or of fine sintered metal particles. The tube itself is also
porous: it forms the support of the membrane.
The pressure difference across the membrane is a few hundred kilopascal.
This forces water (with dissolved salts) through the membrane. The volume flux j through the membrane is small around 10 Pm s-1 (36 L m-2
hr-1). Increasing pressure differences first increase the flux (Figure 5-8).
Figure 5-8 Flux versus pressure drop
The water transports proteins towards the membrane, where they are
stopped. So the protein concentration at the membrane rises. Protein is
mixed back into the main stream by diffusion and turbulence. At a certain
pressure difference the concentration of protein at the membrane becomes
so high that the protein precipitates. It forms a gel layer that is not very
permeable. A further increase of the pressure difference does not increase
the flux only the thickness of the gel layer. This maximum flux is given
by:
c
j k ln G
c

&RQFHQWUDWH
Here k is a mass transfer coefficient that increases with increasing flow
along the membrane. The concentration cG is that where gel formation
starts; the concentration c is that of the protein in the bulk of the liquid.
We see that the maximum flux decreases as the bulk concentration increases. It goes down to zero when c = cG.
In the regime with gel formation we can increase the flux by increasing
the flow along the membrane. This increases the mass transfer coefficient
and the transport of protein back into the bulk fluid. The velocity v along
the membrane is in the range of 1..3 m s-1giving (roughly):
k
105 v
Large velocities require too much pumping power.

In the Lab
The flux as a function of concentration can be measured in the lab in a
single membrane tube (Figure 5-9). The liquid in the stirred vessel is
pumped around through the membrane tube and the permeate is collected
and measured as a function of time. During the experiment, the concentration of the protein increases. So we only need to do a single batch experiment to cover the whole concentration range. The data from this experiment are sufficient to design large installations also those using multiple
stages or continuous flow.
Figure 5-9 Batch UF in the lab
In the Plant
There are two main ways of scaling up. In not-too-large installations, ultrafiltration is done batchwise (as in our lab experiment). Here we simply
need more tubes in parallel in the plant.
Large installations are often operated continuously, using several stages
in series as shown in Figure 5-10. Each stage has its own circulation loop.
The flux through the membrane decreases in time due to fouling. So
membranes often have to be cleaned daily using detergents and other
chemicals. Even then, they have to be replaced after some time.

Figure 5-10 Continuous multistage UF

Yield
The membrane splits the feed into a permeate and a concentrate. We use a
large circulation along the membrane, so the concentration cC of the concentrate is the same everywhere. This then also applies to the concentration cP of the permeate. We can then describe the separation performance
of the membrane by its retention:
R 1
cP
cC
The retention differs for different components, but for a given component
and operation we can regard it as constant. Values are often given by
manufacturers. For proteins the separation is almost complete, so R 1.
For a salt there is little difference between the two concentrations: cP cC
and R 0.
In the following we have to distinguish between batch and continuous operation (Figure 5-11):
Figure 5-11 Batch versus continuous separation
In batch operation we separate a volume of the feed into a volume of concentrate and a volume of permeate. In continuous operation we separate a
flow into two flows. The yield K is described using a concentration factor
D. For the two cases:
continuous
batch
D
VF
VC
R 1

VF
VC
1
D 1 R R
&RQFHQWUDWH
You will find that the batch separation is more effective than the single
stage continuous operation. However, the difference disappears when
many stages are used as in Figure 5-10.
Diafilter
We can remove salts and other small molecules from proteins by diafiltration rinsing out the salts with water (Figure 5-12). We treat a batch of
feed, but keep adding fresh water to compensate for the volume of the
permeate. The membrane has to have a high retention for the protein, but
a low retention for the salt. In diafiltration, the concentration of salt goes
down exponentially with the volume VW of water added:
c0 exp W 1 R
V
F
Here c0 is the initial concentration and VF is the volume of the feed. To

limit the volume of water required, we must concentrate the feed before
diafiltration.
Figure 5-12 Diafiltration (salt removal)
Further Reading
M. R. Ladish: Bioseparations Engineering, Principles, Practice and Economics, Wiley-Interscience, 2001, Chapter 4.
J.A. Wesselingh and R. Krishna Mass Transfer in Multicomponent Mixtures, VSSD 2006, Chapters 14-21

Exercises
5-1 Scale Down a Stirred Vessel
We need a stirred vessel of two cubic metres for precipitation. It is to be
fitted with a downward pumping propeller of one half of the diameter of
the vessel. The filling height is taken equal to the diameter.
(a)
(b)
(c)
(d)
Calculate the diameter of the vessel, assuming that it is cylindrical.

Take the propeller to have one half
of this diameter.
The stirrer is to have a variable
speed drive. This is to provide a
dissipation (power input) between
0.001 and 1 Watt per kilogram.
Calculate the required power and
speed for the highest input. Assume
that the liquid is water.
In the lab we want to do experiments with a vessel with one thousandth of the capacity of the production equipment. We use the
same geometry (shape). Calculate the stirrer diameter and speed for
the maximum dissipation in this vessel.
By which factor does the speed have to be reduced to reduce the
dissipation one thousand times?
5-2 Concentration by Precipitation

A broth contains a protein P. In the lab we find the
following solubilities in ammonium sulphate solutions. We denote the ammonium sulphate as AS.
(a)
(b)
(c)
(d)
CAS
gL
0
-1
CP
g L-1
10.0
130
1.0
How much A do we precipitate from ten cubic
260
0.1
metres of broth with a concentration of 260
g L-1 of P?
400
0.01
The volume of the precipitate is 1/3 cubic me530
0.001
tre. By which factor have we concentrated the
protein?
Which volume of water does the precipitate contain? Use as densities of the water and the protein UW = 1000 kg m-3; UP = 1400
kg m-3.
How much ammonium sulphate does the precipitate contain?
(Assume that the protein contains no AS.)

&RQFHQWUDWH
5-3 Enzyme Concentration with Ultrafiltration
We concentrate an enzyme solution with ultrafiltration and find the fluxes shown in the table as a
function of the concentration factor.
j
-5
D
-1
10 m s
Use the data to estimate the mass transfer coefficient of the protein away from the membrane. Also estimate the concentration factor at which the
protein precipitates. Take the retention of the
membrane for the protein as R = 1.
1.40
1.0
1.15
1.5
0.77
3.0
0.49
5.0
Hint: plot j against ln(D). Determine the slope.

When does the flux go to zero?
0.23
8.0
5-4 Diafiltration of Protease

In this exercise we purify a broth containing proteasea constituent of
many detergents. The protease is a large molecule that is retained by an
ultrafiltration membrane. Small contaminants such as sugars, amino acids
and salts are retained much less. The retentions for protease and the small
molecules are RP = 0.99 and RS = 0.10. The volume of the feed is VF =
10 m3. After pre-concentration with ultrafiltration, the protease has a concentration of cP = 160 kg m-3. For the ultrafiltration module you may take
the mass transfer coefficient and the concentration where gel formation
occurs as k = 10 Pm s-1 and cG = 600 kg m-3. We want to reduce the concentration of the contaminants by 95 %.
(a)
(b)
(c)
How much water do we need?

Which fraction of the protein is lost?
How much membrane area do we need to process a batch in one
hour?

6 Extract
We have released and clarified the product, and may have concentrated it
as well. However, we may still have to separate water from the product.
Three important processes for such separations are liquid/liquid extraction, crystallization and distillation. They are complicated, so we spend a
separate lesson on each.
In the Lab
To extract, we form a dispersion of two liquids that do not mix such as
water and oil. We shake or agitate the two together (Figure 6-1). The
product moves into one of the liquids (the solvent); the contaminants
stay in the other. We then separate the two liquids and finally separate
solvent and product. A single extraction usually does not give a sufficiently pure product: we will see in a moment how to deal with this.
Figure 6-1 An extraction funnel
Small molecules such as citric acid or penicillin are extracted with organic solvents such as one of the higher amines or butyl acetate. It is also
possible to form two non-mixing aqueous phases. An example is a concentrated solution of water with potassium phosphate and polyethylene
glycol (PEG). This separates in two layers: a dense one mainly containing the phosphate, and a lighter one containing nearly all the PEG. Such
mixtures can separate large molecules, including proteins, DNA and cell
debris. However, this two-aqueous phases technique is not used on a
large scale, so we do not further discuss it here.
Separation Funnel
Extraction is carried out in the lab using a separation funnel (Figure 6-2).
A volume VW of the clarified broth goes into the funnel; the initial concentration of citric acid in this feed has a value of c0 kg m-3. Then we add
a volume VE of the pure solvent, close the funnel and shake it vigorously.
Citric acid moves into the solvent. Its concentration in the solvent increases and that in the water decreases. The acid also starts to move back
to the water, and at some point the rates back and forth become equal.
The mixture is then at equilibrium. This might take a minute.

Figure 6-2 Distribution of citric acid
Distribution Coefficient
When we stop shaking, the liquids separate in two layers. We carefully
remove the lower layer using the valve and analyze this layer (or both layers) to determine the concentrations of citric acid. That in the aqueous
phase has a value c; that in the solvent a value C (both in kg m-3). The ratio of these is the distribution coefficient:
C
c
This property is important. It should be neither too small, nor too large.
For a good extraction its value should preferably be larger than one. For
contaminants the value should be much smaller. The distribution coefficient depends on many things:
x
the product and its solubility in the aqueous phase,
x
the acidity and ionic strength of the aqueous phase,
x
the solvent and any additives to the solvent and
x
the temperature.
As said, we will consider the distribution coefficient as constant.
Extraction Factor
Once we know the distribution coefficient, it is not difficult to calculate
how a substance distributes in the two liquids. This ratio is called the extraction factor:
E
substance in solvent
substance in water
VE C
VW c
VE
VW
In this lesson we will make much use of such extraction factors. So we

have included a few tests so that you can train yourself in their use.
Please try: you will need them to understand what follows.
Test 6-1
We fill a separation funnel with two equal volumes of feed and solvent:
VW = VE = 0.1 L. The concentration of citric acid in the feed is cA = 50 kg
m-3 and its distribution coefficient A = 3. The values for a contaminant
are cB = 2 kg m-3 and B = 0.2. How large is the extraction factor for citric acid? And for the contaminant? Which part of the citric acid remains
in the aqueous phase? And of the contaminant? How much of the citric

([WUDFW
acid remains in the water and is lost? How much of the contaminant do
we find in the product?
In calculations it is handy to realize that the amounts distribute as:
substance in solvent E
substance in water
1
So, if an unknown amount u remains in the feed, then the amount originally in the feed is (E + 1) u.
Reversible Equilibrium
In cyclic processes we need reversible equilibriums, which can go back
and forth. For this we use the reversible reaction shown in Figure 6-3.
This is a two phase reaction. To understand it, we need to know the solubilities of the participating components. The water phase consists mainly
of water, with dissolved citric acid. (This is a weak acid that is only partly
dissociated.) The other phase consists mainly of a diluent: a liquid hydrocarbon that has a low viscosity and that hardly dissolves in water. The
amine is dissolved in the diluent. It has hydrocarbon chains attached so
that it also hardly dissolves in water. (The pure amine is too viscous to
form a good solvent.) Citric acid is quite polar, so it hardly dissolves in
the solvent. However, it can react with the amine at the interface of the
two phases, and the product does dissolve in the solvent.
Figure 6-3 Interface reaction of citric acid
The solubility depends on the amine, the diluent, the pH and the temperature. We will be doing calculations using the isotherms from Figure 6-4.
These show how the concentrations in the two phases are related at equilibrium. The isotherm at 25 oC starts in the origin and first rises with a
slope = c /C = 3. At high concentrations of acid in the water, all amine
in the solvent will have reacted, and the isotherm reaches a plateau. The
height of the plateau depends on the concentration of the amine. A high
concentration gives a high plateau, but also a viscous solvent with slow
transfer of citric acid. In our (simplified) calculations we will neglect the
effect of the plateau.
The reaction between citric acid and amine produces heat. This implies
that the reaction tends to reverse at higher temperatures. The distribution
coefficient goes down to about = 0.03 at 100 oC. We use that to reverse
the extraction and send the citric acid back to water.

Figure 6-4 Isotherms of citric acid
Citric Acid
To introduce extraction and crystallization processes, we consider the
production of citric acid (Figure 6-5). Citric acid is an extracellular product of Aspergillus niger in a fermenter, where it has a concentration of
about 50 kg m-3. The broth is clarified using a rotating drum filter. The
clear broth still contains many contaminants in addition to the product. So
we extract the citric acid with a tertiary amine that is not soluble in water.
The citric acid reacts reversibly with the amine and dissolves nicely:
o
25 C
CiH 3 R 3 N
o R 3 NCiH 3
Most of the contaminants do not dissolve. As the two liquids (the

phases) do not mix, they form two layers when we stop mixing. We separate the amine layer and contact it with water at a higher temperature,
where the reaction reverses and citric acid is released into the water:
o
100 C
CiH 3 R 3 N m
R 3NCiH 3
This water from the reverse extraction still contains some contaminants
and remnants of the solvent. To remove these, we vaporize the water so
Figure 6-5 Producing citric acid

([WUDFW
the solution becomes supersaturated and citric acid crystallizes. We separate the crystals on a drum filter. The water from the crystallizer is recycled to the extraction.
Our process contains several cycles. The amine solvent flows round from
extraction to reverse extraction and back again. Water flows from the
crystallizer to the reverse extraction and back again. There is a water cycle from the second filter back to the crystallization. A fourth cycle is less
clear. Traces of the solvent go to the crystallizer, where they evaporate
and return to the reverse extraction. A process engineer might also want to
recycle the water from the first filter to the fermenter, but this is not allowed because of the risk of infection.
Simplify
In our process there are five steps:
x
two rotating vacuum filters,
x
the extraction,
x
the reverse extraction and
x
the crystallization.
The last three are multi-stage steps. All steps are connected and there
are several cycles. Accurate calculations on such systems are the domain of specialists using large computer models. (Their results are not as
accurate as their computer output might suggest). So we simplify:
x
the flows leaving a stage are in equilibrium,
x
the total flows in a step are constant,
x
the distribution coefficient is constant in a given step,
x
we neglect product losses in a step (after checking!) and
x
we use rounded values.
The results will be approximate but good enough to show what is going
on.
In the Plant
Extraction on a large scale is done in different kinds of equipment. Here
we only discuss mixers with settlers (Figure 6-6).
Figure 6-6 Continuous mixer-settler
Mixers with Settlers

These are similar to the mixers and settlers that we have seen earlier.
However, here there are continuous flows of both phases through the

mixer; these phases are separated in the settlers. For a good extraction we
need to disperse the phases through each other. This requires agitation,
but not too much. Too much stirring gives fine dispersions which are difficult to separate in the settlers.
Further on we regard the mixer-with-settler as a single piece of equipment, with two flows entering and two flows leaving. Such a combination
is known as a stage and we draw it as a little square. The arrows for the
flows are drawn such that water and solvent flow in opposite directions.
Why this is, we will see in a moment.
Figure 6-7 shows stage number one. The flows leaving the stage take the
number of the stage. The concentrations of citric acid in these flows are c1
and C1. The aqueous feed enters from the left; it has a concentration c0.
From the right a solvent stream enters: this is the stream that has left stage
2 and it has a concentration C2.
Figure 6-7 Equilibrium stage number one
Equilibrium Stage
We now introduce our simplifications. The first is that the liquids stay in
the mixer long enough to attain equilibrium:
C1
c1
So we consider the stages to be equilibrium stages. The second is that

the two liquid flows do not change in such a stage. (The water phase consists mainly of water, which does not transfer; the solvent phase consists
mainly of solvent and amine, which do not transfer either.) So:
VW VW ,00 | VW ,,11 | VW ,2 etc.
VE VEE,1,11 | VE ,22 | VE ,3 etc.

The ratio of the flows of citric acid leaving a stage is then:
VE C
VW c
VE
VW
We first look at a single stage (Figure 6-8). We take the flows for a plant
with a production of 1 kg s-1 (about 30 000 tons per year).

([WUDFW
Figure 6-8 Calculating the flows
The following data are given:
1kgs
1kg
g -1
x
citric acid only enters via the feed: VW c0
x
the solvent entering contains no citric acid: VE C2
x
the flows have values: VW
20 Ls-11; VE
x
the distribution coefficient has a value:
The extraction factor then has a value:
0 kg s-1
13Ls-1
13
2
20
(Using a rounded value.) We begin with the citric acid flows at the right
side of the stage: the unknown flow u that leaves the stage, and the flow 0
that enters. We calculate the remaining flows in two steps:
x
the flow leaving with the solvent is Eu and
x
the sum of the flows entering equals the sum of the flows leaving:
E
1kg s-1 0
u Eu o u
1
1 E
0.33kg s-1
Series of Stages
Our plant contains three extraction stages (Figure 6-9).
Figure 6-9 Starting calculations with E = 2
With such a series of stages, we begin with the rightmost stage. Also here
the flows are expressed in the unknown value u. In the figure, we have
done this for stages 2 and 3; we leave it to you to do the calculations for
stage 1. You can check your answers at the end of this section.
Test 6-2
Figure 6-10 shows the same three-stage system again to allow you to
practice a bit more. However, here the extraction factor has a value E = 1.
Also here, you are to calculate the flows in terms of u and to find their

values. You will find that the extraction of citric acid is not so good. You
could improve this by adding more stages, but with E = 1 you will need a
lot of them. With less solvent (E < 1) you cannot get complete extraction.
Figure 6-10 Starting calculations with E = 1
The technique of starting with an unknown flow u, then stepping backwards using extraction factormass balanceextraction factor is useful for small numbers of stages. For large numbers you will find it handier to use an equation that goes under the name of Kremser (Figure 6-11).
Figure 6-11 The Kremser equation
For the important case that E = 1 you will find that the equation results in
a division by zero. The complete relation, including this case, is:
if E 1 then
u
VW c0
1
u
else
N 1
VW c0
E 1
E N 1 1
Why Counter-Current?
You may have wondered why we let the two flows go in opposite directions. Figure 6-12 will help you to answer this question. It shows a cocurrent system with two equilibrium stages. We have worked out the first
stage you are to do the second. Then check at the end of the lesson to
see whether you have found the same as we have.
Figure 6-12 Two co-current stages
You should find that the second equilibrium stage does nothing. The two
liquids entering are at equilibrium and they stay that way. So co-current
flow is not a good way to run an extraction. Other flow schemes are possible besides counter- and co-current flow, but none of them is as effective as counter-current flow.
Reverse Extraction
There are several changes in the reverse extraction (Figure 6-13). First of
all, the distribution coefficient has gone down to ' = 0.03. The auxiliary
flow is now a flow of water. This is obtained by condensing water vapor
from the crystallizer as we discuss in the next lesson. This flow has a value of 1.2 L s-1.

([WUDFW
Figure 6-13 Extraction and reverse extraction
When calculating the extraction factor, we always put the auxiliary flow
on top. Here this gives:
Ec
VWc cc
VE C
1 VWc
c VE
1 1.2
2
0.03 13
Test 6-3
Check that the flow u' leaving the reverse extraction is small (Figure 614).
Figure 6-14 Flows and separation in the reverse extraction
Losses in a Cycle
When we set up the calculation for the forward extraction, we assumed
that no citric acid entered with the solvent stream. This is the stream that
leaves the right side of the reverse extraction. You have seen that the flow
is indeed small. However, it is not zero. The streams leaving the cycle are
approximately as given in Figure 6-15. Here u and u' are both calculated
assuming that the entering stream of the auxiliary phase is pure.
Figure 6-15 Fractions of the feed leaving the cycle
The flow leaving at the bottom left forms the feed of the crystallization
section that we discuss in the next lesson. It contains a little less than the
1 kg s-1 of citric acid that we began with. However, we will forget about
that for the moment.

Answers to the Tests
The text contains several figures of an extraction system where you have
been asked to fill in the flows. Our answers are in the figures below.
Figure 6-9
Figure 6-10
Figure 6-12
Figure 6-14
Further Reading
J.D. Seader, Ernest J. Henley and D. Keith Roper Separation Process
Principles, John Wiley & Sons, 3rd ed., 2010
J.C. Godfrey and M.J. Slater Liquid-Liquid Extraction Equipment, John
Wiley & Sons 1994

([WUDFW
Exercises
6-1 Two-step Separation Funnel
This is a similar exercise to that in test 6-1, but now we will use two steps.
We use one half of the solvent in the first step, and remove it after shaking
and settling. Then we add the second half for a second extraction, shake,
settle and separate again.
(a) How large is the extraction factor of citric acid in the two steps?
(b) Which part u of the citric acid remains in the aqueous phase after the
second step? (You can find this using the figure below and filling in the
amounts in terms of u, working backwards from right to left.)
6-2 Langmuir Isotherm

The upper 25 oC line in Figure 6-4 has been calculated using the
Langmuir equation:
C
CM
c
c _ c
The behavior of this equation is best understood by looking at the extremes:
c o0 C
CM
c _ c c
cof C
CM
The figure below shows the line over a larger concentration range.
(a) Which value has been used for the plateau concentration?
(b)
(c)
(d)
The constant c_ does have a meaning. What is it?

(Hint: look at the value of C when c = c_.)
At which value of c does the isotherm begin to diverge more than
10% from the linear equation that we assume in our calculations?
Consider the average extraction factor:
E
VE
VW
Will this increase or decrease at high concentrations? Do you think

our linear assumption will be too pessimistic with regard to the separation? Or too optimistic?

6-3 Penicillin Extraction
Extraction is sometimes governed by a reaction in water. The extraction
of penicillin (Pen) in butyl acetate (BA) is an example. Pen is a fairly
strong acid, which largely dissociates in water:
PenH o Pen- H +
Only the part that is not dissociated dissolves in BA. So here we can control the equilibrium by varying the pH. Penicillin is extracted using
cooled high speed centrifugal extractors with a short residence time
(Podbielniaks):
(a)
(b)
Would you choose a high or a low pH for the extraction? And for
the reverse extraction?
Why would such an expensive extractor be used?
6-4 The Kremser Equation

Repeat the calculations in Figures 6-9, 6-10 and 6-14 to see whether you
get the same results using the Kremser equation.
6-5 Losses in the Cycle
Consider our citric acid extraction.
(a) Using the results from the Answers to the Tests calculate the loss
of product of the complete extraction cycle.
(b) We increase the number of stages in the forward extraction to 5.
Keep the values of E and E' the same. Also keep the number of
stages in the reverse extraction the same. Use the Kremser equation
to find the new value of u. Use this value to calculate the losses in
the new system.
6-6 Extracting a Fungicide
Consider the extraction of the fungicide cycloheximide using methylene
chloride as a solvent. Parameters are:
feed concentration
cF = 15 g L-1
fraction lost
u = 0.02
number of stages
N=4
= 23
feed flow
VW 10 m3 hr -1
Calculate the required solvent flow. (You need to solve a fifth order equation. If you do not have a solver, make a plot to find the relevant root.)

7 Crystallize
Our next separation method is crystallization, where we form a pure solid
out of a liquid mixture. We introduce it using the crystallization of citric
acid behind the extraction that we discussed earlier.
In the Lab
In a crystallizer, the liquid is supersaturated with the product, which precipitates in the form of crystals. In the lab this might be done in a stirred
beaker (Figure 7-1). We separate the crystals using a filter or a centrifuge.
Here, a single stage is often sufficient to get a pure product. However, a
great deal of the product may remain in the liquid.
Two common ways of saturating the liquid are:
x
cooling, to reduce the solubility of the product, and
x
evaporation of the liquid to increase the product concentration.
(The two methods are often used simultaneously. Evaporation then also
cools the product.)
Figure 7-1 A simple crystallization
Crystallization and the precipitation process that we have seen in lesson 4

are similar. However, in precipitation we add a precipitant to supersaturate the solution. In crystallization there is no such addition. (In downstream processing you may see the term reactive crystallization instead
of precipitation.)
(Super)Saturate
The thick line in figure 7-2 gives the solubility of anhydrous citric acid in
water as a function of temperature. The solubility rises gradually with in-
Figure 7-2 Solubility of anhydrous citric acid

creasing temperature. Citric acid has three acid groups, so it is polar. It
will not be surprising that it has a high solubility in water at the higher
temperatures on the right of the diagram, the saturated solution contains
more acid than water.
Below the saturation curve, the solution is undersaturated. If we bring a
crystal of citric acid in such a liquid, it will dissolve. Solutions with a
composition above the saturation line can exist, but only temporarily.
Such supersaturated solutions are unstable. If you leave them for a sufficiently long time, crystals will form, and eventually the supersaturation
will disappear. (However, for small supersaturations this can take a long
time).
If you bring a crystal of citric acid in a supersaturated solution, it will
grow. The rate of growth increases with increasing supersaturation, but is
usually low crystallization is a slow process.
As said above, the diagram is for anhydrous citric acid. Citric acid can
also crystallize as the mono-hydrate, in which each acid molecule carries
a single water molecule into the crystal. The two types of crystals have
different properties. The mono-hydrate is stable at temperatures below 38
o
C; at the temperatures in our process you can expect the anhydrous form.
Size Distribution
Let us look again at our simple lab crystallization (Figure 7-3). We start
with a saturated solution and add small seeds. When water vaporizes, the
solution becomes supersaturated and the seed crystals will grow. If all
crystals would grow at the same rate, we would end up with a product
with all particles having the same size. In reality we can only try to approach this ideal. (The particles in the figure are drawn too large: crystals
produced in industry are mostly in the range of 0.1 to 1 millimetre).
Figure 7-3 Idealized batch crystallization
The most important reason why we get a particle size distribution is the
formation of new nuclei. In high shear regions near a mixer or pump impeller, bits of crystal get knocked off, so new small crystals are formed.
Nuclei can also form on bits of dirt in the solution, or on the solid surfaces
of the crystallizer. This happens especially when high supersaturations are
used. Ways to avoid these problems are:
x
careful, low speed mixing,
x
clean equipment and solutions and
x
low supersaturations.
Unfortunately, these measures are not always practical.

7 Crystallize
Of course, large volumes of crystal are not produced as batches in beaker
glasses. Most industrial crystallizers are continuous vaporizers such as
shown in Figure 7-4. In such vessels, the contents have a residence time
distribution. A part of the fluid entering will leave almost immediately;
other parts may stay quite a long time. This again increases the width of
the particle size distribution. This effect can be counteracted by using
several crystallizers in series. There are also other reasons for doing that,
as we shall see further on.
Figure 7-4 Particle sizes in a continuous crystallizer
Seed the Feed

You do not have to seed a solution to get crystals. If you increase the supersaturation sufficiently, nuclei will form spontaneously. However, this
is a poorly controlled process: it often results in far too many nuclei,
which then give a product that is too fine to handle. To avoid this, many
industrial crystallizers are seeded: we add a certain number of nuclei in
the hope of getting the same number of particles.
To seed, we need a seed solution and you might wonder how this is
made. The first thing to realize is that you only need small amounts of the
seed solution. A seed might have a diameter of ten micrometres. If we
grow this to a millimetre it will have a volume of a million times the original seed. So we can use lab techniques to form seed solutions and lab
techniques are more flexible than large scale ones. We might form the
seed suspension by grinding large crystals and carefully adding water.
The very fine particles will dissolve first and we can control the numbers
through the addition of water.
Grow Crystals
Crystal growth occurs on the surface of the crystal where you will find
steps with the height of one molecule (Figure 7-5). Dissolved molecules
tend to adsorb at kinks in the step. This is a dynamic process in a saturated solution as many molecules adsorb as desorb. In a supersaturated
solution more molecules adsorb and the crystal grows. It will be clear that
the rate of growth increases with increasing supersaturation but it also
depends on the density of kinks on the surface. This in turn depends on
the history of the particle on how it has been formed. Crystallization is
slow because it only happens at the kinks, and these are small in number.

Figure 7-5 Crystallization on a molecular scale
If a different molecule can fit in a kink, it can hamper further growth.

Only a small number of molecules is required to stop adsorption this is
why traces of contaminants or additives can have a large influence on the
rate of crystal growth.
Molecules are added one by one, giving a regular crystal lattice. This regularity gives the high purity of crystalsforeign molecules are excluded
unless they fit closely in the kinks. However, the purity of the whole separation is usually governed by other factors:
x
Some liquid will adhere to the crystals in the filter.
x
At high supersaturations inclusions of liquid can form.
Contaminants may adsorb differently on the different faces of a crystal.
These will then grow at different rates and this will change the shape of
the crystal. Figure 7-6 illustrates this for a hexagonal crystal. If the contaminants adsorb on the top and bottom of the crystal we get flat platelets;
if it adsorbs on the sides we get needles. Both shapes are fragile, and for
most products we do not want these.
Figure 7-6 How different shapes can be formed
The hexagonal form is one out of several possibilities. Do not expect to

see much of these regular forms in industrial batches of crystals. There
the particles are usually battered, broken and deformed.

7 Crystallize
In the Plant
In industrial crystallizers solvent is evaporated to get cooling. The equipment often looks like that in figure 7-7. The crystallizer is the vessel in
the middle filled with crystal slurry. This slurry circulates through the
heater at the lower left; the rate of circulation is high, so that the whole is
well mixed. Water vapor is drawn off from the top surface of the slurry
which is the only part of the vessel where the liquid boils. The water vapor condenses in a separate heat exchanger. The feed, seeding and slurry
removal are continuous.
Figure 7-7 An industrial crystallizer
As we shall see in a moment, crystallization requires a large heat input.

This can be reduced by using multiple stages (Figure 7-8). This works in
the same way as the multi-stage evaporation that we saw in lesson 4. The
vapor (steam) from stage one is used to heat stage two; that from stage
two is used in stage three and so on. The first stage works at the highest
temperature and pressure; the last stage at the lowest. With three stages
we only need about one third of the energy required for a single stage
but the equipment is more expensive. Several stages in series also provide
a pattern that is close to plug flow, giving a narrower distribution of the
crystal sizes.
Figure 7-8 Multi-stage crystallization

Flows in Our Process
We finish with a rough look at the flows in our crystallization step (Figure
7-9). It is perhaps worth noting that this is not a counter-current process
it is simpler. The feed consists of water and acid coming from the reverse
extraction. In the ideal case all water will be evaporated, and all acid will
be removed as a dry powder from the filter. So the acid flow is 1 kg s-1;
the water flow 0.6 kg s-1. Roughly, of course.
Figure 7-9 Flows in our process
In a single stage process, the heat required to vaporize the water is:
Q
mW 'h
0.6 kg s 2.2 MJ kg
-1
-1
1.3MW
Here 'h is the enthalpy of vaporization. In a three stage process, we need

a little more than one third of this flow, so about 0.5 MW. The recycle of
mother liquor from the filter to the crystallizer turns out to be large
(Figure 7-10). The liquids in the crystallizers have compositions that are
close to saturation. From Figure 7-2 we find that the acid concentrations
are 0.81 and 0.71 kg L-1 in the first and last crystallizers. So we crystallize
(0.81 0.71) = 0.10 kg of acid for every litre passing through. To crystallize 1 kg s-1 we need a flow of 10 L s-1.
Figure 7-10 The recycle
Some Problems
In processes with recycle flows, you must be on guard for accumulation
of trace amounts of impurities. Ones that come to mind here are:
(1) Non-polar contaminants can accumulate in the solvent used in the ex
7 Crystallize
traction. These can disrupt the operation of the settlers, and may require
cleaning or replacement of the solvent.
(2) Dissolved gases such as air can accumulate in the water recycle from
crystallization to the reverse extraction. These will disrupt the operation
of the condensers. The gases can be removed with a small vacuum pump.
(3) All sorts of rubbish could accumulate in the recycle from filter to crystallization. You might need a small bleed to keep these under control.
In our calculations we have only considered a single process scheme. For
a real design you would want to consider many alternatives, so that you
can choose one of the better ones. (There is seldom a single optimal designthere are nearly always trade-offs.)
Further Reading
M. R. Ladish: Bioseparations Engineering, Principles, Practice and Economics, Wiley-Interscience, 2001, Chapter 4
Exercises
7-1 In a Crystallizer
We consider a batch crystallizer. It contains one cubic metre of saturated
citric acid solution at 100 oC:
acid concentration
cS = 800 kg m-3
The densities of solid citric acid and the acid in solution are not quite the
same, but we will neglect the difference. Take as densities for the acid
and water:
UW = 1000 kg m-3
US = 1400 kg m-3
(a)
(b)
(c)
(d)
(e)
Calculate the volumes of water and acid.

If we evaporate one half of the water, how much acid will crystallize?
Assume that the crystals are uniform spheres with a diameter of
1 mm. (They are not!) Calculate the number of crystals.
What is the volume of crystals in the slurry?
How many seeds would have been required to achieve this?
7-2 Seeding
Assume that we have spherical seeds with a diameter of ten micrometres.
What will the volume and mass of the seeds be for the crystallization in
the previous exercise?

7-3 Crystal Growth
A warning: the description here of crystal growth is too simple.
Crystals grow in supersaturated solutions. A measure for the supersaturation is:
V
c cS
cS
'c
cS
We will take the growth rate to be the same for all crystals and to be proportional to the supersaturation. We also take the crystals to be spherical.
Their radius increases linearly in time:
r t
106 ms-1
k Vt with k
We again consider the citric acid crystallizer of the previous exercises.

How long will it take to form crystals of 1 mm diameter with supersaturations of 0.1 and 0.2?
7-4 Nucleation
Nucleationthe spontaneous formation of small new crystalsis influenced by many factors. These make it almost unpredictable. However,
there is one factor of such an importance that we spend an exercise on it.
The rate of nucleation (nuclei formed per cubic metre of suspension per
second) often follows an Arrhenius type of equation in the supersaturation:
c
rN k1 exp k2 S
'c
(a)
Plot the rate against the value of 'c using the parameters
k1 109 s-1
(b)
cS
800 kg m-3
In our previous examples we considered a crystallizer of one cubic

metre. This was to form crystals of with a diameter of a millimeter.
The number of seeds and the growth time were:
NS
(c)
k2
5.5 108
tG
2.5 103 s
During growth, new nuclei are formed. At which value of 'c will
the number of nuclei formed be as large as the number of seeds?
Compare what happens with small and with large supersaturations.
7-5 Four-stage Crystallizer

Let us add a fourth stage to the plant in Figure 7-8. This last stage is to
operate at 40 oC. (This might require a rather expensive refrigeration, but
we will not worry about that.) As before, the mass and volume flows into
the system are:
(a)
mW
0.60
0
60 kg s-11
mA
1kg s-1
VW
0 6 Ls-11
0.6
VA
0.6 Ls-1
Assuming that the liquids are saturated, what is the citric acid con-

7 Crystallize
(b)
(c)
centration in the liquid of the first stage? And of the last stage?
The recycle of mother liquid from the last stage to the first stage is
large compared to the vapor flow leaving the crystallizers. So you
may assume that the volume flow through the crystallizers is constant. How large does this flow have to be for the required production of acid?
How much water is to be vaporized? How much heat do we need
for this in our four-stage system?
7-6 Size Distributions

Consider a continuous single stage mixed crystallizer. It is being seeded
with equal spherical seeds that are much smaller than the average size of
the crystals grown. Growth stops when a crystal leaves the crystallizer
and is filtered out of the suspension.
Fluid entering the mixer is rapidly dispersed throughout the vessel. So
some parts are immediately transported to the outlet. Other parts will stay
much longer in the vessel. This can be described by a residence time distribution. For a mixer the fraction of the fluid (and of the particles in it)
having a residence time between t and t + dt is:
t
exp dt
W
Many particles have a short residence time; few a long one. The average
residence time is W.
We assume that the rate of increase of the diameter of all particles is the
same. The diameter as a function of time is then:
kt
The diameter of a crystal with the average residence time is:
d0
kW
We often need diameter distributions. There are many of these. For our
system two common ones are:
x
the number distribution of the diameter
fN d
d
exp
d0
3
x
the volume distribution of the diameter
Take d0 = 10-4 m
(a)
(b)
(c)
(d)
fV d
d
1 d
exp
6 d0
d0
Plot the number distribution over the range 0 < d < 15d0.
Do the same for the volume distribution.
Why do the two distributions differ so much?
At which diameter is the peak of the volume distribution?

8 Distill
Distillation is used to recover solvents used in extraction and crystallization and to purify bio-produced solvents such as the alcohols. In distillation the two phases used are a liquid and a vapor.
Distillation is the only separation process in biotechnology that can be designed without doing experiments. Once you know the composition of the
feed and the desired products, the design can be largely done by a specialist on a computer. Even so, it is sufficiently important that every bioprocesser should understand how it works at least qualitatively.
Methanol-Water
We introduce distillation using the distillation of methanol and water
(Figure 8-1). Methanol is not an important bio-product, but its distillation
is simpler than that of the other alcohols that we consider further on. So it
is a better starting point.
Figure 8-1 Distillation of methanol

Some Terms
To keep things simple, we will assume that the feed consists of water and
methanol only. As you see, the distillation consists of two columns,
known as the stripper and the absorber. The feed, which is liquid at its
boiling point, goes into the top of the stripper (the lower column) and
flows downward. At the bottom, a part is vaporized in a boiler: this goes
upward, giving counter-current flow. The vapor continues upward
through the absorber. At the top it is condensed: the greater part of the
liquid formed flows down through the absorber, again giving countercurrent flow. This liquid, the reflux, then joins the feed at the entrance of
the stripper. This cyclic flow from boiler to condenser and back again,
can be large compared to the product flow.

Figure 8-2 Sieve trays
Sieve Trays
Most distillation columns use a construction known as a sieve tray
(Figure 8-2). This consists of horizontal plates with holes of about a centimetre in diameter. The vapor flows upward through the holes, which
prevents liquid from flowing through the holes. Liquid flows on to the
tray through a horizontal slit at the side, then across into a downcomer.
This leads to the opening slit for the tray below. A column may consist of
tens of trays.
The flow on the trays is wild and turbulent as in a heavy storm. The liquid is continuously flung upwards as drops by the vapor jets issuing from
the holes. This gives a good contact between the two phases.
The boiler and condenser have a construction similar to that of the evaporators that we have seen earlier: we will not further discuss them.
As said, the two columns are counter-current separators. Methanol is
more volatile than water, so it dissolves better in the vapor; it is stripped
out of the liquid. Water is absorbed in the liquid and tends to move
downward. If there are enough trays, the bottoms contain hardly any
methanol but nearly all the water from the feed. In an ideal distillation,
the distillate at the top would contain hardly any water and would be
pure methanol.
Find the Flows
The components in the distillations used in biotechnology have roughly
the same molar heats of vaporization: about 40 kJ mol-1. As a result the
molar flows of vapor and liquid are almost constant throughout a column:
vaporization of methanol is compensated by condensation of an equal
molar amount of water. This is one of the reasons why distillation is usually described in molar terms, so using molar flows and mole fractions for
compositions.
Figure 8-3 shows the symbols of all flows: the feed, the bottoms, the distillate and so on. These symbols are only used in this lesson. The reflux
going back into the top of the upper column is a multiple of the distillate
flow:

8 Distill
Figure 8-3 Flows in symbols
nL
RnD
The multiplication factor is the reflux ratio. The vapor flow into the condenser is:
nV
(R 1))nD
Figure 8-4 Flows in mol s-1
Figure 8-4 shows the molar flows in the system for a feed of 100 mol s-1.
The feed only contains a small fraction of methanol, perhaps one tenth.
This leaves as distillate: the top flow is 10 mol s-1. This about 10 000 tons
of methanol per year. Nearly all the water leaves at the bottom, so the bottoms flow is 90 mol s-1. As we shall see in a moment, we need a high reflux ratio for this kind of separation; we have chosen a value R = 5. This
gives a liquid flow in the absorber of 50 mol s-1; the vapor flow must then
be 60 mol s-1.

The feed to this kind of distillation column is usually brought in as a liquid at its boiling point. This goes down into the stripping column. So the
liquid flow in the stripper is equal to the reflux plus the feed, here 150
mol s-1. The liquid remaining after the bottoms has been drawn off is vaporized. This gives a vapor flow of 60 mol s-1: the same flow that we need
in the absorber.
At this point we can estimate the main energy input required for the process. This is the heat of vaporization required in the boiler. Distillation
uses a lot of thermal energy:
nV 'H
60mol s 40kJ mol

-1
-1
2.4 MW
We can also estimate the cross section of the columns. A gas flow of 60
mol s-1 corresponds to a volume flow about 1.5 m3 s-1. Trays in distillation columns at near atmospheric pressures can handle between 1 and 2
m3 m-2 s-1, so we need a cross section of the order of 1 m2.
Separate
To see why we require a high reflux ratio, we need to look at the vapor liquid equilibriums of methanol-water mixtures. In distillation these are
often summarized in an x-y diagram (Figure 8-5). The mole fraction of
methanol in the liquid is plotted along the x-axis, that in the vapor along
the y-axis. The lower left hand corner with x = y = 0 gives pure water; the
upper right hand corner with x = y = 1 gives pure methanol.
Figure 8-5 Equilibrium x,y-diagram
Methanol has a boiling point of 69 oC; water of 100 oC. In distillation this
is regarded as a fairly large difference. As a result, methanol is more volatile than water over the whole range of compositions and y > x (except in
the corners). Near the water corner, methanol has a low concentration.
Here we can approximate the equilibrium as a straight line, with a slope
of about 10. This is the mole-fraction vapor-liquid distribution coefficient' of small amounts of methanol in water. In the methanol corner, water has a low concentration. There the slope of the equilibrium line has a
value of about 0.4-1. This 0.4 is the mole fraction vapor-liquid distribution coefficient of small amounts of water in methanol.

8 Distill
Figure 8-6 The water absorption column
To understand the separation, we can best regard the two columns as separate processes (Figures 8-6 and 8-7). The top column removes traces of
water from methanol gas by absorbing them in the liquid. The separation
can be described by an absorption factor (which is similar to the extraction factor that we saw earlier). To get a decent separation this has to be
larger than one (and a bit more than that because the equilibrium line is
not really straight). With the chosen reflux of 50 mol s-1, we get an absorption factor of about 2. We have already seen that this is sufficient for
a decent separation in the lesson on extraction.
The number of trays required largely depends on how pure we want the
top product. Each tray can roughly be considered an equilibrium stage.
For this separation you might expect the number of trays to be in the tens.
Figure 8-7 The methanol stripping column
The bottom column strips traces of methanol out of a water flow. Here
the value of the stripping factor is about 4: higher than the absorption factor. So you will not need as many trays in this column.

Complications
The gas-liquid equilibriums of many systems in biotechnology are complicated. We will only consider them semi-quantitatively using methane
and four water-alcohol mixtures as examples. (These are all of some importance in biotechnology.)
Volatility
The key concept is volatility. A component is volatile if it has a low boiling point. It then tends to go into the gas, so it has a high gas-liquid distribution coefficient. The three things that mainly determine the volatility of
a component are:
x
the polarity of the component ,
x
the polarity of the other component (solvent) and
x
the size or molar mass of the molecule.
Polar and Apolar
We start this section looking at a rather extreme mixture: that of methane
and water (Figure 8-8). This is to introduce the concept of polarity: methane plays no role in our distillations. The two molecules have almost the
same size. Even so, methane is far more volatile than water: the boiling
points of the two differ by 261 degrees.
Figure 8-8 Water and methane
This difference is due to local charges (or poles) on the water molecule.
The oxygen atom is electronegative and it pulls electrons away from the
hydrogen atoms. The positive and negative charges on different water
molecules give an extra attraction which decreases the volatility. Water is
a very polar molecule. Methane has hardly any polarity.
Polar and apolar molecules dont like each other. A mixture of the two
immediately falls apart (Figure 8-9). So methane hardly dissolves in liquid water - and water hardly dissolves in liquid methane (which only exists at very low temperatures). In gases, where molecules hardly interact,
methane and water do mix freely.
Figure 8-9 Water and methane demix

8 Distill
Four Alcohols
Alcohols have both polar hydroxyl and apolar alkyl groups (Figure 8-10).
Of the four shown below, methanol is the most polar, butanol the least.
The boiling points of the pure liquids increase with molecular size, but
here the increase is partly counteracted by the decreasing polarity.
Figure 8-10 Four alcohols with their boiling points
The differences in polarity between water and the alcohols already have
an effect in the methanol-water system. The dotted line in Figure 8-11
shows what the behavior might be like if there were to be no differences.
In reality, traces of methanol in water are squeezed out a bit, and so are
traces of water in methanol. This causes the effects shown:
Figure 8-11 Effects of polarity on methanol-water
The difference in boiling point between water and the other alcohols is
less than for methanol. Also, the differences in polarity are larger. As a
result, the volatility of traces of water can invert. This is what happens at
high alcohol concentrations in mixture of water and ethanol, and also in
mixtures of water and propanol (Figure 8-12). In these two mixtures there
is a point where the liquid and gas have the same mole fractions. This
Figure 8-12 Diagrams of four alcohols with water

point is known as an azeotrope. The important thing to know is that you
cannot separate across an azeotrope using simple distillation.
In mixtures of water and butanol the volatility of butanol inverts in the
water corner. This system demixes: it has two liquid phases over a range
of the liquid mole fractions. As we shall see, there is a trick to separate
such a hetero-azeotrope by ordinary distillation.
Bio-ethanol
In real processes there are further complications. We will illustrate these
using the example of the production of bio-ethanol fuel (Figure 8-13).
This is currently one of the largest applications of biotechnology.
Figure 8-13 Separations in bio-ethanol production
Byproducts
Bio processes never produce a pure product. The broth in the fermenter
will contains micro-organisms (here yeast). These can be separated with a
filter or a centrifuge. Some smaller byproducts are shown in Figure 8-14.
Figure 8-14 Some byproducts of bio-ethanol
Despite the two oxygen atoms, carbon dioxide is apolar and volatile. It
will end up in the top of the distillation column - too much of it will disrupt the condenser. Fortunately most carbon dioxide can be removed by

8 Distill
heating the feed just above its boiling point and flashing off some gas.
Acids such as the acetic acid shown are polar - especially at low concentrations where they ionize in water. They end up in the bottoms and can
be a nuisance if the water is to be recycled to the fermentation.
Further problems are by-products such as the higher alcohols of which
pentanol-1 is shown. These are volatile in water in the lower column, but
not volatile in the ethanol in the upper column. So traces tend to accumulate in the middle part of the distillation, and can eventually disrupt operation. The trick is to remove this fusel oil as a small side stream at an appropriate point along the columns.
Pervaporation
As said, ethanol-water has an azeotrope. This is at a liquid mole fraction
ethanol of 0.89 (a mass fraction of 0.96). To remove the last traces of water, you need extra process steps. One possibility is shown above. It uses a
membrane that is permeable for water, but much less so for ethanol. A
vacuum is drawn behind this pervaporation membrane by condensing the
water that passes through. This small water flow (which contains some
ethanol) is returned to the distillation.
We must also mention that the mole fraction of ethanol in the feed is not
high (typically about 0.05) and that high reflux ratios (say 20) are necessary to approach the azeotropic top composition. As a result a large part
of the energy content of the product is required to run the distillation.
Butanol-Water
Production of butanol by fermentation was abandoned in the 1960s because of competition of chemical routes using oil. There is a renewed interest in bio-butanol as oil prices are rising. The process has an interesting
distillation section which makes use of demixing (Figure 8-15).
This process uses three columns. The first two are water-rich, the third is
butanol-rich. In the first two columns butanol is the volatile component: in
Figure 8-15 Distillation of butanol-water mixtures

the third one water. The feed (which contains only a small fraction of butanol) enters above the first column. The two first columns increase the
fraction of butanol in the vapor until condensing causes the formation of
two liquids. One liquid is rich in water, the other in butanol. The two
liquids separate in a settler. The water-rich stream is refluxed to the water
-rich columns, the butanol-rich feed to the water stripper. Here water is
the volatile component, so it concentrates in the upper part of the column
and is then sent to the condenser and settler. Pure butanol is taken out as
bottom product.
Final Remark
Distillation is the most used separation process in the oil and petrochemical industry. It has been the subject of a large amount of research and development over more than a century. So you will understand that there is
more to say about it than we have been able to do in this lesson.
Further Reading
J.D. Seader, Ernest J. Henley and D. Keith Roper Separation Process
Principles, John Wiley & Sons, 3rd ed., 2010
Exercises
8-1 Feeds of a Methanol Distillation
Consider the methanol - water distillation below. The flows shown are
those that have been worked out in the text. The feed is a liquid at its boiling point; it contains a mole fraction 0.1 of methanol. It is fed into the top
stage of the methanol stripper. The water absorber has been designed such
that the liquid composition leaving the bottom stage is the same as that of
the feed entering. In this way, there is no loss of separation at the feed
point.
(a)
(b)
Is the feed to the stripper liquid or vapor? What is the composition?

The feed to the water absorber comes from below, from the top
stage of the stripper. What is its composition? Use the x-y diagram
to find this.

8 Distill
8-2 Methanol Distillation
This exercise considers the same distillation columns as the previous one.
The flows in the columns have the values:
feed
n 100 m
mol s-1
F
methanol stripper nL
150mols-1
150m
nV
60mols-1
60mo
water absorber
50mols-1
50mo
ncV
60mols-1
60mo
ncL
You may assume that the following mole-fraction distribution coefficients

are constant:
methanol in the stripper
KA = 10
water in the absorber
KW = 0.4
The numbers of stages in the two sections and the feed mole fractions are:
absorber
N' = 10
y'0 = 0.58 (water)
stripper
N=5
x0 = 0.10 (methanol)
You are to estimate the flow of water in the top product and that of methanol in the bottom product.
(a)
(b)
(c)
(d)
(e)
(f)
Start with the absorber and calculate the absorption factor.

Calculate the molar flow of water in the vapor feed.
Use the equation given to find the fraction of water left in the top
and from that the water flow.
Now the stripper column. Calculate the stripping factor.
Calculate the molar flow of methanol in the liquid feed.
Use the equation given to find the fraction of methanol left in the
bottom and from that the methanol flow.

8-3 Polarity of Acetone
Compare the properties of propanol and acetone in the figure below.
(a)
(b)
(c)
(d)
How do their molar masses compare?

How do their boiling points compare?
Do you think acetone is more or less polar than propanol?
Would traces of acetone in water have a high or a low volatility?
8-4 Flash Distillation of Carbon Dioxide

The broth from ethanol fermentation contains dissolved carbon dioxide.
This can disrupt the pervaporation unit behind the water absorber, so part
of it is removed in a single stage or 'flash' distillation. This consists of a
heater that brings the feed just above its boiling point and a vessel to separate the vapor and liquid leaving the heat exchanger.
To keep things simple, we take the feed to consist only of water and carbon dioxide. Parameters are:
x0 = 410-4
mole fraction CO2
liquid flow
nF 1000 m
mol s-1
(of CO2 at 100 oC)
K = 5000
8-5 At an Azeotrope
Why cant you distill through an azeotrope?
Consider the stage at the right where the
entering flows have an azeotropic composition.
(a) Are they at equilibrium?
(b) What happens when you contact
them?

8 Distill
8-6 Heat Re-use in Distillation
Distillation requires large amounts of heat. This can often be re-used,
sometimes several times. An example in our bio-ethanol process is the reuse of the heat from the bottom product of the ethanol stripper.
The bottom flow is about 95% of the feed. In the figure above we have
taken it to be 100 % of the feed to keep things simple. (This leads to a
constant temperature difference along the exchanger.) This bottom flow
leaves the column at one hundred degrees Celsius, and the heat content
can be used to heat up the feed to the flash up to - say - 95 degrees. A little steam injection will still be necessary, but much less than would be
otherwise.
(a)
The values of the flow of water through the exchanger and the heat
capacity are given below.
m
(b)
20 kg s-1
4000 J kg -1 oC -1
Calculate the heat flow in the exchanger.

The heat transfer coefficient is D = 1000 W m-2 oC-1. Estimate the
required heat transfer area.

9 Purify (1)
Sometimes we need products with a very high purity. This is so for many
medical applications where bio products are injected into the body. Some
contaminants may have to be removed down to parts per million, or even
parts per billion. Such extreme purifications are usually obtained with
sorption processes processes that use a solid auxiliary phase.
In the Lab
Paper Chromatography
Cut a strip of filtration (or blotting) paper, and draw a blue line across it
with a ball pen. Then dip the lower end of the strip into a little ethanol
(Figure 9-1). The ethanol wets the strip and creeps upward, driven by capillary forces. When the ethanol reaches the line, part of the ink dissolves
and moves along with the liquid. However, the constituents of the ink do
not all move with the same speed; some parts of the dye move more rapidly than others and the line broadens to form a band. You will see that
the colours of the front and back end of the band are different. The deep
blue has colour split. The front part is purple, the back part bluish green.
It looks as if the dye contains at least two components that we have separated with filter paper and ethanol.
Figure 9-1 Paper chromatography
This is a simple example of the sorption or chromatography processes that

are often used in biotechnology to separate and purify. In this example we
needed three things:
x the feed (the line of ball pen ink),
x a porous matrix (the paper) and
x a carrier liquid (ethanol).
We were able to separate because the two colours have different speeds.
This is because they adsorb differently in the paper.
Equilibrium Distribution
You can easily study adsorption with batch experiments in the lab. As
these are simple and only require small amounts of material, you can do
many of these at the same time (Figure 9-2). Make aqueous solutions with
a known concentration of the protein you want to absorb. Put known
amounts of these in stirred beaker glasses. Then add known amounts of
the sorbentthe solid particles in which the protein is to adsorb. The protein goes into the particles, but also starts moving back as its concentra-

Figure 9-2 Batch adsorption experiments
tion in the particle increases. After some time (say ten minutes) the flows
back and forth have become equal and we have equilibrium (Figure 9-3).
Analyse the liquids to find how much protein has entered the particles.
The ratio of the concentrations in and outside the particles is the distribution coefficient:
C
c
The distribution coefficient has a low value if the protein is excluded
from the particles. If the protein adsorbs strongly, it has a high value.
When the concentrations in the solid are low, the distribution coefficient
is often independent of the concentration. The relation between the two
concentrations is then linear:
C
With higher concentrations, all available sites for the protein may fill up,
and the isotherm can reach a plateau with a maximum concentration in
the particles:
C CM
The isotherm can often be described by a Langmuir equation:
C
CM
k1
c
k2 c
Figure 9-3 Distribution between particle and liquid

9 Purify (1)
Here k2 is the liquid concentration at which one half of the sites are occupied and k1 determines the initial slope.
If the protein binds strongly, the maximum concentration is already
reached at a low concentration in the liquid. The isotherm then becomes
step-shaped. Non-linearity of the isotherm has important consequences
for the behaviour of sorption processes.
Protein and Matrix
In biotechnology the matrix is often in the form of little spheres with a
diameter between 10 and 100 micrometres (Figure 9-4). These are
brought into a cylindrical column. The particles are often made out of
gel: we can take this to consist of a random structure of nanofibres. The
space between the fibres is filled with water these pores have dimensions of the order of 10 nm. They are large enough to hold protein molecules..
Figure 9-4 Dimensions in sorption equipment
Proteins travel through the column with different velocities due to different interactions with the matrix. A protein that binds, will have a lower
velocity than one that is repelled. Four important properties of the matrix
that influence the interaction are shown in Figure 9-5:
x the size of the pores,
x the polarity (hydrophilicity) of the pore surface,
x the charge of the pore surface and
x the shape of the sites on the pore surface.
In most separations all of these play a role. Even so, one of them is usually dominant. This leads to a subdivision of sorption processes:
x size exclusion, where molecules are separated on size,
x ordinary adsorption based on differences in polarity,
x ion exchange, based on differences in electrical charge, and
x affinity adsorption, where molecules attach to specific countermolecules.
All these processes use the same principles, but there are practical differences. We will discuss a number of examples.

Figure 9-5 Four separation mechanisms
The interactions between protein and matrix also depend on other factors:
x
the properties of the protein,
x
the concentration of the protein,
x
the concentrations of other substances,
x
the properties of the solvent,
x
the pH and ionic strength of the solvent and
x
the temperature.
Chromatography
In a simple chromatography column (Figure 9-6) there is a constant flow
of carrier liquid. We inject a pulse of the feed at the inlet of the column.
The feed contains two components A and B. B has a higher velocity than
A, so the peaks of the two substances move apart and we can draw them
off separately. The separation is not immediate: the peaks have a finite
width and have to be drawn apart. Also their width increases as they flow
through the column. The separation improves with a longer column and a
lower flow velocity.
A pulse only contains a small amount of feed. If we inject more, we get
bands (Figure 9-7). More column length is needed to separate the bands.
Figure 9-6 Chromatographic separation

9 Purify (1)
Figure 9-7 Ways to increase production
Even so, the rate of production increases when we use bands. Widening
of the bands is less important than that of peaks.
If we increase the feed concentration sufficiently, the process becomes
non-linear and the peaks become asymmetric. Again more column length
is required, but the production per volume increases again. We consider
this situation further on when we have analysed the propagation of concentration profiles through a column.
A final variant of chromatography uses a gradient of the properties of the
carrier liquid. Properties that we can vary are:
x the ionic strength,
x the pH and
x the polarity of the liquid.
Adsorption / Elution
Adsorption / elution (Figure 9-8) gives much higher throughputs per volume of column. This operation is done in several steps. We use it if the
desired substance A binds more strongly to the sorbent than a contaminant
B. In the first adsorption step, feed is pumped through the column. The
fresh column preferentially adsorbs A until the sorbent is saturated. We
wash out the traces of B that do adsorb in a separate washing step. Then
we elute by switching to a liquid with a different composition, pH or
temperature. This eluent or regenerant removes the adsorbed A from
the column. Finally we wash out the eluent. The process then starts anew.
In the adsorption step the different species usually travel through the column with more or less sharp fronts. So they can be separated easily. In
elution there is a tendency for the fronts to tail off as we have indicated
with the diffuse profiles in the figure.

Figure 9-8 Adsorption, washing and elution
Peaks and Bands

In this section we look at how peaks and bands move through a column,
and how they can deform. Parts of our treatment will have to remain qualitative an accurate description would require too much mathematics and
too much time.
There are two groups of phenomena:
x those related to the shape of the isotherm and
x those related to non-uniform flow and diffusion.
When describing the first group, we assume local equilibrium between
particles and surrounding fluid. This description leads to a handy concept:
that of the concentration velocity. The second group considers deviations from equilibrium we only look at these qualitatively.
Concentration Velocity
In sorption processes one needs to understand the slow movement of different species through a column. We first consider a single protein.
Our protein will have a certain concentration profile in the column. During adsorption, the concentration will be high near the column inlet and
low near the outlet. The point where the concentration has a certain value
shifts along the column. The speed of this point is the concentration velocity w, which is derived in the appendix Derivations.
j
w
I 1 I w
To practice using this relation, we look at a few special cases.
Linear Chromatography
In the columns used in the lab for chemical analysis (and in the process
known as size exclusion) we usually work in the low concentration range
of the isotherm. Here the isotherm is straight (Figure 9-9).

9 Purify (1)
Figure 9-9 Peaks with a linear isotherm
Now all concentration velocities have the same value, because the distribution coefficient is constant. So the form of the concentration profile
does not change as the protein moves through the column. A peak remains a peak; only its position changes. (This is not quite true. Exchange
of protein between particle and liquid takes some time. As a result the
concentration change in the particles lags a little behind that in the liquid.
This causes broadening of the peak. We do not further consider this here.)
If = 0, the protein is completely excluded from the particles and
j I
The protein now has the same velocity as the liquid. This happens when
the protein is larger than the pores of the gel, or when the proteins are excluded by electrical charges. Base peaks of such substances can often be
seen in size exclusion and ion exchange.
If >> 0, then nearly all the protein is inside the particles and
j
w
1 I w
The protein now moves much more slowly than the liquid, perhaps a hundred times. The difference increases if the channel volume fraction I is
decreased.
Non-linear Effects
When concentrations in the particles are high, the isotherms become nonlinear. Different concentrations then have different velocities. The slope
of an isotherm is usually (but not always) lower at higher concentrations.
As a result, a high concentration has a larger velocity than a low concentration. The first parts of peaks and bands are then sharpened: they form a
front (Figure 9-10). A symmetrical peak becomes more like a rightangled triangle. This sharpening not only happens in chromatography, but
also in adsorption. It is the reason why adsorption columns often fill up
neatly with the same high concentration of the adsorbed substance
throughout the column.

Figure 9-10 Fronts with a non-linear isotherm
The same reasoning tells us that the concentrations at the back of a peak
or band will be pulled apart (Figure 9-11). So a tail is formed. This tail
does have a finite length.
Figure 9-11 Tails with a non-linear isotherm
Interactions
When two or more substances move through a column simultaneously,
they can influence each other. This can become quite complicated we
only give a few comments on some common situations.
Take a desired product A and a contaminant B. At a certain point in the
column, A has a high concentration. A will then occupy a large part of the
capacity of the solid, so there will be less room for B. This increases the
concentration velocity of B. (There is a similar effect of B on A, but this is
less because of the lower concentration of B.) Figure 9-12 shows a few
situations where A binds more strongly than B. Then B moves ahead of A:
this effect is increased by the high concentration of A. In non-linear chromatography this compresses the triangle of the B so the form resembles
that of a trapezium. The separation improves. In an adsorption column, A
causes B to rapidly move ahead. The concentration of B can then become
much higher than that in the feed. The effect of A also improves the washing out of contaminant B as it increases the rate at which B moves out.
Figure 9-12 might give the impression that interactions always improve
separations. This is not so: in a column where the contaminant binds preferentially, the separation worsens due to interactions.

9 Purify (1)
Figure 9-12 Interactions between A and B
Further Reading
J. C. Giddings, Unified Separation Science, John Wiley & Sons, 1991
Exercises
9-1 Molar Mass and Diameter
Biologists and chemists describe the size of a molecule with its molar
mass. However, in separation technology it is often easier to think in
terms of the diameter of the molecule. The two are related. If we assume
that the molecules are spherical:
S
NA U d3
6
Avogadro number NA = 6.021023 mol-1
protein density U = 1400 kg m-3
M
(a)
(b)
Calculate the diameter in nanometres for molar masses M = 1, 10,

100 and 1000 kDa (kmol).
Compare these with the diameter of a small molecule such as water
with U = 1000 kg m-3 and M = 0.018 kg mol-1.
9-2 Langmuir Isotherm

The isotherm of a protein on a sorbent is given on the next page. We want
to describe this using the Langmuir equation.
(a) Determine the two constants.
(b) A low concentrations the isotherm is linear with a slope . How is
this slope related to the two constants?
(c) Determine the value of the slope.

9-3 Four Separation Mechanisms

You may only be able to finish this exercise well when you have finished
lessons 9 and 10. Even so, we would like you to try now.
Look carefully at Figure 9-5. Try to find mechanisms that obey:
(1) Adsorption on the fibers plays no role or a small one.
(2) The adsorption sites are small, so there can be many of them.
(3) The adsorption sites are large, so their number is limited.
(a)
(b)
Two of the mechanisms can give distribution coefficients that are

smaller than one ( < 1). Name one of them.
With which two mechanisms could you get the highest concentrations in the solid phase?
9-4 Peak Broadening

In the chromatography column below, the feed is a pulse containing two
components A and B. The resulting peaks are separated by their difference in concentration velocity. However, the separation is not immediate
because the peaks broaden in the column. The components to be separated
have to diffuse into and out of the particles; this takes a moment. This is
the main reason for broadening.
The width of the peaks increases with time as:

b t
kdw

t
Df
9 Purify (1)
Here k is a constant, d the particle diameter, w the concentration velocity
and Df the diffusion coefficient of the component in the particles. We can
also write this as a function of position:
b t
kd
wz
Df
For the questions below, use the parameters:

k=1
Df = 10-11 m2 s-1
d = 10 Pm
wA = j/10 wB = j/9.9
Z = 10 cm
j = 1 mm s-1
(a) Calculate the residence times of the liquid, of A and of B in the column.
(b) Find the distance between the peaks when B leaves the column.
(c) What are the widths of the peaks at this moment? Will they have
separated?
(d) Would the separation be better or worse if we use larger particles?
9-5 Concentration Velocity
A chromatography column has the following parameters:
fraction occupied by the channels I = 0.5
liquid flux
j = 210-4 m s-1
column cross section
A = 1 cm2
(a)
(b)
(c)
(d)
Calculate the volume flow of the liquid.

We compress the column contents so that the fraction of the channels decreases to I = 0.4. If we maintain the flow, how large will
the new velocity in the channels be?
What is the consequence for the concentration velocity of a protein
with = 0. Will the concentration velocity increase or decrease?
The same question, but now for a protein with >> 0.
9-6 From Band to Triangle

To do this exercise you need a copy of the next page. On the left side of
this page you see a chromatography column with flow downwards. A
band of red substance is injected at the top of the column. This has a high
concentration, so the isotherm is non-linear and the concentration velocity
increases with increasing concentration. As a result the front of the band
remains sharp and moves forward with the velocity shown. However, the
back is pulled apart, with the different concentrations having different
velocities. We have constructed the concentration profiles for two time
increments.
(a)
(b)
(c)
Continue the construction for the following two time increments.

Check that the red areas are all the same as that of the original band.
Continue the construction for the last time increment. Is the area
still the same? What do you think will happen?

10 Purify (2)
In this lesson we look at some important variants of sorption processes:
x
size exclusion,
x
ordinary adsorption,
x
ion exchange and
x
affinity adsorption.
We also consider how lab experiments can help design large scale versions of these steps in a process.
Size Exclusion
As you might have expected, size exclusion separates molecules mainly
on size. It is often used for the separation of proteins and other large molecules. The most common column materials are gels. These have a distribution of pore sizes. Small molecules can enter the gel and are retarded in
the column. Large molecules are retarded less, so we can sort the molecules on size via their retention times.
Distribution
In size exclusion, one tries to avoid adsorption on the internal surface of
the particles. Proteins remain in the pore liquid. The gel particles are very
open and compressible, so we cannot use long columns or high liquid velocities. Typical values of the parameters of such columns are:
particle diameters:
0.10.2 mm
column length:
0.31 m
liquid flux:
2510-5 m s-1
The gel in Figure 10-1 consists of fibres with a diameter f (for fibre).
The fraction of the particles occupied by the pores is H. (Note that this has
nothing to do with the fraction I of the volume of the column occupied by
channels.) The centre of a spherical protein molecule can only approach a
Figure 10-1 Protein in a size exclusion gel

fibre up to a distance equal to its radius. So only the light blue part of the
figure is accessible to the protein. This part is smaller for a larger protein.
Ogston has derived an equation for the fraction that is accessible for molecules with a diameter d:
2
d
exp 1 H 1
If there is no adsorption on the fibres, this is also the distribution coefficient. Here, this has a value smaller than one; the concentration of protein
is less inside the particle than outside. Note that the distribution coefficient does not depend on the concentration. So molecules with a given
diameter all move with the same velocity. The velocities do vary with the
size of the molecules.
All small molecules distribute throughout the channels and the pores. So
all small molecules undergo the same retardation and cannot be separated
on a size exclusion gel. Something similar applies to very large molecules. These do not enter the gel so they all occupy the same channel volume. A gel can only separate molecules with diameters near the thickness
of the fibres: in the middle part of Figure 10-1. In practice this means a
variation of diameters of a factor two to three, or of ten to twenty in molar
mass. The maximum molar mass that can be separated often forms part of
the name of a gel. For example, SephadexTM G100 can be used up to a
molar mass of 100 kg mol-1.
Separating Bands
In the lab mixtures are often analysed using chromatography. The feed
volume used is then very small: we inject a pulse. When producing biomaterials we need to process larger amounts. One way of increasing the
capacity is to inject a band instead of a pulse. Here we estimate how
much time or column length this separation requires.
The sample in Figure 10-2 contains two proteins: a small one A and a
large one B. We inject the sample in such a way that the flux j remains
Figure 10-2 Separating two bands

10 Purify (2)
constant during injection. This takes a time tI. The small protein has the
lowest concentration velocity wA; the larger one has the velocity wB. After
a separation time tS the front of band A has moved a distance wAtS. The
back of band B enters the column at time tI; after the time tS it will have
moved to position wB(tS tI). The two bands will have separated when
they no longer overlap, so when:
wB
wB tS tI wAtS or tS tI
wB wA
Here we have neglected widening of the band by diffusion. For a large
sample (large tI) we need a long time. We also need a large time if the difference (wA wB) is small. If the two velocities are equal, we cannot separate.
Ordinary Adsorption
A classic example of ordinary adsorption is the remove of traces of apolar
components from water using an apolar sorbent (activated carbon). In biotechnology this is known under the impressive name of hydrophobic interaction chromatography or HIC. With a polar sorbent such as silica gel,
you can remove traces of water from hydrocarbons.
In size exclusion chromatography, ion exchange and affinity chromatography some ordinary adsorption does occur in parallel with the main
mechanism. It is often less selective, so one tries to minimize its effect.
Ion Exchange
The third sorption process that we look at is ion exchange. Here the gel
contains fixed charges. These can be positive groups such as those of quaternary ammonium, or negative groups such as those of sulphonic acid
(Figure 10-3).
Figure 10-3 Structure of ion exchangers
Ions with the same charge as the gel (co-ions) are excluded by the matrix. They play no role in this story. The exchanger contains an equal
charge of counter ions which can move around in the pore liquid. These
counter ions can be small ions such as Na+ or Cl-, but they can also be
charged proteins. They can exchange with other ions in the channel liquid.
The Matrix or Gel
A matrix with a positive charge exchanges anions (which are negative).
Anion exchangers often have the letters A or Q (from quaternary ammoni
um) in their name. Cation exchangers have fixed negative charges: they
exchange positive ions. Their names often contain the letters C (cation) or
S (sulphonic acid).
Ion exchange gels are more open than the gels used in size exclusion:
they do not exclude proteins on their size, but on their charge. Proteins
can form a monolayer around the fibres at this point they might be occupying a tenth of the volume of the particle. The concentration of ions
inside the particle is then usually much larger than that outside: the distribution coefficient is much larger than one. So ion exchange gels have a
much larger capacity than size exclusion gels.
As we have seen earlier, proteins can have a positive or a negative charge.
Below their isoelectric point (pI) they are positive, at a higher pH they are
negative (Figure 10-4). The figure shows the capacity of an anion exchanger for a certain protein as a function of the pH. Below the pI the
protein has a positive charge and is excluded. Above the pI the inverse
holds and the capacity rapidly increases up to a maximum. The maximum
capacity is reached earlier when the ionic strength of the surrounding liquid is low. With a high ionic strength, small ions enter the gel and partly
displace the proteins.
Figure 10-4 Capacity of an anion exchanger
The capacity diagram of a cation exchanger mirrors than of an anion exchanger. Here proteins with a positive charge, so below their pI, are
bound.
The height of the beds used in protein ion exchange tends to be smaller
than that in size exclusion: say 0.2 m. Ion exchange gels can be made
more rigid than size exclusion gels, so we can use higher liquid fluxes,
say up to 10-4 m s-1.
Which Exchanger?
Proteins can bind to either an anion- or a cation-exchanger. Which one we
choose depends on several things. We illustrate this with an example.

10 Purify (2)
Most proteins are only stable in a certain range of pH values. (They deactivate at a lower or higher pH.) We must operate in a region where our
protein is stable. Figure 10-5 shows the stability zones of two proteins A
and B. A is the desired product; it is not important to retain B.
Figure 10-5 Which ion exchanger?
With a cation exchanger we can separate A and B on a column at pH = 5.

This is just inside the region where A is stable. B then binds and A is eluted. We can remove B at a pH > 7. In this operation A is not concentrated.
With an anion exchanger we can bind A at pH = 5. We can recover it at
pH < 4 and as we shall see in a moment we can increase the concentration
above that of the feed.
Concentrating
We can use the characteristics of ion exchange equilibriums to increase
the concentration of a product (Figure 10-6). The figure shows two columns schematically with the solid phase on the left, and the flowing liquid
on the right. We use an anion exchanger.
Figure 10-6 Concentrating
The feed enters at a high pH which gives strong binding of the protein. A
band of bound protein forms. Even with a low protein concentration in
the feed, we get a high concentration in the solid. Then we pass a carrier
liquid with pH near the pI of the protein. The isotherm is now much lower
and the concentration of the protein in the liquid becomes much higher

than originally. Deformation of the band decreases this effect. Even so,
we can often concentrate substantially, perhaps by a factor of ten.
We can also concentrate by playing with the ionic strength of the carrier
liquid. The composition of the carrier liquid can also be changed gradually with such a gradient it is often possible to separate several substances
and to concentrate them at the same time. We do not further consider this
here.
Affinity Adsorption
Proteins often have a specific form. They can fit precisely on certain other
molecules called ligands (often anti-bodies). A gel with ligands fixed to
its surface can adsorb such a protein very selectively it has a high
affinity for a single protein.
Using affinity we can often get a good separation with a simple adsorption-elution process. Here we look at a few characteristics of such processes.
The Sorbent
The gels used in affinity adsorption have even larger pores than those in
ion exchange. They need to hold not only the proteins that are being separated, but also the ligands (Figure 10-7). The ligands have to be able to
move a bit so that they can fit into the protein to be adsorbed. They are
fixed to the pore wall with a little molecular chain or spacer. Making
such a sorbent with spacers and ligands is difficult it is a subject on its
own.
The gel matrix is similar to that used in ion exchange and exclusionthe
gel is such that there is little interaction between the protein and the matrix. We are not free to choose the material for the spacers, so some nonspecific adsorption must be expected on the spacers.
Figure 10-7 An affinity sorbent

10 Purify (2)
Adsorption of Two Proteins
Consider two proteins A and B. A is the desired product, B a contaminant.
We assume that the two adsorb independently on different sites (A on the
ligands, B on the spacers). This is shown in Figure 10-8.
Figure 10-8 Isotherms of two proteins
A binds strongly, B does not. So the isotherm of A is step-shaped; that of

B is a straight line. A forms a band near the inlet of the column where the
concentration in the solid has the maximum (plateau) value:
cA
cA,0 CA | CA, M
The band has a sharp front; the concentrations ahead of the band are:
cA 0 CA 0
After a time t the band reaches the end of the column:
lC
t 1 I A, M
jcA,0
(With a step-shaped isotherm we cannot use the equation for the concentration velocity that we derived earlier. The equation above is derived
from a mass balance of the band of A.) We stop adsorption just before the
band reaches the end of the column (Figure 10-9).
Figure 10-9 Adsorption / Elution
The contaminant B binds much less strongly, so it has a much lower concentration in the solid. As a result, B moves much faster through the column:
j
wB
1 I B

Washing
After adsorption we wash the column with a liquid like the feed, but
containing neither A nor B. The product A is bound quite strongly, so it
hardly moves in this step. However, B has a fairly high concentration velocity. So it moves out of the column in a time:
t
l
wB
1 I B l
j
Eluting
After washing we have a column that only contains A. We now feed an
eluent or regenerant. This liquid must differ from the feed in some important way: it might have a much higher ionic strength, a different pH, or
it might contain a different solvent. The eluent must weaken the bond between A and the ligand, so that A starts moving through the column. The
concentration of A can become much higher than it was in the feed.
Just as in ion exchange, the concentration velocities are higher for high
concentrations than for low ones. So the front leaving the column is
sharp. The liquid leaving the column has the concentration in equilibrium
with the filled adsorbent (Figure 10-10). However, once the highest concentration has left the column, lower concentrations follow and we get a
tail. This tail can be quite long. It can be difficult to find a eluent that
removes the product from the matrix without deactivating it. Elution is
often more difficult than adsorption.
Figure 10-10 Difficulty of elution
After elution we need another washing step to remove eluent. The process can then start again.
Affinity Adsorption, Why Not?
Affinity adsorption seems to be a very attractive process:
x the sorbent is almost completely utilised,
x the product has a high purity and concentration,
x the process uses little washing liquid.

10 Purify (2)
Even so, the number of industrial applications is small. Reasons are:
x we need a specific sorbent for every protein,
x the sorbents are very expensive and
x affinity sorbents are often not stable.
Spacers often detach giving some leakage of the ligands and a deterioration of the sorbent performance. Work is going on to solve these problems.
In the Plant
Scaling up a sorption process is not difficult in principle. We increase
the rate of production by increasing the column cross section (Figure 1011). If you use the same particles, same liquid flux and same column
length you may expect the large column to behave as many small columns
in parallel. So it should give the same performance as the lab column.
Figure 10-11 Scaling up of sorption processes
This approach gives several problems:

x
In the lab we like to use fine particles. These can give a good separation on a small, short column.
x
Fine particles are not handy in a large plant. They clog easily and
tend to be expensive.
x
Short beds have to be widelike a pancake. It is difficult to get a
uniform liquid distribution in such beds.
If you can obtain large particles with the same equilibrium properties as
the fine ones, it is worthwhile to investigate their use, although scaling up
is more complicated (Figure 10-12).
Figure 10-12 Using large particles

When a peak passes through a column, the concentration in the liquid
changes slightly ahead of that inside the particlestransfer by diffusion
takes a moment. This causes broadening of peaks and reduces the separation. To get the same separation on different scales, you need to keep the
following ratio constant:
jd2
Z Df
constant
Here Df stands for the diffusivity of the exchanging substance inside the
particles. This Df will be the same on different scales if the sorbent has
the same structure. You see that you can compensate a larger particle diameter by using a larger column length and a lower velocity.
In production, columns have to run for much longer than in the lab. Their
performance can decrease due to fouling or deactivation of the sorbent.
Cleaning and renewal of the sorbent may be required. This can be sorted
out in the lab, but it requires runs as long as are to be expected in production and with the real feed.
Developing the Process
Sorption processes are complicated. There are many interacting variables,
and their effects are not always easy to understand. Development requires
a lot of trial and error and nearly always takes longer than expected.
Before we start doing experiments and calculations we first have to answer a few questions:
x
Which substance do we need to isolate?
x
How pure does this have to be?
x
Which contaminants give problems?
x
Under which conditions (pH, ionic strength and temperature) is the
product stable?
x
Which kind of sorbent do we hope to use?
x
Do we expect problems with the waste or the chemicals used?
At this point we have to be able to analyze and characterize our product,
otherwise experiments are not possible.
We now decide whether to bind either the product or the contaminant to
the column. If the amount of product is small and that of the contaminant
large, then we preferably bind the product. In this case we can often elute
the product with a higher concentration than it has in the feed. If the feed
only contains small amounts of contaminants, we will try to bind the contaminants.
Then we choose a matrix. The number of types and grades is bewildering:
we might start with a type with which we have experience. This could be
an ion-exchange matrix ion exchange is flexible and not extremely expensive. We will be using the matrix under conditions where the product
is stable. Also it is important that we use a matrix that can be obtained on

10 Purify (2)
the scale that we hope to use. (Matrices used for analytical columns may
have desirable properties, but they are usually too expensive for production.) We then will have to start trying out all kinds of things
Theory can help greatly during planning, experimenting and scaling up.
However, the theory is not sufficiently well developed to allow us to
avoid experimentation. (Usually, we will not know basic data such as
on the equilibriums sufficiently well anyhow.)
Simulated Moving Beds
When you need large amounts of product, it is better to run your process
continuously. (We have already seen some examples, such as in the production of citric acid and bioethanol. Here continuous operation gives a
better utilization of the equipment.) In steps such as extraction, it is not
difficult to pump the solvent around and something similar applies to the
vapors and liquids in distillation. In adsorption, however, the auxiliary
phase consists of solid particles, and it is not easy to move these around.
The solution is to use a simulated moving bed or SMB.
Think of the adsorption / desorption system as consisting of a large number of columns operated in parallel. This is the upper part of Figure 1013; it is similar to Figure 10-9, but with many more columns. Each of the
columns going from left to right, is a little further in the cycle of feed
elutewashfeed Now think of this band of columns being folded
into a cylinder as in the lower part of the figure. We can operate this system in two ways. In the first carrousel system the whole cylinder of columns is moved around, past the inlets and outlets of the system. The system is on a large rotating table. Such a construction is expensive and requires much maintenance.
Figure 10-13 Simulated moving beds (SMBs)

In the second rotating valves system we move the inlets and outlets past
the columns; the columns are stationary. In this system it is not necessary
to have the columns arranged as a cylinder: any arrangement will do. The
most common arrangement is to have all beds in a single vertical column.
The problem here is the complexity of the valves and the piping for the
many flows.
Both systems can be designed using data from batch adsorption / desorption experiments in the lab.
Further Reading
Gel filtration, Principles and Methods, Pharmacia LKB Biotechnology,
1999
Ion Exchange Chromatography, Principles and Methods, Pharmacia SE751 84 Uppsala, Sweden, 2002.
Affinity Chromatography, Principles and Methods, Amersham Pharmacia
Biotech Ed AC 2001
Exercises
10-1 In a Size Exclusion Column
Consider a size exclusion column with the following parameters:
channel volume fraction
I = 0.33
gel volume fraction
(1 - I) = 0.67
gel fibre diameter
f = 3 nm
pore volume fraction
H = 0.95
column cross section
A = 1 cm2
column length
l = 10 cm
liquid flow
V 0.03 m
mLs-1
(a)
(b)
(c)
(d)
(e)
(f)
Draw a bar of 10 cm length and indicate the following volume fractions on the bar:
the channels between the particles
the particles (including the pores)
the pores inside the particles
the fibres inside the particles
What is the average velocity of the liquid in the channels? (You
may neglect flow through the pores.)
Calculate the distribution of small molecules such as water.
Calculate the distribution coefficients of molecules with diameters
of 3, 5, 7 and 10 nm.
Calculate the residence times of these molecules assuming that they
do not adsorb.
Calculate the residence time of a molecule that does adsorb with a
distribution coefficient = 10. Compare!

10 Purify (2)
10-2 A Band Separation
This deals with the same system as in exercise 10-1. Now we wish to separate the smallest protein (with a diameter d = 3 nm) completely from the
other proteins. What is the largest sample volume that we can separate on
the column? Use Figure 10-2.
V 0.03 m
mLs-1
liquid flow
velocity smallest protein
wA = 341 Pm s-1
velocity second protein
wB = 375 Pm s-1
available length
l = 10 cm
10-3 Cation or Anion Exchanger?
You have a mixture of equal amounts of two proteins A and B that you
want to separate using ion exchange. The isoelectric points and the ranges
of pH where the proteins are stable are shown below. The desired product
is B (this is the opposite of the example in the book).
(a)
(b)
(c)
(d)
Would you prefer to have the product or the contaminant on the exchanger? Why?
Would that require an anion or a cation exchanger. Or could you
use either of the two?
At which pH would you adsorb?
At which pH would you elute?
10-4 An Affinity Separation

A feed contains two proteins A and B with concentrations:
cA = 5 g L-1
cB = 10 g L-1
The volume of the feed is
VF = 1 m3
A is the desired product. We want to separate on an affinity adsorbent

with the following properties:
CAM = 50 g L-1
B = 2
The volume fraction of the channels in the column is
(a)
(b)
I = 0.4
Which volume of the sorbent (the solid) is needed to take up all A?

What is the volume of the bed of particles in the column?

(c)
(d)
How much B will there be in the column when all A is adsorbed?

What is the minimal volume of washing liquid needed to remove
all B from the column?
10-5 Scaling-up of Chromatography

In this exercise we look at different ways of scaling up a chromatography
or adsorption column. In the first we use the same particles in the small
and in the large column; in the others we allow other particle sizes.
To get the same separation we need to keep the following ratio constant:
j d2
Z Df
constant
Here Df stands for the diffusivity of the exchanging substance in the particles. This Df will be the same on all scales (assuming that the sorbent has
the same structure, independent of the particle size chosen).
In the lab we get a good separation on a column with the parameters:
particle diameter
D = 20 Pm
liquid flux
j = 100 Pm s-1
column height
Z = 10 cm
column diameter
D = 1 cm
In the plant we need a throughput that is 10 000 times larger.
(a)
(b)
(c)
If we choose the same particle size and velocity in the plant, what
will the height and diameter of the column become?
We now assume that we can choose the particle size at will:
d'' = 100 Pm
We keep the velocity the same again. What do the column height
and diameter become?
To reduce the height of the column we reduce the velocity by a factor of 4. We keep the same particle size as in (b). What happens to
the height and diameter of the column is we want the same separation?

11 Process
Downstream processing usually requires many separation steps. These
have to work together as a whole to give a good product for a good price.
When developing a process you have to designfind a (hopefully) optimal combination out of the almost infinite number of possibilities. In this
last lesson we will give you a first impression of how this is done. We
consider two examples:
(1) the processing of an intracellular enzyme and
(2) the processing of the antibiotic cephalosporin.
The first is artificial and highly simplified. It will allow us to make some
rough quantitative process estimates. The second is part of the story of a
real project in which one of your authors has participated.
Quantify
Our first example is the processing of an intracellular enzyme. The production is to be large this will show the problems of scaling up. We will
describe the process steps using models from the previous lessons. In reality process design is never based solely on such models: most steps
have to be tested in the lab, with lab people and process engineers working together. We do need a computer model (using experimental parameters) to investigate different process variants.
Figure 11-1 The fermenter: mainly filled with water
The Broth
The important ingredients of the broth are shown in Figure 11-1. The
broth contains a large amount of water and very little product. In reality
there are far more components than shown: we have kept the number
down to six to keep this discussion manageable. The six components
(Figure 11-2) are:
x
a small component: water with dissolved salts and nutrients,
x
four different enzymes (of which one is desired) and
x
a large component: cells and cell debris.
The four enzymes have different molar masses, so also different sizes.
They also have a range of iso-electric points. The behavior of components
in a process is most easily understood using volumes: that is why we use

Figure 11-2 The six components
these here. For other purposes you may need masses or molar amounts.
The three are related by:
m
nM
UV
Enzyme 3 is the desired product. We want it to be pure. The table suggests that we might first separate components using size exclusion, followed by a separation on charge using ion exchange. This is quite well
possible in the lab (Figure 11-3). However, we want to process 100 m3 of
broth every eight hours. This yields about 100 tons of product per year.
Figure 11-3 The original process
Cell Disruption
Because the product is inside the cells we first have to disrupt them. We
will do this using a homogenizer. You will soon find that even small homogenizers have a large capacity and that the biggest problem here might
be to get sufficient feed to test the equipment in the lab. The energy consumption in the plant is also considerable, of the order of 3 MW
(megawatt).
Clarification
After the cells have been disrupted, we have to separate cell debris from
the broth. We will do that using a filter. Because the particles are fine
(about 0.1 Pm) filtration is slow. We only get reasonable results when using large amounts of filter aid. This is a porous material (for example diatomaceous earth) of particles that are much larger than the cell debris. Using a filter aid gives a voluminous filter cake that has to be washed to
avoid large losses of product.

11 Process
This filtering and washing is to be done on the drum of a rotating vacuum
filter. Our estimates show that we need a large filter. The biggest problem, however, is the filter cake. We will have to find some way of disposing it, and it does not look suitable for road building or whatever
Size Exclusion
The most important separations in enzyme processes are size exclusion
and ion exchange. (A third one, affinity chromatography, has already seen
a bright future ahead for several decades. It still seems to be too expensive for large scale separations.) We will use both methods, starting with
size exclusion.
Size exclusion separates molecules into different sizes. We need the fraction with the two equal sized enzymes 2 and 3. Calculations tell us that
the most difficult separation here is that between 2 + 3 and 4. This separation determines the length of the column and the amount of carrier liquid.
Ideally, size exclusion does not change the concentrations of the components. Each component leaves the column in the same volume of liquid as
it occupied in the feed. Here we have three fractions, which ideally will
give three times the volume of the feed. However in the real process we
need more liquid we will assume five times.
When size exclusion is scaled up, several things are kept the same:
x
the composition of the feed,
x
the kind of particle and its size and
x
the liquid velocity.
These rules lead to wide, flat beds in which it can be difficult to distribute
the liquid properly. In our process we need a bed area of one hundred and
ninety square metres! Our process would be a bonanza for gel manufacturers.
Ion Exchange
In the ion exchange section, we only need to separate the two enzymes 2
and 3. We will do that using an anion exchanger working at a pH of 5.
Under this condition, 3 binds strongly; 2 binds weakly. We elute with
a pH of 4. The column is chosen to have a capacity of three times the
amount of 3 in the feed: experiments would be needed to see whether
this is sufficient.
An ion exchange gel has a much larger capacity per volume than a size
exclusion gel. So the ion exchange column is much smaller than the size
exclusion column. We have assumed that ion exchange concentrates the
feed ten times. Again, experiments are required to see whether this is realistic.
The results of our estimates are summarized in Figure 11-4.

Figure 11-4 Volumes (m3) in the original process
How to Improve?
We now have managed to get 0.2 m3 of pure product in a solution of 4 m3
of liquid. In doing that we have produced the following amounts of waste:
x
54 m3 of filter cake,
x
500 m3 of carrier liquid from the size exclusion and
x
95 m3 of waste water from the ion exchange.
The waste has more than three thousand times the volume of the actual
product! In addition, we need a huge volume of expensive size exclusion
gel and a power of three megawatt for cell disruption.
Do you have any ideas on how to improve the process? Try to list (say)
five of them before reading on. Your ideas may be better than ours.
Our Suggestions
(1) Use ultrafiltration to remove part of the water in the feed to the size
exclusion. In this way we might be able to reduce the volume flow by a
factor of ten. The size of the size exclusion column goes down by the
same factor (it still remains large!). Also the volumes of the waste streams
are reduced (Figure 11-5).
(2) Change the sequence of size exclusion and ion exchange (Figure 116). The ion exchange will be more complicated, but we now only have to
separate two enzymes on the size exclusion column. The flows to the size
exclusion are again smaller, and the size starts to look acceptable.
(3) If we could modify the micro-organism such that it produces the required enzyme in a higher intracellular concentration, the flows again become smaller. (Figure 11-7 shows the effect of a five times higher concentration.)
(4) The filter cake gives big problems. We might try to adsorb the enzymes directly from the broth on an anion exchanger (so without filtration). The waste water could then be handled in a bio-treater. We look at
this again in the next example.
(5) We can try to flocculate the cell debris. If the resulting particles are
large enough, we may be able to filter them without using a filter aid.
Centrifuges usually pull flocs apart, but we might try special low shear
machines.
(6) We could also avoid the filter aid if our microbiologists could find
a large micro-organism that could be filtered directly and that would also
excrete the enzyme. (Unfortunately, many interesting enzymes degrade
outside the cell on the gas-liquid contact areas in fermenters.)

11 Process
Figure 11-5 The effect of ultrafiltration
Figure 11-6 A new sequence of ion exchange and size exclusion
Figure 11-7 A higher feed concentration
First Conclusion
This discussion has centred on a rather artificial process and the estimates
used have been rough. Even so, we hope that it has shown that:
x
Scaling up a laboratory recipe does not always give a good process.
x
Waste streams have to kept small in large processes.
x
We must try to reduce the volume flows in large processes.
x
We should try to avoid intracellular production and the resulting
filtration problems.

Run the Project

This real story deals with two antibiotics: penicillin and cephalosporin
(Figure 11-8). Both have a E-lactam ring, but the other ring has five members in penicillin and six in cephalosporin. Penicillin can be converted
into cephalosporin with the enzyme expandase.
Figure 11-8 Two antibiotics
Penicillin is produced by the mold Penicillium chrysogenum in high concentrations of 50..60 g L-1. Cephalosporin is produced by Cephalosporium acremonium, but in the much lower concentrations of 10..15 g L-1. We
would like to produce cephalosporin with the same high concentration as
penicillin. The idea is to clone the expandase enzyme into the penicillin
mold, with the hope that we will now get high concentrations of cephalosporin. This turns out to be the case. However, the product also contains
penicillin and this is not allowed in pharmaceutical cephalosporin. So it
has to be removed in the downstream processing.
This is going to be a large development. It will take years, require the
work of large numbers of expensive, specialized people, and cost perhaps
a hundred million euros. With such large projects it is necessary to explore many different alternatives before deciding on the final process.
Form Ideas
Here are a few observations that will lead to ideas before the process development starts. It is not clear beforehand which clarification method
will be best we consider four different ones. These are filtration, micro-
Figure 11-9 An expanded bed

11 Process
filtration, centrifugation and the use of an expanded bed. In the expanded
bed (Figure 11-9) the sorbent is in the form of particles of the order of a
millimetre. These are suspended in the broth, so the product can be adsorbed without using filters or centrifuges a huge advantage. However,
this is a new method and we must expect problems when developing it.
For the removal of the bulk of the penicillin there are several ideas. Penicillin dissolves well in organic solvents at low pH. This is in contrast to
cephalosporin, which forms zwitterions with both positive and negative
charges. So we might use extraction. As both substances can be in the
form of ions, it may be possible to separate using ion exchange. Finally,
penicillin is much less stable than cephalosporin, so it can be de-activated
thermally.
After the bulk separation, we expect to need further purification. As a
zwitterion, cephalosporin does not adsorb well on ion exchangers. Initially, it is thought possible to purify using nanofiltration, but the results are
disappointing. However, the analytical lab finds that it does adsorb on
materials with a low polarity, so it can be purified using hydrophobic interaction chromatography (HIC). This will have to be developed into a
large scale SMB unit.
The final purification will be by crystallization which is preceded by a reverse osmosis unit to reduce the amount of water handled.
Keep Track
The possibilities that we are considering are shown in Figure 11-10. The
number of combinations is 4 3 2 = 24 and each of these will have
Figure 11-10 Process routes considered

many variations. It is out of the question to investigate all of these in detail: we have to get information quickly and cheaply to help decide how
to go on. Sources of information are:
x
quick, simple experiments (often with equipment vendors),
x
rough estimates (as you have learned to make in this course!),
x
experience inside the company (very important!),
x
literature, especially patent literature, and
x
experience from others.
The others might be universities, equipment manufacturers, contractors,
customers, competitors You will have to pay for their experience, but
that is often cheaper and quicker than doing everything yourself.
The number of details to be sorted out is enormous. If you get into development projects you will find that the big challenge is to keep track. That
is: not to get lost in the many possibilities and their details.
The expensive part of the project begins during or after this exploration
phase. We have to get a pilot plant running, equipment tests, designs of
the plant, plant construction As said, you will not be doing this alone.
This project has worked out well: the plant was running after five years
and has been in operation for a decade. It produces more than 500 tons
per year of cephalosporin powder. It may well be in operation for another
decade. Not all projects are that lucky. The flow scheme chosen eventually is shown in orange in Figure 11-10.
Conclusion
In this course you have seen the important steps used in downstream processing. You have considered them in some detail and learned how to
understand them quantitatively. If you are involved in the development
and use of bio processes this is important. However, it is not enough.
In this last lesson we have tried to give an impression of how the subjects
of this course enter into development projects. Such projects are a huge
adventure for all people involved. The outcome is uncertain and there are
challenges, dead ends, pitfalls, rows and triumphs ahead. Dealing with
these is not only a matter of making the right sums You are warned!
Further Reading
J. Krijgsman et al
International patent application number: WO 98/48036
Improved process for the fermentative production of cephalosporin
E. van de Sandt et al
International patent application number: WO 99/50271
Novel process for the fermentative production of cephalosporin

Derivations
Here are a few derivations of equations used in the main text. They concern the models of
(1) the vertically mixed settler equation on page 22
(2) the disc-stack centrifuge sigma on page 25
(3) the filtration equation on page 28
(4) the flux equation for ultrafiltration on page 52
(5) the concentration velocity equation on page 100
Vertically Mixed Settler

In this model we assume rapid mixing in the vertical direction, so the
concentration of particles will not depend on height. However, it will depend on the horizontal position:
c x
We consider a vertical slice with a thickness dx. The flux of particles to

the bottom is given by the product of the local particle concentration c
and the settling velocity u. A balance of the particles entering and leaving
the slice then yields:
hv d c uc d x
dc
c
udx

v H
with the boundary condition

c 0
c0
Solving this equation yields:

c
c0
ul
1 exp
vZ
Disc-stack Centrifuge Sigma

The next figure shows two channels in the stack of a disc centrifuge. We
want to know when a particle that enters at the bottom of a channel (the
upper one) will just be separated. To analyze this, we only need to consider two radial components of the velocity, which are denoted as u and v.
The values u0 and v0 at the channel inlet are taken as references.
There are several simplifications in the analysis here: do not take it to be
fully accurate.

The velocity u is caused by the centrifugal acceleration:
a r
Z2 r
So it varies with position as:

u r
u0
r
r0
The velocity v is caused by the volume flow through

the channel (it is much larger than u). The cross section is proportional to the radius, so this velocity
varies as
r
v r v0 0
r
The two distances travelled in a small time increment have the ratio:
d zU
d zV
u
v
u0 r

v0 r0
The ratio of the total distances travelled is the average of this value:
zU
zV
1 1 u0 r
dr
r0 r1 r0 v0 r0
3
3
1 u0 r0 r1
3 v0 r02 r0 r1
The distances travelled by the limiting particle are:

h
zV r0 r1
zU
tan D
So the limiting velocity is determined by
3
3
1 u0 r0 r1
3 v0 r02 r0 r1
h
1
tan D r0 r1
or
v0
u0
3
3
1 r0 r1 tan D
3
r02h
g is the same as that entering the stack:

The flow through the centrifuge
V
v0 Sr0 Nh
and the velocity u0 of the particle entering is related to that of settling under gravity by
u0
ug
Z2 r0
g
Combining these last three equations gives:
S Z2 3 3
ug
r0 r1 tan D N

3 g

Derivations
in which we recognize the effective settling area or sigma:
S Z2 3 3
r0 r1 tan D N
3 g
The Filter Plot

Here we derive the equations used in Figure 3-17. Starting point is the
relation between the flux and the change in the filtrate volume:
1 dV
v
A dt
As discussed, flux and resistance are related by:
'p
v
K R rh
The experiment is done with a constant pressure difference 'p (one atmosphere or 100 kPa if a vacuum is used). Combining the two relations
and rearranging yields:
dV
A 'p
d t K R rh
The height of the cake increases as the volume of filtrate passed through
goes up:
VC CV
h
A
A
Inserting this yields the equation for the volume of filtrate as a function of
time:
AR
V dV
rC
A2 'p
dt
KrC
Using
a
AR
and b
rC
A2 'p
KrC
this simplifies to:
a V dV
bdt
with the solution:

V
a 2 2bt a
We can rewrite this solution in the form:
t V 2a
V
2b
So, if we plot (t / V) against V, we can expect a straight line. The interception with the vertical axis has the value (a / b); the slope has the value
1 / (2b). From these values we can calculate R and r:
A 'p
intercept r
K

2 A2 'p
slope
KC
Diafilter
We can remove salts and other small molecules from proteins by diafiltration rinsing out the salts with water. We add water, which is then removed with the salts via an ultrafiltration membrane. The membrane has
to have a high retention for the protein, but a low retention for the salt. In
diafiltration, the concentration of salt goes down exponentially with the
volume VW of water added:
c
c0 exp W 1 R
VF
VF is the volume of the feed.
To derive this equation we make the following assumptions:

x
the fresh water flow in equals the salt water flow out,
x
the feed flow is much larger than the permeate flow,
x
so the concentration c of salt is the same through the circulation.
The initial salt in the holding vessel is:
VH c0
The change of the content in time is:
VH
dc
dt
This is balanced by the flow of salt out:
VH
dc
dt
Solving this differential equation yields:
c
c0
Using
VW
gives the final relation:
c
c0

VP cP
VP 1 R c
exp P 1 R t
VH
VPt
exp W 1 R
V
F
Derivations
Flux in Ultrafiltration
The flux in ultrafiltration is assumed to be limited by the formation of a
gel precipitate on the membrane. This forms because protein is dragged
towards the membrane by the permeating liquid, causing an increase in
concentration near the membrane.
We take the liquid near the membrane to be a film with convection from
left to right, and diffusion of protein back into the bulk of the liquid. In a
steady state, the two effects must balance, which leads to the differential
equation:
dc
uc D
dz
with the boundary condition:
c G c0
The solution of this equation gives the concentration profile:
c
dc u
zG
dz o ln u
c
D
D
c0
At the limiting (maximal) flux, the concentration at the membrane (z = 0)
becomes equal to the gel concentration. This determines the value of the
limiting flux:
c
c
G
D cG
o uM
ln G uM
ln k ln G
D
G c0
c0
c0

In sorption processes one needs to understand the slow movement of different species through a column. We only consider a single protein.
Our protein will have a certain concentration profile in the column. During adsorption, the concentration will be high near the column inlet and
low near the outlet. The point where the concentration has a certain value
shifts along the column. The speed of this point is the concentration velocity w.
For slow movements of the protein we can derive a simple equation for
the concentration velocity. We assume that the concentration is uniform
in a given cross section. The concentration is allowed to vary along the
column in the z-direction. We consider a thin slice around a point z. The
slice has a thickness 'z. We need a few more symbols:
A area of the column cross section
c protein concentration in the liquid
C protein concentration in the particles

j flux of the liquid
I fraction of the cross section occupied by channels
1 I
fraction of the cross section occupied by particles
If particles are uniformly distributed in space then the fractions of cross

section and volume occupied by the particles are equal. The volume fraction of particles includes pores inside the particles.
At a certain moment t the concentration profile is given by the thick line
sloping downwards. A short time 't later, the profile will have shifted
over a small distance 'z. In this period the average concentration in the
slice changes from (c 'c / 2) to (c + 'c / 2). A mass balance over the
slice gives:

Derivations
accumulation
flow in flow out
flow in = jAI c 'c / 2 't

flow out = jAI c 'c / 2 't
accumulation
1 I A 'z 'C I A 'z 'c
(Accumulation occurs in both the solid and the liquid.) Overall:

jA 'c 't
or
j 't
1 I A 'z 'C IA 'z 'c

'C
I 'z
1 I
'c
The ratio of the two concentration changes is the slope of the isotherm:
'C
'c
dC
dc
This yields the relation for the concentration velocity:

w
'z
't
1 I w I

Answers
These are bare-boneanswers. Fully worked-out exercises can be downloaded from www.vssd.nl/hlf/d030.htm. They are in two forms:
x
pdf files that can be read with Adobe Reader (and many other programs)
x
mcdx files that can be used live in Mathcad Prime 2.0 or higher.
1-1
volume fraction of cells

mass fraction of solids
IC = 0.262
wW = 0.107
1-2
overall yield
new yield
production increase
increase of profit
y0 = 0.658
yN = 0.701
66 ton/yr
25 %
2-1
95 % release after about 1.8 hours
2-2
the remaining activity goes

through a maximum:
2-3
'p, MPa
'T, oC
60
35
0.52
80
40
0.73
100
45
0.87
120
50
0.93
140
55
0.90
the real pump

mass flow out same as mass flow in
there is no heat flow
energy balance
p
p
m T
T0 0 m T
T1 1 W
U
U
'T = 2.5 oC
temperature rise
batch bead mill
no change in pressure
no mass flows in or out
no heat flows
energy content does change
energy balance
mS S mLL
time to heat up to 40 oC
dT
dt
t = 300 s

3-1
settling velocity
solids fraction feed
volume fraction cake
solids in cake
maximum flow
flow area plug
flow area mixed
3-2
u = 33 Pm s-1
fSF = 0.036
fC = 0.16
fSC = 0.223
VM
3.3
3
3 10
1 4 m3 s-1
A = 83 m2
A = 384 m2
height vs. time
Hc
uH S 1 L
HcS
filling time
t = 5105 s
only the first part fills up
settler fills near the inletreduced capacity
z
3.7
3
7L
Ls-1 13.4 m3 hr -1
3-3
capacity
VM
3-4
sigma value
6 = 5644 m2
3-5
cake ratio
flux
filtrate volume
area required
time (3 cycles)
CR = 0.668
j = 30 Pm s-1
V = 6 m3
A = 78 m2
t = 2.4 hr
3-6
medium resistance
cake.. (per height)
required time
RM = 2.031011 m-1
rC = 3.61014 m-2
t = 2900 s
3-7
cake ratio
flux
volume flow
CR = 0.491
j = 640 Pm s-1
V 19.2 m3 hr -1
medium resistance
cake.. (per height)
RM = 2.141010 m-1
rC = 1.011013 m-2
4-1
water to vaporize
fraction vaporized
mW = 926 kg
f = 0.975
4-2
energy balance
mAhA1 mW hL0
air flow
mA
heat flow
3.93MW
3.93M
50/50 gives
ms
1.0kgs-1
1.0kg
60/40 gives
ms
1.5kgs-1
1.5kg
39.3kgs-1
39.3k

mAhA2 mW hV 2
Answers
T, oC
c/c0
20
1.00
30
1.00
40
0.92
50
0.05
60
0.00
4-3
4-4
(a)
change along
heat flow
'Talong = (90 - 20) oC = 70 oC

Q
280kW
driving difference
area
'T = 10 oC
A = 280 m2
(b)
log mean..
average..
LMTD = 33.7 oC
'Tav = 45 oC
(c)
log mean..
average..
LMTD = 21.6 oC
'Tav = 25 oC
(d)
heat flow
log mean..
area
Q 160kW
LMTD = 21.6 oC
A = 74 m2
5-1
vessel diameter
stirrer diameter
rotational speed
speed in the lab
D = 1.366 m
d = 0.683 m
n = 1.75 s-1
n' = 8.12 s-1
5-2
precipitate
concentration factor
water in product
ammonium sulphate
m = 99 kg
D = 30
VW = 263 L
mAS = 68 kg
5-3
mass transfer coefficient

k = 5.6 Pm s-1
concentration factor where flux is zero: D = 11
5-4
water needed
protein fraction lost
required membrane area
6-1
extraction factor
fraction lost
VW = 33 m3
0.033
210 m2
E = 1.5
u = 0.16

6-2
plateau concentration
cM = 600 kg m-3
c_ is the concentration in water where the concentration in the solid
has reached one half of the maximum value.
The isotherm begins to diverge from linearity when the concentra
tion in water exceeds about 50 kg m-3.
The average extraction factor decreases at high concentrations, so
you may expect our assumption to be too optimistic.
6-3
You need a low pH for extraction; a high pH for the reverse. A low
temperature and short residence times are required because penicillin is unstable at both high and low values of the pH.
6-4
The Kremser equation gives the same results.
6-5
(a) lost fraction u = 0.075

(b) lost fraction u = 0.024
7-1
volume of acid
VS = 0.571 m3
volume of water
VW = 0.429 m3
acid that crystallizes
VA = 0.286 m3
number of crystals
NC = 5.46108
fraction of crystals in slurry IC = 0.364
The number of seeds required is the same as the number of crystals.
7-2
volume of seeds Vs = 0.286 mL

mass of seeds
ms = 0.4 g
7-3
at V = 0.1 it takes 5000 s

at V = 0.2 it takes 2500 s
7-4
(b) The point where nucleation takes over from seeding is when
'c = 190 kg m-3.
(c) With a small supersaturation you can get uniform large crystals,
but they grow slowly. Above a certain threshold, spontaneous nucleation takes over. You then produce a large number of fines.
7-5
citric acid, stage 1

citric acid, stage 4
c1 = 810 kg m-3
c4 = 650 kg m-3
recycle flow
water vaporized
mW
heat flow
7-6
6.25Ls-1
6.25L
0.6kgs-1
0.6kg
0.33MW
0.33M
The number distribution is a falling function; the volume distribution has a peak at 3 times the diameter formed at the average
residence time. This is because large particles count heavily in the
volume distribution.

Answers
8-1
8-2
The feed of the stripper consists of 100 mol s-1 of a boiling liquid
with a mole fraction 0.1 of methanol.
The feed to the absorber consists of 60 mol s-1 of vapor in equilibrium with the feed. So it contains a mole fraction 0.42 of methanol
or 0.58 of water.
absorption factor
water to absorber
fraction water to top
water flow to top
water in distillate
A = 2.083
34.8 mol s-1
u' = 3.410-4
0.012 mol s-1
0.002 mol s-1
stripping factor
methanol to stripper
fraction methanol bottoms
water flow to reboiler
water flow in bottoms
S = 4.000
15 mol s-1
u = 7.310-4
0.011 mol s-1
0.007 mol s-1
8-3
molar masses almost equal

acetone boils 39 degrees lower
acetone is less polar than propanol
traces of acetone in water have a very high volatility
8-4
there is 99 times as much CO2 in the vapour as in the liquid

we need a vapour flow of 19 mol s-1 (2% of the feed)
8-5
the two entering flows are at equilibrium

nothing happens on this stageit does not work
8-6
heat flow
required area
A= 880 m2
9-1
9-2
M, kg mol-1
d, nm
1.3
10
2.8
100
6.1
100
13.1
cM = 100 kg m-3
= CM / c_
9-3
4.4 MW
the diameter of the water

molecule is about 0.4 nm
c_ = 0.5 kg m-3
= 200
There is no adsorption in size exclusion; co-ions (with the same

charge as matrix) are excluded in ion exchange.
There are many small adsorption sites in ion exchange and ordinary
adsorption.

In affinity separations we need large ligands, so the number of sites
is limited.
In size exclusion the sorbate is excluded, so the distribution coeffi
cient is smaller than one.
The large number of sites in ion exchange and ordinary adsorption
allow a high concentration in the solid.
tB = 980 s
bB = 1.01 mm
zB - zA = 2 mm
just separated
9-4
tA = 1000 s
bA = 0.99 mm
9-5
velocity between the particles

v = 0.20 mm s-1
the velocity between the particles increases v' = 0.25 mm s-1
the concentration velocity of a non-adsorbing component increases
the concentration velocity of a strongly adsorbing component de
creases
9-6
The first five profiles clearly have the same area. In the sixth case,
the length of profile exceeds that of the original. So the height must
go down to get the same surface area.
10-1 velocity in channels v = 909 Pm s-1

residence time adsorbing component
W = 2343 s
d, nm
0.4
0.94
0.82
0.70
0.57
10
0.39
10-2 The maximum sample volume is 0.772 mL.

10-3 (a)
(b)
(c)
(d)
We would prefer to have the desired product B on the exchanger. This will allow concentration of the product.
You can use either a cation or an anion exchanger.
Possible ranges for adsorption are 5 < pH < 7 for a cation
exchanger and 7 < pH < 9 for an anion exchanger.
Elution is possible in the ranges 7 < pH < 9 for a cation exchanger and 5 < pH < 7 for an anion exchanger.
10-4 solid volume required

column volume
amount of B left
minimum washing
10-5 lab
(a)
(b)
(c)
Z = 0.1 m
Z' = 0.1 m
Z'' = 2.5 m
Z''' = 0.625 m
VS = 100 L
VC = 167 L
mB = 2.7 kg
VW = 267 L
D = 1 cm
D' = 1 m
D'' = 1 m
D''' = 2 m

j = 100 Pm s-1
j = 100 Pm s-1
j = 100 Pm s-1
j = 25 Pm s-1
d = 20 Pm
d = 20 Pm
d = 100 Pm
d = 100 Pm
Index
absorber
absorption factor
acetone
activation energy
adsorption
adsorption / elution
adsorption equilibrium
adsorption isotherm
adsorption matrix
adsorption, affinity
adsorption, ordinary
adsorption, protein
affinity adsorption
affinity isotherm
affinity matrix
alcohols / water
amine solvent
ammonium sulphate
anion exchanger
apolar
Arrhenius
auxiliary phase
azeotrope
bacteria
balance, energy
balance, mass
balancing
band
band separation
batch operation
batch ultrafiltration
bead mill
bio-ethanol
blowing, filter cake
boiler, distillation
broth
butanol
byproducts, distillation
cake ratio
cake resistance
cake, filter
carbon dioxide
cation exchanger
cell debris
81
85
92
40
99
113
95
96
95, 97
112
109
96
98, 112, 119
113
112
87
61
48
119
86
40
64
87, 92
7
13, 39
13, 43
16
98, 100, 105
108, 119
10
53
10, 17
88
29
81
2, 121
89
88
26
28
26
92
119
19
cell disruption
centrifuge
centrifuge capacity
centrifuge sigma
centrifuge, disc-stack
centrifuge, scale up
centrifuge, swing
centrifuge, tubular
cephalosporin
charge, protein
chromatography
chromatography, interactions
chromatography, linear
chromatography, paper
chromatography, scaling up
citric acid
citric acid, isotherm
clarify
co-current extraction
coefficient, distribution
co-ions
concentrate
concentration velocity
condensate
condenser, distillation
continuous flow
continuous mixer-settler
continuous operation
continuous ultrafiltration
counter ions
countercurrent
countercurrent heat exchange
crystallise
crystallise, batch
crystallise, continuous
crystallise, seed
crystalliser flows
crystalliser, energy balance
crystalliser, industrial
crystalliser, multistage
crystals, growth
cycle time
cycle, losses

12, 122
23
33
24, 129
24, 129
23
23
24
121, 126
47
98
102
101
94
120
62
62
19, 122
66
60
109
35, 46, 111
100, 105
108,134
35
81
24
63
10
54
109
66
41
71
72
73
73, 77
76
76
75
75
73, 74, 78
30
67, 70
cycles, process
cylinder, measuring
de-activation
debris, cell
developing, process
diafilter
diafilter protease
diameter, molecule
diluent, extraction
disc-stack centrifuge
dissipation
distill
distillation boiler
distillation, absorber
distillation, condenser
distillation, energy balance
distillation, flash
distillation, flows
distillation, stripper
distillation, byproducts
distribution, residence time
distribution, size
double layer
drum filter
dryer, spray
eluent
elute
energy
energy balance
energy balance, crystalliser
energy balance, distillation
enthalpy
enthalpy of vaporisation
enthalpy, water
enzymatic release
enzyme
enzyme, ultrafilter
equilibrium stage
equilibrium, adsorption
equilibrium, protein
equilibrium, reversible
equilibrium, size exclusion
ethanol
evaporate
evaporator, film
evaporator, multistage
evaporator, rotary
63
20
40, 44
19
116
55, 132
57
103
61
24, 129
49
80
81
81
81
84
92
82
81
88
60
73
72, 79
47
30, 34
43
99
99, 114
13, 39
13, 39
76
84
13
38
38
8, 15
7
57
64
95
107
61
107
88
35
36
36
35
exchange area
exchange heat
expanded bed
extracellular
extract
extraction diluent
extraction factor
extraction solvent
extraction stage
extraction, co-current
extraction, reverse
ferment
film evaporator
filter
filter aid
filter cake
filter cake washing
filter medium
filter plot
filter resistance
filter, drum
filter, funnel
filter, plate
filtrate
filtration scale-up
flash distillation
floc
flocculate
flow, continuous
flow, heat
flow, plug
flow, work
flows, distillation
flows, crystalliser
flux, filtration
flux, ultrafiltration
formulate
fraction, pore
front
fungicide
funnel filter
funnel, separation
fusel oil
gel
gel layer, UF
genetic manipulation
grow crystals
GTD

40
41, 45
126
2
59
61
60
59
64
66
63, 66
1
36
26
29
26
29
26
131
28
30
26
29
26
26
92
47
124
24
1, 35
21
14
82
76
27
52, 133
2
107
99, 102
70
26
59, 69
89
109
52
7
73, 78
42
Index
heat exchange
heat exchange, countercurrent
heat flow
heat re-use
heat transfer coefficient
hetero-azeotrope
HIC
homogenate
homogeniser
hydrophobic interaction
interactions, chromatography
intracellular
ion exchange
ion exchange, matrix
ionic strength
iso-electric point
isotherm, adsorption
isotherm, affinity
isotherm, citric acid
Kremser equation
Langmuir isotherm
ligand
linear chromatography
LMTD
losses in a cycle
losses in a cycle
losses, product
LTD
mass balance
mass balance, extraction
mass transfer coefficient
matrix
matrix, adsorption
matrix, affinity
matrix, ion exchange
measuring cylinder
mechanical release
medium resistance
medium, filter
membrane
methanol
MF
microbiology
microfiltration
micro-organism
mixer
mixer-settler
molar mass
41, 45
41
14, 35
93
40
88
109
10
11, 16
98, 109
102
3
98, 109, 123
109
48
47, 110
96
113
62
66, 70
69, 96, 103
112
101
42
67
70
40
42
13, 43
65
53
109
95, 97
112
109
20
10
28
26
51
80, 90
51
2
51
7
48, 63
63
103
molecule, diameter
multistage evaporator
nanofiltration
NF
non-linear chromatography
nucleate
operation, batch
operation, continuous
ordinary adsorption
paper chromatography
particle
peak broadening
peaks
penicillin
pervaporation
pI
plate filter
plot, filter
plug flow
plug flow settler
Podbielniak extractor
polar
pore fraction
power
power number
precipitate
precipitate
precipitation steps
pressure, vapour
process
process cycles
process route
process, developing
product
product losses
product release
profit
project
propeller mixer
protease, diafilter
protein adsorption
protein charge
protein equilibrium
protein solubility
pulse
pump
purify
reflux

103
36
51
51
101
72, 78
10
10
109
94
19
104
100
70, 126
89
47, 110
29, 33
131
21
21
70
86
107
14
49
46, 56
56
50
38
3, 121
63
127
116
1
40
9
5
1, 126
48
57
96
47
107
46
98
13, 17
94, 107
81
reflux ratio
regenerant
release, enzymatic
release, mechanical
release, ultrasonic
removal
residence time
residence time distribution
retention
re-use of heat
reverse extraction
reverse osmosis
reversible equilibrium
RO
rotary evaporator
sample size
saturate
scale down, stirred vessel
scale up, centrifuge
scale-up, filtration
scale-up, chromatography
seed, crystallise
separating bands
separation funnel
series of stages
settler
settler, filling up
settler, plug flow
settler, vertically mixed
settling ratio
settling velocity
settling, swarms
shape, crystals
shear
sieve tray
sigma value
simulated moving bed
size distribution
size exclusion
size exclusion
size exclusion equilibrium
SMB
solubility, citric acid
solubility, extraction
solubility, protein
solvent, extraction
solvent, precipitation
sorbent
83
99
8, 15
10
8
22
11, 40
73
54
93
63, 66
51
61
51
35
108
71
56
23
26
120
73, 77
108
59, 69
65
20, 31, 63
32
21
22, 129
21
19
20
74
50
82
24, 33, 129
117
72, 79
107
98, 118, 123
107
117
71
61
46
59
48
95
spacer
specific heat
spray dryer
stage series
stage, equilibrium
stage, extraction
steps, process
steps, precipitation
stirred vessel
stripper, distillation
stripping factor
supersaturate
swing centrifuge
tail, elution
time, cycle
time, residence
track process
tray downcomer
tray, sieve
tubular centrifuge
turbine mixer
turbulence
UF
ultrafilter, enzyme
ultrafiltration
ultrafiltration, flux
ultrasonic release
valve
vapour compression
vapour pressure
velocity, concentration
velocity, settling
vertically mixed settler
void fraction
volatility
wash, adsorption
washing, filter cake
waste
water / alcohols
water, enthalpy
work flow
x-y diagram
yeasts
yield

112
13
43
65
64
64
3
50
48
81
85
71
23
99, 102
30
11, 40
127
82
82
24
48
49
51
57
51
133
8
13
37
38
100, 105
108, 134
19
22, 130
28
86
114
29
124
87
38
14
84
7
5, 34
Content
Separating and purifying the desired product is an important part

of any bio-project. This processing downstream from the fermenter is both difcult and expensive. This book introduces all the
common techniques, starting in the lab and ending in the plant. We
have written it for undergraduates in process engineering (biotechnological, mechanical or chemical) and graduate microbiologists
doing a minor on the technology that is required to scale up their
experiments to a commercial scale.
The Authors
J.A. (Hans) Wesselingh studied engineering physics and worked for the
Shell Group for many years. His last job there was as leader of a research
group developing large scale reactors and separation equipment for the
oil, gas and chemical industries. (In those days Shell also had biotechnology interests.) He later worked as professor on research
and teaching of separation processes at Delft University of
Technology, the University of Groningen and the Technical
University of Denmark. He has been a consultant to many
Dutch process industries.
John Krijgsman is a mechanical engineer who has worked in the food
industry and for many years with the Dutch biotechnology company
Gist-Brocades (now part of DSM). He has been a major contributor to
several large biotechnological projects. John has taught
downstream processing to students at the Delft and Eindhoven Universities of Technology. He is currently working
as a consultant for Biogonal.
Published by
Delft Academic Press
ISBN 97890-6562-319-5
printed version:
ISBN 97890-6562-318-8
http://www.vssd.nl/hlf/d030.htm

h1 PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

h1 PDF

Uploaded by

Copyright:

Available Formats

Downstream Processing

@Delft Academic Press / VSSD

Published by Delft Academic Press / VSSD

diameter (small, particle)

slope of equilibrium line

release, removal, retention

velocity across, particle velocity

mole fraction in liquid

cake, concentrate, cycle

Figure 1-1 Product development cycle

Product development is usually cyclic you cannot really say where it

Figure 1-2 Subdivision of develop and produce

Ferment (using Cells)

Figure 1-3 Contents of a fermenter

Figure 1-4 Sequence of process steps

Intracellular products are released from the cells (lesson 2)

0.9510 = 0.60 after ten steps.

2-1 Micro-organisms and their walls

Bacteria have dimensions of the order of a micrometre. Their walls are

Figure 2-2 Enzymatic release

Figure 2-3 Enzymatic release curve

Figure 2-4 Finding parameters in a plot

A disadvantage of enzymatic release is the high cost of the enzymes.

Figure 2-5 A bead mill

Here k is again a rate constant that has to be determined experimentally;

Figure 2-6 Batch and continuous operation

For continuous operation, the time should be replaced by the residence

Figure 2-7 A homogenizer

The exponent in the pressure ratio is determined from experiments with

Figure 2-8 Disruption of cells

Figure 2-9 Systems and sub-systems

change of enthalpy work heat

As we will see in the exercises, the temperature rise can be substantial. In

Figure 2-10 Balances for three systems

The rate constant is a function of composition:

The homogenate leaves the homogenizer at a higher temperature. It is

2-3 Exercise Your Balancing

Batch Bead Mill

Is there a change in pressure?

The symbols denote:

Figure 3-1 Settling velocities

may expect in biotechnology. Note the logarithmic scale the velocities

Figure 3-2 Measuring the velocity

Figure 3-3 Settler dimensions and flow

Figure 3-4 Plug flow model

Plug Flow Settler

Figure 3-5 Plug flow (2)

moved. The fraction removed or removal becomes:

Figure 3-6 Vertically mixed settler

Vertically Mixed Settler

Figure 3-7 Removal in settlers

Figure 3-8 Swing centrifuge

For scaling up we need the settling velocity ug under gravity. We can

Here r is the average radius of the swing centrifuge.

Figure 3-9 Tubular centrifuge

So the centrifuge behaves like a settling tank with a surface area:

The equipment is continuous, so that the vertically mixed model of the

Figure 3-10 Disc-stack centrifuge

Figure 3-11 Stack geometry

Figure 3-12 Principle of filtration

Funnel Filter (Lab)

Figure 3-13 Funnel filter

2.4 106 2000 T