16 Distillation in the pharmaceutical industry

RF. WILCOX

16.1 Introduction

16.1.1 Overview
In the pharmaceutical industry separation is one of the most widely used
unit operations. Its main uses include the recovery of reactants and solvents
for recycling, the purification of products for sale, and the processing of
byproduct streams for waste minimization. Though many separation techniques
exist, the one most often used is distillation. Distillation is simple, has few
moving parts, is well understood, can be simulated readily, and often has a
low capital cost. On the negative side is the high operating cost associated
with high energy requirements to vaporize and condense.
Most distillation simulations are done on the computer now and many
excellent references exist (see the Bibliography) that give detailed accounts
of the mathematics. The emphasis of this chapter therefore will be on the
various uses for distillation to purify a product, break an azeotrope, and
perform batch distillation. Equilibrium thermodynamics will be discussed in
such a way that the engineer can have a feel for the behavior of a given
system, if only in a general sense. After reviewing this chapter the engineer
will understand why a system behaves as it does, for example, why the phase
is split between water and butanol or why more water will dissolve in benzene
than hexane, and will be enabled to make initial decisions about systems
quickly without detailed simulations.

16.1.2 Molecular interactions

Understanding how a system behaves is more important now than in previous
eras. With the advent of the computer, distillation design is being pushed
closer and closer to the theoretical limit. In the past a distillation column
required less front end engineering, but included more design allowances to
account for uncertainties. This gave rise to distillation units that greatly
exceeded their nameplate capacity. This meant higher capital cost than
required for the original scope of a project. Now with computers and improved
correlations, the design allowances have been cut to reduce capital outlays,
thus requiring accurate predictions. But we as engineers must remember that
the computer does not simulate the real world but only represents reality by

E. Goldberg (ed.), Handbook of Downstream Processing

© Chapman & Hall 1997
418 HANDBOOK OF DOWNSTREAM PROCESSING

using a mathematical model. In this way the engineer will need to know
which mathematical model is best for the system, and just as importantly,
when a model has performed poorly. This can be done with a thorough
understanding of the working of the molecules in a system. Why does the
system of benzene and water separate into two phases? Using common sense
an engineer would think that more energy and less entropy is required to
keep the benzene layer floating on the water. As seen in equation (16.1)
minimizing the Gibbs free energy involves two terms, the entropy and the
enthalpy of a system.

G = H - TS (16.1)

where G = Gibbs free energy, H = enthalpy, S = entropy, and T = temperature.

From thermodynamics we know that mixing components increases entropy,
i.e. the randomness of a system, and thus reduces Gibbs or internal energy
of a system. In some cases this entropy change does not offset the increase
in enthalpy. As we all know, benzene and water do form separate phases
and this is because the Gibbs free energy is at a minimum when the phases
are separated. But does this really tell us why? Using equation (16.1), could
one look at the system of methanol and heptane and tell if it is miscible or
not? Probably not, but with an understanding of the way molecules interact,
a good guess can be made. This sort of understanding can be invaluable in
reviewing the output from a computer distillation run. The engineer may
find that the wrong model for thermodynamic equilibrium was used or better
still would choose the right one first.

16.1.3 Phase equilibria

16.1.3.1 Understanding intermolecular interactions. Molecular material can

exist as either a solid, liquid, or vapor over specific temperature and pressure
ranges. These specific ranges exist because interactions between molecules
cause either the solid, the liquid or vapor phases to be more stable. In other
words there may be an energy saving in having molecules closely packed in
a somewhat rigid arrangement rather then loosely and randomly flying about.
All molecular material, whether a single component or a mixture, will seek
to have the lowest potential energy possible[l]. For thermodynamic equilibrium
to exist between phases in a system, the temperature T and the chemical
potential j of each component must be equal in all phases. The chemical
potential is defined as the molar Gibbs free energy of a component. Generally
the pressure is also equal in all phases but this is not a necessity. For example
when phases are separated by a semi-permeable membrane, thermodynamic
equilibrium exists but both phases are not at the same pressure.l 11
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 419
The most stable form for a system will be that which minimizes the Gibbs
free energy (see equation 16.1) and therefore the chemical potential. At lower
temperatures the Gibbs free energy will be dominated by the enthalpy term
H, which means that a structured system with greater molecular interactions
and order is the most stable, i.e. the solid phase. At high temperature the
Gibbs free energy is dominated by the entropy term TS. In order to have
higher entropy the system must exist in a more random state which would
correspond to a gas. The liquid phase is a balance of entropy and enthalpy
effects which minimize the Gibbs free energy. Figure 16.1 depicts the change
in chemical potential with rising temperature for each phase. One can see
that the melting point Tm marks the point where the chemical potential would
be minimized by following the liquid line instead of the solid and likewise
the boiling point Tb is where the vapor takes the path to minimizing the
chemical potential. Also note that thermodynamic equilibrium exists where
the temperature and the chemical potential in each phase are equal: both at
the melting point and the boiling point.

16.1.3.2 Deviation from ideal. An ideal gas is a gas in which no molecules

interact with any other molecule. The molecules that make up the gas will
exert no attractive or repulsive forces. Ideal gases and ideal solutions are
only concepts. No real molecules are truly ideal. The concept is applied to
molecular interaction to give a basis for comparison. Real molecules can be
described by how much they deviate from the imaginary ideal state.

Solid

~TEMPERATURE

Figure 16.1 Representation of the chemical potential of a system as temperature increases.

420 HANDBOOK OF DOWNSTREAM PROCESSING

An ideal gas has the following properties[l):

• No interaction between like molecules

• No interaction between unlike molecules
• The equation PV = nRT holds true for all components.

Likewise when mixing molecules to form an ideal solution the following

properties will need to exist[1,2):

• No volume change occurs when an ideal solution is produced

• The heat of mixing is zero
• The entropy change is equal to that which would be expected from the
extra degree of freedom. In other words the mixture is perfectly random
in concentration throughout
• The Gibbs free energy change is only provided by the entropy gain
• Ideal solutions are always miscible. Only very non-ideal solutions form
multiple liquid phases.

All molecules deviate from ideality but some as we know have greater
deviations than others. Why is this? The answer is provided by molecular
interactions. In an ideal solution, a molecule does not care who its neighbors
are, but in reality molecules do discriminate against their neighbors, some
more so than others. This discrimination is caused by a molecule being drawn
towards or repelled by a like molecular neighbor more than an unlike
neighbor. When a molecule has a greater attraction for like molecules in a
mixture the system is said to deviate positively from ideal. If on the other
hand a molecule has more attraction for unlike molecules, then the deviation
from ideal is negative. For example, a negative deviation from ideal is seen
in the mixture of acetone-chloroform[1). This deviation is attributed to
hydrogen bonding between the two unlike molecules. Evidence is seen to
support this in the increase in intensity of the C- H stretching band of
chloroform which occurs on addition of acetone. As the new component
acetone is added the acetone's oxygen exerts an attraction towards the
chloroform's hydrogen, thus distorting the bond length slightly. There is also
a change in the carbonyl stretching band of the acetone. This interaction
gives rise to a complex with one acetone and two chloroform molecules as
shown in Figure 16.2[1). These data were obtained from the infrared spectrum
of the mixture.
Positive deviation from ideality is more commonly seen, such as when a
highly polar molecule such as water is mixed with a less polar substance.
Water molecules have a great attraction for one another and water associates
into an ordered liquid with many hydrogen bonds between the molecules
(see Figure 16.3). When a diluent is added it will tend to break apart many
of the hydrogen bonds. The energy of the hydrogen bond in water is
5.7 kcaljmol which means the heat of mixing will be non-zero, since energy
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 421

'" '"
CI CI CI CI
'\ /
C/ C
CI
/
H
~ ~ H/ "'" CI
08
II
H~/C~ /H
C C
/'\ /~
H H H H
Figure 16.2 Acetone-chloroform interaction.

o
H/~H

H H 0
'\ / / ~
OH H
H H o
'" o / H/ '" H
Figure 16.3 Possible hydrogen bonding sites for water molecules. The hydrogen bond is shown
as a dashed line.

must be expended to break these hydrogen bonds and create a cavity between
the water molecules[1,31. If the diluent is small and can replace some of the
broken hydrogen bonds with hydrogen bonds of its own then less energy is
required for it to go into solution with water. But if the cavity required is
large requiring the breaking of many hydrogen bonds, the energy increase
required by the enthalpy term in equation (16.1) will exceed the reduction
in free energy obtained from the entropy term by increasing the randomness.
In this case the system's minimum potential energy is achieved by maintaining
separate phases.
The only class of mixtures that closely resembles ideal is called a regular
solution[21. These mixtures obey the ideal mixing law for entropy but the
heat of mixing is not zero although it is small. The number of solutions that
meet these requirements is very limited.
422 HANDBOOK OF DOWNSTREAM PROCESSING

16.1.3.3 Intermolecular forces. Intermolecular attractive forces are caused

by various factors such as molecular shape, size, electron charge density, and
ionic effects to name a few. Because of the inherent short range over which
these attractive forces are felt the effect is more pronounced in liquids and
solids than in vapors[1,2,4-61. A greater deviation from ideality can be seen
in a liquid mixture of water and heptane than in the vapor mixture of these
same components.
The first attractive force, which is exhibited by all molecules whether polar
or not is the dispersion forces also known as the London forces[1,61. These
dispersions forces are caused by an instantaneous fluctuation in the distribution
of the negative electron cloud over the positive nucleus resulting in the
appearance of dipoles. Although the collective instantaneous dipoles must
average to zero the resulting local attractive forces do not. If the electron
cloud shifts in the other direction (see Figure 16.4) the dipole will reverse
which will reverse the induced dipoles shown. Even though the average over
time of the instantaneous dipole is zero, as it must be for molecules without
permanent dipole moments, the attractive forces it is associated with will not
average to zero. The dispersion force is proportional to R- 6 , the distance
separating the molecular centers, which accounts for the small distance over
which the force is effective. In most non-poplar molecules, dispersion forces
will be the only attractive forces to cause deviation from ideality.
Molecules with permanent dipoles can induce a dipole in a neighboring
molecule by distorting the negative electron cloud in such a way that the
affected molecule will also have a dipole, whether polar or not. The induced
and permanent dipoles will be mutually attracted (see Figure 16.5). This
attractive force is known as the dipole-induced dipole or Debye interaction[1,61.
The dipole-induced dipole attraction can be large depending on the size of
the permanent dipole moment, for example with ions, or small for nearly
spherical molecules like CH 4 , or even zero for inert gas atoms which have
no permanent dipole moments. Since the induced moment will readjust itself
nearly instantaneously, regardless of the molecular rotation of the permanent

1
, c9"
8 8
3

8+
c::5-
NEUTRAL STATE ATTRACTIVE STATE

Figure 16.4 A shift in the electron cloud of molecule 1 induces a shift in molecules 2 and 3
causing attractive forces to set up.
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 423
3

+
+ •

8
+

:. I
"

Figure 165 The electron cloud from the permanent dipole of molecule 1 induces a shift in
molecules 2 and 3 causing attractive forces to be set up.

dipole, the effect is thus independent of temperature as long as the mixture

density and therefore molecular spacing is not affected. The attractive force
is proportional to R - 6, the distance separating the molecular centers.
Dipole-dipole or Keesom interactions occur between molecules in which
both have permanent dipole moments[1.61. This attraction is proportional to
the strength of the contributing dipole moments. As shown in Figure 16.6,
if molecule A rotates the associated molecules 1-5 must also rotate, but there
is an inherent time lag in this rotation. This time lag will decrease the attractive
force, proportional to the amount of time that the optimum alignment is not
achieved. Since the force is an orientation effect it is temperature dependent.
As the temperature rises, all molecules will have more random motion and
the optimum orientation will not be achieved as often. Dipole-dipole
interactions have contributions from all associated molecules. The force is
proportional to R - 3, the molecular separation, which means this effect is
felt over a greater range than the previous forces mentioned.
Electron donor-acceptor interactions are of great importance in explaining
non-ideality. These interactions include Lewis acid-base attractions and the
well-known phenomenon of hydrogen bonding which is a Bronstead acid - base
interaction[l.4 81. These interactions cause greater departures from ideality
because the attractions are much stronger than all the others mentioned

2 2

G
1

~o G
66
~A 5

~O
~

o
3 4 3 4

0 G G
Figure 16.6 Rotation of dipole molecule A and its effect on dipole molecules 1 to 5.
424 HANDBOOK OF DOWNSTREAM PROCESSING

previously. The attractions come in two forms. Association, in which like

molecules are attracted, and complexing, where unlike molecules have the
greater attraction. These molecules form quasi-polymeric systems with
defined structures. Water forms a three-dimensional structure while alcohols
will give rise to either cyclic or linear quasi-polymers of indefinite length.
The strength of hydrogen bonding (though greater than dipole effects) is still
less than regular bonds. The energy of the C-H bonds is 87 kcal/mol, while
typical hydrogen bond energies vary from only 2-8 kcal/mol depending on
the electro negativity of the atom bonded to hydrogen. Hydrogen bonding is
favored by atoms with higher electronegativity than hydrogen and the
strength of the hydrogen bond increases in the following order:
C < I < S < Br < Cl < N < 0 < F.
Any mixture of molecules with hydrogen bonding capabilities will deviate
greatly from ideal. If there is an increase in hydrogen bonding due to mixing
then miscibility is enhanced, but if hydrogen bonding is broken miscibility
is reduced as with water and oil. A special case exists where the hydrogen
bonding molecules fit together in complete oligomer packages. Examples are
dimers such as carboxylic acids and some glycol ethers, and hydrogen fluoride
which forms a hexamer, H6 F 6'
Figure 16.7 shows an example of the acetic acid dimer. This results in
diverse solubility properties such as the ability of acetic acid to dissolve in
water and heptane. In water, acetic acid can open up and interact with water
both through its polar properties and hydrogen bonding. But in non-polar
heptane, acetic acid is a self-contained dimer with no remaining site to
hydrogen bond, which causes it to exhibit non-polar tendencies. This ability
to assume the property of the surrounding environment is termed chameleonic
after the famous camouflaging reptile[61. Since these dimerizing hydrogen
bonding molecules form compact packets they are freer to move about at
higher temperature than the larger structures seen in the quasi-polymers
which tend to break hydrogen bonds as the temperature is increased. Thus
these molecules will stay as dimers even in the vapor phase. This effect must
be accounted for in designing distillation systems involving oligomerizing
molecules.

H O-····+H- 0 H
I I; \ I
H-C-C C-C- H
I ~ ~ I
H O-H+"''-O H

Figure 16.7 Acetic acid dimer with hydrogen bonds indicated by the dashed line.
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 425
The last effect to be discussed is often overlooked. This is the effect of
molecular size. Three molecules, as shown in Figure 16.8, are mixed in equal
molar concentrations, one polar and the other two non-polar; with the only
difference between the non-polar molecules being size. The polar molecule
will interact with the larger molecule in proportion to the size difference in
non-polar molecules. This is based on simple molecular collision statistics.
It is easier and therefore more probable to hit a larger target. But as discussed
earlier the larger molecule will have to create a bigger cavity in the polar
solvent, so these two phenomena will counteract each other. With a highly
polar molecule 'A' that forms a quasi-polymeric network, the size difference
will hinder creation of a cavity in the polar network more than it will facilitate
an increase in collisions. But conversely if molecule 'A' is only slightly polar
or forms packets of dimers and trimers, then collision frequency with the
larger molecule will overshadow its resistance to forming a cavity.
Given this basic understanding it should be possible to make certain
assumptions about the behavior of mixtures. If we take the local composition
model, which says that the effect of a molecule is only felt over a short range,
to the nearest neighbor at the least and marginally to the next row (see Figure
16.9), and assume no concentration gradients exist throughout the mixture,
the interaction between a few molecules can be applied to the whole mixture.
These assumptions will be general and may not apply to mixtures with many
different molecules. Also some factors work for and some against reducing
the system potential energy, as seen with molecular size.
Ewell et a1. did this for the relationship between hydrogen bonding and
azeotrope formation[1.8.91. They concluded that hydrogen can coordinate
between two molecules at the O 2 , N 2 , or F site, or even at C if a number
of highly electronegative atoms are attached to the carbon atom. They
suggested the following classifications of hydrogen bonds as strong or
weak.

o
NON·POLAR

NON·POLAR

o
2

POLAR MOLECULE
'A'

Figure 16.8 Effect of size on molecular interaction.

426 HANDBOOK OF DOWNSTREAM PROCESSING

(:)
(:)
Figure 16.9 Local composition model with interactions AI2 and All and very weak interaction A13.

• Strong bonds
O-HO
N-HO
O-HN
• Weak bonds
N-HN
HCCI
o HCCI 2-CCI
N HCNO
HCCN 2
Similarly liquid mixtures can be classified by functional group, which help
determine the deviation from ideality (8). See Tables 16.1 and 16.2.

16.2 Design of a distillation system

It is not the intent of this chapter to go into the details of the mathematics
involved in the calculations for the various distillation methods. Some key
equations will be given but it is assumed that the reader has access to the
many good references listed in either the bibliography or the reference list.
Furthermore most readers will utilize a computer to perform these calculations.

16.2.1 Distillation methods

16.2.1.1 Simple distillation. Distillation accomplishes separation by

maintaining a composition gradient between the vapor and liquid phases
caused by difference in volatilities of the components. The separating agent
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 427
Table 16.1 Classification of liquid mixtures

Class T Liquids capable of fonning three-dimensional networks of H bonds: water, glycol,

glycerol, amino acids, hydroxylamine, hydroxy acids, polyphenols, amides, etc. These
are 'abnormal' or 'associated' liquids having high dielectric constants and are water soluble.
Class IT Other liquids containing active hydrogen atoms and other donor atoms: acids, phenols,
alcohols, primary and secondary amines, oxides, nitro compounds and nitriles with
H atoms, NH 3 , hydrazine, HF, HCN, etc. The characteristics of class II liquids are
the same as class 1.
Class III Liquids composed of molecules containing donor (0, N, F) atoms but not active H
atoms: ethers, ketones, aldehydes, esters, tertiary amines. These liquids are also water
soluble.
Class IV Liquids composed of molecules containing active H atoms but no donor atoms: CHCI 3 ,
CH 2 C1 2 , CH 3 CHCI 2 , CH 2 CI-CHCI 2 , etc., which show only slight water solubility.
Class V All other liquids, i.e. compounds having no hydrogn bond-forming capabilities:
hydrocarbons, CS 2 , RSH, non-metallic elements, etc. This class of materials shows
essentially no solubility in water.

Table 16.2 Deviations of class combinations from Raoult's law

Classes Deviations H bonding

I+V Always positive with T + V H bonds broken only

IT+V frequently showing limited
solubility
III + IV Always negative H bonds fonned only
I + IV Always positive, H bonds both formed and broken
II + IV I + IV limited solubility (dissociation common)
I +I
I + II Usually positive H bonds both broken
I + III Some negative, giving maximum and formed boiling
azeotropes
II+II
II + III
III + III
III + V Quasi-ideal always positive No H bonds involved
IV + IV or ideal
IV + V Minimum azeotropes
V+V Minimum azeotropes (if any)

is the heat energy suuplied. Simple distillation is usually a continuous process

with the lighter or more volatile fraction leaving the top as the distillate and
the less volatile or heavy fraction recovered as the bottoms product. A portion
of the overhead vapor is condensed and returned as reflux to enrich the
vapor rising through the column. The section above the feed where the vapor
is enriched by reflux thus condensing the heavy portion is called the rectifying
section. The section of the distillation column below the feed where the liquid
428 HANDBOOK OF DOWNSTREAM PROCESSING

is stripped of light components by the rising vapor from heat input is called
the stripping section (see Figure 16.10).

16.2.1.2 Azeotropic distillation. For special cases where in a mixture the

differences in effective volatilities are very small, these systems are known as
close boiling mixtures. In some cases the difference is zero, called an azeotropic
mixture. These two systems require special handling to be separated. In close
boiling mixtures simple distillation is prohibitively expensive and an azeotrope
is impossible to separate. Azeotropic distillation or extractive distillation
could be an option to make the separation possible.
In azeotropic distillation (see Figure 16.11) an additional component is
added which will azeotrope with one or more of the components in the
original mixture. This new azeotrope must be easily recoverable and recycled
for reuse or else one separation problem has been traded for another. Selection
of the azeotrope agent is of great importance and since there are a plethora

Accumulator

f-----J Reflux Distillate

Rectifying section

1·1

1+1

Stripping section

Bottoms product

Figure 16.10 Simple distillation column.

 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 429

--I "COOLING
~ MEDIA
/' PYRIDINE

c:=D
- - - PRODUCT

RECYCLE CAUSTIC

--------------; RECYCLE WATER

HI
~
---

PYRIDINE I AZEOTROPIC UQ/LIQ

ACETIC ACID I DISTIUATION EXTRACTION

FEED C.OlUMN
PYRIDINE
---1/HEATING WATER
()" MEDIA AZEOTROPE

'----_/'f---'-___ A;:aI;U~D

Figure 16.11 Azeotropic distillation of pyridine from acetic acid using water and subsequent
dehydration of pyridine using caustic soda. (Courtesy of Jacobs Engineering.)

of compounds to choose from, a selection technique is necessary to winnow

down the choices.
The azeotropic agent when added to the initial binary azeotrope must
form a minimum boiling azeotrope with one of the components in order to
separate the compounds[lOl. If a maximum azeotrope is formed then the
original azeotrope will rise overhead with no benefit obtained. There are two
types of azeotropes. Those separating into two phases when condensed, which
are known as heterogeneous azeotropes, while those remaining as a single
phase are homogeneous azeotropes. If the agent added forms a heterogeneous
azeotrope then the two phases can be separated by decantation in the reflux
drum with the azeotroping agent-rich phase returning to the column as reflux
while the recovered component-bearing phase is sent on for further purification
if necessary. An example of this type is the dehydration of ethanol using
benzene. Benzene forms a heterogeneous azeotrope with water[lOl. The
water/benzene azeotrope is condensed and decants into two phases. The
water-rich phase is sent for further distillation to remove any residual benzene.
The benzene is recycled as reflux. The bottoms product from the azeotropic
distillation column is anhydrous ethanol. When using an azeotroping agent
which forms a homogeneous azeotrope, breaking the new azeotrope must
be dealt with. A few examples of possible methods to use for this new azeotrope
are pressure swing distillation, extractive distillation, or liquid -liquid extraction.
There should, however be an overriding reason to replace one azeotrope by
another. These reasons could be, for example, because of the nature of the
materials of construction or the toxicity of the component removed in the
azeotrope column, or simply because no other method could be found. A
homogeneous azeotropic distillation scheme requires a minimum of three
430 HANDBOOK OF DOWNSTREAM PROCESSING

distillation columns, perhaps more depending on product purity, whereas

the heterogeneous azeotropic scheme could use a single column if high purity
of the overhead product is not required. The azeotropic distillation scheme
is best used with agents that form heterogeneous azeotropes and the smaller
the mutual solubility the better this scheme becomes.
Table 16.4 shows a survey used to make a first pass selection of an
azeotroping agent. The example is pyridine recovery from acetic acid. This
system forms a fairly strong acid-base complex. Because of this there is
negative deviation from Raoult's law caused by the greater attraction of
acetic acid for pyridine than for itself. This system forms a maximum boiling
azeotrope (see Table 16.3)[11 1. Table 16.1 can also be used to predict the
possibility of a maximum boiling azeotrope. Acetic acid is a class I liquid
and pyridine is a class III liquid which indicates the possibility of forming
a maximum boiling azeotrope (confirmed by Horsley[11 1).
Suitable agents are ones which will azeotrope with either acetic acid or
pyridine not only below the boiling point of the acetic acid-pyridine
azeotrope but also below the boiling point of the pure component which is
not being azeotroped. Finally, the boiling point must be less than that of an
azeotrope that might be formed with any other components present. For
this example, we require an agent which will have a positive deviation from
ideal to give a minimum boiling azeotrope. From Table 16.1 the likely
candidates would be as follows:
• Acetic acid (Class H)-Class I -Water
Class IV - Carbon tetrachloride
Class V - Hexane or xylene
• Pyridine (Class III) -Class I - Water
Class V - Hexane or xylene
From the above candidates, the survey (Table 16.4) is then completed. The
first item to be filled in is if the candidate azeotropes with either of the
components and at what temperature and concentration. A good place to
find this information is Horsley [111. Of the candidates only xylene does not
work as an azeotroping agent because the boiling point of the azeotrope is
higher than the normal boiling point for both acetic acid and pyridine. For
all other candidates the remaining survey information can be completed,
which also has an impact on the design such as the materials of construction,

Table 16.3 Azeotrope conditions for acetic acid and pyridine

Normal boiling point Azeotrope Azeotrope

CC) composition boiling point
(wt'l~) CC)
Acetic acid 118.5 51.1 138.4
Pyridine 115 48.9
Table 16.4 Azeotropic distillation agent survey

Ease of B.P.
Stability Feed separ azeo Wt(>~ Wt~/~ Cost
Normal (Rx) Mat'l comp Azeo Wt(:/~ Type from with agent azco reI to Azeo
Azeotropic Mole RP. (DCMP) of to be wjother agent of azeo agent in azeo camp in first agent
(0C) u.b
agent Formula wt Toxic constr' azeotr'd feeds in azeo azeo d comp (T) comp' agent" agent rank Comments

Carbon
tetrachloride CCl. 153.8 77 NjA Y CS 2 98.5 0 Hard 76 Too much agent required
Water H 2O 18 100 NjA N CS 41 0 Hard 94 Best choice
Heptane C 7 H'6 100.2 98 NjA N CS 2 67.13 0 Hard 92,95 2 Azeotropes boil too close

Xylene CHM,o 106.2 140 NjA Prob CS 2 55 0 122 NjA B.P. of azeotrope is >
other components B.P.

Feed Comp #1 C,H,N 79.1 115 NjA CS Pyridine: Azeocomp 1-2,

138°C

Feed comp #2 C 2 H.0 2 60.1 118.5 NjA N 3.04L Acetic acid

URx indicates whether the components will react at operating conditions. b DCMP indicates whether the component will decompose at operating conditions. 'Material of construction
to handle listed chemical. dType of azeotrope~O homogeneous, 1 heterogeneous. 'For heterogeneous azeotropes at reflux drum temperature.
432 HANDBOOK OF DOWNSTREAM PROCESSING

toxicity, cost of the agent, and the ease of breaking the new azeotrope formed.
This is why it is usually best to look for agents which form heterogeneous
azeotropes since they are easier to separate. One can see from the completed
survey that the best candidate for further testing is water because it is readily
and inexpensively available and is safe to handle and also a low weight
percent of agent is required to azeotrope.
Typically more than one candidate meets the requirement for further study.
These candidates are set-up for testing, if required, or if the data are available,
the system can be simulated and optimized to obtain preliminary equipment
and operating costs as a basis to select which agent is the best.

16.2.1.3 Extractive distillation. Extractive distillation is commonly used

to separate close boiling compounds or azeotropes .. It takes advantage of a
relatively non-volatile component which has a greater affinity for one or
more components in the original mixture[lO] (see Figure 16.12). When this
extractive agent is added to the column this affinity pulls these components
down to the bottom of the column thus allowing the remaining components
to rise overhead very readily. The agent is added a few trays below the top
so the effect can be seen throughout the column. The top few trays are
required to knock back any agent that might be carried up the column.
The extract agent is chosen to induce a greater relative volatility between
the compounds to be separated, but by the same token it must not have so
great an affinity that it becomes immiscibile. Once the agent separates it has
no effect on the remaining extractive compound in the other phase.

-j .,
ft/ COOLING
MEDIA

REFLUX
~ I
.~~~-
NITRIC ACID
PRODUCT

f RECYCLE
I SULFURIC ACID

r.:
NITRIC ACID EXTRACTIVE
WATER DISTILlATION
FEED COLUMN

l - --- VENTTO
HEATING --------G:) CONDENSE~CRUBBER
MEDIA 1___ WATER

'----~~~~~~~~ _ _ ~~ ___ .. _ _ VAPORIZER

WEAK SULFURIC
ACID L-~
Figure 16.12 Extractive distillation of nitric acid from water using sulfuric acid and subsequent
reconcentration of sulfuric acid. (Courtesy of Jacobs Engineering.)
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 433
The selection of an extractive agent is not a straightforward process. The
selected agent must increase the relative volatility of the compounds without
requiring a large ratio of agent to feed. As the relative volatility increases
the ease of separation increases which reduces the number of theoretical
stages required, but the smaller the ratio of agent to feed the smaller the
column diameter. These two factors along with the following assessments,
such as how easy is it to recover and reuse the extract agent, what materials
of construction are required, is it a toxic substance, is it expensive, does it
react at the operating conditions, and what are the operating costs, are used
to make the evaluation. As with azeotropic distillation, using a survey form
allows the engineer to compare all these variables when searching for the
best entrainer without drawing up a complete design for all candidates (see
Table 16.5).
Various methods of evaluation can be applied to select the extractive agent.
One quick method is the Scheibel method in which another member of the
homologous series of either of the components is used [12]. For example, the
search for an extract agent for the separation of the methanol/acetone
azeotrope should start by evaluating ketones such as methylethyl ketone,
methyl n-propyl ketone, or methyl isobutyl ketone. Similarly for the alcohol
series the search starts with ethanol, propanol etc. The logic behind this
approach is to find an agent which behaves as moderately ideal with one
compound and non-ideally with the other. A similar method but not limited
to homologs is to use solubility parameters to select agents which are much
more non-ideal in behavior toward one compound than another, since the
greater the difference in solubility parameter the greater the deviation from
ideality. The best method though is to use actual infinite dilution data and
determine the selectivity. The selectivity is defined as the ratio of the infinite
dilution activity coefficients of each species in the extract agent [13].
(16.2)
where S is the selectivity and yoo is the infinite dilution activity coefficient
for components i and j.
Various sources of infinite dilution data are available and are listed in
Table 16.10. It is very easy to obtain experimental infinite dilution data from
various techniques such as differential ebulliometry and chromatography[14].
If no actual data can be found then the infinite dilution can be determined
using either the modified separation of cohesive energy density model
(MOSCED) or UNIFAC [15,16]. Before an agent can be finalized it must be
proved in the laboratory or simulated using actual binary data.

16.2.1.4 Pressure swing distillation The methods mentioned previously

have all required the addition of another component to permit the separation
of an azeotropic mixture. However, another method which has received little
study is pressure swing distillation. The mechanism behind pressure swing
Table 16.5 Extractive distillation agent survey

Ease of Infinite Infinite

Key Solu- separ. dilution dilution Maxi- Cost
Stability Mat! comp Soluble Ratio bility- from act act mum reI to Extract
Normal (Rx) of to be in key entrainer para- key Latent coetf coetf entrn- first agent Com-
Extractive agent FormulaMole wt B.P. (DCMP) Toxic constr" extractd comp to feed meter comp heat LK comp HK comp factore comp rank ments

LK feed compd

HK feed comp'

"Rx indicates whether the components will react at operating conditions. b DCMP indicates whether the component will decompose at operating conditions. 'Maximum entrainment
factor = (inf dil alpha LK)* (vapor press LK)j, inf dil alpha HK)* (vapor press HK). d LK is the key feed comp which goes overhead. 'HK is the key feed comp which is extracted.
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 435
distillation is the effect of pressure on an azeotrope composition. Figure 16.14
shows that at 150mm the azeotrope point of acetonitrile and water is
92.8 wt% acetonitrile. If the pressure is raised to 760 mm then the azeotrope
is 85.8 wt% acetonitrile. With this information a separation system can be
designed (Figure 16.13). The primary feed and the recycled azeotrope from
the higher pressure column are fed to column 1. The overhead from column
1 is the minimum boiling azeotrope at 150 mm, while the bottom is nearly

AZEOTROPE
A~~O~ILE I I)
COOLING

ZM~EDIA --1--- x.:t t COOLING

MEDIA
I
I AZ;~;~OPE
'-- ~ I ACETONITRILE

~ _' /1,,-2_)_~I
LOWCOLUMN
PRESSUJE
~ . - I - 1--- HIGH-PRESSURE
COLUMN
150mm i ! 760mm

FEED i 1 .~
<90% ":;~RNITRILE
!Lr~
'. COLUM. N 1 COLUMN 2

~/ H:~I:G 0-kr
Ii

i 1
99%
ACETONITRILE
, r PRODUCT
WATER

Figure 16.13 Pressure swing distillation scheme to purify acetonitrile.

Vapor Camp Mol % Acetonitrile

100 .---------------------------------------------------~

80

60

40

20
o ~ __ ____ ____
~ ~ _ L_ _ _ _ ~ ____ ~ _ _ _ _L __ _ _ _L __ _ ~ _ _ _ _ _ L_ _ _ _ ~

Liquid Camp Mol % Acetonitrile

760mm 150mm

Figure 16.14 Vapor-liquid equilibrium for water and acetonitrile at (---) 760 mm and (.....) 150 mm.
436 HANDBOOK OF DOWNSTREAM PROCESSING

pure water. Column 1 distillate is fed to column 2 which operates at 760 mm.
This change in pressure means the feed to column 2 is now on the other side
of the azeotrope point. Thus the distillate is the azeotrope at 760 mm which
is recycled to the first column while the bottom component is nearly pure
acetonitrile. Obviously this technique is only appropriate for azeotropes that
show large enough deviation with changes in pressures. If the change in
azeotrope composition with pressure is small, then the recycle can become
enormous or the pressure change required can be so great as to make the
cost of this scheme prohibitive. A good rule of thumb for preliminary screening
is that the change in azeotrope composition should be at least 7-10 mol%.
The vapor-liquid diagram must also be checked for pinch points as the
concentration approaches the azeotrope. It does little good if the azeotrope
composition changes by 10 mol% if a pinch is formed which makes it nearly
impossible to reach the azeotrope. Once a system has been assessed to be a
candidate for pressure swing distillation the optimum pressure levels need
to be determined.
The number of pressure levels to choose from is endless without some
constraints. These constraints are listed below:
1. At least one column should be at a pressure to allow air or water
condensing at a temperature approximately 20 to 40°F (1O-25°C) above
the cooling medium temperature. This column will be the anchor point
column.
2. Avoid vacuum operation if possible. The increase in equipment size and
need for vacuum equipment makes it costly.
3. For a first trial choose a pressure change in the second or swing column
which gives a 7-10% change in the azeotrope molar composition.
4. Choose a pressure for the high pressure column which will allow the
overhead to reboil in the low pressure column if possible.
A cost estimate is done at selected pressures and then this is compared to
the cost of systems where the pressure of the swing column is 1.5 times higher
and 1.5 times lower. If the selected pressure scheme does not give the minimum
cost then take the minimum and bracket it with a higher and lower swing
pressure at 1.3 times the minimum swing column pressure. Continue until a
local minimum is achieved. This will correspond to the optimum pressure
levels for a pressure swing distillation system.

16.2.1.5 Batch distillation. Batch distillation (see Figure 16.15) is similar

in theory of separation to simple distillation but does not run continuously.
Two types of batch distillation are normally encountered. The first and most
common is batch rectification in which the feed is charged to the column
and boiled up through a refluxing column with distillate product being drawn
off. Conversely batch stripping, also known as the inverted batch column,
has the feed charged into the reflux drum and impurities stripped from the
 DISTILLATION TN THE PHARMACEUTICAL INDUSTRY 437
Cooling
Media

Reflux Product

N
-----,
'-L-r~ Heati.ng
11 ~ Media

--{>1-- 'u-.J 'I----.L.v;}_

Initial charge Residue

Figure 16.15 Batch rectification distillation.

falling liquid producing a bottoms products. Batch distillation is a non-steady

state phenomenon with the composition changing throughout the column
with time as material is drawn off.
Batch distillation is preferred over continuous distillation under certain
circumstances. The following criteria can be used to determine if batch
distillation is the separation option of choice.
t. Excellent for small quantities of materials handled at irregularly scheduled
periods.
2. Excellent for dampening the effects of widely varying feed compositional
changes.
3. Lower capital cost for producing multiple products than continuous
operation. The batch option requires only a single column while the
continuous operation requires N - 1 columns for N products.
4. Tolerant of uncertainties in VLE.
5. Excellent for handling residues which have a tendency to foul column
internals and heat transfer surfaces.
6. Excellent for pharmaceutical and biotechnological process because the
equipment can be cleaned and sterilized between each batch run.
On the down side, batch distillation is more expensive per kilogram of
product and requires more operator attention and instrumentation control
due to the variation of operation with time. After each batch charge the
system must be shut down, drained, possibly cleaned, and then recharged
with feed causing a loss of on-stream time.
438 HANDBOOK OF DOWNSTREAM PROCESSING

The batch distillation can be operated in two modes: It either operates at

a constant reflux rate which will produce a variable product composition
high in the lightest material at the initial starting time and progressively
decrease in purity, making the product composition an average of the
composition throughout the product draw. Alternatively the reflux can be
varied to produce a constant product composition until prevented by material
balance limitation[17]. This method will require constant analysis of the
product to be fed back to the reflux controller to make the changes in the
rate required to maintain product purity. Though these are the two most
common operating methods they are not necessarily the optimum ones [18].
When making pure products regardlesss of the reflux mode there will be
a period of time when the amount of the light component is too small to
meet the product specification but still too great to allow the next product
to reach its purity specification. Therefore a slop cut is taken which removes
the last of the light component and some of the next component. These slop
cuts between products are dependent on product purity. Higher purity
requires larger slop cut. The slop cuts are normally recycled back with the
next charge. Redistillation of these components is expensive so the slop cuts
should be minimized by specifying the minimum product purity needed and
proper operation of the column to obtain the greatest separation possible.
The method of recycling the slops can be optimized to reduce operating costs
[19]. The slops can all be combined and fed with the next feed

charge or they can each be fed continuously at the appropriate time into
the column at an optimum location. Another option is to collect the slop
cuts over a number of batch cycles and distil each slop cut when enough
material has been collected.
A key variable which is encountered in batch distillation but not in
continuous distillation is the effect of hold-up. The size of the hold-up in the
charge drum can have a major impact on the sharpness of separation[20]. A
batch column will achieve a sharper separation if the amount of material in
the charge drum is small at the time of transition from one product to another.
This can be reflected as the larger the charge drum liquid the larger the slop
cut between products for sharp separations. This presents an inherent problem
in scaling up batch distillation from laboratory and pilot sized equipment
since there will be a difference in hold-up which must be accounted
for.

16.2.1.6 Molecular distillation. Molecular distillation differs from ordinary

distillation in a number of ways. First since very low pressure is used, as low
as 0.001 mbar, the temperature the system experiences is moderate and with
the inherent short residence times, heat sensitive high molecular weight
materials can be separated effectively. Secondly the vapor phase is produced
by surface vaporization not boiling. The maximum evaporation rate can be
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 439
assumed to be a function of the pressure, molecular weight, and the
temperature I2 1,22].
v = 5.83 x 1O- 2 p(MWjT) (16.3)
where V is in gjs-cm 2 , p is in mmHg, MW is the molecular weight, and T
is in oK. This rate is based on the provison that none of the vapor leaving
the liquid surface returns. In other words, the molecules that evaporate must
not encounter any other molecules and collide, thus returning molecules to
the liquid. To meet this requirement, a very low vacuum is required to
minimize the molecules per unit volume along with a short travel time to
the condenser which minimizes the time available for collisions. In conventional
vacuum distillation, a molecule on average will be vaporized about a thousand
times before reaching the condenser; for true molecular distillation, every
molecule that escapes the liquid reaches the condenser. In most cases to
obtain this close spacing the condensing and evaporating surfaces are parallel
to each other and separated by a distance of less than the mean free path of
the evaporating molecule. The mean free path is the average distance a
molecule travels between collisions and like the maximum evaporation rate
is a function of pressure and molecular weight.
In molecular distillation boiling does not occur to aid liquid mixing and
both high molecular weight and high viscosity tend to hinder diffusion.
Efficient molecular distillation requires mechanical mixing to eliminate
composition gradients caused by low natural mixing. This is done in three ways:
1. Agitating the bulk liquid
2. Utilizing a falling film
3. Mechanically wiping the film along the wall.
Separation occurs not because of boiling point differences but by the
difference in the evaporation rateI21.22] (see equation 16.3). This driving force
is set only by the molecular weight difference since the temperature and
pressure experienced by each component will be identical.
The vacuum in a typical molecular distillation can be as low as 0.001 mbar
to increase the mean free path to an acceptable value. The mean free path
of a molecular distillation system may be approximately 2-3 cm I21 ]. To
accomplish this a vacuum system must be designed to reach these low
pressures. As a consequence, if there are any volatiles, provision must be
made to degas the feed prior to entering the molecular distillation unit.
Volatiles at these low pressures include many components one would not
normally consider, such as water and even light oils if a wax is being processed.
Care must be taken to allow degassing of the feed. With the low pressures
required it does not take much volatile to overwhelm the vacuum system.
Similarly if care is taken to minimize the chances of collisions in the vapor
space before reaching the condenser one must also make sure that the
molecule that reaches the condenser does in fact condense. This can usually
440 HANDBOOK OF DOWNSTREAM PROCESSING

be achieved by providing a temperature difference of lOOoe between the

evaporating surface and the condensing surface.
Short path distillation differs from molecular distillation by the distance
between the vaporizing and condensing surface. Where this distance is the
molecular mean free path in molecular distillation, it may be twice this
distance for short path distillation. This allows a larger unit to be more easily
and cheaply built for production of up to 10000 kg/h. The drawback is that
more collisions between vapor molecules will occur meaning only 40 to 60%
of the molecules evaporated will reach the condenser without returning to
the liquid, unlike the 90% obtained in molecular distillation.
Molecular or short path distillation is used for viscous heat sensitive
materials which cannot be separated by more conventional methods. Product
degradation can occur with some materials if maintained at high temperature
for an extended period of time. A commercial short path unit will operate
at the lowest theoretically possible temperature and at the shortest possible
residence time presently obtainable. By comparison, the hold-up time for a
batch column is of the order of 400 times longer than a short path unit[221.
This means a minimum of product degradation.

16.2.1.7 Supercriticai fluid extraction. Supercritical fluid extraction (SFE)

is not a new phenomenon as it was first reported in 1879[23]. However, little
motivation until recently existed for its development due to cheap energy
and few if any health regulations. Many studies were carried out to explain
why a system near its critical point vaporizes when the pressure is raised
while the converse, raising temperature, causes condensation. This underlying
phenomenon gives rise to the benefit of supercritical fluid extraction. To
appreciate this effect take the example of supercritical ethylene and naphthalene.
For this system at 100 atm and 12°e the observed solubility of naphthalene
in ethylene is more than 15000 times greater than if ethylene were an ideal gas[241.
The supercritical phenomenon can be explained if it is understood that
under these conditions the compound is no longer a gas but has become a
dense fluid whose density resembles that of a liquid more than a gas. This
explains the seemingly abnormal behavior ofSFE; with an increase in pressure
we see an increase in solubility as the supercritical fluid becomes more dense,
but as temperature is increased, the density decreases therefore the solubility
drops and material leaves the solution. Figure 16.16 shows the change in
density for a pure component near its critical point. While other properties
such as the transport properties, diffusivity and viscosity remain gas-like
above the critical point and surface tension is zero. This allows excellent
penetration by a high density fluid into microporous materials for better
extraction from solid substances.
As mentioned previously the solubility of one substance in another is based
on minimizing the free energy of a system. A substance will only dissolve up
to the point where the system's free energy is a minimum and then no more
 DTSTTLLATION IN THE PHARMACEUTICAL INDUSTRY 441

2.0

" <L0
~,

<La::
1.0

0.1 1.0 10.0

Figure 16.16 Behavior of a typical pure component near its critical point.

can dissolve unless the system changes. Factors affecting the system's free
energy are the intermolecular forces between the like and the unlike molecules.
The intermolecular attractive effects are inversely proportional to the molar
volume which means that as the pressure increases the density increases and
therefore the molar volume decreases. This results in the solubility increasing
nearly exponentially with the density.
The trend in SFE is to use compounds which are non-toxic and whose
critical pressure is usually less than 100 atm and critical temperature is close
to ambient. The main workhorses of the SFE field have been carbon dioxide
and light hydrocarbons. The properties of carbon dioxide, being non-toxic,
non-hazardous, non-corrosive if dry, inexpensive, readily available, and the
most studied compound in the supercritical region, make it the choice for
the food and pharmaceutical industry to produce non-contaminated products.
But both carbon dioxide and the light hydrocarbons are non-polar, non-
hydrogen bonding and therefore can rely only on dispersion as the attractive
force to cause a separation. Many compounds to be separated are polar,
form hydrogen bonds, or are electron donor-acceptor complexes which the
above SFE compounds cannot exploit. To take advantage of these effects, a
cosolvent can be added to the original supercritical fluid to enhance its
attractive forces. The cosolvent cannot usually be used by itself because of
its high critical pressure, which would increase costs to the system, or its
442 HANDBOOK OF DOWNSTREAM PROCESSING

high critical temperature which may cause product degradation. An example

is the extraction of benzoic acid by carbon dioxide at 35°C and 200 atm
which can be more than doubled by adding as little as 3.5 mol% of a hydrogen
bonding compound like methanol or acetone[251.
Little commercial work had been done until recently. The new motivation
for SFE is the following:

• Increase in governmental requirements for removal of toxic, harmful,

and/or unfavorable components without damage to the product
• Desire for higher purity heavy molecular weight products
• Heat sensitive materials
• More stringent pollution control
• No other separation method is viable
• Rise in energy costs making distillation more expensive.

Some industrial applications in commercial use include caffeine removal

from coffee, Kerr- McGee's ROSE process, cholesterol removal from butter,
and oilseed extraction[ 241.
New work in the field consists of both commercial applications and trying
to understand and develop models for supercritical fluid extraction. Work
to date has normally centered around an initial screening of conventional
separation methods. If none of these methods, i.e. distillation, liquid-liquid
extraction, vacuum distillation, proves viable, then some of the more exotic
methods like SFE can be investigated. But to determine the viability of SFE
a feasibility bench scale test and then pilot tests are required, which are time
consuming and expensive. If a proven simulation were available then the
initial SFE screening and economical evaluation could be done cheaper and
faster saving the pilot test for verification and optimization purposes.
Numerous SFE models have been proposed but none has achieved industry-wide
acceptance.
The advantages of SFE are that it can be used to separate heat sensitive
materials of high molecular weight with much better purity than other
commercial methods and with little or no residual solvents. Also SFE is a
function of temperature, pressure, and choice of solvent so these three
parameters can be varied to achieve the desired product. The solvent can be
a safe, non-toxic compound for food and pharmaceutical use. The higher
transport properties and zero surface tension allow excellent penetration into
solids enhancing extraction. With the addition of cosolvents even better
extraction capabilities can be attained. A major disadvantage is the absence
of readily available modeling techniques thus requiring expensive tests.
Though SFE may reduce energy requirements, the capital outlay is high
because of the high pressure equipment. The SFE process may add 15 cents
per pound (33 cents per kg) to the cost of decaffeinated coffee[ 26 1.
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 443
16.2.2 Distillation design criteria
16.2.2.1 Thermodynamic equilibria prediction. There is a plethora of
methods to predict thermodynamic equilibrium. Table 16.7 lists many, with
the most commonly used emboldened, and their applicability.
The engineer responsible for simulating a distillation separation now must
decide which model to use. There are expert systems available such as
CONPHYDE and Simulation Science has one incorporated in its simulating
program. Likewise there are flow charts which help in selecting the right
model[ 27 1. Some simple conditions to take into consideration prior to selecting
a model are:
1. What are the key components to be separated, and based on their vapor
pressures, what is the relative volatility?
2. Are there any minor components which may affect the key component
separation?
3. Do experimental data exist for the composition, pressure and temperature
range of interest?
4. Are there any significant polar interactions?

Table 16.6 Methods for predicting thermodynamic equilibrium

Model Applicability

Raoul!'s law For nearly ideal non-polar systems at low

pressure and high relative volatility
Equation of state (EOS) For non-polar to mildly polar systems up to
Soave-Redlich-Kwong (SRK) very high pressures. Requires little input so
Peng-Robinson (PR) they are widely used in the hydrocarbons
Benedict-Webb-Rubin (BWR) industry. Interaction parameters are
Chao-Seader (CS) required for low relative volatility and the
Grayson and Steed (GS) mildly polar systems to enhance the
Chain-or-rotators (COR) accuracy. Some newer EOSs are being
Perturbed-anisotropic developed to handle highly polar systems
chain theory (PACT) but these are not in wide use. These include
Perturbed hard chain PACT, COR, and PHCT.
theory (PHCT)

Liquid activity coefficient (LACT) For highly polar systems at low to moderate
Margules pressure. Requires experimental data on
Van Laar each binary pair.
Wilson
NRTL
UNIQUAC

Group contribution method For highly polar systems at low to moderate

UNIFAC pressure. Requires no experimental data just
ASOG interaction parameters for the various
groups that make up the molecules. These
group interactions are provided with most
programs.
444 HANDBOOK OF DOWNSTREAM PROCESSING

5. Are there any pinch points, azeotropes, or partial miscibility which affect
the key separation?
6. What is the required product purity?
7. What is the accuracy required? Is it for a preliminary study or the final design?
S. Can an experimental test be run?

From this information and reference material one can make the proper
choice for a thermodynamic equilibrium model, but be aware of possible
pitfalls. The LACT model will be the most accurate but only when used
within the range of the experimental data. Extrapolations in pressure up to
approximately 60% of the critical point on the binary are possible. Beyond
that, the liquid activity is affected by pressure. As the system becomes
compressible in the liquid phase the molar volumes become variable. The
EOSs and group contribution models require little input data which makes
them easy to use but these models are the most likely to give erroneous
results when used on a system for which they are not intended. If there is
any doubt a check can be made on a homologous system for which
information is known. For example, if UNIF AC is the only method short
of running experimental tests for triethylamine and methyl ethyl ketone but
data exist for the binaries triethylamine and acetone and also methyl ethyl
ketone and butylamine, then a check can be made of the validity of the
UNIF AC method for the known binaries. If it reflects the data well, it should
also simulate triethylamine and methyl ethyl ketone well.

16.2.2.2 Design parameters. When designing a distillation system the main

parameters to consider are the operating pressure, product specifications,
number of theoretical stages, minimum reflux, type of internals (i.e. trays or
packing), and stage efficiency. Once these are determined the distillation
system can be sized and a cost estimate performed.
The operating pressure of the distillation column determines the overhead
condensing temperature and the boil-up temperature. If the pressure is set
too low the condensing temperature may require mechanical compression
or refrigeration both of which are very expensive. Likewise, if the pressure
is too high the bottoms component temperature may exceed the heating
medium temperature that is available, or cause product degradation. The
optimum operating temperature should give an overhead temperature that
allows either air or water cooling. This temperature should be approximately
20 to 40°F (10- 2 SOC) above the cooling medium temperature. Thus if cooling
water is available at 90°F (32°C) an operating pressure giving a condenser
outlet of 120°F (49°C) would be preferred. The heating medium should be
the lowest temperature available as this is usually the cheapest but at least
20 to SO°F (1O-4SOC) above the temperature of the bottoms component.
Above 100°F (55T) temperature difference between the heating medium and
the bottoms component may cause film boiling in the reboiler which will
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 445
seriously impede the heat transfer. The operating pressure also affects the
column size. At higher pressures the column diameter required is reduced
because the vapor phase is at a higher density and the diameter is directly
dependent on the volumetric flow rate of the vapor. But as the pressure
increases the relative volatility of the components decreases requiring more
stages to make the separation. During the simulation it is a good idea to
vary the pressure while staying within the applicable temperature constraints
and determine if the diameter can be reduced with little effect of the number
of stages. In pharmaceutical applications the product integrity may override
the above constraints. The product may not withstand the temperature of
the bottoms without degradation while using air or cooling water for
condensing. The system pressure would then be set by the temperature
constraints of the products. This may entail the use of refrigeration and
vacuum equipment.
Determination of the number of theoretical stages, reflux ratio and feed
point should first be done using a short cut method such as the Fenske
equation[28 1, the Underwood equation[30,31 1, and the Erbar-Maddox
equation[29 1• The Underwood method will determine the minimum reflux
given infinite stages. The optimum reflux is around 1.1 times the minimum.
Since the number of stages required between minimum reflux and 1.1 times
the minimum goes from a number less than 200 to infinity it is best, and less
than a 2-6% cost increase, to be higher than 1.1 times the minimum reflux.
The recommended reflux ratio is in the range of 1.15 to 1.25. The Fenske
equation will determine the number of theoretical stages required and the
Erbar- Maddox equation will determine the optimum feed location. These
equations have some very limiting assumptions, such as equal molar overflow.
The values determined can be used as a starting point for rigorous simulations
or for preliminary studies to determine the relative cost difference between
distillation schemes.
The rigorous simulation is performed on a computer. In a simulation every
significant relationship involving material flow, energy transfer, and vapor-liquid
equilibrium is converged under the assumed conditions. There are many
PC-based simulator programs such as Simulation Sciences PRO II, Deerhaven
Technical Software's PD-PLUS, ChemShareAspen and Chemstation to name
a few. The simulation is run using the preliminary reflux ratio, number of
stages, and feed location determined above. After this initial run a brief
parametric study should be performed. The feed location should be raised
and lowered to see the effect on separation. Likewise the reflux should be
lowered until the separation degrades to determine the zone of confidence
that the column would operate in. Once the optimum distillation column
has been simulated the equipment can be sized.
In order to size the distillation system the column internal components
must be determined. Two major classes of internal components exist, either
trays or packing. Table 16.7 is intended to assist in this selection.
446 HANDBOOK OF DOWNSTREAM PROCESSING

Table 16.7 Parameters to be considered in choice of internal components

Parameter Type Rating

Pressure drop per theoretical stage Trays Poor

Random packing Fair
Structured packing Good
Minimal liquid hold-up Trays Poor
Random packing Fair
Structured packing Good
Low liquid rate performance Trays Good
Random packing Poor
Structured packing Poor
High liquid rate performance Trays Poor
Random packing Good
Structured packing Good
Cost of internals Trays Good
Random packing Fair
Structured packing Poor
Cost of column plus internals Trays Good
Random packing Fair
Structured packing Fair
Liquid-to-vapor ratio range Trays Poor
Distillation operations are normally low Random packing Good
but absorber and scrubbers can be high Structured packing Good
Foaming systems Trays Poor
Random packing Fair
structured packing Good
Solids present Trays Good
Random packing Fair
Structured packing Poor
Corrosion resistant materials Trays Fair
Random packing Good
Structured packing Poor
Columns less than 46 em diameter Trays Poor
Random packing Good
Structured packing Fair
Turndown Trays Good
Random packing Fair
Structured packing Fair

16.2.2.2.1 Tray Types. Trays are the most common distillation column
contacting device. They have the highest overall operating range and their
design methods are well known and proven. But as more work is done and
knowledge is gathered the use of trays may decrease in certain applications
as packing becomes more competitive.
The two main tray types used in pharmaceutical distillation today are the
sieve and valve. Other types such as bubble cap, cross-flow, dual-flow, and
baffle exist but are for special applications and will only be mentioned briefly.
Sieve trays are the simplest and, therefore, the least expensive tray. A sieve
tray is made from a flat sheet of metal perforated across the entire surface
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 447
with 0.32 cm (1/8") to 2.54 cm (I") diameter holes. The number and sizes of
these holes are determined mainly by the vapor and liquid volumetric
throughput. Although it is possible to design the tray this task is normally
left to the tray supplier. This is also true for valve and bubble cap trays.
Sieve trays are good general purpose devices but have minimal operating
range. See the tray selection guide in Table 16.8, for characteristics.
Valve trays are similar to sieve trays in that the base is a perforated plate
on which are placed the valves which are either movable disks or rectangular
metal strips over the holes. The rising vapor lifts the coverings in proportion
to its flow rate allowing the pressure drop across the tray to remain nearly
constant until all the valves are fully opened. The valves are constructed with
stops to prevent high vapor flow from lifting them out of the hole. These
valves will also reseat themselves at low vapor rates to prevent liquid from
leaking through the tray perforations as would happen with sieve trays. These
movable valves allow a very wide operating range for this type of tray.
Bubble cap trays have been around for a long time. They are mainly found
in older service or where very high operating range is required and will justify
the extra cost or if a minimum liquid level is required on the tray regardless
of the vapor rate.
In dual-flow trays both the rising vapors and falling liquids must share the
same holes while flowing through the column. Dual-flow trays have no
downcomers. These trays though less expensive than trays with downcomers
have a very narrow operating range, and the efficiency is similar to sieve
trays, but only if operated at the optimum design point after which efficiency
falls rapidly. This limits their use to highly fouling services with minimal
turndown[171.
In baffle trays the falling liquid forms a curtain of liquid as it rains down
from the upper baffle to the one below. The rising vapor comes into contact
with the liquid as it flows through this curtain. These trays have the same
limitations as the dual-flow trays and are used for mainly fouling services.
Sometimes, if only a few stages are required they may be used in vent gas
scrubbing.

Table 16.8 Tray characteristics

Sieve Valve Bubble Dual-flow

Operating range (% of flood) 60-80 25-80 10-80 75-80

Efficiency 1 1 2 4
Fouling resistance 2 3 4 1
Cost factor 2 3 4 1
Foaming systems 2 2 3 2
Vent scrubbing 2 2 3 1
Viscous fluids 2 2 3 3
Low pressure drop 1 2 3 1

1-Excellent, 2-good, 3-fair, 4-poor.

448 HANDBOOK OF DOWNSTREAM PROCESSING

16.2.2.2.2 Packing types. The biggest drawback to the use of packing is

the lack of accurate generalized design methods. Until the late 1970s the
design had been based primarily on past experience or vendor input. This
led most engineers to specify trays because of the comfort factor. However,
with effort by university laboratories, research institutes, and packing
manufacturers there has been an increase in data available to the engineer
to improve the comfort factor when specifying packing. Packing can be
categorized into two groups. Random (or dumped) packing and structured
packing.
The contacting mechanism for packing is such that the falling liquid forms
a film on the packing surface that contacts the rising vapor. With this
mechanism it is very important that the liquid wets the packing material to
improve the formation of a thin film to obtain good mass transfer. In other
words plastic packing would not be the best choice for an aqueous system.
Also as the tortuosity of the vapor path increases the contacting surface
increases thus decreasing the height of packing equivalent to a theoretical
plate (HETP). This could increase the pressure drop and vapor velocity
beyond design limits, so it is important to balance the packing type and size
with the design limits of the system. Detail design and selection of packing
are beyond the scope of this work. There are many excellent sources (see the
Bibliography) to assist in determining the best packing type and size. Much
work is in progress as new types of packing are developed.
Random packing as the name implies is placed in the distillation column
in no set order. As a matter of fact the packing is actually dumped in place.
For breakable packing more care is taken in the dump, such as dumping
into the column filled with a liquid to reduce the settling velocity. The
materials of construction for packing include most metals and plastics but
also ceramics, thus greatly enhancing the range of corrosive material that
can be handled.
The mass transfer for random packing can range from 50-183 cm (20-72").
As with all packing the larger the size the smaller the available surface area
which reduces the transfer efficiency but increases the capacity. Care must
be used in choosing a packing type with low HETP but still having the
required capacity at an acceptable pressure drop. Many packing vendors
have introduced new high efficiency random packing in the past decade. This
has greatly increased the uses for random packing. Typically a packing can
be found to supply an HETP of 76.2 cm (30") and this should be used for
preliminary design and quite possibly when a packing vendor is called and
the design has been firmed up, this number can be lower and also guaranteed.
Structured packing comes in two types, corrugated textured sheets or gauze.
The first type makes use of a constant configuration of geometrically arranged
corrugated sheet throughout the column. The main feature that allows better
mass transfer than random packing is the surface texture that is applied to
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 449
the sheet[32]. The type of texture is unique to the manufacturer. This type of
packing is available in metal, plastic, and ceramic, although the metal has
the best texturing capabilities. Also by increasing the corrugation depth the
capacity can be increased for a constant pressure drop but causes reduced
mass transfer efficiency. Typical values for HETP that can be obtained range
from 30.5-76.2cm (12 to 30"). For preliminary designs 50.8 cm (20") can be used.
Gauze-type structured packing is different in that the corrugated sheets
are made from a woven fabric. This fabric can be either a woven metal or
a plastic fabric. The gauze type packing is characterized by the highest mass
transfer efficiency or as commonly noted, the lowest HETP available. The
HETP value ranges from 45.7 cm (6 to 18") depending on the system, with
30.5 cm (12") being a good average number to be used in preliminary work.
On a cost per cubic meter basis, structured packing is more expensive than
random packing. With higher capacity and better efficiency the cost per
theoretical stage can be less. Structured packing may also be required to
meet a lower pressure drop requirement or for low liquid loadings. Table
16.9 gives characteristics for these three types of packing.

16.2.2.3 Column sizing. Once the type of internal component is selected

the column can be sized. The height of the column is determined by the tray
efficiency or the height equivalent to a theoretical plate (HETP) for packing.
The determination of this value for either packing or trays lags behind all
the advances in computer simulation and thermodynamic equilibrium
prediction. The main reason for the inability to predict this value is the
difficulty in determining the diffusion coefficient of the components in the
liquid and vapor phases. Therefore only simplified charts such as the
O'Connell chart for trays or published data on a system such as those found
in McFarland et al. [33] and Vital et al.[34] for trays or Bolles and Fair [35]
and Vital et al.[36] for packing are available or lastly, the predictive method
can be used. The procedure for calculating the efficiency of a tray is as follows:

Table 16.9 Packing characteristics"

Liquid loading
(GPM/ft 2 )
HETP P /theo
stage
Packing type (in) (mmHg) Min Max

Random 30 1-2 0.4 70 (2")

125 (3.5")
Structured
Textured sheets 20 1-1.5 0.2 12-18
Gauze 12 0.1-0.5 0.1 5-10

"The values for HETP assume median conditions for liquid loading and pressure drop along
with moderate operating conditions.
450 HANDBOOK OF DOWNSTREAM PROCESSING

1. Estimate using the following predictive methods:

For trays- MacFarland[33 1, Sigmond, Van Winkle method and the
AIChE method.
For packing- Bolles and Fair method[ 35 1.
2. Search the literature for information on your system or a similar system.
3. Contact tray or packing vendors for their input.
4. Choose the value from actual data for the same system from a reliable
source. If not, choose the lowest calculated value for tray efficiency or
highest HETP closest to a similar system.
5. Add 5% to the number of trays or the height of packing as a safety margin.
The tray spacing required for good operation is between 45.7-76.2 cm (18
and 30") with 60.9 cm (24") being the norm. The tray spacing is set to minimize
the entrainment of liquid with the vapor to the tray above. For a foaming
system where entrainment is a greater possibility, the tray spacing or the
column diameter is increased. From equation (16.4) or (16.5) below plus the
miscellaneous spacings shown in Figure 16.17 for trayed columns and Figure
16.18 for packed columns the total height of the column can be determined.
H t = 1.05(Nt slefficiency) x TS (16.4)
Hp = 1.05(HETP x N ts ) (16.5)

OVERHEAD VAPOR

--~
... VAPOR DISENGAGEMENT

REFLUX I 1___ 36"

TOP TRAY

i 30" r SPACING AT MANWAY

FEED NOZZLE I 30':

_L
r FEED TRAY SPACE

18"'2~'1
BOnOMTRAY
------t- TRAY SPACING
(TYPICAL)

REBOILER RETURN ~. 36" !

HIGH LIQUID LEVEL

BOTTOMS OUTLET AND REBOILER FEED

Figure 16.17 Typical spacing for a trayed distillation column.

 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 451
0.2
,::"'.5...
0.1 '.0_ r-~

--'"
O.SO
0.06
0.04 r--....
0.25 t-....
0.02
· ... "7

·.
~
f.0.l0
......
.. ...
•.. IJ
0.01 1L.3
':"P.0.05
0.006 ClN
0.004 ..,
N
,:"p. pn. HaO/IIl
""
0.002

0.001
"I'\.r'\
1,1\
1\
0.0006
0.0004
0.004 0.04 0.4 4.0 I. o
0.006 0 .01 0.02 0.06 0.1 0.2 0.' 1.0 2.0 8.0

.J:..
G
[.!!.-] ..
II.

Figure 16.18 Generalized pressure drop correlation. G is the gas mass flow in Ib/ft 2 -s (kg/m2 - s),
G* is the gas mass flow in Ib/ft 2 - h (kg/m 2 - h), L is the liquid mass flow in Ib/ft 2 - h (kg/m 2 - h),
F is the packing factor, Pv is the gas density in Ib/ft 3 (kg/m 3 ), PL is the liquid density in lb/fe
(kg/m 3 ), and v is the kinematic viscosity (cSt).

where HETP is the height equivalent to a theoretical plate, H t is the height

of trayed section, Hp is the height of packed section, N ts is the number of
theoretical stages, and TS is tray spacing, usually 60.9 cm (24"). The trayed
or packed column diameter is determined based on vendor specific equations
or the equation from Fractionation Research Inc. (FRI).
For preliminary sizes the trayed column can be estimated from the following
equations:
Column diameter
(16.6)
Tray efficiency
log eft' = 1.67 - 0.25(log(Il(X)) + 0.30(log (L/V)) (16.7)
Column height
He = H t + misc spacings (16.8)
where He is the total height of column (ft), D t is the total height of column
(ft), L is the liquid rate of tray with maximum vapor rate (lb/h), V is the
452 HANDBOOK OF DOWNSTREAM PROCESSING

OVERHEAD VAPOR

REFLUX 1
FEED NOZZLE f

FLOW DISTRIBUTOR

-1
30"

--- - PACKED SECTION

20' MAX

REBOILER RETURN 1-
3;t
HIGH lieu 10 LEVEL - - - -1-
,
BOTTOMS SUMP FOR
LIQUID HOLD·UP AND
REBOILER HYDRAULICS
IF REQUIRED

BOTTOM OUTLET AND REBOILER FEED

Figure 16.19 Typical spacing for a packed distillation column.

maximum vapor rate (lb/h), J-l is the average liquid viscosity (cP), ex is the
average relative volatility of key components, Pv is the vapor density of V
(lb/ft 3 ), and PI is the liquid density of L (lb/ft 3 ).
Preliminary sizes for a packed column can be determined by the following
method. From Figure 16.18 and the packing factor obtained from the vendor
catalog choose a diameter that allows operation between the 2.54 cm (1.0")
and the 1.27 cm (OS') operating line. The column height can be determined
using an HETP value from packing vendors or from Bolles and Fair[ 35 1 or
Vital et af.l 36 1. Then by adding the miscellaneous spacings from Figure 16.18
the total height can be determined.
These calculations will give conservative sizes for the trayed and packed
columns which will be appropriate for cost comparison and rough order of
magnitude estimates.
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 453
References
1. Murrel, J.N. and Boucher, E.A. (1982). The Properties of Liquids and Solutions, John Wiley,
New York.
2. Hildebrand, J.H. and Scott, R.L. (1962). Regular Solutions, Prentice Hall, New Jersey.
3. Sharp, K.A., Nichollis, A., Fine, R.F. and Honig, B. (1991). Science, 252, 106.
4. Anderson, R., Cambio, R. and Prausnitz, lM. (1962). AIChE J., 8, 67.
5. Anderson, R. and Prausnitz, J.M. (1961). AIChE J., 7, 97.
6. Barton, A.F. (1983). Handbook of Solubility Parameters and other Cohesion Parameters, CRC
Press, Florida.
7. Garner, F.H. and Ellis, R.M. (1951). Trans. Inst. Chem. Eng., 29.
8. Van Winkle, M. (1967). Distillation, McGraw-Hill, New York.
9. Berg, L. (1969). Chem. Eng. Prog., 65,52.
10. Hoffman, E.I. (1964). Azeotropic and Extractive Distillation, Interscience, New York.
11. Horsley, L.H. (1973). Adv. Chem. Serv., 116.
12. Scheibel, E.G. (1948). Chem. Eng. Prog., 44, 927.
13. Null, H.R. and Palmer, D.A., Chem. Eng. Prog., 65, 47 (1969).
14. Walas, S.M. (1988). Chemical Process Equipment Selection and Design, Butterworth, Boston, MA.
15. Bastos, lC, Soares, M.E. and Medina, A.G. (1988). Ind. Eng. Chem., Res, 27, 1269.
16. Thomas, E.R. and Eckert, CA. (1984). Ind. Eng. Chem., Process Res. Dev., 23, 194.
17. Rousseau, R.W., Handbook of Separation Process Technology,
18. Diwekar, U.M., Malik, R.K. and Madhaven, K.P. (1987). Comput. Chem. Eng., 11, 629.
19. Quintero-Marmol, E. and Luyben, W.L. (1990). Ind. Eng. Chem. Res., 29, 1915.
20. Schweitzer, P.A. (1979). Handbook ofSeparation Techniquesfor Chemical Engineers, McGraw-
Hill, New York.
21. Perry, R.H. and Chilton, CH. (1973). Chemical Engineers Handbook, 5th edn, McGraw-Hill,
New York.
22. UIC Inc. Bulletin (1990). Short-path Distillationfrom Laboratory to Production.
23. McHugh, M.A. and Krukonis, J.V. (1989). Supercritical Fluid Extraction Principles and
Practice, AiChe Today Series, New York.
24. McHugh, M.A. and Krukanis, Y.l (1986). Supercritical Fluid Extraction Principles and
Practice, Butterworth, Boston, MA.
25. Dobbs,J.M., Wong, lM., Lahiere, R.I. and Johnston, K.P.(1987). Ind. Eng. Chem., Res., 26, 56.
26. Chemical Week (1987). June 24, p. 66.
27. Wilcox, R.F. and White, S.L. (1986). Chem. Eng., Oct. 27, p. 141.
28. Fenske, M.R. (1932). Ind. Eng. Chem., 24, 482.
29. Ebar, lH. and Maddox, R.R. (1961). Petrol Refiner, 40(5), 183.
30. Underwood, AJ.V. (1945). J. Inst. Petrol, 31, 111.
31. Underwood, A.J.V. (1946). J. Inst. Petrol 32, 598, 614.
32. Brovo, lL., Ashutosh, P.A. and Edgar, T.F. (1992). Ind. Eng. Chem. Res., 31, 604.
33. McFarland, S.A., Sigmund, P.N. and Van Winkle, M. (1972). Hydro Proc., 51, 111.
34. Vital, T.I., Grossel, S.S. and Olsen, p.r. (1984). Hydro Proc., 147.
35. Bolles, W.L. and Fair, R.l (1979). Ind. Chem. Eng. Symp. Series, 56, 33.
36. Vital, TJ., Grossel, S.S. and Olsen, p.r. (1984). Hydro Proc., 75.
37. Ewell, R.H., Harrison, lM. and Berg, L. (1944). Ind. Eng. Chem., 36, 871.

Bibliography
Physical properties

Battino, R. and Clever, H.L. (1966). Chem Rev., 66, 395.

Thorough review and compilation of gas solubilities in liquids.
Boublik, T., Fried, V. and Hala, E. (1973). The Vapor Pressure ofPure Substances, Elsevier, Amsterdam.
Correlation and compilation of vapor pressure data for pure fluids. Normal and low pressure
region.
Chu, J.C, Wang, S.L., Levy, S.L. and Paul, R. (1956). Vapor-liquid Equilibrium Data, J.W.
Edwards, Ann Arbor, MI.
Collection of vapor-liquid equilibrium data.
454 HANDBOOK OF DOWNSTREAM PROCESSING

Claxton, G. (1958). Physical and Azeotropic Data, The National Benzene and Allied Products
Association, London.
Compilation of azeotropic data as well as other physical properties including melting and
boiling points.
Design Institute for Physical Property Data (DIPPR) (1990). Data Compilation Tables of
Properties of Pure Compounds, AICHE, New York.
Source for most pure component properties for over 1000 compounds. Includes regressed
equations for temperature dependent properties. Equations have upper and lower temperature
limits and do not extrapolate well.
Faldix, P., Fortsch. Ve~fahrenstech., 6, 440 (1965); 7, 650 (1967); 8, 625 (1969); 9,561 (1971).
Bibliography for vapor-liquid equilibrium data.
Francis, A.W. (1961). Critical Solution Temperature, ACS, Washington, DC, Vol 31.
Collection of critical solution temperatures for various binary mixtures.
Fredenslund, A., Gmehling, J. and Rasmussen, P. (1977). Vapor-Liquid Equlibria using UN 1F AC,
Elsevier, New York.
Detailed and extensive information on the UNIF AC method for estimating activity coefficients
with application to vapor-liquid equilibria at moderate pressures.
Gmehling, J. and Onken, U. (1977). Vapor-Liquid Equilibrium Data Collection, DECHEMA
Chemistry Data Ser., Frankfurt, Vol. 1, pp. 1-10.
Comprehensive data collection for more than 6000 binary and multicomponent mixtures at
moderate pressures. Data correlation and consistency tests are given for each data set.
Hala, E. et al. (1967). Vapour-Liquid Equilibrium, 2nd English ed., trans. George Standard,
Pergamon, Oxford.
Excellent comprehensive survey, including a discussion of experimental methods.
Hala, E., Wichterle, I., Polak, 1. and Boublik, E. (1968). Vapor-liquid Equilibrium Data at Normal
Pressures, Pergamon, Oxford.
Compilation of experimental data for binary mixtures.
Hicks, c.P. (1978). Bibliography of thermodynamic quantities for binary fluid mixtures, in
Chemical Thermodynamics, M.L. McGlashan, Chemical Society, London, Vol. 2, chap. 9.
Literature references for vapor-liquid equilibria, enthalpies of mixing and volume change for
binary systems.
Horsley, lH. (1952-62). Azeotropic Data, Advances in Chemistry Series, No.6 and No. 35,
American Chern. Soc., Washington, DC.
Possibly the most extensive single source compilation of azeotrope data.
Jordan, T.E. (1954). Vapor Pressure of Organic Compounds. Interscience, New York.
Compilation of vapor pressure data for organic compounds; data are correlated with the
Antoine equation and graphs are presented.
Landolt Bornstein (1975). Zahlenwerte und Funktionen aus Naturwissenschaft und Technik,
New Ser. Group IV. Vol. III, Springer Verlag, Berlin. Also part 4c (Solubility of Gases in
Liquids), 1976.
Vapor-liquid equilibrium, data and vapor pressure data. Updates earlier compilation of
vapor-liquid equilibrium data.
Reid, R.C., Prausnitz, J.M. and Poling, B.E. (1986). The Properties of Gas and Liquids, 4th ed.,
McGraw-Hill, New York, NY.
Most recent edition of the famous series started by Reid, Prausnitz, and Sherwood. Source
for estimation techniques for pure component physical properties along with a compilation
of pure component properties for 618 compounds. Excellent starting point.
Riddick, J.A. and Bunger, W.B. (1970). Organic Solvent, 3rd ed., Wiley Interscience, New York,
NY, Vol. II.
Various information on organic solvents.
Seidell, A. and Linke, W.F. (1958-65). Solubilities of Inorganic and Metal-Organic Compounds,
Van Nostrand, Princeton, NJ, Vols. 1 and 2 and supplement.
Presents solubility data for gas-liquid, solid-solid, and liquid-liquid systems.
Stephen, H.T. (1963). Solubilities of Inorganic and Organic Compounds, Pergamon Press, Oxford.
Solubility tables at various temperatures for binary and ternary systems.
Wichterle, I., Lind, 1. and Hala, E. (1973, 1976, 1982). Vapor-Liquid Equilibrium Data
Bibliography, Elsevier, Amsterdam, Vols. 1,2, and 3.
Thorough compilation of literature sources for binary and multicomponent data; includes
 DISTILLATION IN THE PHARMACEUTICAL INDUSTRY 455
many references to the East European literature. Excellent starting point to find if published
information exists for a system.

Equilibrium computation

Discusses the thermodynamic basis for computer calculations for vapor-liquid equlibria;
computer programs are given.
Fredenslund, A. and Heideman, R.A., Vapor-Liquid Equilibria in Complex Mixtures, I. Chern.
E. Symposium Ser. No. 104.
Review of models for the computation of vapor-liquid mixtures which include either
electrolytes, polymers or polar components.
Prausnitz, J.M., Eckert, C.A., O'Connell, J.P. and Orye, R.V. (1967). Computer Calculations for
Multi-component Vapor-Liquid Equilibria, Prentice-Hall, Englewood Cliffs, NJ.
Ray, P. and Hobson, G.V., The Selection and Use ofVLE Methods and Data, I. Chern. Symposium
Series No. 104.
Discussions and review of VLE model for various applications.

Distillation equipment design

Fair, J.R. and Bolles, W.L., Performance and Design of Packed Distillation Columns, I. Chern. E.
Symposium Ser. No. 56.
Performance evaluation of packed column for the following parameter; flood velocity, pressure
gradient and mass transfer for various commercial packings.
Glitch Inc. Bulletins
King, C.J. (1979). Separation Processes, McGraw-Hill, New York.
Discussion with equations for designing and sizing distillation equipment.
Koch Engineering Bulletins
Norton Chemical Process Products Co. Bulletins
Nutter Engineering Bulletins
Perry, R.H. and Chilton, C.H. (5th). Chemical Engineers Handbook, McGraw-Hill, New York.
Discussion with equations for designing and sizing distillation equipment.
Rousseau, R.W. (1987). Handbook of Separation and Process Technology, John Wiley, New York.
Discussion with equations for designing and sizing distillation equipment.
Schweitzer, P.A. (1979). Handbook ofSeparation Techniquesfor Chemical Engineers, McGraw-Hill,
New York.
Discussion with equations for designing and sizing distillation equipment.
Shinskey, G.F. (1984). Distillation Control, 2nd ed. McGraw-Hill, New York.
Detailed discussion of control method for distillation systems.