Chapter 9

Synthesis of Separation Trains

9.0 OBJECTIVES

Most chemical processes are dominated by the need to separate multicomponent chemical mixtures. In general, a number of
separation steps must be employed where each step separates between two components of the feed to that step. During process
design, separation methods must be selected and sequenced for these steps. This chapter discusses some of the techniques for the
synthesis of separation trains. More detailed treatments are given by Douglas (1995), Barnicki and Siirola (1997), and Doherty and
Malone (2001).
After studying this chapter, the reader should:
1. Be familiar with the more widely used industrial separation methods and their basis for separation.
2. Understand the concept of the separation factor and be able to select appropriate separation methods for vapor–,
liquid–,and solid–fluid mixtures.
3. Understand how distillation columns are sequenced and how to apply heuristics to narrow the search for a near-
optimalsequence.
4. Be able to apply algorithmic methods to determine an optimal sequence of distillation-type separations.
5. Be familiar with the difficulties in and techniques for determining feasible sequences when azeotropes can form.
6. Be able to determine feasible separation systems for gas mixtures and solid–fluid systems.

9.1INTRODUCTION recycle streams when conversion of reactants is incomplete.

Almost all chemical processes require the separation of
mixtures of chemical species (components). In Section 6.4,
three flowsheets (Figures 6.9, 6.10, and 6.11) are shown for
processes involving a reactor followed by a separation system.
A more general flowsheet for a process involving one reactor
system is shown in Figure 9.1, where separation systems are
shown before as well as after the reactor section. A feed
separation system may be required to purify the reactor feed(s)
by removing catalyst poisons and inert species, especially if
they are present as a significant percentage of the feed. An
effluent separation system, which follows the reactor system
and is almost always required, recovers unconverted reactants
(in gas, liquid, and/or solid phases) for recycle to the reactor
system and separates and purifies products and byproducts.
Where separations are too difficult, purge streams are used to
prevent buildup of certain species in recycle streams. Processes
that do not involve a reactor system also utilize separation
operations if the feed is a mixture that requires separation.
Frequently, the major investment and operating costs of a
process will be those costs associated with the separation
equipment, rather than with the chemical reactor(s).
Feed Separation System
As shown in Figure 9.1, the combined feed to a reactor section
may consist of one or more feed streams and one or more
When a
234
feed separation system is needed and more than one feed
enters the process, it is usually preferable to provide separate
separation operations for the individual feed streams before
mixing them with each other and with any recycle streams.
Some industrial examples of chemical processes that require
a feed separation system are:
1. Production of polypropylene from a feed of propylene
and propane. Propane, which is not involved in the
propylene polymerization reaction, is removed from
the propylene by distillation.
2. Production of acetaldehyde by the dehydrogenation of
ethanol using a chromium-copper catalyst. If the feed
is a dilute solution of ethanol in water, distillation is
used to concentrate the ethanol to the near-azeotrope
composition (89.4 mol% ethanol at 1 atm) before it
enters the reactor.
3. Production of formaldehyde by air-oxidation of
methanol using a silver catalyst. The entering air is
scrubbed with aqueous sodium hydroxide to remove
any SO2 and CO2, which are catalyst poisons.
4. Production of vinyl chloride by the gas-phase reaction
of HCl and acetylene with a mercuric chloride catalyst.
Small amounts of water are removed from both feed
235 Chapter 9 Synthesis of Separation Trains
gases by adsorption to prevent corrosion of the reactor usual two-phase flash model. The three-phase model
vessel and acetaldehyde formation. considers the possibility that a vapor phase may also be
5. Production of phosgene by the gas-phase reaction of CO present together with two liquid phases.
and chlorine using an activated carbon catalyst. Both feed In the absence of solids, the resulting phases are separated,
often by gravity, in a flash vessel for the V-L case, or in a
Vapor Recycle Vapor Purge

Combined Reactor Product(s)

Fresh Feed(s) Feed Feed Effluent Effluent
Reactor
Separation Mixer Separation
System
System System
Byproduct(s)

Liquid Recycle
Inert Catalyst Liquid Purge
Species Poisons Figure9.1 Generalflowsheetforaprocess
Solids Recycle

with one reactor system.

gases are treated to remove oxygen, which poisons the
catalyst; sulfur compounds, which form sulfur chlorides;
hydrogen, which reacts with both chlorine and phosgene
to form HCl; and water and hydrocarbons, which also
form HCl.
Phase Separation of Reactor Effluent
In Figure 9.1, the reactor effluent may be a heterogeneous
(two or more phases) mixture, but most often is a
homogeneous (single-phase) mixture. When the latter, it is
often advantageous to change the temperature and/or (but
less frequently) the pressure to obtain a partial separation of
the components by forming a heterogeneous mixture of two
or more phases. Following the change in temperature and/or
pressure, phase equilibrium is rapidly attained, resulting in
the following possible phase conditions of the reactor
effluent, where two liquid phases may form (phase splitting)
when both water and hydrocarbons are present in the reactor
effluent:
Possible Phase Conditions of Reactor Effluent
Vapor Liquid
Vapor and liquid Liquid 1 and liquid 2
Vapor, liquid 1 and liquid 2 Liquid and solids
Vapor and solids Liquid 1, liquid 2, and solids
Vapor, liquid and solids Solids
Vapor, liquid 1, liquid 2,
and solids
With the reasonable assumption that the phases in a
heterogeneous mixture are in phase (physical) equilibrium
for a given reactor effluent composition at the temperature
and pressure to which the effluent is brought, process
simulators can readily estimate the amounts and
compositions of the phases in equilibrium by an isothermal
(two-phase)-flash calculation provided that solids are not
present. When the possibility of two liquid phases exists, it is
necessary to employ a three-phase flash model rather than the
decanter for the V-L1-L2 or L1-L2 cases. For the latter two
cases, centrifugal force may be employed if gravity settling is
too slow because of small liquid-density differences or high
liquid viscosities. If solids are present with one or two liquid
phases, it is not possible to separate completely the solids
from the liquid phase(s). Instead, a centrifuge or filter is used
to deliver a wet cake of solids that requires further processing
to recover the liquid and dry the solids.
Several examples of phase-separation equipment are shown
in Figure 9.2. Each exiting phase is either recycled to the
reactor, purged from the system, or, most often, sent to separate
vapor, liquid, or slurry separation systems as shown in Figure
9.3. The effluents from these separation systems are products,
which are sent to storage; byproducts, which also leave the
process; reactor-system recycle streams, which are sent back to
the reactor; or separation-system recycle streams, which are
sent to one of the other separation systems. Purges and
byproducts are either additional valuable products, which are
sent to storage; fuel byproducts, which are sent to a fuel supply
or storage system; and/or waste streams, which are sent to waste
treatment, incineration, or landfill.
Consider the following examples of phase-equilibria
calculations for industrial reactor effluents:
1. Vapor–liquid Case. The reactor effluent for a toluene
hydrodealkylation process, of the type discussed in
Example 6.7, is a gas at 1,150∘F and 520 psia. When
brought to 100∘F at say 500 psia by a series of heat
exchangers, the result is a vapor phase in equilibrium
with a single liquid phase. A two-phase flash calculation
using the SRK equation of state gives the following
results:
Reactor Effluent Phase Equilibrium for a Toluene
Hydrodealkylation Process
Effluent Vapor Liquid
Component (lbmol/hr) (lbmol/hr) (lbmol/hr)
H2 1,292 1,290 2
CH4 1,167 1,149 18
Benzene 280 16 264
Toluene 117 2 115
 9.1 Introduction 236
Biphenyl 3 0 3 without component separation into a reactor recycle stream and
Total 2,859 2,457 402 a vapor purge stream to prevent buildup of CH 4 whereas the
biphenyl can be separated with the toluene and recycled to
As seen, a reasonably good separation is made between the extinction. These two alternatives are shown in Figure 9.4.
Vapor

Liquid 1

Vapor- Liquid 1-
Liquid Liquid 2
Mixture Mixture
Flash Decanter

Liquid 2
Liquid

Vapor
Slurry

Filter or
Vapor-
Centrifuge
Liquid 1-
Liquid 2
Mixture Liquid 1

Flash-
Decanter
Wet
Cake
Liquid 2
Mother
Liquor Figure 9.2 Examples of phase-separation devices.

Liquid 1 Recycle
Purge Product(s)–Byproduct(s)

Vapor Recycle Vapor Separation Recycle

Separation
System Product(s)–
Byproduct(s)
Liquid 1
Vapor
Separation
Liquid 1 System Purge
Reactor
Feed(s) Effluent Phase
Reactor
Separation Separation Recycle
System
System
Liquid 2 Liquid 2 Purge
Solids or Separation
Slurry System
Product(s)–
Solids Recycle Solids–Slurry Byproduct(s)
Separation
System Separation Recycle

Figure9.3 Processflowsheetshowing
Product(s)–Byproduct(s) separateseparationsystemswith
Purge reactor-systemandseparation-system
Liquid 2 Recycle
recycles.
light gases, H2 and CH4, and the three less-volatile aromatic
hydrocarbons. The vapor is sent to a vapor separation system to
recover CH4 as a byproduct and H2 for recycle. The liquid is
sent to a liquid separation system to recover benzene as the
main product, toluene for recycle to the reactor, and biphenyl as
a fuel byproduct. Alternatively, the vapor can be divided
237 Chapter 9 Synthesis of Separation Trains
Methane Byproduct recover a combined methanol and toluene stream for recycle
to the reactor, ethylbenzene as a byproduct, and styrene as
H 2 Recycle Vapor the main product. The water-rich liquid phase (L2) is sent to
Separation
Section another liquid separation section to recover methanol for
recycle to the reactor and water, which is sent to wastewater
Vapor
H 2 Feed treatment to remove small quantities of soluble organic
Reactor
Reactor Effluent Phase
components. It is important to note that a two-phase flash
Section Separation calculation would produce erroneous results. If in doubt,
Toluene perform a three-phase flash calculation rather than a two-
Feed Liquid
phase flash calculation.
Toluene Recycle Liquid
Separation 3. Vapor–solids Case. Phthalic anhydride is manufactured
Section Benzene Product
mainly by the vapor-phase partial oxidation of orthoxylene
with excess air in a shell-and-tube fixed-bed reactor using
Biphenyl Byproduct vanadium pentoxide catalyst packed inside the tubes.
Typically the reactor feed is very dilute in the orthoxylene
H 2 Recycle Purge
with only 1.27 moles of orthoxylene per 100 moles of air.
The main reaction consumes 80% of the orthoxylene to
Vapor
H 2 Feed Reactor
produce phthalic anhydride and water. The remaining 20% of
Reactor Effluent Phase the orthoxylene is completely and unavoidably oxidized to
Section Separation CO2 and water vapor. Typical reactor effluent conditions are
Toluene 660 K and 25 psia. The reactions are exothermic, with heat
Feed Liquid
removal in the reactor by molten salt of the eutectic mixture
Toluene–Biphenyl
Recycle Liquid of sodium and potassium nitrites and nitrates, which
Separation
Section Benzene Product recirculates between the shell side of the reactor and a heat
exchanger that produces steam. The reactor effluent is
Figure 9.4 Alternative flowsheets for hydrodealkylation of toluene cooled, in a heat exchanger to produce steam from boiler
to benzene. feed water, to 180∘C, which is safely above the primary dew
point of 140∘C, corresponding to condensation of liquid
phthalic anhydride. The effluent then passes to one of two
2. Vapor–liquid 1-liquid 2 Case. The reactor effluent in a
parallel desublimation condensers using cooling water,
styrene production process, involving the reaction:
where the effluent is cooled to 70∘C at 20 psia. Under these
methanol + toluene = styrene + hydrogen + water conditions, the phthalic anhydride desublimes on the outside
of the extended-surface tubes of the heat exchanger as a
with a side reaction of the same reactants that produces solid because the temperature is well below its normal
ethylbenzene and water, is a gas at 425∘C and 330 melting point of 131∘C. The desublimation temperature of
kPa. When brought to 38∘C at say 278 kPa by a series 70∘C is safely above the secondary dew point of 36∘C for
of heat exchangers, the result is a vapor phase in the condensation of water. Two dew points can occur
equilibrium with an organic-rich liquid phase and a because water and phthalic anhydride are almost insoluble in
water-rich liquid phase. A three-phase flash calculation each other. The water vapor will not begin to condense until
using the NRTL method for estimating liquid-phase its partial pressure in the vapor reaches its vapor pressure. At
activity coefficients gives the following results: phase-equilibrium conditions of 70∘C and 20 psia (1,034
torr), a two-phase flash calculation on the reactor effluent is
Reactor Effluent Phase Equilibrium for a Styrene Process performed on the reactor effluent. The calculation uses the
Effluent Vapor Liquid 1 Liquid 2 Clausius–Clapeyron vapor-pressure equation of Crooks and
Component (lbmol/hr) (lbmol/hr) (lbmol/hr) (lbmol/hr) Feetham (1946) for solid phthalic anhydride:
H2 352.2 352.2 0.0 0.0
Log10Ps = 12.249 − 4,632∕T
Methanol 107.3 9.9 31.0 66.4
Water 489.1 8.0 0.5 480.6 where vapor pressure, Ps, is in torr and temperature, T, is in
Toluene 107.3 1.7 105.5 0.1 K. This equation is valid for temperatures in the range of
Ethylbenzene 140.7 0.5 140.0 0.2
30∘C to the normal melting point of 131∘C and predicts a
Styrene 352.2 2.0 350.1 0.1
vapor pressure of 0.000517 torr at 25∘C, which is in good
Total 1,548.8 374.3 627.1 547.4
agreement with the often-quoted value of 0.000514 torr.
In this case, the vapor is 94 mol% H2 for which a vapor
separation section may not be needed. The organic-rich Assuming that the solid phase is pure phthalic anhydride, its
liquid phase (L1) is sent to a liquid separation section to partial pressure in the equilibrium vapor phase is equal to its
 9.1 Introduction 238
vapor pressure. The results of the two-phase flash MgSO 4
• 12 H 2O + MgSO 4

50
calculation are: Solution
Solution Solution +
+ + MgSO 4 • H 2O
Reactor Effluent Phase Equilibrium for a Phthalic Anhydride MgSO 4 • 7 H 2O MgSO 4 •

lb Solution
6 H 2O H 2O
Process—Basis: 100 Moles of Reactor Effluent •
MgSO
4
Eutectic + MgSO 4
•

Effluent Vapor Solids 12 H 2O

Component (moles) (moles) (moles) Solution + MgSO •

4/100
4
12 H 2O
77.70 77.70 0.00 2 MgSO 4
• 6 H 2O
N2 •

Concentration, lb MgSO
25 4
O2 15.05 15.05 0.00
Orthoxylene 0.00 0.00 0.00
CO2 2.00 2.00 0.00 MgSO 4
• 12 H 2O
Solution
H2O 4.25 4.25 0.00 Ice + eutectic
Phthalic anhydride 1.00 0.005 0.995 Ice + solution
Total 100.00 99.005 0.995 Ice

At these equilibrium conditions of 70∘C and 1,034 0

torr, the partial pressure of phthalic anhydride in the 20 120 220
Temperature, °F
vapor is (0.005∕99.005)1,034 = 0.05 torr, which is equal
to its vapor pressure. The partial pressure of water in the Figure 9.5 Phase diagram for the MgSO4⋅H2O system.
vapor is (4.25∕99.005)1,034 = 44.4 torr, which is well
below its vapor pressure of 234 torr at 70∘C. Thus, water
does not condense at these conditions. The amount of activity coefficient of water in a sulfate solution for a
vapor phase of H2O, an aqueous phase of dissolved
solids in the table above corresponds to a 99.5%
sulfate using Figure 9.5 to obtain MgSO 4 solubility as a
desublimation of phthalic anhydride. At 85∘C the percent
function of temperature, and a solid phase of hydrated
desublimation is only 98%, while at 96.4∘C it is only
magnesium sulfate crystals, the following phase-
95%. Thus, the recovery of phthalic anhydride from the
equilibrium conditions are calculated:
reactor effluent is sensitive to the desublimation
condenser temperature. Crystallizer Phase Equilibrium for a Magnesium Sulfate Process
While one desublimation condenser is removing Effluent Vapor Liquid Solids
99.5% of the phthalic anhydride from the effluent, Component (lb/hr) (lb/hr) (lb/hr) (lb/hr)
phthalic anhydride in the other condenser is melted with H2O 9,844 581 7,803 0
hot water at 160∘C flowing inside the tubes and sent to a
MgSO4 4,480 0 3,086 0
liquid separation section for the removal of small
MgSO4⋅7H2O 0 0 0 2,854
amounts of any impurities. Thus, the reactor effluent gas
is switched back and forth between the two parallel Total 14,324 581 10,889 2,854
cooling-water condensers. The vapor leaving the The vapor is condensed without further treatment. The
desublimation condenser is sent to a vapor separation magma of combined liquid and solids is sent to a slurry
section. separation system to obtain a product of dry crystals of
4. Vapor–liquid–solids Case. Magnesium sulfate as Epsom MgSO4⋅7H2O.
salts (MgSO4⋅7H2O) is produced by the reaction of solid
Mg(OH)2 with an aqueous solution of sulfuric acid. A
typical reactor effluent is a 10 wt% aqueous solution of Industrial Separation Operations
MgSO4 at 70∘F and 14.7 psia. The effluent is Following phase separation, the individual vapor, liquid,
concentrated to 37.75 wt% MgSO4 in a double-effect solids, and/or slurry streams are sent to individual separation
evaporation system with forward feed, after which filtrate systems, the most common of which is the liquid separation
from a subsequent filtering operation is added. system. When the feed to a vapor or liquid separation system
Crystallization is then carried out in a continuous is a binary mixture, it may be possible to select a separation
adiabatic vacuum flash crystallizer operating at 85.6∘F method that can accomplish the separation task in just one
and 0.577 psia to produce a vapor and a magma (slurry of piece of equipment. In that case, the separation system is
liquid and solid). By making an adiabatic enthalpy relatively simple. More commonly, however, the feed
balance that accounts for the heat of crystallization, heat mixture involves more than two components.
of vaporization, and the
239 Chapter 9 Synthesis of Separation Trains
Figure 9.6 Specification for butenes recovery system.
Propane
99% Recovery

n-Butane
96% Recovery

Feed
37.8°C, 1.03 MPa(10.2 atm)
Separation Process
kmol/hr
Butenes Mixture
Propane 4.5
95% Recovery
1-Butene 45.4
n-Butane 154.7
trans-2-Butene 48.1
cis-2-Butene 36.7
n-Pentane 18.1 n-Pentane
98% Recovery
307.5
Although some progress is being made in devising
multicomponent separation systems involving a single piece of
equipment, most systems involve a number of units in which
the separations are sequenced, with each unit separating its feed
stream into two effluent streams of different composition. The
separation in each piece of equipment (unit) is made between
two components designated as the key components for that
particular separation unit. Each effluent is either a final product
or a feed to another separation device. The synthesis of a
multicomponent separation system can be very complex
because it involves not only the selection of the separation
method(s), but also the manner in which the pieces of
separation equipment are sequenced. This chapter deals with
both aspects of the synthesis problem.
As an example of the complexity of a multicomponent
separation system, consider the synthesis of a separation system
for the recovery of butenes from a C4 concentrate from the
catalytic dehydrogenation of n-butane. The specifications for
the separation process are taken from Hendry and Hughes
(1972) and are shown in Figure 9.6. The process feed, which
contains propane, 1-butene, n-butane, trans-2-butene, cis-2-
butene, and n-pentane, is to be separated into four fractions: (1)
a propane-rich stream containing 99% of the propane in the
feed, (2) an n-butane-rich stream containing 96% of the nC4 in
the feed, (3) a stream containing a mixture of the three butenes
at 95% recovery, and (4) an n-pentane-rich stream containing
98% of the nC5 in the feed. The C3 and nC5 streams are final
products, the nC4 stream is recycled to the catalytic
dehydrogenation reactor, and the butenes stream is sent to
another dehydrogenation reactor to produce butadienes.
Many different types of separation devices and sequences
thereof can accomplish the separations specified in Figure
9.6. In general, the process design engineer seeks the most
economical system. One such system, based on mature
technology and the availability of inexpensive energy, is
shown in Figure 9.7. The system involves two separation
methods, distillation and extractive distillation. The process
feed from the butane dehydrogenation unit is sent to a series
of two distillation columns (1-butene columns, C-1A and C-
1B), where the more volatile propane and 1-butene are
removed as distillate and then separated in a second
distillation column (depropanizer, C-2) into propane and 1-
butene. Distillation unit C-1 consists of two columns because
150 trays are required, which are too many for a single
column (since the tray spacing is typically 2 ft, giving a 300-
ft high tower while most towers do not exceed 200 ft for
structural reasons). The bottoms from unit C-1A, which
consists mainly of n-butane, the 2-butene isomers, and nC5, is
sent to another distillation column (deoiler, C-3) where nC5
product is removed as bottoms. The distillate stream from
unit C-3 cannot be separated into nC4-rich and 2-butenes-rich
streams economically by ordinary distillation because the
relative volatility is only about 1.03. Instead, the process in
Figure 9.7 uses extractive distillation with a solvent of 96%
furfural in water, which enhances the relative volatility to
about 1.17. The separation occurs in columns C-4A and C-4B
with nC4 taken off as distillate. The bottoms is sent to a
furfural stripper (C-5), where the solvent is recovered and
recycled to unit C-4 and the 2-butenes are recovered as
distillate. The 1-butene and 2-butenes streams are mixed and
sent to a butenes dehydrogenation reactor. Although the
process in Figure 9.7 is practical and economical, it does
involve the separation of 1-butene from the 2-butenes.
Perhaps another sequence could avoid this unnecessary
separation.
The separation process of Figure 9.7 utilizes only
distillationtype separation methods. These are usually the
methods of choice for liquid or partially vaporized feeds unless
the relative volatility between the two key components is less
than 1.10 or extreme conditions of temperature and pressure are
required. In those cases or for vapor, solid, or wet solid feeds, a
number of other separation methods should be considered.
These are listed in Table 9.1 in order of technical maturity as
determined by Keller (1987), except for a few added separation
methods not considered by Keller.
 9.1 Introduction 240

Figure 9.7 Process for butenes recovery; C = distillation column; E = heat exchanger.
As noted in Table 9.1, the feed to a separation unit usually
consists of a single vapor, liquid, or solid phase. If the feed is
composed of two or more coexisting phases, consideration
should be given to separating the feed stream into two phases
by some mechanical means of the type shown in Figure 9.2, and
then sending the separated phases to different separation units,
each appropriate for the phase condition of the stream.
The separation of a feed mixture into streams of differing
chemical compositions is achieved by forcing individual
species into different spatial locations. This is accomplished by
any one or a combination of four common industrial
techniques: (1) the creation by heat transfer, shaft work, or
pressure reduction of a second phase; (2) the introduction into
the system of a second fluid phase; (3) the addition of a solid
phase on which selective adsorption can occur; and (4) the
placement of a selective membrane barrier. Unlike the mixing
of chemical species, which is a spontaneous process, the
separation of a mixture of chemicals requires an expenditure of
some form of energy. In the first technique, no other chemicals
are added to the feed mixture and the separation is achieved by
an energy-separating agent (ESA), usually heat transfer, which
241 Chapter 9 Synthesis of Separation Trains
causes the formation of a second phase. The components are
separated by differences in volatility, thus causing each species
to favor one phase over another. In the second technique, a
second phase is added to the separation unit in the form of a
solvent as a mass-separating agent (MSA) that selectively
dissolves or alters the volatility of certain species of the
mixture. An additional separation step is usually required to
recover the solvent for recycle. The third technique involves the
addition of solid particles that selectively adsorb certain species
of the mixture. Subsequently, the particles must be treated by
another separation method to recover the adsorbed species and
regenerate the adsorbent for further use. Thus, the particles act
as
 242
9.2 Criteria for Selection of Separation Methods

Table 9.1 Common Industrial Separation Methods

Separation Phase Condition Developed or Separation
Method of Feed Separating Agent(s) Added Phase Property
Flash L and/or V Pressure reduction or heat transfer ESA V or L Volatility
Distillation (ordinary) L and/or V Heat transfer or shaft work ESA V or L Volatility
Gas absorption V Liquid absorbent MSA L Volatility
Stripping L Vapor stripping agent MSA V Volatility
Extractive distillation L and/or V Liquid solvent and heat transfer MSA L and V Volatility
Azeotropic distillation L and/or V Liquid entrainer and heat transfer MSA L and V Volatility
Liquid–liquid extraction L Liquid solvent MSA Second L Solubility
Crystallization L Heat transfer ESA S Solubility or melting point
Gas adsorption V Solid adsorbent MSA S Adsorbability
Liquid adsorption L Solid adsorbent MSA S Adsorbability
Membrane L or V Membrane ESA Membrane Permeability and/or
solubility
Supercritical extraction L or V Supercritical solvent MSA Supercritical fluid Solubility
Leaching S Liquid solvent MSA L Solubility
Drying S and L Heat transfer ESA V Volatility
Desublimation V Heat transfer ESA S Volatility
an MSA. The fourth technique imposes a barrier that devising such a separation sequence, it is preferable not to
allows the permeation of some species over others. A separate components that must be blended later to form
mechanical energy loss accompanies the permeation. desired multicomponent products. However, many
Thus, this technique involves an ESA. For all four exceptions exist to this rule. For example, in Figure 9.6, a
techniques, mass transfer controls the rate of migration of six-component mixture is separated into four products,
species from one phase to another. Except for the fourth one of which contains 1-butene and cisand trans-2-
technique, the extent of mass transfer is limited by butene. However, the process in Figure 9.7 shows the
thermodynamic equilibrium between the phases. In the separation of 1-butene from the 2-butenes and subsequent
case of membrane separations, the exiting phases do not blending to obtain the desired olefin mixture. The
approach equilibrium; rather, the separation occurs unnecessary separation is carried out because the
strictly because of differences in the rates of permeation volatility of n-butane is intermediate between that of 1-
through the membrane. butene and the two 2-butene isomers. The process shown
in Figure 9.7 is the most economical one known, as
shown later in Example 9.3. In a multicomponent
9.2 CRITERIA FOR SELECTION separation process, each separation operation generally
separates between two components in which case the
OF SEPARATION METHODS minimum number of operations is one less than the
The development of a separation process requires the number of products. However, there are a growing
selection of (1) separation methods, (2) ESAs and/or number of exceptions to this rule, and cases are described
MSAs, (3) separation equipment, (4) the optimal later for which a single separation operation may produce
arrangement or sequencing of the equipment, and (5) the only a partial separation.
optimal operating conditions of temperature and pressure
for the equipment.
When the process feed is a binary mixture and the task Phase Condition of the Feed as a
is to separate that mixture into two products, a single Criterion
separation device may suffice if an ESA is used. If an
When selecting a separation method from Table 9.1, the
MSA is necessary, an additional separation device will be
phase condition of the feed is considered first.
required to recover the MSA for recycle. For a
multicomponent feed that is to be separated into nearly
pure components and/or one or more multicomponent Vapor Feeds
products, more than one separation device is usually If the feed is a vapor or is readily converted to a vapor, the
required. Not only must these devices be selected but also following operations from Table 9.1 should be considered:
an optimal arrangement of the devices must be sought. In
243 Chapter 9 Synthesis of Separation Trains
(1) partial condensation (the opposite of a flash or partial Consequently, the larger the value of SF, the more feasible is
vaporization), the particular separation operation. However, when seeking
(2) distillation under cryogenic conditions, (3) gas a desirable SF value, it is best to avoid extreme conditions
absorption, (4) gas adsorption, (5) gas permeation with a of temperature that may require refrigeration or damage
membrane, and (6) desublimation. heat-sensitive materials; pressures that may require gas
compression or vacuum; and MSA concentrations that may
Liquid Feeds require expensive means to recover the MSA. In general,
operations employing an ESA are economically feasible at a
If the feed is a liquid or is readily converted to a liquid, a lower value of SF than are those employing an MSA. In
number of the operations in Table 9.1 may be applicable: (1) particular, provided that vapor and liquid phases are readily
flash or partial vaporization, (2) (ordinary) distillation, (3) formed, distillation should always be considered first as a
stripping, (4) extractive distillation, (5) azeotropic possible separation operation if the feed is a liquid or
distillation, (6) liquid– liquid extraction, (7) crystallization, partially vaporized.
(8) liquid adsorption, (9) dialysis, reverse osmosis, When a multicomponent mixture forms nearly ideal
ultrafiltration, and pervaporation with a membrane, and (10) liquid and vapor solutions, and the ideal gas law holds, the
supercritical extraction. A flash and the different types of K-values and relative volatility can be readily estimated
distillation are also applicable for feeds consisting of from vapor pressure data. Such K-values are referred to as
combined liquid and vapor phases. ideal or Raoult’s law K-values. Then, the SF for vapor–
liquid separation operations employing an ESA (partial
Slurries, Wet Cakes, and Dry Solids evaporation, partial condensation, or distillation) is given by
Slurry feeds are generally separated first by filtration or
centrifugation to obtain a wet cake, which is then separated SF = �1,2 = PPs1s2 (9.3)
into a vapor and a dry solid by drying. Feeds consisting of where Psi is the vapor pressure of component i. When the
dry solids can be leached with a selective solvent to separate components form moderately nonideal liquid solutions
the components. (hydrocarbon mixtures or homologous series of other
organic compounds) and/or pressures are elevated, an
equation-of-state, such as Soave–Redlich–Kwong (SRK)
Separation Factor as a Criterion or Peng–Robinson (PR), may be necessary for the
The second consideration for the selection of a separation estimation of the separation factor, using
method is the separation factor, SF, which can be achieved
by the particular separation method for the separation
between two key components of the feed. This factor for the SF (9.4)
separation of key component 1 from key component 2 where �i is the mixture fugacity coefficient of component i.
between phases I and II for a single stage of contacting is For vapor–liquid separation operations (e.g., azeotropic
defined as and extractive distillation) that use an MSA that causes
the formation of a nonideal liquid solution but operate at
C1 C2
SF = I ∕ I (9.1)near ambient pressure, expressions for the K-values of the
key components are based on a modified Raoult’s law that
C incorporates liquid-phase activity coefficients. Thus, the
separation factor is given by
where Cij is a composition (expressed as a mole fraction,
mass fraction, or concentration) of component j in phase i. If
phase I is to be rich in component 1 and phase II is to be rich
SF = �1,2 = ��12LLPPs1s2 (9.5)
in component 2, then SF must be large. The value of SF is
limited by thermodynamicequilibriumexcept
formembraneseparationsthatare controlled by relative rates where �i is the activity coefficient of component i, which
of mass transfer through the membrane. For example, in the is estimated from the Wilson, NRTL, UNIQUAC, or
case of distillation, using mole fractions as the composition UNIFAC equations and is a strong function of mixture
variable and letting phase I be the vapor and phase II be the composition.
liquid, the limiting value of SF is given in terms of vapor If an MSA is used to create two liquid phases, such as
and liquid equilibrium ratios (K-values) by in liquid–liquid extraction, the SF is referred to as the
relative selectivity, �:
SF �1,2 (9.2) �II∕�II
SF(9.6)
where � is the relative volatility. In general, components 1 1 2
and 2 are designated in such a manner that SF > 1.0.

where phase II is usually the MSA-rich phase and
component 1 is more selective for the MSA-rich phase than
is component 2.
In general, MSAs for extractive distillation and liquid–
liquid extraction are selected according to their ease of
recovery for recycle and to achieve relatively large values
of SF. Such MSAs are often polar organic compounds
(e.g., furfural) used in the example earlier to separate n-
butane from 2-butenes. In some cases, the MSA is
selected in such a way that it forms one or more
homogeneous or heterogeneous azeotropes with the
components in the feed. For example, the addition of n-
butyl acetate to a mixture of acetic acid and water results
in a heterogeneous minimum-boiling azeotrope of the
acetate with water. The azeotrope is taken overhead, the
acetate and water layers are separated, and the acetate is
recirculated.
Although the degree of separation that can be achieved
for a given value of SF is almost always far below that
required to attain necessary product purities, the
application of efficient countercurrent-flow cascades of
many contacting stages as in distillation operations can
frequently achieve sharp separations. For example,
consider a mixture of 60 mol% propylene and 40 mol%
propane. It is desired to separate this mixture into two
products at 290 psia, one containing 99 mol% propylene
and the other 95 mol% propane. By material balance, the
former product would constitute 58.5 mol% of the feed
and the latter 41.5 mol%. From equilibrium
thermodynamics, the relative volatility for this mixture is
approximately 1.12. A single equilibrium vaporization at
290 psia to produce 58.5 mol% vapor results in products
that are far short of the desired compositions: a vapor
containing just
61.12 mol% propylene and a liquid containing just 51.36
mol% propane at 51.4∘C. However, with a countercurrent
cascade of such stages in a simple (single-feed, two-product)
distillation column with reflux and boilup, the desired
products can be achieved with 200 stages and a reflux ratio
of 15.9.
Single-stage operations (e.g., partial vaporization or
partial condensation with the use of an ESA) are utilized
only if SF between the two key components is very large
or if a rough or partial separation is needed. For example,
if SF = 10,000, a mixture containing equimolar parts of
components 1 and 2 could be partially vaporized to give a
vapor containing 99 mol% of component 1 and a liquid
containing 99 mol% of component 2. At low values of SF,
lower than 1.10 but greater than 1.05, ordinary distillation
may still be the most economical choice. However, an
MSA may be able to enhance the value of SF for an
alternative separation method to the degree that the
method becomes more economical than ordinary
distillation. As illustrated in Figure 9.8 from Souders
(1964), extractive distillation or liquid–liquid extraction
may be preferred if the SF can be suitably enhanced. If SF
244
= 2 for ordinary distillation, it must be above 3.3 for
extractive distillation to be an acceptable alternative and
above 18 for liquid–liquid extraction.
Unless values of SF are about 10 or above, absorption
and stripping operations cannot achieve sharp separation
between two components. Nevertheless, these operations
are used widely for preliminary or partial separations
where the separation of one key component is sharp, but
only a partial separation of the other key component is
adequate. The degree of sharpness of separation is given
by the recovery factor RF,
n
RF = n FiI
i
(9.7)
where n is moles or mass, I is the product rich in i, and F
is the feed.
The separation of a solid mixture may be necessary
when one or more (but not all) of the components are (is)
not readily melted, sublimed, or vaporized. Such
operations may even be preferred when boiling points are
close but melting points are far apart, as
Minimumα Required for Consideration
Minimumβ Required for Consideration
100
of Liquid–Liquid Extraction
of Extractive Distillation
Min β Liquid–Liquid Extraction
or
10
Min α
Extractive Distillation
1
1.0 2.0 3.0
α for Ordinary Distillation
Figure 9.8 Relative selectivities for equal-cost separators.
Adapted from Souders, 1964. Reproduced with permission of
the American Institute of Chemical Engineers. Copyright ©
1964 AIChE. All rights reserved.
9.2 Criteria for Selection of Separation Methods
is the case with many isomeric pairs. The classic example is
the separation of metaxylene from paraxylene whose normal
boiling points differ only by 0.8∘C but whose melting
points differ by 64∘C. With an SF of only 1.02, as
determined from Eq. (9.2), ordinary distillation to produce
relatively pure products from an equimolar mixture of the
two isomers would require about 1,000 stages and a reflux
ratio of more than 100. For the separation by crystallization,
the SF is nearly infinity because essentially pure paraxylene
is crystallized. However, the mother liquor contains at least
13 mol% paraxylene in metaxylene, corresponding to the
245 Chapter 9 Synthesis of Separation Trains
limiting eutectic composition. When carefully carried out,
crystallization can achieve products of very high purity.
The separation factor for adsorption depends on either
differences in the rate of adsorption or adsorption
equilibrium, with the latter being more common in industrial
applications. For equilibrium adsorption, Eq. (9.1) applies
where the concentrations are those at equilibrium on the
adsorbed layer within the pores of the adsorbent and in the
bulk fluid external to the adsorbent particles. High
selectivity for adsorbents is achieved either by sieving, as
with molecular-sieve zeolites or carbon, or by large
differences in adsorbability. For example, in the case of
molecular-sieve zeolites, aperture sizes of 3, 4, 5, 8, and 10
Å are available. Thus, nitrogen molecules, with a kinetic
diameter of about 3.6 Å, can be separated from ammonia
with a kinetic diameter of about 2.6 Å, using a zeolite with
an aperture of 3 Å. Only the ammonia is adsorbed.
Adsorbents of silica gel and activated alumina having wide
distributions of pore diameters in the range of 20 to 100 Å
are highly selective for water, whereas activated carbon with
pore diameters in the same range is highly selective for
organic compounds. When adsorption is conducted in fixed
beds, essentially complete removal from the feed of those
components with high selectivity can be achieved until
breakthrough occurs. Before breakthrough, regeneration or
removal of the adsorbent is required.
If only a small amount of one component is present in a
mixture,changingthephaseofthecomponents
inhighconcentrations should be avoided. In such a case,
absorption, stripping, or selective adsorption best removes
the minor component. Adsorption is particularly effective
because of the high selectivity of adsorbents and is widely
used for purification, where small amounts of a solute are
removed from a liquid or vapor feed.
For membrane separation operations, SF may still be
defined by Eq. (9.1). However, SF is governed by relative
rates of mass transfer in terms of permeabilities rather than
by equilibrium considerations. For the ideal case where the
downstream concentration is negligible compared to the
upstream concentration, the separation factor reduces to:
PM1 (9.8)of constituents for recycle, a high degree of separation from
SF =
PM2
where PMi is the permeability of species i. In most cases, the
value of SF must be established experimentally. In general,
membrane separation operations should be considered
whenever adsorption methods are considered. Membranes
are either porous or nonporous. If porous, the permeability is
proportional to the diffusivity through the pore. If
nonporous, the permeability is the product of the solubility
of the molecule in the membrane and its diffusivity for
travel through the membrane. An example of the use of
membranes is gas permeation with nonporous hollow fibers
to separate hydrogen, helium, carbon dioxide, and/or water
vapor from gases containing oxygen, nitrogen, carbon
monoxide, and/or light hydrocarbons. For a typical
membrane, the SF between hydrogen and methane is 6.
Because it is difficult to achieve large numbers of stages
with membranes, an SF of this magnitude is not sufficient to
achieve a sharp separation, but is widely used to make a
partial separation. Sharp separations can be achieved by
sieving when the kinetic molecular diameters of the
components to be separated differ widely and when
membrane pore diameter lies between those kinetic
diameters.
Supercritical extraction utilizes the solvent power of a gas
at near-critical conditions. It is the preferred method for the
removal of undesirable ingredients from foodstuffs with
carbon dioxide. The separation factor, which is given by Eq.
(9.1), is difficult to estimate from equations of state using
Eq. (9.4) and is best determined by experiment. Equation
(9.1) also applies for leaching (solid–liquid extraction), often
using a highly selective solvent. As with supercritical
extraction, the value of SF is best determined by experiment.
Because mass transfer in a solid is very slow, it is important
to preprocess the solid to drastically decrease the distance
for diffusion. Typical methods involve making thin slices of
the solid or pulverizing it. Desublimation is best applied
when a sublimable component is to be removed from
noncondensable components of a gas stream, corresponding
to a very large separation factor.
Reason for the Separation as a
Criterion
A final consideration in the selection of a separation method
is the reason for the separation. Possible reasons are (1)
purification of a species or group of species, (2) removal of
undesirable constituents, and (3) recovery of constituents for
subsequent processing or removal. In the case of
purification, the use of an MSA method may avoid exposure
with an ESA method to high temperatures that may cause
decomposition. In some cases, removal of undesirable
species together with a modest amount of desirable species
may be economically acceptable. Likewise, in the recovery
the product(s) may not be necessary.
9.3SELECTION OF EQUIPMENT
Only a very brief discussion of equipment for separation
operations is presented here. Much more extensive
presentations, including drawings and comparisons, are
given in Perry’s Chemical Engineers’ Handbook (Green and
Perry, 2008) and by Kister (1992), Walas (1988), and Seader
et al. (2016). In general, equipment selection is based on
stage or mass-transfer efficiency, pilot-plant tests, scale-up
feasibility, investment and operating cost, and ease of
maintenance.

Absorption, Stripping, and Distillation
For absorption, stripping, and all types of distillation (i.e.,
vapor–liquid separation operations), either trayed or packed
columns are used. Trayed columns are usually preferred for
initial installations, particularly for columns 3 ft or more in
diameter. However, packed columns should be given serious
consideration for operation under vacuum or where a low
pressure drop is desired. Other applications favoring packed
columns are corrosive systems, foaming systems, and cases
where low liquid holdup is desired. Packing is also generally
specified for revamps. Applications favoring trayed columns
are feeds containing solids, high liquid-to-gas ratios, large-
diameter columns, and where operation over a wide range of
conditions is necessary. The three most commonly used tray
types are sieve, valve, and bubble cap. However, because of
high cost, the latter is specified only when a large liquid
holdup is required on the tray, for example, when
conducting a chemical reaction simultaneously with
distillation. Sieve trays are the least expensive and have the
lowest pressure drop per tray, but they have the narrowest
operating range (turndown ratio). Therefore, when flexibility
is required, valve trays are a better choice. Many different
types of packings are available. They are classified as
random or structured. The latter are considerably more
expensive than the former, but have the lowest pressure
drop, the highest efficiency, and the highest capacity
compared to both random packings and trays. For that
reason, structured packings are often considered for column
revamps.
Liquid–Liquid Extraction
For liquid–liquid extraction, an even greater variety of
equipment is available, including multiple mixer-settler
246
units or single countercurrent-flow columns with or
without mechanical agitation. Very compact, but
expensive, centrifugal extractors are also available. When
the equivalent of only a few theoretical stages is required,
mixer-settler units may be the best choice because
efficiencies approaching 100% are achievable in each unit.
For a large number of stages, columns with mechanical
agitation may be favored. Packed and perforated tray
columns can be very inefficient and are not recommended
for critical separations.
Membrane Separation
Most commercial membrane separations use natural or
synthetic glassy or rubbery polymers. To achieve high
permeability and selectivity, nonporous materials are
preferred, with thicknesses ranging from 0.1 to 1.0
micron, either as a surface layer or film onto or as part of
much thicker asymmetric or composite membrane
materials, which are fabricated primarily into spiral-
wound and hollow-fiber-type modules to achieve a high
ratio of membrane surface area to module volume.
Adsorption
For commercial applications, an adsorbent must be chosen
carefully to give the required selectivity, capacity,
stability, strength, and regenerability. The most commonly
used adsorbents are activated carbon, molecular-sieve
carbon, molecular-sieve zeolites, silica gel, and activated
alumina. Of particular importance in the selection process
is the adsorption isotherm for competing solutes when
using a particular adsorbent. Most adsorption operations
are conducted in a semicontinuous cyclic mode that
247 Chapter 9 Synthesis of Separation Trains
includes a regeneration step. Batch slurry systems are favored
for small-scale separations whereas fixed-bed operations are
preferred for large-scale separations. Quite elaborate cycles
have been developed for the latter.
Leaching
Equipment for leaching operations is designed for either
batchwise or continuous processing. For rapid leaching, it is
best to reduce the size of the solids by grinding or slicing. The
solids are contacted by the solvent using either percolation or
immersion. A number of different patented devices are
available.
Crystallization
Crystallization operations include the crystallization of an
inorganic compound from an aqueous solution (solution
crystallization) and the crystallization of an organic compound
from a mixture of organic chemicals (melt crystallization). On
a large scale, solution crystallization is frequently conducted
continuously in a vacuum evaporating draft-tube baffled
crystallizer to produce crystalline particles whereas the
falling-film crystallizer is used for melt crystallization to
produce a dense layer of crystals.
Drying
A number of factors influence the selection of a dryer from the
many different types available. These factors are dominated by
the nature of the feed, whether it be granular solids, a paste, a
slab, a film, a slurry, or a liquid. Other factors include the need
for agitation, the type of heat source (convection, radiation,
conduction, or microwave heating), and the degree to which
the material must be dried. The most commonly employed
continuous dryers include tunnel, belt, band, turbo-tray, rotary,
steam-tube rotary, screw-conveyor, fluidized-bed, spouted-
bed, pneumatic-conveyor, spray, and drum dryers.
9.4 SEQUENCING OF ORDINARY
DISTILLATION COLUMNS FOR
THE SEPARATION OF NEARLY
IDEAL LIQUID MIXTURES
Multicomponent mixtures are often separated into more than
two products. Although one piece of equipment of complex
design might be devised to produce all the desired products, a
sequence of two-product separators is more common.
For nearly ideal feeds such as hydrocarbon mixtures and
mixtures of a homologous series, for example, alcohols, the
most economical sequence will often include only ordinary
distillation columns provided that the following conditions
hold:
1. The relative volatility between the two selected key
components for the separation in each column is greater
than 1.05.
2. The reboiler duty is not excessive. An example of an
excessive duty occurs in the distillation of a mixture with a
low relative volatility between the two key components
where the light key component is water, which has a very
high heat of vaporization.
3. The tower pressure does not cause the mixture to approach
its critical temperature.
4. The overhead vapor can be at least partially condensed at
the column pressure to provide reflux without excessive
refrigeration requirements.
5. The bottoms temperature at the tower pressure is not so
high that chemical decomposition occurs.
6. Azeotropes do not prevent the desired separation.
7. Column pressure drop is tolerable, particularly if operation
is under vacuum.
Column Pressure and Type of Condenser
During the development of distillation sequences, it is necessary
to make at least preliminary estimates of column-operating
pressures and condenser types (total or partial). The estimates are
facilitated by the use of the algorithm in Figure 9.9, which is
conservative. Assume that cooling water is available at 90∘F,
sufficient to cool and condense a vapor to 120∘F. The bubble-
point pressure is calculated at 120∘F for an estimated distillate
composition. If the computed pressure is less than 215 psia, use a
total condenser unless a vapor distillate is required, in which case
use a partial condenser. If the pressure is less than 30 psia, set the
condenser pressure to 30 psia and avoid near-vacuum operation.
If the distillate bubble-point pressure is greater than 215 psia, but
less than 365 psia, use a partial condenser. If it is greater than
365 psia, determine the dew-point pressure for the distillate as a
vapor. If the pressure is greater than 365 psia, operate the
condenser at 415 psia with a suitable refrigerant in place of
cooling water. For the selected condenser pressure, add 10 psia
to estimate the bottoms pressure (for nonvacuum duty,
acceptable column pressure drops are in the range 0.05–0.15
psi/tray), and compute the bubble-point temperature for an
estimated bottoms composition. If that temperature exceeds the
decomposition or critical temperature of the bottoms, reduce the
condenser pressure appropriately.
Number of Sequences of Ordinary
Distillation Columns
Initial consideration is usually given to a sequence of ordinary
distillation columns, where a single feed is sent to each column
and the products from each column number just two, the
distillate and the bottoms. For example, consider a mixture of
benzene, toluene, and biphenyl. Because the normal boiling
 9.4 Sequencing of Ordinary Distillation Columns for the Separation of Nearly Ideal Liquid Mixtures 248

points of the three components (80.1, 110.8, and 254.9∘C,
respectively) are widely separated, the mixture can be
conveniently separated into three nearly pure components by
ordinary distillation. A common process for separating this
mixture is the sequence of two ordinary distillation columns
shown in Figure 9.10a. In the first column, the most volatile
component, benzene, is taken overhead as a distillate final
product. The bottoms is a mixture of toluene and biphenyl, which
is sent to the second column for separation into the two other
final products: a distillate of toluene and a bottoms of biphenyl,
the least volatile component.
Even if a sequence of ordinary distillation columns is used,
Start
Distillate and bottoms
compositions are known
or estimated
Note in Figure 9.10 that it takes a sequence of two ordinary
distillationcolumns toseparate amixtureintothreeproducts.
Furthermore, other sequences can produce the same final
products. For example, the separation of benzene, toluene, and
biphenyl shown in Figure 9.10a can also be achieved by
removing biphenyl as bottoms in the first column, followed by
the separation of benzene and toluene in the second column.
However, the separation of toluene from benzene and biphenyl
by ordinary distillation in the first column is impossible because
toluene is intermediate in volatility. Thus, the number of possible
sequences is limited to two for this case of the separation of a
ternary mixture into three nearly pure products.

Calculate bubble-point P D < 215 psia

pressure (P D ) of (1.48 MPa )
distillate at
120°F (49 °C) Use total condenser
(reset P D to 30 psia
if P D < 30 psia)
P D > 215 psia
T B < bottoms
Calculate dew-point P D < 365 psia Estimate Calculate bubble-point decomposition or critical
pressure (P D ) of (2.52 MPa ) bottoms temperature (T B ) temperature
distillate at Use partial pressure of bottoms
120°F (49 °C) condenser (P B ) at P B

P D > 365 psia T B > bottoms

decomposition or critical
temperature
Choose a refrigerant
so as to operate
partial condenser Lower pressure
at 415 psia P D appropriately
(2.86 MPa)

Figure 9.9 Algorithm for establishing distillation column pressure and condenser type.

p-Xylene
Benzene Toluene m-Xylene
Ethylbenzene

Feed Feed

Figure 9.10 Distillation configurations for

Biphenyl
not all columns need give nearly pure products. For example,
Figure 9.10b shows a distillation sequence for the separation of a
mixture of ethylbenzene, p-xylene, m-xylene, and o-xylene into
only three products: nearly pure ethylbenzene, a mixture of p-
and m-xylene, and nearly pure o-xylene. The para and meta
isomers are not separated because the normal boiling points of
these two compounds differ by only 0.8∘C, making separation
by distillation impractical.
separation of ternary mixtures: (a)
o-Xylene
separation
of a benzene-toluene-biphenyl mixture;
(a) (b) (b)
separation of xylene isomers.
Now consider the more general case of the synthesis of all
possible ordinary distillation sequences for a multicomponent
feed that is to be separated into P final products that are nearly
pure components and/or multicomponent mixtures. The
components in the feed are ordered by volatility with the first
component being the most volatile. This order is almost always
consistent with that for normal boiling point if the mixture forms
nearly ideal liquid solutions, such that Eq. (9.3) applies. Assume
249 Chapter 9 Synthesis of Separation Trains

that the order of volatility of the components does not change as as follows: A–B, B–C, C–D, and D–E. Now let j be the
the sequence proceeds. Furthermore, assume that any number of final products that must be developed from the
multicomponent products contain only components that are distillate of the first column. For example, if the separation
adjacent in volatility. For example, suppose that the previously point in the first column is C–D, then j = 3 (A, B, C). Then P −
cited mixture of benzene, toluene, and biphenyl is to be separated j equals the number of final products that must be developed
into toluene and a multicomponent product of benzene and from the bottoms of the first column. If Ni is the number of
biphenyl. With ordinary distillation, it would be necessary first to different sequences for i final products, then for a given
produce products of benzene, toluene, and biphenyl and then separation point in the first column, the number of sequences
blend the benzene and biphenyl. is NjNP−j. But in the first separator, P − 1 different separation
An equation for the number of different sequences of points are possible. Thus, the number of different sequences
ordinary distillation columns, Ns, to produce a number of for P products is the following sum:
products, P, can be developed in the following manner. For the
P−1
first separator in the sequence, P − 1 separation points are
possible. For example, if the desired products are A, B, C, D,
and E in order of decreasing volatility, then the possible Ns = ∑j=1 NjNP−j = [P2!((PP−−11)]!)! (9.9)
separation points are 5 − 1 = 4,
Table 9.2 Number of Possible Sequences for Separation by Ordinary
Distillation Application of Eq. (9.9) gives results shown in Table 9.2 for
Number of Number of Separators Number of sequences producing up to 10 products. As shown, the number of
Products, P in the Sequence Different sequences grows rapidly as the number of final products
N
Sequences, s increases.
2 1 1 Equation (9.9) gives five possible sequences of three columns
for a four-component feed. These sequences are shown in Figure
3 2 2
4 3 5
9.11. The first, where all final products but one are distillates, is
5 4 14
often referred to as the direct sequence. It is widely used in
6 5 42 industry because distillate final products are more free of
7 6 132 impurities such as objectionable high-boiling compounds and
8 7 429 solids. If the purity of the final bottoms product (D) is critical, it
9 8 1,430 may
D C be
A 10 B 9C A 4,862 B

A B C A B B
B C D B C C
C D C D
D D
D
(direct sequence) (a)
(b)
A C

A A C
B B D
C
D
B D
( c)

C B
A B A

A A B A A A
B B C B B B
C C C C
D D
D D C
(indirect sequence) Figure 9.11 The five sequences for a
(d) (e) four-component feed.
 9.4 Sequencing of Ordinary Distillation Columns for the Separation of Nearly Ideal Liquid Mixtures 250

produced as a distillate in an additional column called a rerun (or

finishing) column. If all products except one are bottoms
products, the sequence is referred to as the indirect sequence.
This sequence is generally considered to be the least desirable
sequence because of difficulties in achieving purity
specifications for bottoms products. The other three sequences in
Figure 9.11 produce two products as distillates and two products
as bottoms. In all sequences except one, at least one final product
is produced in each column.
251 Chapter 9 Synthesis of Separation Trains
EXAMPLE 9.1
Ordinary distillation is to be used to separate the ordered mixture C 2,
C=3 , C3, 1-C=4 , nC4 into the three products C 2; (C=3 , 1-C=4 ); (C3, nC4).
Determine the number of possible sequences.
SOLUTION
Neither multicomponent product contains adjacent components in the
ordered list. Therefore, the mixture must be completely separated with
subsequent blending to produce the (C=3 , 1-C=4 ) and (C3, nC4) products.
Thus, from Table 9.2 with P taken as 5, Ns = 14.
Heuristics for Determining Favorable
Sequences
When the number of products is three or four, designing and
costing all possible sequences can best determine the most
economical sequence. Often, however, unless the feed mixture
has a wide distribution of component concentrations or a wide
variation of relative volatilities for the possible separation points,
the costs will not vary much and the sequence selection may be
based on operation factors. In that case, the direct sequence is
often the choice. Otherwise, a number of heuristics that have
appeared in the literature, starting in 1947, have proved useful
for reducing the number of sequences for detailed examination.
The most useful of these heuristics are:
1. Remove thermally unstable, corrosive, or chemically
reactive components early in the sequence.
2. Remove final products one-by-one as distillates (the direct
sequence).
3. Sequence separation points to remove, early in the
sequence, those components of greatest molar percentage
in the feed.
4. Sequence separation points in the order of decreasing
relative volatility so that the most difficult splits are made
in the absence of the other components.
5. Sequence separation points to leave last those separations
that give the highest purity products.
6. Sequence separation points that favor near equimolar
amounts of distillate and bottoms in each column.
None of these heuristics require column design and costing.
Unfortunately, however, these heuristics often conflict with
each other. Thus, more than one sequence will be developed,
and cost and other factors will need to be considered to
develop an optimal final design. When energy costs are
relatively high, the sixth heuristic often leads to the most
economical sequence. Heuristics 2–6 are consistent with
observations about the effect of the nonkey components on the
separation of two key components. These nonkey components
can increase the reflux and boilup requirements, which, in
turn, increase column diameter and reboiler operating cost.
These and the number of trays are the major factors affecting
the investment and operating costs of a distillation operation.
 9.4 Sequencing of Ordinary Distillation Columns for the Separation of Nearly Ideal Liquid Mixtures 252

EXAMPLE 9.2

Consider the separation problem shown in Figure 9.12a except that

separate isopentane and n-pentane products are also to be obtained
with 98% recoveries. Use heuristics to determine a good sequence of EXAMPLE 9.3
ordinary distillation units.
Consider the separation problem studied by Hendry and Hughes
(1972) as shown in Figure 9.6. Determine the feasibility of two-
SOLUTION product ordinary distillation (method I), select an alternative
Approximate relative volatilities for all adjacent pairs are: separation technique (method II) if necessary, and forbid impractical
splits. Determine the feasible separations and the number of possible
Component Pair Approximate α at 1 separation sequences that incorporate only these separations.
atm
C3/iC4 3.6
iC4/nC4 1.5 SOLUTION
nC4/iC5 2.8 Data for the six species are as follows:
iC5/nC5 1.35 Normal Critical
For this example, there are wide variations in both relative volatility Boiling Critical Pressure
and molar percentages in the process feed. The choice is Heuristic 4, (MPa)
Species Point (∘C) Temperature (∘C)
which dominates over Heuristic 3 and leads to the sequence shown
in Figure 9.12b, where the first split is between the pair with the Propane A −42.1 96.7 4.17
highest relative volatility. This sequence also corresponds to the 1-Butene B −6.3 146.4 3.94
optimal arrangement. n-Butane C −0.5 152.0 3.73
trans-2-Butene D 0.9 155.4 4.12
cis-2-Butene E 3.7 161.4 4.02
n-Pentane F 36.1 196.3 3.31
Figure 9.12 Synthesis problem and separation train for Example 9.2: (a)
paraffin separation problem; (b) sequence developed from heuristics. Because both trans- and cis-2-butenes are contained in the
butenes product and are adjacent when species are ordered by
relative volatility, they need not be separated. All ordinary
Sequencing of General Vapor–Liquid distillation columns can be operated above atmospheric pressure and
Separation Processes with cooling-water condensers. Approximate relative volatilities
assuming ideal solutions at 150∘F (65.6∘C) are as follows for all
When ordinary distillation is not practical for each separator in a adjacent binary pairs except trans-2-butene and cis-2-butene, which
sequence for separating a multicomponent mixture, other types of are not split.
separators must be employed and the order of volatility or other Approximate Relative
separation index may be different for each type. Of great Volatility at 150∘F
importance when other types of separators are used is that the Adjacent Binary Pair (65.5∘C)
number of possible sequences is greatly increased over the value Propane /1-Butene (A/B) 2.45
calculated from Eq. (9.9). Usually the number of different types of
1-Butene /n-Butane (B/C) 1.18
separators to be considered is limited by applicability. If they are all
n-Butane /trans-2-Butene (C/D) 1.03
two-product separators and if T equals the number of different
types, then the number of possible sequences is given by cis-2-Butene /n-Pentane (E/F) 2.50
Because of their high relative volatilities, splits A/B and E/F,
NTs = TP−1Ns (9.10) even in the presence of the other hydrocarbons, should be by
ordinary distillation only. Split C/D is considered infeasible by
ordinary distillation; split B/C is feasible, but an alternative method
When a separator type utilizes a recycled MSA, as, for example, might be more attractive.
with extractive distillation, homogeneous azeotropic distillation, From Hendry and Hughes (1972), the use of approximately 96 wt
liquid–liquid extraction, absorption, stripping, and so on, Eq. % aqueous furfural as a solvent for extractive distillation increases
(9.10) may be applied if the MSA is recovered for recycle in the the volatility of paraffins relative to olefins, causing a reversal in
separator following the separator into which it is introduced and volatility between 1-Butene and n-Butane and giving the separation
if these two separators are counted together as one unit. For order (ACBDEF). Thus, the three olefins, which are specified as one
example, if P = 3 and ordinary distillation, extractive distillation product, are grouped together. They give an approximate relative
volatility of 1.17 for split (C/B)CH4. In the presence of A, this would
with solvent I, extractive distillation with solvent II, and liquid–
add the additional split (A/C), which by ordinary distillation has a
liquid extraction with solvent III are to be considered, then T = 4,
very desirable relative volatility of 2.89. Also, split (…C/D…)II, with
a relative volatility of 1.70, is more attractive than split (…C/D…)I,
according to Figure 9.8.
 Propane iC 4
98% Recovery

C3
Isobutane iC 4
253 Chapter 9 Feed,
Synthesis of Separation Trains 98% Recovery nC 4
37.8°C, 1.72 MPa
nC 4
kmol/hr Separation
and the application of (CEqs. (9.9)45.4and Processgives n32
(9.10)
- Butane
possible C
sequences.3 This is an4 enormous increase iCover
iC 5
the two sequences
Propane 3) 98% Recovery
iC 4 nC 4
resulting in considering
Isobutane only
( iC 4) ordinary
136.1 distillation. nC 4 iC 5
n-Butane (nC 4) 226.8
Equations (9.9) and (9.10)
i-Pentane (iC 5) can be applied to evenPentanes
181.4 more complex problems,
iC 5 such nCas5 those ofiCFigure 9.1, where the feed contains
5
six components but only (nC
n - Pentane four5) products
317.5 are specified.
98If
% two separation
Recovery methods
nC 5 (i.e., ordinary distillation
nC 5 and extractive distillation
907.2
with aqueous furfural) are considered and only the four desired products are produced without blending, then nC 5 T = 2, P = 4, and Eqs.
(9.9) and (9.10) give NTs = 40. If five products,
( a) including the 2-butene isomers, are produced,
(b) with the two butene streams being
blended at the end of the sequence, NTs = 224. If six products are produced with the three butene streams being blended at the end
of the sequence, then NTs = 1,344. Therefore, the total number of sequences possible, if all three product numbers are considered, is
40 + 224 + 1,344 = 1,608.
Because of the enormous increase in the number of possible sequences when separation methods other than ordinary distillation
are considered, it may be worthwhile to forbid certain separations by certain separator types. In this manner, the number of
possible sequences can be greatly reduced. For example, for the separation process shown in Figure 9.6, cis- and trans-2-butene
are adjacent to each other in volatility order for ordinary distillation and for extractive distillation with aqueous furfural. Therefore,
sequences that include their separation are not necessary. This reduces the total number of possible sequences to 264. Further
reduction can be made by excluding other separations. For example, the relative volatility between cis-2-butene and n-pentane for
ordinary distillation is approximately 2.5; accordingly, by using Figure 9.8, extractive distillation for this separation probably need
not be considered.
9.4 Sequencing of Ordinary Distillation Columns for the Separation of Nearly Ideal Liquid Mixtures 254
255 Chapter 9 Synthesis of Separation Trains

The decimal numbers in Figure 9.13 are annualized costs for the modules on separation associated with this book). This is the FUG
separations in thousands of dollars per year. Annualized C costs,
A, for method for which all process simulators have a model (DSTWU
ordinary distillation operations assuming splits of the two key compo-in ASPEN PLUS, Shortcut Column in UniSim®Design/ASPEN
nents (e.g., 99 mol % recovery of the light key component in the dis- HYSYS).
%
tillate and 99 mol recovery of the heavy key component in the 3. Select a tray spacing (typically 2 ft) and compute the height of the
bottoms product) are estimated using the following steps for columnstower,H.
with trays:
U , using the Fair correlation. Set
4. Estimate the flooding velocity,
f
the vapor velocity, = 0.85Uf, and compute the tower diameter,
1. Set column condenser and reboiler pressures, P andP , using
D B U
D , using Eq. (13.11).
the algorithm in Figure 9.9. T
2. Estimate the minimum number of stages, N , at total reflux 5. Estimate the installed cost of the distillation tower and its trays,
mi C , using Figure 16.13 or Eqs. (16.52), (16.57), (16.58), and
using the Fenske or Winn methods. Estimate the minimum B
n
R , using the Underwood equation.
reflux ratio (infinite stages),
mi (16.6
M
N, usingR∕ R = 1.3
Estimate the theoretical number of stages, 6. Estimate the cost of the condenser utility (e.g., cooling water) and
mi
and the Gilliland correlation (see Chapter 13 and the multimedia 6).
n
reboiler utility (e.g., steam).
n

ABCDEF

256.3
1 2 3 4
AB/CDEF I

Fig. 9.13a Fig. 9.13c Fig. 9.13d

AB CDEF

14.5 582.2 68.3

8 9 10
A/B I C/DEF II CDE/F I

A B C DEF CDE F

35.2 521.3

21 22
DE/F I C/DE II

DE F C DE

(b) Figure 9.13(Continued)

9.4 Sequencing of Ordinary Distillation Columns for the Separation of Nearly Ideal Liquid Mixtures 256
 Propane

257 Chapter 9 Synthesis of Separation

Propane Trains
1-Butene
Distillation
Mixed Butenes

1–Butene

Feed trans-Butene -2
Distillation n–Butane cis-Butene -2

n- Butane
trans- Butene
cis- Butene
n- Pentane Distillation
Distillation

Extractive
Distillation

n–Pentane

MSA

Figure 9.14 Lowest-cost sequence for Example 9.3

Marginal Vapor Rate Method A convenient method for determining the molar vapor rate in
an ordinary distillation column separating a nearly ideal system
The last two sections have presented alternative strategies for uses the Underwood equations to calculate the minimum reflux
sequencing ordinary distillation columns. The first of these ratio, Rmin. This is readily accomplished, as in the next example
focused on the usage of heuristics to select cost-effective with a process simulation program. The design reflux ratio is
separation sequences. However, reliance solely on heuristics may taken as R = 1.2Rmin. By material balance, the molar vapor rate,
ABCDEF
lead to conflicting results, and as shown in the last section, it is V, entering the condenser is given by V = D(R + 1), where D is
preferable to employ sequencing methods that rely on column the molar distillate rate. With the assumption that the feed to the
design and, in some cases, cost estimation. As shown, exhaustive column 1047.5
is a bubble-point liquid, the molar vapor rate through
search to calculate the1 annualized cost2 of every sequence 3 can 4
the column will be nearly constant at this value of V. In making
determine the optimal sequence provided that column-operating the calculations
AC/BDEF of MV, the selection of product purities is not
II
conditions are optimized, and may be justified for sequences critical because the minimum reflux ratio is not sensitive to
involving just three or possibly four products. However, less those purities. Thus, to simplify the material balance
rigorous methods areFig.available
9.13a that can produce goodFig.
Fig. 9.13b sequences,
9.13c
calculations, it is convenient to assume nearly perfect
although not always optimal. These methods, which attempt to separations with the light and lighter-than-light key components
reduce the search space, include those of Hendry and Hughes AC BDEF
leaving in the distillate and the heavy and heavier-than-heavy
(1972), Rodrigo and Seader (1975), Gomez and Seader (1976), key components leaving in the bottoms. Column top and
Seader and Westerberg (1977), and the marginal vapor rate (MV) bottom pressures are estimated with Figure 9.9. The column
method of Modi and Westerberg (1992). The latter method 21.1 46.6
14 feed pressure15is taken as the average of the top and bottom
outperforms the other methods and can be applied without the pressures. BDE/F I
A/C I
necessity of complete column designs and calculations of costs. EXAMPLE 9.4
For a given split between two key components, Modi and
Westerberg (1992) consider the difference in costs between the Use the marginal vapor rate method to determine a sequence for the
separation in the absence of nonkey components and the separation of light hydrocarbons specified in Figure 9.12(a) except
separation in the presence of nonkey components, defining this (1) remove the propane from the feed, (2) ignore the given
A C BDE E
difference as the marginal annualized cost (MAC). They show temperature and pressure of the feed, and (3) strive for recoveries of
that a good approximation of MAC is the MV, which (d) is the Figure 9.13(Continued
99.9% of the key components in each ) column. Use a process
corresponding difference in molar vapor rate passing up the simulation program, with the Soave-Redlich-Kwong equation of
state for K-values and enthalpies, to set.
column. The sequence with the minimum sum of column MVs is
selected. TheSequences
Table 9.3 good approximation
for Example 9.3 is due to the fact that vapor
rate is a reliable measure of cost because it is a major factor10.
Annual Cost, C A ($/yr)
inEstimate the cost of sales,COS. For comparison of separa-
Seque
determining column diameter as well as reboiler and condensertion sequences, include only the annual cost of utilities. See
1 – 5 – 16 –28 900,200 SectionSOLUTION
17.2.
areas (thus, columnnce and heat-exchanger capital costs) and
1 – 5 – 17 –29 872,400 11. Compute To the annualized C using
cost,
reboiler and condenser duties (thus, heat-exchanger annual produce A pureEq.
four nearly
= 0.2. of three
(17.10)from
products with the
the four-component
return feed,
operating costs). 1 – 6 –18 1,127,400 on investment,
r
five sequences ordinary distillation columns each are shown
1 – 7 – 19 –30 878,000 Similar steps are used to estimate the annualized cost for the extrac-
1 – 7 –20 1,095,600 tive distillation operation.
8 The lowest-cost sequenceis identified in Table 9.3 and
Figure 9.13(b) with its PFD presented in Figure 9.14 (which is
2 888,200
 in cost than the lowest. If only the
A C II and E F II splits
8 are prohibited as by Hendry and Hughes (1972), the consequence is
Lowest-cost 64unique separations and 227 sequences. However, every one of the
2 860,400
sequence
10– 22 additional 215 sequences is more than 350% higher in cost than the
9.4 Sequencing of Ordinary Distillation
lowest-cost Columns for the Separation of Nearly Ideal Liquid Mixtures
sequence. 258
3 – 11 – 23 –31 878,200
3 – 11 –24 1,095,700
in Figure 9.11. Let A = isobutane, B = n-butane, C = isopentane, and (e) Indirect
D = n-pentane. A total 25 of 10 unique separations is embedded in 844
Figure 9.11. These 3are12listed in Table 9.4 with867,400
the results of the
calculations for the top26column pressure, Ptop, in kPa; the molar
distillate rate, D, in
3 – kmol/hr;
13 –27 and the reflux 1ratio,
,080,100R, using the
shortcut (Fenske–Underwood–Gilliland or FUG) distillation model
4 – 14 –15 1,115,200
of the CHEMCAD process simulation program. This model applies
Reflux Ratio, Vapor Rate, Marginal Vapor (R = 1.2Rmin) V
= D(R+1) (kmol/hr) Rate (kmol/hr)

Table 9.4 Calculations of Marginal Vapor Rate, MV 10.7 1,594 0

11.9 1,757 163
Column Top Distillate Rate, Separation
Pressure (kPa) D (kmol/hr) 13.2 1,934 340
2.06 694 0
A/B 680 136.2 1.55 925 231
A/BC 680 136.2 A/BCD 680 136.2
Following the development of an optimal or near-optimal two interlinked cases (III and IV), five cases that include the use 3.06
sequence of simple, two-product distillation columns, revised of sidestreams (III, IV, V, VI, and VII), and one case (V)
involvsequences involving complex, rather than simple, distillation ing a column with two feeds. All columns in Cases I, II, V, VI,
921
columns should be considered. Some guidance is available from and VII have condensers and reboilers. In Cases III and IV, the a
study by Tedder and Rudd (1978a, b) of the separation of first column has a condenser and reboiler. In Case III, the rectiternary227
mixtures (A, B, and C in order of decreasing volatility) fier column has a condenser only whereas the stripper in Case IV in which
eight alternative sequences of one to three columns has a reboiler only. The interlinking streams that return from the were
considered, seven of which are shown in Figure 9.15. The second column to the first column thermally couple the columns
configurations include the direct and indirect sequences (I and II), in Cases III and IV.
B/C 490 226.8 AB/C 560 362.9 B/CD 490 226.8 2.11 1,129 435 13.5 2,632 0
AB/CD 560 362.9 6.39 3,017 385
C/D 210 181.5 BC/D 350 408.3 4.96 3,245 613
ABC/D 430 544.4 A B A

Complex and Thermally Coupled

C C B
Distillation Columns I. Direct Sequence II. Indirect Sequence
the Underwood equations to estimate the minimum reflux ratio, as
A B A
described by Seader et al. (2016). Column feeds were computed as
bubble-point liquids at Ptop + 35kPa. Also included in Table 9.4 are
values of the column molar vapor rate, V, and marginal vapor rate,
MV, both in kmol/hr. From Table 9.4, the sum of the marginal vapor
rates is calculated for each of the five sequences in Figure 9.11. The
results are given in Table 9.5.
Table 9.5 shows that the preferred sequence is the one that C C B
performs the two most difficult separations, A/B and C/D, in the III. Distillation with IV. Distillation with
absence of nonkey components. These two separations are far more Vapor Sidestream Liquid Sidestream
difficult than the separation B/C. The direct sequence is the next Rectifier Stripper
A
best.

Table 9.5 Marginal Vapor Rates for the Five Possible Sequences
B
Sequence in Marginal Vapor Rate, Figure 9.11 MV
(kmol/hr)
V. Prefractionator
C
(a) Direct 567 with Distillation
(b) 725 (c) 435
A A
(d) 776
259 Chapter 9 Synthesis of Separation Trains B

VI. Distillation B VII. Distillation with drawn. Each column is provided with its own condenser and
with Lower Upper
Sidestream reboiler. As shown in Figure 9.17, eliminating the condenser
C C
Sidestream and reboiler in the prefractionator and providing, instead, reflux
and boilup to that column from the product column can
thermally couple this arrangement, which is referred to as a
Figure 9.15 Configurations for ternary distillation.
Petlyuk system after its chief developer, and is described by
Petlyuk et al. (1965). The prefractionator separates the ternary-
As shown in Figure 9.16, optimal regions for the various mixture feed, ABC, into a top product containing A and B and a
configurations depend on the process feed composition and on an bottom product of B and C. Thus, component B is split between
ease-of-separation index (ESI), which is defined as the relative the top and bottom streams exiting from the prefractionator.
volatility ratio, �A,B∕�B,C. It is interesting to note that a ternary The top product is sent to the upper section of the product
mixture is separated into three products with just one column in column, and the bottom product is sent to the lower section.
Cases VI and VII in Figure 9.15, but the reflux requirement is The upper section of the product column provides the reflux for
excessive unless the feed contains a large amount of B, the the prefractionator, and the lower section provides its boilup.
component of intermediate volatility, and little of the component The product column separates its two feeds into a distillate of
that is removed from the same section of the column as B. A, a sidestream of B, and a bottoms of C. Fidkowski and
Otherwise, if the feed is dominated by B but also contains Krolikowski (1987) determined the minimum molar boilup
appreciable amounts of A and C, the prefractionator case (V) is vapor requirements for the Petlyuk system and the other two
optimal. Perhaps the biggest surprise of the study is the thermally coupled systems (III and IV)
superiority of distillation with a vapor sidestream rectifier, which
is favored for a large region of the feed composition when ESI >
1.6. The results of Figure 9.16 can be extended to
multicomponent separation problems involving more than three
components if difficult ternary separations are performed last.
Case V in Figure 9.15 consists of a prefractionator followed
by a product column from which all three final products are
C C
α
ESI = ____
AB
α
_ BC

II II
III
VII VII
III
V
VI I V VI
B A B A
Expected Regions of Optimality Expected Regions of Optimality
ESI 1.6 ESI > 1.6

Figure 9.16 Regions of optimality for ternary distillation Figure 9.17 Thermally coupled Petlyuk system.
configurations (Adapted from Tedder and Rudd, 1978a, b).
Reproduced with permission of the American Institute of Chemical
Engineers. Copyright © 1978 AIChE. All rights reserved.
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures
in Figure 9.15, assuming constant relative volatilities, constant
molar overflow, and bubble-point liquid feed and products.
Fidkowski and Krolikowski compared the requirements to
those of the conventional direct and indirect sequences shown
as Cases I and II in Figure 9.15 and proved that for all
combinations of feed flow rates of the components A, B, and
C, as well as all values of relative volatilities, that: (1) the
Petlyuk system has the lowest minimum molar boilup vapor
requirements and (2) Cases III and IV in Figure 9.15 are
equivalent and have lower minimum molar boilup vapor
requirements than either the direct or indirect sequence.
Despite its lower vapor boilup requirements, no industrial
installations of a two-column Petlyuk system have been
reported. Two possible reasons for this, as noted by Agrawal
and Fidkowski (1998), are (1) an unfavorable thermodynamic
efficiency when the three feed components are not close
boiling because all of the reboiler heat must be supplied at the
highest temperature and all of the condenser heat must be
removed at the lowest temperature and (2) the difficulty in
controlling the fractions of vapor and liquid streams in the
product column that are returned to the prefractionator as
boilup and reflux, respectively. The Petlyuk system can be
embodied into a single column, with a significantly reduced
capital cost, by using a dividing-wall column (also called
divided wall and column in column), a concept described in a
patent by Wright (1949) and shown by his patent drawing in
Figure 9.18. Because the dividing-wall column makes possible
savings in both energy and capital and because control
difficulties appear to have been solved, it is attracting much
attention. The first dividing-wall column was installed by
BASF in 1985. A number of such columns using packing have
been installed in the past 15 years, and the first dividing-wall
column
260
Figure 9.18 Dividing-wall (partition) column of Wright.
Column Column
1 2
Figure 9.19 Heat-integrated direct sequence of two distillation columns.
using trays was recently announced. Agrawal and Fidkowski
(1998) present other thermally fully coupled (FC) systems of
distillation columns that retain the benefit of a minimum vapor
requirement and afford easier control. Energy savings can also be
achieved by heat-integrating the two columns in a direct
sequence. In Figure 9.19, Column 2 is operated at a higher
pressure than Column 1, such that the condenser duty of Column
2 can provide the reboiler duty of Column 1. Rev et al. (2001)
show that heat-integrated systems are often superior in
annualized cost to the Petlyuk system. For further discussion of
design aspects of heat-integrated distillation columns, see
Section 11.8, and Examples 11.14 and 11.25; control aspects are
presented in Example 20.2 and Example 20S.8.
9.5 SEQUENCING OF OPERATIONS
FOR THE SEPARATION OF
NONIDEAL LIQUID MIXTURES
When a multicomponent fluid mixture is nonideal, its separation
by a sequence of ordinary distillation columns will not be
technically and/or economically feasible if relative volatilities
between key components drop below 1.05 and, particularly, if
azeotropes are formed. For such mixtures, separation is most
commonly achieved by sequences composed of ordinary
distillation columns, enhanced distillation columns, and/or
liquid–liquid extraction equipment. Membrane and adsorption
separations can also be incorporated into separation sequences,
but their use is much less common. Enhanced distillation
operations include extractive distillation, homogeneous
azeotropic distillation, heterogeneous azeotropic distillation,
pressure-swing distillation, and reactive distillation. These
operations are considered in detail in Perry’s Chemical
Engineers’ Handbook (Green and Perry, 2006) and by Seader et
al. (2016), Stichlmair and Fair (1998), and Doherty and Malone
(2001). A design-oriented introduction to enhanced distillation is
presented here.
261 Chapter 9 Synthesis of Separation Trains

In many processes involving oxygenated organic compounds
such as alcohols, ketones, ethers, and acids, often in the presence
of water, distillation separations are complicated by the presence
of azeotropes. Close-boiling mixtures of hydrocarbons (e.g.,
benzene and cyclohexane whose normal boiling points only
differ by 1.1∘F) can also form azeotropes. For these and other
mixtures, special attention must be given to the distillation
boundaries in the composition space that confine the
compositions for any one column to lie within a bounded region
of the composition space. To introduce these boundaries leading
to approaches for the synthesis of separation trains, several
concepts concerning azeotropes, residue curves, and distillation
lines are reviewed in the subsections that follow.
of Raoult’s law, as shown in Figure 9.20a for the benzene-
toluene mixture at 90∘C. Note that the bubble-point curve (P − x)
is linear between the vapor pressures of the pure species (at x1 =
0, 1), and the dew-point curve (P − y) lies below it. When the (x1,
y1) points are graphed at different pressures, the familiar vapor-
liquid equilibrium curve is obtained, as shown in Figure 9.20b.
Using McCabe-Thiele analysis, it is shown readily that for any
feed
1.4
P –x (TEMP = 90.0 C)
1.2
P –y (TEMP = 90.0 C)

Pressure Bar
1

0.8
Azeotropy
0.6
The word azeotrope is derived from the Greek words �����´
(boil) and ��o´�o� (turning) combined with the prefix �- (no) to
give the overall meaning, “to boil unchanged,” implying that the 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
vapor emitted has the same composition as the liquid Molefrac Benzene

(Swietoslawski, 1963). When classifying the many azeotropic ( a)

mixtures, it is helpful to examine their deviations from Raoult’s 1
law (Lecat, 1918). 0.9
When two or more fluid phases are in physical equilibrium, 0.8
the chemical potential, fugacity, and activity of each species is Vapor Molefrac Benzene
0.7
the same in each phase. Thus, in terms of species mixture 0.6
fugacities for a vapor phase in physical equilibrium with a single 0.5
liquid phase, 0.4
f Vj = f Lj j = 1, …,C (9.11) 0.3
0.2 (TEMP = 90.0 C )
Substituting expressions for the mixture fugacities in terms of 0.1
mole fractions, activity coefficients, and fugacity coefficients,
0 0.10.20.30.4 0.5 0.60.70.80.9 1
Liquid Molefrac Benzene
�V
yj j P = xj�jLfjL j = 1, …,C (9.12) (b)

where � is a mixture fugacity coefficient, γ is a mixture activity
coefficient, and f is a pure-species fugacity.
For a binary mixture with an ideal liquid solution (�jL = 1) and
a vapor phase that forms an ideal gas solution and obeys the ideal
�
gas law ( Vj = 1 and fjL = Psj), Eq. (9.12) reduces to the following
two equations for the two components 1 and 2:
y1P = x1Ps1 (9.13a)
y2P = x2Ps2 (9.13b)
where Psj is the vapor pressure of species j. Adding Eqs. (9.13a)
and (9.13b), noting that mole fractions must sum to 1,
Figure 9.20 Phase diagrams for the benzene–toluene mixture at 90∘C,
calculated using ASPEN PLUS: (a) P − x − y diagram: (b) x − y
diagram.
composition, there are no limitations to the values of the mole
fractions of the distillate and bottoms products from a
distillation tower.
However, when the mixture forms a nonideal liquid phase
and exhibits a positive deviation from Raoult’s law (�jL > 1, j =
1,2), Eq. (9.14) becomes

(y1 + y2)P = P = x1Ps1 + x2Ps2 P = x1�1LPs1 + (1 − x1)�2LPs2 (9.15)

= x1Ps1 + (1 − x1)Ps2 Furthermore, if the boiling points of the two components are
= P + (P − P 2)x1
s s s
(9.14) close enough, the bubble- and dew-point curves may reach a
2 1
maximum at the same composition, which by definition is the
This linear relationship between the total pressure, P, and the azeotropic point. Such a situation is illustrated in Figure 9.21a
mole fraction, x1, of the most volatile species is a characteristic for the binary mixture of isopropyl ether (1) and isopropyl
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 262

alcohol (2) at 70∘C. Figure 9.21b shows the corresponding x − bubble- and dew-point curves drop below the straight line that
y diagram, and Figure 9.21c shows the bubble- and dew-point represents the bubble points for an ideal mixture, as
curves on a T − x − y diagram at 101 kPa. Note the minimum- anticipated by examination of Eq. (9.15). Furthermore, when
boiling azeotrope at 66∘C, where x1 = y1 = 0.76. Feed streams the bubbleand dew-point curves have the same minimum, an
having lower isopropyl ether mole fractions cannot be purified azeotropic composition is defined, as shown in Figure 9.22a
beyond 0.76 in a distillation column, and streams having for the binary mixture of acetone (1) and chloroform (2) at
higher isopropyl ether mole fractions have distillate mole 64.5∘C, where x1 = y1 = 0.35. For this system, Figures 9.22b
fractions that have a lower bound of 0.76. Consequently, the and 9.22c show the corresponding x–y diagram and T–x − y
azeotropic composition is commonly referred to as a diagram at 101 kPa. On the latter diagram, the azeotropic
distillation boundary. point is at a maximum temperature, and consequently, the
Similarly, when the mixture exhibits the less-common system is said to have a maximum-boiling azeotrope. In this
negative deviation from Raoult’s law (�jL < 1,j = 1,2), both the case, feed streams having
1.2 1.2

1.15
1.1
1.1
1

Pressure Bar
Pressure Bar

P –x (TEMP = 60.0 C)
1.05 P –y (TEMP = 60.0 C)
0.9
1
0.8 0.95
P –x (TEMP = 70.0 C)
0.7 P –y (TEMP = 70.0 C) 0.90

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Molefrac IPE Molefrac Acetone

( a) ( a)

1 1
0.9 0.9
0.8 0.8
Vapor Molefrac Acetone
Vapor Molefrac IPE

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2 ( PRES = 1.01 BAR )
0.1 ( PRES = 1.01 BAR ) 0.1

0 0.10.20.30.40.50.60.70.80.91 0 0.1 0.20.30.4 0.50.60.70.80.91

Liquid Molefrac IPE Liquid Molefrac Acetone

( b) ( b)

82.5 66

80 64
T –x (PRES = 1.01 BAR)
Temperature ° C

77.5
Temperature ° C

T –y (PRES = 1.01 BAR)

62
75
T –x (PRES = 1.01 BAR)
72.5 60 T –y (PRES = 1.01 BAR)
70
58
67.5

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Molefrac IPE Molefrac Acetone

(c) (c)

Figure 9.21 Phase diagrams for the isopropyl ether-isopropyl Figure 9.22 Phase diagrams for the acetone–chloroform binary
alcohol binary computed using ASPEN PLUS: (a) P–x–y diagram at computed using ASPEN PLUS: (a) P–x–y diagram at 60∘C; (b) x–y
70∘C; (b) x–y diagram at 101 kPa; (c) T–x–y diagram at 101 kPa. diagram at 101 kPa; (c) T–x–y diagram at 101 kPa.
263 Chapter 9 Synthesis of Separation Trains

lower acetone mole fractions cannot be purified beyond 0.35 in unity. Based on the general phase equilibria Eq. (9.12), the
the bottoms product of a distillation column, and streams having criterion for azeotrope formation is given by:
higher acetone mole fractions have a lower bound of 0.35 in the
acetone mole fraction of the bottoms product. yj = �jLfjL =
In summary, at a homogeneous azeotrope, xj = yj, j = 1,…,C, the Kj = xj �Vj P 1 j = 1, …,C (9.16)
expression for the equilibrium constant, Kj, for species j becomes
Figure 9.23 Binary phase diagram at a fixed pressure

1 x1, azeo 1 0 x10, azeo

(a) (b)
2 for (a) homogeneous azeotrope; (b) heterogeneous azeotrope.
where the degree of nonideality is expressed by the deviation residue curve is constructed by tracing the composition of the
from unity of the activity coefficients and fugacities for the equilibrium liquid residue of a simple (Rayleigh batch)
liquid phase and of the fugacity coefficients for the vapor phase. distillation in time, starting from a selected initial composition of
At low the charge to the still using the following numerical procedure.
�V Consider L moles of liquid with mole fractions xj(j = 1,…,C) in
pressure, j = 1 and fjL = Psj so that Eq. (9.16) reduces to
a simple distillation still at its bubble point as illustrated in
Figure 9.24. Note that the still contains no trays and that no
y Psj reflux is provided. As heating begins, a small portion of this
Kj = j = �jL = 1 j = 1, …,C (9.17) xj P
liquid, ΔL moles, is vaporized. The instantaneous vapor phase
Because the K-values for all of the species are unity at an has mole
azeotrope point, a simple distillation approaches this point at dL ; y j

which no further separation can occur. For this reason, an

azeotrope is often called a stationary or fixed or pinch point.
For a minimum-boiling azeotrope, when the deviations from
Raoult’s law are sufficiently large (�jL ≫ 1, usually > 7), splitting
L ; xj
the liquid phase into two liquid phases (phase splitting) may
occur, and a minimum-boiling, heterogeneous azeotrope may
form that has a vapor phase in equilibrium with the two liquid
phases. A heterogeneous azeotrope occurs when the vapor–liquid
envelope overlaps with the liquid–liquid envelope, as illustrated Figure 9.24 Simple distillation still.
in Figure 9.23b. For a homogeneous azeotrope, when x1 = x1,azeo =
y1, the mixture boils at this composition, as shown in Figure fractions yj(j = 1,…,C), assumed to be in equilibrium with the
9.23a; for a heterogeneous azeotrope, however, when the overall remaining liquid. Since the residual liquid, L − ΔL moles, has
liquid composition of the two liquid phases, x1 = x10,azeo = y1, the
mole fractions xj + Δxj, the mass balance for species j is given by
mixture boils at this overall composition, as illustrated in Figure
9.23b, but the three coexisting phases have distinct compositions. Lxj = (ΔL)yj + (L − ΔL)(xj + Δxj) j = 1,…,C − 1 (9.18)
In the limit, as ΔL → 0,
Residue Curves dxj
= xj − yj = xj(1 − Kj{T,P,x,y}) j = 1,…,C − 1
To understand better the properties of azeotropic mixtures that dL∕L and setting d (9.19)
contain three chemical species, it helps to examine the properties t̂ = dL∕L,
of residue curves on a ternary diagram. A collection of residue
curves, which is called a residue curve map, can be computed
and drawn by any of the major simulation programs. Each
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 264

saddles at the vertices of the light (L) and intermediate (I)

dxj j = 1,…,C − 1
= xj − yj = xj(1 − Kj{T,P,x,y}), d ̂t
(9.20)
where Kj is given by Eq. (9.16). In Eq. (9.20), ̂t can be interpreted
as the dimensionless time with the solution defining a family of
residue curves, as illustrated in Figure 9.25. Note that each
residue curve is the locus of the compositions of the residual
liquid in time, as vapor is boiled off from a simple distillation
still. Often, an arrow is assigned in the direction of increasing
time (and increasing temperature). Note that the residue curve
map does not show the equilibrium vapor composition
corresponding to each point on a residue curve.
Another important property is that the fixed points of the
residue curves are points where the driving force for a change in
the liquid composition is zero; that is, dx∕d̂t = 0. This condition is
satisfied at the azeotropic points and the pure-species vertices.
For a ternary mixture with a single binary azeotrope, as in Figure
species.
It is of special note that the boiling points and the
compositions of all azeotropes can be used to characterize
residue curve maps. In fact, even without a simulation program
to compute and draw the detailed diagrams, this information
alone is sufficient to sketch the key characteristics of these
diagrams using the following procedure. First, the boiling points
of the pure species are entered at the vertices. Then the boiling
points of the binary azeotropes are positioned at the azeotropic
compositions along the edges, with the boiling points of any
ternary
Stable Node Unstable Node Saddle

9.25, there are four fixed points on the triangular

H (a) (b) (c)
Stable
Node
Figure 9.26 Stability of residue curves for a ternary system in the
vicinity of a binary azeotrope.
B
Nitrogen
79.2 K
A

1.31 Bar

Octane
398.8 K
A
A C
Unstable Saddle 92.5 K 89.8 K
Node 1.013 Bar Oxygen Argon
L I D ( a) 000
389.1 K
Saddle

Figure 9.25 Residue curves of a ternary system with a minimum-boiling

Acetone
binary azeotrope. C 329.2 K
B
409.2 K E A
408.1 K
Ethylbenzene 400.1 K 2-Ethoxyethanol
( b) 120
diagram: the binary azeotrope and the three vertices. 1.013 Bar
D
Furthermore, the behavior of the residue curves in the vicinity of 328.7 K
the fixed points depends on their stability. When all of the 330.5 K F 337.4 K
G
residue curves are directed by the arrows to the fixed point, it is
referred to as a stable node, as illustrated in Figure 9.26a; when
all are directed away, the fixed point is an unstable node (as in C B
337.7 K E 334.2 K
Figure 9.26b); and finally, when some of the residue curves are Methanol 326.4 K Chloroform
directed to and others are directed away from the fixed point, it is (c) 311-S
referred to as a saddle point (as in Figure 9.26c). Note that for a
Figure 9.27 Maps of residue curves or distillation lines: (a) system
ternary system, the stability can be determined by calculating the
without azeotropes; (b) system with two binary azeotropes; (c) system
eigenvalues of the Jacobian matrix of the nonlinear ordinary with binary and ternary azeotropes (Stichlmair et al., 1989).
differential equations that comprise Eq. (9.20). Reproduced with permission of the American Institute of Chemical
As an example, consider the residue curve map for a ternary Engineers. Copyright © 1989 AIChE. All rights reserved.
system with a minimum-boiling binary azeotrope of heavy (H)
and light (L) species as shown in Figure 9.25. There are four
fixed points: one unstable node at the binary azeotrope (A), one azeotropes positioned at their compositions within the triangle.
stable node at the vertex for the heavy species (H), and two Arrows are assigned in the direction of increasing temperature in
265 Chapter 9 Synthesis of Separation Trains

a simple distillation still. As examples, typical diagrams for Defining h as the dimensionless distance from the top of the
mixtures involving binary and ternary azeotropes are illustrated tower, a backward-difference approximation at tray n is given by
in Figure 9.27. Figure 9.27a is for a simple system, without
azeotropes, involving nitrogen, oxygen, and argon. In this
mixture, nitrogen is the lowest-boiling species (L), argon is the dh dx||||n ≈ xn − xn−1 (9.22)
intermediate boiler (I), and oxygen is the highest-boiling species
(H). Thus, along the oxygen–argon edge, the arrow is pointing to
the oxygen vertex, and on the remaining edges, the arrows point
dx || ≈ − Vn D
away from the nitrogen vertex. Since these arrows point away at | xn y + xD (9.23)
the nitrogen vertex, it is an unstable node, and all of the residue dh |n L n− 1 n L n− 1
curves emanate from it. At the argon vertex, the arrows point to
and away from it. Since the residue curves turn in the vicinity of
this vertex, it is not a terminal point. Rather, it is referred to as a
saddle point. All of the curves end at the oxygen vertex, which is
a terminal point or stable node. For this ternary mixture, the map
shows that pure argon, the intermediate boiler, cannot be
D
obtained in a simple distillation. xD
h

Simple Distillation Boundaries

The graphical approach described here is effective in locating the
starting and terminal points and the qualitative locations of the L n–1
Vn xn – 1
residue curves. As illustrated in Figures 9.27b and 9.27c, it yn

works well for binary and ternary azeotropes that exhibit

multiple starting and terminal points. In these cases, one or more
xn
simple distillation boundaries called separatrices (e.g., curved xn
line DE in Figure 9.27b) divide these diagrams into regions with Rearranging Eq. (9.21) and substituting in Eq. (9.22),
distinct pairs of starting and terminal points. For the separation
of homogeneous mixtures by simple distillation, these Figure 9.28 Schematic of rectifying section.
separatrices cannot be crossed unless they are highly curved. A y2
y1
feed located in region ADECA in Figure 9.27b has a starting y3 x1
x0
point approaching the composition of the binary azeotrope of x2
octane and 2-ethoxyethanol and a terminal point approaching
pure ethylbenzene, whereas a feed located in region DBED has a y4
starting point approaching the same binary azeotrope but a x3
terminal point approaching pure 2-ethoxyethanol. In this case, a
pure octane product is not possible. Figure 9.27c is even more y5
complex. It shows four distillation boundaries (curved lines GC, x4
DG, GF, and EG), which divide the diagram into four distillation
regions. Tie Lines
y6
x5

Distillation Towers Figure 9.29 Distillation line and its tie lines.
When tray towers are modeled assuming vapor–liquid
equilibrium at each tray, the residue curves approximate the
liquid composition profiles at total reflux. To show this, a species At total reflux, with D = 0 and Vn = Ln−1, Eq. (9.23) becomes
balance is performed for the top n trays, counting down the
tower, as shown in Figure 9.28:
dx
Ln−1xn−1 + DxD = Vnyn (9.21) dh ||||n ≈ xn − yn (9.24)

where D and xD are the molar flow rate and vector of mole
fractions of the distillate. Similarly, Ln−1 and xn−1 are for the liquid
Hence, Eq. (9.24) approximates the operating lines at total reflux
leaving tray n − 1, and Vn and yn are for the vapor leaving tray n.
and, because ̂t and h are dimensionless variables and Eq. (9.20)
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 266

is identical in form, the residue curves approximate the operating ρF

δ1
δ2 ρE
lines of a distillation tower operating at total reflux.

E F

Distillation Lines b
a
An exact representation of the operating line for a distillation ρ1 ρC
ρD

tower at total reflux, also known as a distillation line [as defined

D
by Zharov (1968) and Zharov and Serafimov (1975)], is shown B
in A
Figure 9.29. Note that, at total reflux, C

xn = yn+1 n = 0,1,… (9.25) (a) ( b)

Distillation Line
Furthermore, assuming operation in vapor–liquid equilibrium,
Residue Curve
the mole fractions on trays n, xn, and yn+1 lie at the ends of the Tie Line
Liquid Vapor
equilibrium tie lines.
To appreciate better the differences between distillation lines Figure 9.31 Geometric relationship between distillation lines and
and residue curves, consider the following observations. First, residue curves.
Eq. (9.20) requires the tie line vectors connecting liquid
composition x and vapor composition y, at equilibrium, to be
tangent to the residue curves, as illustrated in Figure 9.30.
Since these tie line vectors must also be chords of the Computing Azeotropes for
distillation lines, the residue curves and the distillation lines must
Multicomponent Mixtures
intersect at the liquid composition x. Note that when the residue
curveislinear(asforbinarymixtures),thetielinesandtheresidue Gmehling (1994) provides data on more than 15,000 binary
curve are collinear, and consequently, the distillation lines azeotropes and 900 ternary azeotropes. Undoubtedly, many more
coincide with the residue curves. ternary azeotropes exist as well as untold numbers of azeotropes
involving more than three components. When a process
Figure 9.31a shows two distillation lines (�1 and �2) that
simulation program is used to compute a residue curve map for a
intersect a residue curve at points A and B. As a consequence of
ternary system at a specified pressure, compositions and
Eq. (9.20), their corresponding vapor compositions at
temperatures of all azeotropes are automatically estimated. The
equilibrium, a and b, lie at the intersection of the tangents to the
results depend, of course, on the selected vapor pressure and
residue curves at A and B with the distillation lines �1 and �2.
liquid-phase activity coefficient correlations. For quaternary and
Clearly, the distillation lines do not coincide with the residue
higher systems, the arclength homotopy-continuation method of
curves, an assumption that is commonly made but that may
Fidkowski, Malone, and Doherty (1993) can be used for
produce significant errors. In Figure 9.31b, a single distillation
homogeneous systems to estimate all azeotropes. They find all
line connects the compositions on four adjacent trays (at C, D, E,
roots to the following equations, which define a homogeneous
F) and crosses four residue curves (�C,�D,�E,�F) at these points.
azeotrope:
yj − xj = 0 j = 1,2,…,C − 1 (9.26)
Distillation
Line

yj xj j = 1,2,…,C − 1 (9.27)
∑C
Residue Curve
Q
y–x xj = 1 (9.28)
dx = y – x
___ j=1
Tie Line
dt
C
P ∑yj = 1
y
j=1

x (9.29)
xj ≥ 0, j = 1,2,…,C (9.30)
To find the roots,
they construct the following homotopy to replace Eqs. (9.26) and
O
(9.27) based on gradually moving from an ideal K-value based
Figure 9.30 Residue curve and distillation line through P. on Raoult’s law to the more rigorous expression of Eq. (9.27):
267 Chapter 9 Synthesis of Separation Trains
yj − xj = {((1 − t) + t �jLjV ) PPsj − 1}xj
�
= H(t,xj) = 0 j = 1,2,…,C − 1 (9.31)
Initially, the homotopy parameter, t, is set to 0 and all values of xj
are set to 0 except for one, which is set to 1.0. Then t is gradually
and systematically increased until a value of 1.0 is obtained.
With each increase, the temperature and mole fractions are
computed. If the resulting composition at t = 1.0 is not a pure
component, it is an azeotrope. By starting from each pure
component, all azeotropes are computed. The method of
Fidkowski, Malone, and Doherty (1993) is included in many of
the process simulation programs. Eckert and Kubicek (1997)
extended the method of Fidkowski, Malone, and Doherty to the
estimation of heterogeneous multicomponent azeotropes.
Distillation-Line Boundaries and
Feasible Product Compositions
Of great practical interest is the effect of distillation boundaries
on the operation of distillation towers. To summarize a growing
body of literature, it is well established that the compositions of a
distillation tower operating at total reflux cannot cross the
distillation-line boundaries except under unusual circumstances
where these boundaries exhibit a high degree of curvature. This
provides the total-reflux bound on the possible (feasible)
compositions for the distillate and bottoms products.
y feed, and distillate mole fractions is shown. Figure 9.32a shows
As shown in Figure 9.32a, at total reflux, xB and D reside on a
distillation line. Furthermore, these compositions lie collinear
with the feed composition, xF, on the overall material balance
line. As the number of stages increases, the operating curve
becomes more convex and in the limit approaches the two sides
of the triangle that meet at the intermediate boiler. As an
example, an operating line at total reflux (minimum stages) is the
curve AFC in Figure 9.33a. At the other extreme, as the number
of stages increases, the operating curve becomes more convex
approaching ABC, where the number of stages approaches
infinity (corresponding to minimum reflux). Hence, the operating
line for a distillation tower that operates within these limiting
regimes lies within the region ABCFA in Figure 9.33a. Note that
H
xB
Distillation
xF Line
xD
H
L yD I
Tie Line
( a)
xB
Residue
xF Curve
xD
L I
yD Tie Line
(b)
Figure 9.32 Overall mass balance line with a partial/total condenser.
when a distillation tower operates with a partial condenser as the
feed and product streams are decreased toward total reflux, the
last bubble of vapor distillate has the mole fractions yD, as shown
in Figures 9.32a and 9.32b. Consequently, as total reflux is
approached, the material balance line connecting the bottoms,
y
the distillation line that passes through the xB and D mole
fractions, and Figure 9.32b shows the residue curve that passes
through the xD mole fractions, and approximately through the xB
mole fractions.
Two additional bounds in Figure 9.33a are obtained as
follows. First, in the limit of a pure nitrogen distillate, the line
AFE represents a limiting overall material balance for a feed
composition at point F, with point E at the minimum
concentration of oxygen in the bottoms product. Similarly, in the
limit of a pure oxygen bottoms, the line CFD represents a
limiting overall material balance, with point D at the minimum
concentration of nitrogen in the distillate along the nitrogen–
argon axis. Hence, the distillate composition is confined to the
shaded region ADFA, and the bottoms product composition lies
in the shaded region CEFC. Operating lines that lie within the
region ABCFA connect the distillate and bottoms product
compositions in these shaded regions. At best, only one pure
species can be obtained. In addition, only those species located at
the end points of the distillation lines can be recovered in high
purity. Hence, the end points of the distillation lines determine
the potential distillate and bottoms products for a given feed.
This also applies to the complex mixtures in
Figures9.33band9.33cwherethelocationofthefeedpointdetermines
the distillation region in which the potential distillate and
bottoms product compositions lie. For example, in Figure 9.33b,
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 268

for can be recovered in high purity for feeds in the region LTGCL.
For a feed in the region EDTHGBE, no pure product is possible.
Before attempting rigorous distillation calculations with a
simulation program, it is essential to establish, with the aid of
computer-
generatedresidualcurvemaps,regionsofproductcomposition
feasibility such as shown in Figure 9.33. Otherwise, it is possible
to waste much time and effort in trying to converge distillation
calculations when specified product compositions are
impossible.

Heterogeneous Distillation
In heterogeneous azeotropic distillation, an entrainer is utilized
that concentrates in the overhead vapor and, when condensed,
causes the formation of a second liquid phase that can be
decanted and recirculated to the tower as reflux. The other liquid
phase, as well as the bottoms, are the products from the
feed F, only pure 2-ethoxyethanol can be obtained. When the distillation. This is possible when the entrainer forms a
feed is moved to the left across the distillation-line boundary, heterogeneous azeotrope with one or more of the species in the
pure ethylbenzene can be obtained. In Figure 9.33c, only feed. For example, ethanol and water form a minimum-boiling
methanol azeotrope at 89 mol% ethanol and 1 atm, but by using a suitable
B entrainer, it is possible to devise a process to produce pure
89.8 K
Argon
ethanol as a product. Such a process is described next.
Ethanol
351.5K

351.2K
349.7K
Limit of two liquid
E phases in equilibrium
C

F
347.4K
T W

383.6K 357.0K 373.0K

Toluene Water

337.7 K G 334.2 K
Methanol 326.4 K Chloroform
Figure 9.34 Ternary composition diagram for a mixture of ethanol,
(c)
water, and toluene.

Figure 9.33 Regions of feasible distillate and bottoms product

compositions (shaded) for a ternary mixture: (a) system without
azeotropes; (b) system with two binary azeotropes; (c) system with
binary and ternary azeotropes (Stichlmair et al., 1989). Reproduced with
permission of the American Institute of Chemical Engineers. Copyright
© 1989 AIChE. All rights reserved.
Although toluene is the highest-boiling species, it is an
appropriate entrainer because it forms minimum-boiling
azeotropes with both water and ethanol. As shown in the ternary
composition diagram for the mixture of ethanol, water, and
toluene in Figure 9.34, three binary azeotropes and a single
tertiary azeotrope are formed, dividing the phase plane into three
distillation regions: E within which pure ethanol can be
produced, W within which pure water can be produced, and T
within which pure toluene can be produced. Since the feed
composition lies within distillation region W, purified ethanol
cannot be produced directly from the feed stream, but purified
water can be produced. It is noted that the tertiary azeotrope is a
heterogeneous one because it resides inside the region in which
two liquid phases are in equilibrium.
269 Chapter 9 Synthesis of Separation Trains

These observations guide the construction of the PFD, shown Ethanol

351.5 K
in Figure 9.35a, following the material balance lines positioned B2
on the ternary phase diagram presented in Figure 9.35b. Noting
K
that distillation region E can produce pure ethanol, we begin by M
351.2 K
placing the azeotropic tower, C-2, to recover ethanol as high- 349.7 K
C-2
purity bottoms product, B2, with the overhead vapor of C-2 D1
approaching the ternary, minimum-boiling, heterogeneous M2 C-1
azeotrope, D2. This condenses into two liquid phases, one rich in
S1 M1
toluene (point S2) and the other rich in water (point S1), which F
are separated in the decanter, represented in Figure 9.35b by the D2
347.4 K
tie line S2-D2-S1. As the aqueous stream, S1, is inside S2 S-1

distillation region W, it is mixed with F to produce M1, the feed

to the preconcentrator tower, C-1, which produces high-purity
water as bottoms, B1, and a distillate, D1, just to right of the
simple distillation boundary separating regions E and W. The
organic stream drawn from the decanter, S2, is mixed with D1 to
produce M2, the feed stream to the azeotropic tower, C-2, thus
completing the process.
The distillation sequence shown in Figure 9.35a is only one of
several sequences involving from two to four columns that have
been proposed and/or applied in industry for separating a mixture
by employing heterogeneous azeotropic distillation.
Most common is the three-column sequence from the study of
Ryan and Doherty (1989) as shown in Figure 9.36a. When used
to separate a mixture of ethanol and water using benzene as the
entrainer, the three columns perform the separation in the
Water-rich
S2 Toluene-rich
D1 D2
S1
F C–1 C–2
M1 M2
B1 B2
L B1
383.6 K 357.0 K 373.0 K
Toluene Water
(b) 311-S
Figure 9.35 Dehydration of ethanol using toluene as an entrainer: (a)
process flow diagram; (b) ternary composition diagram.
following manner, where the material-balance lines for Columns
2 and 3 are shown in Figure 9.36b. The aqueous feed, F1, dilute
in ethanol, is preconcentrated in Column 1 to obtain a pure water
bottoms, B1, and a distillate, D1, whose composition approaches
that of the homogeneous minimum-boiling binary azeotrope. The
distillate becomes the feed to Column 2, the azeotropic column,
S2 where nearly pure ethanol, B2, is removed as bottoms. The
overhead vapor from Column 2, V2, is close to the composition
S–1 S1
Decanter
of the heterogeneous ternary azeotrope of ethanol, water, and
benzene. When condensed, it separates into two liquid phases in
the decanter. Most of the organic-rich phase, L2, is returned to
Column 2 as reflux. Most of the water-rich phase, D2, is sent to
Column 3, the entrainer recovery column. Here, the distillate,
D3, consisting mainly of ethanol but with appreciable amounts
of benzene and water, is recycled to the top of Column 2. The
bottoms, B3 from Column 3, is nearly pure water. All columns
operate at close to 1 atm pressure.

Water Ethanol
Preconcentrator Azeotropic
Tower
(a)
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 270

(b) Material Ethanol

balance lines
1.0
Bottoms Composition
from Azeo-Column (B2)

Binary Feed to
Azeo-Column 0.8
( D 1) Distillate Composition from
Entrainer Recovery Column (D3)

Entrainer Recovery Column

0.6
Material Balance Line

Azeo-Column Material
Balance Line xlean
xN
0.4
xrich
D2, xD 2

0.2
Bottoms Composition 0
L2, xR2
Aqueous Feed (F1) from Entrainer
Recovery Column (B3)

0
0 0.2 0.4 0.6 0.8 1.0

Bottoms Composition Water Benzene

from Preconcentrator
( B 1) Overall Vapor Composition from Azeo-Columny(N)
Liquid in Equilibrium with Overhead Vapor from Azeo-Column
Distillate Composition from Entrainer Recovery ColumnxD(3)
Overall Feed Composition to Azeo-Column (D1 + D3)

Simple Distillation Boundaries (Approximate) Figure 9.36 Kubierschky three-column system.

Multiple Steady States

The occurrence of multiple steady states in chemical reactors has
been well recognized for at least 50 years. The most common
example is an adiabatic CSTR for which, in some cases, for the
same feed and reactor size, three possible products may be
obtained, two of which are stable and one unstable. The product
obtained in actual operation depends upon the startup procedure
for the reactor. Only in the past 35 years has the existence of
271 Chapter 9 Synthesis of Separation Trains

multiple steady states in distillation towers been shown by Figure 9.37b as a function of the bifurcation parameter. In the
calculations and verified by experimental data from tower range of bottoms flow rate from approximately 78 to 96 kmol∕hr,
operation. In particular, azeotropic distillation is especially three steady states exist, two stable and one unstable. For a
susceptible to multiple steady states. Disturbances during bottoms rate equal to the flow rate of ethanol in the feed (89
operation of an azeotropic tower can cause it to switch from one kmol∕hr), the best stable solution is an ethanol mole fraction of
steady state to another, as shown by Prokopakis and Seider 0.98; the inferior stable solution is only 0.89. Figure 9.37b shows
(1983). the computed points. In the continuation method, the results of
Methods for computing multiple steady states for homogeneous one point are used as the initial guess for obtaining an adjacent
and heterogeneous azeotropic distillation are presented in a point.
number of publications. Kovach and Seider (1987) computed, by
an arclength homotopy-continuation method, five steady states
for the ethanol–benzene–water distillation. Bekiaris et al. (1993,
1996, 2000) studied multiple steady states for ternary
homogeneous- and ternary heterogeneous-azeotropic distillation,
respectively. Using the distillate flow rate as the bifurcation
parameter, they found conditions of feed compositions and
distillation-region boundaries for which multiple steady states
can occur in columns operating at total reflux (infinite reflux
ratio) with an infinite number of equilibrium stages (referred to
as the 1–1 case). They showed that their results have relevant
Stage 1

Feed 1

Liquid
F = 1.962 kmol/hr Decanter
P = 1 atm
Mole Fractions : Stage 2 Organic Phase
Benzene = 1.0
D
Stage 3 Aqueous Phase
Feed 2 L = 508.369 kmol/hr
Stage 4
Liquid
F = 100 kmol/hr Stage 5
P = 1 atm
Mole Fractions :
Ethanol = 0.89
Water = 0.11

Stage 27 Partial
Reboiler

Stage 28

Bottoms Flow Rate, B

(a)
implications for columns operating at finite reflux ratios with a
finitenumberofstages.VadapalliandSeader(2001)usedASPEN
PLUS with an arclength continuation and bifurcation method to
compute all stable and unstable steady states for azeotropic
distillation under conditions of finite reflux ratio and finite
number of equilibrium stages. Specifications for their
heterogeneous azeotropic distillation example, involving the
separation of an ethanol–water mixture using benzene, are shown
in Figure 9.37a. The total feed rate to the column is 101.962
kmol∕hr. The desired bottoms product is pure ethanol. Using the
bottoms flow rate as the bifurcation parameter, computed results
for the mole fraction of ethanol in the bottoms are shown in
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 272

Although heterogeneous azeotropic distillation towers are

probably used more widely than their homogeneous
counterparts, care must be taken in their design and operation. In
addition to the possibility of multiple steady states, most
azeotropic distillation towers involve sharp fronts as the
temperatures and compositions shift abruptly from the vicinity of
one fixed point to the vicinity of another. Furthermore, in
heterogeneous distillations, sharp fronts often accompany the
interface between trays having one and two liquid phases as
well. Consequently, designers must select carefully the number
of trays and the reflux rates to prevent these fronts from exiting
the tower with an associated deterioration in the product quality.
Although these and other special properties of azeotropic towers
(e.g., maximum reflux rates above which the separation
deteriorates, and an insensitivity of the product compositions to
the number of trays) are complicating factors, they fortunately
are usually less important when synthesizing separation trains,
and consequently, they are not discussed further here. For a
review of the literature on this subject, see the article by
Widagdo and Seider (1996).

Pressure-Swing Distillation
In some situations, azeotropic points are sensitive to moderate
changes in pressure. When this is the case, pressure-swing
distillation can be used in place of azeotropic distillation to
permit the recovery of two nearly pure species that are separated
by a distillation boundary. This section introduces pressure-
swing distillation.
The effect of pressure on the temperature and composition of
the ethanol–water and ethanol–benzene azeotropes, two
minimum-boiling binary azeotropes, is shown in Figure 9.38. For
the first, as the pressure is decreased from 760 to 100 torr, the
mole fraction of ethanol increases from 0.894 to 0.980. Although
not

Bottoms Flow Rate (kmol/hr) (b) bifurcation diagram; branches I and

(b) III—stable, branch II—unstable.
shown, at a lower pressure, below 70 torr, the azeotrope
disappears entirely. The temperature changes are comparable for
the ethanol–benzene azeotrope, but the composition is far more
sensitive. Many other binary azeotropes are pressure sensitive, as
273 Chapter 9 Synthesis of Separation Trains

discussed by Knapp and Doherty (1992), who list 36 systems and consequently F1, are richer in A than the azeotropic
taken from the compilation of azeotropic data by Horsley (1973). composition at P1. Hence, the bottoms product, B1, that leaves
An example of pressure-swing distillation described by Van Column 1 is nearly pure A. Since the distillate, D1, which is
Winkle (1967) is provided for the mixture A − B having a slightly richer in A than the azeotropic composition, is less rich
minimum-boiling azeotrope with the T − x − y curves at two in A than the azeotropic composition at P2, when it is fed to
Column 2, the bottoms product, B2, is nearly pure B. Another
240 example is provided by Robinson and Gilliland (1950) for the
220 dehydration of ethanol, where the fresh-feed composition is less
200
rich in ethanol than the azeotrope. In this case, ethanol and water
are removed as bottoms products also, but nearly pure B (water)
180 is recovered from the first column, and A (ethanol) is
Temperature, °C

160

140 B2
P2
120
Ethanol–water
F2
100 D2
Ethanol–benzene
80
T
60 B1

40 P1
100 1,000 10,000 100,000
D1
System Pressure, torr

( a)

1.0
Pure B F F1 Pure A
0.9 Composition
Ethanol–water
0.8 ( a)
Mole fraction of ethanol

Ethanol–benzene
0.7

0.6 Pressure P 1 Pressure P 2

0.5
D1
D2
0.4
Ethanol–benzene
0.3

0.2
100 1,000 10,000 100,000
System Pressure, torr F F1 F2
1 2
(b)

Figure 9.38 Effect of pressure on azeotrope conditions:

(a) temperature of azeotrope; (b) composition of azeotrope.
pressures shown in Figure 9.39a. To take advantage of the
B1 B2
decrease in the composition of A as the pressure decreases from
P2 to P1, a sequence of two distillation towers is shown in Figure Pure A Pure B
9.39b. The total feed to Column 1, F 1, operating at the lower (b)

pressure, P1, is the sum of the fresh feed, F, whose composition

Figure 9.39 Pressure-swing distillation for the separation of a
is richer in A than the azeotrope and the distillate, D 2, whose minimum-boiling azeotrope: (a) T − x − y curves at pressures P1 and P2
composition is close to that of the azeotrope at P2 and which is for minimum-boiling azeotrope; (b) distillation sequence for minimum-
recycled from Column 2 to Column 1. The compositions of D 2, boiling azeotrope.
9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 274
275 Chapter 9 Synthesis of Separation Trains

Membranes, Adsorbers, and Auxiliary j

= − = … −; ′
(9.33a)
Separators Xj Yj j 1, ,C 1 j j
d�
When operating homogeneous azeotropic distillation towers, a
where = xj∕vj − xj′∕v
convenient vehicle for permitting the compositions to cross a
distillation boundary is to introduce a membrane separator,
adsorber, or other auxiliary separator. These are inserted either Xj − vTxj j′
before or after the condenser of the distillation column and serve vj′ (9.33b)
a similar role to the decanter in a heterogeneous azeotropic = yj∕vj − yj′∕v
distillation tower, with the products having their compositions in (9.33c)
adjacent distillation regions. Yj − vTyj j′
vj′

�= H (vj′ − vTyj )t (9.33d)

Reactive Distillation
Another important vehicle for crossing distillation boundaries is
through the introduction of chemical reaction(s) on the trays of a
distillation column. As discussed in Section 6.3, it is often
advantageous to combine reaction and distillation operations so
as to drive a reversible reaction(s) toward completion through the
recovery of its products in the vapor and liquid streams that leave
the trays. Somewhat less obvious, perhaps, is the effect the
reaction(s) can have on repositioning or eliminating the
distillation boundaries that otherwise complicate the recovery of
nearly pure species. For this reason, the discussion that follows
concentrates on the effect of a reaction on the residue curve
maps. Several constructs must be introduced, however, to
prepare for the main concepts.
For reactive systems, it is helpful to begin with a more
rigorous definition of an azeotrope, that is, a mixture whose
phases exhibit no changes in composition during vaporization or
condensation. On this basis, for vapor and liquid phases with
dxj∕dt = dyj∕dt = 0,j = 1,…,C in the presence of a homogeneous
chemical reaction ΣjvjAj = 0 at equilibrium, the conditions for a
reactive azeotrope can be derived (Barbosa and Doherty, 1988a)
such that yj − xj = d� = � j = 1,…,C (9.32)
vj − xjvT d�
where vj is the stoichiometric coefficient of species j, vT = Σjvj, � is
the extent of the reaction, � is the moles of vapor, and � is a
constant. Furthermore, it can be shown that the mass balances for
simple distillation in the presence of a chemical reaction can be
written in terms of transformed variables (Barbosa and Doherty,
1988b):
dX
≠
� vj′ − vTxj
Here, H is the molar liquid holdup in the still, and j′ denotes a
reference species. Clearly, Eq. (9.33a) corresponds to the mass
balances without chemical reaction [Eq. (9.20)]. By integration
of the latter equation for a nonreactive mixture of isobutene,
methanol, and methyl tertiary-butyl ether (MTBE), the residue
curve map in Figure 9.42a is obtained. There are two minimum-
boiling binary azeotropes and a distillation boundary that
separates two distillation regions.
When the chemical reaction is turned on and permitted to
equilibrate, Eq. (9.33a) is integrated and at long times,
Xj = Yj j = 1,…,C (9.34)
define the fixed point and are the conditions derived for a
reactive azeotrope [Eq. (9.32)]. At shorter times, reactive residue
curves are obtained, as shown in Figure 9.42d, where the effect
of the chemical reaction can be seen. It is clear that the residue
curves have been distorted significantly and pass through the
reactive azeotrope or so-called equilibrium tangent pinch.
Furthermore, the distillation boundary has been eliminated
completely. The reactive azeotrope of this mixture is shown
clearly in an X − Y diagram (Figure 9.43), which is similar to the
x − y diagram when reaction does not occur. Finally, through the
use of a kinetic model involving a well-stirred reactor, it is
possible to show the residue curves as a function of the residence
time (that is, the Damkohler number, Da). Figures 9.42b and
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 276

9.42c show how the residue curves change as the residence time
increases (Venimadhavan et al., 1994).

Separation Train Synthesis

Beginning with the need to separate a C-component mixture into
several products, alternative sequences of two-product
distillation towers are considered in this section. Although the
synthesis strategies are not as well defined for highly nonideal
and azeotropic mixtures, several steps are well recognized and
are described next. It should be mentioned that these strategies
continue to be developed, and variations are not uncommon.
Step 1: Identify the Azeotropes. Initially, it is very helpful to
obtain estimates of the temperature, pressure, and
composition of the binary, ternary,…, azeotropes
associated with the C-component mixture. For all of the
ternary submixtures, these can be determined, as
described above, by preparing residue curve or
distillation-line maps. When it is necessary to estimate
the quaternary and higher-component azeotropes as well
as the binary and ternary azeotropes, the methods of
Fidkowski et al. (1993) and Eckert and Kubicek (1997)
are recommended. When the C-component mixture is
the effluent from a chemical reactor, it may be helpful to
include the reacting chemicals, that is, to locate any
azeotropes involving these chemicals as well as the
existence of reactive azeotropes. This information may
show the potential for using reactive distillation
operations as a vehicle for crossing distillation
boundaries that complicate the recovery of nearly pure
species.
Step 2: Identify Alternative Separators. Given estimates for the
azeotropes, the alternatives for the separators involving
all C species are identified. These separate two species
277 Chapter 9 Synthesis of Separation Trains

Methanol Methanol
(128.5°C) (128.5°C)
1.0 1.0

0.8 0.8

0.6 0.6
Da = 0.0 Da = 0.12

0.4 0.4
113.7°C Saddle
60.2°C
0.2 0.2
Stable
Node
0.0 0.0
0.2 0.0 0.4 0.6 0.8 1.0
0.0 MTBE 0.2 0.4
0.6 0.8 1.0
Isobutene MTBE Isobutene
(122.9°C) (62.0°C) (122.9°C) (62.0°C)
(a) (b)

Methanol Methanol
(128.5°C) (128.5°C)
1.0 1.0

0.8 0.8

0.6 0.6
Da = 0.5 Da = 50.0

0.4 0.4

Kinetic Tangent Pinch

0.2 0.2 Equilibrium
Equilibrium
Tangent Pinch Tangent Pinch

0.0 0.0
0.2 0.0 0.4 0.6 0.0 0.2 0.4 0.6 0.8 1.0 0.8 1.0
Isobutene MTBE MTBE Isobutene
(122.9°C) (62.0°C) (122.9°C) (62.0°C)
(c) (d)

Figure 9.42 Residue curve maps for isobutene, methanol, and MTBE as a function of Da at 8 atm. (Reprinted from Venimadhavan et al., 1994).
Reproduced with permission of the American Institute of Chemical Engineers. Copyright © 1994 AIChE. All rights reserved.
that may or may not involve a binary azeotrope. When no binary
azeotrope is involved, a normal distillation tower may be
adequate unless the key components are close boiling. For close-
boiling binary pairs or binary pairs with an azeotrope separating
the desired products, the design of an extractive distillation
tower or an azeotropic distillation tower should be considered.
The former is preferred when a suitable solvent is available.
Step 3: Select the Entrainer. Probably the most difficult
decision in designing an azeotropic distillation tower
involves the selection of the entrainer. This is complicated
by the effect of the entrainer on the residue curves and
distillation lines that result. In this regard, the selection of
the entrainer for the separation of binary mixtures alone is a
large combinatorial problem complicated by the existence
of 113 types of residue curve maps involving different
combinations of lowand high-boiling binary and ternary
 9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures
azeotropes with associated distillation boundaries. This
classification,
which involves several indices that characterize the various kinds
of azeotropes and vertices, was prepared by Matsuyama and
Nishimura (1977) to aid in screening potential entrainers.
Thus, many factors need to be considered in selecting an
entrainer, factors that can have a significant impact on the
resulting separation train. Two of the more important guidelines
are the following:
(a) When designing homogeneous azeotropic distillation
towers, select an entrainer that does not introduce a
distillation boundary between the two species to be
separated.
(b) To cross a distillation boundary between two species to be
separated, select an entrainer that induces liquid-phase
splitting as in heterogeneous azeotropic distillation.
The effects of these and other guidelines must be considered
as each separator is designed and as the
1.0
0.8
0.6
Y1
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Methanol Isobutene
X1
Figure 9.43 Transformed compositions for isobutene, methanol, and
MTBE in chemical and phase equilibrium (Reprinted from Doherty and
Buzad, 1992). Reproduced with permission of the American Institute of
Chemical Engineers. Copyright © 1992 AIChE. All rights reserved.
separation sequence evolves. More recently, Peterson
and Partin (1997) showed that temperature sequences
involving the boiling points of the pure species and the
azeotrope temperatures can be used to effectively
categorize many kinds of residue curve maps. This
classification simplifies the search for an entrainer that
has a desirable residue curve map, for example, one that
does not involve a distillation boundary.
Step 4: Identify Feasible Distillate and Bottoms-product
Compositions. When positioning a two-product
separator, it is usually an objective to recover at least
278
one nearly pure species or at least to produce two
products that are easier to separate into the desired
products than the feed mixture. To accomplish this, it
helps to know the range of feasible distillate and
bottoms-product compositions. For a three-component
feed stream, the feed composition can be positioned on a
distillation-line map and the feasible compositions for
the distillate and bottoms product identified using the
methods described in the subsection on distillation-line
boundaries and feasible product compositions. For feed
mixtures containing four or more species (C > 3), a
common approach is to identify the three most important
species that are associated with the separator being
considered. Note, however, that the methods for
identifying the feasible compositions assume that they
are bounded by the distillation line at total reflux
through the feed composition. For azeotropic
distillations, however, it has been shown that the best
separations may not be achieved at total reflux.
Consequently, a procedure has been developed to locate
the bounds at finite reflux. This involves complex
graphics to construct the so-called pinch-point
trajectories, which are beyond the scope of
this presentation but are described in detail by Widagdo
and Seider (1996). Because the composition bounds at
finite reflux usually include the feasible region at total
reflux, the latter usually leads to conservative designs.
Having determined the bounds on the feasible
compositions, the first separator is positioned usually to
recover one nearly pure species. At this point in the
synthesis procedure, the separator can be completely
designed (to determine number of trays, reflux ratio,
installed and operating costs, etc.). Alternatively, the
design calculations can be delayed until a sequence of
separators is selected, with its product compositions
positioned. In this case, Steps 2–4 are repeated for the
mixture in the other product stream. Initially, the
simplest separators are considered, that is, ordinary
distillation, extractive distillation, and homogeneous
azeotropic distillation. However, when distillation
boundaries are encountered and cannot be eliminated
through the choice of a suitable entrainer, more
complex separators are considered, such as
heterogeneous azeotropic distillation, pressure-swing
distillation, the addition of membranes, adsorption,
auxiliary separators, and reactive distillation. Normally,
a sequence is synthesized involving many two-product
separators without chemical reaction. Subsequently,
after the separators are designed completely, steps are
taken to carry out task integration as described in
Section 2.3. This involves the combination of two or
more separators and seeking opportunities to combine
the reaction and separation steps in reactive distillation
towers. As an example, Siirola (1995) describes the
development of a process for the manufacture of methyl
279 Chapter 9 Synthesis of Separation Trains
acetate and the dehydration of acetic acid. Initially, a
sequence was synthesized involving a reactor, an
extractor, a decanter, and eight distillation columns
incorporating two mass separating agents. The
flowsheet was reduced subsequently to four columns by
using evolutionary strategies and task integration before
being reduced finally to just two columns, one
involving reactive distillation.
The next example illustrates the usage of the above procedure,
as applied to the separation of a binary mixture exhibiting a
minimum-boiling azeotrope and focuses of the selection of a
suitable entrainer and column operating pressures.
EXAMPLE 9.6 Entrainer Selection and Operating Pressure
The most economical process is sought for the separation of a feed
stream containing 30 mol% of A and 70 mol% of B into highly pure
product streams (at least 99 mol%). The process should utilize one or
more 2-product distillation columns, with each column operating
either at 2 or 10 bar. The process may include the usage of one of two
entrainers, C and D. Data for the ternary systems A-B-C and A-B-D
are provided in Table 9.7. As seen, both feature two homogeneous
azeotropes. Your solution should provide justification for your choice
of entrainer, as well as the material balance lines for the columns that
comprise the separation system you devise, and a PFD for the process.
9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 280
281 Chapter 9 Synthesis of Separation Trains

As illustrated throughout this section, process simulators have and small quantities of isobutene, methanol, and acetone, which can
extensive facilities for preparing phase equilibrium diagrams (T − be disregarded. A separation sequence is to be synthesized to x −
y, P − x − y, x − y,…), residue curve maps, and bimodal curves for produce 99.6 mol% pure DTBP containing negligible water. It may
ternary systems. In addition, related but independent packages be difficult to separate TBA and water. Therefore, rather than
have been developed for the synthesis and evaluation of recovering and recycling the unreacted TBA, the conversion of TBA
distillation trains involving azeotropic mixtures. These include to isobutene and water in the separation sequence should be
SPLITTM, developed by Aspen Technology, Inc., in the early considered. In the catalytic reactor, the TBA dehydrates to isobutene,
which is the actual molecule
2000s, and DISTILTM developed by M. F. Doherty and M. F.
thatreactswithTBHPtoformDTBP.Thus,isobutene,insteadofTBA, can
Malone at the University of Massachusetts (as incorporated into be recycled to the catalytic reactor.
HYSYS and UniSim®Design). The following is an example of the
capabilities afforded by such packages.
SOLUTION
A residue curve map at 15 psia, prepared using ASPEN PLUS (with
EXAMPLE 9.7 Manufacture of Di-Tertiary-Butyl the NRTL option set and proprietary interaction coefficients), is
Peroxide displayed in Figure 9.46a. There are three minimum-boiling binary
azeotropes:
This example involves the manufacture of 100 million pounds per
year of di-tertiary-butyl peroxide (DTBP) by the catalytic reaction of
T, ∘F
tertiary-butyl hydroperoxide (TBHP) with excess tertiary-butyl
alcohol (TBA) at 170∘F and 15 psia according to the reaction DTBP–TBA 177 xTBA = 0.82
TBA–H2O 176 xH2O = 0.38
CH3
CH3 H2O–DTBP 188 xDTBP = 0.47
C OH CH 3 C OOH CH 3 C OOC CH 3
CH3
CH3
and the boiling points of the pure species are 181, 212, and 232∘F,
CH3 H2O
for TBA, H2O, and DTBP, respectively. In addition, there is a
CH3 CH3 CH3 CH3 minimum-boiling ternary azeotrope at xTBA = 0.44, xH2O = 0.33, and
Tertiary-butyl Tertiary-butyl DTBP xDTBP = 0.23, and 174∘F. Consequently, there are three distinct
Alcohol Hydroperoxide distillation regions with the feed composition in a region that does
(TBA) (TBHP)
not include the product vertex for DTBP.
To cross the distillation boundaries, it is possible to take
Assume that the reactor effluent stream contains advantage of the partial miscibility of the DTBP–H2O system as well
as the disappearance of the ternary azeotrope at 250 psia as
illustrated in Figure 9.46b. One possible design is shown in Figure
Ibmol/hr Mole Fraction 9.47, where the reactor effluent is in stream S-107. Column D-102
TBA 72.1 0.272 forms a distillate in stream S-108 whose composition is very close to
H2O 105.6 0.398 the ternary azeotrope and a bottoms product in stream S-109, as
DTBP 87.7 0.330 shown on the ternary diagram in Figure 9.48a. The latter stream,
1.000 containing less than 5 mol% TBA, is split into two liquid phases in
the decanter. The aqueous phase in stream S-111 enters the
distillation tower, D-103, which forms nearly pure water in the
bottoms product, stream S-113. The distillate
9.5 Sequencing of Operations for the Separation of Nonideal Liquid Mixtures 282
283 Chapter 9 Synthesis of Separation Trains

pressure. Note, however, that the material balance line for the tower,
D-102, would preferably be positioned farther away from the
distillation boundary to allow for inaccuracies in its calculation.
Since this design was completed, the potential for DTBP to
decompose explosively at temperatures above 255∘F was brought to
our attention. At 250 psia, DTBP is present in the bottoms product of
tower D-104 at 480.2∘F. Given this crucial safety concern, a design
team would seek clear experimental evidence. If positive, lower
pressures, with corresponding lower temperatures, would be explored,
recognizing that the distillation boundaries are displaced less at lower
pressures.
For additional details of this process design, see the design report
by Lee et al. (1995). Also, see Problem A-IIS.1.10 in the
Supplement_to_Appendix_II.pdf (in the PDF Files folder, which can
be downloaded from www.seas.upenn.edu/∼dlewin/design
projects.html) for the design problem statement that led to this design.
9.6 SEPARATION SYSTEMS FOR GAS
MIXTURES
Sections 9.4 and 9.5 deal primarily with the synthesis of
separation trains for liquid–mixture feeds. The primary
separation techniques are ordinary and enhanced distillation. If
the feed consists of a vapor mixture in equilibrium with a liquid
mixture, the same techniques and synthesis procedures can
oftenbe employed. However, as discussed in Section 9.1, if the
feed is a gas mixture and a wide gap in volatility exists between
two groups of chemicals in the mixture, it is often preferable to
Purge
Vapor Recycle Vapor Separation
System
Partial
Condenser
Reactor Phase
Feed(s) System Split
Liquid
Reactor
Effluent
Liquid Separation
System
Liquid Recycle
Purge
partially condense the mixture, separate the phases, and send the
liquid and gas phases to separate separation systems as discussed
by Douglas (1988) and shown in Figure 9.49. Note that if a
liquid phase is produced in the gas separation system, it is routed
to the liquid separation system and vice versa.
In some cases, it has been found economical to use distillation
to separate a gas mixture, with the large-scale separation of air
by cryogenic distillation into nitrogen and oxygen being the most
common example. However, the separation by distillation of
many other gas mixtures, such as hydrogen from methane or
hydrogen from nitrogen, is not practical because of the high cost
of partially condensing the overhead vapor to obtain reflux.
Instead, other separation methods, such as absorption,
284
9.6 Separation Systems for Gas Mixtures
adsorption, or membrane permeation, are employed. In just the
past 35 years, continuous adsorption and membrane processes
have been developed for the separation of air that economically
rival the cryogenic distillation process at low to moderate
production levels.
Barnicki and Fair (1992) consider in detail the selection and
sequencing of equipment for the separation of gas mixtures.
Whereas ordinary distillation is the dominant method for the
separation of liquid mixtures, no method is dominant for gas
mixtures. The separation of gas mixtures is further complicated
by the fact that whereas most liquid mixtures are separated into
nearly pure components, the separation of gas mixtures falls into
the following three categories: (1) sharp splits to produce nearly
pure products, (2) enrichment to increase the concentration(s) of
one or more species, for example, oxygen and nitrogen
enrichment, and (3) purification to remove one or more low-
concentration impurities. The first category is often referred to as
bulk separation, the purpose of which is to produce high-purity
products at high recovery. Separations in this category can be
difficult to achieve for gas mixtures. The best choices are
cryogenic distillation, absorption, and adsorption. By contrast,
the second category achieves neither high purity nor high
recovery and is ideally suited for any of the common separation
methods for gas mixtures including membrane separation by gas
permeation. To produce high-purity products by purification,
adsorption and absorption with chemical reaction are preferred.
The synthesis of a separation train for a
Vapor Product
gas mixture can be carried out by first
determining the feasible separation methods,
which depend on the separation categories
and the separation factors, and then
designing and costing systems involving
these methods to determine the optimal train.
Vapor The design of equipment for absorption,
adsorption, distillation, and membrane
separations is covered by Seader et al.
(2016). Besides the separation category and
separation factor, the production scale of the
process is a major factor in determining the
optimal train because economies of scale are
most pronounced for cryogenic distillation
and absorption, and least pronounced for adsorption and
membrane separations. For example, for the separation of air into
nitrogen- and oxygen-enriched products, membrane separations
are most economical at low production rates, adsorption at
moderate rates, and cryogenic distillation at high rates.
285 Chapter 9 Synthesis of Separation Trains
Figure 9.49 Process with vapor and liquid significantly in molecular size and/or shape, as characterized by
Liquid Product(s) separation systems. (Source: Modified and reprinted the kinetic diameter, zeolites and carbon molecular-sieve
with permission from Douglas, 1988). adsorbents can be used to advantage because of the strong
selectivity achieved by molecular sieving. These adsorbents have
Membrane Separation by Gas Permeation very narrow pore-size distributions that prevent entry into the
pore structure of molecules with a kinetic diameter greater than
In gas permeation, the gas mixture is compressed to a high the nearly uniform pore aperture. Zeolites are readily available
pressure and brought into contact with a very thin membrane to with nominal apertures in angstroms of 3, 4, 5, 8, and 10. Thus,
produce two products: (1) a permeate that passes through the for example, consider a gas mixture containing the following
membrane and is discharged at a low pressure and (2) a retentate components with corresponding kinetic diameters in angstroms
that does not pass through the membrane and is maintained at in parentheses: nitrogen (<3), carbon dioxide (>3 and <4), and
close to the high pressure of the feed. The separation factor
benzene (>7 and <8). The zeolite with a 3-Å aperture could
defined by Eqs. (9.2) and (9.8) can be applied to gas permeation
selectively adsorb the nitrogen, leaving a mixture of carbon
when the retentate-side pressure is much greater than the
dioxide and benzene that could be separated with a zeolite of 4-Å
permeate-side pressure, if y is the mole fraction in the permeate
aperture. Barnicki (1991) gives methods for estimating kinetic
and x is the mole fraction in the retentate. The relative volatility
diameters. In effect, the separation factor for a properly selected
is replaced by the ratio of the membrane permeabilities for the
sieving-type adsorbent is infinity.
two key components of the feed–gas mixture, sometimes called
Adsorbents made of activated alumina, activated carbon, and
the permselectivity. Most commercial membranes for gas
silica gel separate by differences in adsorption equilibria, which
permeation are nonporous (dense) amorphous or semicrystalline
must be determined by experiment. Equilibrium-limited
polymers. To pass through such polymers, the gas molecules first
adsorption can be applied to all three categories of separation,
dissolve in the polymer and then pass through it by diffusion.
but is usually not a favored method when the components to be
Thus, the permeability depends on both solubility and diffusivity
selectively adsorbed constitute an appreciable fraction of the feed
in the particular polymer at the conditions of temperature and
gas. Conversely, equilibrium-limited adsorption is ideal for the
pressure. The permeability is the product of the solubility and
removal of small quantities of selectively adsorbed impurities. At
diffusivity. Permeabilities are best determined by laboratory
a given temperature, the equilibrium loading of a given
measurements. However, a predictive method given by Barnicki
component, in mass of adsorbate per unit mass of adsorbent,
(1991) for a number of glassy and rubbery polymers, which
depends on the component’s partial pressure and to a lesser
depends on species van der Waals volume and critical
extent on the partial pressures of the other components. For
temperature, can be applied in the absence of data. In general,
equilibrium-limited adsorption to be feasible, Barnicki and Fair
gas permeation is commercially feasible when the ratio of
(1992) suggest that the ratio of equilibrium loadings of the two
permeabilities (permselectivity) for the two components is
key components be used as a separation factor. This ratio should
greater than 15. However, some processes that require only
be based on the partial pressures in the feed gas. A ratio of at
rough enrichments use membranes having permselectivities of
least 2, and preferably much higher, makes equilibrium
only 5. Commercial applications include the recovery of carbon
adsorption quite favorable. However, two other conditions must
dioxide from hydrocarbons, the adjustment of the hydrogen-to-
also be met: (1) the more highly adsorbed component should
carbon monoxide ratio in synthesis gas, the recovery of
have a concentration in the feed of less than 10 mol% and (2) for
hydrocarbons from hydrogen, and the separation of air into
an adsorption time of 2 hr, the required bed height should not
nitrogen- and oxygen-enriched streams.
exceed 20 ft. Equilibrium-limited adsorption is usually the best
alternative for the removal of water and organic chemicals from
Adsorption mixtures with light gases and should also be considered for
enrichment applications.
Adsorption differs from the other techniques in that it is a cyclic
operation with adsorption and desorption steps. However,
adsorption is a very versatile separation technique. To be Absorption
economical, the adsorbent must be regenerable. This requirement
Absorption of components of a gas mixture into a solvent may
precludes the processing of gas mixtures that contain (1) high-
take place by physical or chemical means. When no chemical
boiling organic compounds because they are preferentially
reaction between the solute and absorbent occurs (physical
adsorbed and are difficult to remove during the regeneration part
absorption), the separation factor is given by Eq. (9.2). Thus, if
of the cycle, (2) lower-boiling organic compounds that may
component 1 is to be selectively absorbed, a small value of SF is
polymerize on the adsorbent surface, and (3) highly acidic or
desired. Alternatively, Barnicki and Fair (1992) suggest that
basic compounds that may react with the adsorbent surface. In
consideration of physical absorption should be based on a
some cases, such compounds can be removed from the gas
selectivity, S1,2, defined as the ratio of liquid-phase mole
mixture by guard beds or other methods prior to entry into the
fractions of the two key components in the gas mixture. This
adsorption system.
selectivity can be estimated from the partial pressures of the two
Selectivity in adsorption is controlled by (1) molecular sieving components in the gas feed and their K-values for the given
or (2) adsorption equilibrium. When components differ
 286
solvent. For components whose critical temperatures are greater S1,2 = x2 � (9.35)
than the system temperature,
�
where ∞ is the liquid-phase activity coefficient at infinite
x1 = �2∞p1Ps2
dilution, p is partial pressure, and Ps is vapor pressure. For
components whose critical temperatures are less than the system
9.7 Separation Systems for Solid-Fluid Mixtures

temperature, the selectivity can be estimated from Henry’s law solution crystallization (solutes with high melting points that
constants: are crystallized from a solvent), melt crystallization (crystalx1 = H2p1 (9.36) lization from a mixture of components with low
to moderate

, x Hp
S1 2 = 2 1 2 melting points), precipitation (rapid crystallization from a sol-
where H = yP∕x. For enrichment, the selectivity should be 3 or vent of nearly insoluble compounds that are usually formed greater;
for a sharp separation, 4 or greater. The number of the- by a chemical reaction), desublimation, and/or drying, as well oretical stages
should be at least 5. For the removal of readily as the phase-separation operations of filtration, centrifugation, soluble organic
compounds from light gases, Douglas (1988) rec- and cyclone separation. In addition, because specifications for ommends the use of
10 theoretical stages and a solvent molar flow solid products may also include a particle size distribution, rate, L, based on an
absorption factor, A, for solute of 1.4, where size-increase and size-reduction operations may also be necessary. If particle shape is
also a product specification, certain
L
A= (9.37) types of crystallizers and/or dryers may be dictated. Even when
KV
the final product is not a solid, solid–liquid or solid–gas separa-
V
with = gas molar flow rate. When the partial pressure in the tion operations may be involved. For example, liquid mixtures gas feed
of the component to be absorbed is very small and a of meta- and para-xylene cannot be separated by distillation high percentage of
it is to be removed, physical absorption may because their normal boiling points differ by only 0.8∘C. Instead, not be favorable.
Instead, particularly if the solute is an acid or because their melting points differ by 64∘C, they are separated base, chemical
absorption may be attractive. industrially by melt crystallization. Nevertheless, the final products are liquids. Another example is
phthalic anhydride, which,
Partial Condensation and Cryogenic Distillation although a solid at room temperature, is usually shipped in the molten state.
It is produced by the air oxidation of napthalene The previously discussed separation techniques for gas mix- or ortho-xylene. The
separation of the anhydride from the tures all involve a mass separating agent. Alternatively, thermal reactor effluent gas mixture is
accomplished by desublimameans are employed with partial condensation and cryogenic tion followed by distillation to remove
impurities and produce distillation. Barnicki and Fair (1992) recommend that partial a melt. condensation be considered for
enrichment when the rela-
A common flowsheet for the separation section of a process for
tive volatility between the key components is at least 7. For
large-scale (>10 − 20 tons∕day of product gas) enrichment and themanufactureofinorganicsaltcrystalsfromtheiraqueoussolu-tion is
shown in Figure 9.50. If the feed is aqueous MgSO 4, a typisharp separations, cryogenic distillation is feasible when the cal process
proceeds as follows. A 10 wt% sulfate feed is concenrelative volatility between the key components is greater than 2. trated, without
crystallization, to 30 wt% in a double-effect evapHowever, if the feed gas contains components, such as carbon oration system. The
concentrate is mixed with recycled mother
dioxide and water that can freeze at the distillation temperatures, liquors from the hydroclone and centrifuge before being fed to
those components must be removed first.
an evaporative vacuum crystallizer, which produces, by solution
crystallization, a magma of 35 wt% crystals of MgSO4⋅7H2O, the
9.7 SEPARATION SYSTEMS FOR SOLID-FLUID stable hydrate at the temperature in the crystallizer. The magma
MIXTURES is thickened to 50 wt% crystals in a hydroclone and then sent to
287 Chapter 9 Synthesis of Separation Trains
The final product from many industrial chemical processes a centrifuge, which discharges a cake containing 35 wt % moisis a solid
material. This is especially true for inorganic com- ture. The cake is dried to 2 wt% moisture in a direct-heat rotary pounds, but is also
common for a number of moderate- to dryer. Approximately 99 wt% of the dried crystals are retained on high-molecular-weight
organic compounds. Such processes a 100-mesh screen and 30 wt% are retained on a 20-mesh screen. involve the separation
operations of leaching, evaporation, The crystals are bagged for shipment. Rossiter (1986) presents
Vapor Vapor

Aqueous Overflow
Solution Two-Effect Vacuum
Evaporation Evaporative
System Crystallizer
Magma Hydroclone

Underflow

Mother Centrifugal
Liquor Filter Vapor

Cake
Rotary
Dryer
Dried Figure 9.50 Process for producing
Crystals crystals.
inorganic salt