1

VAPOR/LIQUID SEPARATION PROCESSES: DISTILLATION

1. INTRODUCTION

Distillation is a separation process in which a liquid mixture containing components of

different boiling points is separated by partial evaporation. As in absorption, there will be contact
between a gas and a liquid, but in this case the gas will be close to its dew point, and for this it is
referred to as a "vapor". One new element in this type of separation process is that to partially
vaporize the mixture, we will need to provide energy in the form of heat.
As an industrial process, distillation is possibly the most important of all separation processes
due to its widespread use and multiplicity of applications. In a report (Chemical Engineering
Progress, 1998), it was estimated that there were 40,000 active distillation columns in the US that
represent as a whole $8 billion in capital investment, with an average cost of $200,000 per
column. In terms of investment and volume of production, one of the most important
applications of distillation is in petroleum refining. One of the first steps to which crude oil is
subjected in its processing is a distillation operation, in which the crude is separated into various
fractions of different boiling points.
To illustrate the basic principles underlying distillation, consider the following example. We
start with a liquid that is a mixture of water and methanol. At 1 atm, the boiling points of these
two compounds are: water: 100°C, methanol: 65°C. If we heat up the liquid to the point at which
we produce a vapor/liquid mixture, we expect the vapor to be rich in methanol and the liquid to
be rich in water. If we set up this operation as a continuous process taking place in a tank/heat
exchanger, we will have what is illustrated in Figure 1. Note that the relative amounts of vapor
and liquid produced will depend on the amount of heat provided to the system. If the heat flow is
large enough that all the liquid evaporates then there will be no separation since the vapor
produced will have the same composition as the liquid fed to the system. Therefore, separation
can be accomplished only through partial vaporization of the liquid.

Figure 1. Boiling a water/methanol mixture will produce a vapor that is rich in methanol and
a liquid that is rich in water. The process is achieved by providing heat (Q) to the mixture. This
separation process is called Binary Distillation.

The thermodynamics of the type of system depicted in Figure 1 are different from those used
in absorption/stripping. In what follows we will review concepts related to the thermodynamic
description of vapor/liquid equilibrium.
 2

Vapor/Liquid Equilibrium
Consider a binary mixture subjected to conditions at which a vapor and a liquid phase are at
equilibrium. We will use the methanol/water system as an example. Both components will be
distributed between the two phases, as depicted in Figure 2. To determine the number of degrees
of freedom, we apply the phase rule:

F=C–P+2 (1)

Figure 2. Example of vapor-liquid equilibrium involving two components. Both components

will be distributed between the two phases but at different mole fractions. Since methanol is
more volatile (it has a lower boiling point), one would expect the vapor phase to be richer in
methanol than the liquid phase (y>x). According to the phase rule, this system has only two
degrees of freedom.

The system has two degrees of freedom. This means that, if two thermodynamic variables are
specified, then the thermodynamic state of the system is unique and completely defined. For
example, if we specify P and T, the compositions of both phases (i.e., x and y) are automatically
set. Normally, the pressure, P, is set, and one additional variable is needed to specify the
equilibrium state.
Vapor/liquid equilibrium in binary systems is represented by means of a phase diagram in
which temperature is plotted as a function of mole fraction for a fixed total pressure. An example
of such representation is given in Figure 3. Three regions are distinguished in the diagram: a
region in which the mixture is a liquid, another region in which the mixture is a gas (vapor), and
a region in which vapor and liquid coexist at equilibrium. In the vapor/liquid equilibrium region,
setting the temperature (e.g. T=T0), since the pressure is already set (P), means that there will be
a liquid with a specific mole fraction of component 1 (x0) at equilibrium with a vapor that has a
specific mole fraction of component 1 (y0).
To better understand the representation in Figure 3, consider the process sketched in Figure
4. We start with a sub-cooled liquid (i.e., a liquid at a temperature below its boiling point) that
has a mole fraction of component 1 equal to xF (point A in Figure 3). We then add heat to this
liquid at constant pressure until we reach an equilibrium state at which a first bubble of vapor
forms (point B). This is the boiling point and the liquid is said to be "saturated". The heat that
has been added has increased the temperature of the liquid. If we add more heat to the liquid,
vaporization will occur. The thermal energy added will be used to vaporize part of the liquid as
well as to continue increasing the system temperature. If we stop at point C (Figure 3), there will
be a vapor/liquid mixture at equilibrium. At that point, the liquid (which will be saturated) will
 3

have a mole fraction of light component equal to xC whereas the gas will have a mole fraction of
light component equal to yC (see figure). Further heat addition from point C will lead to a
complete vaporization of the mixture until the last drop of liquid evaporated (point D). At this
point, we have a saturated vapor and the mixture is at its dew point (meaning that any extraction
of heat will result in condensation). Finally, heat added beyond point D will result in temperature
increases. The final stage (point E) will be a super-heated vapor.

Figure 3. Typical T-x-y diagram for a binary mixture at a fixed pressure P. See text for a
description. The two components are designated so that component 1 is the most volatile (i.e., its
boiling point is lower than that of component 2, Tb1<Tb2); in common notation, component 1 will
be known as the "light" component, whereas component 2 will be the "heavy" component (note
that this notation has nothing to do with density).

Since all steps in the process shown in Figure 4 are equilibrium stages, the process is
reversible in the sense that, if we start at point E and remove heat from the system, we will
successively achieve points D, C, B and A.
The vapor/liquid equilibrium data in Figure 3 can be represented also in a y vs. x diagram:
for temperatures between Tb1 and Tb2, the values of x and y of the liquid and vapor can be found
from the plot (such as x0 and y0). Note that, for T=Tb1, y=x=1, and, for T=Tb2, y=x=0. Plotting
the equilibrium values of y vs. x leads to a representation such as that illustrated in Figure 5.
Note that, since x and y are mole fractions of the light component, y>x and the equilibrium line
is above the y=x line.
 4

Figure 4. Successive equilibrium steps in the heating of a binary mixture at constant pressure
from a sub-cooled liquid state to a super-heated vapor state. Each step corresponds to an
equilibrium state. The process is represented in the phase diagram shown in Figure 3. See text for
details.
 5

Phase diagrams (such as Figure 3) and y-x equilibrium diagrams (such as Figure 5) can be
generated by means of independent measurements. These diagrams will be a fundamental tool in
the design of distillation processes. Sometimes, equilibrium diagrams are generated by
thermodynamic models. A particularly simple model corresponds to the case in which the liquid
is an ideal solution and the vapor is an ideal gas. In this case, the partial pressure of a component
in the vapor phase equals its mole fraction in the liquid times its vapor pressure at the
temperature of the equilibrium state. This relation applies, not only to binary systems, but also to
multicomponent systems, and it is called Raoult's law. For the binary system considered here,
Raoult's law for each of the two components can be expressed mathematically as follows,

Figure 5. Vapor/liquid equilibrium line in a y vs. x diagram. Each point corresponds to a

different temperature.

P1 = Pv1 (T ) x1 = Pv1x (2)

P2 = Pv 2 (T) x 2 = Pv 2 (1 − x ) (3)

where Pv1 and Pv2 are the vapor pressures of components 1 and 2, respectively. Recall that vapor
pressures are properties of pure substances that can be obtained from physical property data
sources (such as Perry, R.H. and D.W. Green, Chemical Engineers' Handbook, 7th edition,
McGraw-Hill, 1991).
 6

Since y1=y=P1/P and y2=1-y=P2/P, equations (2) and (3) can be written as

Pv1
y= x (4)
P

Pv 2
1− y = (1 − x ) (5)
P

These equations can be solved for x and y to obtain

P − Pv 2
x= (6)
Pv1 − Pv 2

P − Pv 2 Pv1
y= (7)
Pv1 − Pv 2 P

These equations can be used to generate either the T-x-y or the y-x diagram for the mixture. To
do this, temperatures in the range Tb1<T<Tb2 are selected. The vapor pressures are evaluated at
these temperatures, and then x and y are calculated from equations (6) and (7).
Systems that do not follow Raoult's law might have complex equilibrium relations whose
description is outside the scope of this course. However, in some cases, equilibrium relations are
expressed as

y = K1 (T ) x , component 1 (8)

1 − y = K 2 (T)(1 − x ) , component 2 (9)

where K1 and K2 are called "partition coefficients" or "distribution coefficients", and they are, for
a certain class of substances, only functions of temperature. Note that Raoult's law can be
considered a particular case of these representation with (compare with equations 4 and 5)

Pvi
Ki = , i=1,2, if Raoult's law applies (10)
P

The ratio between distribution coefficients is called relative volatility,

K1
α= (11)
K2

Note that, if Raoult's law is valid, then α would be the ratio of vapor pressures. Note also that,
since component 1 is the more volatile component, α>1.
Typically, distribution coefficients will experience appreciable variations in the whole
vapor/liquid equilibrium range of temperatures. However, their ratio (i.e., the relative volatility)
 7

tends to have a lower sensitivity on temperature. In fact, for certain mixtures (e.g. substances
belonging to homologous chemical series like, for example, alkanes, and with close molecular
weights) α can be approximated as being a constant. In this particular case, the y vs. x
equilibrium relation can be generated by using α as the only parameter: dividing equations (8)
and (9) and using equation (11) we get:

y x
=α
1− y 1− x

which implies

αx
y= (12)
1 + (α − 1) x

2. SINGLE-STAGE BINARY DISTILLATION (FLASH DISTILLATION)

The simplest scheme to perform a distillation process in continuous mode (see Figure 1) is
the steady-state, single-stage equilibrium process. This process is usually referred to as "flash
distillation". A feed stream with a molar flow rate F and a mole fraction of light component xF is
heated or expanded (or both) to form a vapor/liquid mixture that is allowed to reach equilibrium
and then separated. A scheme of a flash distillation unit is shown in Figure 6.

Figure 6. A flash distillation process. The feed (1), which is typically a liquid, is heated (2)
and/or its pressure is reduced in an expansion valve (3) to produce a vapor/liquid mixture that is
then separated (4). The vapor produced (distillate) is rich in light component whereas the liquid
produced (bottoms) is rich in heavy component.

To quantify a flash distillation process we start by performing mole balances. A total mole
balance at steady state yields

F= D+B (13)

where D and B are the molar flow rates of distillate and bottoms product, respectively. To
quantify the relative rates of products we will use a parameter that represents the fraction of the
feed that is vaporized in the flash, defined as
 8

D
f= (14)
F

Dividing equation (13) by F we can show that

B
= 1− f (15)
F

Typical design problems for flash distillation can be formulated as follows:

INFORMATION AVAILABLE:

F, xF, PF, TF  what we want to treat

P  operating pressure
Equilibrium  from Thermodynamics, with knowledge of P
yD or T or f  specification or goal of process

DESIGN PARAMETERS:

yD or T or f  the two parameters that are not set

Q  heat transfer rate required

To solve any of these design problems we start with a mole balance in light component:

Fx F = Dy D + Bx B (16)

Dividing both sides of this equation by F, and using equations (14) and (15) yields

x F = fy D + (1 − f ) x B (17)

For any of the three possibilities in the design problems formulated above, this equation will
always have at least two unknowns. But we also know that xB and yD are at equilibrium (see
Figure 7).
Let us consider the three different cases formulated above:

(1) If yD is known:
Find xB from either the T-x-y or y-x diagrams (see Figure 7)
Find T from the T-x-y diagram
Calculate f from equation (17)

(2) If T is known:
Find xB and yD from the T-x-y diagram
Calculate f from equation (17)

(3) If f is known:
Equation (17) and the equilibrium line in the y-x diagram must be solved
 9

simultaneously to find xB and yD

Find T from the T-x-y diagram

Figure 7. The distillate and bottoms product of a flash distillation operation are at equilibrium,
so that they can be related through the use of a T-x-y phase diagram, or a y-x equilibrium line.

Case 3 can be solved graphically. Consider equation (17), replacing yD by a generic y, and xB
by a generic x, and then solving for y, we then find the following linear function

1− f x
y=− x+ F (18)
f f

This is a straight line in the y-x diagram that has the following characteristics:

(1) Its slope is -(1-f)/f.

(2) For x=xF, then y=xF (i.e., the line intersects the y=x line at xF).
(3) For x=xB, then y=yD.
 10

For case 3, the straight line given by equation (18) can be plotted, and its intersection with
the equilibrium line yields the point (xB,yD). This graphical procedure is illustrated in Figure
8.The slope of the line given by equation (18) is always negative, since 0<f<1. Figure 8 shows
the extreme cases: if f→0 (vertical line), then there is almost no vapor produced and xB→xF. In
this case, the small amount of vapor produced will have a mole fraction of light component equal
to yM, which represents the maximum y that can be achieved in the flash. On the other hand, as
f→1 (horizontal line), then there is almost no liquid produced and yD→xF. In this case, the small
amount of liquid produced will have a mole fraction of light component equal to xm, which
represents the minimum x that can be achieved in the flash.
The final design parameter in the problem considered is the heat transfer rate required to
perform the separation. This parameter can be found by performing a thermal energy balance in
the flash unit. The principle of conservation of energy for the flash at steady state, and neglecting
changes in kinetic energy, can be written as

thermal energy per unit time rate of heat sup plied thermal energy per unit time 
 + = 
entering with the feed  to the system  leaving with distillate and bottoms

Figure 8. Graphical solution to the Case 3 design: knowing f and xF, equation 18 is plotted. Its
intersection with the equilibrium line yields the distillate and bottoms product compositions. The
graph also shows the extreme cases of no vaporization (f=0), and complete vaporization (f=1).

The thermal energy of feed and products is the total enthalpy of the flows, since they are
 11

moving fluids. We will use molar enthalpies, H (enthalpy per unit mole of mixture) to quantify
these interactions. The thermal energy balance can be expressed as

FH F + Q = DH D + BH B (19)

where Q is the heat rate supplied to the flash. Using the total mole balance, equation (13), we can
rewrite equation (19) as follows,

Q = D(H D − H F ) + B(H B − H F ) (20)

The calculation of the enthalpy differences in this equation will depend on the thermodynamic
condition of the feed (i.e., if the feed is a liquid, a vapor, or a vapor-liquid mixture). Note that the
distillate is a saturated vapor at T, and the bottoms product is a saturated liquid at T. Therefore,
we have:
H D − H B = ∆H vap (21)

where ∆Hvap is the molar enthalpy of vaporization of the mixture.

Generally, the feed is a liquid and the temperature difference between the feed and the flash
is low enough that the molar heat capacity of the liquid, CpL is approximately independent of
temperature and composition. Since the bottoms product is also a liquid, the enthalpy difference
between bottoms and feed can be expressed by

H B − H F = C pL (T − TF ) (22)

The enthalpy difference between distillate and feed can be determined as follows:

HD − HF = HD − HB + HB − HF

which, using equations (21) and (22), becomes

H D − H F = ∆H vap + C pL (T − TF ) (23)

Substituting equations (22) and (23) into equation (20) yields

Q = D∆H vap + (D + B)C pL (T − TF )

Using equation (13), dividing by F and then using equation (14) leads to

Q
= f∆H vap + C pL (T − TF ) (24)
F

This equation gives the amount of heat to be supplied to the flash per unit mole of feed (Q/F).
 12

3. MULTISTAGE BINARY DISTILLATION

The degree of separation that can be achieved in a flash unit is limited by the equilibrium
relation (see Figure 8 and its discussion). If better separations are required, a multistage process
can be used. In what follows we will explore the conformation of a countercurrent, multistage
distillation process. We start with the original flash process, with a new notation, shown in
Figure 9. To increase the degree of separation, we can take L2 to be, not the feed to the process,
but the bottoms product of a second flash unit, as shown in Figure 10. Note that the final liquid
product in this scheme has a lower mole fraction of light than the liquid product of the single
flash, which means that the degree of separation has improved. However, we face two problems
with this new scheme: first, we produce two vapors, one of which (V1) is not as rich in light
component as the other one (V2) (i.e., V1 is not a desirable vapor product); second, we are
required to supply heat to the two units, which makes the process substantially more expensive,
and difficult to carry out in a tray column, since heating the liquid in a tray is very difficult in
practice. On the other hand, the vapor V1 is at a higher temperature than flash 2, and contains
more heavy component than V2. This means that, if we introduce V1 into flash 2, we will be
providing this flash with thermal energy, not only product of its higher temperature, but also as a
result of condensation of the heavy component. This release of thermal energy will serve to
partly vaporize the liquid feed, L3, so that an external heat supply will not be required. This
rearrangement leads to the scheme shown in Figure 11.

Figure 9. A flash distillation process and its representation in a y-x diagram.

More stages can be added to the scheme shown in Figure 11 to increase separation. With this
process we will get a liquid product that can be very concentrated in heavy component, but the
vapor product composition (V2 in Figure 11) will still be limited by the separation that can be
accomplished from the feed in only one flash, since both feed and vapor product are flows
associated with the first flash. To concentrate more the light component in the vapor product, we
could introduce the feed at an intermediate stage in the multistage process, as is shown in Figure
12 for an n stage process with the feed introduced at stage number F. The problem with this
scheme is that we need a liquid, Ln+1, that must be richer in light component than the feed. The
solution to this problem is to condense the vapor Vn and use part of the liquid produced as Ln+1,
as shown in Figure 13.
 13

Figure 10. Two flash processes in series: the bottoms product from flash 2 is the liquid feed
to flash 2. Note that T1>T2.

Figure 11. A two-stage distillation process that is energy-efficient: the thermal energy of the
vapor V1 is used to accomplish the partial vaporization of the liquid feed to flash 2.
 14

Figure 12. A multistage process that would result in a vapor product (Vn) that is rich in light
component.

Figure 13. The vapor product produced in the last stage of the process in Figure 12 is
condensed by extracting from it a heat rate QC in a heat exchanger (total condenser). Part of the
liquid obtained is produced as distillate (rich in light component), and part is recycled to the
multistage process (reflux).
 15

The scheme proposed in Figure 13 not only solves the problem of the liquid required in stage
n, but also allows us to produce the distillate as a liquid, which, in general, will be easier to
handle than a vapor.
The whole process described above can be carried out in a tray column, except for stage 1,
since a heat exchanger is needed to provide heat to the mixture in this first stage. The external
heat exchanger where Q is added to the system is called reboiler. Figure 14 shows a schematic
diagram of a distillation column with N trays. The portion of the column that is above the feed
tray is called rectifying or enriching section, whereas the region below the feed is called stripping
section. The column has N trays, but, since the reboiler is really a flash, there would be N+1
equilibrium stages in this distillation process.
Tray distillation columns are identical in construction to tray absorption columns (see Part
II). Notice that, in both cases, the trays serve to contact a gas phase with a liquid phase in
countercurrent flow. There is, however, a fundamental difference: in an absorption tray, there is
mass transfer from the gas to the liquid, as the solute is absorbed into the liquid phase, whereas
in a distillation tray, the heavy component condenses from the vapor into the liquid while the
light component evaporates from the liquid into the vapor. In fact, under certain circumstances,
the molar amount of material that evaporates equals the molar amount of material that
condenses. This happens commonly in practice (as an approximation), when the following
assumptions and conditions are satisfied:

(1) The tray is adiabatic (i.e., thermally insulated from the surroundings).
(2) The molar enthalpy of vaporization of both components is approximately the same, i.e.,

∆H vap,1 ≈ ∆H vap, 2 (25)

where the subindices 1 and 2 denote light and heavy component, respectively.
(3) Thermal energy exchange due to temperature differences is negligible with respect to
evaporation and condensation.

Condition (1) can be met in practice with a good design. Condition (2) is the most restrictive,
and is discussed in more detail below. Condition (3) is typically met in practice: over a relatively
narrow range of temperatures, the heat required to change the temperature of a fluid will be
much lower than the heat required to evaporate the same amount of material. For example, for
water (100°C, 1 atm), the heat capacity of the liquid is cpL=76 kJ/kmol K and the enthalpy of
vaporization is ∆Hvap=40600 kJ/kmol. For a temperature difference of 10 K, which is rather high
compared to what is observed in a distillation tray, the ratio between the heat required to heat up
water this temperature difference and the heat required to evaporate the same amount of water is
76x10/40600=0.018 (1.8%). This calculation indicates that condition (3) is commonly met in a
distillation tray.
To explore the consequences of the conditions listed above, consider material and thermal
energy balances in a distillation tray. Schematically, the processes occurring in a distillation tray
are represented in Figure 15. Mole balances of liquid and vapor phases yield:
 16

Figure 14. Scheme of a tray distillation column.

 17

Figure 15. Elements for a mole balance in a distillation tray. In the tray, WH moles of heavy
component condense per unit time, and WL moles of light component evaporate per unit time.

L in + WH = L out + WL (26)

Vin + WL = Vout + WH (27)

Under the conditions specified above, a thermal energy balance can be written as:

WH ∆H vap, 2 = WL ∆H vap,1 (28)

Combining this equation with equation (25) yields

WH = WL (29)

i.e., the mass transfer process in a distillation tray corresponds to an equimolar counterdiffusion
process. In this case, equations (26) and (27) simplify to:

L in = L out (30)

Vin = Vout (31)

which states the molar flow rates of vapor and liquid do not change across a distillation tray. This
simple result is known as "constant molal overflow".
The most restrictive condition necessary to achieve constant molal overflow is the equality of
heats of vaporization, equation (25). However, this condition will be satisfied in practice in a
wide variety of situations. For example, if the two components to be separated have similar
chemical structures, such as substances belonging to a homologous chemical series, such as
alkanes, and have close molecular weights, equation (25) will be, in general, a good
approximation. If the two components are very different from a molecular structure standpoint,
 18

or if they belong to the same series but have very different molecular weights, good separation
might be achieved in a flash, so that a distillation column would not be required. In the case that
the conditions stated above are not satisfied, then a thermal energy balance must be done around
each tray to quantify the changes in vapor and liquid flow rates. In the analysis that we will
perform here, we will assume that constant molal overflow is valid.

Calculation procedure for binary distillation in a tray column

Under the assumption of constant molal overflow, the molar flow rates of vapor and liquid
are uniform in each section of the column (rectifying and stripping sections). They change
around the feed tray because of the addition of the feed.
The calculation procedure to be described here is based on mole balances and the equilibrium
relation. The change in flow rates at the feed tray indicates that each section of the column
should be treated separately. We will start with overall mole balances, using the whole column as
a control volume (see Figure 14):

Total: F= D+B (32)

Light component: Fx F = Dx D + Bx B (33)

Mole balances in the rectifying section (control volume CV1 in Figure 14) yield

Total: V = D+L (34)

Light component: Vy n +1 = Dx D + Lx n (35)

Note that equation (34) is also a total mole balance in the condenser, where the vapor
produced at the top of the column (V) is totally condensed to a liquid that is then divided in two
parts: a distillate (D), and a reflux (L). Note that both these flows have the same composition
(x0=xD). The reflux is usually quantified by means of the reflux ratio, defined by

L
R= (0<R<∞) (36)
D

This is a very important parameter in the design of distillation processes. Note that the reflux can
be manipulated; i.e., it is the designer's or operator's choice to decide how much of the liquid
produced in the condenser is returned to the column. Using equations (36) and (34) we can
express the liquid and vapor flow rates in the rectifying section in terms of R:

L = RD (37)

V = (R + 1)D (38)

Dividing by V and using these last two relations, equation (35) can be expressed as follows,
 19

R x
y n +1 = xn + D (39)
R +1 R +1

This equation relates the mole fractions of light component in the vapor and the liquid at a given
level in the rectifying section (yn+1 and xn). Therefore, we will use it to define the Rectifying
Section Operating Line (RSOL),

R x
y= x+ D (40)
R +1 R +1

Note that this line passes through the point (xD,xD), its slope is R/(R+1) (always less than 1), and
it intersects the y-axis at y=xD/(R+1). A representation of this line is shown in Figure 16.

Figure 16. Representation of the Rectifying Section Operating Line (RSOL, equation 40). The
line can be drawn by joining the point corresponding to x=xD on the y=x line with the
intersection of the y axis y=xD/(R+1).

Mole balances in the stripping section (control volume CV2 in Figure 14) yield

Total: L=V+B (41)

 20

Light component: Lx m = Vy m +1 + Bx B (42)

Note that, for m=N, these two equations become balances in the reboiler. Dividing equation (42)
by V and substituting equation (41) leads to

L Bx B
y m+1 = xm − (43)
L−B L−B

This equation relates the mole fractions of light component in the vapor and the liquid at a given
level in the stripping section (ym+1 and xm). Therefore, we will use it to define the Stripping
Section Operating Line (SSOL),

L Bx B
y= x− (44)
L−B L−B

Note that this line passes through the point (xB,xB), its slope is always greater than 1. A
representation of this line is shown in Figure 17.

Figure 17. Representation of the Stripping Section Operating Line (SSOL, equation 44). The
line can be drawn from the point corresponding to x=xB on the y=x line with a slope equal to
L /( L − B) .
 21

In a typical design problem (see below), R, D and xD will be known, so that the RSOL can be
drawn. For the SSOL, however, B and xB will be known, but L must be calculated. The liquid
and vapor molar flow rates in the stripping section are determined by balances around the feed
tray. Let us consider that the feed may be: (1) a saturated liquid, (2) a saturated vapor, or (3) a
vapor-liquid mixture at equilibrium (the cases of a subcooled liquid or a superheated vapor feed
will be considered later in this chapter). Let q the fraction of the feed that is a liquid (moles of
liquid per total moles of feed). When the feed enters the feed tray, its liquid portion, qF, will join
the liquid that comes from the rectifying section into the stripping section, whereas its vapor
portion, (1-q)F, will join the vapor that comes from the stripping section into the rectifying
section (Figure 18). Mole balances can then be written as follows:

L = L + qF (45)

V = V + (1 − q )F (46)

Notice that, if L and V are calculated from R and D (equations 37 and 38), then these last two
equations can be used to calculate L and V .

Figure 18. Distribution of molar flow rates in the feed tray.

When the feed enters the feed tray, it mixes with the vapor-liquid mixture that is being
contacted in the feed tray. When this mixing occurs, part of the feed might vaporize or condense,
but energy interactions indicate that the same number of moles that vaporize will condense as a
result of the fact that the enthalpies of vaporization of both components are equal. Therefore,
equations (45) and (46) will be always valid under the conditions considered here.
There is a graphical representation of the balances around the feed tray. Consider the
operating lines given by equations (40) and (44), expressed in terms of molar flow rates of liquid
and vapor in each section,

L D
y= x + x D (rectifying section) (47)
V V
 22

L B
y= x − x B (stripping section) (48)
V V

Let (xI,yI) be the point of intersection of these two lines. Multiplying each equation by the
corresponding vapor molar flow rate, evaluating both equations at the point of intersection, and
subtracting, leads to

(V − V ) y I = (L − L ) x I + Dx D + Bx B (49)

Using equations (33), (45) and (46), this equation becomes

(1 − q ) y I = −qx I + x F (50)

Based on this equation, we define the following straight line:

−q x
y= x+ F (51)
1− q 1− q

This line is called the feed line (FL), and it has the following characteristics: (1) it passes through
the intersection between the RSOL and the SSOL, (2) it passes through the point (xF,xF). Note
that the line can be drawn from a knowledge of q and xF, which are properties of the feed that
always will be known. Figure 19 illustrates how the feed line can be used in the graphical
representation of the operating lines for a distillation calculation.
With the information that we have discussed, we can develop a procedure to determine the
number of theoretical stages required for a multistage binary distillation process. To do this, we
will formulate a typical design problem as follows:

INFORMATION AVAILABLE:

F, xF and q  feed
P  operating pressure
Equilibrium  from Thermodynamics, with knowledge of P
D, xD  goals
R  reflux ratio

DESIGN PARAMETERS:

N  number of equilibrium trays (and feed location)

QR  heat transfer rate required in the reboiler (reboiler duty)
QC  heat transfer rate required in the condenser
 23

Figure 19. Construction of operating lines from a knowledge of xF, xB, xD, R and q: the RSOL
operating line is drawn by joining the point on the y=x line at xD with the y axis at xD/(R+1); the
feed line is drawn from the point on the y=x line at xF with slope -q/(1-q); finally, the intersection
point between the RSOL and the FL is joined with the point on the y=x line at xB to find the
SSOL.

With the information given, the following parameters can be calculated:

B can be calculated from equation (32)

xB can be calculated from equation (33)
L can be calculated from equation (37)
V can be calculated from equation (38)
L can be calculated from equation (45)
V can be calculated from equation (46)
 24

In addition, the operating lines can be drawn in a y vs. x diagram following the sequence
indicated in Figure 19. At this point, the McCabe-Thiele method can be used to determine the
number of theoretical stages required for the specified process, as shown in Figure 20.

Figure 20. The McCabe-Thiele method for binary distillation.

The application of the McCabe-Thiele method starts with the distillate point on the RSOL.
Note that this point corresponds to the liquid and vapor compositions at the top of the column, x0
and y1, which are both equal to xD (refer to Figure 14). The value of x1 can be found from y1 and
the equilibrium line, and so forth. The last stage drawn is the reboiler, since it acts as a flash. For
the particular example represented in Figure 20, the column has 5 equilibrium trays plus a
reboiler, and the feed enters the column at tray 4 (see below). As was the case for multistage
absorption, the number of equilibrium stages might be a fractional number.
 25

In the following sections we will address some practical design issues associated with binary
distillation calculations.

Feed location
In the example illustrated in Figure 20, we located the feed in the tray that passed the
intersection between the two operating lines. This is not the only possible solution to this design
problem: the feed may be located in other trays. In fact, the analysis in the previous section
indicates that the feed tray is merely the tray at which we need to change operating lines.

Figure 21. The same design case as in Figure 20, but locating the feed in tray 6 (see text for
discussion).

For example, in Figure 21 we show the application of the McCabe-Thiele method to the same
system in Figure 20, but, instead of introducing the feed into the column in tray 4, we extend the
 26

use of the RSOL until tray 6, and then we change operating lines. This means that the feed is
introduced in tray 6. This design requires 6 equilibrium trays plus reboiler (actually, a little less;
notice that the reboiler slightly overshoots xB), so that the result is not as good a design as that in
Figure 20. In fact, locating the feed at the intersection between operating lines yields the
optimum feed location, since this minimizes the total number of trays required.

Choice of reflux ratio

In the design procedure presented above, we assumed that the reflux ratio, R, was known.
However, R is a parameter that can be manipulated in a design operation. Recall that R controls
the amount of liquid in the rectifying section of the column, since L=RD. Let us explore extreme
values of R. Suppose first that R is very large (R→∞). Physically, to have finite flow rates within
the column and achieve this condition, we require D→0. An infinite reflux is obtained by
shutting off the distillate completely. However, since we must operate at steady state, this means
that both F,B→0. This mode of operation, illustrated in Figure 22, is called total reflux. In this
case, the fluids are continuously recirculated inside the column: all the vapor condensed in the
condenser is returned to the column, and all the liquid reaching the reboiler is evaporated and
returned to the column. Even though no products are drawn, the potential distillate would have a
mole fraction of light component xD and the potential bottoms would have a composition xB.
Note that the reboiler has ceased to be an equilibrium stage (flash) since now the feed to it is
completely vaporized.

Figure 22. Operation at total reflux: there are no feed or products, but a degree of separation
is obtained between the potential distillate (xD) and the potential bottoms product (xB).

The operation at total reflux is used in practice to start up and shut down distillation columns.
 27

A typical start up operation for a distillation column consists of the following steps: (1) load a
certain amount of liquid into the reboiler, (2) start up the heating element in the reboiler, (3) once
enough vapor is generated, start up the cooling element in the condenser, (4) wait for steady
state, and (5) once steady state is achieved for the total reflux operation, start feed and products
flows.
If we specify xD and xB for a given separation, a design calculation can be done for the case
of total reflux. Consider the rectifying section operating line (equation 40),

R x
y= x+ D
R +1 R +1

For a given xD, if we let R→∞ in this equation, we obtain

y=x, operating line for total reflux (52)

Furthermore, in this case the column has only one section (since there is no feed). Using this
operating line, we can determine the number of equilibrium trays necessary to achieve the
desired separation by applying the McCabe-Thiele method, as shown in Figure 23.
In practice, a design calculation at total reflux is not very interesting, since we will never
build a column to operate it without getting any products. However, a design calculation at total
reflux gives the minimum number of trays necessary to achieve a specified separation. To
understand this, compare Figure 20 (finite reflux) with Figure 23 (infinite reflux). As R→∞, the
distance between the operating line and the equilibrium line is maximized. As this distance
increases, the separation that can be achieved in any tray increases. The operating line that can
have the greatest separation form the equilibrium line is y=x (total reflux), and, therefore, it will
yield the lowest possible number of trays.
The calculation of the minimum number of equilibrium trays at total reflux operation gives a
lower bound for the number of trays that will be required and, consequently, it is a calculation
that is convenient from a practical stand point.
As the reflux ratio is decreased from the point of total reflux, the number of stages obtained
increases until the operating lines touch the equilibrium line at a point (Figure 24). Under this
condition, the number of stages required would be infinite. This point, therefore, yields a
minimum reflux for the specified separation. Notice that the minimum reflux condition is
qualitatively similar to the minimum liquid flow rate condition in absorption calculations.
The shape of the equilibrium curve might affect the determination of minimum reflux.
Sometimes the first contact between operating lines and equilibrium line will not occur at the
point of intersection between operating lines, but somewhere else. This type of situation is
illustrated in Figure 25.
Summarizing, the choice of reflux ratio for a specified separation must be made between the
following two limits:

R=Rmin, which implies N→∞

R→∞, which implies N=Nmin

 28

Figure 23. McCabe-Thiele method for total reflux operation. In this case (same as Figures 20
and 21), 4 trays are required to achieve the desired separation.

In practice, the choice of reflux ratio is dictated by costs. Figure 26 shows a typical
qualitative curve of cost variations with reflux ratio. Since the costs increase without bounds
when R→Rmin (since N→∞), or when R gets very large (high energy costs due to the large
amount of liquid that the reboiler needs to evaporate), there is a point of minimum cost (Ropt) that
should be selected as design point. For many practical systems, this optimum reflux ratio is
approximately given by

R opt ≈ 1.5R min (53)

For design purposes in this course, we will consider that this is always the optimum, but a
 29

detailed economic evaluation would be necessary in a real design situation.

Figure 24. Minimum reflux: as R decreases, the operating lines get closer to the equilibrium line.
When the equilibrium line is first touched by the operating lines (which, in this case, happens at
the point of intersection between FL, RSOL and SSOL), the number of stages required is infinite.
This operating condition corresponds to the minimum reflux condition.

Feed condition
In the analysis of tray column design that we have carried out so far, we have used the
condition of the feed to define the feed line and the optimum location of the feed tray. The
thermodynamic condition of the feed has been quantified by q, a parameter that has been defined
as the fraction of the feed that is a liquid. We have allowed three possible cases:

q=0, the feed is a saturated vapor (i.e., it is at its dew point)

 30

q=1, the feed is a saturated liquid (i.e., it is at its boiling point)

0<q<1, the feed is a vapor-liquid mixture

Figure 25. Example of determination of minimum reflux for a case in which the operating lines
touch the equilibrium line at a point different from the point of intersection of operating lines.
The operating lines represent the first value of R obtained (as R is decreased) for which N→∞.

However, the feed can also be a superheated vapor or a subcooled liquid. In fact, the latter
case is the most common in practice. In this section we will redefine q, and the feed line, to
accommodate these two cases. Our new definition will be consistent with the meaning of q
employed in previous sections.
Consider the two new cases (Figure 27). Energy interactions between the feed and the
mixture in the feed tray will result in a net evaporation for the case of superheated vapor feed
(Figure 27a), and a net condensation for the case of subcooled liquid feed (Figure 27b). To
 31

determine the changes in vapor and liquid flow rates around the feed tray, an energy balance is
necessary. We will consider that the feed tray is adiabatic. In this case, the only thermal energy
interactions around the feed tray will be quantified by the enthalpies of the flows.

Figure 26. Typical Cost vs. R curve for a tray distillation column. The total cost includes the
cost of building the column plus the cost of energy supplied in the reboiler. At low reflux ratios,
a high number of trays is required and the column cost becomes dominant. At high reflux ratios,
the number of trays approaches Nmin, and the energy costs become dominant. The point of
minimum total cost is used to select the optimum reflux ratio.

Consider a thermal energy balance in the control volume shown in Figure 28:

FH F + LH L + VH V = LH L + VH V (54)

We will define q by means of the following equation

L = L + qF (55)

Note that this equation is identical to equation (45), so that the meaning of q for vapor-liquid
mixtures is retained. However, now we will allow q to be greater than 1 (subcooled liquid feed)
or less than zero (superheated vapor feed). To find the relation between vapor flow rates, we use
a total mole balance in the CV of Figure 28:

L+F+ V = L+V (56)

 32

Figure 27. Flow rate changes in the feed tray. When the feed is a superheated vapor (a), all of it
will contribute to the vapor of the rectifying section, but it must lose some thermal energy to
become a saturated vapor; this thermal energy evaporates part of L. When the feed is a subcooled
liquid (b), all of it will contribute to the liquid of the stripping section, but it must gain some
thermal energy to become a saturated liquid; this thermal energy is obtained by condensing part
of V .

Figure 28. Control volume (CV) for the thermal energy balance in the feed tray.

Substituting equation (55) into equation (56) leads to

V = V + (1 − q )F (57)

which is identical to equation (46). Now we substitute equations (55) and (57) into equation (54)
to obtain, after manipulations,

F[H F − (1 − q )H V − qH L ] = L(H L − H L ) + V(H V − H V ) (58)

 33

The liquid flows around the feed tray are, ideally, saturated liquids, and the vapors are saturated
vapors. Even though their temperatures and compositions change somewhat, their enthalpy
differences will be negligible compared to the enthalpies of vaporization. Based on this, it is
usually safe to assume that:

HL ≈ HL
HV ≈ HV

Using these two approximations, and recalling that

∆H vap = H V − H L

allows us to simplify equation (58). Solving then for q leads to

HV − HF
q= (59)
∆H vap

We will use this equation to calculate q for subcooled liquids or superheated vapors, as follows:

(1) Superheated vapor.

Since both the feed and the vapor are gases, their molar enthalpy difference can be evaluated
using the molar heat capacity of the vapor, CpV. The temperature of the vapor at the feed tray
cannot be determined but, if the optimum feed tray location is used, the composition of the feed
will be similar to the composition of the vapor leaving the feed tray, and this vapor is at its dew
point. Therefore, as an approximation, the temperature of the vapor leaving the feed tray can be
taken as the dew point of the feed (Tdew), and this means that

H V − H F = C pV (Tdew − TF ) (60)

where we are assuming that CpV is a constant in the given range of temperature. The temperature
Tdew can be obtained from a T-x-y diagram at the feed composition. Substituting equation (60)
into equation (59) yields

C pV (TF − Tdew )
q=− (61)
∆H vap

Note that, since the feed is a superheated vapor, TF>Tdew, and, therefore, q<0.

(2) Subcooled liquid.

In this case, since the feed is a liquid, to calculate the enthalpy change needed in equation
(59), we use the following equation,

H V − H F = (H V − H L ) + (H L − H F )
 34

The first term on the right hand side of this equation is the enthalpy of vaporization, while the
second term represents a change of enthalpy of a liquid, which will be quantified in terms of the
molar heat capacity of the liquid, CpL, to get:

H V − H F = ∆H vap + C pL (Tboi − TF ) (62)

where we have assumed that CpL is a constant. By virtue of the same reasons used before, the
temperature of the saturated liquid will be taken as the boiling point of the feed (Tboi).
Substituting equation (62) into equation (59) leads to

C pL (Tboi − TF )
q = 1+ (63)
∆H vap

Note that Tboi>TF and hence q>1.

Since the definition of the feed line was derived from the mole balances around the feed tray,
which are still the same in terms of the new definition of q, the feed line is still given by equation
(51). A representation of the various possible cases is shown in Figure 29.

Figure 29. Possible orientations of the feed line for various values of q. The slope of the feed
line is -q/(1-q).
 35

Condenser type
In the distillation process that we have analyzed so far, the distillate is produced as a liquid,
since the vapor coming out at the top of the column is completely condensed to produce the
distillate and the reflux (see Figures 13 and 14). The condenser used in this case is called total
condenser. In some cases, it is not convenient to produce the distillate as a liquid, since
condensing all the vapor at the top of the column, which is rich in light component, might require
very low temperatures. In this case, a partial condensation is achieved in a heat exchanger. The
condensed liquid constitutes the reflux, and the vapor produced is the distillate. This scheme is
called partial condenser, and it is represented in Figure 30. The use of partial condensers is
especially important in multicomponent distillation processes in which the mixture to be distilled
contains a small portion of very light components that have very low boiling points.

Figure 30. Partial condenser. In this scheme, the vapor leaving the column is not completely
condensed (as opposed to a total condenser, see Figure 13). The liquid produced in the condenser
is the reflux and the vapor is the distillate. The partial condenser is equivalent to a flash
distillation unit.

The determination of the number of stages required by a process with a partial condenser
proceeds the same as before, but now the fact that the partial condenser is an equilibrium stage
must be considered. For example, for the calculation depicted in Figure 20, the partial condenser
would take the place of tray 1, and the column would need 4 theoretical trays (the six equilibrium
stages required would be 4 trays, a partial condenser, and a reboiler).

Thermal energy interactions

Two important design parameters are the heat transfer rates in the condenser (QC) and the
reboiler (QR). In fact, QR will be an important portion of the total energy requirements of a
distillation process, which in turn represents a sizable contribution to operating costs. In typical
industrial-scale distillation columns, the condenser is a heat exchanger whose cold-side fluid is
cooling water, whereas the reboiler is a heat exchanger in which the hot fluid is steam whose
condensation releases the thermal energy that constitutes QR.
The heat required in the condenser can be evaluated by a thermal energy balance in the
 36

condenser. Regardless of the type of condenser used, this thermal energy balance can be written
as (see Figures 14 and 30):

VH V = Q C + LH L + DH D (64)

where QC>0. For a total condenser, both distillate and reflux are saturated liquids: HD=HL, and,
since V=L+D=(R+1)D and ∆H vap = H V − H L , equation (64) becomes

Q C = D(R + 1)∆H vap , total condenser (65)

On the other hand, for a partial condenser, HD≈HV (both are saturated vapors), and equation (64)
leads to

Q C = DR∆H vap , partial condenser (66)

The heat transfer rate in the reboiler (sometimes called "reboiler duty") can be calculated
from a thermal energy balance in the reboiler. This balance leads to (see Figure 14):

LH L + Q R = BH B + VH V (67)

Since B is a saturated liquid, H L ≈ H B , and using L = B + V and ∆H vap = H V − H L , we get:

Q R = ( L − B)∆H vap (68)

To put this equation in terms of the basic information provided in a design calculation, we can
substitute equation (55) and L=RD to get

Q R = (RD + qF − B)∆H vap (69)

Note that equations (65) or (66) and equation (69) can be used to calculate the heat transfer rates
in the condenser and the reboiler from the basic information available in a typical design
problem.
Equations (65) and (66) will be valid in general, but equation (69) has the assumption of
adiabatic trays built-in. If the column is not well insulated, QR must be calculated from equation
(68), recognizing that L would be the liquid flow rate going into the reboiler (which would not
be uniform throughout the stripping section).
When the column is not well insulated, thermal energy losses can be evaluated by means of a
thermal energy balance in a control volume that encompasses the whole column. This balance
can be written as

FH F + Q R = DH D + BH B + Q C + Q L (70)

The left-hand side of this equation represents all the thermal energy per unit time going into the
 37

column, and the right-hand side represents all the thermal energy leaving the column, including
the rate of heat losses (QL). Equation (70) is used to determine energy losses, and it is usually
expressed in terms of enthalpy differences by using F=D+B to get:

Q L = D(H F − H D ) + B(H F − H B ) + Q R − Q C (71)

The enthalpy differences in this equation must be evaluated by different means, depending on the
thermodynamic condition of the feed and the distillate (which depends on the type of condenser).

Tray efficiency
In the design procedure described above, we have considered that the trays in the distillation
column behave as equilibrium stages. In practice, a reboiler and a partial condenser will behave
as equilibrium stages, but the trays will not. Formally, a tray that acts as an equilibrium stage
would do so if that mass transfer between the vapor and the liquid is fast enough that equilibrium
is reached. This would mean that the vapor leaving the tray is at equilibrium with the liquid
leaving the tray. For the tray in Figure 31, this would mean that yn is at equilibrium with xn.

Figure 31. Flow rates and compositions around a distillation tray. If the tray were an
equilibrium stage, yn and xn would be at equilibrium.

In general, in real trays there are mass transfer limitations and equilibrium is not always
achieved. Since the vapor phase is being enriched in light component in the tray, we have that
yn>yn+1. If equilibrium is not achieved, the real yn would be lower than the equilibrium value. Let
y*n be the vapor-phase mole fraction that would be at equilibrium with xn. If equilibrium is not
achieved, then we would have: y*n >yn>yn+1. The departure from equilibrium is usually quantified
in terms of the Murphree tray efficiency, EM, defined as the ratio between the change in
composition achieved in the real tray and the change of composition that would be achieved in
an equilibrium tray; i.e.,

y n − y n +1
EM = (72)
y*n − y n +1
 38

Note that EM<1. Sometimes EM is specified as a percentage instead of a fraction.

In a y vs. x diagram, the Murphree efficiency can be interpreted geometrically, as shown in
Figure 32. This geometrical interpretation suggests a modification of the McCabe-Thiele method
that allows for the calculation of the actual number of trays that would be necessary for a given
separation, for the case in which EM is known and is the same for all trays. In this case, a line can
be drawn through intermediate points between the equilibrium line and the operating lines such
as point B in Figure 32. These points can be located using the geometric relation in Figure 32
and the known value of EM. This new line, which we will call pseudo-equilibrium line,
represents a line where the real yn's should be located, instead of the equilibrium line. The
McCabe-Thiele method can then be used to determine the number of real (non-equilibrium) trays
in the column, by using the pseudo-equilibrium line as the end point for the drawing of an
individual tray, in the manner shown in Figure 33.

Figure 32. Geometrical interpretation of the Murphree efficiency in a y-x diagram. Portions
of the equilibrium and operating lines are shown (this could be either in the rectifying or
stripping sections of the distillation column). The Murphree efficiency is the ratio of distances
between points A and B, and points A and C.

The Murphree efficiency is a measure of the impact that vapor-liquid mass transfer rates have
on tray performance. For this reason, the value of EM depends on the same parameters that a
convective vapor-liquid mass transfer coefficient would depend on; namely, physical properties
of the phases, tray flow patterns and tray geometry.
The use of Murphree efficiencies is the most rigorous way to calculate the actual number of
trays that a distillation column requires. However, as a rough estimate, and in the absence of
information on Murphree efficiencies, we can use empirical correlations to calculate the overall
column efficiency, defined as

N
E0 = (73)
NA

where N is the number of equilibrium trays and NA is the number of actual trays. If E0 is known,
N can be obtained by means of the McCabe-Thiele method, and then the number of actual trays
can be calculated. An example of a correlation for overall efficiency is the version of O'Connell's
correlation for distillation columns, presented in Figure 34. To use this correlation, the following
 39

facts should be taken into account:

Figure 33. The McCabe-Thiele method used to determine real trays when the Murphree
efficiency is known. The pseudo-equilibrium line is drawn by joining points like B, determined
from the geometric relation in Figure 32. This plot corresponds to EM=0.7. The column designed
would have 7 trays plus a reboiler and a total condenser. Note that the reboiler is drawn going all
the way to the equilibrium line (a partial condenser would be drawn the same way), because it is
considered an equilibrium stage.

(1) The feed viscosity must be a liquid viscosity. If the feed is a liquid, then µF is the
viscosity of the feed, but if the feed is a vapor or a vapor-liquid mixture, µF is the viscosity of the
saturated liquid having the same composition as the feed.
(2) If the equilibrium is given in terms of a y vs. x equilibrium line, the relative volatility can
 40

be calculated by solving equation (12) for α:

y /(1 − y)
α= (74)
x /(1 − x )

Figure 34. O'Connell's correlation for overall efficiency of tray distillation column. The
parameters are: µF is the viscosity of the feed as a liquid (in cp), and α is the average relative
volatility of the light component.

where y and x are at equilibrium. Since the relative volatility will change along the column, an
arithmetic average must be used,

αT + αB
α= (75)
2

where αT and αB are the relative volatilities calculated at conditions at the top and bottom of the
column, respectively.
 41

Short-cut design guidelines for tray distillation columns

A. Information available:

(1) Feed flow rate, composition, and thermodynamic state: F, xF, q or TF.

(2) Pressure at the top of the column (P).

(3) Compositions of distillate and bottom products (xD and xB).

(4) Information on vapor-liquid equilibrium (T-x-y and y-x diagrams), enthalpies and heat capacities.

B. Information that might be available... ... what to do if it is not available

(5) Relative feed location Use optimum feed location

(6) Reflux ratio (R) Use R=1.5Rmin

(7) Murphree (EM) or Overall (E0) tray efficiency. Estimate E0 using O'Connell's correlation

C. Report design parameters:

(1) Determine number of trays using McCabe-Thiele method: if EM is known, calculate number of actual trays
with pseudo-equilibrium line; if E0 is known, calculate number of equilibrium trays N, and then determine
number of actual trays as N/E0.

(2) Column diameter: assume a gas superficial velocity vg=1 m/s, and determine diameter (d) from

πd 2 R'T
vg = Qg = V
4 P

where Qg is volumetric flow rate. The vapor molar flow rate (V) and temperature (T) to use here should be
selected at a tray such that the calculated diameter is the greatest possible. In this equation, R' is the universal
gas constant.

(3) Column height: assume that top and bottom disengagement sections require a height H0 (usually, H0≅2-3
m), and then use an incremental height of 0.7 m/tray.

(4) Total pressure drop: for sieve trays, assume pressure drop is 3 in H2O per tray.

(5) Calculate condenser and reboiler heat transfer rates.

 42

4. INTRODUCTION TO MULTICOMPONENT DISTILLATION

A large amount of distillation columns in industry deal with systems that contain more than
two components. If a multicomponent mixture is to be separated into purified forms of all of its
components, it is important to acknowledge that this cannot be accomplished in a single
distillation column. In fact, for a mixture of n components, n-1 distillation columns are required
to achieve purification. For example, for a ternary mixture of components A, B and C, where the
order of volatility is: A>B>C, a possible scheme to purify the three components by distillation is
shown in Figure 35.

Figure 35. Example of the use of distillation to purify a ternary mixture: two columns would
be required.

Distillation columns might have side exits (i.e., liquid or vapor extracted from an
intermediate tray in the column). For the ternary mixture considered, one could propose to use
one column to produce purified A in the distillate, C in the bottoms product, and then have a side
 43

exit to produce B. However, it should be clear that B cannot be produced at a high purity level in
a side exit: in the rectifying section, the mole fraction of A will increase as we move upward, so
that any side exit should be richer than the feed in A, whereas in the stripping section, C is
concentrated as we move downwards, so that any side exit should contain more C than the feed.
Hence any side product, even if it manages to recover a sizable amount of B, will have either A
or C in significant proportions.
Sometimes the quality of a product is not determined by its purity but by other
characteristics. For example, in petroleum refining, the objective of a distillation operation is to
separate the crude oil into different fractions with different volatilities, each of which has a
specific application. In this case it is appropriate to use side exits to produce fractions of
intermediate volatility, as shown in Figure 36.

Figure 36. One of the first steps in petroleum refining is to distill the crude oil to produce
several products with different volatilities. Each product shown has a relatively large number of
components, but they have a specific market value. Note that a partial condenser is used, due to
the low boiling point of the lighter fractions (natural gas).

Even when there are more than two products (apart from distillate and bottoms), the design
of a multicomponent distillation column generally will be determined by the degree of separation
required between distillate and bottoms. In a design, it is important to be aware of the limitations
regarding the number of parameters that can be specified to set the goal of the process.
Multicomponent distillation columns do not provide additional degrees of freedom over binary
columns, since the number of operating parameters available are the same (i.e., reflux ratio,
number of stages, location of feed tray). In what follows we analyze a typical example of
 44

specifications that can be made for a design problem.

To have enough information to determine the number of equilibrium trays required for a
specific operation involving n components, we could specify the following:
(1) Characteristics of the feed: F, xFi (for n-1 components), thermodynamic condition of the
feed.
(2) Operating pressure: P. This would be the pressure at the top of the column.
(3) Specify that the feed is to be placed at its optimum location; i.e., the feed location that
would minimize the number of stages required.
(4) Reflux ratio: R.

At this point, two additional specifications need to be made to define the problem
completely. These specifications will constitute the goals of the distillation process. It is
customary to relate the two additional specifications to two components in the mixture. We then
should identify two components on which we will base the degree of separation of the process.
In general these two components will be defined in such a way that one of them will be
substantially recovered in the distillate (this components is called the "light key") while the other
one will be substantially recovered in the bottoms product (this component is called the "heavy
key"). If the n components present in the feed are numbered in order of decreasing volatility, let
LK and HK be the number of the light and heavy key components, respectively. The two final
specifications could be:

(5) Mole fraction of the light key in the distillate: xDLK.

(6) Mole fraction of the heavy key in the bottoms: xBHK.

Providing that these two specifications yield a high recovery of each component in each product,
we would expect the product distribution shown in Figure 37: there should be no appreciable
amount of components lighter than the light key in the bottoms product (since there will be only
a small amount of light key there), and there should be no appreciable amount of components
heavier than the heavy key in the distillate (since there will be only a small amount of heavy key
there). Note that the only components present in both products in measurable amounts will be the
keys and all components with volatilities between the keys (if any). This type of product
distribution represents the simplest case of specifications in a multicomponent distillation
process.
We will start our quantitative analysis of multicomponent distillation with the case of single
stage (flash) distillation, and then we will treat multistage, multicomponent distillation by means
of approximate methods.

Multicomponent flash distillation

Consider the flash distillation of a multicomponent mixture (Figure 38). A total mole balance
in the flash is identical to that obtained for a binary flash (equation 13). To characterize the total
mole balance, we will use once again the fraction of the feed that is vaporized, f, defined by
equation (14). A typical problem might be formulated as follows:

INFORMATION AVAILABLE:

F, xFi, TF  what we want to treat

 45

P  operating pressure
Thermodynamics  equilibrium, enthalpies
T or f  flash operating goal

DESIGN PARAMETERS:

yDi, xBi  products composition

T or f  parameter that is not specified
Q  heat transfer rate required

Figure 37. Split of a multicomponent feed with two identifiable key components in a
conventional distillation column. The components are arranged by order of volatility (1-most
volatile, n-least volatile). See text for discussion.
 46

Figure 38. Multicomponent flash distillation.

Now we can write mole balances for each component in the mixture. However, since we are
making use of the total mole balance, only n-1 of these mole balances will be independent:

Fx Fi = Dy Di + Bx Bi , i=1,2,…n-1 (76)

Using equations (14) and (15), these mole balances may be written as follows,

x Fi = fy Di + (1 − f ) x Bi , i=1,2,…n-1 (77)

Before we explore the solution strategy, let us verify that the system is completely specified,
from a thermodynamic equilibrium point of view. We can calculate the number of degrees of
freedom using the phase rule:

F = C-P+2 (78)

In this case, the number of components is C=n, the number of phases is P=2, which implies F=n.
This means that n degrees of freedom must be specified to set the equilibrium in the flash. The
variables involved in the equilibrium are: P, T, f, and the compositions xBi and yDi. If f is known,
the n-1 mole balances (equation 77) set n-1 degrees of freedom so that the specification of one
additional variable defines the equilibrium. This would be P in our example. If f is not known,
the n-1 mole balances set n-2 degrees of freedom and two additional variables need to be known.
These would be T and P. In other words, setting P plus one additional parameter (either f or T)
completely defines the equilibrium state.
Now we will concentrate in solving the problem stated before for the two possible cases
(depending on whether T or f are known). First, we will assume that the equilibrium can be
characterized by distribution coefficients that are only functions of temperature; i.e.,

y i = K i (T ) x i , at equilibrium (79)

(recall that Ki=Pvi/P if Raoult's law is valid).

Since distillate and bottoms product are at equilibrium, we have

y Di = K i x Bi , i=1,2,…n (80)

Equations (77) and (80) will be used to solve the problem. The solution can be found by either of
 47

the two procedures described in what follows.

Procedure 1
Solve equation (80) for xBi and substitute into equation (77) to get

y Di
x Fi = fy Di + (1 − f ) (81)
Ki

This equation is valid for all components (i=1,2,…n). We can solve this equation for yDi and then
require that

n
∑ y Di = 1 (82)
i =1

The result is

n x Fi
∑( ) =1 (83)
i =1 1− f
f+
K i (T )

This equation provides a direct relation between f and T: if one of them is known, the other one
represents the only unknown in this equation. Solving equation (83) for its only unknown leads
to a knowledge of both f and T. At this point, the yDi's can be evaluated from equation (81) and
then the xBi's can be calculated from equation (80), which solves the problem.
An interesting limiting case of procedure 1 occurs when f→0. In this case equation (83)
reduces to

n
∑ K i (T ) x Fi = 1 (84)
i =1

The temperature resulting from solving this equation is the boiling point of the feed.

Procedure 2
Substituting (80) into equation (77) yields

x Fi = fK i x Bi + (1 − f ) x Bi (85)

This equation is valid for all components (i=1,2,…n). We can solve this equation for xBi and then
require that

n
∑ x Bi = 1 (86)
i =1

The result is
 48

n x Fi
∑( ) =1 (87)
i =1 fK i (T ) + 1 − f

This equation provides a direct relation between f and T: if one of them is known, the other one
represents the only unknown in this equation. Solving equation (87) for its only unknown leads
to a knowledge of both f and T. At this point, the xBi's can be evaluated from equation (85) and
then the yDi's can be calculated from equation (80), which solves the problem.
An interesting limiting case of procedure 2 occurs when f→1. In this case equation (87)
reduces to

n x Fi
∑ =1 (88)
i =1K i (T )

The temperature resulting from solving this equation is the dew point of the feed.

Multicomponent, multistage distillation: short-cut methods

The rigorous design of a multicomponent distillation process is a complex calculation that
can be performed with commercial software packages. In some cases, however, it is convenient
to have simple methods to estimate design parameters for a multicomponent distillation process,
in order to carry out a preliminary design. The main purposes of those methods are:

(1) To provide a rough but quick estimate of the final design that can be used, for example, to
discard unfeasible options.
(2) To provide the designer with adequate bounds for the design parameters, which can be
starting points for more rigorous designs.

In this section, we will explore simple calculation methods to estimate:

(1) The minimum number of trays required for a specific separation.

(2) The minimum reflux required for a specific separation.
(3) The total number of theoretical trays.
(4) The location of the feed tray.

We will establish the assumptions and principles behind every short-cut method.

(1) Minimum number of trays - the Fenske equation

As was the case in binary distillation, the minimum number of trays required for a specific
separation corresponds to an infinite reflux ratio (R→∞) which can be achieved in practice when
there are no products, and no feed is introduced to the column.
Our approximate analysis will be based in two assumptions:

1. Constant molal overflow (i.e., vapor and liquid molar flow rates do not change across a
tray).
2. Constant relative volatilities.
 49

There are systems for which these assumptions are good enough that our analysis will
provide an accurate estimate for the minimum number of trays. However, our intention is to use
the results for any system as an estimate. Consider the diagram shown in Figure 39. Mole
balances around the condenser lead to

Figure 39. Operation at total reflux of a multicomponent distillation column.

L=V (89)

y i1 = x Di (90)
 50

A mole balance in component i for the control volume shown in Figure 39 yields

Vy i 2 = Lx i1 (91)

which means that the two mole fractions are equal, by virtue of equation (89). In fact, vapor and
liquid mole fractions will be the same at any level in the column (notice that this is the equivalent
to the operating line y=x for total reflux in binary systems):

y ik +1 = x ik , k=1,2,…N (92)

To quantify the equilibrium for this system, we will use relative volatilities. For two arbitrary
components i and j in the mixture, we have the following equilibrium relations

y i = K i (T ) x i (93)

y j = K j (T ) x j (94)

The ij relative volatility is defined as follows,

Ki
α ij = (95)
Kj

so that the equilibrium relations (93) and (94) can be expressed as

yi x
= α ij i (96)
yj xj

Going back to the total reflux case (Figure 39), we see that xi1 and xj1 are at equilibrium with yi1
and yj1. Therefore, we have

y i1 x
= α ij i1 (97)
y j1 x j1

But, since yp1=xDp for any component p and xp1=yp2, this equation becomes

x Di y
= α ij i 2 (98)
x Dj y j2

Now, since xi2 and xj2 are at equilibrium with yi2 and yi2, the application of equation (96) leads to

x Di x
= α ij2 i 2 (99)
x Dj x j2
 51

Following this procedure sequentially down the column, we will eventually arrive at

x Di x y
= α ijN iN = α ijN iN +1 (100)
x Dj x jN y jN +1

The liquid in the reboiler (with composition xBi) is the potential bottoms product, and is at
equilibrium with the vapor. Therefore, using equation (96) once again, equation (100) becomes

x Di x
= α ijN +1 Bi (101)
x Dj x Bj

This equation is valid for any pair of components i and j. The distillate and bottoms composition
will not be known, in general, for all the components in the mixture. However, they might be
known for the light and heavy keys as specifications in a design problem. Therefore, letting
i=LK and j=HK, allows one to use equation (101) to obtain N:

x DLK N +1 x BLK
= α LK , HK (102)
x DHK x BHK

Solving for N and recalling that this would be the minimum number of stages for the specified
separation, since this is a total reflux calculation (N=Nmin) yields

x DLK x BHK
ln( )
x DHK x BLK
N min = −1 (103)
ln(α LK, HK )

This equation is known as Fenske's equation. Once the number of trays is calculated from this
equation, equation (101) and mole balances can be used to find the distribution of non-key
components.
The previous analysis is based on the assumption that relative volatilities are constant. In real
systems, they will be a function of temperature and will change from tray to tray. For varying
relative volatilities, it has been suggested that an average value be used in Fenske's equation.
This average has been defined as a geometric average between distillate and bottoms,

α ij = α ijD α ijB (104)

where α ijD and α ijB are the relative volatilities evaluated at distillate and bottoms temperature,
respectively.

(2) Minimum reflux ratio - Underwood's method

For a given separation, decreasing the reflux ratio from the total reflux condition (R→∞)
requires an increasing number of theoretical trays. This trend will continue until a value of R is
 52

reached for which N→∞. This value of R is the minimum reflux ratio, Rmin. In binary distillation,
the minimum reflux condition was characterized by the first contact between the operating and
equilibrium lines (pinch point). Typically, the pinch point coincided with the feed location, as
illustrated in Figure 40. Since the pinch point cannot be reached, infinite equilibrium stages
would be required. Depending on the shape of the equilibrium line, the pinch point can also
occur away from the feed. Consider the example in Figure 41, where the pinch point occurs in
the rectifying section.
For multicomponent mixtures, the behavior is more complex than for binary mixtures. The
type and location of pinch points generally depends on how the components are distributed
between the distillate and bottoms product. Based on this, we will define two types of systems:

Type I systems, for which all the components in the feed are present in appreciable amounts
in both distillate and bottoms. In this case, there are no clear key components or, alternatively,
component 1 (the most volatile) would be the light key and component n would be the heavy
key. This occurs whenever the mixture has a very narrow distribution of boiling points. For this
type of systems, rigorous calculations show that the pinch point region generally occurs around
the feed tray, similar to the behavior of the binary system illustrated in Figure 40. However,
Type I systems are not very common in practice.

Figure 40. Minimum reflux condition in a binary system with the pinch point at the feed: this
separation would require infinite theoretical trays in a region around the feed tray.
 53

Figure 41. Minimum reflux condition in a binary system with the pinch point in the
rectifying section: this separation would require infinite theoretical trays in a region within the
rectifying section.

Type II systems, for which light and heavy keys are clearly defined, so that the distribution of
components follows the scheme in Figure 37. For these systems, rigorous calculations show that
there are two pinch-point regions: one in the rectifying section and one in the stripping section,
as illustrated in Figure 42. The two pinch point regions appear simultaneously as the minimum
reflux is approached. In what follows, we will restrict our analysis to minimum reflux conditions
for Type II systems.
 54

Figure 42. Minimum reflux condition for multicomponent Type II systems. Two pinch points
occur simultaneously. In the region between the feed and the rectifying section pinch point,
components HK+1 to n are completely removed as we move upwards. In the region between the
feed and the stripping section pinch point, components 1 to LK-1 are removed as we move
downwards.

Within each of the pinch-point regions in Figure 42, each tray does not achieve a significant
separation since the compositions of vapor and liquid fed to the tray are near equilibrium.
Therefore, in the pinch-point regions, we can assume, as an approximation, that vapor and liquid
mole fractions are not changing from tray to tray, as shown in Figure 43.
A mole balance in component i for the control volume CV 1 in Figure 43 yields

Vy ∞i = Lx ∞i + Dx Di (105)

Since y∞i and x∞i are at equilibrium, we can state

y ∞i = K i (T∞ ) x ∞i (106)

where T∞ is the temperature of the rectifying section pinch-point region (which does not vary
from tray to tray within the region).
 55

Figure 43. Minimum reflux condition for multicomponent Type II systems. Within each
pinch-point region, compositions do not change from tray to tray since equilibrium has been
achieved and N→∞ within each region.

The equilibrium relation can be expressed in terms of the relative volatility of component i
with respect to a reference component r (usually taken as the heavy key, in which case r=HK):
 56

Ki
α ir = (107)
Kr

so that equation (106) can be written as

y ∞i = α ir K r (T∞ ) x ∞i (108)

Solving for x∞i and substituting into equation (105) yields

Ly ∞i
Vy ∞i = + Dx Di (109)
α ir K r

We can express the molar flow rates of vapor and liquid in terms of the reflux ratio,

V = (1 + R min )D (110)

L = R min D (111)

Substituting these two equations into equation (109), and solving for y∞i leads to, after
manipulations,

α ir x Di
(1 + R min ) y ∞i = (112)
R min
α ir −
(1 + R min )K r (T∞ )

Let

R min
Φ= (113)
(1 + R min )K r (T∞ )

n
Now we sum equation (112) for all the components (i=1 to n). Recalling that ∑ y ∞i = 1 we
i =1
obtain

n α ir x Di
1 + R min = ∑ (114)
i =1 α ir − Φ

Note that the parameter Φ is the same for all components, but it is related to the unknown T∞.
A similar analysis for the stripping section pinch-point region (using a mole balance in CV 2,
Figure 43) leads to
 57

L n α ir x Bi
1− =∑ (115)
B i =1α ir − Φ

where

L
Φ= (116)
VK r ( T∞ )

and T∞ is the temperature of the stripping section pinch-point region.

Using the assumptions of constant molal overflow and constant relative volatilities,
Underwood showed, by performing an analysis of the zone between pinch-point regions,
including the feed tray, that a solution for which Φ = Φ = θ represented an exact link between
the regions (i.e., the temperatures in both regions must be such that the parameters defined by
equations 113 and 116 are the same). Knowing this, equations (114) and (115) can be expressed
in terms of the common parameter θ as follows,

nα ir x Di D
D+L = ∑ (117)
i =1 α ir − θ

α ir x Bi B
n
B−L = ∑ (118)
i =1 α ir − θ

Since the reflux ratio is not known, these two equations have three unknowns: θ, L and L
(assuming that the mole fractions of all components in distillate and bottoms are known).
However, adding these two equations and making use of the balances

F= D+B

Fx Fi = Dx Di + Bx Bi

L = L + qF

leads to

n α ir x Fi
1− q = ∑ (119)
i =1α ir − θ

which has only one unknown (θ). Once the value of θ has been found from this equation (see
below), the minimum reflux ratio can be calculated from equation (114), expressed as

n α ir x Di
1 + R min = ∑ (120)
i =1α ir − θ
 58

Underwood showed that equation (119) has multiple roots for θ. In fact, it is possible to show
that there is one root between any adjacent pairs of relative volatility values. The way to apply
equations (119) and (120) to determine the minimum reflux ratio depends on whether more than
two components (apart from the heavy and light keys) distribute between distillate and bottoms
product. The latter will happen, in general, when there are components with volatilities between
those of the keys. The rules, as determined by Underwood, are as follows:

Case 1. The keys are the only components that distribute in distillate and bottoms. This will
happen when: (1) there are no components with intermediate volatility between LK and HK, and
(2) the degree of separation is such that the rest of the components do not distribute (problems
might be encountered if, for example, the component LK-1 has a relative volatility that is very
close to LK). For Case 1, Underwood showed that, if r=HK (i.e., the relative volatilities are all
referred to the heavy key; note then that αHKr=1), the appropriate root of equation (119) lies in
the interval α LKr > θ > 1 . This root is evaluated and then used in equation (120) to determine
Rmin. Note that, in this case, all the distillate mole fractions are known.

Case 2. Other components, besides the light and heavy keys, distribute between distillate and
bottoms. If the number of these other components is m, then equation (119) is used to find m+1
roots, whose values must be between the relative volatilities of the components that distribute
and the light and heavy keys. For example, if only one additional component distributes (m=1),
and the relative volatility of this component (let us call it component number c) is between those
of the keys, i.e., α LKr > α cr > 1 , then two roots θ1 and θ2 of equation (119) are found, such that

α LKr > θ1 > α cr > θ 2 > 1

Then, equation (120) is written m+1 times (one for each root), and the resulting system is solved
for Rmin and the xD's of the m distributing components. This procedure must assure that
n
∑ x Di = 1 , which can be achieved by adjusting the total flow rate of distillate.
i =1

(3) Number of equilibrium stages - the Gilliland correlation

An estimate for the number of theoretical trays required for a given separation can be found
by using an empirical correlation developed by Gilliland. The correlation gives an estimate for
the number of equilibrium stages (Ns) required for a given separation, and it is based on a wide
range of conditions. In graphical form, the correlation is shown in Figure 44.
 59

Figure 44. Gilliland's correlation for the estimation of the number of equilibrium stages (Ns)
required for a specific separation.

The total number of stages required depends on the degree of separation. Note that the
independent parameters required to evaluate Ns from Gilliland's correlation do not have an
explicit reference to the degree of separation. However, the presence of Nmin and Rmin in the
correlation indirectly quantifies the degree of separation, since these values are affected by the
specifications set in formulating a design problem. The fact that Nmin and Rmin are required to
calculate Ns gives more relevance to Fenske's equation and Underwood's method.
An analytical equation that fits the curve in Figure 44 is also available. Let

R − R min
X= (121)
R +1

N s − N min
Y= (122)
Ns + 1

Then, Gilliland's correlation is fitted by the following equation

 1 + 54.4X  X − 1 
Y = 1 − exp    (123)
 11 + 117 . 2 X  X 

Finally, it is important to mention that the number of equilibrium trays can be obtained from the
 60

number of equilibrium stages as N=Ns−1 for a column with a total condenser and a reboiler, and
N=Ns−2 for a column with a partial condenser and a reboiler.

(4) Optimum feed tray location - the Kirkbride correlation

An empirical correlation proposed by Kirkbride can be used to estimate the optimum location
of the feed in a multistage distillation process with clearly defined light and heavy key
components. The correlation expresses the ratio between the number of stages in the rectifying
section (above the feed), NR, and the number of stages in the stripping section (below the feed),
NT, as follows,
0.206
NR x x  B
2
=  FHK  BLK   (124)
NT  x FLK  x DHK  D 
 

Once the ratio is calculated from this equation, the number of stages in each region is determined
by knowing that Ns=NR+NT. The number of theoretical trays in the stripping section will be NT-1
(one stage allowed for the reboiler), and the number of theoretical trays in the rectifying section
will be NR for a total condenser, or NR-1 for a partial condenser.
 61

5. BATCH DISTILLATION

The distillation processes studied in the previous sections were all based on continuous flows
and steady state conditions. Continuous processes are appropriate when the amounts of materials
to be handled are relatively large. Certain applications require distillation separations that are
more efficiently carried out in small-scale batch processes. This is especially common in
industries with relatively low outputs, such as pharmaceuticals and other biotechnology
applications, or in laboratory-scale separations. The most typical example of such small-scale
distillation processes is based on the scheme presented in Figure 45. A liquid is placed in a still
pot and heated to its boiling point. The vapor that is generated by boiling is continuously passed
through a total condenser and the condensed liquid is collected in a product reservoir. This
process is an example of batch distillation and it is usually known as differential distillation.

Figure 45. Differential distillation set up.

Differential distillation
In this section we will analyze the differential distillation of binary mixtures. We will
represent compositions once again in terms of the mole fraction of light component in the
mixture. Before analyzing quantitatively the process, let us consider qualitatively how the
separation is achieved.
As liquid evaporates from the still pot, the amount of liquid remaining will decrease with
time. Let W(t) represent the total number of moles of liquid in the still pot at time t. This
parameter will decrease with time from an initial value W0 that corresponds to the total moles of
 62

liquid initially loaded in the pot. As the liquid boils, the vapor that is generated will be richer in
light component than the remaining liquid; i.e., yD>xw. In fact, the liquid will always be at its
boiling point, and the vapor will be saturated. As time passes, the mole fraction of light
component in the still pot will decrease with time. The liquid remains at its boiling point during
the whole process. If the system pressure is kept constant, this means that its temperature must
increase with time, because the boiling point of the liquid increases as its composition in light
component decreases. The progression of the process is shown schematically on a T-x-y
equilibrium diagram in Figure 46.

Figure 46. Progress of a differential distillation process with respect to time. Initially, the
liquid in the still contains a mole fraction of light component xw0 at a temperature Tw0. The initial
vapor produced will be highly concentrated in light component (yD0). As time passes, the liquid
will always remain at its boiling point so that xw will decrease with time and so will yD. To keep
the liquid boiling, Tw will increase with time.

Let us now analyze the process quantitatively. First, we will assume that the condenser is
operating at steady state. Total and light component mole balances in the condenser than indicate
that the distillate rate D (moles per unit time) is equal to the rate of vapor generated: D=V and
xD=yD. Next we perform unsteady mole balances selecting as control volume the liquid in the
still pot. A total mole balance is
 63

dW
= −D (125)
dt

and a mole balance in light component yields

d ( Wx w )
= −Dy D (126)
dt

Substituting D from equation (125) and expanding the derivative on the left-hand side leads to

dW dx dW
xw +W w = yD
dt dt dt

Multiplying by dt and rearranging yields

Wdx w = ( y D − x w )dW

Since yD represents the composition of the vapor that is at equilibrium with xw, yD is a function
of xw, so that the previous equation can be separated and integrated as follows

xw W
dx w dW
∫
xw 0
= ∫
y D − x w W0 W

where the lower limit of integration corresponds to the initial state (t=0). Integrating yields

xw
dx w W
∫ yD − x w
= ln(
W0
) (127)
xw0

This equation is known as the Rayleigh equation. By expressing yD in terms of xw using the
appropriate equilibrium relation for the system being distilled, the left-hand side integral can be
found, and the result yields an equation that expresses the composition of the liquid remaining in
the still (xw) in terms of the amount of liquid that remains (W). The amount and composition of
collected distillate at that point can be found from mole balances between the initial state of the
system and a given time, as follows. Neglecting material retained in pipes, connections and
condenser, a total mole balance yields

W0=W+S (128)

and a mole balance in light component is

x w 0 W0 = x w W + x s S (129)

So, for a given amount of liquid remaining (W), S can be found from equation (128) and xs from
 64

equation (129).
The evaluation of the integral in equation (127) is done by using the y vs. x equilibrium
relation. For example, if the relative volatility is a constant (independent of temperature), we can
apply equation (12) as follows,

αx w
yD = (130)
1 + (α − 1) x w

Substituting this relation in equation (127) yields

x
1 w
[1 + (α − 1) x w ]dx w W
α −1 x ∫ x w (1 − x w )
= ln(
W0
) (131)
w0

The integrand can be decomposed in partial fractions as follows:

1 + (α − 1) x w A B
= +
x w (1 − x w ) xw 1− xw

which leads to

1 + (α − 1) x w = A(1 − x w ) + Bx w

Equating independent terms and terms proportional to xw we find:

1=A

(α − 1) = −A + B ⇒ B=α

so that

1 + (α − 1) x w 1 α
= +
x w (1 − x w ) xw 1− xw

Substituting into equation (131) we find

x x
1 w
dx w α w
dx w W
α −1 x ∫ xw
+ ∫
α −1 x 1− x w
= ln(
W0
)
w0 w0

or
 65

1  xw 1 − x w0  W
ln( ) + α ln( ) = ln( ) (132)
α − 1  x w0 1− xw  W0

This equation can be further rearranged to give

1 α
x 1 − x w 0 α −1 W
( w ) α −1 ( ) = (133)
x w0 1− xw W0

For more complex y vs. x equilibrium relations, the integral in equation (127) could be done
numerically using methods such as the trapezoidal rule (see Chapter II, equations 129 and 131).
As an example, consider the differential distillation of a mixture n-pentane/n-heptane at 1
atm, for which the relative volatility of the mixture is approximately equal to 5.4. An initial load
of 100 moles of a mixture containing 60% n-pentane is placed in the still pot. In this case, we
have:

W0=100 mol
xw0=0.6
α=5.4

Equation (133) can be written as

1 5.4
x 0.4 4.4
W = 100( w ) 4.4 ( )
0.6 1− xw

This relates the instantaneous value of xw to the remaining amount of liquid, W. The
corresponding collected distillate is given by equation (128):

S=100-W

and the composition of the collected distillate is found from equation (129):

60 − x w W
xs =
100 − W

Using the previous equations, we generate the plot shown in Figure 47. Note that, due to the high
relative volatility, the separation obtained at the beginning of the process is relatively good: the
distillate initially contains almost 90% n-pentane. Towards the end of the process, as most of the
mixture is vaporized, the composition of the collected distillate drops and it eventually reaches
the initial composition of the still pot (60% n-heptane) when everything is vaporized. If the
process is stopped when half of the mixture has been distilled (W=S=50 moles) we obtain the
two products: xs=0.84 and xw=0.36.
 66

1
0.9
xs
0.8
0.7
0.6
0.5 xw
0.4
0.3
0.2
0.1
0
0 20 40 60 80 100
SS
S (moles)

Figure 47. Mole fraction of n-pentane in still pot (xw) and collected distillate (xs).

Differential distillation with feed of heavy component

An alternative mode of operation in differential distillation consists of a continuous feed of
heavy component to the still pot. A control system is used to maintain a flow rate of pure heavy
component that equals the instantaneous flow rate of distillate that is being produced. This keeps
the number of moles constant in the still pot (i.e., mass conservation implies that W=constant). A
mole balance for the light component is still given by equation (126), but now W can be taken
out of the derivative,

dx w
W = −Dy D (134)
dt

Let L be the molar flow rate of liquid fed to the still pot. In this case we have L=D, and these
flow rates are set by the rate at which the still pot is heated. A total mole balance on the reservoir
for collected distillate yields

dS
=D (135)
dt

Separating equation (134), replacing W=W0, and using equation (135) leads to

dx w 1
=− dS
yD W0

Integrating from t=0 to a generic time t yields

xw

∫
S dx w
=− (136)
W0 yD
xw0
 67

For systems with constant relative volatility, we can relate yD to xw by equation (130), which
leads to

xw
[1 + (α − 1) x w ]dx w
∫
S
=−
W0 αx w
xw0

Integrating and rearranging leads to

S 1  x  (α − 1)
= ln w 0  + (x w 0 − x w ) (137)
W0 α  x w  α

The composition of the collected distillate can be found from equation (129).