4

Distillation, Absorption and Stripping

.Jean-Paul Gourlia

4.1 Introduction
Whether to prepare the feed or to separate the effluents into usable products,
all industrial processes use separation operations. Their ultimate end is to iso-
late components while complying with constraints on the recovery ratio’ and
the composition expressed in mole, weight or volume fraction. A lot of differ-
ent technologies can be used: membranes, adsorption, absorption, liquid-liq-
uid extraction, supercritical fluid extraction, etc., all have their particular area
of application. Distillation alone, whose characteristics we will present in this
chapter, is the most frequently utilized. We could qualify distillation as the uni-
versal technique since, using a cascade of vapor-liquid or even liquid-liquid-
vapor equilibria we can separate almost all mixtures according to ranked
boiling point temperatures.
The operating principle of a distillation column is simple (see Fig. 4.1).
While running at constant pressure it displaces the phase equilibria by means
of a temperature gradient created by a cold source (the condenser) which
causes a flow of cold liquid moving downward in the column, and a hot source
(the reboiler) which generates a flow of hot vapor moving upward. To achieve
transfers between the two streams, heat and mass-transfer contact devices are
located inside a tower. The contact elements can consist of trays, whose basic
principle is to bubble vapor through a liquid retained by a damming effect on
the liquid phase (see Fig. 4.2a), or packing, which usually consists of a metal
latticework dispersing the two phases and providing a good exchange area
between the fluids circulating in counterflow (see Section 5.4). The reboiler
works with an outside fluid (steam, hot oil, process stream). Usually cooled by

1. The recovery ratio of a component i is the ratio between the partial flow rate of the
component in the product and its partial flow rate in the feed.
94 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPlNG

air or water, sometimes by a process stream, the condenser works in different

conditions (see Fig. 4.2b) depending on whether a light cut, called distillate is
desired in a liquid or vapor form. As soon as there are several coexisting
phases, their composition and flow rate are roughly speaking governed by
thermodynamic equilibria. For example, for some refining applications the
presence of noncondensable components in the feed leads to simultaneous
production of a liquid distillate and a vapor distillate. In other circumstances,
thermodynamic equilibrium leads to a liquid-liquid or a liquid-liquid-vapor
equilibrium. This is found for certain azeotropic distillations or in atmospheric
columns for crude oil distillation where stripping steam is injected (see
Chapter 5).
When the column runs continuously, the feed is let into the tower at a point
that depends on the characteristics of the feed. The feed is then separated into
a residue at the bottom of the column and a distillate at the top. The part
located below the feed inlet is called the stripping zone and its objective is to

Condenser

Reflux Distillate
*

=I-
Feed

I Trav

Vapor -4== . Liquid

I Reboiler

Residue

Figure
4.1 General flowchart of a distillation column with trays.
 Chapter 4. DISTILLATION,
ABSORPTION
AND STRIPPING 95

General flowchart
showing a tray.

- Liquid

Vapor

-
Reflux Reflux

Figure
4.2b Partial condensation and total condensation.

lower the light product content in the residue. The top of the column or recti-
fying zone is designed to eliminate heavy products from the distillate. This def-
inition immediately suggests that the stripping zone can be less efficient than
the rectifying zone if the quality requirements are lower for the residue than
for the distillate.
In batch operation, the reboiler also serves to hold the feed. The column is
therefore reduced to a rectifying zone and the quality of the distillate alone can
be controlled. In contrast, the residual liquid in the boiler will contain rela-
tively light products as can be shown with a simple material balance.
96 Chapter 4 DISTILLATION, ABSORPTION AND STRIPPING

If the basic principles are easy to understand, practical implementation

and comprehension of the phenomena observed are much less so.
Consequently we will endeavor to present a synthesis of the problems in order
to provide the reader with a series of readily usable implements.
A number of reference works are available in the literature and give theo-
retical and practical information on distillation. We suggest that our readers
consult the writings of Holland, King, Kister, Van Winkle and Wuithier which
are listed in the bibliographic references at the end of this volume.

4.2 The Basics of Distillation

4.2.1 Phase Equilibria
Even though Chapter 2 was totally devoted to this topic, a few reminders of the
basic principles are required to introduce the concepts indispensable for a
proper understanding of distillation.
Let us consider a closed vessel of variable volume such that it remains iso-
baric, filled with a binary liquid mixture whose composition is characterized
by the molar composition z of the component with the lower boiling point.
After heating the vessel, three characteristic domains can be observed:
1 . Below the temperature where the first bubble of vapor appears, the mix-
ture is entirely liquid; it is “subcooled”. The temperature is called the
bubble point temperature of the mixture.
2. Continuing to heat the vessel, the vapor fraction increases to the tem-
perature where the last drop of liquid disappears. This dew point tem-
perature of the mixture marks the end of the two-phase domain.
3. Above the dew point temperature, the vapor phase alone remains. It is
“superheated”.
In the two-phase domain, the molar composition of the liquid phase x and
the vapor phase y is determined by thermodynamic equilibria that can be
described by the following conventional equations:
~ = K , x for the light component
(1 - y) = K2 (1 -x ) for the heavy component
or:

Kl and K2 stand for the equilibrium coefficients, or K values, of each of the

components. As mentioned in Chapter 2, the most general expression is given
by the equation below:

(2.12)
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 97

The quotient PT/P is the K value obtained by applying Raoult’s law. This
way of expressing the K value is well suited when the pressure and the vapor
pressure of the component under consideration are low and when the solution
is ideal. The second quotient merges the corrective terms, generally close to
one, which depend mainly on the deviations from the ideal gas laws in the
vapor phase. Finally, the activity coefficient is crucial since it represents the
singularities specific to the liquid mixture. It depends on the nature of the com-
ponents, the temperature and the composition. The influence of the pressure
can usually be disregarded when pressure is close to atmospheric. This leads
to the following as an initial approximation:

(2.13)

By varying the composition of a binary mixture, the bubble and dew point
temperatures vary as indicated in Figure 4.3. The curves divide the diagram up
into three areas: to the left of the two phase region any mixture is entirely liq-
uid; to the right it is totally vapor. The part inside the region corresponds to
two-phase conditions. Any feed, such as the one represented by point F, is bro-
ken down into a vapor phase V of composition y and a liquid phase L of com-
position x. The respective quantities of the two phases are determined by the
lever rule. This example corresponds to a regular mixture. Other equilibrium
diagram types are found when the components form azeotropes or het-
eroazeotropes. This problem will be dealt with in the relevant section.
For the same temperature, the x and y abscissas represent the composition
of the liquid and vapor phases at equilibrium. Based on Figure 4.3, a diagram
showing the direct relationship between x and y can be drawn by eliminating
the temperature parameter. This yields the representation in Figure 4.4, which
additionally shows the influence of the ratio K,/K,, conventionally called the
relative volatility and written a.It can be seen that even at high values for rel-
ative volatility, the difference in composition between the two phases remains
small. It increases, however, with increasing relative volatility. When a is inde-
pendent of, or not very dependent on the temperature, the equilibrium curve
is described by Eq. 4.2:

The parameter a,which defines the relative value of the equilibrium coef-
ficients, or K values, is a key criterion in predicting the separating power of a
vapor-liquid equilibrium and even more so of a distillation column.

4.2.2 Separating Power of a Vapor0Liquid Equilibrium

and Distillation
The flash drum in Figure 4.5 is fed by a binary mixture F of molar composition
z for the light component. Its temperature is regulated by means of a heat
exchanger to vaporize (or condense) the feed partially in order to get a vapor
phase (V, y) and a liquid phase (15,x).
98 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

Isobaric phase diagram

L for a binary mixture
(the mole Fraction
refers to the more
volatile component).

I I I

0 0.2 0.4 0.6 0.8 1

Mole fraction
LIVE GRAPH
Click here to view

Liquid phase mole fraction (x)

Figure
Relationship between the equilibrium composition o f the liquid phase and that
o f the vapor phase for differentvalues of relative volatility Q = K,/K,.
 Chapter 4. DISTILLATION,
ABSORPTION
AND STRIPPING 99

In this type of process, the exiting vapor is considered at its dew point and
the liquid at its bubble point. More generally speaking, the two phases are con-
sidered at equilibrium. Taking the general rule for writing balances into
account (Eq. 4.3):
[input] + [production rate] = [output] + [accumulation] (4.3)
the principal equations describing how the equipment works are written as fol-
lows in the transient state:
Overall mass balance:
dML
F = L + V+ -
dt
Mass balance for each of the components:
d (MLx)
Fz=Lx+ Vy+ - dt

F(l- Z) = L ( l - x ) + V(l- y) + d [ML(1 - x ) ]

dt
In these expressions, the accumulation term was considered to involve the
liquid phase mainly.
The last partial balance does not provide any information since it is redun-
dant with the first two. The result is very general. A mass balance is estab-
lished with n equations of partial balance or an overall balance plus n - 1
partial mass balances, if n is the number of components.

Figure
4.5 Representation of a flash drum.
M L= liquid retention. M V = vapor retention.
100 Chapter 4. DISTILLATION, AESORPTDNAND STRllPplNG

Enthalpy balance:
d [ML h L]
Q + Fh, = L h L + VhV+
dt
Equilibrium equations:
Y=Ki(T,P,x,y)x
(1-y)=K,(T,P,x,y)(l-x)
In steady state conditions, the differential terms with respect to time are
zero, thereby reducing the previous system to:

F=L+V (44.4)
Fz=Lx+ v y (4.5)
Q + Fh, (TF, P, Z) = LhL(T,P, X) + VhV(T,P, y ) (4.6)
Y = K (~T , P , x , y ) x (44.7)
(l-y)=K,(T,P,x,Y)(l-x) (4.8)
These five equations are insufficient in number to allow identification of all
the parameters (Eq. 4.10). Under-representation means that five values, con-
ventionally called the degrees of freedom of the system, must arbitrarily be
fixed. In this specific case it is necessary to assume we know the feed condi-
tions, the flash drum pressure, and another condition to be chosen between
the vapor flow rate V, the liquid flow rate L,the thermal load Q, the tempera-
ture Tor the composition of one of the two effluents (x or y). For example the
bubble or dew point conditions for a given pressure and feed correspond to
the temperatures that reduce the vapor and liquid flow rates respectively to
zero. They are expressed in the followingequations:
Bubblepoint
K1 ITbubble, p, 212 + K2 [Tbubble p, z ] (1 -2) - 1 = 0
Dewpoint

These relations can be generalized to cover multicomponent mixtures and

are written:

(4.10)

Based on these equations, it can be advantageous to study the separating

power of a cascade of separator drums such as the ones described by the
 Chapter 4 DISTILLATION, ABSORPTION AND STRIPPING 101
LIVE GRAPH
Click here to view

preceding figure. Figure 4.6 illustrates this calculation for a binary mixture of
equimolecular composition characterized by a relative volatility of 1.5. In
order to have a mole fraction of the more volatile component of approximately
0.7 in the vapor phase, the mass yield for vapor phase is only 8%.This value
corresponds to a mole fraction of the more volatile component of 48%in the
liquid phase. The only way to increase this yield is to recycle the liquid phases
as shown in Figure 4.7A. Although this configuration increases the recovery
yield of the light component, it gives no control of the composition of the liq-
uid phase which remains related to the working conditions of the feed drum
(these conditions can be readily found by means of a simple overall material
balance). To achieve control, a cascade of separator drums where the vapor is
recycled is placed below the feed inlet (Fig. 4.7B). This configuration summa-
rizes the operating principle of a distillation column (Fig. 4.8).
In order to separate a feed into two products of specified quality, a cascade
of vapor-liquid equilibrium drums or theoretical trays must be utilized. The
flow rate of the two phases that circulate in counterflow can be regulated
either by heat exchangers placed in each drum or at each end of the column.
In this case the condenser creates the liquid reflux and the reboiler the vapor
reflux. The part located between the feed inlet and the top of the column, or
rectifying zone, eliminates the heavy components from the light product, or
distillate. The area located between the feed inlet and the bottom of the

0 20 40 60 80 100
Vapor phase yield (%)

Figure
4.6 Variations in the composition of the vapor phase versus the number of drums
in cascade (on the vapor phase). Feed made up of a binary equimolecular
102 Chapter 4. DISJILLAJION,
ABSORPJION
AND SJRlPPlNG

t >
t > 1/
t > 1/
F
----f
> .c A

- Figure
4.7 Cascade of flash drums.
A. Enriching the vapor phase in the light component.

column, usually called the stripping zone, strips the bottom product, or
residue, of light components. We will see later on that an important operating
parameter for each of the equilibrium stages or drums is the ratio between the
liquid and vapor flow rates. It is called the internal reflux ratio to differentiate
it from the external reflux ratio created by the condenser of the column and
usually measured.
One of the important conclusions from this brief introduction is the limited
separation efficiency of separator drums. Whatever the value of the relative
volatility, it is impossible to remove all traces of heavy product from the vapor
phase and inversely. This must also be applied to distillation, where it is
impossible to obtain very clearly defined cuts. Even if concentrations remain
very low, the distillate is always polluted by heavy products and the residue
by light products.
 Chapter 4. DISTILLATION, ABSORPTION
AND STRIPPING 103

Cascade of flash drums

combining enrichment
of the vapor phase in
the light component
and enrichment of the
liquid phase in the
heavy component.

4.2.3 Continuous Binary Distlllation.

The McCabe and Thiele Method
4.2.3.1 Characteristic Distillation Column Balances
for a Binary Mixture
As we have seen earlier, the basic configuration of a distillation column con-
sists of a series of theoretical separation trays where a vapor generated by a
reboiler and a liquid, moving downward in the column by gravity and gener-
ated in the condenser, circulate in counterflow. Balances can be established
for the column a s a whole, for a tray taken individually or for an operating zone
in the column. Convention dictates numbering the trays from the top down for
the rectifying zone and from the bottom up for the stripping zone, with the
condenser and reboiler numbered 1 and 1’ respectively. The balances can be
written as follows:
104 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

Balances for the whole column

Overall mass balance
F=D+R
Partial mass balance
FZ = D x , + Rx,
Enthalpy balance
Fhp + Q R = D h b + R h k + Qc
Balance for a tray in the rectifying zone
Overall mass balance
vN+l+ LN-1 = VN + LN
Partial mass balance
VN+IYN+I + LN-IXN-I = VNYN+ LNXN
Enthalpy balance
VN+ih;+, + LN-lhk-1 = VNh: + LNhk
Equilibrium relations
Y N = Kl.NxN
- YN) = K2,N (l - xN)
Balance by zones
a) Rectifying zone (defined by the envelope surrounding the condenser and
tray N)
Overall mass balance
VN+l= D + LN
Partial mass balance
VN+IYN+I = D X D+ LNXN
Enthalpy balance
VN+1h&+1= D h b + LNhk + Qc
Equilibrium relations
Y N = KI,NxN
c1 - Y N ) = K2,N(1 - xN)
b) Stripping zone (defined by the envelope surrounding the reboiler and
tray N’)
Overall mass balance
I!,’,,+, = R + VN,
Partial mass balance
L‘,,+1 xN’+1= RX, + VN*J”,
Enthalpy balance
L‘,t+lhfir.+l= Rhk + V’,,h$ + QR
Equilibrium relations
yNt = K I ~ , X N *
(1 -y~t)= K ~ A N- X’ N( .~>
 Chapter 4. DISTIllATION, ABSORPTION AND STRIPPING 105

The preceding equations in conjunction with the ones that are required to
estimate the K values and the enthalpies must be solved simultaneously in
order to identify the way in which the column runs.

4.2.3.2 The McCabe and Thiele Method

So as to avoid the above-mentioned difficulty, the simplified McCabe and
Thiele method (1925) is based on the assumption that liquid and vapor flows
are constant in each section of the column. This assumption is known as the
Lewis hypothesis (1922) and means that the enthalpy balance is disregarded,
considering that the heats of vaporization of the components are equal and
that the energy contained in the form of sensible heat is negligible. Such an
assumption is acceptable only if the substances distilled are very close to one
another (isomers, for example). It is not applicable when the enthalpy proper-
ties of the components are significantly different (by way of an example, the
molar heat of vaporization of water is twice that of ammonia). In order to
adapt the McCabe and Thiele method to all cases, some authors have pro-
posed to work with pseudocompositions defined by standardizing the molar
compositions by latent heat of vaporization.
The advantage of the hypothesis proposed by McCabe and Thiele is that it
simplifies the balance equations by zones. If we call the reflux ratio r, i.e. the
ratio LID, and the stripping ratio s, i.e. the ratio V / R ,they are transformed into
linear relations linking the composition of tray Nto that of tray N+ 1. The prop-
erty of these straight lines is that they pass through point (xD,xD)representa-
tive of the distillate and through point (xR,xd representative of the residue
respectively.
The parameters r and s, which define the characteristics (slope and ordi-
nate at the origin) of the operating lines of each of the sections of the column
are dependent and connected by the enthalpy balance of the column: when
the reboiler output increases and s therefore increases, it is necessary to
increase the flow rate L in order to maintain the balance (r increases).

I Rectifying zone

Partial mass balance

I Stripping zone

Partial mass balance

VyN+l= Dx, + LxN (4.1 1) L”X,y+l = RxR + V‘yW (4.13)

Operating line Operating line

r 1 1 S
YN+l = r+l x N + - X, (4.12)
r+ 1
X”+1 = - xR+
s+ 1
- YN,
s+ 1
(4.14)
106 Chapter 4. DISTILLATION, ABSORPTION
AND STRIPPING

We are now going to use these recurrence relations and the equilibrium
laws, or their graphic representation at least, to describe the way a total con-
denser column works (Figs. 4.9a and b).
The overhead vapors coming from tray 2 and arriving in the condenser
have the same composition as the distillate as condensation is total (xD= y2).
From this representative point on the diagonal, a horizontal straight line seg-
ment is drawn so that it intersects the equilibrium curve at point (x2,y2), r e p
resentative of the vapor-liquid equilibrium of tray 2. A vertical straight line
segment is used to intersect the operating line at point (x2,y3). Another hori-
zontal segment intersects (x3,y3), representative of tray 3. The step-wise con-
struction continues as shown in Figure 4.9a. It should be noted that in this
configuration the condenser does not bring about any vapor-liquid equilib
rium and is not accounted for as a tray.
The same type of construction is built from a point describing the residue.
Note that the beginning of the construction is different from the one in the rec-
tifying zone. In the configuration indicated in Figure 4.9b, X , = x l # .
To get y,., a vertical segment must be drawn from the point on the diagonal
(x,, xR>.From the point representative of the first tray, we intersect the reboil-
ing operating line, which provides x2-and so on.
The feed tray marks the limit between the rectifying and the stripping
zones. Theoretically, the two constructions described above converge on the
feed inlet tray whose operation must not be thrown off by injection of the feed.
In other words, the molar composition of the liquid and vapor flows from the
feed tray must be identical to the composition of the liquid and vapor phases
of the feed. This allows the following mass and enthalpy balances to be writ-
ten (Fig. 4.10):
F = FL + Fv
FZ = FLxF + FvyF
L'=L+FL
V=V'+Fv
FhF+ LhL+ V'hV = L'hL + VhV
And from these expressions:

(4.15)

It can be deduced that the operating characteristics of the feed inlet tray
(xF,yF) and the point defining the composition of the feed (z, z) are on a
straight line, the feed line, with the following equation:
y = p x + (1 - p ) z (4.16)
Table 4.1 defines the value of the slope p according to feed conditions.
This table requires some comment. When the feed is subcooled or super-
heated, the liquid and vapor flow is disturbed since the feed inlet tray
becomes a direct contact exchanger that allows the two phases to be placed
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 107

L2
0.9

I
I I I
I
I
I, ,I I 0.5
0.5 XN X N - ~ 0.8 X2 0.9 XD=X~ 1
Liquid phase mole fraction

L,x, DJ,

1
Figure
4.9a McCabe and Thiele graphic construction. Rectifying zone.
108 Chapter 4. DISTILL4TlON, ABSORPTION
AND STRlPPlNG

I
I
I
I
I
I
I
I
I

0 xR '2' 0.3
,
0.4
I
XNm
-
Liquid phase mole fraction

Figure
4.9b McCabe and Thiele graphic construction. Stripping zone.
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 109

Convergence
the three stra Feed straight line

Feed tray

- Figure
4.10 Feed inlet tray conditions.

Enthalpy nature Variation in internal

Value of p flow rates
of the feed
~

Subcooled liquid hF< h L L'>L V < v'

Liquid at bubble point hF=hL L'=L+F v= V'
Partially vaporized feed hL < h, < hV L'= L + F~ V = V ' + F v
((p = vaporized fraction)

Vapor at dew point hF=hV L'=L V=V'+F

Superheated vapor hF > hV L'<L V > v'

1 14.1 Value o f the slope o f the feed line according to the enthalpy nature o f the feed.
110 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

in the normal operating conditions of a distillation column (liquid at its bub-

ble point and vapor at its dew point). This disturbance leads to irreversibili-
ties and affects the efficiency of the column. Inversely, the best feed conditions
occur when the feed is in vapor-liquid equilibrium conditions. However, this
basic principle can be called into question because of process or thermal inte-
gration constraints.
As we have seen earlier, once the feed characteristics are known (compo-
sition and enthalpy content) and the product specifications are stipulated, we
can define the operating conditions of the rectifying and stripping zones after
having chosen a reflux ratio (which defines the stripping ratio), starting from
the top and from the bottom of the column. However, for the calculation to be
valid, the conditions of convergence on the feed inlet tray must be complied
with. They can be summed up as follows: the operating lines in the rectifying
and stripping zones and the feed line must intersect at a characteristic tray
operating point obtained by the stripping and rectifying zone construction
(Fig. 4.11).
For this particular example, 3 theoretical separation stages are necessary
for rectifying and 5 in stripping (including the reboiler), plus the feed inlet tray.
Beyond the assumptions mentioned earlier, note that the McCabe and Thiele

z = x,

Figure
Location o f the feed inlet tray. In the case pre-
sented, x, = z since the feed is liquid at its bub-
ble point. Rectifying zone: 3 trays + total
condenser.
Stripping zone: 4 trays + reboiler. The feed inlet
tray is noted 5.
 Chapter 4. ~ I S J I L U T I O N ,ABSORPTION AND STRIPPING 111

method can not be used to fix the number of trays required for a separation
beforehand. This characteristic of the column, which can be used to define
investments, is a consequence of the choices made about operating condi-
tions, which fix the operating costs. Despite the strict convergence criterion
which validates each calculation, there is no single solution to a given prob-
lem. There are a number of couples of reflux ratio/number of theoretical trays
that can solve the problem. When a column is designed several cases must be
worked out to identify the economic optimum. A double iterative calculation
is required if an existing column’s operating conditions are to be readjusted.
The first ensures convergence of calculations while the second deals with the
search for a number of theoretical trays equal to the value identified for the
unit.

4.2.3.3 Influence of the Reflux Ratio.

Concept of Minimum Reflux and Total Reflux
Figure 4.12 represents two case studies A and B which correspond to increas-
ing reflux ratios. When rincreases, the slope of the operating line approaches 1.
It can clearly be seen that a tray’s separation work, defined as the difference
in concentration between the incoming and the outgoing liquid, increases as
the reflux ratio does. Therefore, for a fixed number of trays the product qual-
ity improves as the reflux ratio increases. Symmetrically, the number of trays

1 Figure
Influence o f the reflux ratio on the number o f
trays for the same performance.
112 Chapter 4. IS TILL AT ION, ABSORPTION AND STRIPPING

-
Figure
4.13 Minimum number of trays (infinite reflux).

required to reach a specified quality decreases as the reflux ratio increases. In

extreme conditions where the reflux ratio is very high or infinite, the number
of theoretical stages is minimal. In this case the operating lines in the rectify-
ing and stripping zones are the same as the first bisecting line in the equilib
rium diagram (Fig. 4.13). The value of the minimal number of trays is defined
by the Fenske equation (1932), which has a very general application. It will be
demonstrated in the section on multicomponent distillation.

In this equation, Nminstands for the minimum number of vapor-liquid equi-

librium stages in the column.
Inversely, when the reflux ratio is reduced, the number of trays approaches
infinity. In what conditions does this situation occur? The tray construction in
the McCabe and Thiele method shows that the separation work of the trays
decreases with the distance between the equilibrium curve and the operating
line. When the distance becomes zero, an infinite number of stages is required.
Figure 4.14 illustrates the most general conditions for the existence of minimal
reflux. The pinch point is created by the intersection of the feed line and the
equilibrium curve. When the reflux ratio is chosen such that the rectifying line
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 113

intersects the equilibrium curve at the pinch point, the minimal reflux condi-
tion has been reached. It can be written as follows:

(4.18)

If the feed is at its bubble point, yFis related to X, by the general equation
of the equilibrium curve below:

a relationship between rmin,a,X, and z is obtained.

In other circumstances the pinch point is due to an irregularity in the equi-
librium curve (Fig. 4.15). Here the minimal reflux is defined by the physical
characteristics of the mixture that is being separated.
A basic property of distillation has just been highlighted: column perfor-
mance is defined by a couple of parameters, the reflux ratio and the number of
theoretical trays.
Accordingly, when the operating conditions of the process are defined
(feed composition and vaporized fraction, product composition), the reflux

Z xD

14141 Minimal reflux ratio (infinite number of trays).

The feed line has a slope p = m since the feed is
here at its bubble point.
114 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

K \
Operating straight I
line defining minimal
I
I
I
I
I

1
1
Figure
Minimal reflux ratio
with an equilibrium
curve exhibiting a con-
reflux ratio cave shape.
I
I
I
I
I
I
I I I I I

Z xD

ratio and the number of theoretical trays are related by a univocal relation lim-
ited by two extreme conditions:
for an infinite reflux ratio (consequently infinite operating costs), the
number of trays is minimal;
for an infinite number of trays (infinite investment), the reflux ratio is
minimal.
During the design phase, the objective is to search for the best compromise
between investment and operating costs by identifying the best pair of param-
eters. The optimum is usually considered to be approximated by the following
rules of thumb:
r = 1.2 rmin
N = 2 Nmin

4.2.4 Limits of the McCabe and Thiele Method.

Other Calculation Methods for Binary Distillation
The McCabe and Thiele method’s limits of applicability are inherent to the
underlying hypotheses which assume that the liquid and vapor flow is con-
stant in each zone. However, even for such trivial separations a s between
isopentane and n-pentane, variations in traffic of 10% are often found.
Furthermore, the method can be used only for a binary mixture, seldom found
 Chapter 4. DISTILLATION, ABSORPTION
AND STRIPPING 115

in the industry, and this severely limits the area where it can be used as an
implement in calculating columns. Additionally, it has become so easy to find
distillation column simulation software for PCs or work stations that it is out-
dated to use limited methods.
Yet the McCabe and Thiele method offers the advantage of giving a graphic
representation of the distillation process. This advantage can be exploited to
diagnose the operation of a column and analyze the relevance of design
choices. Other authors have used this simple tool to create grey-box type
models for predictive column control.
When the Lewis conditions are no longer applicable at all (mixture of water
and ammonia, for example), it is possible to use the Ponchon (1921) and
Savarit (1923) method. It uses both the enthalpy and the equilibrium diagrams
to build the step-like construction characteristic of the operation of each of
the zones according to a procedure similar to McCabe and Thiele’s. Here the
operating lines have become operating curves.

4.3 Theoretical Trays, Real Contacting Equipment

The preceding development assumed that contacts between the liquid and
vapor were perfect and that thermodynamic equilibrium was achieved
between the two phases. This assumption implies that the residence time on
the trays is very long, or that the coefficients of mass and energy transfer are
very large. Reality is, of course, a different story. The technology used limits
interfacial areas, mass and heat transfers and residence time, so contact effi-
ciency is also reduced. Even though the subject was developed in Chapter 3,
it is still advisable to briefly go over the concepts required to understand the
phenomena involved in general and exploit simulation software efficiently.
However the contact may be implemented (trays or packing), the basic
principle consists in dispersing the liquid and vapor phases in order to maxi-
mize the interfacial area and facilitate mass and heat transfers. From the stand-
point of phenomenology, contacting equipment operation is often represented
by the double film theory. It assumes (Fig. 4.16) the existence of two continu-
ous phases separated by an interface. Boundary layers are formed for each of
the two continuous phases where mass transfers take place by diffusion, and
heat transfer by conduction. The values of transfer coefficients depend on the
physical properties of the phases but also on the thickness of the layers,
which is in turn dependent on hydrodynamic conditions. The species are also
considered to be at thermodynamic equilibrium at the interface.
If a slice of volume dv is considered in the model in Figure 4.16, the balance
equations for one of the components of the mixture are written:
V(Yi + d ~ ; =) VY;+ k;,vQCYiinterface - Y ; ) do (4.19)
L ( x ~- dxi) + k i , , ~(xi - xPterface)
do = L x ~ (4.20)
116 Chapter 4. DISTILLATION,
ABSORPTION
AND STRlPPlNG

where:
ki, is the film transfer coefficienk2on the vapor phase side (mol/m*.s)
kj, is the film transfer coefficient, on the liquid phase side (mol/m2.s)
Q is the interfacial area (m2/m?
L are the vapor and liquid molar flow rates (mol/s)
du is the element of volume under consideration (m?

Interface

Liquid phase, x,

;ilm on vapor side Film on liquid side

I
I
I

I
I
I

I
I
I /

\
\
I
\ I '
\
\ I I'
-
Figure
4.16 Transferof mass between phases. Double film model.

2. To express the transfer coefficient in m/s, we have k; = k v q / p v and k l = KLK/pL,with

= mean molecular mass (kg/mol) and p = density (kg/m3>.
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 117

Since there is no mass accumulation at the interface, the transfer flows

must be equal. As a result:
Ni (mol/(m3.s)) = k , , Q (yjnterface - yi) = kj,,a(xi - xj”terface)
= X;,va 0, -y;) (4.21)
with:
Ni the flow of component i per unit of volume (mol/m3.s)
the global mass transfer coefficient, related to the vapor phase
(mol/(m2.s))
The last term of the equation defines the mass transfer rate based on the
description of the concentrations in the continuous phases. y: is the compo-
sition of component i in the vapor phase that would be in equilibrium with the
composition xi in the liquid phase.
This assumption means that we have two equilibrium equations y: = Kixi
and yPterface = Kixpterface.
The equality of Kvalues can be assumed for them.
The basic equations defining the mass transfer rate and equilibrium rela-
tion therefore give:
1 - 1 K;
--- +- (4.22)
%,v ki,v k,L
and:
dY;
-- - -
Xi,, Q

(4.23)
Yi* -Y; v
Their integration in the whole contact volume can be used to estimate the
behavior of the contacting equipment.
First of all, a number of comments should be made:
The transformation performed means that the vapor phase is in fact con-
sidered as the reference phase. When the laws on efficiencies are
deduced, they will characterize the difference between the real composi-
tion of the phase obtained with the contacting equipment and the theo-
retical composition defined by the thermodynamic equilibrium with the
liquid phase. This approach has become widely accepted in the technical
literature. Similar concepts could have just as well been developed by
taking the liquid as the reference phase.
As a reminder, the transfer coefficients and ki,Lare functions of the dif-
fusion coefficients and the thickness of the boundary layer. Their values
are therefore specific to each component. Accordingly, a differentiation
has to be possible between each component’s behavior in order to simu-
late distillation columns accurately. In the conventional approach
employing the concept of efficiency for the trays and that of height equiv-
alent to a theoretical plate (HETP) for packing, this difference is disre-
garded thereby causing sometimes large errors especially for absorption
columns. Figure 4.17 gives an example of variations in tray efficiency in a
column for different components involved in the separation.
118 Chapter 4. DISTILLATION,
ABSORPTION AND STRIPPING LIVE GRAPH
Click here to view

1.05

1.oo

0.95

0.90
=-.
c
0
._
(I)
0
$ 0.85 BTX,
P
t-
0.80

0.75

0.70

0.65
10 20 30 40 50
Tray number

1 Figure I
1 1 4.17
Tray efficiency for different components (curves smoothed after Biddulph,
1977).

In a relatively wide operating domain the transfer coefficients are almost

constant. In contrast, when the liquid or vapor flow goes beyond admis-
sible limits, malfunctions of hydrodynamic origin occur causing separa-
tion performance to deteriorate rapidly. With respect to both design and
operation, distillation column operating conditions must be held within
the limits defined by Figure 4.18 in order to get satisfactory performance.

4.3.1 Trays
4.3.1.1 Local Efficiency
As shown in Figure 4.19, the gas and the liquid meet in cross current. The liq-
uid stream coming from the inlet weir sweeps over the vapor stream and
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 119

>
8 I Entrainment I
Satisfactory
~ Dump operation Flooding
zone

- Figure
4.18 Column operating domain.
Dump point: the liquid flow rate is not high enough to maintain sufficient liq-

comes out the other side of the tray by the exit weir. At a specified point in the
active zone, an efficiency can be defined by integrating the preceding equa-
tions over the total height of the liquid.
If we assume that the liquid phase in the volume under consideration is
thoroughly mixed and that the vapor phase is in plug flow, we have:
Y*p-Yp
Y*p -Yp+l
= exp (- XV7 a
0) = exp (-Nv) (4.24)

(With X' in m/s, I/' must be expressed in m3/s).

and:
Yp - Yp+l
= 1 - exp ( - A + ) =Eos
Y; -Yp+l

N, is called the number of transfer units, related to the vapor phase

Eos is the local efficiency
120 Chapter 4. DISTILLATION, ABSORPTION
AND STRIPPING

-1
Figure
4.19 Illustration of the con-
cept of local efficiency
(vapor phase in plug
flow when it goes
through the perfectly
mixed liquid phase).

The second equation represents the ratio of the real difference between the
composition of the incoming and exiting vapor phase and the theoretical dif-
ference defined by a vapor phase at thermodynamic equilibrium with the liq-
uid. This efficiency, for which an algebraic equation can be defined in
particular hydrodynamic conditions, therefore represents the local conse-
quence of transfer phenomena.

4.3.1.2 Murphree Efficiency (1925)

The preceding definition is not completely satisfactory as it does not describe
the overall working conditions of a tray. When the liquid phase goes through
the active zone, it is stripped of light products all along the way. If we assume
that the composition of the vapor yp+lis constant, the concentration potential
governing transfer decreases all along the active zone, thereby reducing local
efficiency. The overall tray efficiency (conventionally termed Murphree effi-
ciency) is consequently an average, weighted by hydrodynamic phenomena,
of the local efficiency over the whole operating zone or active area. By analogy
with local efficiency, Murphree efficiency is defined by the following equation:
Yi,p -Yi,p+l
Ei, MV = with yTp = K i P x i p (4.25)
Yt p -Yi,p+l

In this equation, xiprepresents the composition of the liquid exiting the

tray, yip+land yi,pstand for the average composition of the vapor phase enter-
ing and exiting.
This definition is not homogeneous since it compares a local value to a
mean value in the vapor phase. The efficiency can therefore by nature be
greater than 1. Lewis (1936) compared the consequences on this parameter of
different flow configurations (Figs. 4.20 and 4.21). It can be noted that the most
effective configuration is when the flows are of the plug flow type and the liq-
uid phase always sweeps the vapor phase in the same direction. This situation
 Chapter 4. DISTILLATION, ABSORPTION
AND STRIPPING 121

is perfectly logical since it prevents any remixing and follows concentration

gradients. It can also be seen that the efficiency reaches a value of 2 in ideal
conditions. The most unfavorable conditions of course occur when the flow
conditions are identical but the liquid phase sweeps the vapor phase in a back
and forth alternating fashion.

L Ii
Mixed
vapor

Case I Case IV
(Gautreaux and 0'Connell)

c-
L

Vapor in
plug flow

Case II Case 111

1 I
4'20
Flow configurations discussed by Lewis (1936) and Cautreaux and O'ConneN
(1955).
122 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

2.0

Figure
1.5
Influence OF local eFFi- 4.21
ciency on Murphree
efficiency in the three
cases studied by Lewis
(1936).

1 .o
0.5 1 .o
Local efficiency Eos

1
1.6
Figure
4.22 Influence of the number
L of mivers on Murphree
efficiencyin the
1.3 Gautreaux and
O%onnell model (1955)
(A = I, Eog = I).

1.o
0 6 12 18 24
Number of mixers
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 123

Gautreaux and O'Connell (1955) (Figs. 4.20 and 4.22) studied the influence
of remixing the liquid phase when the vapor phase is perfectly mixed. It should
be noted that here the direction of liquid phase flow has no importance. The
study is interesting because it is representative of the problems that crop up
with trays operating in the industry. It underscores the advantages to be
gained by reducing the dispersion of the liquid phase by means of systems
that guide the liquid. As for the vapor phase, sizing conditions are such that
the flow is extremely turbulent. To reduce remixing phenomena, barriers must
be included to separate the stream into various sub-streams.
Chan and Fair (1984) proposed a model that takes the dispersion of the
liquid phase into account and assumes a perfectly mixed vapor phase
(Eq. 4.26):

q= [; JZg- 11 (4.26)

A = absorption factor (see Eq. 4.43)

L2
Pe= -
144D,tL

D, = 0.013 u ~+ 0.025
. ~h, -~ 0.061

In these equations, L is the length of the liquid's path defined as the dis-
tance between the two weirs in feet, tLrepresents the residence time of the liq-
uid in seconds, h, is the height of the clear liquid expressed in inches, D,
stands for the turbulent diffusion coefficient in ft2/s and u is the velocity of the
vapor through the active zone in ft/s.

4.3.1.3 Overall Efficiency

Let us consider a real column with a known number of trays (Nr,J To simu-
late it using a method with theoretical stages, Nthtrays are required to fit the
model with reality. The overall efficiency of the column is identified by the sim-
ple ratio between these two values (Eq. 4.27). To be absolutely rigorous, this
parameter has no physical meaning since it integrates a large number of phe-
nomena. Estimating it experimentally may be biased by faulty design or any
uncertainty about the thermodynamic model (Fig. 4.23).
 124 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

Error in overall efficiency (“YO)

- 90

- 60

7 /
- 30

+ 90
5 10 15
Error in relative volatility a (“YO)

Figure
4.23 1 Error concerning the estimated value of overall efficiency versus uncertainly
about the value of the relative volatility of the two key components.

LIVE GRAPH 1
Click here to view

0.8

1
IFigure 1%
6
c
.$ 0.6

Relationship between
overall efficiency and
Murphree efficiencyac-
cording to the value o f
4.2‘
J: p 0.4

the absorptionfactorA 0.2

0 0.2 0.4 0.6 0.8 1

Murphree efficiency
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 125

However, in certain particular conditions, defined by the Lewis hypothe-

ses, the overall efficiency and Murphree efficiency are related by relation 4.28,
where A is the column’s absorption factor (Fig. 4.24):

In [ 1 + EM, (A - l)] AEoverall -1

Eoverall= or EMv= (4.28)
In A A-1

As shown in Figure 4.24, there is a significant difference between the two

types of efficiency.
Correlations have been proposed to define the overall efficiency empiri-
cally. We will mention the O’Connell correlation (1946) which was developed
for bubble cap trays, its application to other tray technologies is conservative:

Eoverall= 0.492 (pLa)-0.245 (4. 29)

with:
pL viscosity of the liquid (mPa.s)
a relative volatility of the two key components
As we will see later on, almost all present-day simulators can be used to cal-
culate distillation columns by integrating the Murphree efficiency for each of
the zones in the column. This option, which provides simulation tools with
greater accuracy, particularly for existing columns, remains limited by the
nondifferentiation of each of the components. Additionally, some ambiguity
may result when it is used, since engineers have become accustomed to rea-
soning in terms of overall efficiency. Among other examples, there is the fact
that the overhead temperature is different from the dew point temperature.

4.3.2 Packing
Assuming plug-type flow for the liquid and vapor phases, Eqs. 4.19 and 4.20,
which describe local processes, can be integrated over the whole height of the
packing and yield:

with:
Lo
Z=H
dY;
yi*-y;-
%,vQ sH
v
H the height of the bed of packing (m)
S the cross-sectional area of the bed of packing (m2>
i.e.:

H= - dY;
(4.30)

or:
H=NTUxHTU (4.31)
126 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

The number of transfer units, NTU, which is defined by the integral term,
measures the difficulty of separation by establishing a ratio between the
change in composition of the vapor phase and the average value of the trans-
fer potential. NTU is similar to the number of theoretical trays in a tray-type
column.
The height of a transfer unit (HTU) measures the height of packing required
to bring about a change in composition corresponding to one transfer unit.
The height of a transfer unit for packing should be compared with tray effi-
ciency, the two parameters vary inversely.
These concepts are relatively difficult to use in practice, since the height
varies depending on the component under consideration. They are difficult to
use in the transition to the concept of theoretical stages. In fact, when an accu-
rate representation of packing operation is sought, it is preferable to work
directly with transfer coefficients and to use suitable simulation models (see
Section 4.5.2).
Otherwise, to return to the conventional concept of theoretical stage, all
that is needed is to use the idea of theoretical stage (HETP), which is the sim-
ple ratio between the contacting equipment height and the number of theo-
retical trays required to represent the operation of the column. This
parameter, which characterizes the performance of the packing, has a mean-
ing similar to the overall efficiency used for trays. It has no more physical
meaning, since, the HETP is influenced by the height of the bed, for example.
In certain conditions it is possible to relate the height equivalent to a the-
oretical plate to the height of a transfer unit by means of Eq. 4.32:

In A
HETP = HTU -
A-1
(4.32)

Correlations exist to predict the HETP simply, among them let us mention
the Harrison and France correlation (1989). They propose the equation below
for structured packing:

1200
HETP= -+ 4 (4.33)
QP

with:
Q~
the specific area in ft2/ft3
HETP in inches

This correlation is conservative because of the constant term that has been
added.
Table 4.2 gives orders of magnitude for the HETP of packing, with data
extracted from Kister (1992), where the reader can find a body of references
concerning experimental data.
 Chapter 4. IST TILL AT ION. ABSORPTION AND STRIPPING 127

HETP
Packing
I Application Pressure
inches) Source

5/8” Pall rings Random Ethylbenzene/styrene 100 mmHg 8-13 BASF

1” Pall rings Random Isopentaneln-pentane Atmospheric 18 Norton
1” Pall rings Random Ethylbenzene/styrene 100 mmHg 13-19 BASF
1“ Pall rings Random Light gasoline/heavy naphtha 100 mmHg 24-39 Norton
1.5“ Pall rings Random Deethanizer (rectifying) 21 bar 35 Norton
1.5” Pall rings Random Depropanizer (rectifying) 19 bar 38 Norton
1.5” Pall rings Random Depropanizer (stripping) 19 bar 29 Norton
1.5” Pall rings Random Debutanizer (rectifying) 6 bar 29 Norton
1.5” Pall rings Random Debutanizer (stripping) 6 bar 24 Norton
2” Pall rings Random Ethylbenzene/styrene 100 mmHg 22 Glitsch
2” Pall rings Random Deethanizer (stripping) 21 bar 40 Norton
IMTP # 25 Random Isobutane/C, cut 6 bar 16-18 Norton
IMTP # 25 Random Methanol/water Atmospheric 22-25 Norton
IMTP # 50 Random Cyclohexane/n-heptane Atmospheric 27-29 Norton
CMR # 2 Random Et hylbenzene/styrene Atmospheric 18-24 Glitsch
Nutter rings # 2 Random Cyclohexane/n-heptane 1.7 bar 28-30 FRI
1” lntalox saddles Random Light gasoline/heavy naphtha 100 mmHg 28-32 Norton
1.5”Intalox saddles Random Iso-octane/toluene Atmospheric 22-23 Norton
Koch Sulzer CY Structurec Chlorobenzene/ethylbenzene 20-300 mmHg 3.2-4.5 Sulzer
Koch Sulzer BX Structurec DEA/TM 15-18 mmHg 10.4 Koch
Mellapak 250Y Structurec Chlorobenzene/ethylbenzene 76 mmHg 15 Sulzer
Gempack 2A Structurec Ethylbenzene/styrene 119 mmHg 17 Glitsch
Gempack 1.5A Structurec Naphtha/kerosine 1.6 bar 28.8 Glitsch
Gempack 1.5A Structurec Kerosine/gas oil 1.6 bar 25.5 Glitsch
Gempack 1.5A Structurec Gas oil/atmospheric residue 1.6 bar 25.5 Glitsch
Gempack 0.5A Structurec Residue stripping 1.6 bar 53 Glitsch

Table

4.3.3 Comparison between Packing and Trays

Above and beyond problems of efficiency, it is useful to compare the advan-
tages and drawbacks of each of the technologies.

4.3.3.1 Factors in Favor of Packing

Packing has much lower pressure drop than trays and it is therefore extremely
advantageous for vacuum columns. This criterion is important, and another
important quality is its low liquid retention. Low liquid retention is essential
for:
operations when residence time must be minimized (risk of side reac-
tions);
batch distillation when retention affects the quality of separation.
128 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

Packing is also advantageous in the following cases:

small diameter column: installing contacting equipment is much easier:
foaming system: since the contact between the vapor and the liquid is
much less violent, the tendency to foam is minimized with packing.

4.3.3.2 Factors in Favor of Trays

Trays offer advantages in the following cases:
Liquid loaded with solids: the solids accumulate less readily on trays
which are therefore easier to clean.
Complex columns: for any feed injection or product draw off, gathering
and redistribution equipment must be installed. Trays, which solve the
problem naturally, are therefore very advantageous. This is true when
several feed inlet points are planned in order to take changes in feed qual-
ity into account.
Large diameter column: the efficiency of packing depends greatly on the
quality of liquid distribution. It is increasingly difficult to solve this prob-
lem the larger the column diameter. Unless other criteria suggest pack-
ing, then trays are better suited.
Overly high/low liquid flow rate: the flexibility of tray design (optimization
of the number of perforations, use of caps, multiflow technology) makes
them more adaptable than packing when liquid and vapor traffic is highly
unbalanced.
Stop and go operation: when the column is stopped and started fre-
quently, mechanical phenomena are much better attenuated.
Weight:using trays minimizes column weight.
Column inspection: as long as the trays can be equipped with manholes,
column inspection becomes much easier.
Draining and washing: draining and washing operations are easier in a
tray column.

4.4 Distilling Complex Mixtures

Binary mixture distillation is a useful case for introducing the principles of dis-
tillation but is not often found in industrial processes. In most instances, the
products processed contain a large number of components. Though the main
principles discussed in the preceding section remain valid, some adjustments
must still be made.
 Chapter 4. DISTILLATION, AND STRIPPING129
ABSORPTION

4.4.1 Number of Degrees of Freedom

In listing the number of equations and the number of variables describing the
balances and equilibria involved in the column, the number of degrees of free-
dom is seen to be equal to n + 4 + L, where L is the number of streams exiting
the column. For example, in the simplest columns where there is only a distil-
late and a residue, L is equal to 2. The operation of a column is determined
after fixing a number of parameters as mentioned below:
1. The feed flow rate, molar or mass composition and enthalpy set n + 1
parameters. Whether in design or operation, these data are available
and do not represent any real degree of freedom for the column.
2. The column’s pressure. The choice of this operating parameter is essen-
tial during column design. During operation, it may be advantageous to
vary the value although the range is much more limited. We will come
back to this point later on.
3. The number of trays in the rectifying zone and in the stripping zone.
They are fixed for an existing column, but during design the number of
trays and the location of the feed inlet are crucial.
4. L specifications, which can be total flow rates, partial flow rates, recov-
ery ratios or mole or mass fractions. This concept is extremely impor-
tant, it shows that when L operating parameters are chosen, all the other
variables describing the column are fixed. For example for a column with
one distillate, one residue and one sidedraw, the following series of
parameters can be chosen: reflux flow rate, reboiler power and side
stream flow rate.

4.4.2 Distributed and Nonedistributed Substances

Let us consider a mixture of R components with increasing relative volatilities
(defined with respect to the heaviest component) al,a2,a3,...,a,,-l,an.
When this mixture is distilled, the components are separated in different
ways:
1. Part of the light products end up only in the distillate: these are the non-
distributed light substances.
2. Part of the components end up both in the distillate and in the residue:
these are the distributed substances.
3. Some of them are present only in the residue: these are the non-
distributed heavy substances.
In these circumstances, even though the nondistributed substances take
part in phase equilibria occurring on the trays, they do not “do” anything
but transit through the corresponding zone (rectifying or stripping). In con-
trast, distillation behavior is governed by the distributed components. To
characterize it quantitatively two reference substances, or key substances, are
130 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

conventionally mentioned for "binary" distillation, i.e. columns that have only
one distillate and one residue. The light and the heavy key of the separation
are accordingly defined by reference to the relative volatilities. However, a
recovery ratio with a value of 1 must never be employed as it provides no
information on separation.

4.4.3 Minimum Reflux

Assuming the Lewis hypotheses, also used in the McCabe and Thiele method,
and the constance of relative volatilities all along the column, Underwood
(1947) demonstrated that the minimum reflux ratio for a distillation column is
given by the system of R ~ linear
- ~ equations of the following type:

i=nd
aiDi
c
i=l
-
( a i - e j ) = Vmin= (1 + rmin)D
(4.34)

where:
Di is the partial molar flow rates of the different distributed components
in the distillate
D is the total flow rate of distillate
nd is the number of distributed substances
and the 8, are the roots of Eq. 4.35, included among the relative volatilities:

(4.35)

where cp is the feed vapor fraction.

Also assuming initially that all the substances are distributed, the solution
of Eq. 4.35 defines R - 1 roots or pseuderelative volatilities 8, included among
the relative volatilities of the components. By inserting these values in
Eq. 4.34, the linear system of n - 1 equations can be solved if two values are
fixed. It affords access to partial flow rates Di and minimum reflux ratio rmin.
The fixed variables can be two partial flow rates or a partial flow rate and the
total distillate flow rate. Nondistributed substances are detected easily from
the results obtained. When the partial molar flow rate Di in the distillate is
higher than the partial molar flow rate Fiin the feed, Di is taken as equal to 4.
This decision implies that the corresponding substance is not distributed.
After eliminating it from the problem, Eqs. 4.34 and 4.35 are solved for R - 1
components. This procedure is applied symmetrically when the partial flow
rate Di is negative. The component is then considered to be found entirely in
the residue. Methods have been proposed in the literature to identify non-
distributed substances beforehand.

Example of Applying the Underwood Method

Let us consider a ternary mixture whose characteristics are given by the val-
ues below and for which the minimum reflux ratio is sought. A recovery ratio
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 131

of 90% of light product in the distillate (ql) and 90% of heavy product in the
residue (1 - q3) are the target figures.
Total feed flow rate: 100 mol/s; vaporized fraction: 10%.

Relative volatility in
relation to heavy product Mole fraction

Light (1) 1.7 0.2

Middle (2) 1.5 0.5
Heavy (31 1.o 0.3

The two roots of Eq. 4.35, included among the relative volatilities, are
respectively 1.6422 and 1.1215, which give a set of two equations with two
unknowns:
a,%% + %D, + a3Fz3173
= Vmin
(al - el> (a2- el) (a3- el)
a,%rl1 + %D2 + a3Fz3rl3
= Vmin
(a1 -0 (a, - (a3 - 02)
i.e.:
a, x 100 x 0.2 x 0.90 a3x 100 x 0.3 x 0.10
(a1 - 1.6422)
+ (012 -%D,
1.6422)
+

(a3 - 1.6422)
= Vmin

a, x 100 x 0.2 x 0.90 a3x 100 x 0.3 x 0.10

+ (a2-a2D2
+

(al - 1.1215) 1.121 5) (a3- 1.1215) =

The solution of this system yields the two unknowns (the partial molar flow
rate of the intermediate component in the distillate and the minimum flow rate
of overhead vapor), i.e.:
D, = 33.96 mol/s
Vmin = 162.867 mol/s
The minimum reflux ratio is therefore 2.962.

4.4.4 Total Reflux

Let us consider the balance equations by zone established in Section 4.2.3.1.
Total reflux represents operating conditions where the distillate flow rate is
zero or very small in relation t o the internal flow rate in the column. The terms
related to the distillate can therefore be eliminated. The new system is
reduced to the following equations for a tray p in the rectifying zone:
Overall mass balance: Vp+l = Lp
Partial mass balance: yi,p+l=
Equilibrium relations: yi,, =
132 Chapter 4. DISJILLATION, ABSORPTION AND STRIPPING

Assuming a total condenser, the application of these last equations for two
key substances with subscripts i and j gives:
K1.2 . -x. -x.
. XJ.2
Ki,2xi,2 = xi,l = xi,D - 1.1 - J P
K i , 3 X i . 3 = xi,2 q,3xiv3= x j . 2
Ki,4Xi,4 = xi,3 qV4xj,4= xj,3
Ki,N-I xi,N-l = xi,N-2 K / , N - I X ~ , N - I= x i f l - 2
. X.r f l = K i., NX r. , U -- x .1.N-1
K r,N K1.N
. XJ. f l -- K J
. JXJ.. R -- x .J ~ N - ~

By multiplying all these equalities member by member, we obtain:

K1,2 Ki,3 Ki,4 *** Kifl-l Ki,N = xi,D and %,2 q v 3K/,4 qfl-1 K/,N = xj,D
If the relative volatility between substances i and j is constant, these two
equations are reduced to Eq. 4.36:

(4.36)

This is the Fenske equation as it was described for binary mixtures.

Expression N - 1 was obtained by assuming that the condenser, included in the
number of trays, was a total condenser, which implies the absence of vapor-
liquid equilibrium. As a result, according to the way the trays are counted,
depending on the number of trays where liquid/vapor equilibria are located,
the formalism of the Fenske equation can vary. This fact is of little importance
if we remember that the superscript involved in the equation is equal to the
number of trays where occur vapor-liquid equilibria. Establishing this relation
does not presuppose any other assumption than the Constance of relative
volatilities, contrary to the Underwood equation which assumes the validity of
the Lewis hypotheses. This equation defined for the key substances yields the
minimum number of trays to reach the specified objectives. Inversely, when N
is fixed, its application to all the components allows the overall mass balance
and product composition to be identified.
A number of approaches have been proposed to establish the average
value of the relative volatility involved in the Fenske and Underwood equa-
tions. No very general recommendations can be made, however. We will sum-
marize them in Table 4.3.
To take the fluctuations in relative volatility into account more accurately,
Winn (1958) has suggested a modification in the Fenske equation (Eq. 4.37):

(4.37)

pi,. which is a pseudo-relative volatility and Qid are determined by solving

the two Eqs. 4.38 for the top and the bottom of the column:

(4.38)
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 133

,
I Values assessed Value of relative volatility

Top temperature (To,,)

and bottom temperature (Tbotto&
%op, &bottom
%op, &bottom
Bubble point temperature of feed
4.3 %~p,~feed,~bottom

Table
Estimates of relative volatiliw.

4.4.5 Gilliland Relation. Simplified Study

of a Distillation Column
For a specified feed and product quality we know that there is a univocal rela-
tion between the number of trays and the reflux ratio, with these two variables
varying between a minimum value determined by the Underwood or the
Fenske equations and an infinite value. This univocal relation specific to each
problem has been established in the form of a statistical correlation by
Gilliland (1940). It is represented in Figure 4.25. Some authors have proposed
an analytic representation. When r - rmin/r+ 1 is greater than 0.01, Eduljee
(1975) proposes Eq. 4.39, while Molokanov (1971) suggests Eq. 4.40:

(4.39)

r - rmin 1
1 + 54.4- r - rmin --
N - Nmin = 1 -exp r+l r+l

-J2
(4.40)
N+1 r-rmin
11 + 117.2 -
r +1

The simpler but less accurate first equation seems sufficient to the extent
that the Gilliland correlation is an approximate method to get a rough idea of
a distillation problem. A search for great accuracy at this stage would proba-
bly be superfluous.
After determining the minimum number of trays by one of the preceding
equations and the minimum reflux by the Underwood equation, it becomes
very easy to identify the optimum design conditions by utilizing the Gilliland
correlation. As Figure 4.26 shows in qualitative terms, operating costs (curve T)
increase as the reflux ratio increases, while investments (curve IT) start to
decline due to the rapid decrease in the size of the condenser and reboiler,
134 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

between the number of

trays and the reflux
ratio.

0.5 1
r - rmin
r+ 1

before beginning to grow regularly with the size of the column. This variation
shows that there is a solution with a minimum investment, and designing a col-
umn below this point would be senseless. Adopting a depreciation time N
leads to curve I11 which gives the total cost I + NE, where I represents the
investment and E the yearly operating costs. This curve exhibits a minimum
zone which defines the optimum design conditions.
This approach can be replaced by rules of thumb which provide orders of
magnitude for optimum values. For columns working with water or air cooled
condensers, the optimum reflux ranges between 1.2 rminand 1.5 rmin.The num-
ber of optimum theoretical trays is close to 2Nmin.For cryogenic columns, the
optimum reflux ratio depends of course on the temperature. It may reach val-
ues a s low a s 1.05 r,, for a temperature of -100°C.
The feed inlet still remains t o be positioned. As we saw in the section on
binary distillation, there is an optimum inlet point which minimizes distur-
bances t o the system. For binary distillation with feed at the bubble point, the
optimal tray is the one where the liquid phase has the same composition a s
the feed. For multicomponent distillation, this criterion can not usually be met,
so the tray must be found where the composition ratio of light and heavy keys
is identical t o that of the feed. Since this information can only be obtained after
calculating tray by tray, some authors have suggested relations that are inte-
grated in the preliminary sizing tools. The most frequently used is the one
defined by Kirkbride (1944) (Eq. 4.41):
 Chapter 4. DISTILLATION, ABSORPTION
AND STRIPPING 135

-0
ln
a,
.I-
c
0

.
C
B
.4-

8
0

t
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

-
Figure
4.26 Variations in costs versus r - rmin/r+ 1.

(4.41)

In Eq. 4.41, i a n d j represent the respective subscripts of the light and the
heavy keys, R and D stand for the flow rates of the residue and the distillate.
N, and N, are the number of trays in the rectifying and the stripping zones.
We could just as well use the Fenske relation to identify this optimum ratio:

Xi,D
- zi
= a!! - zi = ah's -
- xi, R
(4.42)
Xi,D hl zj Zj ii Xj,R

4.4.6 Advantage of Simplified Methods

All the preceding methods are approximate since they depend on a number of
simplifying hypotheses. This characteristic must not cause the engineer to dis-
regard them. In fact, for a rough analysis of new problems or existing columns,
these easy to implement tools provide a preliminary idea. In design they
rapidly give an idea of the economic stakes and for existing columns, they can
be used to check operation. Even with present-day computerized tools, it is
highly recommended to use them. In this way no time is wasted making prob-
lems converge without having established the orders of magnitude for operat-
ing parameters beforehand.
136 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

4.5 Calculation Methods for Distillation Columns

4.5.1 Static Simulation of Distillation Columns
4.5.1.1 Structural Methods
These methods are mentioned for historical reasons, they were the first
attempts to calculate distillation columns before the advent of the computer-
ized tools we can avail ourselves of now. To reduce the calculations required
for each of the steps to a minimum, the methods used the structural proper-
ties of columns as we saw during the discussion of the McCabe and Thiele
method.

a. The Thiele and Geddes Method (1933)

Let us consider the partial mass balance for the component i in the vicinity of
a rectifying zone including the first N trays:

vN+IY;,N+I= Dxi,D + LNxi,N

VN+IK;,N+IX;,N+I = Dx;,D+ LNX;,N

Let AN be the absorption factor of tray N defined by the ratio LN/K;nVN.The

preceding equation is reorganized to lead to a recurrence relation between
tray N and tray N + 1.
LN+IX;,N+I
= [ LN+I ][ LNX;,N 1]
+

DXi,D Ki,N+I vN+l Dxi,D

(4.43)

Since the condenser is assumed to be total, the term [ L , X ~ , ~ / D X is~ , ~ ]

reduced to r (r is the reflux ratio). Application of the recurrence relation for the
whole rectifying zone leads to relation 4.44:

The same type of approach is possible for the stripping zone. It leads to
definition of a stripping factor Si,Nr= VL,KjJ,/LL,.
This recurrence relation is used as a basic tool in the Thiele and Ceddes
method and has also been utilized by Edmister (1957) for a simplified calcula-
tion model for multicomponent columns. If we assume that the absorption and
stripping factors are constant throughout the corresponding zone, Eq. 4.44 in
fact becomes a simple algebraic equation which is written:

(4.45)
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 137

Based on the same construction, the recurrent equation is developed for

the stripping zone and leads to Eq. 4.46:

(4.46)

+ 1)
Si = VSi,2J(Si,N, + 0.25 - 0.5 (4.47)
Once the parameters hi and Si have been calculated with the laws recom-
mended by Edmister (Eq. 4.47), it becomes very easy to solve the 2n equations
to identify the composition of the distillate and of the residue when the num-
ber of trays is known in each of the zones.
Thiele and Geddes do not use the approximation suggested by Edmister.
Once the hypotheses have been chosen as to the temperature profile, the liq-
uid and vapor traffic and the product composition (which fixes the absorption
and stripping factors), tray composition is estimated beginning at the top and
at the bottom by means of recurrence relations. This composition profile then
serves to estimate the bubble point temperature of the liquid and the compo-
sition of the vapor phase. Then the enthalpy balances yield the new
liquid/vapor profile. All these values, which are useful in updating the absorp-
tion and stripping factors, are reutilized during the following iteration. The
assumptions about product composition are modified according to mass bal-
ance differences on the feed tray.
This iterative procedure is characteristic of all calculation methods. A vec-
tor of variables is chosen as decision-making variable, and this allows all the
dependent values to be calculated. It is then estimated iteratively until it sat-
isfies a convergence criterion.

b. The Lewis and Matheson Method (1932)

The Lewis and Matheson method is very close to the preceding one in spirit.
Instead of using absorption factors, it utilizes partial and total mass balances in the
stripping zone along with the enthalpy balance. Accordingly, calculation of the
series of trays 1t o p - 1 determines the characteristics of the incoming vapor (flow
rate Vpand composition Y ; , ~ Tray
) . p is calculated then by a flash calculation to esti-
mate the equilibrium temperature and the composition of the liquid. The series of
partial mass balances and the enthalpy balance f i i the n + 1 unknowns (Lp, Vt+l,
Y ~ , ~ . Calculation
+~). of each of the zones is ended when the ratio of the mole fraction
of the key substances becomes greater than that of the feed (Fig. 4.27).
In the most general case, the compositions determined from the stripping
zone are different from those identified from the rectifying zone. The assump-
tions are revised using the formula:

where p and p' represent the last trays in rectifying and stripping.
138 Chapter 4. DISTILLATION, ABSORPTION
AND STRIPPING

light key
heavy key

-
Figure
4.27 Determining the number of trays in the stripping and rectifying zones (Lewis
and Matheson method).

This approach is strictly suited to simple distillation columns but can read-
ily be coded if the user has the laws required to calculate the liquid/vapor
equilibria and the enthalpies.
It should also be noted in Figure 4.27 that separation work decreases as the
distance from the top increases for example. This is true for normal mixtures
exhibiting a regular equilibrium two phase region and shows the parallel
between the equilibrium curve and the operating curve.

4.5.1.2 Numerical Methods

Contrary to structural methods, which allow distillation columns to be calcu-
lated by moving tray by tray from the top and from the bottom of the column,
global numerical methods attempt to solve all the equations descriptive of
tray operation simultaneously. They can deal with any type of structure. As
shown in Figure 4.28, feed inlet and liquid or vapor sidedraws can be consid-
ered, as well as a heat transfer if needed. The advantage is that they allow sim-
ulation of interconnected columns such as those currently used in distillation
of crude oil.
The basic equations are of course the material and enthalpy balances
along with the tray operation equations (theoretical or based on the Murphree
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 139

Lp- 1 VP

Numerical method for

calculating distillation
Rvp
I
columns.
R = retention (mol).
S = sidedraw flow rate
(mol/s).

equation). It should be noted here that the use of the last criterion is specific
to distillation columns with trays.
The various balances available for tray p are described by the equations
below:
Total mass balance:
Fp + L p - l + Vp+l = (V, + s:) + (Lp + sb)

Partial mass balance:

FpZi + L,-lXi,pl + Vp+lYi,p+l = (V,, + s : ) Y ; . p + ( L p + s ; ) x i p

Enthalpy balance:
FphF,p+L,-lhkl+ V , , + ~ h V , ~ = ( V , , + S ~ ) : ) h ~++S(3Lh, b
with: h,,,, = f(T'@, PFp, 2 ) h:= g(T,,, P p , 7) hb = g'(T,,, P,,, X)
_ _ _
where x,y , t represent composition vectors.
Tray balance chosen between:
- equilibrium relation, for theoretical trays:
v,,,pp9
Y;,,,= K;,,, Xi@, %m i ,

- tray relation, for trays characterized by a Murphree efficiency:

Yip = Ei,MV,pKip(Tp* 'p* x i p , ' 9 7)xi,p + (l -Ei,MVp)Yi,p+l

One of the relations must be chosen according to the chosen tray model.
140 Chapter 4. DlSTILLATiON, ABSORPTION AND STRlPPlNG

Constraints:
. .
c
. .

=1 and Y;,=
~ 1
i= 1 i= 1

This set of equations represents 2n + 3 equations, since the constraint

equations are redundant with the partial mass balances and the total mass bal-
ance. If we consider the pressure of the tray, the liquid and vapor sidedraw
flow rates, the feed flow rate and the thermal power transferred on the tray as
data, they correspond to 2n + 3 unknowns 0.e. the composition vectors 7 and
X,the operating temperature of the tray and the liquid and vapor flow rates).
The calculation methods for distillation columns differ depending on the
way the descriptive equations of the column are handled. As a reminder, for
distillation producing only one distillate and one residue, the number of
degrees of freedom of the whole system is only 2, once the pressure, the feed
characteristics, the number of trays, and the feed tray have been fixed.

a. The Tridiagonal Matrix

If we assume theoretical trays, the mass balance is rewritten substituting the
value Ki,p for the giving the following equation:

Lp-1 x;,p-l- vp + s:) Ki,p + (Lp + S , L ) I x i , p + vp+lKi,p+l x i , p + l = - F p z i

i.e.
aip1xip-1- b i p x i p + Ci,p+lxip+l= -Fpzi

As a result, for each component i, the balance for all the trays is summed
up by the matrix equation:
bi.1 - c ~ , I ... 0 0 0 0 0
+i,2 bi,2 0 ...
-c,,a 0 0 0 0
0 -ai,3 bi,3 -q3 0 0 ... 0 0

. . . .
. . . . 4ifl-2 bi~v-2 - ~ i , t - 2 0
. . . . 0 diJ-1 biJ-1 -Ci,N-l
0 0 ... . 0 0 -ai, bi,

If we assume as known the parameters b L j ,q i , which means stating

hypotheses on the liquid and vapor flow rates as well as on the K values, then
the mole fractions of component i can be calculated by solving the linear sys-
-7 6
tem A.?; = wherexi and stand for the composition vectors of component i
on the trays and the partial inlet molar flow rate vector respectively. There are
of course as many systems as components. The structure of matrix A is pecu-
liar since only the principal diagonal and the parallels contain non-zero values.
It can be used for an extremely easy numerical solution by elimination meth-
ods, such as the one proposed by Thomas, but modified to avoid accumulat-
ing rounding errors.
 Chapter 4. DISTILLATION,
ABSORPTION
AND STRIPPING 141

Most of the algorithms that will be presented use it during the calculation
procedure either to compute partial mass balances or to initialize column cal-
culation.

b. The Bubble Point Method

Based on an estimate of liquid and vapor traffic and tray temperatures, com-
positions are calculated with the tridiagonal matrix, thereby allowing the bub-
ble point temperature of the liquid phase to be estimated and the liquid and
vapor traffic to be revised based on enthalpy balances. In order to do this, the
balances are worked out starting at the top and the bottom of the column
using an overall balance of the zone including the tray under study and the
corresponding extremity.
This method is well suited to mixtures that deviate little from ideality and
exhibit a narrow boiling temperature range.

c. The Holland 0 Method (1981)

The Holland method can only be used to calculate columns specified by the
flow rates of distillate and reflux. It is based on a convergence promotor, the 8
parameter, and a method of estimating tray operating temperatures, the K,
method, to make calculations faster and more reliable.

Correcting Partial Flow Rates in the Column

Following several calculations by the tridiagonal method, nonconvergence
makes itself felt particularly by errors in partial mass balances for the column.
This fact can be utilized to define the convergence parameter 8, which is used
to revise partial flow rate assumptions so as to approach a solution where it is
equal to 1.
By definition, we have:

( ?)mod if i d = ( ?)calculated

If F;.represents the sum of partial molar flow rates feeding component i into
the column, it becomes easy to write an equation from which the value of 8 can
be extracted, since the flow rate D is specified:
i=n
5 and D- c F, =O
Di,rnodified =

If we assume that the partial flow rate Di must be modified by weighting it

by a factor p i , the flow rate Ri is weighted by the factor 8 p i , thus giving the
expression to estimate p i :
Fi
Pi =
(Di)calculated + (Ri)calculated
142 Chapter 4. DISTILLATION, ABSORPTION
AND STRIPPING

This last parameter is then used to revise the internal flow rates and the
compositions:
(Li,p)rnodified = Pi (Li,p)calculated (I/r,p)modified = Pi(vi,p)calculated

('%,p)rnodified
(Xi,p)rnodified = i= ,
(Li.p)modified
i=1

Correcting Temperatures
Tk+l for the iteration k + 1 is estimated with a simplification of the equilibrium
model, based on a law of the ClausiusClapeyron type:

Kp,k = Ap,k exp [ Tp,k ]

-Bp,k
Kp,k+l = I= 1 (Xi,p)modified (%,p)k

The coefficients Ap,k and Bp,kidentified in this way are used to reestimate
the new temperature.

Correcting Total Flow Rates

Liquid and vapor flow is revised by utilizing the total mass balances and
enthalpy balances for each tray progressing from the top of the column.

d. Methods Using the Newton-Raphson Technique

Including liquid and vapor fugacities and tray pressure drops (assumed to be
known), the total number of variables to be identified is 6nN + 3n + 8N + 9 + L,
where L is the number of streams exiting the column. Only 6nN + 2n + 3N + 3
equations are available. To solve the problem, the partial feed flow rates
( n values), the feed enthalpy contents, the column operating pressure and L
specifications must be fixed.
The most frequently used method of solving this type of system is the
Newton-Raphson method which consists in approximating the solution by
extrapolating the k iteration point by a tangent plane. As a result, the new
point is defined by the equation:
-
Jk(xk+l - xk) =- k k):(
where Jk(called Jacobian) is the matrix of partial derivatives of all the func-
tions with respect to all the variables. The Jacobian is a matrix of X dimension,
where X is the number of independent variables. In this matrix, many terms
are equal to zero. This hollow matrix structure is judiciously used to reduce
the size of tables.
In most cases, the analytic derivatives of the equations are not available
and the values are approximated numerically by variational calculation.
 Chapter 4 DISTILLATION, ABSORPTION AND STRIPPING 143

Estimating a Jacobian and its inverse becomes extremely long. In order to cut
down on the time, most methods:
do not calculate J, at each iteration, but use an updating procedure pro-
posed initially by Broyden (1965);
use a factoring method to solve the linear system.
For some software, the analytic derivatives of thermodynamic laws have
been coded, and this enhances calculation speed and accuracy.
Because of its extrapolation process, the Newton method is not very robust
when systems are highly non-linear or when initial values are too different
from the solution. The quality of initialization and the weighting procedures
for the acceleration factor condition software performance. Among the more
recent approaches, “homotopy” and “inside-out” algorithms provide different
ways of solving this problem. They are both satisfactory. Thermodynamic
equations, a source of non-linearity, are approximated by a simplified model
which is used to ensure the convergence of the Newton method. The model is
then revised, either by an external loop (inside-out) or by a transition proce-
dure toward the true model (homotopy).

N Dimension Methods
In order to minimize the size of the problem to be solved, some methods uti-
lize the interactions between the different trays in a column. The methods are
based on choosing an independent variable and an objective function per tray.
The other variables are revised with the other equations.
One of the methods is called “Sum of rates” and uses tray temperatures as
independent variables and the enthalpy balance as an independent equation.
It is ideally suited to modeling absorbers and strippers and mixtures with a
wide boiling temperature range. It can not be applied as effectively to distilla-
tion columns whose condenser and reboiler operation has to be estimated
separately. The algorithm is as follows:
1. Give an estimate of the concentration profile and vapor traffic.
2. Calculate the liquid flow rates with total mass balances.
3. Apply the tridiagonal matrix to revise the partial molar flow rates in the
vapor phase and the absorption factor for partial molar flow rates in the
liquid phase.
4. Use the Newton-Raphson technique to recalculate the temperature with
functions 4 ( T I ,T2...,TN).
5. Return to step 1 until the functions 4 ( T I , T2,...TN)have reached a mini-
mum value defined as the convergence threshold.

2N Dimension Methods
Two variables and two equations are used for each tray. Tomich (1970) pro-
posed using the enthalpy balance and the composition normation equations in
the following form: i=n

’
i=1
i=n
xi,p- C
i = l Yi,p =0
144 Chapter 4. DISTILLATION,
ABSORPTION
AND STRIPPING

The independent variables are total vapor flow rate and temperature for
each tray. Holland amended this initial proposal by replacing the normation
equation by a bubble point equation and the vapor flow rates by convergence
parameters
The methods are well suited to distillation involving products with an aver-
age or wide boiling temperature range.

Global Methods
Global methods solve partial mass balance, enthalpy balance and vapor-liquid
equilibrium equations simultaneously by putting them all in a normalized form
as suggested by Naphtali and Sandholm (1971):

i= 1 i= 1

---
KipLi,p
i=n
Vi,p
i=n
=o
C
i= 1
Li,p C Vi,p
i= 1

The independent variables are of course the temperatures and partial flow
rates in the liquid and vapor phases. Solution becomes direct and does not
need to pass through other loops the way the preceding methods do. The nat-
ural specifications are the condenser and reboiler thermal loads, which fix the
mass balance and the reflux ratio. When other specifications are chosen, the
enthalpy balance of the corresponding stage is replaced by the specification
equation and thermal loads calculated in an external loop.
The Goldstein and Stanfield method (1970) is an alternative solution which
uses the temperature, total liquid and vapor flow rates and the composition of
the liquid phase as decision-making variables. The matrix handling procedure
allows temperatures and vapor flow rates to be modified before changing the
liquid flow rate and the composition.
Though very similar, the methods have slightly different areas of applica-
tion because of the specific procedures involved. The first is well suited to
columns processing few components, even if the mixtures are highly non-ideal,
but with a lot of trays. The second is better suited to columns with few trays
but processing mixtures with a lot of components.
Gallun and Holland (1976) have improved the NaphtaliSandholm method
by developing techniques for preparing and solving the Jacobian, which is
a sparse matrix, by integrating the activity coefficients in the independent
 Chapter 4. DISTllLATION, ABSORPTION AND STRIPPING 145

functions ("Almost band method). They have provided the needed flexibility
to add or remove variables or to regroup them in order to adjust problem
preparation as well as possible to the application. The method's area of appli-
cation is very wide and it is robust enough for non-ideal distillation.
The analysis of distillation equations shows that only phase equilibrium
equations have this characteristic. This is why methods have recently been
proposed to minimize the impact of phase equilibrium non-ideality.
Accordingly, "insidequt" methods determine simplified models for the Kij and
the enthalpies. These laws specific to each tray are revised in an external loop
which utilizes composition data coming from the internal loop.
The methods perform well because this system allows easy convergence of
the internal loop and great insensitivity to non-ideality.
To define simplified thermodynamic models, Boston et al. (1972, 1974,
1978) use as a basis the K, technique described in the Holland method. The
phase equilibrium model is represented by a series of equations where the mi

I
stand for weighting factors specific to the Boston method:

Ki.p = Kb,p %,pY?p

rtp = a i , p + bi,pXi,p
Ki,p,C"me"t

= Kb,p,ref

where F is a reference temperature.

The enthalpy laws are given by the following equations, where the coeffi-
cients are determined from tray properties and the h" terms, the ideal gas
ent halpy:
h: = hi" + Cp- Dp(Tp- T * )
hb = h i L+ Ep- Fp(Tp- T*)

All of these parameters are therefore calculated from column characteris-

tics at a given iteration then revised. To solve the internal loop Boston has pro-
posed a variation on the "Sum of rates" method by means of a change in
variables:

Kb.p ' P
c Fpqp
p= 1
Rp = Kb,p ' P + L~ = 1 + Rp + R P q p
146 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

All the physical variables describing the tray are deduced from it:
Pi,P a i , pPi.p
x. =
1.P I
-
=n Yip = j z n

i =1
c Pi# c
i=1
ai,ppi,p

Kbp= ;=n P i p

c ai,pPi,p
Tp = [In (K;; -4 + 4 -1

i=1
i=n
Lp = (1 - R c
1i = l P i , p
Since the calculation method is not suited for handling other specifications
than the mass balance and the reflux ratio internally, Boston added an inter-
mediate loop to adjust these parameters and attain more elaborate specifica-
tions.
The Boston method is suited to any type of mixture (wide or narrow boil-
ing temperature range) and performs extremely well for sharp separation.
Russell (1983) has modified this approach by choosing the relations below
for the weights 0;
ti d [In ( ~ i , p , c " r , n 3 1
a.=- with ti
I ;=n
d tl/T,l
c
i= 1
ti

He then applies a Newton-Raphson N method, calculating the partial molar

flow rates in the liquid phase with the tridiagonal matrix, and the total flow
rates by summing the preceding values. The independent equations are the
enthalpy balances for each of the trays and the independent variables are the
logarithm of the stripping factors or of the sidedraw rate when necessary. The
great advantage of this solution is that it makes convergence relatively insen-
sitive to initialization of variables.

Other Methods
A whole category of methods use dynamic simulation to more easily get oper-
ating conditions more severe than initial conditions. They do not currently
seem to provide much improvement since they compound static simulation
difficulties and dynamic ones.
Another approach, which consists in using the "homotopy" principle, looks
extremely interesting. The principle's theoretical approach has been
described by Seader (1985) and Wayburn and Seader (1987) among others.
The basic idea is to use a simplified F(x) model to initialize the problem and to
approach the rigorous model, which is difficult to solve, by means of a transi-
tion parameter t. The solved function is then defined by one of the following
equations. The parameter t is equal to 0 at the beginning of the simulation and
to 1 at the end.
 Chapter 4. DISTILLATION,
ABSORPTION
AND STRIPPING 147

H(F, t) = (1 - t) F(k) + tG(F) =0

H(F, t) = F(5)l-I G (X)‘ = 0
Differentiation of the first in relation to t gives:

):(
H? - + H t = O

After solving the system with the preceding methods for the simplified
model (t = 0) and choosing a At step for the parameter t, equation H(F, t) = 0 is
solved at each iteration by taking the value identified by the equation
-
x , , ~ = 5,+ At (dF/dt), as the initial value of 2.
The advantage of such a procedure is that it provides a gentle transition
between simplified thermodynamics of the ideal type and rigorous thermody-
namics. Along the same lines, some authors suggest using tray efficiency as a
transition parameter. In the beginning the Murphree efficiency is taken equal
to zero, which means that the composition of the liquid on all the trays is the
same as that of the feed. As efficiency is increased, separation gets “harder”
and approaches the solution of the problem.

4.5.1.3 Methods of Distillation Column Calculation

by Transfer Model
All of the preceding models used the concept of separation stage, theoretical
stage or real stage when the Murphree efficiencywas included. We mentioned
that the last approach was inaccurate since it was necessary to provide a
value representing an average of the transfer phenomena involved on the tray
under consideration for all the components. This kind of modeling is even less
well suited for packing-type columns, because it means that the continuous
phenomenon involved in the contacting equipment is discretized by means of
the HETP value.
A new method is being developed to help overcome these difficulties,
which consists of coming directly back to the description of transfer phenom-
ena with the double film theory (see Section 4.3 on contacting equipment). To
avoid integration in relation to the state variable, contacting equipment is rep-
resented by a series of finite volumes (cells) such as the one described by
Figure 4.29. The equations characterizing the operation of a cell where there
are no chemical reactions are established below (the liquid and vapor phases
are assumed to be homogeneous)? This assumption about hydrodynamics
causes no problems in theory since a large number of them are arranged in a
cascade, thereby leading to simulation of plug flow. By adjusting the number
of cells any level of dispersion can be simulated. However, model construction
becomes complex when all types of hydrodynamics are represented in detail.

3. A chemical reaction is dealt with simply by adding into the equations the disappear-
ance/production streams of the relevant chemical species associated with the corre-
sponding enthalpy variations (see Section 4.10).
148 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

/Interface
J
I

Vapor
phase
I
I ILj- I

Liquid
phase
 Chapter 4. IS TILL AT ION, ABSORPTION AND STRIPPING 149

Mass Balances
Vapor phase Vi,j+l - Vj,,+ FY, - SYj + NYj = 0
Liquid phase Li,j-l - L , , + F k j - S k j + N k. J. -- O
Interface N \1.J! . = N 1.1
C . = N L. J.
In these expressions, Lij-l and Lip Vij and Vij+l stand for the partial flow
rates of component i in the liquid and in the vapor phase at the inlet and out-
let of cellj (mol/s). Fij and Sjjare the partial inlet and offtake flow rates of com-
ponent i (mol/s). The terms NY, and Nk, are the mass transfer streams in the
liquid and in the gaseous phase (mol/s). They are of course the same as there
is no accumulation of matter at the interface.

Enthalpy Balances
Vapor phase:

Liquid phase:

where:
h)' is the enthalpy of a mole of vapor phase (J/mol)
h ) is the enthalpy of a mole of liquid phase (J/mol)
The terms Q, and I$ ,
represent the heat flow supplied from outside and the
heat flow due to the transfer respectively. They are of course the same in the
liquid and in the vapor phase.

Transfer Equations
N . . = k v. .(I. (u. . - y interface) = k L . v.a.(x interface - xi,j )
kJ 1,) J J L J J J 4J

Q I.J. = %\!
J v.Q.( J - Tinterface)
J J T\! J = %k J vJ. aJ . ( T y f a c e - T))
where:
Xy is the heat transfer film coefficient, vapor phase side (W/(m2.K)
X) is the heat transfer film coefficient, liquid phase side (W/m2.K)
k yj is the mass transfer film coefficient, vapor phase side (mol/(m2.s))
kt, is the mass transfer film coefficient, liquid phase side (mol/(m2.s))
i' is the volume of the cell j (m3>
is the volume interfacial area (m2/m3>
Ti" is the temperature of the vapor phase (K)
Ti" is the temperature of the liquid phase (K)
Tinterface
I is the temperature of the interface (K)
150 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

Equilibrium Relations at the Interface

interface = Kinterface Xj3terface at the temperature Tjnterface
Y iv; 1.1

Constraint Relations

This model is solved by a global Newton method, which sometimes poses

convergence problems when initialization is insufficient or when thermody-
namic models are highly non-linear. Above and beyond these mathematical
problems, the use of transfer models is advantageous only if there are accurate
correlations to estimate the interfacial area and the mass and heat transfer
coefficients. Krishnamurthy and Taylor (1985) have established recommenda-
tions for these estimates.
These calculation methods are starting to be available in commercial soft-
ware. Their complexity means that use is restricted to problems where trans-
fer phenomena are limiting. This mainly involves packing-type distillation
columns, absorbers (with trays, but also often with packing), and thermal
exchange columns.

4.5.1.4 Simulation Software

The commercial supply of simulation software is plentiful and diversified.
Specific software can be found to calculate a unit operation, size equipment
and simulate processes. These products, whose primary purpose is to estab-
lish mass and enthalpy balances of all the nodes of a system, provide repre-
sentations of unit operations with a greater or lesser degree of accuracy.
Distillation is one of the operations with the most complete calculation tools.
Most of them offer the possibility to access:
Simplified methods, which can be used in a rapid rough approach to a
distillation problem.
Tray-by-tray calculation methods using the concept of theoretical stage
or real tray defined with the Murphree efficiency. This latter option is sat-
isfactory since it allows more accurate models of distillation columns to
be worked out. However, the fact of assigning one single efficiency per
tray remains restrictive.
Some software offers the possibility of simulating columns by transfer mod-
els. Here again, the model’s possibilities remain limited with respect to the
hydrodynamic phenomena involved in each of the phases (no description of
liquid and vapor distributors). At this chapter’s publication date, the three
best-selling software packages4 are:
PRO11 by Simulation Sciences (USA)

4. See “References”section.
 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING 151

ASPEN Plus by Aspen Tech (USA)

HYSIS by Hyprotech (Canada)
Constant evolution of products and competition means that other software
must not be disregarded. It is more difficult to attempt to indicate their advan-
tages and drawbacks as definitely as their sales. We would just advise poten-
tial software buyers to check how well the product fits their technical needs
and verify the quality of the thermodynamic data bank in the areas they are
involved in. Buyers might even want to proceed to calculation tests to validate
the system.

4.5.1.5 Comments on the Use of Simulation Software

The use of simulators offers numerous possibilities of defining the specifica-
tions that determine distillation column operation. We know that it is neces-
sary to fix the same number of values or specifications as there are products
in order to determine the system. The choice must be thought through in such
a way that the problem can be solved or that convergences are facilitated. First
and foremost, it is useful to define a general approach. When the user does not
have orders of magnitude for a new problem, he is advised to carry out an ini-
tial calculation, defining specifications that ensure a solution, such as the
reflux ratio and the mass balance. Analysis of the first solution orients any
changes to be made.

Feed Position
Improper feed location very often causes a composition pinch. In concrete
terms, this means flat temperature and composition profiles. The feed inlet
point must be moved toward the composition pinch tray. When this tray is
located at the feed inlet, the column is near the minimum reflux for the speci-
fied problem. If the product quality is satisfactory, this offers the possibility of
carrying out the same separation with fewer trays at the expense of a very
small increase in operating costs.
When the feed inlet zone is defined, the optimum tray can easily be identi-
fied by defining several feed inlet trays with a stream distributor, or splitter,
upstream from the column. By optimizing the operation of the distributor, the
best solution is found. Some software packages can find the optimum tray
themselves.

Feed Temperature
When the feed temperature is too high or too low, the temperature and com-
position profiles in the column are very often perturbed. A zone can be found
where the temperature increases toward the top of the column. Except for the
specific case where a stripper is simulated, this phenomenon shows that the
feed is overly subcooled. This is why it is preferable to place the feed in bub-
ble point conditions during a design study in order to avoid any superposition
with the problems of defining the feed inlet point. The temperature should be
taken as an optimization variable during the solution’s refining phase.
 Next Page

152 Chapter 4. DISTILLATION, ABSORPTION AND STRIPPING

Specification on the Mass Balance

The fact of specifying the flow rates of all the streams exiting the column
causes over-specification due to the existence of the overall balance equation.
This type of configuration should be rejected, one of the exiting flow rates
must be left free and another specification must be laid down.
Purity or Recovery Ratio Specifications
This type of specifications is more difficult to handle, since one of the greatest
risks is taking incompatible performance criteria:
Several specifications on the same stream: if we take one of the quality
criteria as a reference, the partial flow rates of the other products are
dependent through the descriptive equations of the column. Using such
an approach is extremely risky and can only succeed in certain excep
tional cases. However, it is frequent and risk-free to place a definite value
on the sum of the partial flow rates in an exiting stream for a subset of
components.
Several specifications on the same compound: even though it is theoret-
ically possible, such a choice often poses numerical problems when the
partial mass balance is solved and leads to difficulties in the solution.
Extreme purity or recovery ratio: direct definition of specifications can
pose problems because of the proximity of mathematical or physical con-
straints. It is preferable to choose specifications on other products in
order to reach the target result. This can be done by fixing the impurity
rate or the recovery ratio of an adjacent component.
Purity specifications on sidestreams: sidestream composition depends
mainly on the liquid and vapor traffic that has been set and less sensi-
tively on the offtake flow rate. It is extremely dangerous to use a purity
specification on a sidestream. It should be contemplated only to refine a
solution determined by fixing the offtake flow rate.
Other Specifications
Tray temperature: this means that choice of temperature must be perti-
nent and the tray must be sensitive enough to orient separation.
Condenser and reboiler thermal duty: these specification types do not
cause any problems provided that they are compatible.
Partial condenser: for a given vapor quality, the only degree of freedom
available is the equilibrium temperature in the condenser. The composi-
tion of the two phases is then fixed. This very strong relationship leads
to giving pride of place either to the liquid or to the vapor distillate and
making the remaining one dependent.

4.5.1.6 Criteria for Choosing among

the Different Simulation Methods
The principal characteristics of the different methods described in Sections
4.5.1.1 and 4.5.1.2 have been discussed in the text. The objective of Table 4.4
is to present a synthesis.