Unit Operations of Chemical Engineering: January 2018

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/322364940
Unit Operations of Chemical Engineering
Book · January 2018
CITATIONS READS
10 55,600
1 author:
Dennis Prieve
Carnegie Mellon University
129 PUBLICATIONS 5,602 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Total Internal Reflection Microscopy View project
Teaching View project
All content following this page was uploaded by Dennis Prieve on 10 January 2018.
The user has requested enhancement of the downloaded file.

Class Notes
Unit Operations
of Chemical Engineering
by
Dennis C. Prieve
Department of Chemical Engineering
Carnegie Mellon University
Pittsburgh, PA 15213
An electronic version of this book in Adobe PDF® format was made available
to students of 06-361, Department of Chemical Engineering,
Carnegie Mellon University, Spring Semester, 2017.
Copyright © 2017 by Dennis C. Prieve

06-361 page ii Spring, 2008
Table of Contents
CHAPTER 1. COURSE INTRODUCTION .................................................................. 1
Syllabus Summary .........................................................................................................................1
Instructor ................................................................................................................................................ 1
Textbook.................................................................................................................................................. 1
Web Site .................................................................................................................................................. 1
Grading ................................................................................................................................................... 1
What Are “Unit Operations”? ......................................................................................................2
Objective of Course ................................................................................................................................ 2
Overview of ChE Curriculum ................................................................................................................. 3
CHAPTER 2. HEAT EXCHANGE EQUIPMENT ........................................................ 3
double-Pipe Hxer ...........................................................................................................................3
Shell-and-Tube Hxer .....................................................................................................................4
CHAPTER 3. FUNDAMENTALS OF HEAT TRANSFER: REVIEW .......................... 5
Log-Mean Driving Force ...............................................................................................................5
Co-Current versus Counter-Current Flow ........................................................................................... 12
Non-Linear Relation Between q(x) and ∆T(x)....................................................................................... 12
Resistances in Series ....................................................................................................................16
Overall Coefficients (U) vs. Film Coefficients (h) ................................................................................ 16
Fouling Factors .................................................................................................................................... 20
Correlations to Estimate h’s .......................................................................................................21
hi for Turbulent Flow In Long Pipes .................................................................................................... 22
Effect of Duct Shape on hi .................................................................................................................... 23
CHAPTER 4. SHELL-AND-TUBE HEAT EXCHANGERS ....................................... 24

Tube Pitch .....................................................................................................................................24
Baffles ............................................................................................................................................25
ho for Shell-and-Tube Exchanger ..............................................................................................26
Multipass Construction ...............................................................................................................27
"True" Mean Temperature Difference .................................................................................................. 28
An Example:*........................................................................................................................................ 30
Appendix: Derivation for Sb and Sc ..................................................................................................... 37
CHAPTER 5. EVAPORATION ................................................................................. 39

Equipment for Evaporation ........................................................................................................40
Long-Tube Vertical Evaporator............................................................................................................ 40
Forced-Circulation Evaporator ............................................................................................................ 41
Evaporator Performance.............................................................................................................41
Example: Simple Evaporator Problem ......................................................................................42
Steam Economy..................................................................................................................................... 46
Overview of Evaporator Design .................................................................................................48
Complications ...............................................................................................................................49
Boiling-Point Elevation ........................................................................................................................ 49
Heat of Dilution .................................................................................................................................... 52
Lever-Arm Rule for Heat of Dilution.........................................................................................53
Multiple Effects ............................................................................................................................54
Approximation Solution ........................................................................................................................ 56
Performance ......................................................................................................................................... 58
Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

06-361 page iii Spring, 2008
Example: Simple Multiple-Effect Evaporator Problem...........................................................59

Effect of Boiling Point Elevation .......................................................................................................... 61
Rigorous Solution (incl. BPE and sensible-heat effects) ...................................................................... 62
Vapor Recompression ..................................................................................................................72
CHAPTER 6. VAPOR-LIQUID EQUILIBRIUM ......................................................... 78
No. of Phases Present ...................................................................................................................78
One Component .................................................................................................................................... 78
Two Components .................................................................................................................................. 79
xy and Txy Diagrams ...................................................................................................................82
Ideal Solutions and Constant Relative Volatility ......................................................................83
Constant Relative Volatility Approx. .........................................................................................84
Ideal Binary Mixtures..................................................................................................................85
Nonideal Behavior........................................................................................................................88
Batch (Differential) Distillation ..................................................................................................89
CHAPTER 7. FLASH DISTILLATION ...................................................................... 93
Approximate Solution..................................................................................................................94
Lever-Arm Rule on Txy Diagram.......................................................................................................... 96
Rigorous Solution .........................................................................................................................98
Lever-Arm Rule on Hxy Diagram ....................................................................................................... 101
CHAPTER 8. MULTISTAGE OPERATIONS ......................................................... 101
Counter-Current Cascade .........................................................................................................103
CHAPTER 9. EQUIPMENT FOR TRAY TOWERS ................................................ 104
CHAPTER 10. PERFORMANCE OF A C-C CASCADE .......................................... 109
Analysis of a C-C Cascade ........................................................................................................109
Constant Molal Overflow ..........................................................................................................110
A Rectifying (Enriching) Cascade ............................................................................................112
A Stripping Cascade ..................................................................................................................115
Steam Distillation .......................................................................................................................119
CHAPTER 11. MCCABE-THIELE METHOD ........................................................... 120
Location of Feed Plate ...............................................................................................................123
Convergence of Operating Line and Equilibrium Curve ......................................................124
Choosing the Reflux Ratio.........................................................................................................131
1) Existence of Rmin ........................................................................................................................... 131
2) Total Reflux: R→∞ gives Minimum Stages .................................................................................... 133
Calculating Flowrates and Condenser/Reboiler Heat Duties ................................................134
Role of q-Line .............................................................................................................................137
Plate Efficiency ...........................................................................................................................138
Determining Number of Real Stages ........................................................................................140
CHAPTER 12. MULTI-COMPONENT DISTILLATION ............................................ 142
Sharp Split ..................................................................................................................................142
Product Distribution for a Sharp Split ....................................................................................143
Minimum Number of Trays ......................................................................................................148
Review of Relevant Thermo ................................................................................................................ 151
Dew Point Evaluation ......................................................................................................................... 153
Bubble Point Evaluation ..................................................................................................................... 153

06-361 page iv Spring, 2008
Minimum Reflux Ratio ..............................................................................................................155

Method 1) Pseudo-Binary Method ...................................................................................................... 155
Method 2) Underwood's Method ........................................................................................................ 156
Number of Ideal Plates at Operating Reflux ...........................................................................161
Method 1) Underwood's method ......................................................................................................... 161
Method 2) Erbar & Maddox correlation ............................................................................................ 161
Method 3) Gilliland correlation ......................................................................................................... 162
Feed-Plate Location ...................................................................................................................162
Rigorous Solution of MESH Equations ...................................................................................163
Testing the McCabe-Thiele Method .................................................................................................... 164
Testing Underwood & Fenske ............................................................................................................ 169
CHAPTER 13. GAS ABSORPTION/STRIPPING..................................................... 172
Equipment for Absorption/Stripping.......................................................................................172
A Typical Absorber Design Problem .......................................................................................174
Solution Overview: ............................................................................................................................. 175
Detailed Solution: ............................................................................................................................... 176
Equilibrium Curve .....................................................................................................................177
Operating Line ...........................................................................................................................178
Only One Transferable Component .................................................................................................... 178
Interfacial Mass Transfer: Review ...........................................................................................181
Definitions of Transfer Coefficients .................................................................................................... 184
Determining the Interfacial Concentrations: (xi , yi ) ........................................................................ 185
Diffusion-Induced Convection ............................................................................................................ 188
Case I: Equimolar Counter-Diffusion of Gases ..............................................189
Case II: Diffusion of A Through Stagnant B ..................................................190
Height of a Packed Tower .........................................................................................................193
Evaluation of NTU's ..................................................................................................................197
Special Case: ya<<1 with concave downward EC. .......................................................................... 199
Transfer Unit ..............................................................................................................................200
Analogy with Double-Pipe HXer ..............................................................................................202
Relationships among HOx, HOy, Hx and Hy ..........................................................................203
Pressure Drop in Packed Beds ..................................................................................................204
Flooding According to McCabe, Smith & Harriott, 5th Ed. ................................................................ 207
Flooding According to Geankoplis ..................................................................................................... 207
Tower Diameter .........................................................................................................................210
CHAPTER 14. MEMBRANE SEPARATION ............................................................ 215
Examples .....................................................................................................................................215
Gases .................................................................................................................................................. 216
Liquids ................................................................................................................................................ 216
Equipment ..................................................................................................................................216
Separation of Gases....................................................................................................................220
Porous Membranes ............................................................................................................................. 220
Nonporous Polymer Membranes ........................................................................................................ 222
Example: Production of Enriched Air .....................................................................................224
Stage Cut............................................................................................................................................. 228
CHAPTER 15. LIQUID-LIQUID EXTRACTION ........................................................ 232
Equipment ..................................................................................................................................233
Triangle Diagrams .....................................................................................................................234

06-361 page v Spring, 2008
Nomenclature .............................................................................................................................237
Rectangular Diagrams ...............................................................................................................237
Flash Mixing of Two Liquids ....................................................................................................238
Single-Stage Extractor ...............................................................................................................240
Multi-Stage Extractor ................................................................................................................242
Operating Lines for Extractors ........................................................................................................... 243
Difference-Point Method .................................................................................................................... 244
Minimum Solvent Flowrate ................................................................................................................. 246
TABLE OF CONTENTS.................................................................................................. II
INDEX VI

06-361 page 1 Spring, 2017
Chapter 1. Course Introduction
SYLLABUS SUMMARY
Instructor
My name is Dennis Prieve and my office is DH 3120. I encourage each of you to come and
see me when you have any problems. Right after class is the best time; right before class is the
worst time (I’m usually preparing for class). If you don’t find me in my office, contact me by
email.
Textbook
C. J. Geankoplis, Transport Processes and

Separation Process Principles, 4th Ed., PTR
Prentice-Hall, Englewood Cliffs, NJ (2003). This
is a required text. You will need your own copy to
take exams (which are open book). I will also be
publishing my own notes on BlackBoard which is
a form of textbook in preliminary form.
Web Site
I will be making extensive use of BlackBoard

in this course. You will find the following there:
• lecture notes
• homeworks and solutions
• old exams from previous years
• Mathcad modules to help with homeworks
Grading
Your final letter grade will be assigned based primarily on your numerical average of the
following:
50% Two in-class exams (Feb. 22 and Apr. 5 – both are Wednesdays)
25% Final Exam (TBA)
10% Homework
10% Project Report
5% Attendance

06-361 page 2 Spring, 2017
Most of the grade will come from three exams: two in-class exams and a final exam given during
final-exam week. They will count 25% each. All of these exams are open-book and open-notes.
You will find examples of old exams on BlackBoard which can also be used as practice exams.
We will be having 11 weekly homework assignments usually due on Wednesday. The first
homework will be due a week from today and I’ll have more to say about homeworks in a later
lecture. Collectively, the homeworks will count 10% of the final average.
Attendance will count 5%. I consider regular attendance to be required for learning by the vast
majority of students. Of course, I also expect your full attention during class.
The remaining 10% comes from a term paper or project report which will be due near the end of
the semester. This is a 10-page written summary of some unit operation of your choosing, not
one of the unit ops covered in lectures. We will not have a homework assignment collected one
week so you can work on this project.
WHAT ARE “UNIT OPERATIONS”?

A typical process which a chemical engineer might work with is the production of gasoline
from crude oil. Any process can be subdivided into a number of steps which are performed in
sequence to go from some initial starting material (crude oil, in this case) to some final material
(gasoline). For example, we might start by heating the crude oil to lower its viscosity, then pump
the oil to the distillation column, where we then separate various components of the crude. A
unit operation is one of the steps of this sequence. This course is concerned primarily with
physical transformations (as opposed to chemical transformations; e.g. chemical reactor) of the
input stream.
process = sequence of “unit operations”
Objective of Course
The objective of this course is to learn the principles needed to size equipment used for
physical transformation of process streams. In particular, we will begin to appreciate the role of
mass transfer and phase equilibrium. Equipment which we will learn how to size includes:
• shell-and-tube heat exchangers (change the temperature of a stream)

• single and multiple-effect evaporators (remove water from liquid)
• batch still (separate a mixture of miscible components)
• tray columns for distillation (separate a mixture of miscible components)
• packed columns for gas absorption and stripping (separate a mixture of miscible
components)
• membranes for gas separation (separate a mixture of gases)
• liquid-liquid extractors (separate a liquid mixture of miscible components)

06-361 page 3 Spring, 2017
Overview of ChE Curriculum
In this course, we will apply the principles learned in the engineering science courses you
have had during the last three semesters; these will be applied to the design of equipment for
physical transformations. We might divide up the courses in the curriculum as follows:
Course Topics
intro conservation of mass
1) conservation of energy
thermo
2) criteria for phase/chemical equilibria
rates of approach to thermal equilibrium (heat

heat/mass transfer
transfer) or phase equilibrium (mass transfer)
unit operations design equipment for physical transformations
1) rate of approach to chemical equilibrium

kinetics
2) design equipment for chemical transformations
divide process into sequence of steps

process design
(synthesize flowsheet)
Chapter 2. Heat Exchange Equipment
DOUBLE-PIPE HXER
One particularly simple heat-exchange device is the double-pipe heat exchanger, shown in
the figure at right. It is called this because the basic component is constructed from two pipes,
one inside the other. One fluid (Fluid B)
flows inside the inner pipe while the second
fluid (Fluid A) flows in the annular space
between the two pipes. Of course, one fluid is
hotter than the other and heat flows through
the wall of the inner pipe from the hot fluid to
the cold fluid. To obtain larger heat exchange
area, several pipes are arranged side-by-side
and fittings are attached to allow the fluids to
contact the pipes in series so the overall flow
is the same as if only one very long pipe were
used rather than several shorter ones.

06-361 page 4 Spring, 2017
SHELL-AND-TUBE HXER
If very large heat exchange areas are required, the double-pipe heat exchanger design
becomes impractical. Instead, the shell-and-tube heat exchanger design is used.
Shown above is a bundle of small-diameter tubes which are arranged parallel to each other and
reside inside a much larger-diameter tube called the “shell”, much like strands of uncooked
spaghetti come in a tube-shaped container. The tubes are connected to a common space (called a
“manifold”) at either end so that the “tube fluid” enters the left side and is distributed equally
among all the tubes. At the right side, the fluid exits from each tube, is mixed together in a
second manifold, then leaves as a single stream. The second fluid, called the “shell fluid”, flows
in the space in between the outside of tubes. Baffle plates inside the shell force the shell fluid to
flow across the tubes repeatedly as the fluid moves along the length of the shell.
In the design above (called a 1-1 hxer), the flow is from left to right inside the tubes and from
right to left for the shell-side fluid. Another common design is shown below (called a 1-2 hxer).

06-361 page 5 Spring, 2017
Half of the tubes have flow from left to right while the other half have flow in the opposite
direction.
Below is a third design, called a 2-4 hxer.
Chapter 3. Fundamentals of Heat Transfer: Review
LOG-MEAN DRIVING FORCE

In the heat & mass transfer course, we learn that the rate of heat transfer q [=] J/s♠ is
proportional to the driving force ∆T = T' – T (where T' is the temperature of the hot fluid and T is
the temperature of the cold fluid) and proportional to the area A of contact between them.
=
q UA ∆T (1)
where the proportionality constant U is some heat-transfer coefficient.
In any real heat exchanger, the local driving

force ∆T will vary along the length of the
exchanger: say it varies between ∆T1 at one end
of the exchanger to ∆T2 at the other end. The
graphs below were produced by the Mathcad
document Hxer T profiles
*
(countercurrent).mcd which simulates the
temperature profiles in a counter-current double-pipe heat exchanger.
♠ An equal sign inside square brackets (i.e. “[=]”) should be read “has units of” rather than “equal to.”
* Mathcad documents and other files referred to in the Notes are usually available on the Blackboard website for this
course: under Handouts / Student Resources.

06-361 page 6 Spring, 2017
p j y p
R
Tho = 344.3 K <-----<-----<-----<-----<-----<-----<-----<-----<----- Thi = 377.6 K x
n This ratio should be nearly unity if integration is
400 = 0.99930
LT accurate
∆T1 = 6.402 K These two ∆T's must both be positive for the
simulation to be realistic; if one is negative,
350
Thi ∆T2 = 55.5 K change mc or mh; if both are negative check
the T's
Tci
300 3 kg
mh = 7.26 × 10 Thi = 377.6 K Tho = 344.3 K
hr
3 kg
mc = 2 × 10 Tci = 288.8 K Tco = 371.198 K
250 hr
0 0.2 0.4 0.6 0.8
Tci = 288.8 K ----->----->----->----->----->----->----->----->----->
xi Tco = 371.198 K watt D = 5 in L = 32.314 m
U = 653
2 ∆Tlm = 22.733 K
LT m ⋅K
60 Nonzero values must be supplied for the following heat capacities:

3 J 3 J
cph ≡ 2.85⋅ 10 ⋅ cpc ≡ 4.181⋅ 10 ⋅
kg ⋅ K kg ⋅ K
One of the following variables must be left unknown (assign it zero
40
value and it will be calculated from an overall energy balance):
∆T i kg
mh ≡ 7260⋅ Thi ≡ 377.6⋅ K Tho ≡ 344.3⋅ K
hr
20
kg
mc ≡ 2000⋅ Tci ≡ 288.8⋅ K Tco ≡ 0
hr
One of the following variables must also be left unknown (assign it
0 zero value and it will be calculated from the design equation):
0 0.2 0.4 0.6 0.8
watt
xi U ≡ 653⋅ D ≡ 5in L≡0
2
LT m ⋅K
After entering or changing values, press Ctrl-F9 to recalculate
sheet.
Since ∆T is a continuous function of position inside the heat exchanger, we will have to
integrate a differential form of this equation. At any axial position x inside the heat exchange,
the local heat flux (rate per unit area) is given by
dq
heat
 = = U ( x ) ∆T ( x )
flux
 (2)
J
dA
m2 − s
where the heat transfer coefficient as well as driving force in general depend on x.
Mathematically, dq is
dq = q(x+dx) – q(x) [=] J/s = W

pipe wall
where q(x) is the rate of heat transfer through the
subset of the heat exchanger between one end T'1 hot fluid T' T'2
q(x)
(denoted as x=0) and some arbitrary position x along
the length of the heat exchanger. Physically, dq T1 T T2
represents the rate of heat transfer through the slice
cold fluid
of heat exchanger having the area dA.
x=0 x=x x=L
dA is the heat exchanger area in this slice between x

and x+dx. If the inner pipe (of a double-pipe hxer) is a

06-361 page 7 Spring, 2017
cylinder of diameter D, then the heat exchange area of this slice is
dA = πD dx and A(x) = πDx
We are tempted to substitute this expression for dA into (2) and try to integrate both sides with
respect to x. This turns out to difficult because functions U(x) and especially ∆T(x) are unknown
functions. A better approach turns out to be solving the heat flux equation for dA:
dq
dA =
U ∆T
We then integrate over the entire heat exchanger (from x = 0 to x = L) to obtain the total area of
required for this heat duty, treating q as the independent (or integration) variable instead of x:
=A ( L ) A=
T q ( L ) qT
dq
=AT ∫ dA
= ∫ U ( q ) ∆T ( q )
=A ( 0 ) 0=q ( 0) 0
q ( L ) = qT
dq
or AT = ∫ U ( q ) ∆T ( q )
(3)
q ( 0) = 0
This serves as the general design equation for heat exchangers. To evaluate this integral, it’s
useful to think of q [= q(x)] as being the independent variable instead of x.
Example: To show how this done, let’s consider a simple example in which the flow of hot and
cold streams is countercurrent and no phase change is occurring in either fluid.
Consider steady-state countercurrent flow through a pipe wall

double-pipe heat exchanger. A steady-state energy
balance on the hot fluid (shown in orange color) T'1 hot fluid T' T'2
q(x)
gives
T1 T T2
rate in = rate out
(4) cold fluid
m=′H ′ m′H1′ + q
where m is the mass flowrate of the stream (units are x=0 x=x x=L
kg/sec), H is the enthalpy per unit mass (units are
J/kg), the subscripts and primes have the following meaning:
• unprimed refers to the cold stream,

• prime ( ' ) refers to the hot stream,
• “1” refers to the left end of hxer, and
• “2” refers to the right end of hxer.
A similar balance on the cold fluid (shown in blue color) gives

06-361 page 8 Spring, 2017
rate in = rate out

(5)
mH1 + q =mH
Solving (4) and (5) for q:
m  H ( x ) − H1  = q ( x) = m′  H ′ ( x ) − H1′  (6)
 
rate of energy rate of energy
gained by lost by
cold stream hot stream
Now, let’s relate each enthalpy to the appropriate temperature. In general, enthalpy changes can
have both latent heat (caused by change in phase) and sensible heat (caused by change in
temperature) contributions.
=
dH c p dT + λ dy
 
sensible latent
heat heat
where
• dH is the change in enthalpy (energy/mass)

• cp is the heat capacity (energy/mass/degree) of the fluid, also known as “specific heat”
• dT is the change in temperature (degree)
• λ is the heat of vaporization (energy/mass) > 0
• dy is the change in mass fraction of fluid vaporized
When the fluid consists of two phases (liquid and vapor), which have different heat capacities
cpL and cpV, we use weighed average of the two heat capacities*
cp = (1-y)cpL + y cpV
where y is the mass fraction of vapor present and 1-y is the mass fraction of liquid present. For
the time being, let’s assume:
Assumption #1: no phase changes occur in either fluid (dy = 0)
Now we only have to worry about contributions from changes in the temperature (sensible
heat):
dH = cp dT
Integrating over the temperature change allows us to evaluate the change in enthalpy
* A vertical line drawn on the left side of a paragraph denotes material which was omitted in class

06-361 page 9 Spring, 2017
using using
assumption assumption
Hb #1 Tb #2
∫ dH
= ∫=
c p dT c p (Tb − Ta )
 
H
a
Ta δT
δH
In general, the specific heat (cp) depends on temperature. But if the temperature change is not
too large — say 10-20° — then the specific heat doesn't change too much. Let's assume we can
neglect these small changes:
Assumption #2: constant specific heat for both streams
Then we can factor cp out of the integral and the integration becomes trivial:
H – H1 = cp(T – T1)
and H ′ − H ′1= c ′p (T ′ − T ′1 )
where cp and c'p are the specific heats of the cold and hot streams. Thus (6) can be rewritten as:
at x=x: mc p T ( x=
) − T1  m′c ′p T ′ ( x )=
 − T ′1  q ( x ) (7)
For example,
at x=0: ( )
mc p (T1 − T1 ) = m′c ′p T ′1 − T ′1 = q ( 0 ) = 0
In particular, we are interested in the total rate of heat transfer over the entire length L of the
exchanger:
at x=L: ( )
mc p (T2 − T1 )= m′c ′p T ′2 − T ′1 = q ( L ) ≡ qT
Using (7), I can show that the temperature of either stream [i.e. T(x) or T’(x)] is a linear
function of the amount of heat transferred, q(x):
T
 1 
T ( x=) T1 +   q ( x) T' vs q
∆T2
 mc p 
  T vs q 1
 1  ∆T1
T ′ ( x=
) T1′ +   q ( x)
mcp
 m′c ′p  ∆ T vs q
 
∆T1 ∆T2
Since both temperatures are linear functions of q, q qT
0
their difference ∆T is also a linear function of q. A
straight line has the very nice property that its slope

06-361 page 10 Spring, 2017
is the same at every point:
d ( ∆T ) ∆T2 − ∆T1
=
slope = const
= (8)
dq qT − 0
I can easily evaluate the driving force (∆T) at the two ends of the heat exchanger, and I know the
total heat duty (qT), so I can calculate this slope. The dq and d(∆T) which appear in (8) represent
the incremental rate of heat transfer and the change in driving force ∆T which occurred over a
differential length dx of the heat exchanger(from x to x+dx).
Assumption #3: U is constant
Now we are in a position to evaluate the integral in (3) which becomes
qT
1 dq
AT =
U ∫ ∆T ( q )
0
Next, we multiply and divide the integrand by d(∆T):
qT
1 dq d ( ∆T )
AT =
U ∫ d ( ∆T ) ∆T
0
Since the slope is a constant, so is the inverse of the slope dq/d(∆T) which can be factored out of
the integral. Its value is given by the inverse of (8). After substituting this constant value, we
have
∆T ∆T
1 dq 2
d ( ∆T ) 1 qT 2
d ( ∆T ) 1 qT ∆T
=AT = ∫ = ∫ ln ( ∆T ) ∆T2
U d ( ∆T ) ∆T U ∆T2 − ∆T1 ∆T U ∆T2 − ∆T1  1
∆T ∆T
1 1
∆T2
ln ∆T2 − ln ∆T1 =
ln
∆T1
In the above we have also changed the limits of integration to reflect the fact that we are now
integrating with respect to ∆T instead of q: on the left end of the heat exchanger where q(0) = 0,
the ∆T = ∆T1; whereas on the right end of the heat exchanger where q(L) = qT, the ∆T = ∆T2.
This leaves us with
∆T2
ln
q ∆T1
AT = T
U ∆T2 − ∆T1

06-361 page 11 Spring, 2017
∆T − ∆T1
Solving for qT: =
qT UAT ∆Tlm where ∆Tlm =2 (9)
∆T
ln 2
∆T1
is called the log-mean temperature difference (LMTD). It represents the average driving force
for heat transfer.
Summary: The general design equation for heat exchangers is (3). (9) is a simplified form which
is valid whenever 1) U(q) can be replaced by a suitable constant and 2) ∆T is a linear function of
q. For most of the design problems in this course, we will use (9) as the design equation.
We have just shown that these two conditions are met by making the following assumptions:
• U and cp's are constant

• hot and cold streams flow countercurrently
• δH was all sensible heat for both streams (i.e. no phase changes)
In the homework, you will show that (9) is also correct if the hot and cold streams flow co-
currently; you will also show that δH can also be all latent heat and the LMTD is still the
appropriate average driving force. For example, if you use condensing steam as your hot fluid,
as long as the steam remains saturated:
for saturated steam: T’(x) = const = Tbp
where Tbp is the boiling point of the steam at some pressure.
Lecture #2 begins here
One last detail: in (1) we said A is the area of the inner pipe. Owing Di
to the finite thickness of the pipe wall, the inner and outer diameter of Do
the inner pipe are different (Di < Do), which implies that the outer
surface Ao of the pipe wall is larger than the inner surface Ai of the pipe
wall. Either area can be used to calculate the heat duty, but a different
heat transfer coefficient must also be used to obtain the same heat duty: Ui and Uo are defined
such that:
qT = U o Ao ∆Tlm = U i Ai ∆Tlm
where Ai = πDiL and Ao = πDoL
From the figure, it is clear that Do>Di. This implies that Ao> Ai so that Uo<Ui.

06-361 page 12 Spring, 2017
Co-Current versus Counter-Current Flow
Given a choice between cocurrent or countercurrent flow, which is preferred? It turns out
that countercurrent flow is usually more efficient. To see why, let’s look at an example:
Example: A cold fluid is being heated from 15.8ºC to 62.9ºC while the hot fluid is cooled from
104.6ºC to 71.3ºC. Compare ∆Tlm for co-current flow to that for countercurrent flow.
104.6ºC → 71.3ºC
Solution: For co-current flow, we have the adjacent temperatures
shown at right. This corresponds to the following ∆T’s at the two ends
of the heat exchanger: 15.8ºC → 62.9ºC
∆T1 = 104.6 – 15.8 = 88.8ºC

∆T2 = 71.3 – 62.9 = 8.4ºC
The LMTD given by (9) as
8.4°C − 88.8°C
( ∆Tlm )cocurrent =
8.4°C
=
34.1°C
ln
88.8°C
71.3ºC ← 104.6ºC
For counter-current flow, we the adjacent temperatures shown at right.
This corresponds to the following ∆T’s at the two ends of the heat
exchanger: 15.8ºC → 62.9ºC
∆T1 = 71.3 – 15.8 = 55.5ºC

∆T2 = 104.6 – 62.9 = 41.7ºC
The LMTD given by (9) as
41.7°C − 55.5°C
( ∆Tlm )countercurrent =
41.7°C
=
48.3°C
ln
55.5°C
Notice that, for exactly the same heat duty, the average driving for countercurrent flow is higher
than for co-current flow. This means that less heat exchanger area is needed if the flow is
countercurrent.
Non-Linear Relation Between q(x) and ∆T(x)
The general design equation (3) can be replaced by the much simpler form of (9), provided 1)
U(x) can be taken to be a constant over the entire exchanger and 2) ∆T is a linear function of q.
In the analysis at the beginning of this chapter, we showed that the latter is a reasonable
approximation if the flow is counter-current and no phase change occurs. In the homework, you
will show that this can be extended to 1) co-current flow with no phase change and 2) isothermal
phase change in one fluid. So you might well ask “when can’t (9) be used?

06-361 page 13 Spring, 2017
To give one example, suppose the hot fluid is superheated steam. Over part of the heat
exchanger surface, we have cooling of the hot vapor to its saturation temperature. Over the rest
of the surface, we have condensation of the saturated steam. The result is a nonlinear
relationship between ∆T(x) and q(x).
Tho = 417.567 K <-----<-----<-----<-----<-----<-----<-----<-----<----- Thi = 500 K

500 1
Asub Asub+ A
AT AT
450
Thot or Tcold
q(x)/qT
Tbp ( ps)
400 0.5
Hot fluid T
350
Cold fluid T
q(x)/qT
300 0
Tci = 3000 K 0.2 0.4 0.6 0.8 Tco = 400 K
----->----->----->----->----->----->----->----->----->
x/L
Figure 1
Figure 1 above shows the hot and cold temperature profiles and the cumulative heat duty
q(x). Figure 2 below shows the corresponding ∆T(x) and ∆T(q) functions. Notice that neither
function is linear over the entire length of the exchanger. But note that ∆T(q) consists of two
linear segments: one for the region in which superheated vapor is cooling to its saturation
temperature and a second line for the region in which saturated vapor is condensing.

06-361 page 14 Spring, 2017
150
deltaT vs q/qT
deltaT vs x/L∆T2
delta T (degC)
100 ∆T1
50
0
0 0.2 0.4 0.6 0.8
q/qT or x/L
Figure 2
Since ∆T(q) is piecewise linear, we might evaluate (3) by breaking the integral into two parts
and taking advantage of the linearity over each region:
qT q1 qT
dq dq dq
=AT ∫=
U ( q ) ∆T ( q ) ∫ U ( q ) ∆T ( q ) + ∫ U ( q ) ∆T ( q )
0 0 q1
q1 q − q1
or =AT + T (10)
U1∆Tlm1 U 2 ∆Tlm 2
where q1, U1 and ∆Tlm1 are the heat duty, overall heat-transfer coefficient and log-mean driving
force for the region of the hxer surface in which saturated vapor is condensing; qT – q1, U2 and
∆Tlm2 are the same quantities evaluated for the other region in which superheated vapor is being
cooled to its saturation temperature.
Example: evaluate the area required for the situation shown in Figure 1 in which 1,000 lb/hr of
water is being heated from 300 K to 400 K by superheated steam at 500 K and 4 atm pressure
using U = 700 btu/ft2-hr-ºF.♣
Solution: using the steam table,♥ we look up the enthalpies of steam:
Hhi = 1223 btu/lb (4 atm, 500K) and Hvap = 1178 btu/lb (4 atm, 418K)
♣ Condensing steam has a different U compared to cooling of a vapor. We should really use two different U’s for
the two different regions but in this example we will use the same U for both regions.
♥ Appendices A.2-9 (saturated steam) and A.2-10 (superheated steam).

06-361 page 15 Spring, 2017
Hho = Hliq = 216 btu/lb (4 atm, 418K)
By taking the outlet enthalpy to equal the enthalpy of saturated liquid, we are assuming that the
steam leaves the shell side as condensate at its boiling temperature and at the same pressure at
the inlet (4 atm). The flowrate of steam can be calculated from an overall energy balance:
mh(Hhi – Hho) = qT = mccpc(Tco – Tci) = 52.75 kW
Substituting mc = 1,000 lb/hr, cpc = 1 btu/lb-ºF, Tco = 400 K and Tci = 300 K along with the
values of the enthalpies above, we obtain qT = 52.75 kW and mh = 187 lb/hr. The part of the heat
duty associated with condensing steam is
q1 = mh(Hvap – Hliq) = 50.27 kW
leaving qT – q1= 2.48 kW associated with the cooling of superheated vapor.
To calculate the ∆Τ’s appearing in (10), we need to evaluate the hot and cold temperatures at
the point in the heat exchanger where the superheated steam becomes saturated steam. We will
denote the temperatures at this point Th1 and Tc1. Th1 = 417.6 K is the temperature of saturated
steam at the inlet pressure (4 atm). The other temperature can be calculated from
q1 = mccpc(Tc1 – Tci) = 50.27 kW
which yields Tc1 = 395.3 K. We now summarize the temperatures:
hot fluid: 417.6 ← 417.6 ← 500

cold fluid: 300.0 → 395.3 → 400
The corresponding three ∆T’s are
∆T1 = 117.6 K, ∆T3 = 22.3 K and ∆T2 = 100 K
The two log-mean ∆T’s are
∆Tlm1 = (117.6 K, 22.3 K)lm = 57.3 K
∆Tlm2 = (22.3 K, 100 K)lm = 51.8 K
Substituting all results into (10):
AT = 2.38 ft2 + 0.13 ft2 = 2.51 ft2
Comment #1: If instead, we blindly applied (9) using the two end-point ∆T’s, we would have
calculated
∆Tlm = (117.6 K, 100 K)lm = 108.5 K

06-361 page 16 Spring, 2017
which considerably overestimates the average driving force, leading to an large underestimate of
the area required: 1.32 ft2 instead of 2.51 ft2 (almost a factor of 2 error).
Comment #2: A much better approximation would be to ignore any superheating or subcooling
of the steam: just assume the steam enters and leaves at the saturation temperature.
hot fluid (approx): 417.6 ← 417.6

cold fluid: 300.0 → 400
This leads to ∆Tlm = (117.6 K, 17.6 K)lm = 52.7 K which lies in between the two exact ∆T’s. In
the next chapter (Evaporation), this is exactly the approximation we shall make in preliminary
designs.
RESISTANCES IN SERIES
Overall Coefficients (U) vs. Film Coefficients (h)
We said the design equation for heat exchangers is given by (9):
=
qT UAT ∆Tlm (9)
We have just discussed how to calculate ∆T in this equation. Now we will turn our attention to
calculating U. Heat transfer coefficients are usually estimated from empirical correlations of
experimental data. However the correlations usually do not give U directly; instead, the
correlations are usually for the “film coefficients” hi and ho, which are related to U by the
following formula:
1 1 D x D 1
= o + w o +
U o hi Di kw Dlm ho
We will now review the theory of heat transfer which leads to this result.
Consider heat transfer through the inner pipe of a double-pipe heat exchanger. Let’s take a
thin axial slice (of length dL) out of the heat exchanger and examine the temperature profile
along a radius to the pipe. The labelled temperatures in Figure 3 below assume (somewhat
arbitrarily) that the hot fluid flows in the inner pipe while the cold fluid flows in the annular
space between the two pipes.

06-361 page 17 Spring, 2017
Figure 3
One can identify three regions in which the temperature varies sharply. These correspond to the
three resistances to heat transfer: 1) the pipe wall itself, 2) the fluid film inside the pipe and 3)
the fluid film outside the pipe.
Ui and Uo are called overall heat-transfer coefficients because when each is multiplied by
the overall driving force and the appropriate area, they give the heat flux: ♣
definition of U=
i: dq U i dAi (Th − Tc ) (11)

πDi dL
definition of=
Uo: dq U o dAo (Th − Tc ) (12)

πDo dL
The overall driving force Th – Tc can be decomposed into the sum

of the three separate drops across each of the three regions
♣ The temperature measured at the center of the pipe (T in Figure 3) differs slightly from the “mixing cup”
i
temperature Th which instead is used to characterize the average temperature of the fluid inside the pipe. Similarly
the temperature measured at the outer edge of the annulus (To in Figure 3) differs slightly from the mixing cup
temperature Tc of the cold fluid. The mixing-cup temperature is what would be measured if a sample of the fluid
were collected in a cup and allowed to come to a uniform temperature without any addition or loss of heat. The
mixing cup temperature is also what is used in evaluating the enthalpy of a stream for purpose of an energy balance.

06-361 page 18 Spring, 2017
contributing resistance to the overall heat transfer (see Figure 3 above):
Th − Tc = (Th − Twh ) + (Twh − Twc ) + (Twc − Tc ) (13)
The temperature drop across the fluid film on the inside of the pipe wall is Th – Twh. The
film heat-transfer coefficient hi is defined as the proportionality constant between heat flux
dq/dAi and this driving force:
definition of hi: =dq hi dAi (Th − Twh ) (14)


πDi dL
Since we are calculating the flux through the inner film, we use the area dAi of the inner surface
of the pipe wall. Similarly, the film coefficient ho is defined as the proportionality constant
between heat flux dq/dAo and the driving force across the outer film:
definition of ho: =dq ho dAo (Twc − Tc ) (15)


πDo dL
Film coefficients and their definitions (14) and (15) are used to describe convective heat-
transfer rates across fluid films undergoing turbulent fluid flow. By contrast, heat transfer
through the pipe wall occurs by pure conduction. Nonetheless, the relationship between rate of
heat transfer, driving force and heat transfer area can be written in a similar form. For steady-
state conduction through an annulus, the relationship is:
kw
=
conduction through pipe wall: dq dAlm (Twh − Twc ) (16)
xw
where kw is the thermal conductivity of the pipe wall and xw is the wall thickness
Do − Di
xw = ro − ri =
2
dAo − dAi π ( Do − Di ) dL
and where dAlm = = = πDlm dL

ln 
dAo   Do 
dAi 
ln 
  Di 
is the log-mean of dAo and dAi. A relationship similar to (16) (including the log-mean area) was
probably derived in your heat-and-mass transfer class.*
* See Section 4.2B in Geankoplis; in particular, see eq. (4.2-10)

06-361 page 19 Spring, 2017
Notice that the same dq is used as the rate of heat transfer in (11) through (16). This
sameness is a consequence of assuming steady-state heat transfer. At steady state, there can be
no accumulation of heat at any location.
The two areas dAo and dAi do not differ by very much so neither do the values Uo and Ui.
The choice of which area to use is arbitrary. We will proceed using the outer area. Then (12)
can be written as
dq
= Th − Tc
U o dAo
Substituting (13), (14), (15), and (16):
dq
=Th − Thw + Thw − Tcw + Tcw − Tc
U o dAo       
dq dq dq
hi dAi kw ho dAo
dAlm
xw
Dividing through by dq dAo leaves:
1 1 D x D 1
= o + w o + (17)
U o hi Di kw Dlm ho
after substituting the expressions relating the dA’s to D’s and dL. Starting with (11) instead of
(12), the analogous expression for Ui is
1 1 xw Di 1 Di
= + +
U i hi kw Dlm ho Do
If we divide (17) by the outer area (Ao = πDoL) we obtain what Geankoplis calls the total
resistance to heat transfer:
1 1 xw 1 1
= + + =
U o Ao hi Ai kw Alm ho Ao U i Ai
    
total resistance of resistance of resistance of total
resistance inner film pipe wall outer film resistance
To see why 1/UA might be called the resistance, recall again the design equation (1):
1
=
q U A ∆T or ∆
T=q
I
V UA
R

06-361 page 20 Spring, 2017
The heat duty q is like the electric current I in Ohm’s law (V = IR) and the driving force ∆T is
like the voltage drop V across a resistor having resistance 1/UA. In this expression the total
resistance is either 1/UiAi or 1/UoAo. Since the same current is passing through all three
resistors, the resistors must be arranged in series.
Fouling Factors
The usefulness of analogies is that we can extend our knowledge in one area using
knowledge in other. For example, if there are additional layers through which heat must be
conducted, then we can just add their resistances to get the total. A situation involving additional
layers is:
fouling — deposition of dirt, scale or any solid deposit on one or both sides of the tube
Fouling generally reduces heat transfer by imposing an additional resistance. It's effect is
accounted for by use of
k
fouling factors: hd = d
xd
which are calculated from simple conduction theory, assuming the fouling layer is planar film of
thickness xd; compare with coefficient in (16).
There may be a fouling factor for both the inside and the outside of the pipe. Adding these
resistances to the others:
1 1 1 x D 1 Do 1 Do
=+ + w o + +
U o ho hdo kw Dlm hdi Di hi Di
 
 
fouling fouling
outside pipe inside pipe
By means of the electrical analogy, we are able to guess the form of the correction without
having to go through the analysis.

06-361 page 21 Spring, 2017
CORRELATIONS TO ESTIMATE h’S

In some of the problems in Hwk #1, we are given the values of the 1-phase heat transfer
coefficients (the hi and ho). Now we are going to turn our attention to how these can be
estimated if they are not otherwise known. The empirical correlations appropriate for a given
situation usually depend on three questions:
1. What is the geometry? (e.g flow inside pipe, flow across bank of tubes, flow around sphere)
2. Is the flow turbulent or laminar?
3. Does a phase change occur?
Our book presents a number of common correlations which cover the most common situations:
Sect. 4.5, 4.6 - without phase change

laminar flow inside tubes – Sect. 4.5B (not covered in class)
turbulent flow inside tubes – Sect.4.5C-F♣
Sect. 4.8 - with phase change (not covered in class)
For flow inside long tubes, the criteria for turbulent flow is simple: just calculate the
Reynolds number:
ρ v Di GDi
=
N Re =
µ µ
where ρ = fluid density

µ = fluid viscosity
v is the cross-sectional average fluid velocity
Q
The latter quantity is defined as v =
Si
where Q = volumetric flowrate through tube (units of volume per time)

Si = πDi2/4 = cross-section area of tube
Another term which is sometimes used to calculate Reynolds number is the mass velocity:
ρQ m
G≡ρ =
v = (units of mass per area per time)
Si Si
The mass velocity is often used in problems involving gas flow where both density and velocity
depend on pressure, but their product is independent of pressure (since it represents mass
♣ You should make a special effort to read these pages before working the HWK problems, since I don’t have time
in class to present all the details.

06-361 page 22 Spring, 2017
flowrate and at steady state, a mass balance requires the mass flowrate to be the same at all axial
positions). However Reynolds number is calculated, the criteria for turbulent flow is as follows:
NRe > 6000 ⇒ flow is turbulent

NRe < 2100 ⇒ flow is laminar
2100 < NRe < 6000 ⇒ flow is in transition
hi for Turbulent Flow In Long Pipes
For turbulent heat transfer (NRe > 6000) of high Prandtl number (0.7 < NPr < 16,000) fluids
in long (L/D > 60) pipes, the individual film coefficient inside a pipe can be estimated using the
Sieder-Tate equation [eq. (4.5-8) in G]:♦
0.8 1/ 3 0.14
hi Di  GDi   c pi µi   µi 
= 0.023       (18)
ki µi 
  ki   µiw 
  
N Nu N Re N Pr
where ki = thermal conductivity of fluid inside tube

cpi = heat capacity (specific heat) of fluid inside tube
µi = viscosity of fluid inside tube
ρQ m
and where G≡ρ =
v =
Si Si
is called the mass velocity (units of mass per area per time). Unless specifically indicated, all
fluid properties are evaluated at the mixing-cup temperature of the fluid:
µi, ki, cpi evaluated at T of fluid inside tube
The subscript “iw” indicates the property is to evaluated at the inner wall temperature:
µiw evaluated at the wall temperature Tiw
Experimentally, one finds that heat transfer coefficients can be different for heating and cooling
over the same temperature range. The reason is that the viscosity at the wall is very important in
determining transport and µw can be very different from µ. This is accounted for in the µ/µw
factor.
♦ The coefficient of 0.023 in this equation is the value which appears in McCabe, Smith & Harriott. Geankoplis
gives a value of 0.027 instead. In homeworks and exam, you can use either value. The Notes use 0.023 because
most of the homework problems using this equation were taken from MS&H.

06-361 page 23 Spring, 2017
Which T do you use for property evaluations?
When the inlet and outlet T’s of a heat-transfer fluid differ, at what T do you use to evaluate
the properties? Strictly speaking, you should use the local value of T(x), which then leads to
different U at each x. In turn, (3) must be used as the design equation. This is significantly more
work than using (9).
In most cases, U does not vary significantly with local temperature. Then a good
approximation is to treat U as a constant evaluated using the arithmetic mean values for T and
Tw:
T + To T + Two
T = i and Tw = wi
2 2
where the subscript “i” refers to the inlet and “o” refers to the outlet.* Since the wall temperature
Tw depends on the heat transfer coefficient, the determination of h is often a trial-and-error (see
Example 4.5-2).
Water is often used in heat-transfer equipment. The following simplified equation contains
the variation in physical properties with temperature and can be used in the temperature range 4
< T < 105°C:
v 0.8
for liquid water: hL =1429 (1 + 0.0146T °C ) (19)
D 0.2
The two coefficients in this equation have some weird units (involving fractional exponents).
Rather than giving those weird units, we just point out that the constants were chosen to work
with SI units: the equation yields hL in W/(m2-°K) if v is expressed in m/s and D is expressed in
meters. Similar simplified correlations exist for air at 1 atm or for organic liquids:
v 0.8
for air at 1 atm: hL = 3.52
D 0.2
v 0.8
for organic liquids: hL = 423
D 0.2
Effect of Duct Shape on hi
Although the Seider-Tate equation (18) was developed

for turbulent flow inside pipes having circular cross-section,
it can also be used to estimate heat transfer coefficients inside
* This represents a change in notation: previously “i” stood for inner surface while “o” stood for outer surface.

06-361 page 24 Spring, 2017
pipes (or ducts) having noncircular shape: for example, the annular space between the inner and
outer pipes of a double-pipe heat exchanger (see blue region in figure at right).
The main question is: what is used in (18) for Di? Note that an annulus has two diameters,
which are labelled DA and DB in the figure above. The answer is to use the hydraulic diameter
which is defined as
S
DH ≡ 4
P
where S is the cross-sectional area of the duct and P is its wetted perimeter. For a circular pipe
of inside diameter Di, the area and wetted perimeter are
π 2
Scircle = Di and Pcircle = πDi
4
Note that the hydraulic diameter is just the usual diameter for a circular pipe. For the annulus in
the figure above, the area and wetter perimeter are
=
Sannulus
π 2
4
( 2
DB − DA ) and π ( DB + DA )
Pannulus =
DB2 − DA2
so the hydraulic diameter is D=
H = DB − DA
DB + DA
Chapter 4. Shell-and-Tube Heat Exchangers

The value of ho for shell-and-tube heat exchangers depends on arrangement of tubes and
baffles inside the shell. We will first describe typical arrangements and then give a common
correlation for ho.
TUBE PITCH
The bundle of tubes in a shell-and-tube heat exchanger can be stacked in one of two different
ways

06-361 page 25 Spring, 2017
The use of triangular pitch allows the tubes to be more tightly packed -- more tubes and therefore
more area per unit volume of shell. This makes the shell cheaper. On the other hand, the square
pitch has the advantage that it is easier to clean. You can get a long brush in between two rows
or between two columns. Of course, the shell has to be pulled off of the tube bundle before you
can get at it in either case.
Regardless of which geometry is used, the center-to-

center distances between adjacent tubes is called the tube
pitch.
BAFFLES
To some extent, it is advantageous to reduce the cross-sectional area available for flow. For
the same volumetric flowrate, this will increase the mass flux or the mass velocity of the fluid
flow, which will increase ho. Of course, this also increases the pressure drop and the pumping
cost.
A common technique for increasing ho

is to install baffle plates, which partially
block the cross-sectional area on the shell
side. A baffle plate is a disk whose
diameter is the inside diameter of the shell:
Ds = diameter of shell
with holes drilled through it for the tubes

to pass through. Part of the disk is cut off
to form what is called the baffle window,
which is the cross section available to the shell-side of the fluid. When the height of the window
is 1/4 of the shell diameter, we call the plate a 25% baffle.
A number of these plates is placed along the length of the exchanger with the location of the
baffle window alternating between the top and bottom of the plate.

06-361 page 26 Spring, 2017
A second advantage of these plates is that they prevent the formation of large stagnant regions
(having low heat transfer rates) which usually accompany “channelling” of the fluid in a
direction path connecting inlet to outlet. The spacing between the baffle plates is called the
baffle pitch:
P = baffle pitch
Typical baffle pitch is a fraction of the shell diameter:
0.2 < P/Ds < 1
Besides reducing the area available for flow on the shell side, the baffles also force the fluid
across all of the tubes, which normally will enhance the heat transfer coefficient.
hO FOR SHELL-AND-TUBE EXCHANGER
Reference:♠ McCabe, Smith & Harriott, 5th ed., p432-436
An approximate but generally useful correlation for estimating the shell-side coefficient in a
shell-and-tube heat exchanger is the Donohue equation:
0.6 0.33 0.14

ho Do D G   c po µo   µo 
= 0.2  o e     
ko  µo   ko   µ wo 
(20)
where µo, ko and cpo are properties of the shell-

side fluid and Do is the outside diameter of the
tubes. The form of this equation is very similar to
the Sieder-Tate equation, except the exponents
♠ A PDF file containing a scan of these pages can be found under “Handouts” in Blackboard.

06-361 page 27 Spring, 2017
and coefficients are different. Another difference is that the mass velocity Ge is a weighted
average
Ge = Gb Gc (geometric mean) (21)
of the mass velocity through the baffle window♣
m Kg
=
Gb =[ ] (22)
Sb m2s
and the mass velocity for crossflow perpendicular to the tubes
m
Gc = (23)
Sc
where m is the mass flow rate (Kg/s), Sb is the open area available for flow of the shell-side
fluid through the baffle window and Sc is the open area available for crossflow perpendicular to
the bank of tubes at the widest point in the shell. The formulas for calculating these two areas
are given in the Appendix starting on page 37.
MULTIPASS CONSTRUCTION
Just as we can improve the heat transfer coefficient on the shell side by reducing the baffle
pitch, we can improve the heat transfer coefficient on the tube side by restricting the flow to half
of the tubes. Both have the effect of reducing the area available for flow, thereby increasing the
fluid velocity and the heat transfer coefficient.
Figure 4: a 1-2 S&T Hxer
Suppose we partition the manifold in such a way that the inlet fluid can only enter half of the
tubes. The other half of the tubes are used for return flow. We say that such an exchanger has
two tube passes.
♣ An equal sign enclosed in square brackets (i.e. [=]) should be read “has units of …”

06-361 page 28 Spring, 2017
Although two tube passes improves hi, one of the passes will be co-current. A 2-2 exchanger
improves both hi and ho, and in addition, the flow is more nearly countercurrent. We could also
partition the shell side with a horizontal baffle in addition to the vertical baffles, producing two
shell passes.
Figure 5: a 2-2 S&T Hxer
The most commonly used designs are:
• 1-2 exchanger
• 2-4 exchanger
where the first number is the number of shell passes and the second number is the number of
tube passes.
An even number of tube passes is desirable because then input and output of the tube-side
fluid occur on the same end of the exchanger. This allows the other end of the exchanger to
"float" in response to thermal expansion of the metal, which can otherwise be a serious problem
when the shell changes temperature during startup or shutdown.
"True" Mean Temperature Difference
Recall that the “design equation” for heat exchangers is
qT = Uo Ao ∆T (24)
For double-pipe heat exchangers, the appropriate average driving force ∆T is usually the log-
mean ∆Tlm of the two ∆T evaluated at either end of the heat exchanger. For multipass shell-and-
tube heat exchangers, the flow is neither counter-current nor co-current. Instead, the flow is
some combination of counter-current, co-current and cross flow. As a consequence, the average

06-361 page 29 Spring, 2017
driving force is no longer the log mean. Instead the appropriate average driving force is called
the true mean temperature difference (MTD) and is somewhat less than the log-mean:
∆Tm = FT ∆Tlm (25)
FT is a correction factor which is unity for:
pure co-current flow 


pure counter-current flow  
→ FT = 1
1-1 shell-and-tube heat exchanger 

For pure crossflow to a bank of tubes, or for a mixture of crossflow and counter-current flow, the
factor is less than unity:
FT = FT (Y, Z, type of flow) ≤ 1
where the “type of flow” includes:
 1-2 shell-and-tube
 2-4 shell-and-tube
 cross flow to a bank of tubes
where Y and Z are ratios of temperature differences:
T −T T − Tho
Y = co ci and Z = hi Tho Thi
Thi − Tci Tco − Tci
where the subscript “c” refers to the cold fluid Tci Tco
while the subscript “h” refers to the hot fluid and
the subscript “i” refers to the inlet while “o”
refers to the outlet.* Using this same notation,
the log-mean ∆T is calculated from
(Thi − Tco ) − (Tho − Tci )

∆Tlm =
T − Tco
ln hi
Tho − Tci
This is the log-mean ∆T for counter-current flow: it is used for all shell-and-tube heat
exchangers, regardless of type (1-1, 1-2 or 2-4).
* This represents a change in notation: previously “i” stood for inner surface while “o” stood for outer surface.

06-361 page 30 Spring, 2017
FT is usually plotted as a function of Y for

different values of Z. Such a plot is sketched at
right. More accurate plots can be found in
Geankoplis’s book:
1-2 HX → Fig. 4.9-4a
2-4 HX → Fig. 4.9-4b
Note that as Z→0, FT → 1 over the entire range of

Y<1. This special case is applicable for condensing
vapors. As long as the vapor is condensing, its
temperature stays at the boiling point:
Thi = Tho = Tbp
so that Z = 0 and FT = 1
for condensing steam as the hot fluid. Similarly if we have boiling water as the “cold” fluid, its
temperature doesn’t change (i.e. Tci = Tco = Tbp) and then we have
Y = 0 and FT = 1
Although the flow in the exchanger is complex, we calculate ∆Tlm in (25) as if the flow were
counter-current. The deviation caused by not being counter-current is included in the values of
the empirical correction factor FT.
An Example:*
A steel mill needs to cool 3,000 gal/hr of a dilute acid from 200°F to 130°F using 5,000 gal/hr of
80°F river water. A shell-and-tube heat exchanger is available (see description below).
Determine if this heat exchanger is sufficient for this task.
Description of available heat exchanger:
type: 1-2
tubes: 158 1-inch, 14BWG stainless tubes, 16 ft long, stacked in 1¼ inch square pitch
baffles: 25% with 6 inch pitch
shell: 20 inch inside diameter, mild steel
Solution: The strategy is to compare the required heat duty to the heat duty which can be
expected under these conditions. If the former is larger (smaller), the heat exchanger is not (is)
capable of performing this task.
* Shell&tube hxer example.mcd

06-361 page 31 Spring, 2017
First we calculate the required heat duty: the rate at which heat must be removed from the
dilute acid is
 gal   lbm   hr   btu 

qT =  h c ph (Thi − T
,reqd m =ho )  3000   8.34   1  ( 70°F )
hr   gal   3600 sec   lbm-°F 

lbm
6.94
sec
btu
= 484
= 510 kW
sec
In making this calculation, we have assumed that the properties (density and heat capacity) of
dilute acid are essentially those of pure water. This is to be compared with the available heat
duty, which is calculated from the design equation:
from (24): qT, available = Uo Ao ∆Tm
True Mean ∆T. One of the four temperatures is not given: we have to determine the outlet
temperature of the cold stream. The rate of energy lost by the hot stream is the rate of energy
gained by the cold stream
qT m h c ph (Thi − T
= =ho ) m
 c c pc (Tco − Tci ) (21)
Assuming the physical properties of dilute acid and river water are the same as pure water, we
can take cph = cpc so that heat capacity can be cancelled out of the second equation above,
leaving
Tco − Tci m h c ph m h 3000 gal hr

= ≈ = = 0.6 (22)
Thi − Tho m c c pc m c 5000 gal hr
From the problem statement we have
Thi = 200°F Tho = 130°F Tci = 80°F
Substituting these into (22) and solving for Tco, we obtain
Tco = 122°F♦
Next, we calculate the two quantities needed to evaluate the true-mean driving force:
♦ Such a high temperature in the river could be harmful to fish and other wildlife. Engineers need to be mindful of
the impact of their designs on the environment as well as on the health and safety of people in the vicinity. Faced
with this result, the engineer should suggest altering the design (e.g. increase the flowrate of cooling water to obtain
a lower increase in temperature — most states allow no more than a 10ºF increase). For the purpose of this
academic example, we will ignore this objection and proceed through the analysis.

06-361 page 32 Spring, 2017
Thi − Tho 1 Tco − Tci

=
Z = = 1.67 =
and Y = 0.35
Tco − Tci 0.6 Thi − Tci
0.86
0.35
Since this is a 1-2 exchanger, we read FT from Fig. Fig. 4.9-4a:
FT = 0.86
Although the flow in the exchanger is complex, we calculate ∆Tlm as if the flow were counter-
current:
∆T1 = Thi-Tco = 78°F and ∆T2 = Tho-Tci = 50°F
∆T − ∆T2
∆Tlm = 1 =63°F
∆T1
ln
∆T2
Finally ∆Tm = FT ∆Tlm =54°F
Overall heat transfer coefficient Uo. We will obtain the overall heat transfer coefficient by first
estimating the two film coefficients and then using the resistances in series formula (17):
1 1 D x D 1
= o + w o +
U o hi Di kw Dlm ho
The shell-side coefficient (ho) can be estimated using Donohue’s correlation (20):

06-361 page 33 Spring, 2017
0.6 0.33 0.14

ho Do D G   c po µo   µo 
= 0.2  o e      (20)
ko  µo   ko   µ wo 
Since acid is corrosive, we would be well-advised to place it inside the stainless tubes, leaving
river water on the shell side (shell is mild steel). With this choice, the mass flowrate on the shell
side is
 gal   lb  4 lb
m  5000
=   8.34 =  4.17 × 10
 hr   gal  hr
The mass flux Ge appearing in Donohue’s correlation (20) is the geometric mean [defined by
(21)] of Gb and Gc defined by (22) and (23). To calculate these last two mass fluxes, we need to
divide the mass flowrate m by either of two different areas Sb and Sc whose formulas are
derived in the Appendix (starting on page 37).
The area of the baffle window is calculated from (30):
 πD 2   πD 2 
= Sb fb  s  − Nb  o  (28)
  4   4 
area of    
blue region
where Ds = diameter of entire shell

Nb = number of tubes in baffle window
fb = area fraction of baffle plate that is window
For a 25% baffle plate,♦ fb is given by (31) as
fb = 0.1955
Let’s assume that the fraction of tubes which pass through the baffle window equals the fraction
of the shell cross-section represented by the baffle window. In other words:
Nb = fb×158 = 0.1955×158 = 30.8
where 158 was given as the total number of tubes in the shell. Now everything in (28) is known:
 πD 2   πD 2 
Sb = fb  s  − Nb  o  =0.258 ft 2
 4   4 
   
♦ A 25% baffle plate has a window whose height is 25% of the diameter of the shell (as shown in the figure above).

06-361 page 34 Spring, 2017
The mass velocity through the baffle window is
m lb
G=
b = 1.62 × 105
Sb 2
ft -hr
For crossflow, the maximum available area for flow can be calculated from (29):
 D 
=Sc PDs 1 − o 
 p 
p = tube pitch
P = baffle pitch
using P = 6 inch, Ds = 20 inch, Do = 1 inch (see Appendix A.5-2), and p = 1.25 inch. The cross-
flow area is
 D 
Sc = PDs 1 − o = 0.167 ft 2
 p 
and the mass velocity for cross-flow is
m lb
G=
c = 2.5 × 105
Sc ft 2 -hr
and the effective mass velocity is
lb
=
Ge =
Gb Gc 2.01 × 105
ft 2 -hr
Before we can estimate ho from (20), we need some fluid properties. Since we have no idea
what the wall temperature is, let’s first get a rough estimate of ho by setting µ/µw = 1. The other
properties (besides µw) are evaluated at the average temperature of the shell-side fluid [T =
0.5(80°F + 122°F) = 101°F (38.3°C)]. Using Appendix A.2-11:

06-361 page 35 Spring, 2017
µ = 0.458×10-3 lbm/ft-s, k = 0.363 BTU/ft-hr-°F, ρ = 62.1 lb/ft3, cp = 0.999 BTU/lb-°F
(20) yields ho = 363 BTU/ft2-hr-°F
Since the thermal properties of dilute acid are very nearly those of pure water, we will estimate
the inside film coefficient using (19):
v 0.8
hL =1429 (1 + 0.0146T °C )
D 0.2
The tubes are one-inch 14 BWG. Properties can be found in table A.5-2:

06-361 page 36 Spring, 2017
Do = 1 in Di = 0.834 in xw = 0.083 in Ac = 0.00378 ft2
Since there are two tube passes, the entering acid is distributed among half of the 158 tubes. The
average velocity in any one tube is then
 gal   ft 3 
 3000   
 hr   7.48 gal  ft m
=v = 0.372 = 0.113
158
2
(
0.00378 ft 2 )
s s
(19) can then be evaluated at the average acid temperature: (130+200)/2 = 165°F = 74ºC.
hi = 1125 W/m2-°C
The overall heat transfer coefficient can be evaluated from the sum-of-resistances formula (17).
We just a thermal conductivity. The tubes are constructed of stainless steel. Table A.3-16 gives
thermal conductivities of a number of metals. For 304 stainless at 100°C:
kw = 16.3 W/(m-°K)
from (17): Uo = 104 BTU/ft2-hr-°F = 0.591 kW/m2-ºC
Total Heat Transfer Area. The heat duty which can be supplied by the available heat exchanger
is given by (24). The outside area of all 158 tubes (L = 16 ft) is

06-361 page 37 Spring, 2017
A0 =158πDo L =662 ft 2 = 61.5 m2
from (24): qT = Uo Ao ∆Tm = 3.7×106 btu/hr = 1093 kW
The required heat duty was evaluated at the beginning as 510 kW. Since the available heat duty
exceeds what’s needed, the available heat exchanger is capable of performing the job. Had the
two heat duties been more nearly equal, we would have had to go back and include the correction
µ/µw in both hi and ho. This is a trial-and-error.
Appendix: Derivation for Sb and Sc
In case you are curious as to how these formulas we obtained, I have put their derivation in
this appendix.
widest
row of
tubes
P
p
Ds Do
side view area for end view

cross-flow
(shaded yellow)
Sc is the area open to fluid flow across the widest row of tubes in the shell and between two
adjacent baffle plates. First the total area (ignoring tubes) is
total area = PDs
where P is the baffle pitch. To calculate the fraction of PDs which is open. let's subdivide Ds
into repeating unit cells:
length of unit cell = p (tube pitch)
length which is open = p-Do
fraction open = (p-Do)/p = 1 - Do/p
Multiplying this fraction by the total area gives the area open:

06-361 page 38 Spring, 2017
 D 
=Sc PDs 1 − o  (29)
 p 
Of course, we have ignored some end effects which could be significant if the number of tubes is
not large.
Likewise the open area of a baffle window is the total area of the window minus the area
blocked off by tubes:
 πD 2   πD 2 
=Sb fb  s  − Nb  o  (30)
 4   4 
   
where fb = 0.1955 (for a 25% baffle)
and Nb = no. of tubes thru window
To get this value of fb, we have to do a little geometry. First, recall that the area of a circle of
radius r is πr2.
Acircle = πr 2
A “sector” is a portion of a circle (see figure at right): sector

(heavy outline)
segment
Asector 1 r 2
=
2
(
2θ ) (dark shading)
total angle y
subtended
by sector x
θ r
where 2θ is the total angle subtended by the arc of the

sector. Note that if we substitute 2π for the 2θ, we get the
Asector = Acircle. The area of the triangle whose sides
are x,y, and r equals
= 1=
Atriangle bh 1 xy
2 2
Finally, the area of the segment is calculated from
Asegment = Asector - 2Atriangle
For a 25% baffle, the height of the segment, r-x, will be 25% of the diameter -- or 50% of the
radius:

06-361 page 39 Spring, 2017
x π 3
= 0.5 ⇒ θ= cos −1 ( 0.5 )= 60°= ⇒ y= r 2 − x 2= r
r 3 2
1  r  3  3 2
=
Atriangle   = r = r 0.217 r 2
2  2   2  8
π
Asec tor =θr 2 = r 2
3
π 2  3 2 π 3 2
Asegment = r − 2  r  =−
 0.61418r 2
 r =
3  8  3 4 
Now fb is the fraction of the total area of the circle which is represented by this segment:
π 3 2
 − r
Asegment  3 4  0.61418
=fb = = = 0.1955 (31)
Acircle πr 2 3.1416
Chapter 5. Evaporation
evaporation -- concentrating a solution of a nonvolatile solute by boiling away the solvent
(usually water)
Examples: vapor
• production of orange juice concentrate

V
• production of concentrated H2SO4 feed heat
• production of distilled water L
Usually the product is the concentrated solution (called

the “liquor”) and the vapor is discarded.
liquor
liquor - the concentrated solution
An important exception is the production of distilled

water from tap water. Here the condensed vapor is the product and the more concentrated
solution of minerals or (in the case of desalination) salts is discarded.
Evaporation is not the only unit operation that involves boiling. Some others that involve boiling
include:

06-361 page 40 Spring, 2017
• distillation - involves two or more volatile components (evaporation has just one)
• drying - product is a solid (in evaporation, product is usually a more concentrated liquid
solution)
• crystallization - product is a slurry of crystals precipitated from solution (not a simple
solution)
Usually these differences are sufficient to require different equipment to perform the operation.
EQUIPMENT FOR EVAPORATION

There are several types of evaporators, but they all have the following parts:
Part #1: HX (for adding the latent heat of vaporization to the liquid feed). Usually HX is a
shell-and-tube with condensing steam as the hot fluid on the shell side (because steam
is free of minerals which form scale).
Part #2: vapor space or vapor head (larger chamber in which liquid, entrained in the vapor as
droplets or foam, can be separated from the vapor, usually by impinging the two-phase
mixture onto a plate.)
You also need something to move the liquid through the heat exchanger. Sometimes a pump is
used and sometimes it's just left to gravity, either in the form of a falling film on the outside of
the HX tubes or the entrainment of liquid by rising bubbles.
Part #3: liquid mover
• pump
• gravity - falling film
• gravity - entrainment by rising bubbles
Long-Tube Vertical Evaporator
When the heat exchanger tubes are oriented

vertically, the bubbles produced by boiling rise
and carry liquid up with them. For this reason,
this is also called a “climbing film” evaporator.
Tubes are typically 1 or 2 inches in diameter and
12 to 32 feet in length. The liquid entrained in the
vapor coming out of the HX is richer in solute
than that entering the HX. Vapor and fine liquid
droplets are separated in the vapor head by
placing a baffle plate in the path of the two-phase
fluid coming out of the HXer.
Notice the vent at the top of the shell-side of

the HX. Usually a very small fraction of the

06-361 page 41 Spring, 2017
steam flow is allowed to bleed out of the shell. This is done to allow air (which is often present
with the steam) to escape. Otherwise non-condensable gases like air will accumulate in the shell,
causing the pressure to rise and decreasing the steam flowrate.
Long-tube vertical evaporators are especially effective in concentrating liquids that tend to
foam. Foam is broken when the two-phase mixture hits the baffle plate.
Forced-Circulation Evaporator
Many concentrated solutions (e.g.

orange juice) are highly viscous. For
example, a concentrated sugar solution
can have a viscosity that is 106 times
larger than water. If the concentrated
solution is highly viscous, the velocity of
rising bubbles will be very small and the
heat transfer coefficient will suffer.
For very viscous liquids, a pump is

often used to improve the circulation. Of
course a pump also raises the pressure
significantly which can prevent the liquid
from boiling. In a forced-circulation
evaporator, boiling usually occurs near the
end of the HX where the pressure is lower
or in the pipe leading from the HX to the
vapor space.
HX tubes are now oriented horizon-

tally so they are easier to clean. Again a
baffle plate is used to separate vapor from
entrained liquid.
EVAPORATOR PERFORMANCE
Regardless of which type of evaporator we use, there are two standard measures of performance:
capacity – pounds water evaporated per hour
economy – lb water evaporated per lb steam used (usually somewhat less than unity)

06-361 page 42 Spring, 2017
EXAMPLE: SIMPLE EVAPORATOR PROBLEM

Suppose we are trying to
concentrate a solution of orange
juice from 10 wt% solids to 50 wt%
in a long-tube vertical evaporator.
Our past experience has shown that a
reasonable heat transfer coefficient.
is
Given: U = 500 Btu/h-ft2-°F
To keep the boiling T low enough,

we keep the vapor space under vacuum:
p = 4 inHg
If the feed is 55,000 lb/hr at 70°F, what area of heat exchanger is required and what flowrate of
15 psig steam?
Find: m s , A
Solution: First we do a mass balance to determine the flowrates. A mass balance on the solids
yields:
10%(55,000 lb/hr) = 50%( m )
so m = 11,000 lb/hr (liquor)
is the flowrate of the liquor. The corresponding flowrate of the vapor must be (from an total
mass balance):
m f – m = 44,000 lb/hr (vapor)
An enthalpy balance around the entire evaporator requires:
rate in = rate out
feed + steam = vapor + liquor + condensate
or q = steam – condensate = vapor + liquor - feed
The heat duty of the HXer is just the heat lost by the condensing stream, or the heat gained by
the feed. In terms of enthalpy per mass of the streams, the heat gained by the feed is:
q = m s (Hs – Hc) = ( m f – m )Hv + m H – m f Hf (32)

06-361 page 43 Spring, 2017
For high-molecular-weight substances like orange juice and other “organic colloids”♥ we can
usually neglect boiling-point elevation and heat of dilution. Then the enthalpies of the feed and
the liquor are essentially the same as pure water under those conditions. The enthalpy of liquid
water and water vapor can be read from the Steam Tables found in Appendices A.2-9 (saturated
steam) and A.2-10 (superheated steam).♦
at 70 °F:♠ Hf = 38.09 btu/lbm
But what is the temperature of the liquor? Well, since the liquor is in equilibrium with the vapor,
their temperature and pressure should be the same. Since the vapor is just saturated steam at 4
inHg, we look up its properties in the steam table:
pv = 4 inHg = 1.97 psia
from App. A.2-9: T = 125.1 °F
♥ “Organic colloids” are solutions of high molecular-weight polymers or dispersions of fine particles. Then the
molar concentration of these solutions is negligibly small and their properties are close to that of pure water: see
example on page 48. An example of an organic colloid is orange juice.
♦ For consistency with later examples in which heat of dilution effects are important, we read all the enthalpies
from the steam tables, which have a reference state (where H is defined to be zero btu/lb) corresponding to saturated
liquid water at its freezing point. In Example 8.4-1, Geankoplis works a very similar problem in which all
enthalpies are calculated using the conditions of the liquor as the reference state. Then (by definition) H in (32) is
zero; the enthalpy of the feed Hf differs from that of the liquor just in sensible heat, which can be estimated as (8.4-
9); and the enthalpy of the vapor Hv differs from that of the liquor just in latent heat (i.e. Hv is just the heat of
vaporization at the T,p of the vapor space). See the discussion on “steam economy” on page 45. Try re-working
Example 8.4-1 evaluating the enthalpies appearing in (8.4-7) using the steam tables (as above): you should get very
nearly the same answer.
♠ Strictly speaking, the enthalpies tabulated in A.2-9 are for saturated liquid, meaning that the pressure on the liquid
equals the vapor pressure. Here the pressure on the feed is probably 1 atm (or 14.7 psia) rather than 0.3622 psia (the
vapor pressure). Pressure has a very weak effect on the enthalpy of liquid water and the effect of pressure is usually
neglected. From thermodynamics (see 4.2-24 on p169 of Sandler):
 ∂H   ∂V 
 ∂P  = V − T  ∂T 
 T  P
which identically vanishes for an ideal gas ( PV = RT ). For most liquids like water, the second term is negligible
compared to the first. Then the effect of a change on pressure equal to 1atm on the enthapy of liquid water can be
estimated as
 cm3  btu
∆p V ≈ (1 atm ) 1
∆H T = =0.044
 g  lb m
which is equivalent to a change in temperature of only 0.044 °F. As long as we are dealing with much larger
sensible temperature changes, we can neglect the effect of pressure on the enthalpy of liquid water.

06-361 page 44 Spring, 2017
Interpolating the tables,♣ we obtain
at 125.1 °F: Hv = 1115.7 btu/lbm and H = 93.1 btu/lbm
Substituting back into (32):
 lb   btu   lb   btu   lb   btu 

q=  44, 000  1115.7  + 11, 000   93.1  −  55, 000   38.09 
 hr   lb   hr   lb   hr   lb 
btu
= 4.91 × 107 + 0.10 × 107 − 0.21 × 107 = 4.80 × 107
hr
Now, using the design equation, we can calculate the required area:
q = UA∆Tm (33)
In a typical evaporator problem, steam is condensing on the shell side while water is evaporating
on the tube side. When a phase change occurs with a single component (water in this case), the
temperature stays at the boiling point. Thus the driving force ∆T is practically constant along the
length of the heat exchanger.
However, some sensible-heat changes might still Steam enters

occur at either end of the heat exchanger. For superheated
example, steam might enter superheated, meaning that Ts
Temperature
it’s temperature is above the boiling point. Until the Steam leaves
temperature drops to the boiling point, condensation subcooled
does not occur. Similarly, so much heat might be T
removed from the steam, that it completely condenses Feed solution
and is then subcooled below the boiling point. Finally, enters subcooled
the feed solution has to be heated up to its boiling point 0 L
before evaporation occurs.♥ The temperature profile Position in Hxer
at right illustrates the most general situation.
For a preliminary design (which is all we will do in this course), it is customary to

approximate the “true mean” MDT as:♠
♣ In working the homework assignments, you will have frequent need to perform this interpolation. A Mathcad
document has been prepared to assist you: Steamtab.mcdx. It can be found at our Blackboard website: see the
“Student Resources” folder under “Handouts”.
♥ The heat required to preheat the feed from 70°F (the feed conditions) to 125°F (the boiling point at the pressure in
the vapor space) is
 lb   btu btu  btu
 55000   93.1 − 38.09 =0.30 × 107
 hr   lb lb  hr
This is only 6.3% of the heat duty q. We are willing to tolerate this size of error in our preliminary design.
♠ For the alternative to making this assumption, see the example beginning on page 14.

06-361 page 45 Spring, 2017
∆Tm ≈ Ts - T
where Ts = T of saturated steam at p in steam chest
T = Tbp of concentrated liquor at p in vapor space
“Proof”: Assuming that the sensible heat associated 1) with superheating the steam, 2) with sub-cooling the
condensate and 3) with pre-heating the feed up to its boiling point are all negligible compared with the latent heat of
vaporization of either stream, then the fraction of the heat exchanger area experiencing a driving force different from
Ts – T will be negligible; thus the entire heat exchanger can be approximated as having Ts – T as the driving force: in
particular,
∆T − ∆T2
∆T1 = ∆T2 = Ts – T and ∆Tlm =1 = Ts – T
∆T
ln 1
∆T2
When ∆T1 and ∆T2 have identical values, it seems reasonable (almost obvious) that their mean should have the same
value; however the expression above for log-mean is mathematically indeterminate: ∆Tlm = 0/0. This indeterminacy
can be resolved if we consider the asymptotic limit of this expression as the two quantities being averaged approach
each other. Define
∆T1 = ∆T2 (1 + ε )
then ∆T1 − ∆T2 = (1 + ε ) ∆T2 − ∆T2 = ε∆T2
∆T1 1 1 1 ε→0 1
and ln =ln (1 + ε ) =ε − ε 2 + ε3 − ε 4 +  → ε − ε 2
∆T2 2 3 4 2
where second equation represents a Taylor series expansion. Finally, we take the limit as ε→0:
∆T1 − ∆T2 ε∆T2  1  ∆T1 + ∆T2

lim ≈ = 1 + ε  ∆T2 =
∆
ε − ε2 
ε→0 T 1 2  2
ln 1
∆T2 2
When the two numbers being averaged are nearly the same, the log-mean and the simple mean yield the same result.
Back to our simple evaporator problem: Knowing the pressure of the steam, we can find its
temperature in the steam tables:
at 15 psig = 29.7 psia: Ts = 250°F
The heat of vaporization λs is the difference between the enthalpy of saturated vapor Hs and the
enthalpy of the condensate Hc leaving (assumed here to be saturated liquid):
λs = Hs – Hc = 1164.2 btu/lb - 218.6 btu/lb = 945.6 btu/lb
The driving force for heat transfer is the difference in temperature of saturated steam and
saturated vapor:

06-361 page 46 Spring, 2017
∆Tm ≈ Ts-T = 250-125 = 125°F
Now everything in (33) is known except for A:
Btu  Btu 
4.77 × 107
=  500 2  ( A )(125°F )
h  h − ft − °F 
or A = 768 ft2
The amount of steam required is just the heat duty divided by the latent heat of the steam:
Btu
4.77 × 107
q h= 50,800 lb
s =
m=
λs Btu hr
945.6
lb
Recall that our two main performance indicators were “capacity” and “economy”. Capacity is
just the rate of evaporation, which will be the flowrate of the vapor. For this example,
lb
capacity = m f − m
=44,000
hr
The economy is the mass of vapor produced per mass of steam consumed:
lb
m f − m 44, 000 hr
= =
economy = 0.87
m s lb
50,800
hr
Steam Economy
Why is the economy less than one? The answer becomes clear, if (instead of looking up the
enthalpies of water and its vapor in a table) we estimate them as sensible and latent heat. First,
let’s select the liquor as the reference state for the calculations of the enthalpies. Then
choose liquor as reference: H=0
The vapor is at the same temperature as the liquor so there are no sensible heat changes, but the
vapor has a much higher enthalpy owing to its heat of vaporization:
Hv = λ
Since the feed is also a liquid (like the liquor), its enthalpy differs from the liquor only in
sensible heat:

06-361 page 47 Spring, 2017
Hf = cp(Tf - T)
Substituting these enthalpies in (35):
( m f ) ( )
− m λ − m f c p T f − T = m s λ s
or
m f − m λ s m f c p T − T f
= −
( ) (34)
m s λ m s
 λ
<1
Generally the heat of vaporization of water decreases with increases in temperature. This is a
consequence of the specific heat of vapor being less than that of liquid water. To show this,
recall that the heat of vaporization λ is the difference between the enthalpies of saturated vapor
and liquid:
=
λ HV − H L
Differentiating with respect to temperature, holding pressure fixed:
 ∂λ   ∂HV   ∂H L  ( 0.42 − 1) BTU = BTU

  =   −  =
c pV − c pL = −0.58
 ∂T  P  ∂T  P  ∂T  P lb m -°F lb m -°F
The enthalpy derivatives are heat capacities, which have been substituted in the second equation
above. Now the heat capacity of steam is about half of that for liquid water, which leads to the
numerical values in the third and fourth equations above. In the above calculation, we have held
pressure constant whereas the heat of vaporization is usually evaluated at the vapor pressure of
the liquid which increases significantly with temperature. However the effect of pressure on
enthalpy is generally negligible (see footnote ♠ on page 43). Thus we have shown that
Alternatively, We can use the steam table A.2-9 on p964 of Geankoplis to directly evaluate
the changes in λ with temperature.
at 150°F: λ = 1126.1 – 117.96 = 1008.1 btu/lbm
at 250°F: λs = 1164.2 – 218.59 = 945.6 btu/lbm
The drop in latent heat in this example is about 62.5 BTU/lbm which is close to 58 BTU/lbm
obtained by multiplying cpV-cpL by 100ºF (or 58 BTU/lbm).
Since the steam must be hotter than the water being boiled, this means that λs < λ. So the
first term in (34) is already less than one. The second term decreases it further; this second term
represents the energy required to preheat the feed up to the temperature of the liquor where
boiling occurs. So a pound of steam boils less than one pound of water because some of the
energy of the steam is required to preheat the feed. However, even if the feed were at the same

06-361 page 48 Spring, 2017
temperature as the liquor, a pound of steam would boil less than one pound of water because the
latent heat of the hot steam is less than the latent heat of the water being boiled.
OVERVIEW OF EVAPORATOR DESIGN

The design equation for evaporators can be summarized as
heat duty: (

m f − m ) H v − m f H f + mH
  
 =
 m s ( H s − H c )
q=
 
(35)
liquor
vapor feed steam
HXer area: q = U A ∆Tm
For preliminary designs, we can approximate
∆Tm ≈ const. = Ts - T
where T is the boiling point of the liquor at the pressure in the vapor space. This is approximate
when there is significant BPE because T will vary along the length of the heat exchanger tube as
the concentration of solute also changes.
• Evaluation of enthalpies with negligible BPE and negligible heat of dilution (example on
page 42 and also Example 8.4-1). Assuming the solvent is water, we can read the enthalpies
of all three streams from the steam tables.
read T, Ts, H, Hf and Hv from steam tables
• Evaluation of enthalpies with complications (Example 8.4-3):
read H, Hf from special table or chart

read T from Dühring chart; Ts, Hv from steam tables♣
♣ Since enthalpies are being taken from more than one source, we should take care that the reference state is the
same for all tables and charts used in a given problem. The steam tables in Geankoplis use saturated liquid water at
32°F and 1 atm as the reference state. The enthalpy chart for NaOH solutions in Geankoplis also use 0 wt% NaOH
at 32°F and 1 atm as the reference state.

06-361 page 49 Spring, 2017
COMPLICATIONS
Boiling-Point Elevation
If the solution is more concentrated (in a mole-fraction sense), complications arise because
the enthalpy and partial pressure of a liquid more generally depend on the solute concentration as
well as the temperature. Whenever a solute is added to water, the partial pressure of the water is
reduced. Recall Raoult's law from thermodynamics.
Roault’s law: pi(xi,T ) = xi pio(T )
where pi = partial pressure of component i

xi = mole fraction component i in liquid
pio = vapor pressure of pure i at some T
Note that adding any solute with water causes xw<1 so that
pw < pwo
In general, this implies that I must heat the solution up to a higher temperature (i.e. higher than
the b.p. of pure solvent) before it will boil. This increase in boiling point due to addition of
solute is called
boiling-point elevation (BPE)♣ - increase in b.p. due to addition of solute
In effect, the boiling point of a solution depends, not only on the total pressure, but also on the
concentration of solute:
Tbp = Tbp(p, c)
This effect can be quite significant for concentrated solutions of inorganic salts. For
example, for at a pressure of 1 atm, the boiling point of pure water is
1 atm, 0% NaOH in water: Tbp = 212°F
If the pressure is other than 1 atm, the temperature can be read from the steam tables. Given this
boiling point for pure water, we can look up the boiling point of solutions of NaOH from a
Dühring chart (see Fig. 8.4-2 of Geankoplis):
♣ This is called “boiling-point rise” in Geankoplis.

06-361 page 50 Spring, 2017
Notice that the lines (which correspond to a particular mass fraction) are nearly parallel. If they
were perfectly straight and parallel, then BPE would be independent of pressure. From the
Dühring chart, we can read the boiling point of a concentrated NaOH solution:
50wt% NaOH in water: Tbp = 290°F at 1atm
Note that the boiling point of the solution is significantly greater than the boiling point of pure
solvent. BPE is a colligative property, which means
colligative property -- depends on molar (not mass) concentration
Example: Use Raoult’s law to estimate the partial pressure of water at 100°C of a) a 50wt%
NaOH solution and b) a 50wt% suspension of “organic colloids” (use MWaverage = 1000 g/mol).
Estimate the BPE for each.

06-361 page 51 Spring, 2017
Solution: To calculate the mole fraction of water in the solution, take 1 gram of solution as a
basis. This 1 gram mixture consists of 0.5 g of water (MW = 18) and 0.5 g of NaOH (MW =
40). We calculate the corresponding number of moles of each and the mole fraction:
NaOH (MW=40) Organic Colloids
(MW=1000)
0.5g 0.5g
g g
18 18
mol = 0.690 mol = 0.982
xw
0.5g 0.5g 0.5g 0.5g
+ +
g g g g
18 40 18 1000
mol mol mol mol
At 100°C (the b.p. of pure water), the vapor pressure of pure water is 1 atm (= pwo). Using
Raoult’s law, the partial pressure is estimated as
pw 0.690(1 atm) = 0.690 atm 0.982(1 atm) = 0.982 atm
To cause the solution to boil, the partial pressure must be raised to 1 atm. For the same mole
fraction of NaOH, this means raising the vapor pressure to
1 atm 1 atm
pwo = 1.450 atm = 1.018 atm
0.690 0.982
Using the steam tables (Appendix A.2-9), we find that the temperature has to be
T 110.0°C 100.8°C
Thus using Raoult’s law, the BPE is estimated to be
BPE from 110.0°C - 100°C = 10.0°C 100.8°C - 100°C = 0.8°C

Raoult’s law
Of course, this is only an estimate. Raoult’s law is only asymptotically correct as xw→1. The
actual BPE for NaOH is 39°C.
BPE 283°F - 212°F = 71°F = —

determined 39°C
from Fig.
8.4-2
Although BPE is significant for a low-molecular-weight solute like NaOH, the same 50wt%
concentration of an "organic colloid" (e.g. orange juice) may have a negligible BPE because the
molecular weight is much higher. Thus the same wt% gives a much smaller molar concentration
of a high-molecular-weight solute and therefore a much smaller BPE.

06-361 page 52 Spring, 2017
Heat of Dilution
If you dilute concentrated sulfuric acid by adding water, large quantities of heat can be
released. If the mixing is adiabatic and both fluids are initially at room temperature, the final
temperature after mixing can by high enough to causing boiling. You were probably warned to
use great caution when diluting sulfuric acid in chemistry lab. The problem is that the enthalpy
of a solution depends not only on its temperature, but also on its concentration:
H = H(c,T)
which means you then need more than just a heat capacity to calculate sensible heat changes.
Figure 6 gives the enthalpy-concentration diagram for NaOH solutions.
Figure 6: Enthalpy of NaOH solutions. Taken from Fig. 16.8 in MS&H.
Example: Suppose we dilute 1-lb of 50wt% NaOH solution at 70°F with 1-lb of pure water, also
at 70°F. What is the final temperature if the mixing is adiabatic?
Solution: read the enthalpies of the two solutions from 8.4-3:
H(50%, 70°F) = 123 BTU/lb

H(0%, 70°F) = 40 BTU/lb

06-361 page 53 Spring, 2017
Multiplying each by the corresponding weight and adding, we obtain the final enthalpy of the
mixture (“adiabatic” just means that no energy is lost during mixing):
(123 BTU/lb)(1 lb) + (40 BTU/lb)(1 lb) = 163 BTU
Dividing by the mass of the final mixture, we obtain the enthalpy per pound:
163 BTU BTU

= 81.5 = H=( 25%, T ? )
2 lb lb
This represents the specific enthalpy of the final mixture, which we also know is 25wt%. Again
using Figure 6 (Fig. 8.4-3 in Geankoplis), to get this H, we need to choose the temperature as
T = 123°F
Thus the final solution is warmer than the two initial solutions, due to the “heat of dilution.”
Question: when diluting sulfuric acid (which also exhibits a significant heat of dilution), why is
is wiser to add (concentrated) acid to water, rather than water to (concentrated) acid?
LEVER-ARM RULE FOR HEAT OF DILUTION

In the previous example, the two solutions being mixed and
the mixture correspond to three points in Figure 6: (x1,H1),
(x2,H2) and (xm,Hm). In that example, equal amounts M1 = M2
= 1 lb of the concentrated and dilute solutions are being mixed
and the point representing the mixture (xm,Hm) turns out to be
the midpoint of a straight line connecting (x1,H1) and (x2,H2).
This is not a coincidence. If the amounts of the two

solutions being mixed were not equal, conservation of mass and
energy would require that the mixing point (xm,Hm) still lies Figure 7
somewhere along a straight line connecting (x1,H1) and (x2,H2).
Label the three points “1”, “2” and “m”. Mass and energy balances further require that point
“m” be located such that
M1
2m = (36)
m1 M2
where 2m and m1 are the lengths of the two lines drawn in the schematic of Figure 7: 2m is the
length of the line connecting point “2” with point “m” while m1 is the length of the line
connecting point “m” with point “1”. Note that decreasing the amount of M1 relative to M2
shifts point ‘m’ closer to ‘2’ (i.e. it decreases 2m ). In the limit that M1 = 0, point ‘m’ coincides

06-361 page 54 Spring, 2017
with ‘2’, which makes sense. Armed with (36), we can solve such problems graphically without
doing any balances.
Proof of (36): conservation of mass and energy require
x1M1 + x2M2 = xm(M1 + M2) and H1M1 + H2M2 = Hm(M1 + M2)
By collecting together terms multiplied by either M1 or M2, these two relationships can be
rearranged to show
H 2 − H m x2 − xm M1 2m
= = = (37)
H m − H1 xm − x1 M 2 m1
This can be rearranged to further show
H 2 − H m H m − H1 rise
= = = slope
x2 − xm xm − x1 run
where the ratio represents the slope of either of the two line segments in Figure 7. That the two
line segments have equal slopes and share a common point means that they are part of the same
straight line. This serves a proof that the mixing point “m” lies on a straight line connecting “1”
and “2”.
To find the exact location of the mixing point on this straight line, consider the two shaded
right triangles shown in Figure 7. Since the slope of the two hypotenuses is the same, and the
other sides are either horizontal or vertical, the corresponding angles of the two triangles must be
the same for all three angles defining either triangle. This means the two triangles are similar in
the geometric sense. For two similar triangles, the ratio of lengths of corresponding sides are the
same for all three sides. (37) gives the ratio of the heights (i.e. ∆H) as M1/M2; therefore the ratio
of the lengths of the two hypotenuses must also equal M1/M2, which completes the proof of (36).
(36) is called the lever-arm rule. As a useful

analogy, think of M1 and M2 as being masses being
balanced on a beam, resting on a fulcrum (see sketch at
right). The further the mass is from the pivot, the
smaller the mass can be. At (mechanical) equilibrium:
M1 l2
M1l1 = M 2 l2 or =
M 2 l1
Notice the inverse relationship between the mass and the distance from the pivot.
MULTIPLE EFFECTS
We will now look at two methods to improve steam economy of an evaporator:

06-361 page 55 Spring, 2017
1. multiple “effects”
2. vapor recompression
In the simple evaporators we have dealt with so far, one of
?
the product streams is water vapor — also known as steam.
Steam has a utility cost associated with it. Instead of
discarding this vapor product, why not use this vapor in place
of steam (which would save this utility cost)? The answer is
V
simple: if the vapor replaced the steam, there would be no
driving force for heat transfer because the “steam” and the L
liquid would have the same temperature: Ts = T and ∆Tm = 0.
To get a driving force in the right direction, we either have

to 1) raise the temperature of the vapor or 2) lower the
temperature of the liquor. In the next section, we will consider
raising the temperature of the vapor by mechanical compression. In this section, we will
consider lowering the temperature of the liquor by dividing the heat duty among several
evaporators operating in series. By operating downstream evaporators at lower pressure, we can
lower the boiling temperature and thereby use the vapor produced by the upstream evaporator.
This reduces the steam usage and improves economy.
You can have any number of evaporators connected together in this way. Each evaporator in this
sequence is called an effect, and the entire sequence is called a multiple-effect evaporator. In
the figure above, we show a double-effect evaporator. In order to maintain a driving force for
heat transfer in each effect, we must have
T2 < T1 < Ts
This can be accomplished by having♣ p2 < p1 < ps
♣ Assuming we have saturated liquid and vapor in each effect, the pressure in any effect is just the vapor pressure at
the temperature for that effect. Recall that vapor pressure increases with temperature.

06-361 page 56 Spring, 2017
This requires some device to lower the pressure in the last effect. While a mechanical vacuum
pump (a compressor applied in reverse) could be used, it is usually much cheaper to use a
barometric condenser (see description in section 8.6C of Geankoplis).
p2 < p1:♦ ∆T 2 = T1 - T2 > 0
q2 = U2A2∆T 2 (38)
Of course, we still have to supply steam to the first effect
∆T 1 = Ts - T1 > 0
q1 = U1A1∆T 1 (39)
Approximation Solution
For simple designs where BPE and ∆Hdil are negligible and where the sensible heat needed
to preheat the feed up to its boiling point is negligible, the same latent heat added to the vapor in
the first effect is recovered by condensing in the second effect:
neglecting sensible heat: q1 ≈ q2 (40)
Usually the pressure in the last effect and the pressure of the steam are known. Using the steam
tables, we can convert these pressures into temperatures. The unknowns are the temperatures
and pressure in the other effects.
Example: given Ts and T2, find T1. Solution: (38) and (39) into (40) yields
U1A1(Ts – T1) = U2A2(T1 – T2)
This represents one equation in one unknown (T1), assuming that the U’s and A’s are known.
If instead, we have N effects, the analog of (38), (39) and (40) yield N-1 equations in N-1
unknowns: equating the heat duties
q1 ≈ q2 ≈ ... ≈ qN
or U1A1(Ts – T1) = U2A2(T1 – T2) = … = UNAN(TN-1 – TN)
♦ The expression for ∆T which follows assumes that the vapor produced in the first effect condenses in the steam
chest of the second effect at the same temperature at which it boiled. If BPE is significant, the vapor will condense
at a lower temperature than it boiled (even if the pressure is the same). See discusion of BPE starting on page 57.

06-361 page 57 Spring, 2017
Now Ts and TN are known, but T1, T2 … TN-1 represent N-1 unknowns. These equations are
strongly coupled. A simpler approach is to try and solve for the ∆Ti’s instead of the Ti’s. We
will illustrate this by re-solving the last example. From (38) and (39), we have
q
∆T1 = 1 (41)
U1 A1
q
and ∆T2 = 2 (42)
U 2 A2
2 q1 q
Adding (41) and (42): ∑ ∆T=i + 2
U1 A1 U 2 A2
(43)
i =1
Dividing (41) by (43):
q1 1
q1 = q2
∆T1 U1 A1 U1 A1
= ⇒ (44)
∑ ∆T q1 U1 A1
q
+ 2
U 2 A2
1
U1 A1
+ 1
U 2 A2
Since Σ∆T can be easily calculated from the info given in the problem statement:
∑ ∆T = ∆T1 + ∆T2 = (Ts − T1 ) + (T1 − T2 ) = Ts − T2 (45)
we can easily calculate ∆T1 from (44):
1
U1 A1
=∆T1 (Ts − T2 )
1 + 1
U1 A1 U 2 A2
The analog of (40) - (45) for N effects are
q
(41): ∆Ti = i for i = 1,2, ... N (46)
U i Ai
qi
∆Ti U i Ai
=
∑ ∆T q1 U1 A1
q
+ 2
U 2 A2
q
+ + N
U N AN
Applying the analog of (40) (q1 = q2 = ... = qN ≡ q), we can cancel out the qi, leaving:

06-361 page 58 Spring, 2017
1
∆Ti U i Ai
= (47)
∑ ∆T 1U1 A1 + 1U 2 A2 +  + 1U N AN
When BPE is negligible, the sum of the driving forces can be calculated explicitly without
knowing the temperatures in the intermediate effects (i.e. without knowing Ti for i = 1, 2 … N–
1):
∑ ∆T = ∆T1 + ∆T2 +  + ∆TN

= (Ts − T1 ) + ( T1 − T2 ) +  + (TN − 2 − TN −1 ) + ( TN −1 − TN ) (48)
= Ts − TN
Notice that the all of the unknown temperatures cancel out. Substituting (48) into (47):
1
U i Ai
∆Ti = (Ts − TN ) for i = 1,2, ... N
1 + 1 + + 1
U1 A1 U 2 A2 U N AN
In the simple case in which all the A’s are equal and all the U ’s are equal, this last result reduces
to
1
∆=
Ti (Ts − TN ) for i = 1,2, ... N (49)
N
In short, the total driving force is divided equally among each of the N effects.
Performance
Each effect results in the evaporation of water at a rate equal to q/λi. Summing the rate of
evaporation in each effect, we obtain the total rate of evaporation:
N q
mV = ∑ i

λ
i =1 i
Steam is used only in the first effect. The rate of steam consumption in the first effect is
q
m s = 1
λs
The steam economy can then be calculated as

06-361 page 59 Spring, 2017
N q
∑ λi
m V i =1 i q1= q2= = q N N
1
= ⇒ λs ∑ ≈N
m s q1
i =1
λ i
λs
If we neglect changes in the latent heat with temperature and pressure, then all the λ’s are equal
and the sum reduces to N. If changes in the latent heat with temperature are considered, the sum
will be less than N.
Summary: For the same capacity, steam requirements are significantly less for multiple-effect
evaporators, than for single-effect evaporators: generally, the economy increases with the
number of effects. The price that is paid for this increased steam economy is greater equipment
costs: generally the cost of the equipment is proportional to the number of effects.
For a single effect: economy < 1
For N effects: economy < N
The reason that the economy increased is that we need to supply steam only to the first effect.
The source of heat for 2nd, 3rd, ... Nth effects is supplied by condensing the vapor produced in the
previous effect.
EXAMPLE: SIMPLE MULTIPLE-EFFECT EVAPORATOR PROBLEM

Let’s re-solve the orange-juice concentrate problem on page 42 this time using a 3-effect
evaporator. The information given in that earlier example has been added to the schematic
below for a 3-effect evaporator.
For simplicity, let’s assume the heat-exchanger area and heat-transfer coefficient are equal for
each effect:

06-361 page 60 Spring, 2017
BTU
U=
1 U=
2 U=
3 500 2
ft -°F
and A=
1 A=
2 A=
3 ?
The driving force is each effect is estimated from (49):
T − T3 250 − 125
∆T1 =∆T2 =∆T3 =s = =41.7 °F
   3 3
Ts − T1 T1 − T2 T2 − T3
Thus we can calculate the temperature in the first two effects:
T1 = 250 - 41.7 = 208.3 °F and λ1 = 973 btu/lbm
T2 = 208.3 - 41.7 = 166.6 °F and λ2 = 998 btu/lbm
T3 = 125 °F and λ3 = 1023 btu/lbm
Ts = 250 °F and λs = 946 btu/lbm
Although the heat duty in each effect is the same, the amount evaporated is not quite equal
because the heat of vaporization depends on temperature. The total of evaporation rates in each
effect is
q q q lbm
+ + =
44, 000
λ1 λ 2 λ3 hr
Solving for the q heat duty in each effect:
btu
q 3.66 × 106
=
hr
Using the same U as in the previous example, we can estimate the hxer area:
q
=A = 704 ft 2
U ∆Ti
Of course, we have three heat exchangers, each of this area. The steam requirements can be
estimated as
q lbm
=
S = 1.547 × 104
λs hr

06-361 page 61 Spring, 2017
F − LN
and the steam economy is = 2.84
S
Comment: The economy is slightly less than the number of effects, but significantly greater than
unity. While the improved steam economy is the main advantage of using multiple effects, the
tradeoff is greater heat exchanger costs: we need 3×704 = 2112 ft2 compared to only 768 ft2 for
a single effect evaporator. Of course, the heat-exchanger cost is a capitol expense, which you
pay only once, whereas steam is an operating cost which is ongoing. As the cost of energy rises,
the use of multiple effects becomes more attractive.
Effect of Boiling Point Elevation
Consider what happens to the temperature of the vapor leaving the first effect as it condenses
in the hxer of the second effect. When it leaves the first effect, the vapor is at the boiling point
of the solution in the first effect
boiling point of solution in effect 1 = T1
But this same vapor condenses at the boiling of pure water (not the boiling point of the solution
from whence it came). The difference between the two temperatures is the BPE for the first
effect BPE1.
condensation temperature of vapor from first effect = T1 – BPE1
This means that the driving force for heat transfer in the second effect is not T1 – T2. Instead:
∆T2 = T1 – BPE1 – T2
where boiling point of solution in effect 2 = T2
The same correction must be made for each of the subsequent effects. Thus the effect of boiling
point elevation is to reduce the driving force for heat transfer in the downstream effects
∑ ∆T = ∆T1 + ∆T2 +  + ∆TN

= (Ts − T1 ) + (T1 − BPE1 − T2 ) +  + (TN −1 − BPEN −1 − TN ) (50)
N −1
=Ts − TN − ∑ BPEi
i =1
To the extent that sensible heat is still negligible compared to latent heat, we can still claim the
heat duties of each effect is the same: q1 = q2 = ... = qN. (46) and (47) still hold but (48) must be
replaced by (50). The final result is

06-361 page 62 Spring, 2017
N −1 1
  U i Ai
∆Ti =  Ts − TN − ∑ BPEi  (51)
  1
 U1 A1 + U 2 A2 +  + U N AN
1 1
 i =1
Boiling point elevation thus lowers the driving force for each effect and lowers the amount of
evaporation which occurs in each effect. Unfortunately, BPEi cannot be evaluated until Ti is
known. So the problem is now trial-and-error.
Rigorous Solution (incl. BPE and sensible-heat effects)
Equality of the heat duties (i.e. q1 = q2 = ... = qN) represents a first approximation. Sensible
heat is generally not entirely negligible and this causes the heat duties to vary from one effect to
another. If the heat duties are not equal, they will not cancel out as in (47). This leads to the
following generalization of (51):
qi
 N −1  U i Ai
∆Ti =  Ts − TN − ∑ BPEi  (52)
  q1 q q
 i =1  U A + 2 U A + + N U A
1 1 2 2 N N
Unfortunately, the individual qi’s cannot be calculated until the temperatures are known, so the
problem requires a trial-and-error solution, which is described in the following recipe.
We will illustrate the solution using an example: Ex. 8.5-1 in Geankoplis.
Example: A triple-effect evaporator is being used to concentrate 50,000 lb/hr of a 10wt% sugar
solution at 80ºF to 50wt% sugar. BPE (in ºF) and heat capacity cp (in btu/lb) can estimated
from♥
BPE = 3.2x + 11.2 x2 and cp = 1.0 – 0.56x
where x = wt fraction of sugar. BPE and cp can be assumed to be independent of pressure.

Saturated steam at 29.8 psia is used in the first effect.
ps = 29.8 psia
The pressure in the third effect is controlled at 1.94 psia.
p3 = 1.94 psia
The overall heat transfer coefficients are
♥ Unlike orange juice, sugar is generally a low-molecular weight compound. For example, sucrose has a molecular
weight of 342 g/mol.

06-361 page 63 Spring, 2017
U1 = 550, U2 = 350 and U3 = 200 btu/ft2-hr-ºF
The areas of the three effects are the same
A1 = A2 = A3 = ?
Calculate the area of each effect and the steam requirements. The flow diagram containing some
of the input data follows:
Given: U1, U2 … UN, F, xF, TF, p
Find: S and A1 = A2 = … = AN
General recipe for solution with N effects:
Step 1) Knowing the pressures pN and ps and the composition xN in the last effect, use a Dühring
chart or other thermo data provided to determine the boiling point of the solution TN and then use
the steam tables to determine the temperature of the steam Ts.
From the steam table, the boiling point of pure water at 1.94 psia (or 29.8 psia) is found as
124.921ºF (or 250ºF). Add the boiling-point elevation BPE(0.5) = 4.400ºF gives the boiling
temperature of the liquor in the last effect as
T3 = 124.921 + 4.400 = 129.321ºF and Ts = 250ºF
Step 2) calculate final liquor flowrate and total capacity from overall mass balances
x F N
LN = F
xN
and ∑ V=i F − LN
i =1

06-361 page 64 Spring, 2017
3
0.1  lb  lb lb
=L3 =  50, 000  10, 000
0.5  hr  hr
and ∑ Vi = 50, 000 − 10, 000 = 40, 000 hr
i =1
Step 3) As a first guess, assume the vapor flowrates from each effect are equal:
F − LN
Vi = for i = 1,2 … N (53)
N
lb
40, 000
V= hr= 13,333 lb
1 V=
2 V=
3
3 hr
Step 4) calculate the liquid flowrate and composition leaving each effect using total mass and
solute mass balances:
x F
Step 4a) L1= F − V1 and x1 = F for i = 1
L1
x L
Step 4b) =
Li Li −1 − Vi and xi = i −1 i −1 for i = 2, 3 … N–1
Li
lb 50, 000
L1 = 50, 000 − 13,333 = 36.667 = = 0.1364
and x1 0.5
hr 36, 667
lb 36, 667
L2 = 36, 667 − 13,333 = 23.333 = = 0.2143
and x2 0.1364
hr 23,333
We don’t need to do the calculations for i = N (except possibly as a check) because the results for
L3 and x3 are already known.
Step 5) Knowing all the concentrations xi we can now estimate the boiling-point elevation for
each effect. Although we don’t know the pressures (except in the last effect), BPE is practically
independent of pressure. This is why the lines of a Dühring chart are nearly parallel.
BPEi ≈ BPE(xi)
Using the known xis and the BPE formula provided in the problem statement, the BPEs can be
calculated:
BPEi = 3.2xi + 11.2 xi2 = (0.645, 1.200, 4.400) ºF
Step 6) Estimate the driving force in each effect using (52) in which the equal areas cancel out:

06-361 page 65 Spring, 2017
qi
 N −1  Ui
∆Ti =  Ts − TN − ∑ BPEi  for i = 1,2 … N (54)
  q1 q q
 i =1  U + 2 U + + N U
1 2 N
Although the q’s will turn out to be unequal in Step 10), the first time through this loop, we will
assume them to be equal (in order to get started):
N −1 1
  Ui
∆Ti =  Ts − TN − ∑ BPEi  (55)
  1 + 1 + + 1
 i =1  U1 U2 UN
Comment: Geankoplis corrects the value of ∆T1 “for the cold feed”. This correction (if done)
takes the form:
 Fc pF (T1 − TF ) 
( ∆T1 )corrected =
1 +  ∆T1
 V1λ s1 
where the heat capacity cpF, heat of vaporization λs1 and Ti are evaluated below. This increase
in ∆T1 is then subtracted equally from the remaining values so that the sum of the ∆T’s is
unchanged.
( ∆T1 )corrected − ( ∆T1 )

( ∆Ti )corrected =∆Ti − for i = 2, 3 … N-1
N −1
At the end of the first full iteration, we will calculate each of the q’s including sensible heat
effects; the ∆T’s will be re-calculated from (54) instead of (55). If such rigorous calculations are
planned, the main effect of making this correction is perhaps to reduce the number of iterations
by one (is it worth the effort?). On the other hand, if we are willing to settle for an approximate
solution without iteration, then this correction is probably worthwhile. In the example which
follows, I will not make this correction.
With an eye to eventually evaluating (55): First, the sum of the driving forces is calculated as
2
Ts − T3 − ∑ BPE j = 250 − 129.3 − ( 0.6 + 1.2 ) = 118.8ºF
j =1
Notice that the BPE in the last effect is not included in the above summation. Next, we will
calculate the sum of 1/Ui:
1 1 1 1 1 1 ft 2 ⋅ hr ⋅ °F
+ + = + + = 0.00968
U1 U 2 U 3 500 350 200 BTU

06-361 page 66 Spring, 2017
Using the given overall heat transfer coefficients we can now estimate the driving force in each
effect from (55):
1
∆T1= 118.8 = 22.3°F
500 × 0.00968
Similarly for other two effects:
∆Ti = (22.331, 35.092, 61.411) ºF
In this example we will pass on making the correction for the cold feed.
Step 7) Calculate temperatures of the hot and cold streams in each effect. The hot fluid
temperature is the temperature at which the vapor from the previous effect condenses. If there
were no BPE (as with saturated steam used in the first effect), then the hot fluid temperature in
the ith effect is just be the boiling temperature from the previous effect. But if there is BPE, then
the hot fluid temperature in the ith effect is the condensation temperature of the vapor from the (i-
1)th effect, which in turn is the boiling temperature of the solution in the (i-1)th effect minus the
BPE:
Th1 = Ts and Thi = Ti-1 – BPEi-1 for i = 2, 3 … N (56)
The cold fluid temperature is the hot-fluid temperature minus the driving force:
T=
i Thi − ∆Ti for i = 1, 2 … N–1 (57)
There is no need to evaluate again TN in the above calculation, because its calculation in Step 1)
is exact; however repeating the calculation might be worthwhile as a check on other calculations.
The hot- and cold-stream temperatures in each effect can be estimated from (56) and (57) as
Th1 = Ts = 250ºF and T1 = Th1 – ∆T1 = 250 – 22.331 = 227.669ºF
for i = 2: Th2 = T1 – BPE1 = 227.669 – 0.645 = 227.024ºF
T2 = Th2 - ∆T2 = 227.024 – 35.092 = 191.932ºF
for i = 3: Th3 = T2 – BPE2 = 191.932 – 1.200 = 190.732ºF
T3 = Th3 - ∆T3 = 190.732 – 61.411 = 129.321ºF
Step 8) Knowing the temperature and composition of all the streams, we now determine their
enthalpies from given thermo data. All vapor streams are pure water, so we use the steam tables
to evaluate their enthalpies. We look for the enthalpy of saturated vapor at the boiling
temperature:

06-361 page 67 Spring, 2017
H vi = H sat'd vapor (Ti ) for i = 1,2 … N♦
Remember the temperature of the vapor leaving the ith effect is Ti.
Since steam tables will be employed in most calculations, we adapt the reference state used
there (H ≡ 0 for pure liquid water at 0ºC). If enthalpy data is available for the concentrated
solution (for example, H for NaOH solutions is given in Figure 6 or Fig. 8.4-3 in Geankoplis),
then use that:
Hi = H(Ti, xi) for i = 1,2 … N
Important: If you use more than one source of enthalpy data, make sure all sources employ the
same reference state for enthalpy. We have already noted that the steam tables from Geankoplis
define H ≡ 0 for saturated liquid water at 0ºC. By inspection of Figure 6 we see that H = 0 for 0
wt% NaOH at 32ºF. So this second source is OK.
When heat of dilution is negligible, enthalpy data for solutions can be estimated by
specifying the heat capacity as a function of concentration. Then enthalpy can be calculated
from
or Hi = cp(xi) (Ti – Tref) for i = 1,2 … N
where the reference temperature for enthalpy (that employed in the steam tables) is 0ºC or Tref =
273ºK.
We will also need the heat of vaporization of the “steam” used in the ith effect. Since some
of the vapors are superheated, we need to correct as follows:
λs1 = Hsat'd vapor(Ts) – Hliquid(Ts) (58)
λsi = Hv,i-1 – Hliquid(Thi) for i = 2, 3 … N (59)
Knowing their temperatures and compositions, we can now estimate the enthalpy of each stream.
The liquid heat capacities and enthalpies♣ are found to be
cpF = 0.944 btu/lb-ºF
♦ Owing to BPE, the vapor produced by boiling the solution is superheated (the pressure of the vapor is lower than
the vapor pressure of pure water at this temperature); we are neglecting the effect of pressure on vapor enthalpies
(see justification in footnote ♠ on page 43).
♣ Liquid enthalpies were calculated as c (T - 32°F). This assumes that all sugar solutions (regardless of sugar
pi i
concentration) have zero enthalpy at the freezing point of pure water. Actually the reference condition for enthalpy
is pure water at 32°F. There can be a slight effect of concentration: note that the isotherms in Figure 6 for NaOH
solutions at 40°F and 50°F are nearly horizonal, but not exactly.

06-361 page 68 Spring, 2017
and HF = cpF(TF – Tref) = 0.944(80 – 32) = 45.9 btu/lb
Similarly for the other liquid streams:
cpi = (0.924, 0.880, 0.720) btu/lb-ºF and Hi = (181.3, 141.3, 70.5) btu/lb
The vapor enthalpies are read directly from the steam table using the boiling temperature Ti:
Hvi = (1156.3, 1142.8, 1117.3) btu/lb
To evaluate the enthalpy of the condensate leaving the ith effect, we need the condensation
temperatures calculated from (56):
Thi = (250, 227.0, 190.7) ºF
The heat given up by condensing the superheated vapor from the previous effect can be
estimated from (58) and (59):
λsi = (945.6, 961.0, 984.1) btu/lb (60)
Step 9) Recalculate all the flowrates using mass and energy balances:
F = L1 + V1
HFF + λs1S = H1L1 + Hv1V1
Li-1 = Li + Vi for i = 2, 3 … N
Hi-1Li-1 + λsiVi-1= HiLi + HviVi for i = 2, 3 … N
These are 2N equations in 2N unknowns: L1, L2 … LN-1, S, V1, V2 … VN. The steam flowrate S
replaces LN in this list of unknowns since LN is known. Collecting terms involving the unknown
flows on the left-hand side and know flows on the right-hand side, gives the following matrix
equation to solve (shown for N=3):
 H1 0 −λ s1 H v1 0 0   L1   FH F 

 − H1 H2 0 −λ s1 Hv2 0   L2   0 
 0 −H2 0 0 −λ s3 H v3   S   − L3 H 3 
   =   (61)
 1 0 0 1 0 0   V1   F 
 −1 1 0 0 1 0  V2   0 
    
 0 −1 0 0 0 1  V3   − L3 

06-361 page 69 Spring, 2017
The first three rows are the three energy balance equations and the last three rows are the mass
balances. While this could be solved by hand using elimination, it can be very easily solved in
Mathcad (or other software) using the LSOLVE function.
In preparation for doing mass and energy balances we construct the following two matrices
M and B in (61):
 181.3 0 −945.6 1156.3 0 0   L1   2.294 × 106 

 −181.3 141.3 −  L   
 0 961.0 1142.8 0   2  0 
 0 −141.3 −984.1 1117.5  S   
0 0
=  −7.050 × 105 
  
 1 0 0 1 0 0   V1   50, 000 
 −1 1 0 0 1 0  V2   0


  
 0 −1 1   3 
V  
 −10, 000 
0 0 0
This matrix equation MX = B was then solved for X using the LSOLVE function in Mathcad.
The result for X is
 L1   37, 611
L   
 2  24, 222 
S 19,934 
=
X =    lb/hr
 V1   12,389 
V2  13,389 
   
V3  14, 222 
Step 10) Calculate the heat duties for each effect from
q1 = λs1S and qi = λsiVi-1 for i = 2, 3 … N (62)
The vapor flowrates V1, V2 … VN calculated in Step 9) will generally differ from those
approximated by (53). If this difference is only a few percent, you can stop here and skip to Step
12). If the difference is significant (and it almost always is significant the first time you reach
this step), you need to iterate in Step 11) until the difference from the previous step is negligible.
The heat duties of each effect are then calculated from (62) using the modified latent heats λsi
from (60):
qi = (18.8, 11.9, 13.2) 106 btu/hr
Notice that the heat duties are not all the same [as we had assumed in Step 6 – see (55)]. This
means we have to iterate at least once more. We will iterate until the vapor flowrate converge
within some tolerance.
Step 11) Repeat Step 4b) through Step 10) using the updated ∆Ts until the flowrates converge.

06-361 page 70 Spring, 2017
repeat Step 4) We next recalculate the xi’s and BPEi’s:
xi = (0.133, 0.206, 0.500) and BPEi = (0.62, 1.14, 4.40)
repeat Step 5&6) We then recalculate the driving forces, using the qi’s:
∆Ti = (30.4, 30.1, 58.4) ºF
repeat Step 7-10) Calculate a new set of flowrates (all three sets are shown):
 L1  36, 667   37, 611 37,507 

 L   23,333   24, 222   24,157 
 2      
S  ?  19,934  19, 722 
 = → → 
 V1  13,333  12,389  12, 493  lb/hr
V2  13,333  13,389  13,350 
       
V3  13,333  14, 222  14,157 
initial guess 1st iter 2nd iter
Notice that the changes are relatively small after the 2nd iteration, so we stop.
Step 12) Once converged, the areas can be calculated from the Hxer design equation:
qi
Ai = (63)
U i ∆Ti
The area should be the same for each effect, because that was what was assumed in (54).
Calculate the heat exchanger areas from (63):
Ai = (1,128.5, 1,128.4, 1,128.6) ft2
The areas should be equal, but are not quite equal because the mass and energy balances are not
perfectly converged. Additional iterations would make them more nearly equal. The area to use
in design is the average of the three:
A = 1129 ft2
The steam requirements are S = 19,722 lb/hr
40, 000
The economy is = 2.03
19, 722

06-361 page 71 Spring, 2017
Although it was not asked for, you might be interested to know how the results depend on the
number of effects. Here is a summary:♦
N
S VN ∑ Ai (ft2) economyAs
(lb/hr) (lb/hr) i Economy Total Area
1 44,843 40,000 703 0.89 expected,
steam 4.5
2 25,575 20,450 1,454 1.56
3 18,984 14,000 2,208 2.11 consumpti 4
4 15,717 10,800 2,978 2.55 on S 3.5
decreases
3
with N (or economy increases). Somewhat surprising is the
large increase with N in total area of all N heat exchangers: 2.5
the total area is essentially proportional to N. The reason 2
for this large increase can be rationalized as follows. Recall 1.5
the design equation for any of the heat exchangers: 1
0.5
qi
Ai = 0
U i ∆Ti
1 2 3 4
Let’s again neglect sensible heat effects and BPE as well as
the T-dependence of the heat of vaporization. Then
F − LN
∑ qi ≈ λ
i
F − LN T − TN
and qi ≈ and ∆Ti ≈ s
Nλ N
Substituting these results back into the expression for Ai and summing:
F − LN
qi Nλ = F − LN F − LN
Ai = ≈
(Ts − TN ) λ
and ∑ Ai ≈ N (T ∝N
U i ∆Ti Ts − TN
i s − TN ) λ
N
This leads to the prediction that the total area is proportional to N, as in our more rigorous
calculation in the example. Thus the substantial increase in steam economy comes at the price of
a substantial increase in heat exchanger area. Is it worth it? The answer depends on the relative
cost of heat exchanger area and steam.
♦The economy for N=3 is slightly different from what was calculated in the detailed example above. This is
because in varying N I have used U = 500 for all hxers, whereas in the above example, each hxer had a different
value.

06-361 page 72 Spring, 2017

VAPOR RECOMPRESSION
In the previous section, we saw how lowering the pressure on the vapor space of downstream
effects allowed us to use the vapor produced in upstream effects in place of steam. An
alternative to create a driving force to transfer energy from the vapor back to boil the liquor is to
raise the pressure on the vapor. In this section, we examine vapor recompression.
Recall that the problem with using the vapor in

place of steam is that the temperature of the vapor is
the same as the temperature of the liquor: the liquor
and vapor leaving the same effect are already in
thermal equilibrium with each other. In other words,
there is no driving force for heat transfer.
without compression: ∆T = Ts - T = 0
One way to increase the temperature of the vapor is to

compress it, as in the figure at right.
Adiabatic compression of water vapor (treated as an ideal gas) increases its temperature
according to♣
1− 1
Tout  pout  γ
=  (64)
Tin  pin 
where “out” and “in” refer to the outlet and inlet of the compressor and where γ is the heat
capacity ratio, which is
ideal
c p gas cp water
γ≡ ≈ = 1.324 or 1− 1 =0.245
cv R γ
cp −
M
for water vapor using

cp = 0.45 Btu/lb-°F = 0.45 cal/g-ºC,
R = ideal gas constant = 1.986 Btu/lbmol-R = 1.986 cal/mol-K
M = mol. wgt. = 18 lb/lbmol = 18 g/mol
♣ Eq. (3.3-16) in Geankoplis.

06-361 page 73 Spring, 2017
psia
4.7
Temperature (degR)
=1
r p in
n fo
1000 ssio
pr e
C om
ab atic
t Adi
Boiling Poin
1 10 100
Outlet Pressure (psia)
Figure 8: Red curve show how temperature increases with adiabatic compression, starting with
saturated vapor at 14.7 psia. For comparison, the boiling point is shown by the black curve.
This increase in compressor outlet temperature is plotted in Figure 8. Note that the outlet
temperature rises faster with increases in outlet pressure than the boiling point increases with
pressure. Thus adiabatic compression of a saturated vapor nearly always produces a superheated
vapor. When heat is removed from this compressed vapor, it will eventually condense at its
boiling point for that pressure.
Increasing the pressure of the vapor also increases its boiling point, so that as the vapor
condenses, it stays at a high temperature Ts so we always have some driving force.
With compression: ∆T = Ts - T > 0
Example: Suppose we compress the vapor above a solution of organic colloids (negligible BPE)
from pin = 1 atm to pout = 2 atm. Evaluate ∆T for use in the Hxer design equation.
Solution: The inlet temperature for the compressor is the boiling point for 1 atm: Tin = 212ºF.
The outlet temperature for the compressor can be calculated from (64): Tout = 339ºF. But this is
not the temperature of the hot fluid which should be used. The compressed vapor is superheated.
The boiling point at pout = 2 atm is Ts = 250ºF. It is this second temperature which should be
used in preliminary designs:
∆T = 250 – 212ºF = 38ºF
Unfortunately, raising the temperature and pressure of the vapor actually lowers its latent
heat per pound
λs < λ (because Ts > T)

06-361 page 74 Spring, 2017
so the vapor no longer has enough latent heat to boil the same mass of liquid. Thus make-up
steam is required, although generally much less steam is needed than if the vapor’s latent heat is
not recycled.
The reduction in steam is not without cost. We still have to pay the utilities (electricity or
steam turbine) to operate the compressor. Generally, compressors are expensive to buy,
maintain, and operate; so you have to do some economic calculations to see if using a
compressor is economical. At the usual conditions of an evaporator, the vapor can be treated as
an ideal gas. Then the work required for adiabatic compression can be estimated from♦
 
1− 1 γ 
γ RTin p
 = Btu
=W pr out
 − 1 [ ] (65)
γ −1 M  pin   lb
 
where the subscripts “in” and “out” refer to conditions in the inlet or outlet to the compressor and
R is the universal gas constant. Now (65) gives the work needed to compress a unit mass of gas,
under ideal conditions (no irreversible losses of energy). To get the power requirements, we
must multiply by the mass flow rate and divide by the mechanical efficiency (η) of the
compressor [typically η ≈ 0.7]:
=P
(=
m f − m ) W pr
[=]
Btu
[ ] kW
η hr
By compressing the vapor to raise its temperature, we are lowering its latent heat of vaporization.
This means that, even if we completely condense all the vapor, we still won’t have enough latent
heat to boil the liquor at its lower temperature. The difference must come from make-up steam.
To estimate the make-up steam requirements, let’s return to the equation (32) for the total
heat duty added to the cold fluid:
energy content energy content

of cold fluid leaving
 of cold fluid entering

( 
)
q = m f − m H v + mH
 −

 liquor
m f H f

vapor feed
♦ See Eq. (3.3-15) in Geankoplis.

06-361 page 75 Spring, 2017
In the simple case where BPE and ∆Hdil are negligible (i.e. for inorganic colloids), Hf and H
differ from the enthalpy of saturated liquid only in sensible heat. To the extent that sensible heat
is negligible,♣ we can substitute the enthalpy of saturated liquid for Hf and H:
neglecting sensible heat: ( )

q =m f − m H v − m f H f + m 

H
≈H
≈ HL L
( ) (
= m f − m H v − m f − m H L)
( m f
= − m ) ( H v − H L )
heat to boil vapor: (

q ≈ m f − m λ ) (66)
This heat duty is partially offset by the latent heat recovered by condensing the compressed
vapor. The remainder comes from make up steam:
=
q (
m f − m ) λ s
 
+ m s λ s
 (67)
compressed vapor make-up steam
where we are assuming we are compressing the vapor to the same pressure as the make-up
steam. Eliminating q between (66) and (67), we can solve for the steam requirements:
 λ 
(
m s ≈ m f − m  − 1)
 λs 
Notice that any compression allows us to recover all the latent heat of the vapor, thus
significantly reducing the steam consumption which would be otherwise be required without
compression. However greater compression actually increases steam consumption (because λs
decreases with pressure). Rearranging this expression, we can calculate the economy:
m f − m λs
≈ >> 1
m s λ − λs
which is now very much greater than unity. However, this way of computing economy is unfair
because the reduction in steam consumption is at the cost of operating the compressor. In
computing the “economy” for an evaporator with vapor recompression, it is customary to add an
♣ Sensible heat is often negligible compared to latent heat. To see why, note that the latent heat of vaporization of
water at 1atm, 212°F is 970.3 btu/lb, where 1 btu is the amount of sensible heat required to raise the temperature of
one lb of water 1°F. This means that, for sensible heat to equal latent heat, I need to raise the temperature of water
970°F. A temperature rise of only 30°F, say, represents a sensible heat change of 30/970 or 3% of the latent heat of
s.
water. This is an estimate of the relative error made in this estimate of m

06-361 page 76 Spring, 2017
equivalent steam rate for the compressor to the make-up steam rate before dividing into the
capacity.
• If the plant uses a steam-powered turbine or if the plant uses steam generators to make its
own electricity, then the apparent steam usage is
P btu hr lb
( m s )app =
m s + [=
] [=
]
λ ′s ηs btu lb hr
where P is the electrical power supplied to the compressor, λ ′s is the latent heat of the steam
used (the prime is added because the pressure of the steam used here might be different from that
used in the evaporator) and ηs is the efficiency of electricity generation from steam (typically
0.35).
• On the other hand, if electricity is purchased from a utility to drive the compressor, then it
makes sense to convert the electricity usage into an equivalent amount of steam (on a
financial basis):
$
( m s )app = m s + P ×
$
kW-hr
lb of stm
If P is in kW then the last term yields lb/hr of steam. Regardless of which method of converting
the power usage is appropriate, the economy is computed as:
m f − m
economy ≡
( m s )app
In Hwk #4, Prob. 1, we will consider

the trade off of operating costs (direct
4 .10
5
steam usage plus electricity for
14.7 23.2
Annual Costs ($)
compressor) and capital costs (heat

exchanger area) as a function of the
output pressure of the compressor. The
2 .10
5
results are summarized in the graph at
right, which shows that an optimum
pressure exists at which total costs are
minimized.
0
15 20 25 30
Notice that both electricity costs for
P ressure at compressor exit (psia)
operating the compressor as well as Tot al
steam costs increase with pressure. The Heat Exchanger Area
steam usage goes up because as the Electricit y
pressure increases, the temperature of the Steam
steam and compressed vapor increase

06-361 page 77 Spring, 2017
which decreases the latent heat of vaporization, requiring more make-up steam. Only when the
cost of the heat exchangers is considered does an optimum exist. The cost of the heat exchanger
decreases with pressure because a larger driving force for heat transfer leading to smaller area.
Comment: It might appear surprising that the amount of steam increases with the pressure of the
compressed vapor. After all, we are arguing that compression reduces steam requirements; you
might expect that more compression reduces steam requirements more – but it doesn’t. The
reason is simple: the higher the pressure of the compressed vapor, the more its latent heat of
vaporization λs is reduced (see example on page 47) and the more make-up steam is required
from (67) for the same vapor flowrate mf –
m and the same heat duty q. Cos ts of Simple Ev aporato r
2 .10
6
14.7 23.5
At the minimum, the total costs with
vapor recompression are about $159,000
1.5 .10
6
per year. If we use all steam instead of
Annual Costs ($)
vapor recompression plus make-up steam,
the cost of the steam alone would exceed 1 .10
6
$1.4 million per year. Clearly vapor

recompression is the way to go in this
5 .10
5
problem.
0
15 20 25 30
Pres sure of Steam (p sia)

Total w/o compressor
Heat Exchanger Area
Electricity
Steam
Total with vapor recompress ion (for comparison)

06-361 page 78 Spring, 2017
Chapter 6. Vapor-Liquid Equilibrium

Reference: Smith & Van Ness, 4th Ed., Sect. 10.5
Sandler, 3rd Ed., Sect. 8.1
The next unit operation we will consider is distillation in which we attempt to separate a
mixture of two miscible components by creating a second phase: either partially vaporizing a
liquid or partially condensing a vapor. This is quite different from evaporation because both
components are present in both liquid and vapor phases (whereas evaporation involves an
nonvolatile solute).
In general, the liquid and vapor phases (created by partially vaporizing a liquid feed or
partially condensing a vapor feed) will have different concentrations of the two components.
This is the basis of separation by distillation. We will start by reviewing the principals of vapor-
liquid equilibrium, which limits the amount of separation that can be accomplished.
NO. OF PHASES PRESENT

Suppose I have an apparatus like that shown in the figure at 1 atm
right in which I can place an arbitrary mixture of components
and control the temperature and pressure. How many phases
will be present at equilibrium?
z
One Component
T
The phase diagram for water is shown below:
105
piston
104
Liquid
103
Pressure (kPa)
102
Solid
101
Vapor
100
10-1
10-2
-50 0 50 100 150 200
Temperature (oC)
Figure 9: phase diagram for pure water (no air).

06-361 page 79 Spring, 2017
To comprehend the meaning of such phase diagrams, it’s helpful to recall Gibbs phase rule:♦
P+F=C+2
where P is the number of phases present, F is the number of degrees of freedom and C is the
number of components. For example, in the phase diagram above, we have only one component
present (water♥), so C = 1 and Gibbs phase rule says
for C = 1: P+F=3
Suppose we are interested in portions of the diagram having only one phase present at
equilibrium: P = 1. Then Gibbs predicts F = 2; i.e. two degrees of freedom. This means that we
can vary both pressure and temperature over some range and still have only one phase present.
In other words, single phase regions are represented on this P-T chart as 2-D areas, which of
course are marked with the words “liquid”, “vapor” or “solid” representing the single phase
found under those conditions.
Conditions in which two phases co-exist at equilibrium correspond to P = 2 and F = 1: or one

degree of freedom. This means that you can vary either temperature or pressure, but not both. In
other words, if you specify a particular temperature, there is only one pressure at which two
phases will co-exist. Once you pick the temperature you are not “free” to also specify pressure.
Two-phase regions on the phase diagrams are lines; the two phases present are indicated by the
words marking the regions on either side of the line.
Conditions in which three phases co-exist at equilibrium correspond to P = 3 and F = 0: or no

degree of freedom. In other words, you can arbitrarily specify neither temperature nor pressure.
Three phase regions are isolated points on the phase diagram. For example, water has its triple
point at T = 0.01ºC and p = 0.611 kPa (0.00603 atm) where ice, water and steam co-exist at
equilibrium.
Suppose we put one pound of water into our

piston at 25ºC and 1 atm (101.3 kPa). Of course, this
places us in the interior of the region marked “liquid”
and we observe a single phase. As heat is added, the
temperature and vapor fraction increases as shown in
the sketch at right.
Two Components
Now suppose I put two components in the pot:

let’s say a mixture of n-pentane (C5) and n-octane
♦ The somewhat silly mnemonic I use to remember this is “Police Force = Cops + 2”.
♥ It’s important to exclude air which might otherwise be present. Air (itself a mixture of nitrogen and oxygen)
behaves like a second or third component and will affect phase equilibria.

06-361 page 80 Spring, 2017
(C8). The vapor-liquid phase behavior of two-component systems is a little more complicated
because I have an additional variable — composition — as well as T and P. Thus we need a 3-D
graph to represent the phase diagram for two component mixtures. The vapor-liquid portion of
the phase diagram for two components is shown in Figure 10.
250 250 Vapor
200
Vapor 200
L&V
Temperature (degC)
Temperature (degC)
150 150
100 100
50 50
0 L&V 0
Liquid
-50 -50
1
(a tm) 1
(a tm)
0.2 sure 0.8 sure
0.4
0.6
0.8 0.1
Pres 0.6
0.4
0.2 10
Pres
1.0 0.0
Mole Fraction Mole Fraction
Pentane Pentane
Figure 10: Phase diagram for mixtures of pentane and octane. The two views are from the low-
pressure, pentane-rich side (left) or from the high-pressure, pentane-lean side (right).
Suppose we fix the pressure inside our piston. Instead of a single temperature at which both
liquid and vapor can co-exist (as in Figure 9), there is a range of temperatures. This can more
easily be visualized by taking a vertical slice though the 3-D phase diagram of Figure 10 (a
P=const plane).
The result is a Txy diagram like that shown at right, where

T refers to the temperature and x and y refer to mole fractions
of the more-volatile component in either the liquid (x) or vapor
(y) phase. To understand the meaning of this diagram,
consider what happens if we charge the piston with a mixture
of heptane and octane having a particular composition,
indicated by the vertical dashed line in the figure at right. If
the temperature is low enough (i.e. T < Tbub), we will have just
one phase: liquid.
As we heat this mixture to the temperature corresponding

to the lower curve, we begin to see bubbles of vapor formed in
the liquid as more heat is added. This temperature at which
the bubbles first form is called the
bubble point — when heating a subcooled liquid, the

temperature at which the first bubble forms

06-361 page 81 Spring, 2017
The composition (y) of this first bubble can also be read from the Txy diagram. In general, the
vapor will be richer than the liquid in one component and leaner in the other. This difference in
composition of the vapor and liquid can serve as the basis for separating two or more
components of a mixture. By completely condensing the vapor, I produce a liquid which is
richer in one component than the original liquid. Thus I have achieved a degree of separation of
the two components.
For a binary system having two phases, we denote the composition of the liquid and vapor
using x and y:
x = mole fraction of more volatile component♣ in liquid phase
y = mole fraction of more volatile component in vapor phase
Keep in mind that liquid and vapor must have the same temperature at equilibrium. The line
connecting the two points representing vapor and liquid is called a tie line. It’s horizontal on a
Txy diagram.
As we continue to heat the mixture, the liquid becomes

less rich in the more volatile component since the more of the
more volatile component is leaving for the vapor phase. The
compositions of the liquid and vapor move along the heavy
solid line shown in the figure at right. The highest
temperature at which liquid remains is called the
dew point — when cooling a superheated vapor, the

temperature at which the first drop of dew
forms
The temperature, vapor fraction and composition are

shown at right as a function of the amount of heat
added. Eventually, as we continue to heat, all the
mixture ends up in the vapor.
♣ By “more volatile component” we mean that component which

has the higher vapor pressure at a given temperature.

06-361 page 82 Spring, 2017
XY AND TXY DIAGRAMS
The Txy diagram is a convenient graphical method

for summarizing vapor-liquid equilibrium (VLE) data at
a particular pressure. For binary mixtures, we usually
denote the mole fraction of the more volatile component
as x in the liquid phase as y in the vapor phase. For this
particular choice of x and y, the Txy diagram will always
slope downward as shown at right.
From the Txy diagram, one can construct an “xy”

diagram, which is the one we will use in distillation
calculations. The xy diagram is shown at right. For xy
diagrams which resemble this one, the mathematical
relationship between x and y can be sometimes be
represented by a single number: the relative volatility,
which is defined generally as
K
αij ≡ i
Kj
where Ki is the distribution coefficient for the ith

component (also known as the K-Factor); this quantity is
defined as:
y
Ki ≡ i
xi
The subscripts i and j refer to components in the two-phase mixture. Generally the most volatile
component will have the largest distribution coefficient and will have its concentration in the
vapor increased (compared to the liquid) by the largest percentage. The ratio of the distribution
coefficients of two components is the relative volatility.
For a binary mixture we don’t need subscripts because x and y denote the mole fractions of
the more volatile component and 1-x and 1-y are the mole fractions of the other component.
Combining the two equations above for this binary mixture:
K more vol y x
α≡ =
K less vol (1 − y ) (1 − x )
αx
Solving for y(x): y= (68)
( α − 1) x + 1
Although relative volatility is generally a function of temperature [this implies that α = α(x) in
(68)], for ideal solutions of components having nearly equal molar heats of vaporization, the
relative volatility is nearly constant (see “Constant Relative Volatility Approx.” on page 84).

06-361 page 83 Spring, 2017
Then the VLE for an N-component mixtures can be entirely represented using only N-1 relative
volatilities. For a binary mixture, only one constant is needed and the xy diagram can be
represented by (68).
IDEAL SOLUTIONS AND CONSTANT RELATIVE VOLATILITY

Reference: Smith & Van Ness, Example 10.1 (p305)
Sandler, Illustration 8.1-3 (p493).
To see why relative volatility might be independent of temperature, let’s look at ideal
solutions whose partial pressure pi is given by Raoult’s law:
Raoult's law: pi = xiPi'
where Pi' = vapor pressure of pure i
Recall that partial pressure of a component is also defined as the mole

fraction times the total pressure:
pi ≡ yiP = partial pressure of component i
P = total pressure
where xi,yi = mole fractions in liq or vapor at equil.
Substituting xi from Raoult’s law into the definition of partition coefficient, we have
yi pi P Pi′
Ki ≡ = = (69)
xi pi Pi′ P
Ki P′
and αij= = i (70)
K j Pj′

06-361 page 84 Spring, 2017
For a homologous series of components (e.g. the

alkanes), the log of the pure component vapor
pressure versus temperature is a set of uniformly
spaced curves: see figure at right. Recall that
log(P'i) - log(P'j) = log(P'i/P'j)
Thus if the vertical distance between any two of

these vapor pressures curves (i.e. the difference
in the logs) is independent of temperature, then
the ratio of the vapor pressures must also be
nearly independent of temperature and we have
nearly constant relative volatility according to
(70).
CONSTANT RELATIVE VOLATILITY

APPROX.
Time out: In case you are wondering why
constant relative volatility is such a good
approximation (at least when λi ≈ λj), it turns out
there is a good thermodynamic reason. Vapor
pressure depends very strongly on temperature: recall the Clausius-Clapeyron equation (S&VN,
p182):
d ln Pio λ d ln Pio λ
= i or = − i
dT RT 2 d (1 T ) R
where λi is the molar heat of vaporization (in the past we have used λ to denote the heat of
vaporization per unit mass, rather than per mole) of the pure component. Integrating this with
respect to T (ignoring the weak temperature dependence of λi) yields♣
λi
i′ Ai −
ln P= (71)
RT
♣ A somewhat more accurate but empirical approximation is often used to summarize vapor pressure data in
handbooks (e.g. Lange’s Handbook of Chemistry, 13th edition, pages 10-29 to 10-54):
Bi
i ′ Ai −
ln P=
T + Ci
It is called Antoine’s equation and it accounts for the variation of λi with T. Other handbooks (e.g. CRC Handbook
of Chemistry & Physics, 57th edition, pages D-183 to D-215) give vapor pressures at several temperatures and leave
you to develop your own interpolation scheme.

06-361 page 85 Spring, 2017
 λ 
or Pi′(T ) Ai′ exp  − i 
= (72)
 RT 
where Ai is just some integration constant. Since vapor pressure depends strongly on
temperature, so does the partition coefficient Ki given by (69). However, the relative volatility α
is much less dependent on temperature. The reason becomes clear when we substitute (72) into
(69):
P ′ Ai′  λi − λ j 
αij =
≡ i exp  − 
Pj′ A′j  RT 
Note that when the two molar heats of vaporization are equal, the temperature-dependence
vanishes. Even if they are only approximately equal, the temperature-dependence of the relative
volatility will be much less than of the vapor pressures. For this reason, the assumption of
constant relative volatility is frequently used.
IDEAL BINARY MIXTURES

Now let’s try to predict the Txy diagram for the simplest case: an ideal solution in
equilibrium with an ideal gas. What we have said so far about ideal solutions applies to any
number of components. Let's now restrict our attention to systems with only two components,
which we will call A and B. The mole fractions must sum to one for each phase, so we can
express the mole fraction of B in terms of that for A:
xB = 1-xA
yB = 1-yA
Applying Raoult's law to each component:
pA = yAP = xAPA' (73)
pB = (1-yA)P = (1-xA)PB' (74)
Adding the partial pressures from (73) and (74) we obtain the total pressure:
P = pA + pB = xAPA'(T) + (1-xA)PB'(T)
Solving for the mole fraction xA: xA = (P-PB')/(PA'-PB') (75)
from (73): yA = xAPA'/P (76)

06-361 page 86 Spring, 2017
Keep in might that the vapor pressures in these equations depend on temperature. So if you
specify a T and P, I can determine the vapor pressures and calculate the composition of the vapor
and liquid which will be at equilibrium.
Given: T,P
Look up: PA'(T) and PB'(T)
Calculate: xA,yA from (75) and (76)
Thus I can construct a complete phase diagram for the mixture using only vapor-pressure data
for the pure components.
Example.♥ n-heptane and n-octane form a nearly ideal mixture. Use Raoult's law to construct a
phase diagram at constant pressure (1 atm).
Solution. We first look up the normal boiling points of the two pure components.
component Tbp (°C) denoted by

n-heptane 98.4 A
n-octane 125.6 B
We choose the "low boiler" as T (°C) P'A P'B x y α

component A since it has the larger
(mmHg) (mmHg)
vapor pressure over this range of
98.4 760 333 1.000 1.000 2.28
temperatures. Next, we choose a
105 940 417 0.656 0.811 2.25
temperature between the two boiling
110 1050 484 0.488 0.674 2.17
points, look up vapor pressures of the
115 1200 561 0.311 0.492 2.14
pure components and substitute into
120 1350 650 0.157 0.279 2.08
(75) and (76). The results are
125.6 1540 760 0.000 0.000 2.03
summarized in the table at right and the
graph below.
♥ Illustration 9.1 from Treybal. A similar example is 10.1 from S&VN (p305).

06-361 page 87 Spring, 2017
126
124
122
120 Vapor
Temperature (deg C)
118
116
114
112
L&V
110
108
106 Liquid
104
102
100
98
0 0.2 0.4 0.6 0.8 1
mole fraction of heptane

06-361 page 88 Spring, 2017
NONIDEAL BEHAVIOR
Fig. 1. Txy and xy diagrams for nonideal systems. On left is carbon disulfide in acetone at one
atmosphere pressure. On right is acetone in chloroform. Taken from Treybal, 2nd ed., pages 290 and
295.
Examples Txy and xy diagrams for binary mixtures which display significant nonideal
behavior are illustrated in Fig. 1. In particular notice that the curve on the xy-diagram can cross
the 45° line. This point of crossing is called an azeotrope:
azeotrope - vapor and liquid have the exactly the same composition at equilibrium (i.e. y = x)
(other than the trivial case of pure components)
Notice also that the boiling temperature of the azeotrope is not bounded by the boiling points
of the two pure components. The example on the left is called a minimum-boiling azeotrope

06-361 page 89 Spring, 2017
while the example on the right is called a maximum-boiling azeotrope. The occurrence of
azeotropes can pose difficulty for distillation since producing a new phase by condensation or by
evaporation produces no separation of components.
BATCH (DIFFERENTIAL) DISTILLATION

Now let’s apply these concepts about vapor-liquid
equilibria to distillation. The oldest device for
distillation consisted of a kettle into which a liquid
mixture was placed. By heating the kettle either by fire
or steam or electrically, the liquid is partially
vaporized. After condensing the vapor, you have a
product which is richer in the more volatile component.
If both of the components are volatile (e.g. ethanol

and water), it wouldn't make sense to vaporize all of
the liquid in the kettle. A mass balance can tell you
that, if you vaporize the entire charge, the cumulative
condensate composition will be the same as the initial charge. Oh, perhaps you have removed
some nonvolatile minerals which might have been in the water, but you won't enrich the ethanol
concentration in this way. And enriching the composition of the more volatile component is the
objective of distillation.
The solution is to evaporate only partially, because the first vapors which come off will be
the richest in the more volatile component. Vapors produced later will only dilute the
condensate. This process is called batch distillation because we process one batch at a time
rather than continuously processing as with flash distillation.
Let's now try to predict how the composition of the liquid and vapor change as we evaporate
more and more of the charge.
Let nA(t), nB(t) = moles of component A or component B left in liq.
n(t) = nA(t) + nB(t) = total moles of liq. left in kettle at time t
n (t )
x (t ) = A = mole fraction of A in liq. remaining (77)
n (t )
y(t) = instantaneous mole fraction of vapor produced
Suppose we evaporate a small amount of liquid at time t over a time interval dt:
n(t+dt) - n(t) = dn < 0
-dn = total moles evaporated in time dt
A fraction of this is component A:

06-361 page 90 Spring, 2017
-dnA = y(-dn) (78)
(77) into (78): -d(xn) = y(-dn)
xdn + ndx = ydn
ndx = (y-x)dn
dn dx
=
n y−x
Integrating from t=0, at which time x,n have their initial values:
x(0) = x0 and n(0) = n0
to some later time, at which time x,n have values:
x(t) and n(t)
n (t ) x (t )
dn dx
we obtain ∫ n
= ∫ y−x
n0 x0
x0
n dx
Integrating: ln 0 =
n (t ) ∫ y−x
x (t )
 x0 
 dx 
Solving for n(t): =(
n t ) n0 exp −
 y − x ∫ (79)
 x (t ) 
which is called Rayleigh's equation. This serves as the design equation

for batch distillation. To use this equation, we must recall that the vapor
coming out of the kettle at any time is in equilibrium with the remaining
liquid:
at equilibrium: y = f (x)
For each x between x(0) and x(t), we can look up the corresponding y(t)
and compute the integrand in Rayleigh's equation 1/(y-x). The integral is the area under the
curve between these two limits.

06-361 page 91 Spring, 2017
Example: A 50mol% mixture of heptane and

octane is placed in a batch still and heated. a)
Calculate the fraction of the initial charge which
must distilled to reduce the concentration
remaining in the still to 20mol% heptane. b)
Calculate the corresponding composition of the
accumulated condensate.
At the pressure in the still, the vapor-liquid

equilibrium can be reasonably represented by a
constant relative volatility of α = 1.7 (for a discussion on when this might arise, see p84). For
binary solutions, it's customary to define x and y to be the mole fraction of the more volatile
component (heptane in this case); this choice yields relative volatilities greater than unity.
Solution: We evaluate n/n0 from Rayleigh's equation (79)
Step 1: Assign a value to x(t) in the interval 0.2 - 0.5, say x=0.2.
Step 2: Calculate the corresponding y. We expect the vapor to be richer in the more volatile
component. For constant relative volatility, the equilibrium curve is given by (68):
αx
y= 1
( α − 1) x + 1 α = 1.7
This function is evaluated for α=1.7 and plotted in the

figure at right. For x=0.2, this yields, y=0.298.
y
0.5
Step 3: Calculate the corresponding value of the integrand

in Rayleigh’s equation
1 0
= 10.2 0 0.5 1
y−x x
Step 4: Repeat Steps 1-3 for various values of x(t) in the

interval 0.2 to 0.5.
Step 5: Evaluate the integral appearing in 15
0.2 0.5
n (t )  0.5 dx 
= exp  −
n0  ∫
y − x 1
10
 0.2  y
i
x
i
5
The integrand is plotted in the figure at right. The area
under the curve between the two dotted lines is
0
0 0.5 1
x
i

06-361 page 92 Spring, 2017
0.5
dx
∫ y−x
= 2.45
0.2
n (t )  0.5 dx 
Rayleigh's equation yields
n0
exp  −
=
 ∫ =
y − x
0.086
 0.2 
This represents the fraction of the initial charge which remains in the still as liquid. The fraction
of the initial charge which has been distilled is just unity minus this
n (t )
1− =
0.914
n0
0.8
Step 6: Calculate the condensate concentration
If all the condensate is collected in a container

0.6
and mixed together, the average concentration of
the condensate will be
0.4
moles of A depleted
x =
cond total moles depleted
0.2
x0 n0 − xn
= = 0.528
n0 − n
0
0 0.5 1
The graph at right shows what typically happens x(t)
fraction of initial charge distilled =
as the initial liquid is more and more distilled. y(t) [n(0)-n(t)]/n(0)
xbar_condensate
Notice that the vapor is initially richer than the x(0)
liquid. Since the leaving vapor is richer, the
remaining liquid becomes leaner. Thus both x and y become leaner as distillation proceeds.

06-361 page 93 Spring, 2017
Chapter 7. Flash Distillation

The simplest device for continuous distillation of some
liquid feed is the “flash unit.” This is just a heat exchanger
to add enthalpy to the feed and a chamber to allow the
resulting vapor and liquid to separate. The schematic at
right is the same as that used for evaporation; however the
vapor now contains all of the components — not just the
“solvent”.
Given the feed composition and the amount of heat

added to or removed from the feed, we would like to
determine the composition and flowrates of the distillate
and bottoms products.
General Flash Problem:
Given: ♠ F, zF, HF, Q and thermo data
Find: D, B, yD and xB
Solution: the mole and enthalpy balances are:
F=D+B (80)
zFF = yDD + xBB (81)
HFF + Q = HDD + HBB (82)
where F,D,B are total molar flowrates of each stream and x and y are mole fractions of the more
volatile component. The remaining equation comes from requiring the vapor and liquid products
to be at equilibrium:
(xB, yD) lies on equil.curve (83)
(80)-(83) represent four equations in the four unknowns. Of course, HF, HD and HB are also
needed. In general, for a given pressure (which we are assuming to be held constant), the
enthalpy of any stream depends on its temperature and composition. For a 2-component system,
having two phases at equilibrium, temperature and composition are not independent. For a given
value of yD (for example), there is a single temperature at which liquid coexists and its
♠ We use x for mole fraction of the liquid and y for mole fraction of the vapor. More generally, when we have a
two-phase mixture (or if we are talking more generally and don’t wish to specify which phase), we will use z for
mole fraction of the mixture (i.e. total moles of the more volative component in both phases divided by the total
number of moles in the two phases).

06-361 page 94 Spring, 2017
composition is xB. Thus for a particular pressure and yD, the values of HD, HB, xB and
temperature are unique and can be found from the given “thermo data.” The details will become
clearer in upcoming examples.
Using (80) to eliminate B in (82) and then solve for D:
D HF − HB Q F
f=
≡ + (84)
F HD − HB HD − HB
The ratio D/F represents the (molar) fraction of the feed which is vaporized. This fraction is
defined as f. Dividing (80) by F and substituting (84):
B D
=1 − =1 − f (85)
F F
Dividing (81) by F and substituting (84) and (85):
z F = yD f + xB (1 − f ) (86)
f −1 z
Solving for yD: =yD xB + F (87)
 f f
<0
These relationships will be used in the example which follow.
APPROXIMATE SOLUTION
The solution to the four equations (80)-(83) can be obtained in a number of ways. In some
mixtures (especially of homologous series* like the alkanes), the molar heat of vaporization is
nearly the same for both components. For this reason, we tend to solve all distillation problems
using mole fraction rather than mass fractions, and express enthalpies per unit mole rather than
enthalpy per unit mass. If you can assume that the molar heat of vaporization of the mixture is
completely independent of its composition, then solution is quite easy:
• Assume HD-HB = const, say λ [=] Btu/mol (independent of yD)

• neglect sensible heat compared to latent heat
• feed is all liquid
* A “homologous series” is a set of very similar items, which differ in only one property. In the case of alkanes, the
series is a set of components of similar molecular structure which differ only in the number of methyl groups in the
chain.

06-361 page 95 Spring, 2017
Let's choose saturated liquid having the composition of the feed as our reference state for
enthalpies and neglect sensible heat effects. Then we can re-state our example problem:
Example #1: liquid feed
Given: F, xF, HF, Q>0, xy data and λ
Approx. Solution: if we neglect sensible heat, the two liquid streams have the same enthalpy. If
we choose the saturated liquid as the reference state for enthalpies, then that enthalpy is zero:
HF = HB = 0
The vapor stream has a higher enthalpy equal to the heat of vaporization
HD = λ
D 0 Q F Q
When HF = HB, (84) reduces to: f ≡ ≈ + =
F λ λ λF
(87) is a straight line (relating the two unknowns xB and yD) which can now be
constructed on the xy diagram. Knowing the value of f we first calculate the slope:
f −1
slope = <0
f
We just need one point to finish the construction. Of course, we could use the y-intercept: yD =
zF/f = xF/f for xB = 0. But this value is frequently off-scale (e.g. yD = ∞ for f = 0 and xB = 0). A
better choice for an arbitrary point is to substitute xB = xF. Then (87) yields
f −1 xF f −1 + 1
=
yD xF + = xF = xF
f f f
Thus one point on the line given by (87) is
one point: (xB, yD) = (xF, xF)
Knowing one point and the slope, we can draw the

line. In effect, this line represents the relationship
between yD and xB which is imposed by the mass and
energy balances for a given Q (or a given f ) and for
the given assumptions on calculating enthalpy. We
call this line the

06-361 page 96 Spring, 2017
operating line: relationship between yD and xB which is imposed by the mass and
energy balances
To find which point on this line is the solution for a given Q, we need also to satisfy (83), which
represents phase equilibrium. This relationship is a curve on the xy diagram, known as the
equilibrium curve: relationship between yD and xB which is imposed by vapor-

liquid equilibria
The intersection of operating line and the equilibrium curve satisfies all four of the original
equations and yields the answer. Note that the vapor mole fraction yD actually exceeds the feed
xF which in turn exceeds the liquid product mole fraction xB:
yD > xF > xB
So we have achieved some degree of separation of the two components in the feed.
In the previous example we assumed the feed was only liquid and we added heat to create a
vapor. Suppose instead, the feed is only vapor and we remove heat (i.e. Q < 0).
Example #2: vapor feed
Given: F, zF, HF, Q<0, xy data and λ
Approx. Solution: In terms of the enthalpies, the only differences compared to Example #1 are
1) that Q<0 (we are removing heat) and 2) the feed is now a vapor so
HF = λ
D HF − HB Q F λ Q F Q
(84) gives: = + =+ =
1+ <1
F HD − HB HD − HB λ λ Fλ
which will be <1 for Q<0. But if we stick with the original definition of f as
D Q
f ≡ =+
1 <1
F Fλ
then (85) – (87) still hold. In particular, the construction of the operating line on the xy-diagram
is the same as in the previous example. Its intersection with the equilibrium curve gives the
answer to this flash problem.
Lever-Arm Rule on Txy Diagram
There is also a simple graphical procedure for estimating the fraction vaporized on a Txy

06-361 page 97 Spring, 2017
diagram. Suppose you are given a feed composition and (instead of specifying the fraction
evaporated) the final temperature of the mixture at equilibrium.
Example #3: use of Txy diagram
Given: zF, T and the Txy diagram
Find: yD, xB and f
Solution: Locate the point (zF, T) on the Txy diagram. A

horizontal line through this point intersects the vapor and
liquid curves at the composition of those phases: (xB, T) and
(yD, T).
As it turns out (see “Proof” below), the amount of vapor

produced is inversely proportional to the distance between
the feed point (xF, T) and the vapor point (yD, T). Similarly,
the amount of liquid remaining is inversely proportional to
the distance between the feed point (xF, T) and the liquid
point (xB, T). More precisely, the ratio of vapor D to feed F flowrates in an continuous flash unit
is given by (see “Proof” below):
D z F − xB B yD − z F
=
f = and = (88)
F y D − xB F y D − xB
where D, B and F are proportional to the distances indicated on the Txy diagram above. These
relations are another example of the lever-arm rule (see page 54).
Proof:
D z F − xB
Solving (86) for f: f ≡ = (89)
F y D − xB
which confirms the first equation in (88). Substituting (89) into (85):
B D z − xB
=1 − =1 − F =
y D − xB − z F − xB( ) = yD − z F (90)
F F y D − xB y D − xB y D − xB
which confirms the second equation in (88).
In light of the approximations made in Examples 1&2, we need to emphasize that no

approximations have been made in the solution of Example 3. The approximations in Examples
1&2 dealt with the evaluation of enthalpy. Enthalpies were not needed in Example 3. Instead,
by specifying the flash temperature and by having the Txy diagram, we could immediately

06-361 page 98 Spring, 2017
determine the corresponding point (xB, yD) on the xy-diagram; we didn’t need to evaluate f first.
The downside of this method of solution is that I don’t know what Q is required; to estimate Q, I
need to evaluate enthalpies and then I am forced to make approximations like those in Examples
1&2, unless I have full enthalpy data (see next section).
RIGOROUS SOLUTION
Our solution to Example #1 (the interaction of the operating line and the equilibrium curve in
Error! Reference source not found. on page Error! Bookmark not defined.) was approximate
because of the two assumptions we made about the enthalpies. Although this approximate
solution works reasonable well for some mixtures, for others it might not be good enough. Then
you will need a lot more thermodynamic data besides latent heat and heat capacity: you will need
an Hxy diagram.
An Hxy diagram is at least superficially similar

to a Txy diagram in that it gives you some idea of
the thermal condition of the mixture as a function
of composition, except that instead of temperature
T on the y-axis, we have the molar enthalpy H. The
two red curves in the figure at right represent the
enthalpies of saturated liquid HB(xB) (lower curve)
and saturated vapor HD(yD) as a function of the
composition of those phases.
Unlike a Txy diagram, the saturated liquid and

saturated vapor curves do not connect at the ends or
at any other composition. The vertical distance
between these two curves represents the latent heat
of vaporization at a particular composition, which is usually substantial for all compositions.
Unlike temperature, enthalpy of liquid and vapor will be quite different, so the tie lines are not
horizontal like on a Txy diagram. In fact, tie lines tend to be more nearly vertical.
As we shall soon see, we will need to determine the tie line through a particular point. No matter
how many tie lines your graph has, it would be a real stroke of luck if one passes through the
particular point specified. So we need some way to interpolate tie lines. This is accomplished
by means of:
auxiliary line -- a curve on an Hxy diagram for interpolating tie lines
If we form a right triangle out of each tie line, using the tie line as the hypotenuse, then the locus
of the points where the right angle is located is the auxiliary line. Of course, the ends of the
auxiliary line coincides with the ends of the saturated vapor line. By reversing this procedure,
we can locate more tie lines.
Now let's see how this diagram can be use do solve our flash distillation problems. First, I will
give you the recipe and then later I will show why the recipe works.

06-361 page 99 Spring, 2017
Example #4:
Given: zF, HF, Q/F and Hxy diagram
Find: xB, yD and D/F
Solution (recipe):
• locate the feed point (zF, HF)
• locate a second point Q/F above the feed

point (zF,HF+Q/F). We will call this the
“heated feed point”
• find the tie line which passes through this

second point
• ends of the tie line are (xB,HB) and (yD,HD).
• f can be determined with the help of (89):

z F − xB D
= f =
y D − xB F
Proof: now let's show that this procedure really works.
The main idea which we need to prove is that the point (zF, HF+Q/F) [call this point “F”] lies on
the tie line. We know that (xB, HB) [call this point “B”] and (xD, HD) [call this point “D”] lie on
the tie line since the two streams leaving an equilibrium flash unit are assumed to be in phase
equilibrium with each other. Our approach will be show that the slope of the line drawn from B
to F is the same as the slope of the tie line drawn from F to D.
Substituting (80) into (81): zF(D+B) = yDD + xBB
B yD − z F
Dividing by D and solve for B/D: = (91)
D z F − xB
Q D B
Divide (82) by F: HF + = HD + HB
F F F
Using (80) to eliminate D:
Q  B B B
HF + = 1 −  H D + H B = H D − ( H D − H B )
F  F F F

06-361 page 100 Spring, 2017
 Q
HD −  HF + 
Solving for B/F:
B
=  F
(92)
F HD − HB
If instead, we had used (80) to eliminate B:
 Q
 HF +  − HB
D  F
= (93)
F HD − HB
 Q
HD −  HF + 
(92) ÷ (93):
B
=  F
(94)
D  q
 HF +  − HB
 F
Equating (91) and (94), we could show that:
 Q  Q
HD −  HF +   HF +  − HB
 F
=  F
yD − z F xF − xB
Note that both sides of this equation have the form of rise/run on an Hxy diagram; thus each side
represents the slope of a line drawn between two points on an Hxy diagram. These two slopes
involve three distinct points:
(xB,HB); (zF,HF+Q/F); and (yD,HD)
The equality of these two slopes means that these three points must lie on the same line. That
line connecting B and D must be a tie line. The conclusion is that
(zF,HF+Q/F) lies on tie line
This is the basis for the recipe.

06-361 page 101 Spring, 2017
Lever-Arm Rule on Hxy Diagram
Equations (91) - (94) provide a basis for a lever-

arm rule which is similar to (88). The total difference
in enthalpy between saturated vapor and saturated
liquid is HD – HB.♦ This difference can be partition-
ed into (HF+Q/F) – HB and HD – (HF+Q/F). Thus the
red tie line can be divided into two parts which are
proportional to this difference in enthalpies. (94) says
that ratio of the lengths of these two parts of the tie
line equal the ratio of molar flowrates B/D.
Chapter 8. Multistage Operations

In both batch distillation and flash distillation, the separation of the two components is
limited by equilibrium. For example, suppose I have a 50-50 mixture of hexane and octane. If I
vaporize half of this mixture, I produce a vapor having:
xF = 0.5, f = D/F = 0.5 → yD = 0.57, xB = 0.43
which we read off of an xy diagram. Suppose you want a

95% pure product? If I vaporized a smaller fraction of the
feed, that would increase the composition of the distillate, but
the best I can do is:
xF = 0.5, f ≈ 0 → yD = 0.63, xB = 0.5
This helps a little but it's still a long way from 0.95 and you
don't generate much product with such a small fraction
vaporized.
One solution is to take the distillate product of the first flash and partially condense it. This
will produce a liquid phase which will be leaner in the more volatile component, leaving the
remaining vapor richer:
♦ H -H is not the “heat of vaporization” (since the composition of these two streams is different). Instead heat of
D B
vaporization is the heat to totally vaporize one mole of feed. On the Hxy diagram, heat of vaporization is the length
of a vertical line between the two curves.

06-361 page 102 Spring, 2017
In the schematic above, we have repeated this step N-1 times which is required to obtain
some arbitrary enrichment of the more volatile component. We have also separated the heat
exchanger from the vapor space of the flash (for reasons which will become apparent in the next
section).
Suppose we design the heating (or cooling) rates such that in

each flash unit, the distillate product rate is 50% of the feed to that
unit. Then each successive flash will produce a distillate product
which is richer in the more volatile component than the feed. In the
xy-diagram at right, we have stepped off the distillate and bottoms
compositions for several stages.
Although we can produce a stream of the desired composition

(e.g. yN = 0.95), it turns out that we are "throwing out" a lot of the
desired component in the "waste" streams -- the bottoms or Bis.
From the first flash unit alone, we are losing 25% of component A
in the initial feed [(1-f)x1 ≈ 0.25]. If this component A is valuable, we don't want to throw it
away.
• good news: success at achieving high yD

1
• bad news: low recovery♣ (recovery ≈ )
2N
• more bad news: requires N+1 heat exchangers
Another problem is that each flash unit requires a heat exchanger and then we will need one
additional heat exchanger to condense the product from the final flash unit.
♣ “Recovery” is fraction of component A in feed which ends up in final product stream.

06-361 page 103 Spring, 2017
COUNTER-CURRENT CASCADE
Both of these drawbacks (very low recovery and very high heat-exchanger costs) can be
removed by a slight modification: we recycle the bottoms of one flash unit into the feed of the
previous flash unit.
To condense the vapor arriving from the i-1th flash unit, we need to transfer some of its latent
heat to the liquid recycled from the i+1th flash unit. For this to work, the temperature of the
vapor must be greater than that of the liquid. Let’s see if this inequality is satisfied: generally,
the vapor will get richer in the more volatile component as we move left to right:
y1 < y2 < ... < yN
Assuming that the vapor and liquid streams leaving each flash are
in equilibrium, then the corresponding inequalities for tempera-
ture can be deduced from the typical Txy diagram at right:
T1 > T2 > ... > TN
So the recycled liquid will be cooler than the vapor it's mixed
with. Thus the liquid will absorb some of the heat of the vapor. This condenses some of the
vapor and vaporizes some of the liquid.
Although they are not shown in the schematic of the counter-current cascade shown above,
we still need two heat-exchangers: 1) to vaporize a liquid stream to produce the feed stream
labelled V0 and 2) to partial condense VN to generate the stream labelled LN+1. Now that we
have shown that most of the heat exchangers might be eliminated, let’s return to a discussion of
the countercurrent cascade.
The product recovery for a countercurrent cascade is much better than for the series of flash
units without recycle. We have only one waste stream (the bottoms product from the first unit,
labelled L1) -- instead of one from each flash.
stage -- one unit of a cascade (e.g. one flash unit)
ideal stage -- a stage whose exit streams are at equilibrium

06-361 page 104 Spring, 2017
Chapter 9. Equipment for Tray Towers

In our schematic of a countercurrent
cascade, we had the streams moving
horizontally. Generally, a pump would
then be required between each stage to
overcome friction. With vapor and liquid,
the density difference is sufficiently large,
that we can use gravity to move the fluids
if we stack the stages vertically rather than
horizontally. A vertical stack of stages is
called a tray tower.
At right we see a schematic of a typical

tray tower. Each stage or flash unit is
represented by one tray. Five trays are
evident in Fig. 2. Liquid flows down from
the tray above and flows across the tray
where in comes into contact with vapor
flowing upward from the tray below.
After mixing, the liquid and vapor are
separated, with the new liquid flowing
down to the tray below and the vapor
flowing up to the tray above. Through the
act of mixing, the composition of the
liquid and vapor streams has changed.
Fig. 3. An expanded view of a single

“bubble cap”. Taken from Fig. 6.6 of
Treybal.
♦R.E. Treybal, Mass-Transfer Operations, 2nd Ed.,

Fig. 2. A typical tray tower. Taken from Fig. 6.2 of
McGraw-Hill, 1968.
Treybal.♦

06-361 page 105 Spring, 2017
The contacting of vapor and liquid on any given tray is facilitated by bubble caps, one of
which is shown in more detail in Fig. 3. Each tray contains many bubble caps arranged in a
array covering the central region of the tray (see Fig. 4).
Fig. 4. Top view of arrangement of bubble caps on a tray. Taken from Fig. 6.6 of Treybal.
In the particular design shown in Fig. 3, vapor penetrates the tray through a circular hole 2.47
inches in diameter, upward through a 3-inch length of pipe before being diverted downward by
the cap and forced through the liquid outside the pipe.
Fig. 5. Flow patterns under normal and abnormal operating of a bubble cap. Taken from Fig. 6.7 of
Treybal.
The flow pattern under both normal and abnormal operation is shown in Fig. 5.

06-361 page 106 Spring, 2017
An alternative tray design is the sieve-plate tray (see Fig. 6). Sieve-plate trays are much
cheaper to fabricate compared to bubble-cap trays since the tray consists of a circular plate with
holes drilled through for the passage of vapor. These holes are generally much smaller in
diameter than the bubble caps.
Fig. 6. Side view of a “sieve plate” tray. Taken from Fig. 6.8 of Treybal.
A disadvantage of sieve plates is that they are much more prone to weeping and flooding
(conditions corresponding to poor contact of liquid and vapor). “Weeping” is the leaking of
liquid down through the small holes of the sieve plate. “Flooding” is the failure of the gas phase
to flow through the small holes; instead gas flow may stop completely.
To convey some idea of the size of distillation tower, see the photo on the next page. This
photo is of a DuPont plant (in Texas) which purifies ethylene. Ethylene is used (for example) in
the production of polyethylene sheet.
The photo shows six separate towers: five are distillation columns and one is an gas-
absorption column (we will cover gas absorption in Chapter 13). The five distillation towers
range in height from 67-197 feet and are 3 or 4 feet in diameter.

06-361 page 107 Spring, 2017

06-361 page 108 Spring, 2017

06-361 page 109 Spring, 2017
Chapter 10. Performance of a C-C Cascade
ANALYSIS OF A C-C CASCADE

At right is a schematic representation of a countercurrent
cascade. It ignores some of the details of plumbing inside the
stages and, of course, in reality there are no pipes to carry the
liquid and vapor between the adjacent stages. But these are
unimportant details as far as mass and energy balances are
concerned. The schematic preserves the essential feature -- which
is that the flow of liquid and vapor occurs counter-currently.
Since we have so many streams to deal with, some convention

for numbering the stages and labelling the streams will be helpful.
Geankoplis uses the following conventions:
tray numbers: increase in direction of liquid flow (1 ≥ i ≥ N)
stream labels: subscripts denote stage from which the stream is

leaving
• Li, Vi are molar flowrates of total (A+B) liquid or total vapor leaving the ith stage.
• L0 and VN+1 are the feed flows at the top and bottom of the cascade
• xi, yi are the mole fraction of the more volatile component in the liquid and vapor leaving the
ith stage.
Suppose we know everything about the two input streams: their flowrate, composition, and
enthalpy. Suppose we also have all the thermodynamic data: Hxy diagram and xy diagram.
Given: L0, x0, VN+1, yN+1, Hxy and xy diagrams
We want to determine the flowrate, composition and enthalpy of the two output streams and of
all the intermediate streams:
Find: Li, xi, Vi, yi for i=1,...,N ( = 4N unknowns)
I haven't bothered to list the enthalpies or temperatures as unknowns, because if I know the mole
fraction and pressure, I can look up the enthalpy and temperature of saturated liquid or vapor.
As our system, let's select the top "i" stages. An overall mole balance for this system
requires:

06-361 page 110 Spring, 2017
total: L0 + Vi+1 = Li + V1 for i=1,...,N (95)
while a mole balance on the more volatile component requires:
component A: L0x0 + Vi+1yi+1 = Lixi + V1y1 for i=1,...,N (96)
Finally, we can also perform an enthalpy balance:
enthalpy: L0Hx,0 + Vi+1Hy,i+1 = LiHx,i + V1Hy,1 for i=1,...,N (97)
Do I have enough equations to calculate the 4N unknowns? How many independent equations
do I have?
equations: (95), (96), (97) × N ( = 3N eqns)
The additional equation is obtained by assuming that the two streams leaving any given stage are
in thermodynamic equilibrium with each other (same temperature and same chemical potential
for each component). When this assumption applies, we say the stage is ideal. Mathematically,
this means
(xi, yi) lies on the equilibrium curve of xy-diagram ( = N eqns)
So, at least when I am dealing with ideal stages, I have now as many equations as unknowns.
Many distillation towers have N=20 stages or more, so I could easily be talking about 80
equations -- a formidable problem. Fortunately, there are assumptions which can make this
problem tractable so that it can be solved easily with pencil and paper.
CONSTANT MOLAL OVERFLOW

In our discussion of flash distillation, we noted that many pure
components have the same heat of vaporization per mole. If you
can assume that the molar heat of vaporization of the mixture is
completely independent of its composition, then the solution is
greatly simplified.
Assumptions:
1) Hy,i – Hx,i = λ for all i.

2) neglect sensible heat compared to latent heat
Let's choose saturated liquid having the composition of the feed as

our reference state for enthalpies. Since the only difference
between the various liquid streams is sensible heat, we take them
all to be equal -- equal to zero (our reference state for enthalpy):

06-361 page 111 Spring, 2017
Hx,i = 0 and Hy,i = λ for all i
Substituting these enthalpy values, (97) simplifies to:
Vi+1 = VN+1 for all i (98)
Substituting this result into the total mole balance about the top i stages (95) yields:
L0 = Li for all i
In other words, all of the vapor streams have the same molar flowrate VN+1 within the cascade
(although their composition and mass flowrate will change) and all the liquid streams have the
same molar flowrate L0:
V1 = V2 = ... = VN = VN+1 ( = V, say)
L0 = L1 = L2 = ... = LN ( = L, say)
This state is called constant molal overflow. Since the input flowrates L0 and VN+1 are known,
we know all the flowrates. Thus we have just halved the number of unknowns from 4N to 2N.
We can then drop the subscripts on the flowrates and forget the enthalpy balance (97) and the
total mole balance (95). This leaves just the component mole balance (96):
Lx0 + Vyi+1 = Lxi + Vy1.
L  L 
which can be rearranged to: yi +1 = xi +  y1 − x0  (99)

V V 
m b
This is a straight line on an xy diagram and is called the:
operating line — yi+1 vs. xi; a graphical representation of the mass and energy balances.
Another important curve on an xy diagram is the:
equilibrium curve — yi vs. xi; a graphical representation of VLE
In the next section, we will see how these two curves can be used to determine the number of
equilibrium stages required to achieve some desired degree of separation.

06-361 page 112 Spring, 2017
A RECTIFYING (ENRICHING) CASCADE

In a typical distillation tower, the liquid fed to
the top of the enriching cascade is a portion of the
vapor product which has been condensed and
refluxed back into the tower, as shown at right.
The ratio of the amount refluxed to the amount not
refluxed is called the reflux ratio
L
R≡ (100)
D
When all of the vapor stream leaving the top of the

cascade is condensed (as shown in the figure at
right), the three streams have the same
composition:
xD = x0 = y1 (101)
Then the operating line (99) can be simplified to♠
L  L
yi +1= xi + 1 −  xD (102)
V  V
The ratio of the flowrates can be expressed in terms of the reflux ratio R using (100):
L L RD R
= = = (103)
V L + D RD + D R + 1
L R R +1− R 1
1− =1 − = =
V R +1 R +1 R +1
Substituting these relations into (102), the operating line becomes
R 1
=
yi +1 xi + xD (104)
R +1
 +
R
 1
slope intercept
Thus the slope of the operating line (a plot of yi+1 versus xi) is R/(R+1). One point on the
operating line is xi = yi+1 = xD. In other words, if we substitute xD for both xi and yi+1, (104) is
automatically satisfied:
♠ Alternatively, we could “start over” and perform a component mole balance on the top i stages of the cascade
(indicated by the blue line in the sketch).

06-361 page 113 Spring, 2017
one point: (xD, xD)
R 1
=
yi +1 xD + xD
R +1 R +1
Proof:
R +1
= xD = xD
R +1
Armed with the slope and one point we could plot the operating
line on an xy diagram (see figure at right).
However, it is usually easier to plot a line if you are given

two points. A second point is the y-intercept. Substituting xi
= 0 into (104):
 xD 
second point:  0, 
 R +1
R 1 xD
Proof: y ( 0=
) ×0+ x=
D
R +1 R +1 R +1
See figure at left.
EQUATION SECTION 1Lecture #11 begins here
Example #1. Suppose I have a feed stream containing 50% alcohol in water which I want to
enrich to 95% alcohol. How many stages do I need if I use a reflux ratio of unity and what will
be the flowrate of this product?
Given: xF = 0.5, xD = 0.95, F and R = 1
Find: N and D
Solution: The first step is to draw the operating line on the xy-diagram. One point is (xD,xD) or
 x 
(0.95, 0.95). The second point is  0, D  or (0, 0.475). Knowing two points, we can draw the
 R +1
operating line (as in figure above left).* Once the operating line is plotted on the xy diagram
along with the equilibrium curve, we can determine the composition of other streams as follows.
*The operating line is not drawn to scale in these sketches. For that matter, neither is the equilibrium curve.
However, given the equilibrium curve and operating line as drawn, the method of stepping off the number of stages
is correct.

06-361 page 114 Spring, 2017
The two streams leaving a given ideal stage are at

equilibrium, thus the compositions of the two streams leaving a
given stage must lie on the EC. Since we already know the
composition of the vapor stream leaving stage 1 (y1 = xD =
0.95), we can determine x1 by moving horizontally from one
point on the operating line (xD, xD) to one point on the
equilibrium curve (EC) and then read off the value of x1:
(x1, y1 = xD) on EC: gives x1
On the other hand, the operating line (OL) represents a

component mole balance about an arbitrary number of stages
measured from the top. In particular, for the top stage, the
operating line (102) gives me a relationship between x1 and y2:
(x1, y2) on OL: gives y2
Knowing x1, I can determine y2. In general, the compositions of

the two streams between a given pair of adjacent stages must lie
on the OL. Now I can repeat the process:
(x2,y2) on EC: gives x2
(x2,y3) on OL: gives y3
and so on. Continuing to step

off stages in this manner until
the vapor mole fraction
equals or drops below xF,
requires four steps:
answer: N=4
Comment #1: Although the feed might be a liquid, it’s the

vapor mole fraction which must be at or below the feed
composition, because the actual feed to the bottom of the
cascade is a vapor (see stream whose composition is labelled
yN+1 in the flowsheet on page 110). The horizontal line y = xF
(= yN+1) on the xy-diagram is called the feed line.
Comment #2: After taking 4 steps (or any integer number), we don’t end up exactly at the feed
composition (except by some extreme stroke of luck). Instead we end up stepping below the
feed composition. If the trays really are perfectly ideal, and the feed composition is 0.5, the
distillate composition would turn out to be slightly richer than 0.95. Although any real column
has an integer number of trays, it is customary to report the answer as a fraction; for example, to

06-361 page 115 Spring, 2017
go from a feed of 0.50 to a distillate of 0.95, we require 3.7 ideal stages (or 3 + 0.7, where 0.7 is
the fraction of the last step which corresponds to the feed composition) in the current problem.
Comment #3: Suppose that our feed composition had been a little lower, say yb'. No matter how
many steps we took, we could never get a vapor composition lower than that at the intersection
of the operating and equilibrium curves. In that case, any solution for the specified reflux ratio is
impossible. Instead the reflux ratio R needs to be increased, which will rotate the operating line
counter-clockwise about the point (xD, xD). This will lower the composition at the intersection
point and make the separation possible. It turns out there exists a minimum value of the reflux
ratio which allows a particular distillate composition xD to be produced from a given feed
composition xF. This called the minimum reflux ratio.
Let’s finish this problem by calculating the flowrates. Since the vapor flowrate leaving each
tray is the same, and because we are creating the vapor to the first tray by completely vaporizing
the feed, we can say that
V=F
The ratio of liquid to vapor flowrate inside the cascade is given in terms of the known reflux
ratio by (103):
L R R =1 1 1 1
= = or =
L = V F
V R +1 2 2 2
Since the bottoms stream B is one of the liquid streams
1
B= L= F
2
Finally, the distillate flowrate can be calculated from the definition of the reflux ratio: recall
(100):
L R =1 1
=
D = = L F
R 2
A STRIPPING CASCADE
Use of a single cascade with 3.7 stages allows us to
convert a feed of 50mol% alcohol into one containing
95mol%. But a significant amount of the alcohol is being lost
in the second product stream. It’s possible to strip the alcohol
from this stream by feeding it to a second cascade.
In a stripping cascade, liquid is fed into the top. Liquid out

the bottom of the cascade is partially vaporized and the vapor

06-361 page 116 Spring, 2017
is recycled into the bottom of the cascade. The vapor stream at the top represents the second
product stream. The analog of the reflux ratio for a stripping cascade is the boil-up ratio
V
RB ≡
B
Because the bottoms product is only partially vaporized to form the vapor stream which is
recycled to the cascade, its composition is not the same as the bottoms product. In other words
xN ≠ yN+1 ≠ xB
So no analogy can be made to (101). However we still have a operating-line relationship like
(99) or (102) which is imposed by mole balances. From a balance over the entire stripping
cascade on the more volatile component, we know that
=
Fx = BxB + Vy1
Lx
F 0 
in out
so Vy1 – Lx0 = –BxB
Substituting this result into (99), the operating line for a stripping cascade becomes♠
L B
y=
i +1 xi − xB (105)
V V
The ratio of the flowrates can be expressed in terms of the boil-up ratio:
B B 1
= =
V RB B RB
L V + B RB B + B RB + 1
= = =
V V RB B RB
RB + 1 x
The operating line (105) becomes:=
yi +1 xi − B (106)
RB RB

slope
Like all the operating lines we’ve encountered so far, this one
has a slope of
♠ Alternatively, we could “start over” and perform a component mole

balance on the bottom N-i stages of the cascade (indicated by the blue line in
the sketch).

06-361 page 117 Spring, 2017
L RB + 1
slope: = >1
V RB
which is always greater than unity for a stripping cascade. One point on this operating line is
one point: (xB, xB)
Proof: substituting yi+1 = xi = xB automatically satisfies (106). Having one point and the slope,
we can draw the operating line. Alternatively, a second point is
 RB + xB 
second point:  R + 1 ,1
 B 
 R +1 xB
Proof: Substituting y = 1 into =
(106): 1  B x−
 RB  RB
RB  x  RB  RB + xB  RB + xB
= 1 + B=  = R +1
 
RB + 1  RB  RB + 1  RB
then solving for x: x
 B
Example #2. Suppose I have a feed stream containing 50% alcohol in water which I want to
strip the alcohol from to produce a stream containing only 5% alcohol. How many stages do I
need if I boil half of the liquid leaving the bottom of the cascade and return it as vapor?
Given: xF, xB and RB = 1
Find: N
Solution: Once again, vapor and liquid streams leaving the same stage are assumed to at
equilibrium; thus (xi,yi) lies on the equilibrium curve. On the other hand, the composition of
adjacent streams (xi,yi+1) lies on an operating line for this cascade. One point on this operating
line is xi = yi+1 = xB. A second point is:
 RB + xB   1 + 0.05 
= ,1 = ,1 ( 0.525,1)
 RB + 1   1 + 1 
Knowing two points, we can immediately draw the operating

line as shown in Error! Reference source not found. or at
right.♣ The liquid product being drawn off at xB is in
equilibrium with the vapor being sent back into the cascade;
thus (xB, yN+1) lies on the equilibrium curve:
♣ Clearly the figure at right is not drawn to scale (since the y=1 intercept is located well above 0.5 on the x-axis).

06-361 page 118 Spring, 2017
on EC: (xB, yN+1)
On the other hand, (xN, yN+1) are adjacent streams in the cascade and so their compositions must
lie on the operating line:
on OL: (xN, yN+1)
Similarly, (xN, yN) lies on the equilibrium curve, while (xN-1, yN) lies on the operating line.
We continue to step off stages in this manner until we step past a liquid composition equal to
the feed xF; in other words, we continue stepping until we cross the blue vertical feed line in
Error! Reference source not found.. In this case, three steps are required (actually only 2.3).
But one of these steps is the reboiler; only two additional equilibrium trays are required:
answer: N = 1.3 plus reboiler
Comment #1: Notice that the “feed line” for a stripping cascade is vertical, whereas the feed line
for a rectifying cascade was horizontal (see page 114). This is because the feed into the bottom
of a rectifying cascade is a vapor, whereas the feed to a stripping cascade is a liquid.
Comment #2: Suppose that our feed composition had been a little higher, say x'F, which is above
the intersection of the operating and equilibrium curves. No matter how many steps we took, we
could never get a liquid composition higher than that at the intersection of the operating and
equilibrium curves. In that case, any solution for the specified boil-up ratio RB is impossible.
Instead RB would have to be increased, which will rotate the operating line clockwise about
point (xB, xB). This will raise the liquid composition at the intersection point and make the
separation possible. It turns out that, just like the minimum reflux ratio for a rectifying cascade,
there exists a minimum boil-up ratio RB which allows a particular bottoms composition xB to be
produced from a given feed composition xF.
Let's finish the problem by determining the flowrates. Once again, all the flowrates can be
deduced from the equimolal overflow♠ assumption and the boil-up ratio. All the liquid flowrates
correspond to the flowrate of the liquid fed to the top:
L=F
Since I’m boiling half of this liquid to return to the cascade as vapor, all the vapor flowrates are
V = L/2 = F/2
The bottoms flowrate is what remains after half of the liquid to the reboiler is boiled:
B = L – V = F/2
♠ “Equimolal overflow” is another name for the “constant molal overflow” approximation: see page 110.

06-361 page 119 Spring, 2017
STEAM DISTILLATION
Instead of using a heat exchanger to partially boil the liquid
stream leaving the bottom stage to produce the vapor entering the
bottom stage, we could directly inject steam as the vapor. This has
the advantage that it avoids the cost of a heat exchanger and you are
getting direct contact between the hot and cold fluids.
Of course, use of steam as the vapor is only useful if water is one

of the two components being separated (e.g. alcohol/water) and
further, it must be the less volatile component. If instead you are
trying to separate heptane from octane, injecting steam would be
produce a second liquid phase (since oil and water are not miscible).
This produces the operating line shown at right. Note that

one point on the operating line is (xB, 0) since the “vapor”
entering the bottom is pure steam (water). Of course, part of the
operating line lies below the 45° line (x=y). But there are no
restrictions that y must be larger than x. As with the stripping
cascade, we continue to step off stages between the operating line
and the equilibrium curve until our liquid composition exceeds
xF.
Comment: Suppose that our feed composition had been a little

higher, say x'F. No matter how many steps we took, we could
never get a liquid composition higher than that at the intersection of the operating and
equilibrium curves. In that case, any solution for the specified L/V ratio is impossible. Instead
the L/V ratio would have to be decreased, which will rotate the operating line clockwise about
the specified point. This will raise the composition at the intersection point and make the
separation possible. It turns out there exists maximum L/V, or a minimum steam rate V which
allows a particular bottoms composition xB to be produced from a given feed composition xF and
liquid flowrate L.

06-361 page 120 Spring, 2017
Chapter 11. McCabe-Thiele Method

A typical distillation tower will consist of at least two
cascades -- one to enrich the feed in the more volatile
component and a second one to strip the more volatile
component out of the bottoms stream before disposing of
it.
rectification -- enrichment of vapor by contact with liquid

reflux
With enough trays, we can usually make D as pure as we

like (an exception occurs for an azeotropic mixture).
However, the liquid stream at the bottom of the cascade
might still contain a significant fraction of the more
volatile component. This is where a second cascade is
useful:
stripping -- removal of the more volatile component from

a liquid by contact with reboiled bottoms (or direct
injection of steam)
A typical problem in designing a distillation column is the

following:
Given: zF, xD, xB, R and HF
Find: N and N
where N and N are the number of stages in enriching and stripping sections.
Solution: Having two cascades is a little trickier than one, so let me first tell you how to solve
the problem by giving you the recipe, and later we will rationalize some of the steps in this
recipe.
This recipe is called the McCabe-Thiele Method and it works when you have only two
components and you can assume equal molal overflow.
Step 1) Plot equilibrium data on an xy diagram.
Step 2) Locate xB, zF, and xD on 45° line
We use “zF” to denote the mole fraction of the more volatile component (A) in the feed, ignoring
which phase the feed is in:

06-361 page 121 Spring, 2017
moles of A in liquid + vapor

zF ≡
total moles in liquid + vapor
Step 3) Calculate q (a dimensionless variable representing

the thermal condition of the feed):
HV − H F
q≡
HV − H L
where HF = enthalpy of feed
HV = H of sat'd vapor with y=zF
HL = H of sat'd liq. with x=zF
Step 4) Plot q-line, which is a line given by the following equation (QL):
 q  zF
=y  x+ (107)
 q −1 1− q
one point: (zF, zF)
q
slope:
q −1
which is zero (horizontal) for a saturated vapor (q = 0) or infinite (vertical) for a saturated liquid
(q = 1). Otherwise, it’s easier to plot two points. A second point is where the q-line crosses the
y-axis (i.e. at x=0). Substituting x=0 into (107), we calculate the y-coordinate of the second
point:
2nd point: (0, ) zF

1− q
Should this y-coordinate be larger than one or less than zero, an alternative second point is where
the q-line crosses the y=1 line. Substituting y=1 into (107), we solve for the x-coordinate of the
alternative second point:
3rd point: ( q −1+ z F

q )
,1
Of course, the name "q-line" is given because this line depends on q — the thermal condition
of the feed. For example, suppose the feed is ...

06-361 page 122 Spring, 2017
q
superheated vapor: HF>HV → q<0 → 0< <1
q −1
sat'd vapor:
q
HF=HV → q=0 → = 0 (horizontal line)
q −1
2-phase mixture:
q
HL<HF<HV → 0<q<1 → −∞ < <0
q −1
Figure 11: an xy diagram
In this case, q represents the fraction of the feed which is liquid. showing q-lines for feeds having
different thermal conditions (i.e.
q different HF).
sat'd liquid: HF=HL → q=1 → = ∞ (vertical line)
q −1
q
subcooled liquid: HF<HL → q>1 → 1< < +∞
q −1
As it turns out, the two operating lines for the two cascades
intersect somewhere along this line: for a proof, see Role of q-
Line below on page 137.
Stop 5) Plot the operating line for the rectification section

(ROL):
R x
=y x+ D (108)
R +1
 R +1
L
V
one point: (xD, xD)
 xD 
another point:  0, 
 R +1
Step 6) Plot the operating line for the stripping section (SOL):
RB + 1 x
=y x− B (109)
RB RB

L
V

06-361 page 123 Spring, 2017
one point: (xB, xB)
another point: intersection of ROL and QL
Comment #1: equations (108) and (109) for the two operating lines share a common feature also
found in the generic form of the operating line given by (99): the slope of any operating line is
the ratio of liquid to vapor flowrates in that cascade: L V for the rectifying cascade and L V for
the stripping cascade.
Comment #2: The ROL was plotted in Step 5) using the same two points used in Example #1
(for an isolated rectifying cascade on page 113). But for the SOL, we used only one of the two
points used in Example #2 (for an isolated stripping cascade on page 117): namely (xB, xB).
Instead of using the second point in Example #2, Step 6) of the recipe above used the
intersection of the QL and the ROL.♦ While the second point used in Example #2 is also
correct here, it would have required us to determine the boil-up ratio RB. The procedure
suggested in this recipe avoids this determination.
Note that the boil-up ratio RB and the reflux ratio R cannot both be specified in the problem
statement (we specified only R). Instead these two values are related by the constraint that the
ROL and SOL must intersection along the QL.
Step 7) Step off stages
You could start at either end (x=y=xD or x=y=xB), although it is customary to start at the top.
LOCATION OF FEED PLATE

At some point during the stepping process (starting from the top at x=y=xD), you will have a
choice of either operating line to step down to. Either operating line can be used, but once you
switch operating lines, you cannot go back to the original line: switching operating lines
corresponds to switching from one cascade to the other. The stage at which we switch operating
lines is the
feed plate: stage at which you switch operating lines
♦ A proof that ROL and SOL intersect along the QL is offered below in Role of q-Line on page 136.

06-361 page 124 Spring, 2017
Either of the above two stepping schemes is OK, but one scheme leads to fewer total trays than
the other.
optimum feed plate: to get the fewest number of total stages, you switch operating lines when
that allows you to take larger steps
In other words, if you have a choice of which operating line to use, use that one which takes you
furthest from the equilibrium curve. This choice gives the biggest step and fewest total stages.
CONVERGENCE OF OPERATING LINE AND EQUILIBRIUM CURVE

In some of the homework problems, the distance between the operating line and equilibrium
curve becomes very small near the point of their intersection. When this happens, a very large
number of steps might be required to make any progress on separation.
Rule: an infinite number of steps is required to get to an intersection of the operating line and
equilibrium curve.
Proof: when the size of your steps are no longer large compared to the width of the pen or pencil
you are using to draw them, you should "blow up" the drawing to reduce drawing errors.
Figure 12
To demonstrate the infinity of steps required, let me present the results of stepping off stages in a
particular example:

06-361 page 125 Spring, 2017
Example: given the xy-diagram for acetone and methanol at 1 atm pressure (has an azeotrope at
x = y = 0.8000), step off stages for the following conditions:
Given: zF = 0.25, xD = 0.7999, xB = 0.05, q = 0.3, R = 8.618
Find: N and N for the optimum feed location
Solution: This is a particularly difficult separation because the desired distillation composition
xD = 0.7999 is very close to the azeotrope. We anticipate a large number of steps. Available
online is a Mathcad document♣ which steps off stages automatically using the optimum feed
location.
The results are shown in the figure above. The program reports the total number of steps and the
feed location. In this figure, there are a total of 64 steps with the feed introduced on the 60th
from the top. In the upper right corner of the xy-diagram, the steps are becoming so small as to
be invisible.
Below on the left is a blow-up of the region from 0.7 to 0.8. Even on this blow-up the steps
are becoming invisible in the upper right corner. So the second graph below on the right is
another blow-up: this from 0.79 to 0.80. Still the steps are becoming vanishingly small. Notice
that in these blow-ups, both the equilibrium curve and the operating line are very linear.
♣ McCabe-Thiele.mcdx can be found in the “Student Resources” folder under the “Handouts” heading.

06-361 page 126 Spring, 2017
0.78 0.798
0.76 0.796
0.74 0.794
0.72 0.792
0.7 0.79
0.7 0.72 0.74 0.76 0.78 0.79 0.792 0.794 0.796 0.798
Over a narrow range, any continuous

curve becomes virtually linear. When both
the operating line and equilibrium curve are
straight lines (see xy-diagram at right), an
analytic expression exists for calculating the
number of steps:

log 
( ya* − ya ) 

N=

 ( yb* − yb  )
 (110)

log 
( ya* − yb* ) 

 ( ya − yb ) 
 
which is called Kremser's equation.♥ The subscripts a and b denote streams at the top and
bottom of the cascade, respectively (i.e. “b” is for “bottom”). The y*s are vapor compositions in
equilibrium with the liquid at the top and bottom of the cascade.
Proof: As we step down the cascade of stages (see schematic at

right), (xi, yi) for any stage #i is a point on the EC which is now
assumed to be linear:
yi = mxi + b (111)
♥ A. Kremser, “Theoretical Analysis of Absorption Process,” Nat’l Petrol. News

22, 42-52 (1930).

06-361 page 127 Spring, 2017
On the other hand, (xi, yi+1) is a point on the OL, which is always linear (assuming equimolal
overflow) [recall (99) on page 111]:
L  L 
yi +1 = xi +  ya − xa 
V  V 
where we have substituted y1=ya and x0=xa. Using (111) to eliminate xi the OL equation
becomes:
L yi − b L
=
yi +1 + ya − xa (112)
 m
V 
V
mA mA
Now we define a new parameter, which represents the ratio of the slope of the operating line
(L/V) to the slope of the equilibrium curve (m):
L
A≡
mV
In terms of A, (112) becomes
yi +1 = A ( yi − b ) + ya − Amxa = Ayi + ya − A ( mxa + b )

 
y *a
From (111), we see that mxa+b is just the vapor concentration in equilibrium with liquid at
concentration xa, which we denote at y*a. Making this substitution, the above equation becomes
yi +1 =Ayi − Ay *a + ya (113)
This serves as a recursion formula, which we can apply repeatedly to obtain the concentration on
successive stages, starting at the top. For i=1, recalling y1 = ya, (113) yields
i = 1: y2 = A y1 − Ay *a + ya = ya (1 + A ) − Ay *a

ya
y3 =Ay2 − Ay *a + ya =A  ya (1 + A ) − Ay *a  − Ay *a + ya
( ) ( )
i = 2:
= ya 1 + A + A2 − y *a A + A2
where the second equality above results from substituting y2 from the previous step. Now we
begin to see a pattern developing:

06-361 page 128 Spring, 2017
i = n: ( ) (
yn +1= ya 1 + A + A2 +  + An − y *a A + A2 +  + An
 
) (114)
A (1+ A + A2 + + An −1 )
Each sum is a partial sum of a geometric series, whose total is known analytically:♣
1 − An + 1
1 + A + A2 +  + An =
1− A
Substituting this result into (114):
1 − An + 1 1 − An
=yn + 1 ya − y *a A
1− A 1− A
For the bottom stage of the cascade, n=N and yN+1 = yb:
1 − A N +1 1 − AN
=yb ya − y *a A
1− A 1− A
Multiplying through by 1-A and collecting terms multiplied by A or by AN+1:
yb (1 − A )= ya (1 − A N +1 ) − y *a A (1 − A N )
yb − Ayb =ya − A N +1 ya − Ay *a + A N +1 y *a
N +1
yb − ya + A ( y *=
a − yb ) A ( y *a − ya ) (115)
This can be further simplified. Recall that A is defined as the ratio of the slopes of the operating
line to the equilibrium curve. This ratio can be calculated from the concentration defined on the
xy-diagram above:
( ya − yb )
slope of OL ( xa − xb ) ya − yb
A≡ = = (116)
slope of EC ( y *a − y *b ) y *a − y *b
( xa − xb )
We can use this result to obtain ya – yb = A(y*a – y*b), which will now be substituted into (115):
♣ see (2.6) in M.D. Greenberg, Foundations of Applied Mathematics, Prentice-Hall, 1978. Available electronically
since 2013 as a Dover Edition.

06-361 page 129 Spring, 2017
A ( y *b − y *a ) + A ( y=
*a − yb ) A N +1 ( y *a − ya )
  
( )
A y *b − y *a + y *a − yb= A ( y *b − yb )
After dividing through by A and solving for AN:
 y * −y 
log  b b 
AN =
y *b − yb
or N=  y *a − ya 
y *a − ya log A
Taking the log of both sides, recalling that log(AN) = N logA, then solving for N we obtain the
second expression above. To obtain the form of Kremser's equation given as (110), we substitute
(116) for A.
Comment #1: In the xy-diagram shown next to (110), the equilibrium curve lies above the
operating line. This is always the arrangement when we are dealing with gas absorption
(transport from gas to liquid). In order to obtain N>0, the following inequalities must be obeyed:
distillation: ya* > ya yb* > yb ya > yb
If you inadvertently (but consistently) interchange the subscripts a and b, you will obtain a
negative value for N.
Comment #2: If the operating line and equilibrium curve should cross (which is physically
impossible), when evaluating (110) you will find yourself trying evaluate the log of a negative
number.
Comment #3: The situation in which the operating line and equilibrium curve are parallel (i.e.
A = 1) is a special case: (110) becomes indeterminant. However, since the driving force y*-y is
constant, the number of steps is just the total change in concentration ya-yb divided by the height
of one step y*a-ya:
ya − yb
for A = 1: N=
y *a − ya
Now we will return to discuss the motivation for Kremser's equation: approaching the
intersection of OL and EC:
Notice that, as xb approaches the point of intersection (see Figure 12 on page 124), yb* → yb
and
( ) 
log 
= 0  log [ ∞ ]
=N →∞
log [ ] log [ ]

06-361 page 130 Spring, 2017
Thus you can never reach the point of intersection for any finite number of steps. In particular, a
very large number of stages is required as we increase the purity of either of our two product
streams. For example, as xD→1, the end-point (xD, xD) of our operating line approaches (1,1),
which is always one point on the equilibrium curve: as xD→1, N→∞.
Example: in the previous example in which xD is close to the azeotrope, both the OL and EC
appear to be straight in the interval 0.7900 ≤ x ≤ 0.7999 (see right-hand figure on page 126). Use
this fact and Kremser’s equation to calculate the number of stages in this interval.
Solution: we need to calculate the four y’s which appear in (110). Two of these y’s are given by
the operating line. One of the points on the operating line is
xa = ya = xD = 0.7999
The second point on the operating line (104) is:
R 1
yb = xb + xD =0.791029
R +1 R +1
Substituting xb = 0.79 and R = 8.618, we obtain the value given in the second equation above.
To evaluate points on the equilibrium curve, we linearly interpolate two points at each end of this
interval: (0.790, 0.792146)♠ and (0.800, 0.800).
0.800 − 0.792146
y * −0.792146
= ( x − 0.790 )
0.800 − 0.790
For xa = 0.7999, we obtain ya* = 0.799921 and for xb = 0.79, we obtain yb* = 0.792146.
Substituting these numbers into Kremser’s equation, we obtain N = 30.1.
♠ The value of y* = 0.792146 was obtained using the cubic-spline interpolation [i.e. yeq(0.79)] of the tabular data
for this methanol-acetone system found in McCabe-Thiele.mcdx, except one point was omitted from the table: (1,
1). Values reported above were calculated by McCabe-Thiele near pinch (in-class example).mcd. Omission of this
point – which is outside the interval of interest – alters slightly the result of the interpolation: including (1, 1) gives
yeq(0.79) = 0.791899; excluding (1, 1) gives yeq(0.79) = 0.792146.

06-361 page 131 Spring, 2017
 y* − y 
log 
a a  ( )
 log  0.799921 − 0.799900 
=N = 
yb
*
− yb
 (  )
 0.792146 − 0.791029 

log
 y* − y* 
 a b  ( )
 0.799921 − 0.792146 
log 
 0.799900 − 0.791029 
( y − y ) 
 a b 

 0.000021 
log  
= =  0.001117  log ( 0.0019 ) −1.72
= = 30.1
 0.007775  log ( 0.876 ) −0.057
log  
 0.008871 
Stepping off the stages with Mathcad, we require 33 steps to get the liquid mole fraction below
0.79.
CHOOSING THE REFLUX RATIO

An important part of the design of any distillation column is the choice of reflux ratio. There
are a few qualitative considerations which must be kept in mind.
1) Existence of Rmin
The first thing you should know is that there is a minimum

value that the reflux rtio can assume and still achieve the
desired separation. Recall that the slope of the ROL is
R
slope of ROL=
R +1
Notice that, as R is reduced, the intersection of the ROL and

the Q-line moves toward the equilibrium curve. As it does so, the minimum number of stages
required approaches infinity for the reasons mentioned above: you
can never get to the intersection of the operating line and the
equilibrium curve.
Usually this minimum reflux ratio can be determined using

the point of intersection between the q-line and the equilibrium
curve. On some peculiar xy diagrams the ROL or SOL
constructed in this way might cross the equilibrium curve at
points other than where the Q-line crosses. Then the reflux ratio
must be increased until neither operating line crosses equilibrium
curve before reaching the q-line. Of course, the operating line can
be tangent to the equilibrium curve.

06-361 page 132 Spring, 2017
This point of tangency is called a pinch point. In stepping off the stages, it will take an
infinite number just to reach the pinch, but if the reflux ratio is increased ever so slightly, then
the separation can be achieved, although the number of stages might be large.
minimum reflux ratio -- smallest value of R for which the desired separation can be achieved
with a finite number of stages.
The pinch point might also occur on the SOL as shown at right.
Regardless of where the pinch occurs, once you have drawn the q-
line and operating lines, the minimum reflux ratio (thus determined
graphically) can be calculated from:
x − y′
R= D (117)
y ′ − x′
where (x', y' ) is any point on the ROL except (xD, xD). For
maximum accuracy in a graphical solution, choose a point as far as
possible from (xD, xD); usually this would be the y-intercept of the ROL.
Proof: (117) was derived by setting xi = x' and yi+1 = y' in (104) on page 112:
R x
=
yi +1 xi + D (104)
 R +1  R +1
y′ x′
and then solving for R. The point (x', y' ) can be chosen for convenience: for example, you might
want to choose the y-intercept.
To determine the minimum reflux ratio using the McCabe-

Thiele method, we might imagine drawing the ROL and SOL
on the xy-diagram starting with a high reflux ratio (where ROL
and SOL coincide with 45° line) and then decrease the reflux
ratio until the first of three things happens:
1. the ROL becomes tangent to the equilibrium curve
2. the SOL becomes tangent to the equilibrium curve
3. the intersection of the ROL and SOL occurs on the

equilibrium curve
An important thing to notice about the minimum reflux ratio is that, as we approach it, the
number of stages approaches infinity:
as R → Rmin, N → ∞

06-361 page 133 Spring, 2017
Clearly you don't want to operate near the minimum reflux ratio because it would cost too much
to build a column with so many stages. Generally as you increase the reflux ratio, both operating
lines move away from the equilibrium curve so that the steps become larger. Then fewer stages
will be needed.
as R↑, N↓
2) Total Reflux: R→∞ gives Minimum Stages
In the limit that R approaches infinity, the y-intercept of the ROL becomes zero. The ROL
becomes the 45° line. Of course, the SOL and ROL always intersect along the q-line, which
means that the SOL also becomes the 45° line. This is the furthest the operating lines can ever
be from the equilibrium curve, so this condition gives the minimum number of stages:
as R→∞: ROL → SOL → 45°-line
and N → Nmin
Although it might appear that this is the best value of the reflux ratio, it turns out that there is a
"dark side" to this solution. The problem is what is happening to flowrates inside the column.
Recall R ≡ L/D
Now to get infinite reflux ratio, one of two things has to happen:
R → ∞: L→∞ or D→0
In most designs, you are interested in processing a fixed amount of material, this implies that
F,D, and B are fixed, which eliminates the second alternative above. If V = (R+1)D, L = RD and
D remains fixed as R→∞, we see that both L→∞ and V→∞:
L → ∞: V → ∞, qc → ∞, qr → ∞, tower diameter → ∞
This means that the heat exchanger for both the

condenser and the reboiler are getting very
large as are the cost of utilities (steam and
cooling water).
Thus the cost of building the tower (these

are called “fixed costs”)♣ becomes large either
near the minimum reflux ratio (where the
height of the tower becomes large) or near total
♣ The term “fixed costs” is used in accounting because the depreciation of construction costs is the same every
month regardless of how the equipment is operated. In contrast, the cost of utilities (electricity, steam, cooling
water) are termed “variable costs” because they vary month by month, depending on how the equipment is used.

06-361 page 134 Spring, 2017
reflux (where the diameter of the column becomes large). The costs of utilities increase about
linearly with reflux ratio.
The total costs will experience a minimum at some reflux ratio, which we call the
optimum reflux ratio: minimizes total costs/year
Typically the optimum occurs for:
1.1 < Ropt/Rmin < 1.5
However most chemical plants actually operate at slightly larger R since the additional cost is not
very much and it buys you flexibility in case you decide later you want to change xD or in case
the feed composition changes, you have some room to maneuver:
1.2 < R/Rmin < 2.0
CALCULATING FLOWRATES AND

CONDENSER/REBOILER HEAT DUTIES
To compute the heat duties of the condenser and reboiler,
we will need to know the flowrates in various parts of the
column. We begin by determining the output flows as
illustrated in the following example
Given: F, zF, xD, xB and R
Find: qr and qc
Solution: performing a total mole balance and a component

mole balance over the entire distillation column (the blue
system in and figure at right), we obtain two equations in the
two unknowns.
F=D+B
zFF = xDD + xBB
These are the same pair of equations we solved earlier for flash distillation [see (80) and (81) on
page 93]. The solution is given by (89) and (90):
x − xB x − xF
D= F F and B= D F (118)
xD − xB xD − xB

06-361 page 135 Spring, 2017
Now we’ll determine the internal flows by starting at the top of the column and work our
way down. First we use the definition of reflux ratio:
L = RD and V = L+D = (R+1)D (119)
In a total condenser, we are condensing all of the overhead vapor. Assuming the condenser
output is a saturated liquid, the condenser heat duty is just
qc = V λ
x − xB
qc= ( R + 1) Dλ= ( R + 1) F Fλ
xD − xB
where in the second and third equations, we have substituted (119) and (118), respectively.
To calculate the reboiler heat duty qr, we need to

evaluate the molar flowrate through it: V . A look at
the streams entering the feed plate (see figure at right),
we realize that we have at least two unknowns L and
V . A total mole balance gives:
V +F +L=V +L (120)
which represents only one relationship in two

unknowns. The remaining relationship is obtained from an enthalpy balance about the feed tray:
FH F + LH L + VHV = LH L + VHV (121)
Our analysis of countercurrent cascades is based on the equimolar overflow assumption,

which contains assumptions about enthalpy: see "Constant Molar Overflow" on p110. In
particular, we neglect sensible heat and assume that all components have the same latent heat of
vaporization, then we can evaluate the enthalpies of all saturated liquid or saturated vapor
streams as:
HL = 0, HV = λ
Recall the definition of q from page 121. This allows us to express the enthalpy of the feed in
terms of q:
HV − H F λ − H F
q≡ = or HF = (1 – q)λ
HV − H L λ
Substituting these values for enthalpy into (121) and solving for V gives:
F(1-q)λ + V λ = Vλ or V = V − (1 − q ) F (122)

06-361 page 136 Spring, 2017
Substituting this into (120) and solving for L gives:
L= L + qF (123)
Thus introducing the feed onto a plate causes a discontinuity in the flowrates of liquid and vapor
in the column. Although the flowrates are the same for each stage in the cascade above the feed,
and the same for each stage in the cascade below the feed, the flowrates are different in the two
cascades.
For example, for a saturated liquid feed, q=1, the liquid leaving the feed tray has a higher
flowrate than the liquid entering by an amount equal to the feed flow rate:
for q=1: L –L = qF = F and V– V = (1–q)F = 0
whereas the vapor flow is unchanged.
On the hand, if the feed is saturated vapor, then the vapor flowrate leaving
the stage will be higher than entering by an amount equal to the feed rate.
for q=0: L –L = qF = 0 and V– V = (1–q)F = F
The flows for other thermal conditions of the feed are summarized graphically
below:
When the feed is a superheated vapor (q<0), not only does all of its flow join the vapor stream,
but it also boils up some of the liquid. When the feed is a two-phase mixture (0<q<1), then the
liquid fraction joins the liquid stream and the vapor fraction joins the vapor stream. When the
feed is a subcooled liquid (q>1), all of its flow enters the liquid and some of the vapor is also
condensed and joins the liquid stream too.
In any case, given F, L and V, we can calculate L and V from (122) and (123). Knowing
all the flowrates, we can now calculate the reboiler heat duty. The reboiler has to boil all the
vapor sent into the bottom cascade. For a partial reboiler (where the bottoms product is
withdrawn from the liquid remaining in the kettle), there can be no superheating of the vapor (it
must be a saturated vapor). Then the reboiler heat duty is
qr = V λ

06-361 page 137 Spring, 2017
 x − xB   x − xB 
q=
r V − (1 − q ) F =
λ ( R + 1) F
xD − xB
F − (1 − q ) F = λ ( R + 1) F
xD − xB
+ q − 1 F λ
      
 V 
where the second and third equations result from substituting

(122), then (119) and (118).
ROLE OF Q-LINE
Recall (see “Step 4” on p121) that we said one point of the
SOL is the intersection of the q-line (QL) and the ROL. In this
section, we present the arguments supporting this claim.
The equations of the operating lines are obtained by

performing a mole balance on component A between two pairs
of streams: one pair in the rectifying cascade and the second
pair in the stripping cascade (balances use the systems denoted
by the red lines):
ROL: Vy = Lx + DxD (124)
SOL: V y = L x - BxB (125)
We claim that the q-line represents the intersection of these two lines. Think of these as two
equations in two unknowns (x, y). The point of intersection of these two lines is that point (x, y)
which simultaneously satisfies both equations.
Now we are not going to try to solve these two equations for (x, y). But we can more easily
find another linear relationships between x and y by combining these two equations. One linear
combination is particularly convenient in that it yields an equation which we can plot without
knowing any of the flowrates that appear in (124) and (125):
(125) minus (124): ( V -V)y = ( L -L)x - (BxB+DxD)
Notice that the last term in this equation represents the total rate of
component A leaving the tower. This must equal the total rate at
which A enters the tower. So the above becomes:
( V -V)y = ( L -L)x - FzF
Dividing by F gives:
V −V L−L
= y x − zF
F F
q −1 q

06-361 page 138 Spring, 2017
Substituting (122) and (123): (q-1)y = qx - zF
 q  zF
or =y  x+
 q −1 1− q
which is the equation of our q-line. This equation represents a linear combination of the two
equations which are satisfied at the intersection of the two operating lines. Thus this equation
must also be satisfied at that point. It turns out that the QL represents the locus of intersections
of the ROL and SOL for different reflux ratios.
PLATE EFFICIENCY
So far we have always assumed that the vapor and liquid
streams leaving a given stage are at equilibrium. In stepping off the
stages, we have always gone all the way from the operating line to
the equilibrium curve. If the contacting of the two phases on the
tray is good (that is, if the area of contact is large enough and the
time of contact is long enough) we can approach this ideal state, but
we will never quite reach it. For a non-equilibrium stage, our
horizontal step will be a little shorter than for an ideal stage. Of
course, we still end up at a point on the operating line, because this
represents a mole balance which must still be satisfied.
Because the real steps are shorter, it will take more of them to achieve the same separation.
The ratio of the number of ideal to real stages required will be less than one and is called the:
overall (column) efficiency -- Nideal / Nreal ≡ Eoc < 1
While this efficiency is easy to use, the value of the efficiency is hard to
determine. A more fundamental quantity is the
Murphree (tray) efficiency –
Tray efficiencies can be based either on changes in liquid composition or

changes in vapor composition. The definition based on liquid
composition is given by
x −x width of real step

EMx ≡ n −1 n = (126)
xn −1 − xn* width of ideal step
where xn = actual x leaving stage n

06-361 page 139 Spring, 2017
xn* = x in equil with yn
See equilibrium curve and operating line sketched at right. Basically the definition is the size of
a real step from the operating line toward the equilibrium curve divided by the size of an ideal
step (which would take us all the way to the equilibrium curve).
The corresponding definition for the efficiency based on the vapor mole fractions is
yn − yn + 1 height of real step

≡ EMy = (127)
yn* − yn +1 heigth of ideal step
This definition is based on the height of a step (involves changes in y) rather than based on the
width of a step (changes in x). Unless the equilibrium curve and operating lines are parallel,
these two definitions are not equivalent, but the differences are probably within experimental
error.
Some typical values of plate efficiencies are given in Table 14-12 (taken from Perry’s
Chemical Engineering Handbook, 8th edition, 2008).♦
♦ CMU users can access the entire book online at http://site.ebrary.com/lib/cmu/detail.action?docID=10211725.

06-361 page 140 Spring, 2017
DETERMINING NUMBER OF REAL STAGES

Suppose we have a value for EMx or EMy, how do we use it to determine the number of real
stages required? The procedure is similar to that for ideal stages:
Steps 1-6) same as for ideal stages
We locate the two operating lines and the q-line, just as before. But before we step off the stages
we use the efficiency to determine a new curve between the operating lines and the equilibrium
curve.
Step 7) Plot pseudo-equilibrium curve using

either (127) or (126).
The blue curve in the xy-diagram at right

was drawn using (127). At several x-values
between xB and xD, we draw a vertical line
between the lowest operating line and the
equilibrium curve. On each of these line
segments, we locate a point which is a
fraction of the distance up the line segment
from the bottom, where that fraction equals
the plate efficiency. The locus of these
points is the pseudo-equilibrium curve.
Finally, we step off the stages as before, but

using the pseudo-equilibrium curve in place
of the true equilibrium curve.
Step 8) Step off stages between operating

lines and pseudo-equilibrium curve.
Comment #1: In locating the blue pseudo-

equilibrium curve in the figure above, we
used five vertical lines with each yielding one
point on the pseudo-equilibrium curve. The
lengths of vertical lines represent changes in
vapor mole fraction; thus this pseudo-
equilibrium curve corresponds to a particular
value of EMy. Had we instead been given a
value for EMx, we would use horizontal lines
to locate the pseudo-equilibrium curve (see
green curve in figure at right).
Comment #2: Unlike the true equilibrium

curve, which depends solely on the vapor-
liquid equilibrium, the pseudo-equilibrium

06-361 page 141 Spring, 2017
curve depends both on the vapor-liquid equilibrium and on the operating lines. Thus if we
change the reflux ratio, for example, we will have to replot the pseudo-equilibrium curve.

06-361 page 142 Spring, 2017
Chapter 12. Multi-Component Distillation

The objective of distillation is the separation of the feed into streams
of nearly pure products. In binary distillation, we have two components
and two product streams. With more than two components in the feed,
there is no way to achieve this with a single column (a single column
might have a third product stream, but it’s composition would never be
nearly pure).
To separate a mixture of three components into three

nearly pure streams, we will need at least two separate
distillation columns to get three streams. In general, if
there are Nc components, we will need Nc-1 columns.
Nc components
Nc-1 columns
This assumes that each column can achieve a sharp separation between two components of
adjacent volatility. “Alternative #1” represents one scheme for separating a 3-component feed
into 3 nearly pure products.
Alternatively, the designer might try to first split off

component C in the first column — rather then A. The
choice between these two alternatives is a matter of
experience. One “rule of thumb” is to perform the easiest
separations first. In this course, we will not delve into the
strategy for deciding the overall scheme for multiple
distillation columns. Instead, we will focus on the design
of single columns after this decision has been made.
SHARP SPLIT
The first step in the design of a single column is to specify the product distribution. This can
be estimated easily by assuming the column achieves a “sharp split” between two components.
To some extent, the choice of which components to split is up to the designer who can tune the
reflux ratio and feed location to obtain different sharp splits.

06-361 page 143 Spring, 2017
Let A,B,C,... denote components ranked in order of decreasing

volatility.
sharp split -- distillate contains only components A-M, while

bottoms contains only components N-Z
Comment #1: In any real distillation, all components will be

found in all product streams in at least trace amounts. When we
say “only” components A-M are found in the distillate, we
usually mean that the mole fractions of the remaining
components (i.e. N-Z) can be neglected when summing the
mole fractions to unity for the distillate stream. Similarly, we can neglect the mole fraction of
components A-M in the bottoms.
Comment #2: Sharp separations are not always possible, but we will focus on them in this
introduction, for the sake of simplicity.
In the description above, components M and N play a special role and are called
key components -- M,N (for sharp split)
The more volatile of the two is called the “light key,” while the other is the “heavy key”:♣
light key = M
heavy key = N
PRODUCT DISTRIBUTION FOR A SHARP SPLIT

Example #1: A mixture with 33% hexane, 37% heptane and 30% octane is to be distilled to give
a distillate product of 1% heptane and a bottoms product of 1% hexane. Estimate the remaining
compositions of the distillate and bottoms.
given: zF6 = 0.33, zF7 = 0.37, zF8 = 0.30, xD7 = 0.01, and xB6 = 0.01
find: D/F, B/F, and xDi, xBi for the other components
Solution:
Step 1) Rank components by volatility
♣ Here “light” and “heavy” suggest weight, as in molecular weight: in a homologous series, lower molecular weight
compounds are more volatile so the light key is more volatile than the heavy key.

06-361 page 144 Spring, 2017
In this example, ranking the volatilities is Component Formula

easy since we have a homologous series of hexane C6H14 most volatile
alkanes. Since vapor pressure generally heptane C7H16
drops with molecular weight in any octane C8H18 least volatile
homologous series, we don't need to
consult tables of boiling points or vapor pressures to order the components.
If we had a more complex mixture (i.e. not all components members of the same homologous
series), then we would look up the boiling points of the pure components at the pressure in the
distillation column. The ranking then is from lowest boiling point (i.e. most volatile) to highest
boiling point (least volatile).
Step 2) Identify the key components
We are told that the bottoms is 1% hexane and the

distillate is 1% heptane. Note that hexane is more volatile
than heptane so the vast majority of the hexane ends up in
the distillate and the vast majority of the heptane ends up in
the bottoms. Since octane is less volatile than heptane, we can reasonably assume that all of the
octane must be in the bottoms too. Thus we have a sharp split between hexane and heptane, with
hexane going to the distillate and the remaining two components going to the bottoms.
Let's take a basis of 100 moles of feed (F=100) and try to fill out the following table:
Component FzFi zFi DxDi xDi BxBi xBi
hexane 33 0.33 -- -- -- 0.01 light key
heptane 37 0.37 -- 0.01 -- -- heavy key
octane 30 0.30 -- -- -- --
totals = 100 1 D 1 B 1
Comment #1: In this example the sharpness of split was

specified by giving the mole fractions xD,HK and xB,LK.
Alternatively, we might have been given the fractional
recovery of the two key components in either product
stream: DxD,HK / FzF,HK is the recovery of the heavy key in
the distillate and BxB,LK / FzF,LK is the recovery of the light
key in the bottoms.
Step 3) Apply “rules” Figure 13. A perfectly sharp split

between C6 and C7 would achieve the
As an initial guess of the product distribution for a separation shown here.
sharp split, we apply the following rules:
1. Components heavier than the HK are assumed to end up in entirely in the bottoms
2. Components lighter than the LK are assumed to end up in entirely in the distillate
In essence, we are then making a sharp split between hexane and heptane with the “1%” values
specifying the sharpness of the split desired. In other words, hexane has been chosen as the light

06-361 page 145 Spring, 2017
key and heptane as the heavy key. If we assume that there is virtually no octane in the distillate
(octane is heavier than the heavy key), then we can finish specifying the composition of the
distillate stream. In this case, this means 1) that the mole fraction of octane in the distillate is
zero and 2) the molar flowrate of octane in the bottoms is equal to that of the feed: 30. Finally,
to have the sum of the mole fractions of the distillate equal unity, the mole fraction of hexane can
be deduced to be 1 – 0.01 = 0.99:

hexane 33 0.33 -- 0.99 -- 0.01
heptane 37 0.37 -- 0.01 -- --
octane 30 0.30 0 0 30 --
totals = 100 1 D 1 B 1
Step 4) Apply mole balances
Knowing the concentration of hexane in both of the product streams allows us to determine the
flowrates from mole balances:
component C6: 0.99D + 0.01B = 33
total: D + B = 100
Solving simultaneously: D = 32.65, B = 67.35

hexane 33 0.33 32.32 0.99 0.67 0.01
heptane 37 0.37 0.01 --
octane 30 0.30 0 0 30 --
totals = 100 1 32.65 1 67.35 1
Now using the fact that the balance of the distillate and bottoms streams must be heptane, we can
fill out the table:

hexane 33 0.33 32.32 0.99 0.67 0.01
heptane 37 0.37 0.33 0.01 36.67 0.5445
octane 30 0.30 0 0 30 0.4454
totals = 100 1 32.65 1 67.35 1
Comment #2: The only additional information that is needed if the number of components is
larger than 3 is the mole fraction of those components in the feed. Although there will be more
rows in the table (if there are more components), the additional unknown mole fractions in the
distillate and bottoms streams equals additional constraints obtained by applying the “rules.”
Example #2: A mixture with 25% butane, 25% pentane, 25% hexane and 25% heptane is to be
distilled to recover 95% of the butane in the distillate and 95% of the pentane in the bottoms.
Estimate the remaining compositions of the distillate and bottoms.

06-361 page 146 Spring, 2017
Given: zF4 = zF5 = zF6 = zF7 = 0.25, DxD4 = 0.95×FzF4 and BxB5 = 0.95×FzF5
Find: D/F, B/F, and xDi, xBi for the other components
Solution:
Step 1) Rank components by volatility
Step 2) Identify key components
A vast majority of the butane ends up in the distillate and

the vast majority of the pentane ends up in the bottoms.
Since the remaining components are less volatile than
pentane, we can reasonably assume that all of them must be in the bottoms too. Thus we have a
sharp split between butane and pentane.
Recall that 95% recovery refers to the component flowrate in the distillate or bottoms as a
fraction of the component flowrate in the feed:
DxD4 = 0.95×FzF4 = 0.95×25 = 23.75
Similarly BxB5 = 0.95×FzF5 = 23.75
Let's take a basis of 100 moles of feed (F=100) and try to fill out the following table:

butane 25 0.25 23.75 -- -- -- light key
pentane 25 0.25 -- -- 23.75 -- heavy key
hexane 25 0.25 -- -- -- --
pentane 25 0.25 -- -- -- --
totals = 100 1 D 1 B 1
Step 3) Apply “rules”
Components heavier than the heavy key end up entirely in the bottoms. This gives us the
flowrate of the hexane and pentane components.

butane 25 0.25 23.75 -- -- -- light key
pentane 25 0.25 -- -- 23.75 -- heavy key
hexane 25 0.25 0.00 -- 25.00 --
pentane 25 0.25 0.00 -- 25.00 --
totals = 100 1 D 1 B 1
Step 4) Apply mole balances
Knowing the component flowrate of any component in one of the product streams allows us to
determine that components flowrate in the other product stream:

06-361 page 147 Spring, 2017
component C4: BxB4 = FzF4 – DxD4 = 25 – 23.75 = 1.25
component C5: DxD5 = FzF5 – BxB5 = 25 – 23.75 = 1.25

butane 25 0.25 23.75 -- 1.25 -- light key
pentane 25 0.25 1.25 -- 23.75 -- heavy key
hexane 25 0.25 0.00 -- 25.00 --
pentane 25 0.25 0.00 -- 25.00 --
totals = 100 1 D 1 B 1
Adding up the component flowrates gives us the totals:

butane 25 0.25 23.75 -- 1.25 -- light key
pentane 25 0.25 1.25 -- 23.75 -- heavy key
hexane 25 0.25 0.00 -- 25.00 --
pentane 25 0.25 0.00 -- 25.00 --
totals = 100 1 25.00 1 75.00 1
Knowing the flowrate of each component in each stream, we can divide by the total flowrate to
component the mole fraction of each component in each stream:

butane 25 0.25 23.75 0.95 1.25 0.017 light key
pentane 25 0.25 1.25 0.05 23.75 0.317 heavy key
hexane 25 0.25 0.00 0.00 25.00 0.333
pentane 25 0.25 0.00 0.00 25.00 0.333
totals = 100 1 25.00 1.00 75.00 1.000
Comment: Notice that the extra unknowns introduced by an additional component were matched
by an equal number of extra constraints arising from applying the “rules.”
Proof: To prove this statement, let’s do an accounting of unknowns and equations.
Unknowns: xDi and xBi for i = 1, 2…N ............................ 2N

B and D ..................................................... 2
total unknowns .......................................... 2N+2

06-361 page 148 Spring, 2017
Equations: mole balances .................................................. N

ΣxDi = 1 and ΣxBi = 1 ....................................... 2
xD,HK and xB,LK (or 2 product
recoveries) are specified ............................. 2
“rules”♥ ..................................................... N−2
total equations ........................................... 2N+2
MINIMUM NUMBER OF TRAYS

Now let's see if we can get some idea of how many trays we will need. This is difficult to
calculate precisely (see Rigorous Solution of MESH Equations below on page 163). What can be
done relatively easily is to estimate the number of trays needed at total reflux -- this is the
minimum number of trays and it can be estimated by Fenske’s equation:
x x 
ln  Di Bi 
 xDj xBj 
N min =  
ln αij ( )
To obtain such a simple analytical result requires some gross assumptions:
• equal molal overflow: L=const and V=const for all trays

• total reflux: L≈V
• constant relative volatilities: αij=const for all trays
In case you have forgotten, the definition of relative volatility can be found in the next
section and the “proof” that it is nearly constant for ideal solutions of components having the
same molar heat of vaporization can be found on page 83.
Proof:
Ki
Recall αij ≡ (128)
Kj
is the relative volatility of component i compared to j, where
y
Ki ≡ i (129)
xi
♥ We have one rule for each component lighter than the light key and heavier than the heavy key, which is one real
for each component less the two keys.

06-361 page 149 Spring, 2017
is the distribution coefficient for component i (or j). With these assumptions, we can start at the
top of the column and work out the compositions of streams leaving every stage.
For a total condenser:
y1i = x0i = xDi (130)
for i=1,...,Nc. The first subscript is the stage number (as in

binary mixtures) and the second subscript is the component.
With the composition of the vapor stream leaving stage 1 known,
we can determine the composition of the exit liquid stream from
the known values of the relative volatility:
y1i
Ki x
αij ≡ = 1i
K j y1 j
x1 j
x1i 1 y1i y
or = = α ji 1i (131)
x1 j αij y1 j y1 j
x1i x
Substituting (130): = α ji Di (132)
x1 j xDj
for each component i. Notice that if i=j, (132) yields
x1i x
for i=j: = αii Di or 1=1
x1i  xDi
 1 
1 1
which doesn’t tell us anything new; thus we have only Nc-1 independent relations of this type
instead of Nc. Choosing one element as the reference (say j=Nc), we can write Nc-1 relations for
the Nc-1 independent mole fractions. The missing equation is Σx1i = 1. Thus the composition of
the liquid leaving tray 1 has been completely determined.
Next we determine y2i by mole balances (system is enclosed by red rectangle in figure
above).
component i: Vy2i = DxDi + Lx1i
For total reflux D≈0 so that L≈V, leaving:
y2i = x1i (133)

06-361 page 150 Spring, 2017
for i=1,...,Nc.
Knowing the composition of the vapor leaving stage 2, we can calculate the composition of
the liquid leaving stage 2 from the known volatilities. Rewriting (131), replacing the stage
number 1 by 2:
x2i y
= α ji 2i
x2 j y2 j
x2i x
From (133): = α ji 1i
x2 j x1 j
x2i x
Using (132): = α 2ji Di (134)
x2 j xDj
From (132) and (134), we begin to see a trend. The generalization of this trend yields:
xni x
= α nji Di
xnj xDj
In particular, we are interested in the number of steps required to get from the distillate to the
bottoms:
xBi N x
= α jimin Di
xBj xDj
The exponent represents the minimum number of steps on the xy diagram. One of these steps is
the partial reboiler. Solving for Nmin:
x x 
ln  Di Bi 
 xDj xBj 
N min =  
( )
ln αij
Now in any real system, the relative volatility will not be constant. One way to patch things
up is to use some kind of average volatility. The usual choice is the geometric mean of the
volatilities evaluated at the two extreme temperatures in the column:
αij = α Dij α Bij

06-361 page 151 Spring, 2017
where αDij and αBij are evaluated at the

temperatures of the top and bottom stages,
respectively. Assuming a total condenser, αDij
should be evaluated at the dew-point of the
vapor leaving the top stage, which has the
same composition as the distillate. This is not
the same as the temperature of distillate stream
which is instead at its bubble point. On the
other hand, the bottom stage is actually the
partial reboiler. Thus αBij is evaluated at
boiling temperature inside the reboiler, which is the bubble-point of the bottoms.
Review of Relevant Thermo
The relative volatility of component i compared to j is defined as (recall page 82)
Ki
αij ≡ (135)
Kj
y
where Ki ≡ i (136)
xi
is the distribution coefficient for component i (or j). The latter quantity describes how a given
component distributes itself between vapor (y) and liquid phases (x). For mixtures of
components which form ideal solutions, Raoult's law applies:
pi = xi Pio (T ) (137)
where Pio (T ) is the vapor pressure of pure component i at the same temperature as the mixture.
Recall that partial pressure of a component is defined as the mole fraction times the total
pressure:
pi ≡ yiP = partial pressure of component i (138)
where P is the total pressure, which also equals the sum (over all Nc components) of the partial
pressures:
Nc
P = ∑ pi
i =1
Substituting xi from (137) and yi from (138) into the definition of partition coefficient (136), we
have

06-361 page 152 Spring, 2017
pi
yi P Pio (T )
Ki ≡ = = (139)
xi pi P
Pio
Finally, we substitute this result into (135):
Ki P o (T )
αij= = i (140)
K j Pjo (T )
Example #3. Calculate the Nmin for the separation in Example

#1 if the column is operated at 1.2 atm. The remaining stream
compositions were determined in Example #1 and are
summarized in the figure at right.
Solution: We need to evaluate the relative volatilities in the

distillate and bottoms streams knowing their composition.
Vapor pressure data for pure components is available in the
CRC handbook. C6, C7 and C8 form a nearly ideal mixture.
So the relative volatilities can be calculated from the vapor pressures using (140). We used
vapor pressure data from Lange’s Handbook of Chemistry, 13th edition, pages 10-45 thru 10-49,
where constants Ai, Bi and Ci for each component can be found.
component A B C T range (ºC)

hexane 6.87601 1171.17 224.41 -25 to 92
heptane 6.89677 1264.90 216.54 -2 to 124
octane 6.91868 1351.99 209.15 19 to 152
Using these constants, we then can calculate the vapor pressure at any temperature from:
Bi
log10  Pio (T )=
 A −
 i (141)
T + Ci
In this equation, the temperature T is

expressed in ºC and the vapor
pressure is obtained in mmHg (torr).
This semi-empirical equation is called
Antoine’s equation. At right is a
graph showing Pio (T ) for all three
components.

06-361 page 153 Spring, 2017
Dew Point Evaluation
Recall that, as temperature drops, the “dew point” is the temperature at which the first drop
of dew is formed in the vapor. Usually, the composition of the vapor is specified as yi for i = 1, 2
… Nc. At the dew point, the vapor having composition yi is in equilibrium with that first drop of
dew having composition xi. Knowing yi, we can calculate xi at any T (for which two phases
coexist) from (136):
yi
xi =
Ki (T )
but T must be chosen so that the mole fractions add up to unity:
Nc
y
∑ K (Ti ) = 1
i =1 i dew
This is a trial-and-error for T. For a particular guess of T, the K-values can either be read from a
graph (like Fig. 11.7-2 in Geankoplis) or they can be calculated from vapor pressure data using
(139). Substituting Ki from (139) into the equation above and dividing through by P:
Nc
yi 1
∑ =
o
(
i =1 Pi Tdew ) P
To evaluate the dew point of the distillate, we choose yi = xDi and P = 1.2 atm. The dew
point is found to be 75.2°C [see Bubble point and dew point of multicomponent mix.xmcd].
This Mathcad document uses vapor pressures calculated from (141). At this temperature, the
vapor pressures are evaluated and the relative volatility is calculated from (140). The results are
summarized by
Component yi = xD Pio (atm) αiD, HK

C6 0.990 1.219 2.547
C7 0.010 0.478 1.000
C8 0.000 0.192 0.401
Bubble Point Evaluation
As temperature rises, the bubble point is the temperature at which the first bubble of vapor is
formed. Usually, the composition of the liquid is specified as xi for i = 1, 2 … Nc. Knowing xi,
we can calculate yi at any T from (136):

06-361 page 154 Spring, 2017
yi = Ki (T ) xi
but T must be chosen so that the mole fractions add up to unity:
Nc
∑ Ki (Tbub ) xi = 1
i =1
Multiplying this equation through by P and substituting (139), we obtain an alternate equation:
Nc
∑ Pio (Tbub ) xi = P
i =1
Once again, solving either equation for Tbub is a trial-and-error. In this example, the bubble
point of the bottoms is calculated taking xi = xBi and P = 1.2 atm [see Bubble point and dew
point of multicomponent mix.xmcd]. The result is 114.2°C and the results are summarized by
the following table:
Component xB Pio (atm) αiB, HK

C6 0.010 3.436 2.213
C7 0.545 1.553 1.000
C8 0.445 0.718 0.463
Now let’s return to applying Fenske’s equation to this example.
Application #1: Calculate Minimum Stages
To calculate Nmin, Fenske’s equation is applied to the heavy key and the light key. So we are
interested in αLK,HK
α=
67 2.547 × 2.213
= 2.374
Fenske’s equation gives
x x 
ln  Di Bi  ln  0.99 0.01 
 xDj xBj 
=N min =   =0.01 0.544 
9.82
ln αij ( )
ln ( 2.374 )
So at an minimum, we need 10 ideal steps, or 9 ideal trays plus the reboiler.

06-361 page 155 Spring, 2017
Application #2: Check on Negligible Mole Fractions
To the extent that the assumptions made during its derivation are valid, Fenske’s equation
should yield the same value for Nmin using any choice of component i and j. Once the value of
Nmin is known, we can choose an i or j corresponding to a component whose mole fraction was
assumed to be zero for the purpose of estimating the product distribution. Then Fenske’s
equation can be used to calculate the actual mole fraction of that component as a check to see
whether the mole fraction is indeed small enough to be neglected.
For example, in Example #1, we assumed that the octane fraction in the distillate is
negligible. Let’s now check to see what the octane mole fraction actually is. We use i=8
(octane) in Fenske’s equation rather then i=6 (hexane, the light key):
α=
87 0.401 × 0.463
= 0.431
x 0.445 
ln  D8 
= 9.82
N min =  0.01 0.545 
ln ( 0.431)
Solving for xD8, we obtain xD8 = 1.93×10-6
which indeed is negligible compared to the other mole fractions in the distillate.
MINIMUM REFLUX RATIO
Method 1) Pseudo-Binary Method
Assuming a saturated liquid feed (q=1) and constant relative volatility, the minimum reflux
ratio turns out to be
xD 1 − xD
− α AB
z 1 − zF
Rmin = F (142)
α AB − 1
where xD and zF are the mole fraction of the light key in the pseudo-binary mixture and 1-xD and
1-zF are the mole fraction of the heavy key in the pseudo-binary mixture:
z F , LK z F , HK
zF = and 1 − z F =
z F , LK + z F , HK z F , LK + z F , HK

06-361 page 156 Spring, 2017
where zF,LK and zF,HK are the mole fractions in the multi-
component mixture.*
Proof: We can obtain the minimum reflux ratio from the point of
intersection of the q-line and the equilibrium curve. Recall (117)
from page 132:
x − y′
Rmin = D (143)
y ′ − x′
which was derived using two points on the ROL to determine the
slope and knowing that the slope is R/(R+1). For a saturated
liquid feed, the q-line is vertical so that
x′ = zF (144)
The y-coordinate of this point is determined knowing that ( x ′, y ′ ) lies on the equilibrium curve.
For constant relative volatility, we can express the equilibrium curve as a simple equation using
the definition of relative volatility:
K A y A x A y (1 − y )
α AB ≡ = =
K B y B xB x (1 − x )
αx
Solving for y: y=
1 + ( α − 1) x
x x=′ z F , we can calculate

Substituting =
α AB z F
y′ = (145)
1 + ( α AB − 1) z F
Substituting (145) and (144) into (143) leads to (142).
Method 2) Underwood's Method
This method makes assumptions similar to those used in Fenske’s equation. In particular, the
assumptions made by Underwood’s method are:
1) equimolal overflow
2) constant relative volatility
3) a sharp split with only one “pinch”
* To derive the relations above, take a basis of 1 mole of feed. then z

F,LK and zF,HK represent the number of
moles of LK and HK in the feed.

06-361 page 157 Spring, 2017
First, we will just state the method and then we will take a quick look at its proof.
Recipe:
Step 1) Find the root (denoted φ) in the interval α HK < φ < α LK which satisfies the following
polynomial (order of polynomial = number of components)
Nc
αz
∑ α i −Fiφ = 1− q (146)
i =1 i
Step 2) Once we have φ, we can compute the minimum reflux ratio from
αx
Rmin + 1 =∑ αi −Diφ (147)
i i
where the sum is just over those components present in the distillate (i.e. we ignore components
heavier than the heavy key).
Proof: A mole balance on component i about the top n stages of the

rectification section yields:
Vn+1yn+1,i = Lnxn,i + DxD,i (148)
Recall that the distribution coefficient is defined as:
yn , i
Ki = yi/xi or K n, i =
xn, i
We added the tray subscript n to K because the value of K depends

on the temperature, which varies from tray to tray. Thus (148) can
be rewritten as:
yn , i
Vn +1=
yn +1, i Ln + DxD, i
K n, i
For constant molal overflow, the flowrates will be the same for each
tray in the rectifying cascade, just as with binary mixtures. Thus we can drop the subscripts on
flowrate.
yn , i
Vy=
n +1, i L + DxD, i (149)
K n, i
When we approach the minimum reflux ratio, we will encounter a "pinch" somewhere in the
column. At the pinch, we take an infinite number of steps without changing the mole fractions:

06-361 page 158 Spring, 2017
at a pinch: yn+1,i = yn,i (150)
(150) into (149) and solving for yi:
D xDi
yi = (151)
V∞ L∞
1−
V∞ K ∞, i
We have added the subscript “∞” to each flow rate to emphasize that this corresponds to the
minimum reflux, which requires an infinite number of trays. If we require (151) to be satisfied
for each component and sum over all components:
D xDi
∑ yi = 1= ∑
V∞ i L∞
(152)
i 1−
V∞ K ∞, i
Now we can write Ki in terms of the relative volatility:
K
αi = i
K ref
While the flowrate ratios in the rectifying section are related to the reflux ratio by (103):
L∞ Rmin D L∞ R 1
= and= =
V∞ Rmin + 1 V∞ V∞ Rmin + 1
αx
(152) becomes: ∑ αi=
Di
−φ
Rmin + 1 (147)
i i
Rmin
where φ≡
K ref ( Rmin + 1)
If Kref were known, then Rmin could be determined as the root of this equation. Unfortunately, to
calculate Kref we need to know the temperature at the pinch, which is not known.
Generally there is also a pinch in the stripping section of the tower. For sharp splits, the
temperature at this second pinch will be the same. We can then derive a second equation like
(147) by performing our balances about the bottom n stages of the column to obtain a second
equation.

06-361 page 159 Spring, 2017
B xBi
∑ yi = 1= ∑
V∞ i L∞
i −1
V ∞ K ∞, i
where the overbars denotes flowrates in the stripping section. Then we will have two equations
and two unknowns: Kref and Rmin. Their solution is the value of φ which satisfies the following
equation:
αi z F , i
∑α = 1− q
i i −φ
where zF,i is the composition of the feed and q is its thermal condition. There are several φ's
which satisfy this equation. It turns out that we want:
αHK < φ < αLK
Once φ is determined, we can calculate Rmin from (147). This is called Underwood's method.
Example #4. Calculate the Rmin for the separation in Example #1 if the feed is 60% vapor (q =
0.4).
Solution: Recall the feed and product compositions from

Example #1 (see sketch at right). Using Underwood’s method,
we first determine the root of (146).
α z
∑ α i −Fiφ = 1− q
i i
Substituting into known values:
2.399 × 0.33 1 × 0.37 0.434 × 0.30

+ + =
1 − 0.4
2.399
− φ 1 − φ
0.434 − φ

f ( φ)
which is a cubic equation. The left-hand side of this equation [defined as f (φ)] is plotted below.

06-361 page 160 Spring, 2017
This equation has three singular points at which the function diverges to ±∞:
φ = α1, α2 and α3
These are shown as the 3 vertical lines in the plot. Notice that there are also 3 different φ’s for
which
f (φ) = 1 – q = 0.6
We are interested in the root in the interval:
α HK < φ < α LK or 1 < φ < 2.399
which turns to be φ = 1.739. Once we have the value of φ, we calculate the reflux ratio from
(147):
αx
Rmin + 1 =∑ αi −Diφ
i i
where the sum only includes those components present in significant quantities in the distillate.
In this case, octane (which is heavier than the heavy key) was neglected in the distillate for
Example #1. It should also be neglected in this sum, although in this case, it’s inclusion does not
change the result.
2.399 × 0.99 1 × 0.01

=
Rmin + 1 =
+ 3.59
2.399 − 1.739 1 − 1.739
Answer: Rmin = 2.59

06-361 page 161 Spring, 2017
NUMBER OF IDEAL PLATES AT OPERATING REFLUX

Ref: McCabe, Smith & Harriott, Unit Operations of Chemical Engineering, 5th Ed. p588-609
Method 1) Underwood's method
Underwood's method can be extended to calculate the number of plates analytically when the
equal molal overflow assumption and constant relativity are appropriate. But even then, the
calculations are sufficiently involved that a computer would be helpful. If we are going to use a
computer, we might as well do the rigorous tray-by-tray calculations (see Method 3); thus we
will not extend Underwood’s method further in these Notes.
Method 2) Erbar & Maddox correlation
An empirical but much easier to use method is the Erbar-Maddox correlation. This is an
empirical relation between the number of ideal stages and the operating reflux ratio. You also
need to know the minimum number of ideal stages (at total reflux) and the minimum reflux ratio.
Because of its simplicity, this correlation is widely used for preliminary estimates.
R N R 
= f  min , min 
R +1  N Rmin + 1 
The function is presented in Geankoplis at Fig. 11.7-3 (except subscript “min” has been replaced
by “m”).

06-361 page 162 Spring, 2017
Method 3) Gilliland correlation
A second type of empirical correlation is the Gilliland Correlation:
N − N min  R − Rmin 
= f 
N +1  R +1 
The function is presented graphically just below.
lim N →∞
R → Rmin
R − Rmin lim N → N min

R →∞
R +1
FEED-PLATE LOCATION
The Kirkbride equation provide an approximate method for estimating the feed plate
location once the total number of trays Ne+Ns has been estimated for the operating reflux ratio:
 2  0.206
N e  z F , HK B  xB, LK 

= × ×  (153)
N s  z F , LK D  xD, HK 

 
where Ne is the number of ideal stages in the enriching section (above the feed) and Ns is the
number of ideal stages in the stripping section (below the feed).
In previous section, we presented a method for estimating the total number N of stages,
which also equals

06-361 page 163 Spring, 2017
N = Ne + Ns (154)
Thus (153) and (154) represent two equations in two unknowns: Ne and Ns. Of course, Ne
represents the number of the feed tray (counting from the top of the column).
RIGOROUS SOLUTION OF MESH EQUATIONS

Reference: Henley & Seader, Equilibrium-Stage Separation Operations in Chemical
Engineering, Chpt. 15.
The final design of multistage equipment to multicomponent distillation usually requires

rigorous determination of temperatures, flow rates, and compositions at each stage. A rigorous
mathematical description of a multistage cascade involves material and energy balances and
some representation of the vapor-liquid equilibrium. The resulting set of algebraic equations is
called the MESH Equations, for reasons which will soon become apparent. Their solution must
be done numerically by computer.
Consider a general stage in the cascade, which

we will label stage n. It has the usual liquid and
vapor streams entering and leaving. It might also be
a feed stage, in which case Fn ≠ 0. If it is not a feed
stage, we just set Fn = 0. There might also be liquid
or vapor product streams Un or Wn as well as a heat
removal Qn.
The MESH equations result from balances done

about this arbitrary stage.
• M Equations are Mole balances for each component i (Nc equations for each stage):
Ln −1 xi, n −1 + Vn +1 yi, n +1 + Fn zi, n = ( Ln + U n ) xi, n + (Vn + Wn ) yi, n
Note that the first subscript is the component i and the second subscript for the stage n.
• E Equations describe the liquid-vapor Equilibrium relations among the two outlet
streams, which are assumed to be at equilibrium (Nc equations for each stage):
yi, n = Ki, n xi, n
where Ki,n is the distribution coefficients.
• S Equations are the mole fraction Summations (2 equations for each stage):
∑ xi, n = 1 and ∑ yi, n = 1

i i

06-361 page 164 Spring, 2017
• H Equations is the energy balance (H stands for enthalpy) for each stage (1 equation for
each stage):
Ln −1 H L, n −1 + Vn +1 HV , n +1 + Fn H F , n = ( Ln + U n ) H L, n + (Vn + Wn ) HV , n + Qn
If there are a total of Ns stages in the column and Nc components, then the number of equations
is
number of equations = Ns(2Nc +3)
Now we consider the following to be unknowns:
xi,n, yi,n, Ln, Vn, Tn for i = 1,...,Nc and n = 1,...Ns
where Tn is the temperature of the liquid and vapor streams leaving stage n.
number of unknowns = Ns(2Nc +3)
Of course, there are more variables involved in the equations. We assume that the feeds and any
side draw-off flows are specified:
given: Fn, zi,n, HF,n, Un, Wn, Qn
Also, we assume that the thermodynamic relations exists so that we can compute the enthalpy of
saturated liquid and saturated vapor, given the composition, pressure and temperature:
known functions: (
H L, n = H L, n Tn , Pn , x1, n  xNc , n)
HV , n = HV , n (Tn , Pn , y1, n  y N , n )
c
Ki, n = K (Tn , Pn ; x1, n  xN , n )

c
Testing the McCabe-Thiele Method
In this section we will use the McCabe-Thiele* solution for the number of stages required to
meet a specified distillate and bottoms compositions and solve the MESH equations to see if the
specified distillate and bottoms compositions can actually be obtained using the number of trays
predicted.
*Warren Lee McCabe was Head of the Department of Chemical

Engineering at Carnegie Institute of Technology from 1938-1947.

06-361 page 165 Spring, 2017
As our test case, we are going to start with the

McCabe-Thiele solution to Prob. 3 on Hwk. #6. This
problem requires us to separate a mixture of acetone
and methanol. The solution we found on the home-
work is summarized in the figure at right.♣ Now we
will try to check that solution using a numeric solution
of the MESH equations.♦
The main assumption of the McCabe-Thiele

solution is "equimolal overflow." In Constant Molal
Overflow starting on page 110, we showed that if 1)
you neglect sensible heat changes compared to latent
heat and 2) you assume that the latent heat of
vaporization is independent of composition, then the molar flowrates of liquid and vapor streams
do not vary from stage to stage within either of the two cascades composing the column (of
course the flows do change at the feed tray).
The plot at right shows the enthalpies (on

the horizontal axis) versus stage number (on
the vertical axis). For equimolal overflow, the
enthalpy profiles on this plot should be two
vertical lines. Clearly, they are not vertical
lines. The departure arises because the heat of
vaporization of acetone and methanol are not
equal. The following values were read from
the ChemSep library:
λacetone = 29.5 kJ/mol
λmethanol = 35.3 kJ/mol
How important are these differences?

Let’s look at the flowrates. The graph below
compares the flowrates obtained from the
Chemsep simulation (shown as the points) with those calculated from the equimolal overflow
assumption (shown as lines).
♣ Actually, the problem whose solution is presented here is slightly different from Hwk6 Prob 3. The
thermodynamic model used by Chemsep produces a slightly different equilibrium curve. In particular, the azeotrope
occurs at a mole fraction of 0.76 instead of 0.80. Thus the distillate composition of 0.78 specified in the original
problem cannot be achieved. So we adjusted the distillate composition to 0.715 and have imported Chemsep’s VLE
data into Mathcad.
♦ Our solution is obtained using CHEMSEP, a process simulator available through the CACHE of AIChE. For any
student who is interested, this is available at www.chemsep.org.

06-361 page 166 Spring, 2017
The discontinuity of flowrates on the 17th stage occurs because this is the feed tray. Because the
feed is 30% liquid, the liquid flowrate in the bottom cascade is higher than in the top; because
the feed is 70% vapor, the vapor flowrate in the top cascade is higher than in the bottom.
Beyond the expected changes in flowrates at the feed tray, there is a continuous drop in flowrate
of both streams as we move down either the rectifying cascade or the stripping cascade. These
changes in flowrates within a cascade are NOT consistent with the equimolal overflow
assumption.
These variations in flowrate within a cascade stem from the fact that the enthalpy of saturated
liquid is not independent of temperature or acetone content as is required for "equimolal
overflow."
How important are these flowrate differences? On the McCabe-Thiele diagram, these
differences show up as curvature in the operating lines. In the graph below we compare the
slopes of the operating lines.

06-361 page 167 Spring, 2017
While the variations in slope look significant, actually plotting the operating lines themselves
shows the difference is not too great:

06-361 page 168 Spring, 2017
The plot above shows points (the red diamonds) taken from the operating lines in the
rigorous ChemSep separation and compares them with the ROL and SOL calculated using the
McCabe-Thiele method. These calculated operating lines are the straight, solid lines on the plot.
Connecting the red diamonds with a smooth curve would yield the actually operating lines.
There is very little difference; indeed all the points appear to lie on one of the calculated
operating lines.
The main prediction for this problem is what the distillate composition and flowrate will be
xD D/F
Model
McCabe 0.715 0.3404
Chemsep 0.7131 0.3414
The difference between the two is insignificant. This insignificant difference is typically the
case when the molar heats of vaporization differ by less than 20%.
Another interesting profile which can be

plotted by ChemSep is temperature. The plot at
right is obtained for the current problem. Notice
that the hotter temperature occur at the bottom of
the column. This is because the bottoms contain
the less volatile component which has the higher
boiling point.
To summarize, the McCabe-Thiele method

predicted that the desired split (0.715 in distillate
and 0.01 in bottoms) could be obtained with a
reflux ratio of 8.62 and 23 ideal stages, with the
feed on the 17th stage. Using a better
thermodynamic model for enthalpies, Chemsep
predicted that this configuration would produce a
distillate only slightly less rich: 0.713 instead of
0.715. This essentially confirms the validity of
the McCabe-Thiele method for estimating the
number of stages required for a given reflux ratio.
The main error made by the equimolar overflow assumption is estimating the heat duties
which are proportional to the vapor flowrates. Recalling the flowrates graph on page 166, we
find (for a feed flowrate of 100):
Model V V
McCabe 327.5 257.5
Chemsep 328.3 210.5
Chemsep predicts about the same flowrate through the condenser, but predicts a reboil rate about
18% below that obtained using the equimolar overflow assumption.

06-361 page 169 Spring, 2017
Testing Underwood & Fenske
Next we will use Chemsep to test the

predictions of the simple equations for the number
of trays required for multicomponent distillation.
We will use Prob. 11.7-4 which you will also solve
on Hwk #8. A quaternary mixture of alkanes is to
be separated at 4 atm using a reflux ratio of 1.3
times the minimum such that a sharp split is
achieved between pentane and hexane. Other
information is summarized in the sketch at right.
You estimated the product distribution for this

problem in Hwk #7 Prob. 3. The results are
summarized in the table below.
Component FzFi zFi DxDi xDi BxBi xBi α

butane 40 0.40 40.00 0.6178 0 0 6.71
pentane 25 0.25 23.75 0.3668 1.25 0.0355 2.50
hexane 20 0.20 1.00 0.0154 19.00 0.5390 1
heptane 15 0.15 0 0 15.00 0.4255 0.421
totals = 100 1 64.75 1 35.25 1
Also shown in the table above are the average relative volatilities. The values shown were
determined from the thermodynamic model of Chemsep, rather than the vapor pressure
correlations presented in the problem statement (Hwk #8, Prob. #1), in order to provide a fair
comparison between Fenske’s equation, Underwood’s method and the more rigorous predictions
of Chemsep. However as a consequence of using a different thermodynamic model here, the
remaining results will differ slightly from your solution to the homework problem.
Using Fenske’s equation, we predict Nmin = 6.42. Using this value of Nmin we can then use
Fenske again to check on the two ‘zeros’ of mole fractions in the table. The results are
xD,heptane = 5.1×10-5 and xB,butane = 1.2×10-4
These are small enough to be ignored. Underwood’s method predicts
Rmin = 0.403
Using an operating reflux of 1.3 times this, we have
R = 0.524
Using the Erbar-Maddox correlation we predict that 15.6 (round

to 16) ideal stages are required while the Kirkbride equation
suggests that 8.5 (round to 9) of them are required in the

06-361 page 170 Spring, 2017
rectifying section. All of this info is summarized by the sketch at right.
Chemsep was supplied with most of this info and asked to calculate the product compositions
for a reflux ratio of 0.524 and a recovery of 95 of the pentane in the bottoms. The main results
are compared below with our predictions:
Distillate Bottoms
Component Hwk7 Chemsep Hwk7 Chemsep
butane 0.6178 0.6183 1.2×10-4 4×10-5
pentane 0.3668 0.3663 0.0355 0.0370
hexane 0.0154 0.0155 0.5390 0.5382
heptane 5×10-5 1×10-5 0.4255 0.4248
total flowrate 64.75 64.69 35.25 35.31
The agreement is very good. Let’s take a closer look at some of the assumptions. A major
assumption in both Fenske’s equation and Underwood’s method is equimolar overflow. This is
an assumption about the enthalpies. Below is a plot of the enthalpies of the liquid and vapor
streams
Both streams show variations in enthalpies, especially below the feed. Most of the changes in
temperature and composition occur below the feed:

06-361 page 171 Spring, 2017
Notice from the graphs above that the changes

in composition are not always monotonic.
The main consequence of the enthalpy

assumption is that the liquid flows are the
same for each stage of each of the two
cascades in the column. Same for vapor
flows. The figure at right shows that the flows
are not quite constant.
The second major assumption in both

Fenske’s equation and Underwood’s method is
that relative volatilities are constant. The
figure at left tests that assumption. Clearly,
there are significant variations in relative
volatility over the height of the column.

06-361 page 172 Spring, 2017
Chapter 13. Gas Absorption/Stripping

gas absorption -- 2 components of a gas are separated by contact with
a liquid (in which one component is preferentially soluble)
stripping -- 2 components of a liquid are separated by contact with a

gas
An example of gas absorption is the removal of

ammonia from air by contact with liquid water.
Ammonia is very soluble in water whereas air is only
slightly soluble. Stripping is the opposite of
absorption. An example of stripping is the removal
of CO2 from water by contact with nitrogen gas.
While some water also evaporates, this is usually
negligible.
Both gas absorption and stripping involve at least

three components. Usually only one of these
components crosses the phase boundary. In the
example of ammonia and air, ammonia is the
component whose molar flowrate changes by the
largest percentage of the inlet value. Although some
air will also dissolve in water, and some water will
evaporate into the air, the molar flowrates of air and
water change by negligible fractions: their flows can
usually be considered constant.
These are the two main differences between

absorption/stripping and distillation: 1) at least 3
components, and 2) often only "one transferrable
component". By contrast, in distillation, all of the
components are present in both phases.
EQUIPMENT FOR ABSORPTION/STRIPPING

Liquid and gas streams for absorption or stripping
could be contacted using a tray column like that used
in distillation. Instead of tray towers, we are going to
look at the design of packed towers. Packed towers
are a reasonable alternative to tray towers in

06-361 page 173 Spring, 2017
situations in which the tray efficiency is low (perhaps only 5% instead of 50%). Because of this
very low efficiency, very large numbers of trays would be required -- perhaps 100's or 1000's.
Fabrication costs just become prohibitively expensive. Fortunately, a viable alternative exists:
the packed tower.
A packed tower is simply a tube or pipe, which is filled with some sort of "packing." The
packing typically consists of particles around an inch in diameter. In commercial packed towers,
the usual choice are particles with one of three different shapes:
• Raschig ring (which is just a piece of pipe which has been cut into segments, whose length
and diameter are about the same)
L ≈ D ≈ 1 to 1 1 inches
2 2
• Berl saddle
• Pall ring
Although, in a pinch, almost anything you have laying around would do -- ping-pong balls, golf
balls, etc. The purpose of the packing is to promote good contact between the liquid and vapor
streams which are being brought together to permit interfacial mass transfer. In particular, what
is desired is a large interfacial area per unit volume.
The liquid stream is usually fed into the top of the tower while the vapor is fed into the
bottom. Thus we have countercurrent flow of the two streams, which has the same advantages
for mass transfer as it did for heat transfer.
The packing promotes good contact between the phases by dividing the two feed streams into
many parallel interconnected paths. Ideally, you would like the liquid to flow downward as a
thin film over the surface of the packing. This would give the maximum surface area of contact
between the gas and liquid.

06-361 page 174 Spring, 2017
If you just pour the liquid from the end of the pipe onto the top of the packing in tower
having much larger diameter than the pipe, most of the packing will not even be wet. Only some
of the channels will be carrying flow. This is called:
channeling -- maldistribution of liquid flow
So some sort of device to distribute the flow over the entire cross section of the tower is needed.
This device is called a distributor.
Even if the flow is evenly distributed at the top of tower, channeling might still develop as
the fluid trickles down. When two thin films converge they tend to form a thick film and a dry
patch, which results in a reduction in contact area. So redistributors are placed every 10-15 feet
along the length of the tower.
A TYPICAL ABSORBER DESIGN PROBLEM

To motivate the next few lectures, let’s pose a typical design problem. For this example, I'm
going to take Prob. 22-1 from McCabe, Smith & Harriott.
Problem: treat 500 SCFM of air containing

14 mol% acetone to remove 95% of the
acetone by absorption in liquid water in a
packed bed operating at 80°F and 1 atm, with
1-inch Raschig rings. The feed water contains
0.02% acetone and the flowrate is 1.08 times
the minimum.
As the designer, you must choose the

following:
• flowrate of water
• diameter of tower
• height of packing

06-361 page 175 Spring, 2017
Solution Overview:
Choice of liquid flowrate: L = 1.1 to 1.5 times Lmin
Just as with distillation, there is a minimum L/V required to achieve the desired separation.
In typical operation of an absorber, the liquid rate is chosen to be just above the minimum.
Tower diameter: GG = 0.5 to 0.7 times GG,flood
The diameter of the tower is usually chosen on the basis of gas mass velocity GG. Generally,
for a particular L and V, the smaller the column diameter, the larger the mass velocities will be,
and the larger the pressure drop. Generally, large mass velocities are desirable because they give
high mass transfer coefficients, but too large mass velocities cause “flooding” which severely
decreases mass transfer rates. The gas mass velocity above is about as close to flooding as you
dare get.
b y
V S dy
Tower height:* ZT =
Kya ∫ y − y*
 ya
  
H Oy
NOy
The height of the tower is determined by mass transfer rates. Basically, the gas and liquid
phases need to be in contact for a certain time for the acetone to have time to diffuse from the gas
phase into the liquid. The equation above is called the "design equation". The integral
yb
dy yb − ya
NOy ≡ ∫ y − y*
=
∆y
ya
is called the number of transfer units, where y is the mole fraction of the transferable
component, y – y* is the local driving force for mass transfer and ∆y is some average driving
force (perhaps the log-mean of the driving force at the top and bottom). The number of transfer
units is loosely analogous to the number of ideal trays required: both can be thought of as
measures of the difficulty of the separation.
Since NOy is dimensionless, to get units of height, the remaining factor in the design equation
(defined to be HOy) must have units of height and is called the height of one transfer unit:
V S
H Oy ≡
Kya
* This particular formula is for dilute solutions, which is probably not really applicable to the current example, but it
is applicable in the homework problems.

06-361 page 176 Spring, 2017
where Kya is the overall mass transfer coefficient times the interfacial area per unit volume of
packing.
Detailed Solution:
First of all, what is “SCFM”? This stands for “standard cubic feet per minute”
SCFM = standard (0°C, 1 atm) cubic feet per minute
where “standard” means that the volume is evaluated at standard conditions. We can convert
SCFM into moles. Starting with the ideal gas law,
PV = nRT
V RT R ( 273° K ) liters ft 3
or = = = 22.4 = 359
n P (1 atm ) gmol lbmol
depending of which set of units are used to express R. Recall that a gmol is the quantity of
material whose mass equals the molecular weight in grams. Similarly, a lbmol is the quantity of
material whose mass equals the molecular weight in pounds. For example, air has a molecular
weight of 29:
g lb
MW= = 29
of air 29
gmol lbmol
Thus we can convert 500 SCFM into a molar flowrate:
500 ft 3 min lbmol

= 1.39 = V
359 ft 3 lbmol min
Now let’s move on to determine the liquid flowrate required. Recall that in distillation the
slope of the ROL is given by
L R
=
slope of ROL =
V R +1
You should also recall that there exists a minimum value of the reflux ratio for any given
combination of product and feed specifications. This minimum value of R translates into a
minimum allowable L/V. There is a similar minimum L/V for gas absorption. To find the
minimum reflux ratio in distillation, we needed to plot an operating line and an equilibrium
curve.

06-361 page 177 Spring, 2017
EQUILIBRIUM CURVE
In Prob. 22-1, McCabe gives us equilibrium data in the form of partial pressure of acetone in
the gas phase, pA, as a function of the mole fraction of acetone in the liquid, x:
o
p=
A PA γ A x (155)
PAo = vapor pressure of acetone at 80°F = 0.33 atm
γA = activity coefficient for liquid mixture
Activity coefficient is a measure of nonideality of the liquid phase. For ideal solutions the
activity coefficient is unity:
γA = 1 (ideal solution)
For ideal solutions and γA = 1, (155) reduces to Raoult’s law. Our solution of acetone in water is
not ideal, but McCabe kindly gives us a model
A 1.95 (1 − x )
2
ln γ=
Note that γA→ 1 (i.e. lnγA → 0) as x → 1. This is a 0.18
general rule: pure components always behave 0.16

0.14
ideally since the partial pressure pA exerted by a
0.12
pure component is its vapor pressure PAo [for (155) 0.1
y
to predict this for x=1 we must require that γA=1 0.08
0.06
for a pure component]. We can obtain an
0.04
expression for the mole fraction of acetone in the
0.02
gas just by dividing the partial pressure in (155) by
0
the total pressure:
0 0.02 0.04 0.06 0.08 0.1 0.12
x
× xe ( )
2
p A 0.33 atm 1.95 1− x
=
y =
P 1 atm
where we use P = 1 atm. Repeating this for different x’s to obtain y’s up to 0.14 (the feed
concentration), we obtain the curve at right, where
x = mole fraction of acetone in the liquid (2nd component is water)

y = mole fraction of acetone in the gas (2nd component is air)

06-361 page 178 Spring, 2017
OPERATING LINE
Suppose we are trying to absorb acetone from a mixture of air and acetone by contacting the
air mixture with water. Let
L,V = total molar flowrates
x,y = mol. frac. of transf. comp.
where the transferrable component is acetone in this

example. Performing a component balance on the
transferrable component about the top section of the
tower yields:
Laxa + Vy = Lx + Vaya (156)
This is identical to (96): its counterpart in

distillation (see page 110). Solving for y:
L V y − La xa
=
y x+ a a
V V
This relationship y(x) between the mole fractions in the liquid and gas streams is again known as
the operating line, since the relationship is imposed by a component mole balance.
Unlike distillation (where the equimolal overflow assumption makes the ratio L/V constant
within any cascade), L and V are not constant over the height of the absorber: acetone is
transferred from the gas to liquid, making L and V change. In particular, their ratio changes.
Thus the operating line is not straight. This makes it difficult to determine the minimum water
flowrate.
There is one limiting case where we can easily predict the change in L and V:
Only One Transferable Component
It is often possible to assume that only one component is undergoing transfer between the
liquid and gas streams. For example, in our problem acetone is being transferred from the air to
the water:
transferable: acetone
non-transferable: air, water
Although some of the water will evaporate when it contacts the air and some of the air will
dissolve in the water, the molar rates of transfer of these components can often (but not always)

06-361 page 179 Spring, 2017
be neglected compared to rate of acetone transfer. When you can neglect the molar transfer rate
of all but one component, then significant simplification can be made. If we can neglect
evaporation of water and dissolution of air, then the moles of water (the nontransferable
component) in the liquid stream must be the same at all elevations:
L′
(1-x)L ≡ L' = const so L= (157)
1− x
In the above equation and what follows, a prime (' ) is used to denote a flowrate or mole
fraction of the nontransferable component. Similarly, the moles of air in the gas stream must be
the same at all elevations:
V′
(1-y)V ≡ V' = const so V= (158)
1− y
L ′xa V ′y L ′x V ′ya
(157) and (158) into (156): + = + (159)
1 − xa 1 − y 1 − x 1 − ya
where L',V' are constants. Solving this equation for y(x):
(1 − ya )( x − xa ) R + ya (1 − xa ) (1 − x )
y ( x) =
(1 − ya )( x − xa ) R + (1 − xa ) (1 − x )
L'
where R≡
V'
The figure at right shows several operating lines

corresponding to different values of R (equally
spaced from R = 3.82 for the upper most red curve
to R = 1.91 for the lowest). The equilibrium curve
is shown in black. All curves were drawn with the
help of Mathcad. While it is hard to tell from the
figure, none of the curves in is a straight line.♦
Then it becomes a trial-and-error process to
determine Rmin. There is an easier way, which
avoids this trial-and-error.
If we express concentrations in terms of mole

ratios instead of mole fraction: Figure 14
x moles of A in liquid
X ≡ =
1 − x moles of non-A in liquid
♦ To confirm that the lines are not straight, align a straight-edge (e.g. a ruler or the just the edge of a paper
envelope) with any of the red lines.

06-361 page 180 Spring, 2017
y moles of A in gas
Y≡ =
1 − y moles of non-A in gas
then the operating line (159) becomes very simple:
L'Xa + V'Y = L'X + V'Ya
L′  L′ 
or Y= X +  Ya − X a  (160)
V′  V′ 
Note that mole ratios need not be smaller than unity:
0≤x≤1
but 0≤X≤∞
although they are in this example. Since

L'/V' = constant, this is the equation of a
straight line. Thus for the special case of
one-transferable component, the operating
lines are straight on mole ratio
coordinates.
One point on the operating line is

(Xa,Ya). We are told that the inlet water
contains a small amount of acetone:
xa
xa = 0.0002 →=
Xa = 0.0002
1 − xa
The concentration of acetone in the gas

phase is determined from the specification
that we want to remove 95% of the
acetone from the feed:
yaVa = 0.05 ybVb
V′ V′
ya = 0.05 yb
1 − ya 1 − yb
Since V' (the molar air flowrate) is the same at

either end of the column, we can cancel it out,
leaving

06-361 page 181 Spring, 2017
= =
Ya 0.05 Yb 0.00814

0.14
= 0.1628
1− 0.14
Ya
or =
ya = 0.00807
1 + Ya
The second point on the operating line

must lie somewhere along the line
Y = Yb = 0.1628
When you plot this up, you find that the

equilibrium curve is below the operating
line. This is generally true for gas
absorption and makes sense when you
realize that the gas must be richer in the
ammonia than at equilibrium — otherwise
the acetone would not spontaneously
absorb into the liquid.
Drawing a straight line through (Xa,Ya)

= (0.0002, 0.0081), we decrease the slope (starting at vertical) until the line touches the
equilibrium curve at any point in the interval Ya ≤ Y ≤ Yb. While the “touch” or “pinch” usually
occurs at Y = Yb, this is not necessary. In this particular case, the “pinch” occurs at a point in the
middle of the interval Ya ≤ Y ≤ Yb.
Extending this straight line from the pinch up to Y = Yb, we locate a second point on the line:
(Xb,max,Yb) = (0.081, 0.1628). Using the two points (Xa,Ya) and (Xb,max,Yb), we can calculate
the slope corresponding to the minimum water flowrate:
 L′  0.1628 − 0.00814
=
 ′ = 1.91
 V min 0.081 − 0.0002
′ =
Lmin 1.91V ′ = {
1.91 × 1.39
lbmol
min
(1 − 0.14 ) =}
2.29
lbmol
min
=4.94
gal
min
Multiplying this by the molecular weight of water (18 lb/lbmol) and dividing by the density of
water (8.33 lb/gal) yields a minimum water flowrate of 4.94 gal/min.
INTERFACIAL MASS TRANSFER: REVIEW

An important design parameter is the depth of packing ZT required, which we will soon see
can be calculated from a design equation like the following:

06-361 page 182 Spring, 2017
yb
V S dy
ZT =
Kya ∫ y − y*
ya
where V is the molar flowrate of gas, S is the cross-sectional area of the tower and Kya is a mass
transfer coefficient. The denominator of the integrand, y-y*, is the local driving force for mass
transfer; on an xy diagram, it is also the vertical distance between the operating line y(x) and the
equilibrium curve y*(x).
In distillation, the tower height was determined by the number of plates required times the
plate separation (which is usually 1-2 feet). The number of plates required for a given separation
is determined by the operating and equilibrium lines. For packed towers, the height also depends
on the operating and equilibrium lines, but in addition it is inversely proportional to the mass
transfer coefficient, which in distillation plays only a minor role in determining plate efficiency.
Now that we have established the importance of interfacial mass transfer in packed towers,
let's talk about modelling mass transfer across a phase boundary. There are two main differences
between interfacial heat transfer (as employed in Hxer design) and interfacial mass transfer:
1. overall driving force

2. reference frame for flux
Let's start by recalling the driving force for heat

transfer across an interface. Suppose I contact a hot gas
with a cold liquid. The temperature profile near the
interface will look something like that shown at right.
There are two characteristics of this sketch which are
important:
1. heat flows from high to low temperature
2. temperature is continuous across the interface

06-361 page 183 Spring, 2017
Interfacial mass transfer is similar to interfacial heat

transfer, but it is also different.
1. mass transfer occurs in the direction from high to

low chemical potential (not necessarily from high
to low concentration*)
2. chemical potential is continuous across interface

(concentration is generally not continuous)
The main difference is evident in the sketch of the

concentration profile near the interface. Note the
discontinuity in the mole fraction at the interface.♣
yi ≠ xi
T yi = T xi = T i
The reason for this discontinuity in concentration across the interface has to do with
thermodynamic criteria for phase equilibrium. Recall:
phase equilibrium: µjV = µjL for j=1,...,Nc
thermal equilibrium: TV = TL
where µj is called the chemical potential which plays

the role of temperature in mass transfer. Unfortunately,
there exists no “thermometer” for measuring chemical
potential. Instead, we are forced to measure chemical
concentration. While chemical potential usually increases with concentration within any given
phase (this is why diffusion of a solute occurs from high to low concentration), when comparing
chemical potentials between two phases, there is no general correlation between chemical
potential and concentration.
To illustrate this, recall the simplest case of VLE: an ideal gas mixture in equilibrium with an
ideal solution. This leads to Raoult's law:
o
=
p j y=
j P x j Pj
* Within a single phase, transport is usually from high to low concentration.
♣ Instead of component i, the superscript “i ” will be used to denote quantities evaluated at the interface between
two phases. In place of i, we will use j to denote components.

06-361 page 184 Spring, 2017
y j Pjo
Note that: = ≠1
xj P
This ratio is never unity (except for pure components).
Thus yj ≠ xj for VLE
even in the simplest case of vapor-liquid equilibrium, the mole fractions are not equal, except in
the trivial case when you have only one component and the total pressure is the vapor pressure.
Definitions of Transfer Coefficients
Recall from the first part of this course, the local heat flux through the interface can be
related to the local temperatures using any one of three types of local heat transfer coefficients:
rate of heat transfer

= hy (Th − Ti )= hx (Ti − Tc )= U (Th − Tc ) (161)
interfacial area
where hx, hy = local, one-phase heat transfer coefficients
U = local, overall coefficient.
Basically these equations say that the rate of heat transfer is proportional to the driving force
(which is the departure from equilibrium) and the proportionality constant is the heat transfer
coefficient.
Similar definitions of transport coefficients can be made for mass transfer:
molar rate of transfer

= k x ( xi − x=
) k y ( y − yi=
) K x ( x * − x=
) K y ( y − y*) (162)
interfacial area
where kx, ky = local, one-phase mass transfer coefficients
Kx, Ky = local, overall mass transfer coefficients
Comparing the mass transport expressions in (162) with their heat-transfer analogues in (161),
there is a good deal of similarity — especially in the one-phase relations. For a single phase, the
driving force is the difference between the concentration or temperature in the bulk and in the
concentration or temperature at the interface (but in the same phase).
But the driving force for the overall coefficients look a little different. The overall driving
force for heat transfer is just the difference between the bulk temperatures of the hot and cold
fluid
Th - Tc

06-361 page 185 Spring, 2017
By simple analogy, you might expect the overall driving force for mass transfer to be the
difference in concentration of the two phases:
y - x ≠ overall driving force
Instead y - y* and x* - x
appear in (162) as the overall driving force. The *s are

defined as shown on the figure at right. Basically y - y* and
x* - x are two different measures of the distance between the
operating line and the equilibrium curve at that elevation in
the packed bed where the liquid and gas concentrations are
(x,y).
x-y does not represent the overall driving force for

mass transfer, because the transport rate does not go to
zero when x-y = 0. The transport rate is zero only at
equilibrium, and x-y ≠ 0 at equilibrium.
Determining the Interfacial Concentrations: (xi , yi )
Although overall mass-transfer coefficients are the

easiest to use in design, correlations are usually more
available in the form of single-phase coefficients. Then
we need to calculate overall coefficients Kx and Ky from
the single-phase coefficients kx and ky. As the first step,
we will need to evaluate the interfacial concentrations xi
and yi, which appear in the definitions above.
Example #1
Given: x, y, kx, ky and the equilibrium curve
Find: xi and yi
Solution: Consider the rate of interfacial transport at some arbitrary

elevation in the absorber, where the local concentrations in the liquid and
gas are x and y and (x, y) is a point on the operating line. At steady state,
the flux of acetone through the gas film must equal the flux of acetone
through the liquid film:
NA = kx(xi-x) = ky(y-yi) (163)
If the fluxes were not equal, we would have acetone building up at the
interface. We will denote this interfacial flux by NA.

06-361 page 186 Spring, 2017
Now the bulk compositions — x, y — are known, together with the two single-phase mass
transfer coefficients. Think of the two interfacial concentrations as two unknowns. We need
two equations. One equation is provided by the requirement that (xi, yi) must lie on the
equilibrium curve. The second relation is (163), which can be re-written as a linear relationship
between yi and xi:
k k
yi =− x xi + y + x x (164)
ky ky
  
slope intercept
One point on this line is (xi, yi) = (x, y) and the

slope is –kx/ky. The intersection of (164) and the
equilibrium curve gives the interfacial concentra-
tions.
Example #2: Next, let’s determine the value of the overall coefficient which leads to the same
flux for these interfacial concentrations.
Given: kx and ky
Find: Ky
Solution: Using the definitions of the k's and of Ky,
NA = kx(xi-x) = ky(y-yi) = Ky(y-y*) (165)
The relationship among the concentrations is shown in the figure above. Adding and subtracting
yi from the overall driving force y-y*:
y − y * = y − y + yi − y * (166)
 i
NA NA
Ky ky
Using (165) we can assign meaning to two of the three differences appear above, leaving:
NA NA
= + yi − y * (167)
Ky ky  
( xi − x ) m
The remaining difference yi-y* can be related to xi-x if we multiply and divide the last term by
xi-x:

06-361 page 187 Spring, 2017
y − y * NA
yi − y* = ( xi − x ) i = m (168)
  xi − x kx
NA  
m
kx
In the second equality above, xi-x is expressed in

term of the remaining single-phase mass transfer
coefficient using (163). Substituting (168) into
(167) and dividing by NA:
1 1 m
= + (169)
K y k y kx
y − y*
where m≡ i = avg. slope of EC
xi − x
is the average slope of the equilibrium curve in the region of the xy diagram between x and xi.
Of course, if the equilibrium curve is straight (as it will be in dilute solutions), then m is its true
slope over the entire range.
Comment: This is similar to the expression for overall heat transfer coefficients for a double-pipe
heat exchanger: ♣
1 1 1
Di = Do: = + (170)
U hx hy
We said that 1/U represented the total resistance to heat transfer through the two phases, which is
just the sum of the resistances of each phase. The main difference between (169) and (170) is
the appearance of m in (169). For heat transfer, the slope of the equilibrium line is unity (m = 1)
because at thermal equilibrium Ty = m Tx = Tx.
In short, (169) states that the total resistance to mass transfer equals the sum of the
resistances of the gas and liquid phases. We could also have showed, in a similar fashion, that
1 1 1
= + (171)
K x m′k y k x
♣ In our earlier treatment of heat transfer, we also distinguished between inner and outer coefficients (i.e. U and
i
Uo). This was because the inside area and outside area of a pipe were somewhat different. We will drop this
distinction here because the inside area and outside area are equal for an interface. In other words, the ‘pipe wall’
thickness is zero.

06-361 page 188 Spring, 2017
y − yi
where now: m′ ≡
x * − xi
is the average slope of the equilibrium curve in the region

of the xy diagram between xi and x*. If the operating is
curved, this slope might be slightly different from the
average between x and xi; thus we add the prime to the
label. On the other hand, if the equilibrium curve is
straight, these two slopes would be the same ( m′ = m ) and
we could relate the two overall mass transfer coefficient:
1 m
straight EC: =
K y Kx
Diffusion-Induced Convection
We have just seen that the overall driving force for interfacial mass transfer is y-y* or x*-x (it
is not y-x). This is one of two major differences between interfacial mass transfer and interfacial
heat transfer. The second major difference arises because diffusion of mass usually induces
convection of the fluid (i.e. motion of the center of mass).
Recall Fourier’s law of heat conduction:
dT energy
qz =
−k = ]
heat flux [ =
dz area × time
If there is also bulk fluid motion (due to pressure-driven flow, for example), an additional term is
added to the heat flux to account for convection of heat:
energy vol
  
qz = −k
dT
(
+ ρc p T ( z ) − Tref vz

)
dz

conduction convection
where vz is the local z-component of velocity of the fluid and ρcp(T–Tref) is the enthalpy per unit
volume. Usually heat conduction does not induce flow.*
By contrast, diffusion of mass nearly always induces flow. Fick’s law of diffusion must be
modified by adding a contribution from convection of mass:
*An important exception is “natural convection” which arises because the nonuniform temperature of the fluid
causes nonuniform density of fluid which in a gravitational field can cause flow.

06-361 page 189 Spring, 2017
mol vol
dc A  N +N
N Az =
− DAB + cA Az Bz (172)
 dz
 c

diffusion v = mol average
velocity

convection
where c = cA+cB is the total molar concentration. The second term represents convection of
matter which arises because the center of mass moves with time.
Case I: Equimolar Counter-Diffusion of Gases
Suppose we use a capillary tube to connect two

gas reservoirs having the same total pressure P but
different partial pressures of gases A and B (see Fig.
6.2-1 at right). Further suppose that the total
pressure P (and temperature) inside the two reser-
voirs remains fixed. According to the ideal-gas law,
the total molar concentration of gas c is given by
n P
=
c = = const
V RT
which must also remain constant as the gases

counter-diffuse (i.e. diffuse in opposite directions).
Then no net transfer of moles is occurring: the
molar flux of A from left to right must equal the molar flux of B from right to left (i.e. NAz
= -NBz). This is called equimolar counter-diffusion of A and B.
Actually there is second and an even stronger argument for equimolar counter-diffusion
when the two reservoirs are rigid and finite in volume. Any departure from equimolar counter-
diffusion of A and B would cause the pressure in one reservoir to increase with time and the
pressure in the other reservoir to decrease. Any difference in pressure between the two
reservoirs would cause pressure-driven flow of the gas mixture toward the lower pressure (i.e.
convection). Unless the diffusion length is very short, convection generally dominates any
diffusion. This is why diffusion boundary layers tend to be very thin. Convection then would
quickly erode any pressure-difference.
In this way, the steady mole-average velocity v = 0 and the total pressure in both reservoirs
remains constant. For NAz = –NBz, (172) gives
dc A
N Az = − DAB (173)
dz
See Example 6.2-1 in Geankoplis for details leading to a linear profile of partial pressure or of
molar concentration of either gas. For 1-D steady equimolar counter-diffusion (from a reservoir

06-361 page 190 Spring, 2017
having cA1 to a second reservoir having cA2), taking the flux NAz in (173) to be independent of z,
(173) integrates to yield cA varying linearly with z. Substituting this result into (173) yields
DAB cDAB
=
N Az ( c A1 −=
c A2 ) ( y A1 − y A2 )
z2 − z1 z2 − z1 (174)
≡ k y′ ( y A1 − y A2 )
where z2-z1 is the length of the capillary through which diffusion is occurring. The last equation
is the definition of k'y, the single-phase (gas) mass transfer coefficient to be used with mole
fractions and equimolar-counter diffusion.
Case II: Diffusion of A Through Stagnant B
If instead, we have diffusion of benzene vapor above its liquid through air (which is virtually
insoluble in the liquid), the air must remain stagnant (i.e. NBz = 0) since it cannot enter the liquid
(see Fig. 6.2-2a). This is diffusion of A through stagnant B. For NBz = 0, (172) gives
dc cA dc dy
N Az =
− DAB A + N Az or (1 − y A ) N Az =
− DAB A =−cDAB A (175)
dz c dz dz
yA
Solving this differential equation for the partial pressure P

profile yields a nonlinear profile (see Example 6.2-2) as
shown in the figure at right. The flux of benzene is
eventually calculated as pA
cDAB
=N Az (y − y )
( z2 − z1 )(1 − y A )M A1 A2 (176) pB
≡ k y ( y A1 − y A2 )
z1 z2

06-361 page 191 Spring, 2017
(1 − y A1 ) − (1 − y A2 )
where (1 − y A )M ≡
(1 − y A1 )
ln
(1 − y A2 )
represents the log-mean of 1-yA evaluated at either end of the diffusion path. Once again, the
second equation in (176) is the definition of ky, the single-phase (gas) mass transfer coefficient to
be used with mole fractions and diffusion through stagnant film of B.
Proof: At steady-state, ∂c A ∂t must vanish and the 1-D continuity equation
∂c A ∂N Az
+ =
0
∂t ∂z
then requires the flux NAz to be constant with respect to z. Taking NAz and cDAB as constants,
(175) is a first-order, ordinary, differential equation in yA(z), with constant coefficients, whose
general solution is
 N 
y A ( z ) = 1 + A exp  Az z (177)
 cDAB 
where A is an integration constant. Applying the boundary condition yA(z1) = yA1, we could
evaluate the integration constant A:
 N 
A =− (1 − y A1 ) exp  − Az z1 
 cDAB 
Substituting this result into (177), then rearranging yields
 N 
(1 − y A1 ) exp  Az ( z − z1 )
1 − yA ( z ) = (178)
 cDAB 
Further matching yA(z2) = yA2 allows us to calculate the flux
cDAB  1 − y A2 
N Az = ln   (179)
L  1 − y A1 
where L = z2 – z1. Using this result to eliminate NAz in (178):
1 − y A ( z )  1 − y A2 
z L
= 
1 − y A1  1 − y A1 

06-361 page 192 Spring, 2017
Finally, if we multiply and divide (179) by the driving force yA1–yA2 across the capillary tube,
then we obtain (176).
Comparing (174) and (176), we see that
k y′
=ky > k y′ (180)
(1 − y A )M
For the same driving force (yA1 – yA2), the fluxes NAz are different. Since 1-yA is always less
than one, we see that equimolar counter-diffusion (174) is slower than diffusion through a
stagnant fluid (176). This can be qualitatively understood as follows. Suppose that to get to
class, you need to walk down a corridor that's crowded with other students. If everyone else was
standing almost still (stagnant fluid), it would be easier to walk around them than if everyone is
walking toward you (counter diffusion).
Although (180) was derived for one particularly simple case (no pressure-driven or gravity-
driven convection), it applies much more generally. Similar relations exist for diffusion in the
liquid phase:
k x′ (1 − x A1 ) − (1 − x A2 )
kx = where (1 − x A )M ≡
(1 − x A )M (1 − x A1 )
ln
(1 − x A2 )
and between the overall mass-transfer coefficients for equimolar counter-diffusion and diffusion
through a stagnant film:
K y′ K x′
Ky = and Kx =
(1 − y A )*M (1 − x A )*M
(
(1 − y A )*M =
1 − y*A ) − (1 − y AG ) (1 − x*A ) − (1 − x AL )
(1 − x A )*M = *
(1 − y*A ) (1 − xA )
where and
ln ln
(1 − y AG ) (1 − x AL )
and where yAG is the mole fraction in the bulk of the gas, y*A is the mole fraction which would
be in equilibrium with the bulk liquid having a mole fraction of xAL, and x*A is mole fraction in
the bulk of the liquid which would be in equilibrium with the bulk gas having a mole fraction of
yAG.

06-361 page 193 Spring, 2017
HEIGHT OF A PACKED TOWER

The analysis which follows has as its goal the determination
of the height of packing required. The approach is similar to that
used in the design of double-pipe heat exchangers in which the
goal is the determination of the area of heat-exchange surface
required. Because the driving force y-y* varies over the height of
the column, we have to chop up the tower into pieces which are
small enough so that the driving force is virtually uniform
throughout each piece. Since the compositions change only with
z, we chop up the tower in such a way that we produce pieces
which have z ≈ const, which is a thin horizontal slice.
Now let's take a closer look at what happens inside a

particular slice of the tower whose lower surface is z = z and
whose upper surface is z = z + ∆z. The slice contains solid
packing as well as liquid and gas streams. In what follows, we
will ignore the solid (since it is stationary) and treat the liquid
and gas phases as if they were completely separated, rather than
interspersed in each other. Let's do a mass balance on the
acetone in the gas phase only contained within our slice of
column. Besides the liquid streams entering and leaving the slice, we have acetone crossing the
interface between the gas and liquid phases. The rate of absorption per unit area can be
expressed in terms of the local overall mass transfer coefficient:
molar rate of transfer

=
NA = K y ( y − y*) (181)
interfacial area
We could have used any one of the other expressions in (162). Just like the overall heat transfer
coefficient Uo depends on the flowrates of both fluids being contacted in the heat exchanger, the
overall mass transfer coefficient Ky depends on both flowrates
K y = K y ( L, V )
To obtain the rate of transfer across the interface in our slice of column, we need to multiply
(181) by the interfacial area in that slice. In a heat exchanger, the heat transfer surface is fixed
by the geometry of the equipment selected: it is just the area of the pipe wall or the tubes. In
particular, the heat transfer area does not depend on the flowrates of the hot and cold streams.
On the other hand, the boundary between liquid and gas in a packed bed is very complex. Most
importantly, the area also depends on the flowrates of the gas and liquid streams (L,V). The
interfacial area is usually expressed as a, the interfacial area per unit volume of packing:

06-361 page 194 Spring, 2017
interfacial area
≡a=a ( L, V ) (182)
volume of packing
One empirical correlation relating area to flowrates is the Schulman Correlation (see Table 6.3
of Treybal above):
a = mGxn G yp
where Gx and Gy are the mass flowrates of the liquid or gas stream divided by the cross-
sectional area of the tower
mass time
G≡ =
mass velocity
πD 2
4
and where m, n and p are constants which depend on the type and size of packing used.

06-361 page 195 Spring, 2017
Multiplying (181) by (182):
rate of transfer interf area molar rate of transfer

× = = K y a ( y − y*) ≡ r (183)
interf area vol. of packing volume of packing
 
K y a ( y − y*) a
Since Ky and a usually cannot be measured independently, what is usually found in correlations
of mass transfer rates is their product Kya.
If we now multiply (183) by the volume of this slice of the

packed column, we will obtain the rate at which acetone crosses
the interface in this slice:
rate of transfer = r S ∆z
is the rate of transfer in a thin slice of column, where
πDT2
S = empty tower cross-section = (184)
4
S ∆z = volume of slice
At steady state, the rate of transport of acetone into the vapor must equal the rate out:
in = out
(Vy)z = (Vy)z+∆z + rS∆z
(Vy ) z + ∆z − (Vy ) z
or = −rS
∆z
Taking the mathematical limit as ∆z → 0, this becomes
d (Vy )
= −rS (185)
dz
Now the total molar flowrate (V) will change as the gas phase looses acetone, but the
flowrate of acetone-free air (V' ) does not change with z:
V′ V′y
V= → Vy =
1− y 1− y

06-361 page 196 Spring, 2017
d d  V′y  d  y  dy dz V dy
(Vy=)  = V′  = = V′
  =
dz dz  1 − y  dz  1 − y  (1 − y )2 1 − y dz
V dy
(185) becomes: =−rS =− K y a ( y − y *) S (186)
1 − y dz
The second equation is obtained by substituting our expression for the rate of absorption r from
(183). We can now solve this equation for the height dz of the slice needed for a certain change
dy in concentration:
−Vdy
dz =
(1 − y )( y − y *) ( K y a ) S
So if we know the local mole fractions y and y* and the change in the mole fraction dy of the
gas phase which occurred between the top and bottom of this slice, we can calculate the height of
the slice. The total height of packing is the sum of the height of each slice:
∫ ∫
ZT ya
−Vdy
=
ZT =dz
0 yb (1 − y )( y − y *) ( K y a ) S
Factoring out the S, which is constant along the entire height
of the packing, and changing the order of integration (since
yb > ya):
∫
yb
1 Vdy
ZT =
S ya (1 − y )( y − y *) ( K y a )
The equation above represents the most general form of
the design equation. Unfortunately, it hard to give any
meaning to the integral in this general form. So let’s examine some limiting cases. For the case
of a dilute gas stream, we know that
y << 1 1 − y ≈ 1=V ≈ V ′ const K y a ≈ const
After making these approximations, we have
V S 
∫
yb
dy
ZT =   (187)
 Kya 
 y y − y*
avg 
  a
H Oy NOy
which is a little simpler. We have added the subscript “avg” to suggest that, in case V and Kya

06-361 page 197 Spring, 2017
are not quite constant, we might still evaluate this quotient at each end of the packed height (i.e.
at y = ya and at y = yb) and average the two values.
EVALUATION OF NTU'S
To better understand how to
evaluate this integral, we will continue
the example we started a couple of
lectures back (see page 174). The
problem is summarized in the figure at
right. The water flowrate given in the
figure is 1.08 times the minimum,
where the minimum water flowrate L'
was determined on page 181:
L′  L′ 
= 1.08
=  ′ 2.06
V′ 
V min
 
1.91
We need to evaluate integral labelled NOy in (187).

The y in this integral is the mole fraction of acetone in
the absorber at some particular elevation. If the
corresponding mole fraction of acetone in the water is
x, then two are related by the operating line given by
(160):
 x x   y y 
L′  − a  = V ′ − a  (188)
 1 − x 1 − xa   1 − y 1 − ya 
In other words, solving (188) for y yields y(x);

alternatively, solving for x yields x(y). To obtain y = yb = 0.14 with L'/V' = 2.06, xa = 0.0002 and
ya = 0.00807, we must take x = xb = 0.07. Given a value of y in the interval ya ≤ y ≤ yb, we can
use the operating line to solve for x(y).
On the other hand, y*(x) is the mole fraction of acetone in the gas, which would be in
equilibrium with a liquid having mole fraction x. Recall (155) and the text which follows:
′
pace Pace
y * ( x=
) x 0.33 x exp 1.95 (1 − x ) 
2
= γ= (189)
P P 
 

06-361 page 198 Spring, 2017
Plots of y(x) and y*(x) are shown at right. Neither

curve is a straight line on these coordinates, although
the operating line is close.
To evaluate the integral, we need to 1

y x from OL y* from EC
compute the integrand at various points in y − y*
the domain. ya = 8.07E-03 2.00E-04 4.64E-04 131.401
1) First we partition the interval (ya, yb) 0.022 7.18E-03 0.016 168.635
0.036 0.014 0.031 204.747
into 10 equally spaced intervals This
0.05 0.021 0.045 225.474
gives the 11 y-values in the first
0.063 0.028 0.059 219.902
column of the table at right.
0.076 0.035 0.071 192.77
2) For each y, we calculate the
0.089 0.042 0.083 158.147
corresponding x from the OL (188).
0.102 0.049 0.094 126.193
These are shown in the second column
0.115 0.056 0.105 100.38
at right.
0.128 0.063 0.115 80.583
3) For each x, we calculate the
yb = 0.14 0.07 0.125 65.615
corresponding y* from the EC (189).
This is the third column at right.
4) Finally, we calculate the integrand: 1/(y-y*).
A plot of the integrand is shown at right. The area

under this curve represents the integral NOy:
yb
∫y
dy
= 21.0
a y − y*
To evaluate this integral in Mathcad (or other software),

it convenient to pose y and y* as functions of x: y(x) is
the operating line given by (188) and y*(x) is the
equilibrium curve given by (189). Transforming the
integral with respect to y into an integral with respect to
x. This is accomplished by substituting dy = y'(x) dx, where y'(x) is the derivative dy/dx of y(x).
Of course, the limits of integration must also be transformed.

06-361 page 199 Spring, 2017
dy
=
dy = dx y ′ ( x ) dx
yb dx xb y′ ( x )
∫ ∫
dy
= dx
ya y − y * xa y ( x ) − y * ( x )
The maximum value of the integrand occurred at the point in the column where the operating
line comes closest to the equilibrium curve. The denominator y-y* is a measure of this distance.
Because the equilibrium curve turns downward in this case, the "pinch" point occurs in the
middle of the range. In other cases (e.g. NH3 in water), the equilibrium curve turns upward
instead.
Special Case: ya<<1 with concave downward EC.
The figure at left below shows the OL and EC obtained in the solution of MS&H5 Prob. 22.2
involving the absorption of NH3 in water. The problem calls for removing 99% of the ammonia
from the air by contacting the ammonia-air mixture with pure water. This corresponds to (xa, ya)
= (0.0000, 0.0025) and (xb, yb) = (0.090, 0.200). Notice also that the equilibrium curve (red) is
concave up (unlike the previous example). The diluteness of both streams at the top of the tower
plus the concave-up shape of the EC cause a very tight “pinch” to occur at the left end of the OL
and EC
From the corresponding plot of the integrand vs y (plot on right side above), the denominator of
the integrand becomes very small near the lower limit; thus the integral has much of its area
under a “spike” located at the lower limit.
To integrate such a function accurately, you will need a very small grid spacing — which
means a lot of points — or at least very small spacing under the spike — which means a
nonuniform grid. One solution is to transform the integral by multiplying and dividing the
integrand by y:
yb yb yb
∫ ∫ ∫
dy y dy y
= = × d ln y
ya y − y * ya y − y * y ya y − y *

06-361 page 200 Spring, 2017
Thus we plot y/(y-y*) versus lny. Since both the

numerator and the denominator become small at
the lower limit, the quotient tends to be more near
constant with y, allowing the integration to be
done with fewer points and a uniformly spaced
grid.
TRANSFER UNIT
Even if the integrand varies significantly over
the domain of integration, we can still pull it
outside the integral, provided we replace it with
some appropriate constant, which can be thought of as the mean value for this range. Recall the
Mean Value Theorem of Calculus:
yb dy yb yb − ya
∫ ∫
1
= = dy (190)
ya y − y * ( y − y *)avg ya ( y − y *)avg
Of course, it is not usually known what type of average to use, but for the present purpose, the
precise value is not important. The integral can be thought of as the total change in gas mole
fraction divided by the average driving force.
(Gas Phase) Transfer Unit -- a slice (not necessarily thin) of an absorber in which the gas
undergoes a change in y equal to ( y − y *)avg which represents the average driving force
over the entire absorber.
No. of TU's -- the number of these it takes to make the entire absorber (not necessarily an
integer)
Height of TU -- height of a slice of an absorber corresponding to one TU
From (190), we see that the integral clearly represents the number of transfer units:
yb
∫
dy
NOy = (191)
ya y − y *
More specifically, this integral is called the number of overall gas-phase transfer units since it is
based on the overall driving force expressed as the gas-phase mole fraction. Since the product in
(187) represents the total height of packing, the coefficient of this integral must represent the
height of a transfer unit:
V S 
H Oy =  
 
 K y a avg

06-361 page 201 Spring, 2017
So that our design equation can be rewritten as:
ZT = HOyNOy
which is called the HTU method for sizing an absorber. Throughout the derivation above we
made use of one type of mass transfer coefficient:
molar rate of mass transfer

= r = Kya(y-y*)
packed volume
Of course, other types of mass transfer coefficient are commonly encountered and can also be
used to compute the height of packing required:
r = Kxa(x*-x) = kxa(xi-x) = kya(y-yi)
to mention a few. Of course, there are still more definitions based on driving forces expressed as
differences in molar concentration or partial pressure (see Table 7.2-1 in Geankoplis). Any one
of these mass transfer coefficients can be used in the HTU method:
ZT = HOyNOy = HOxNOx = HxNx= HyNy
where
xa dx L S
=NOx
∫ =
xb x * − x
H Ox  
 K x a avg
xa dx L S
=Nx
∫
=
xb xi − x
Hx  
 k x a avg
yb dy V S 
=Ny
∫
=
ya y − yi
Hy  
 kya 
 avg
Note that the Ns are not equal to each other; thus the number of transfer units depends on which
driving force you use.
Ny ≠ NOy
It’s still not apparent what “average” to use to compute ( y − y *)avg . There is one case in
which we can provide a simple answer: For very dilute solutions,* this integral in (190) can be
evaluated analytically (i.e. numerical integration is not required). If both x<<1 and y<<1 hold,
* Up to now, “dilute” has meant y and x are small enough (compared to 1) so that V and L are virtually constant
throughout the tower. This makes the OL straight, but not necessarily the EC. When x and y are very small
compared to unity, the EC will also be straight: it should have the form of Henry’s law: y* = mx.

06-361 page 202 Spring, 2017
then the molar flowrates of the two streams won't change much as the solute is transferred across
the phase boundary. In particular, the ratio of the flowrates will be constant:
then L/V ≈ L'/V' = const
This means that the operating line will be straight, even on mole fraction coordinates:
y = (L/V)x + [ya - (L/V)xa]
Also, for sufficiently dilute solutions, Henry's law always applies so that
y* = mx
will also be a straight line. When both operating and equilibrium curves are straight, then the
integration can be done analytically instead of numerically. The result is:
yb − ya
straight OL & EC: NOy = (192)
( y − y *) L
where ( y − y *) L is the log-mean of yb-yb* and ya-ya* which represent the driving force at the
bottom and top of the tower, respectively.
ANALOGY WITH DOUBLE-PIPE HXER

Comparing (192) with (190), we see that for very dilute solutions, the average driving force
for mass-transfer in an absorber is the log-mean of ∆y at the top and bottom of the absorber.
( y − y *)avg =
( y − y *) L
(
=
yb − yb* ) − ( ya − ya* )
yb − yb*
ln
ya − ya*
This result is quite analogous to our design equation for double-pipe heat exchangers. To see
this analogy, first recall the design equation for double-pipe heat exchangers: see (9) on page 11:
qT = UAT∆TL
where qT is the total heat duty — in other words, the total rate of heat transfer. The analogous
quantity for gas absorption is the total rate of ammonia transfer over the entire packed tower,
which can be calculated from a mole balance as the difference between the rate of ammonia into
the tower (via the gas phase) minus the rate of ammonia leaving the tower:
qT = V(yb-ya) (193)
Recall the design equation (187). After substituting (192) into (187) we have:

06-361 page 203 Spring, 2017
V S yb − ya
= =
ZT H Oy NOy
K y a ( y − y *)
L
Solving for V(yb-ya): V ( yb − ya ) K y a

= SZT ( y − y *) L

total volume
of packing
where SZT is the total volume of packing in the absorber and a is the interfacial area per unit
volume of packing. Thus
aSZT = total interfacial area in entire absorber ≡ AT
and substituting in (193) yields: qT = KyAT ( y − y *) L
which is analogous to: qT = UAT∆TL
Thus, for dilute solutions, the design equation for an absorber is identical in form to that for heat
exchangers: with the overall mass transfer coefficient Ky analogous to the overall heat transfer
coefficient U and the log-mean of y-y* at the two ends of the absorber analogous to the log-mean
∆T.
RELATIONSHIPS AMONG HOX, HOY, HX AND HY
In (169) on page 187, we obtained a relationship between the overall mass-transfer

coefficient and the single-phase mass-transfer coefficients. This relationship can be easily
modified to relate the corresponding heights of transfer units, which are often used in the
literature to report correlations of mass transfer rates.
Dividing both sides of (169) by a (the interfacial area per volume of packing), we obtain:
1 1 m
= + (194)
K y a k y a kx a
where m = (yi-y*)/(xi-x) = avg. slope of E.C.

06-361 page 204 Spring, 2017
Of course, if the equilibrium curve is straight (as it will be in dilute solutions), then m is its slope.
If we now multiply (194) by V/S we have the height of an overall gas-phase transfer-unit on the
left-hand side:
V S V S mV S V S V LS
= + = +m
K ya kya kx a kya L kx a
  
H Oy Hy Hx
Recalling the definitions for various heights of a transfer unit we end up with
V
= Hy + m
H Oy Hx
L
Had instead we started with (171) and multiplied by L/S we would obtained
L
=
H Ox H y + Hx
m ′V
PRESSURE DROP IN PACKED BEDS

Reference: Sects. 3.1C and 10.6C in Geankoplis
In our original motivating example (see page 174), we said the designer of an absorber needs
to size three quantities: the liquid flowrate, the tower height and the tower diameter. We have
already discussed the basis for choosing the first two. In this section, we address the third
quantity: tower diameter. The selection of this quantity involves controlling the pressure drop
across the tower. A typical plot of pressure-drop vs. the mass velocities of the gas and liquid
phases is shown in Figure 7:

06-361 page 205 Spring, 2017
Fig. 7. Pressure drop across a packed bed of 1-inch ceramic Intalox saddles as a function of air and
water flowrates. Taken from Fig. 22-4 of MS&H5.
Usually, flow in a packed-bed absorber is countercurrent.

Gravity causes the liquid to trickle down through the packing,
whereas a small pressure drop drives the gas flow upward.
Pressure drop, like the mass-transfer coefficients, depend most
directly on the mass velocities of the gas and liquid Gx and Gy –
rather than the corresponding molar flowrates L and V. Mass
velocities are defined as
lb/hr of liq M L
Gx = mass velocity of liq = = x
cross-section of tower S
lb/hr of gas M yV
Gy = mass velocity of gas = =
cross-section of tower S
where Mx and My are the average molecular weights of the liquid and gas, respectively, and S is
the cross-sectional area of the tower given by (184). The tower diameter DT exerts its influence
on packed-tower design through S in these equations for the mass velocities.

06-361 page 206 Spring, 2017
Without any liquid present, the pressure-drop

caused by flowing gas through dry packing
increases approximately with the square of the gas
flowrate (see curve labelled “dry” in Fig. 7). Once
liquid is flowing down the column while gas is
flowing up, some of the space between the packing
particles is taken up by liquid, leaving less space
for the gas to flow. Forcing the same gas flow
through a smaller opening will increase the
frictional contribution to the pressure drop. Thus
the curves for increasing liquid flow are above that
for dry packing although the curves tend to be parallel at low gas velocities.
Notice in Figure 7 that the curves with nonzero liquid flow (i.e. “wet” packing) curl up at
high gas flow rates. This can be explained as follows:
The upward flow of the gas exerts a upward shear force on the liquid, retarding its downward
motion. To get the same flowrate with a slower velocity, you need a thicker film. Thus liquid
holdup increases with increased gas flowrate. As the gas flowrate increases so does this shear
force. When the shear force becomes comparable to gravity, the liquid flow might slow to zero
(rightmost figure above) and flooding occurs in the tower.
The amount of liquid residing or accumulating in the tower is called:
(liquid) holdup -- fraction of interstitial volume occupied by liquid
intersticial volume -- "empty" space between and inside packing particles; that volume
occupied by liquid or gas.
loading -- an increase in liquid holdup caused by an increase in gas flowrate
For a given liquid flowrate, there is a maximum flowrate of gas which can be forced through the
column. If you exceed this maximum, then the shear forces on the liquid are so high that they
exceed the weight of the liquid. Then the net force on the liquid is upward and you blow the
liquid back out the top of the column. This is called:
flooding -- liquid down-flow is essentially stopped by high gas up-flow

06-361 page 207 Spring, 2017
Since you can't get the liquid through the column, flooding is clearly a condition to be
avoided. Generally, to get a high area of contact between liquid and gas, you want to operate
near but comfortably below the flooding velocity:
Flooding According to McCabe, Smith & Harriott, 5th Ed.
at flooding: ∆Pflood ≈ 2 inch H2O/ft of packing
loading begins: ∆P ≈ 0.5 inch H2O/ft
normal operation: ∆P ≈ 0.25 to 0.5 inch H2O/ft
Normal operating pressure drops are just below those for which loading begins.
Flooding According to Geankoplis
Flooding occurs at a pressure drop depending on the packing factor:
0.115 F 0.7 for F < 60 ft -1

 p p
∆Pflood =
 (195)
2 for Fp > 60 ft -1
where Fp is the packing factor for the particular packing used in the column (see Geankoplis
Table 10.6-1 reproduced as Fig. 10 below). This formula gives ∆Pflood in inches of water per
foot of packed height. Once Gyat flooding is determined then
loading: Gy/Gy,flood ≈ 0.65 − 0.7
normal operation: Gy/Gy,flood ≈ 0.5 – 0.7
Notice (from Fig. 7) that when we increase the liquid flowrate, flooding occurs at lower gas
flows. If we increase the size of the packing particle, we can tolerate higher gas flows. While
higher gas flows can be obtained with larger packing particles, the amount of interfacial area is
generally less (i.e. a is decreased). These effects are summarized by Fig. 8 (below).

06-361 page 208 Spring, 2017
Fig. 8. Air flowrate at which flooding occurs for ceramic Intalox saddles at different water flowrates.
Taken from Fig. 22.5 in MS&H5.
Fig. 9. General correlation of pressure drop across packed beds for random packings. Taken from
Fig. 10.6-5 of Geankoplis. Units of ordinate assume units given in table below.

06-361 page 209 Spring, 2017
Fig. 7 and Fig. 8 are for a particular packing (1” Intalox saddles) and for a particular
temperature and pressure (20 °C and 1 atm). A more general correlation of pressure drop is
provided by Fig. 9 where
mG
GG = = mass velocity of gas [=] kg-s-1-m-2
S
G
vG = G = superficial velocity of gas [=] m/s
ρG
µL = dynamic viscosity of liquid [=] kg-m-1-s-1♦
µ
ν = L = kinematic viscosity of liquid [=] m2/s*
ρL
The y-coordinate of this graph in Fig. 9 is not dimensionless, so Symbol Units

you need to use the units summarized in the table at right. GL, GG lb/ft2-s
µL centipoise
The size and type of packing is accounted for in this
G
correlation through the parameter vG = G ft/s
ρG
Fp = packing factor Fp ft-1
µ
ν= L centistokes
Values of the packing factor for various types of packing are given ρL
in the following table (Fig. 10).
♦ 1 poise = g-cm-1-s-1 and 1 centipoise = 1 cp = 10-2 g-cm-1-s-1 (approximate viscosity of water at room
temperature).
* 1 stokes = 1 cm2/s and 1 centistokes = 10-2 cm2/s. Yes, the “s” belongs at the end here: it does not denote the
plural but is part of a man’s name (George Gabriel Stokes, who gave us “Stokes law”).

06-361 page 210 Spring, 2017
Fig. 10. Characteristics of common packings. Taken from Table 10.6-1 of Geankoplis.
TOWER DIAMETER
Once the total molar liquid and gas flowrates (L and V) are known, we can choose the
diameter of tower we need. The diameter of the tower is usually chosen such that the pressure
drop is some prescribed value below flooding:
choose DT such that: GG/GG,flood ≈ 0.5 – 0.7

06-361 page 211 Spring, 2017
where the flooding velocity GG,flood is found from Fig. 9 such that the pressure drop is given by
(195).
 mass flowrate of liquid  MLL MLL

where GL ≡  = =
 cross-sectional area of column  S πDT2
4
and where ML is the average molecular weight of the liquid.
M GV
Similarly, GG = (196)
S
So to get a certain pressure drop for a certain set of flowrates, we must choose a particular
value for the tower diameter. Although it looks like you have to guess the diameter DT in order
to calculate GL and GG to get ∆p, a trial-and-error can be avoided by noting that the abscissa
(x-coordinate) of Fig. 9 does not depend on the diameter:
MLL
GL S M L
= = L
GG M GV M GV
S
Note that the unknown cross-sectional area S conveniently cancels out. What remains is known.
The procedure is illustrated by the following example:
Given: L, V
Find: DT
Solution: using Geankoplis’ method:
Step 1) calculate the flow parameter

(abscissa) of Fig. 9
The total molar flowrates of gas and

liquid are calculated by dividing the
flowrate of the nontransferable component by the mole fraction of the nontransferable
component:
L′ lbmol V′ lbmol
=
Lb = 3.09 and =
Vb = 1.619
1 − xb min 1 − yb min
We will also need the average molecular weights of the two streams at the bottom (calculated by
taking a basis of one mole of feed):

06-361 page 212 Spring, 2017
gm
M G yb M ace + (1 − yb ) M
= = 33.1
  air
mol
58.1 29
gm
and M L xb M ace + (1 − xb ) M=
= 20.8
 H
2 O
mol
58.1 18
Finally we need the mass densities of the two streams. Water’s density we can look up (we
ignore the small amount of acetone dissolved in it).
gm
ρL =
1.00
cm3
The density of gases can be calculated using the ideal gas law:
P gm lb
=
ρG M G= 0.001343= 0.084
RT cm3 ft 3
Now we are in a position to calculate the flow parameter:
GL ρG M L Lb ρG
= = 0.044
GG ρ L M GVb ρ L
Step 2) calculate ∆Pflood from (195).
We have 1-inch Raschig rings♠ as packing material (see original problem statement on p174).
According to Table 10.6-1, this packing has a packing factor of
Fp = 179 ft −1
Since Fp > 60 ft−1, (195) yields
inch water
∆Pflood =
2
ft of packing
Step 3) locate the point on the curve of Fig. 9 corresponding to ∆Pflood which also has
the calculated value of the flow parameter (abscissa)
♠ Named for its inventor, the German chemist Friedrich Raschig (1863 – 1928). In 1891, he started his own
chemical company which exists today as Raschig GmbH (www.raschig.de).

06-361 page 213 Spring, 2017
Step 4) read the corresponding capacity parameter (ordinate of this point) from Fig. 9
The capacity parameter (y-coordinate) corresponding to this point (see above) is
ρG
vG Fp ν 0.05 =
1.9 (197)
ρ L − ρG
where ν is the kinematic viscosity of the liquid (in centistokes). The kinematic viscosity is
gm
0.0085
µL cm-sec cm 2
=
ν = = 0.0085 = 0.0085 stokes
= 0.85 centistokes
ρL gm sec
1.00
cm3
Step 5) calculate GG which gives this value for the capacity parameter. This gives
GG,flood
First solve (197) for vG and make substitutions:
ρ L − ρG −0.5 −0.05
=vG 1.9
ρG
( Fp ) ν
0.5
 1.00  ( −0.5
( 0.85 )−0.05 3.9 ft
=
1.9   179 )
 0.001343  sec
The final units were assigned using the table located below Fig. 9. Multiplying this velocity
times the density gives the corresponding mass velocity of the gas phase:

06-361 page 214 Spring, 2017
 lb  ft  lb
GG ,flood =
ρG vG =
 0.084 3   3.9 sec  =
0.327
2
 ft  ft -sec
Step 6) calculate the operating value of GG using GG/GG,flood ≈ 0.5 – 0.7. We will take
the lower end of this range (to be on the safe side).
 lb  lb
= =
GG 0.5GG ,flood 0.5  0.327=  0.164
 ft 2 -sec  ft 2 -sec
Step 7) calculate the desired tower diameter.
πDT2 M GV
= S=
4 GG
First calculate the desired cross-sectional area of the tower from (196):
 lb   lbmol 
 33.1   1.619 
M GVb  lbmol   min 
=S = = 5.5 ft 2
GG  lb   sec 
 0.164 2   60 min 
 ft -sec 
which corresponds to the following diameter:
4S
=
DT = 2.6 ft
π

06-361 page 215 Spring, 2017

Chapter 14. Membrane Separation
Reference: Geankoplis Chpt. 13
In distillation, gas absorption, and extraction (the latter yet

to be covered in 0) we have two immiscible phases which we
bring into intimate contact.
Operation Phases contacted

distillation liquid and its vapor
gas absorption liquid and another gas
S/L extraction solid and a liquid
L/L extraction two immiscible liquids
Another important unit operation for mass-transfer is the membrane separator. Here the two
phases are miscible — usually they are both gases or both are aqueous solutions. Because of
this, they cannot be intimately contacted. Instead, a membrane is placed between them. This
membrane serves two functions:
1. keeps phases from mixing
This is also the primary function of the pipe wall in a heat

exchanger; however, the membrane has a second, even more
important, function. Whereas the pipe wall plays a passive role in
heat transfer, the membrane plays an active role:
2. controls selectivity
Because of differences in molecular size or solubility of the gas in the membrane material,
different species diffuse through the membrane at different rates. This difference is the basis of
separation.
EXAMPLES
With the proper choice of membrane material, a membrane separator can achieve significant
separations in a single stage. Indeed, as we shall see, it usually turns out that multistage
operations with membranes are not economical (compared to alternatives like cryogenic
distillation). Most industrial applications of membranes do not require very high purity products.

06-361 page 216 Spring, 2017
Gases
• separation of 238UF6 from 235UF6. This was a very important part of the atom bomb
project in World War II.♠
• separation of O2 from N2 in air. A number of industrial processes require a chemically

inert gas (like nitrogen). Membranes can produce 95-99% nitrogen from air (79%
nitrogen, 21% oxygen) in a single stage. Sometimes its the oxygen-rich stream that
useful: 40-50% oxygen can be obtained from air, which is useful for medical patients
having respiratory conditions.
• recover H2 from purge streams in ammonia, methanol, and hydrogenation plants.
• separation of He from CH4 (natural gas). Because of its very small molecular size, half
of the helium in a feed stream containing <1% He can be recovered in a single stage,
resulting in a permeate stream contain 30 times greater concentration of He.
Liquids
• dialysis (or liquid permeation). Separation of urea and other small molecules from protein
and cells in whole blood is accomplished in artificial kidneys. Recovery of caustic
(NaOH) from hemicellulose produced in the manufacture of rayon.
• electro-dialysis. Brackish water (containing less salt than sea water) can be made potable
by applying electric field across membrane.
• reverse osmosis. Brackish water can also be made potable by applying pressure to force
water through a membrane which is impermeable to ions.
In the 1950’s, President Eisenhower was convinced that the easy availability of drinking water
was the key to world peace in many parts of the world. Eisenhower’s successor as President was
John Kennedy. In a famous speech, President Kennedy expressed his intention to “land a man
on the moon and make the deserts bloom.” The second part of this quote has been obscured by
the first part. Nonetheless, Kennedy did follow through by creating the Office of Saline Water,
which provided financial support for much of the early research on membranes.
EQUIPMENT
In our schematics of membranes, the membrane will be drawn as if it is a simple planar
sheet. The problem with this geometry is that you are very limited about how much area you can
get in a given volume, sort of like the double-pipe heat exchanger. One way to get a greater area
♠ See Henry De Wolf Smyth, Atomic Energy for Military Purposes, “Chapter X: The Separation Of The Uranium
Isotopes By Gaseous Diffusion” (available at http://www.atomicarchive.com/Docs/SmythReport/smyth_x.shtml ).
For a less rigorous summary, see https://en.wikipedia.org/wiki/Gaseous_diffusion under “History”

06-361 page 217 Spring, 2017
in a given volume is to stack the sheets using spacers which allow for inflow and outflow to each
sheet in the stack. If you roll-up the stack of sheets, you have the spiral-wound membrane (see
below).
This is the first of two geometries used. The second is a variation on the shell-and-tube heat
exchanger, namely a bundle of tubes. Suppose instead of sheets, we can construct the membrane
as a hollow tube. Since we want to operate the membrane under very high pressure drops, the
membrane “tubes” are actually fibers with a diameter of 50-200 microns (1000 microns = 1
mm):

06-361 page 218 Spring, 2017
Above is a photomicrograph of one such hollow fiber. The actual membrane of this fiber is
only 2 microns thick. The thickness of the annulus representing the actual membrane is marked
by the two red arrows. The rest of the fiber (the spongy annulus) is a porous support for the
membrane which allows very large pressure drop (perhaps 1000 psi) to occur without exploding
the membrane.
tube sheets
Hundreds of these fibers are then arranged in a bundle with each of their ends sealed in a disk
which will separate the tube-side flow from the shell-side flow much like a tube-sheet does in a
shell-and-tube heat exchanger (see schematic above). In heat exchangers, the metal tubes would
be welded to the tube sheet, one at a time, usually by hand. These hollow fibers are made of a
polymer (plastic) as is the tube sheet. In order to make these hollow-fiber bundles economically,
a technology was developed to seal all the fibers in the tube sheet at once. This is perhaps the
most proprietary part of making these devices.
right tube sheet (end of each

fiber is open to porous disc)
left tube sheet (end of

each fiber is sealed)

06-361 page 219 Spring, 2017
Above is a schematic of a commercial cartridge containing hollow fiber membranes. This

particular cartridge is a Permasep™ cartridge made by DuPont.♣ A similar product called a
Prism™ cartridge is made by Monsanto (now marketed by Air Products♥).
The figure above shows the radial-flow pattern inside bundle (end view). Shown below are
two possible axial-flow patterns inside bundle (side views).
♣ DuPont patented the first commercial reverse osmosis membranes for treating brackish water in 1969 and
improved Permasep to the point that it was capable of desalinizing seawater in 1974. See
http://www2.dupont.com/Phoenix_Heritage/en_US/1969_a_detail.html.
♥ http://www.airproducts.com/products/Gases/supply-options/prism-membranes/prism-membrane-products.aspx

06-361 page 220 Spring, 2017
(a) shows hollow fibers with right ends sealed and left ends open and exiting to the manifold. V
denotes the flowrate of all gas permeating through the membrane whereas L2 denotes the
remainder of the feed which does not permeate (i.e. the “retentate”). (b) shows hollow fibers
with both ends open and exiting to the manifolds.
SEPARATION OF GASES
Porous Membranes
Reference: Sections 6.2E and 7.6B in Geankoplis.
Perhaps the simplest basis for separation of two gases is molecular size. Gases of different
molecular weights have different diffusion coefficients and will diffuse at different rates through
the membrane.
For equimolar counter-diffusion of gases A and B, the Chapman-Enskog theory predicts the
diffusion coefficient is given by*
 1 1 
T3  + 
 M A MB 
DAB = 0.0018583
pσ2AB Ω D, AB
* This is (6.2-44) in Geankoplis, although different units are assumed.

06-361 page 221 Spring, 2017
where Mi is the molecular weight of species i, p is pressure, σAB is the collision diameter (sum of
radii of molecules A and B), ΩD,AB is a function of kT/εAB, and εAB is the depth of the Lennard-
Jones well for the interaction of molecules of A and B. Note that larger molecules diffuse more
slowly.
For membranes having pores whose diameter is comparable or smaller than the mean-free
path of gas molecules, the diffusion coefficient is inversely proportional to the square-root of
molecular weight:
T
for Knudsen diffusion:♦ DA = 9700 r (198)
MA
where r = pore radius (cm)
T = absolute temperature (K)
MA = molecular weight of diffusing gas A
DA = diffusion coefficient (cm2/s)
Notice once again that larger molecules (i.e. larger MA) have smaller diffusion coefficients.
Suppose we have diffusion in a long narrow

capillary connecting two large reservoirs of fixed
concentration CA1 and CA2. The steady state
concentration profile will be linear and the steady
state flux is given by
dC A C (0) − CA ( L )
J Az =
− DA DA A
=
dz L
where the flux JAz is the rate of diffusion per

cross-sectional area of the capillary.
For membranes having fine pores through which the gas can
diffuse, the fluxes of two gases, A and B, through a porous
membrane are given by
C A (0) − C A ()
J Az = DAe (199)
τ
♦ This is (7.6-2) in Geankoplis, although different units are assumed.

06-361 page 222 Spring, 2017
CB (0) − CB ()
J Bz = DBe
τ
where l is the thickness of the membrane and τ (which is greater than unity) is a dimensionless
factor called the tortuosity which accounts for the fact that the pore length is generally larger
than the thickness of the membrane. Also the flux is the rate of diffusion per unit of total
membrane area; if only a fraction ε of this membrane area is open to diffusion, then the effective
diffusion coefficients are a fraction of the true molecular diffusion coefficient:
DAe = ε DA
where ε is called the porosity of the membrane. Even if there are equal driving forces for the two
components to diffuse across the membrane, they will have unequal concentrations one the
downstream side because of their different diffusion coefficient. This is the basis for separation
in a porous membrane.
The best known example of gas separation by a porous membrane is separation of 238UF6
from 235UF6. Most gases have much smaller molecules than these and polymer membranes are
used instead of porous membranes
Nonporous Polymer Membranes
Many polymer membranes do not have pores.

Instead, the gas must first dissolve in the polymer. Then
the dissolved gas diffuses through the membrane, which
can be thought of as a liquid-like film. As a result, the
concentration profile of the diffusing component A is not
continuous: see the figure at right, which shows a
discontinuity in the concentration of A at either interface.
Generally the mole fraction of the dissolved species

is very small so that the equilibrium solubility of the gas in the solid can be described by Henry's
law:
C A = S A p A and CB = S B pB (200)
where CA = molar concentration of A in membrane

pA = partial pressure of A in gas
SA = solubility constant*
Again, at steady state the concentration profile inside the membrane will be linear and Fick’s law
yields:
* This is the reciprocal of the usual Henry’s law constant.

06-361 page 223 Spring, 2017
dC A C (0) − C A () p (0) − p A ()

J Az =
− DA =
DA A =
DA S A A (201)
dz  
For the purpose of comparison, the flux through a porous membrane [given by (199)] can
also be written in terms of the partial pressures. For an ideal gas, the molar concentration CA can
be related to the partial pressure of the gas PA by
p AV = n A RT
nA p A
C=A =
V RT
DA p A (0) − p A ()
(199) becomes: J Az = porous membrane
RT 
p (0) − p A ()
(201): J Az = DA S A A polymer membrane

Although these two expressions have the same form, the coefficients are different. In particular,
the solubility constant SA varies much more (between one gaseous solute and another) than does
the diffusion coefficient DA. This implies that polymer membranes tend to be much more
selective in separation various gas species than porous membranes. Unfortunately, diffusion
coefficients in solids and liquids are much smaller than for gases so this increase in selectivity is
often at the expense of lower permeation rates.
The product DASA is known as the
permeability: PA ≡ DA S A [=] Barrer
which has units of flux per unit pressure gradient. Since these complex units occur frequently in
such problems, a new unit is defined, which is known as the Barrer:
cm3 (STP) 1 cm cm3 (STP)

1 Barrer ≡ 10−10 × × 10−10
=
s   cm 2  cmHg

s-cm-cmHg
rate
   dz dP
1/area
flux
Often the thickness of the commercial membranes is not accurately known. Then it is easier to
measure the flux per unit pressure drop, rather than per unit pressure gradient. The flux per unit
pressure drop is known as the
DA S A
permeance: QA ≡


06-361 page 224 Spring, 2017
EXAMPLE: PRODUCTION OF ENRICHED AIR

It turns out that most polymers are more
permeable to O2 than to N2. Let's see how
this fact can be exploited to enrich the oxygen
content of air. Consider the apparatus at right,
which has one feed and two product streams.
One of the product streams has been pulled
through a polymer membrane with a vacuum
pump.
The feed is atmospheric air which contains

21% oxygen and 79% nitrogen. We will denote these as components A and B. From our
analysis of steady diffusion through the membrane, we know the fluxes of the two species must
satisfy:
J A= QA ( p A1 − p A2 )= QA ( xp1 − yp2 )
Because we are using a vacuum pump to pull the gas through the membrane, the pressure on the
permeate side of the membrane will be much below atmospheric, thus:
for vacuum (on side “2”): p2 << p1
J A ≈ xQA p1 (202)
J B ≈ (1 − x ) QB p1
This leads to the approximate expressions for the fluxes (above).
Since the entire permeate stream must come through the membrane, the mole fraction of A in
the permeate stream is determined by the relative amounts of the two species which come
through. In time t the number of moles of species A which comes through a membrane of area A
is JAAt. Similarly, the number of moles of species B which comes through is JBAt. The mole
fraction of A is the number of moles of A divided by the total number of moles. The area and
time cancel leaving
J A At JA
=y = (203)
J A At + J B At J A + J B
Substituting (202) into (203):
xQA p1 xQA
=y =
xQA p1 + (1 − x ) QB p1 xQA + (1 − x ) QB
Notice that the total pressure cancels out.

06-361 page 225 Spring, 2017
Finally, we divide the numerator and denominator by the permeability of gas B:
αx αx
=y = (204)
αx + (1 − x ) ( α − 1) x + 1
QA
where α≡
QB
is called the separation factor. It's role is similar to relative volatility in distillation [see (68) on
page 82] so we give it the same symbol.
Example: a fairly typical value for the selectivity for the separation of oxygen and nitrogen in a
polymer membrane is α = 7.
Given: x = 0.21, α = 7 and R = 0
Find: y
Solution: (204) yields y = 0.65
If we did not make the assumption of a vacuum on the downstream side, the analysis
becomes a little more complicated, but it is still tractable. More generally, (202) becomes
=J A QA p1 ( x − Ry )
(205)
J B QB p1 (1 − x ) − R (1 − y ) 
=
p
where R ≡ 2 <1
p1
and (203) yields a value for y which is the root of the following quadratic equation:
R ( α − 1) y 2 − {( α − 1) ( x + R ) + 1} y + αx = 0 (206)
Roots for different values of α and R are plotted below:

06-361 page 226 Spring, 2017
1
R=0
0.1
0.9 0.2
0.3
0.4
0.8
0.5
1
Permeate Mole Fraction, y
0.7
0.6
0.5
0.4
0.3
0.2
1 .10 1 .10
3 4
1 10 100
Permeability Ratio, alpha
Note that if R = 0, (206) reduces to (204). On the other hand, when the pressure ratio
approaches unity (i.e. R→1), nonextraneous root of the quadratic corresponds to the permeate
having the same concentration as the feed:
lim y = x
R →1
Comment: Mathematically, this case is indeterminate. If we substitute R = 1 and α = QA/QB

into (205), we obtain
αQB p1 ( x − y )
JA =
=J B QB p1 ( y − x )
Then (203) yields

06-361 page 227 Spring, 2017
JA α QB p1 ( x − y ) α( x − y)
=y = = (207)
J A + J B α QB p1 ( x − y ) + QB p1 ( y − x ) ( α − 1) ( x − y )
While it is tempting to cancel out the common factor of x-y in the numerator and
denominator, this leads to y = α/(α-1) which is a mole fraction larger than unity (recall α>1).
Actually, by cancelling out the common factor, we obtain the extraneous root of the quadratic [in
y(x,α)] which (207) represents before cancelling the common factor. Instead of cancelling the
common factor x-y, suppose we multiply both sides by the denominator: (α–1)(x–y). This results
in a quadratic for y(x,α). The nonextraneous root of this quadratic is y=x and then (207) is
indeterminant ( 0/0 ).
Above is a plot of y versus α for several different R’s. A couple of trends are apparent. First,
lim y = x
α→1
α=1 means there is no difference in permeability of the membrane to oxygen or nitrogen: then
the permeate has the same concentration as the feed. Second, note that the permeate tends to
pure oxygen if the permeability ratio is high enough:
1 for R ≤ x

lim y =  x
α→∞  R for R > x
A closer look reveals that pure oxygen is only obtained if the pressure ratio is small enough. In
particular, the ratio must be smaller than the feed mole fraction. When R<x, then y→1 as α→∞.
This behavior can be rationalized as follows: if oxygen is to permeate the membrane, the
partial pressure of oxygen in the permeate can never exceed that of the feed:
yp2 ≤ xp (208)
 1
pO2 in permeate pO2 in feed
but p2 = Rp1 (209)
x
(209) into (208): y≤
R
If this inequality is not met, we would have diffusion uphill. The upper bound in oxygen partial
pressure is reached when α → ∞:
 x
lim y = min 1, 
α→∞  R
Of course, the mole fraction must always be less than unity as well as meeting this constraint.

06-361 page 228 Spring, 2017
Stage Cut
Although we did not state it as an explicit assumption, the analysis in the previous section
only applies if the fraction of the feed permeating the membrane is negligibly small. If the
fraction permeating the membrane is significant, a number of complications arise. For example,
the concentration of the gas remaining (in the “residue” or “concentrate” or “reject” or
“retentate” stream)♦ will be different from the feed. This requires the gas on the high pressure
side of the membrane to vary continuously in concentration from the feed end of the membrane
to the residue end. It is no longer clear what average should be used as x in the formulas above.
The fraction of the feed which permeates through the membrane is known as the stage cut:
total molar permeate rate Vout

θ≡ =
total molar feed rate Lin
When the stage cut is very small, then very little of

the feed passes through the membrane
Co-current flow through separator
Lout ≈ Lin
This also means the composition of the feed doesn’t change much as a result of passing though
the separator, so that
stage cut << 1: xout ≈ xin
Then every point on the membrane surface

sees the same feed concentration x. This is
what we have implicitly assumed in all of the
above analysis. Under these conditions, it Countercurrent flow through separator
doesn’t matter whether we operate the
separator co-currently or counter-currently; we get the same y for both flow patterns. But this is
not a very economical way to operate: very little of the feed is converted into product.
If the stage cut is not small, we will partially deplete the feed in the more permeable
component: oxygen and x then varies locally along the surface of the membrane.
for significant stage cut: xout < xin ♣
This complicates the analysis, because now there

will be a continuous decrease in x and L as the feed gas
moves along the surface of the membrane toward the
♦ These are all names applied to the same stream: the outlet stream
for gases which do not permeate the membrane.
♣ This inequality assumes x is the mole fraction of the more
permeable component.

06-361 page 229 Spring, 2017
residue (a.k.a. “reject”). The simplest case (see figure at right) is to assume the gas composition
is uniform on both sides of the membrane despite the reject composition being different from the
feed. The stirrer in the schematic suggests that the gas is well mixed like a well-stirred tank.
The concentrations on either side of the membrane were also assumed to be uniform in the
derivation of (206), which still applies. The only remaining question is what is x in this situation
with significant stage cut? xin or xout? In a well-stirred tank, the steady-state concentration in
the tank is the same as the concentration xout of the output stream, which might differ from that
of the feed xin. This suggests that x should be replaced by xout:
R ( α − 1) yout
2
− {( α − 1) ( xout + R ) + 1} yout + αxout = 0 (210)
Now let’s look at the effect of stage cut on the permeate concentration.
Given: α, R, xin and θ
Find: xout and yout
Solution: xout, xin and yout are also related by a mass balance:
=
Lin xin Lout xout + Vout yout
Dividing through by Lin and relating the flowrate ratios to the stage cut θ:
Lout V
xin= xout + out yout= (1 − θ ) xout + θyout
Lin Lin
 
1−θ θ
x − θyout
Solving for xout or yout: xout = in (211)
1− θ
x − (1 − θ ) xout
yout = in (212)
θ
Substituting (211) into (210) and collecting terms:
( α − 1) θ + R (1 − θ ) yout

2
− {αθ + ( α − 1)  xin + R (1 − θ )  + (1 − θ )} yout + αxin =0

06-361 page 230 Spring, 2017
The root of the new quadratic

0.8
gives us yout as a function of stage
cut for different R. Figure 15 shows  0 
α = 10  0.1 
that the permeate concentration  
(calculated for α = 10) is a
R=
0.2 
0.6
monotonic decreasing function of  0.3 
stage cut (see the red curves). yout  0.4 
 
Mole Fraction
also approaches xin as the pressure
 0.5 
ratio R approaches 1. 0.4
Having a value for yout, we can

then calculate xout from (211). The
figure also shows that the reject 0.2 x
composition also decreases mono-
tonically with stage cut (see the blue
curves).
0
The minimum values of both yout 0 0.2 0.4 0.6 0.8 1
and xout occurs for θ=1. This limit is Stage Cut
of interest when the less permeable permeate
component is the more valuable one. reject
For example, we might wish to R=0
produce a nitrogen rich stream to use R=0
as a “nitrogen blanket” (used to
cover liquids which might be Figure 15. Effect of stage cut on product compositions
degraded by exposure to oxygen). calculated for α = 10.
The maximum nitrogen concen-

tration corresponds to the minimum oxygen concentration which, in the figure, corresponds to
the reject stream for θ=1. Substituting θ=1 into (212):
lim yout = xin

θ→1
Of course, this makes sense because if the entire feed stream permeates the membrane; then the
permeate must have the same composition as the feed (by a mass balance). To obtain the
minimum value of xout, we substitute yout = xin into (210) and solve for xout:
xin 1 + R ( α − 1) (1 − xin ) 

lim xout = (213)
θ→1 α − ( α − 1) xin

06-361 page 231 Spring, 2017
Example: If air is used as feed in a membrane having α=7, what is the richest concentration of
nitrogen that can be obtained?
Solution: we want the smallest value of xout predicted by (213). Since α>1, the smallest xout
corresponds to R=0 (see Figure 15). Substituting R=0, α=7 and xin=0.21, (213) predicts xout =
0.0365 or 1-xout = 0.964 or 96.4% nitrogen.
Figure 13.3-4 shows several different

idealizations. Analysis of the other three
flow configurations in Fig. 13.3-4 is more
complicated because the gas composition
is not uniform. Fig. 13.8-2 shows the
effect of stage cut on the permeate
concentration for each of the 4 flow
configurations illustrated in Fig. 13.3-4.
These are calculated by carving the
membrane into a large number of
differential slices and then integrating the
resulting equations.
Note that, in general, the permeate

concentration decreases with stage cut.
This occurs because the local concentration
on the feed side of the membrane is
everywhere lower than at the inlet of the feed.

06-361 page 232 Spring, 2017
Chapter 15. Liquid-Liquid Extraction

Reference: Geankoplis §12.5 – 12.7.
liquid-liquid extraction — separation of two components of a liquid (the “feed”) by contact with
a second immiscible liquid (the “solvent”)
Extraction is used primarily when direct distillation is not economical (since the solvent usually
has to be recovered by distillation). This might be the case with mixtures having very similar
boiling points, or substances that cannot withstand the temperatures of distillation, even under
vacuum. Examples include:
• extraction of penicillin from fermentation broth by contact with amyl [CH3(CH2)4–]

or butyl [CH3(CH2)3–] acetate [–COOCH3]. Penicillin cannot withstand the
temperatures required for distillation.
• recovery of acetic acid from dilute aqueous solutions by contact with ethyl acetate
[CH3CH2COOCH3] (fingernail polish solvent) or ethyl ether. Extraction greatly
reduces the amount of water which has to be distilled.
• separation of aromatics (rings) from aliphatics (straight chains) by contact with

triethylene glycol. These organics have very similar volatilities.
• separation of high-MW fatty acids from vegetable oil by contact with liquid propane

06-361 page 233 Spring, 2017
EQUIPMENT
Contacting of the two liquids might be
performed in a countercurrent cascade of stages.
Sometimes this can be accomplished in a sieve-
tray tower like that used in distillation (see figure
at right), where the difference in density drives
the heavy fluid down while the lighter one floats
upward.
More often a mechanical agitator is required

to form the emulsion of the two immiscible
liquids at each stage. One common design is the
mixer-settler (shown above). The two liquid
phases are fed to the tank where the mixer
imparts mechanical energy and thus disperses one
of the phases as fine droplets in the second phase
(usually continuous). This state of mixtures is
called an emulsion.
This emulsion is then transferred into a settler

tank where the droplets begin to coalesces after
the solute has redistributed across the liquid-
liquid interface. Once the two phases have
separated, the lighter one is drawn off the top
(shown as “extract” in figure above) of the settler
while the heavier one is drawn off the bottom
(“raffinate”).
To understand why these names are used, consider the most common application: to extract a
solute from an aqueous solution by contacting the solution with some oil phase. Since the role of
the oil phase is to extract the solute, the oil-rich product stream is called the “extract.” What
remains of the aqueous phase might be considered to be “refined” so the water-rich product

06-361 page 234 Spring, 2017
stream is called the “raffinate.” The word raffinate comes from the French word raffiner which
means to refine. Since oils are generally less dense than aqueous solutions, the extract will likely
rise to the top of the settler.
Mixers-settlers can also be arranged as a countercurrent cascade of stages, as shown in the figure
above.
TRIANGLE DIAGRAMS
Many extraction systems involve three components which form two immiscible liquid
phases. For example, acetone and water are miscible in all proportions. When an organic
solvent, methyl isobutyl ketone (MIK) is added to a water-rich acetone solution, two phases are
formed: a water rich phase and an MIK-rich phase, both phases containing all three components.
The composition of the two phases can be graphically represented by means of tie lines on a
“triangle diagram.”
Although gas absorption also involved three components, only two of them appeared in
either phase (e.g. acetone in air or acetone in water).♣ Then we could represent the equilibrium
curve on an xy-diagram as in distillation of binary mixtures. By contrast, L/L extraction has
three components in both phases. This means that two independent mole fractions must be
specified in order to know the composition of either phase. A simple xy diagram is no longer
sufficient to represent the phase diagram.
♣ Most of our analysis of gas absorption was based on the assumption of “one-transferrable component” which was
present in both product phases. The gas and liquid phases each contained a second component which was different
for the two phases, for a total of three components. For example, liquid mixtures of NH3 and H2O were separated
by contact with air. The assumption of one-transferrable component means that the mole fraction of H2O in the air
phase and the mole fraction of air in H2O phase are negligible; NH3 is the transferrable component present in both
phases.

06-361 page 235 Spring, 2017
An equilateral triangle has the geometric property that the

sum of the lengths of the three normals drawn from any interior
point to each side equals the height of the triangle, regardless of
which interior point is chosen. If we normalize the lengths of
these three normals by the height, the normalized length can
represent a mass fraction or a mole fraction of one of the three
components. The geometry of the equilateral triangle assures
that
x A + xB + xC =
1
Thus each point inside the triangle represents the composition of a possible three-component
mixture. A point at one of the three corners represents a pure component, while the normal
distance between an interior point and the side opposite the corner for pure A represents the mole
fraction or mass fraction of component A in the mixture.
The phase diagram for a mixture of acetone,♣ methyl isobutyl ketone (MIK) and water at
25°C is shown above. The region below the curve represents compositions for which two
immiscible liquid phases form, while the region above the curve are compositions for which only
one phase forms. The nearly horizontal lines inside the lower region represent tie lies connecting
the compositions of two liquid phases that co-exist at equilibrium.
♣ acetone is also known as dimethyl ketone.

06-361 page 236 Spring, 2017
In this system, acetone is completely miscible in both water and MIK, but MIK is only
slightly soluble in water and water is only slightly soluble in MIK. This is an example of a Type
I triangle diagram: the one-phase region completely includes two of the three sides of the
triangle.
Notice in the Type I triangle diagram that as the fraction of acetone gets higher, the tie lines
get shorter. Eventually, the tie line becomes infinitesimally short, such that the composition of
the two immiscible phases become equal at the
plait point — point on the triangle diagram at which the composition of the two immiscible
phases are equal.
A very different looking triangle diagram results when one of the liquids is not completely
miscible with either of other two. The diagram below is for the system aniline/heptane/methyl-
cyclohexane.♦ This is an example of a Type II triangle diagram: the one phase region
completely includes only one of the three sides of the triangle.
♦ aniline is a benzene ring with a NH group.

2

06-361 page 237 Spring, 2017
NOMENCLATURE
Under normal operation we contact the “feed” stream
with a “solvent” stream, producing two immiscible liquid Solvent Extract
phases which are separated to form two streams. The two MIK
product streams are called the “extract” and the “raffinate.”
extract — product stream which is richer in the solvent Raffinate Feed

(the remains of the solvent stream after the acetone + water
solute has joined it)
raffinate — product stream which is leaner in the solvent (what remains of the feed after the
solute has been extracted)
RECTANGULAR DIAGRAMS
While triangle diagrams are
commonly used to present the
phase diagram for three-compo-
nent mixtures, the geometry of an
equilateral triangle complicates
construction. Three-component
phase behavior can also be
presented using a rectangular
diagram, an example of which is
shown at right for a ternary
system of Type I.
As with an equilateral tri-

angle, the point at each apex of
this right triangle represents one
of the three components in the
pure state. Each of the three sides
of the triangle represents mixtures
in which one of the three compo-
nents is completely absent. For
example, the base of the right
triangle at right represents mix-
tures of acetic acid and water
which do not contain any ether.
The hypotenuse of the right

triangle corresponds to
xA + xC = 1 or xB = 0

06-361 page 238 Spring, 2017
where the subscripts denote components. Lines parallel to the hypotenuse correspond to
xA + xC = const < 1 or xB > 0
if the line lies below the hypotenuse and
xA + xC = const > 1 or xB < 0
if the line lies above the hypotenuse. Clearly, the latter situation is impossible so that points
outside the right triangle are extraneous.
The main advantage of a rectangular diagram (relative to an equilateral triangle diagram) is

that it’s easier to construct. The disadvantage is that the scale factor for component B is different
than for components A and C.
FLASH MIXING OF TWO LIQUIDS

Suppose we take two immiscible liquids (formed of
the same three components: A, B and C) and mix them.
What will be the composition of the mixture? Let
L, V = total mass flowrate of the two liquid input streams
M = total mass flowrate of the output stream (ignoring the phase distribution)
xA, xC = mass fraction of components A,C in the stream having flowrate L
yA, yC = mass fraction of components A,C in the stream having flowrate V
xAM, xCM = mass fraction of components A,C in the stream having flowrate M
EXAMPLE #1: Given: L, xA, xC, V, yA and yC
Find: M, xAM and xCM
Solution: Here is a recipe for a graphical solution using a

rectangular diagram (see figure at right).
Step 1: Locate points L and V on the rectangular diagram

[coordinates are (xA, xC) and (yA, yC)].
Comment: In the text, we will use non-italicized letters L, M or

V to denote points on the diagram while italicized letters L or V
refer to the total mass flowrates. In the some of the figures
(like that the one at right), the labels for points are also
italicized but it is obvious that these points are not flowrates.

06-361 page 239 Spring, 2017
Step 2: Locate point M on a straight line connecting L and V using the lever-arm rule:
L
VM = (214)
LM V
where LM or VM denote the length of the lines connecting points L and M or V and M,
respectively.
Step 3: Read coordinates of point M as (xAM, xCM).
Proof:♥ A total mass balance is
L+V=M (215)
and mass balances on components A and C yield:
LxA + VyA = MxAM (216)
LxC + VyC = MxCM (217)
Substituting (215) into (216): LxA + VyA = (L+V)xAM
Collecting terms, we obtain: L(xA – xAM) = V(xAM – yA)
which can be rearranged to give:
L x AM − y A
= (218)
V x A − x AM
Similarly combining (215) and (217), we obtain
L xCM − yC
= (219)
V xC − xCM
x AM − y A xCM − yC
Equating (218) and (219): =
x A − x AM xC − xCM
After cross-multiplying and rearranging, this becomes
♥ This proof is similar to that for Lever-Arm Rule for Heat of Dilution on page 53.

06-361 page 240 Spring, 2017
xCM − yC xC − xCM
= ≡ tan θ
x AM − y A x A − x AM
 
slope of line VM slope of line LM
where θ is the angle measured between line segment LV and the horizontal. Since line segments
LM or VM have the same slope and share a common point, the point M must lie on line segment
LV. Moreover, by dividing numerator and denominator of (218) by cosθ, we convert the lengths
of the adjacent sides of the two right triangles into the lengths of the hypotenuses:
( x AM − yA )
=
L
= cos θ VM
V ( x A − x AM ) LM
cos θ
where the overbar quantities denote the length of the line. This
last relationship proves (214). Notice that the two lengths could
be measured on an actual graph with a ruler. The units used for
length cancel out.
Comment: While we usually define the “lever-arm rule” as (214), in applications, it is often
more convenient to use (218) or (219). The three equations give the same value for L/V.
SINGLE-STAGE EXTRACTOR
If point M in the previous example lies in the 1-phase region of the phase diagram, then we
are done because we know the composition (xAM, xCM) of the single phase leaving the mixer. If
the point M lies in the two-phase region and the two phases are allowed to reach equilibrium and
separate in the settler, we will have (not one, but) two product streams. The additional equations
required to solve for the extra unknowns associated with the second product stream can be
obtained by assuming the two product streams are in phase-equilibrium with each other; in other
words, the mixer-settler unit performs as an ideal stage.
Let’s re-label the two input and two output streams as in the
figure at right. In a typical extraction problem, stream L0 is an stage
aqueous solution containing a solute we are trying to extract by #1
contact with an organic solvent. We will use the following
convention when referring to components:
A = solute
B = water
C = organic solvent
So let’s assume that the second feed stream (labelled V2) is pure solvent. Let’s try to determine
the composition of the two product streams.

06-361 page 241 Spring, 2017
EXAMPLE #2: Given: L0, xA0, xC0 = 0, V2, yA2 = 0 and yC2 = 1
Find: L1, xA1, xC1, V1, yA1 and yC1
Solution: Here is a recipe for a graphical solution

using a rectangular diagram (see figure at right).
Step 1: Locate points L0 and V2 on the rectangular

diagram [coordinates are (xA0, xC0=0) and (yA2=0,
yC2=1)]
Step 2: Locate the mixing point M as in Example #1
Step 3: Find the tie line which passes through M
Step 4: The endpoints of this tie line represent L1 and

V1 [coordinates are (xA1, xC1) and (yA1, yC1)]
Step 5: The flowrates L1 and V1 can be determined using the reverse of the procedure for
determining the mixing point M in Example #1. In particular, (214) is used
L1 V1M
= and L1 + V1 = L0 + V2
V1 L1M
which represent 2 equations in the two unknowns.
Comment #1: Essentially the mixing point gives the composition of a hypothetical single-phase
mixture formed from the two feed streams. According to a mass balance, this is the same
mixture which would be formed by mixing the two product streams. In other words, M is the
mixing point for either L0 and V2 or L1 and V1.
Comment #2: The mixing point M must lie somewhere

along the line connecting L0 and V2. The location
depends on the flowrate of the solvent V2 relative to the
flowrate of the solution L0. Increasing the solvent
flowrate V2 shifts the mixing point closer to the point V2
(see Example #1). The mixing point must remain inside
the two-phase region if we are to have any separation of
solute from water. For the phase diagram shown, there is
a minimum solvent flow to keep the mixing point M
inside the two-phase region; and there is also a maximum
solvent flow. This implies that range of solute
concentrations which can be achieved in the raffinate

06-361 page 242 Spring, 2017
stream are limited:
xA1,min ≤ xA1 ≤ xA1,max
MULTI-STAGE EXTRACTOR
If we are satisfied with the amount of solute we have been able to extract from the solution in
the single-stage extractor, then we are done. If we desire to extract more of the solute (i.e. we
wish to achieve a smaller value of xA1) then multiple stages might be used. Let’s specify an
arbitrarily small value for xAN, the concentration of solute remaining after N stages.
EXAMPLE #3: Given: L0, xA0, xC0 = 0, VN+1, yAN+1 = 0, yCN+1 = 1 and xAN
Find: LN, xCN, V1, yA1 and yC1
Solution: The two output streams LN and V1 involve a total of six unknowns (two mole fractions
plus the flowrate for each stream). One of those mole fractions is given (xAN) and we can
perform a total of three independent mole balances (when there are three components in the
mixture). The remaining two constraints arise from the fact that LN and V1 are products of a
equilibrium mixer-settler stage (but not the same stage, so they are not at opposite ends of the
same tie-line). Therefore the two points LN and V1 must lie somewhere on the equilibrium curve
(the curve separating the 1-phase region from the 2-phase region).
Here is a recipe to solve this problem graphically:
Step 1: Locate points L0 and VN+1
Step 2: Locate their mixing point M. Read coordinates (xAM, xCM)

and calculate M = L0 + VN+1.
This will also be the mixing point for the two product streams LN and
V1. Even though these streams are not coming from the same stage,
they contain the same material as in the two feed streams and
therefore, they must have the same mixing point.
Assuming all the stages in the cascade are equilibrium stages, we also know that the points
LN and V1 lie somewhere on the equilibrium curve, but not necessarily on opposite ends of the

06-361 page 243 Spring, 2017
same tie line. Since these streams are not coming from the same stage, they don’t have to be in
equilibrium with each other.
Step 3: Using the known value of xAN, locate the point LN on the lower portion of the
equilibrium curve (i.e. on the raffinate portion of the curve). Read xCN.
Step 4: Draw a straight line connecting LN and M; then extend this line until it crosses the
equilibrium curve again.
Step 5: The intersection of this line with the equilibrium curve is point V1. Read the coordinates:
(yA1, yC1).
Step 6: Calculate the flowrates LN and V1 using the inverse lever-arm rule:
LN M V1
=
V1M LN
This gives us the ratio of the two unknown flowrates. We also know their sum:
total mass balance: LN + V1 = M
Knowing the ratio and the sum, we can solve simultaneously for the two unknowns: LN and V1.
Operating Lines for Extractors
To determine the concentration and flowrates of the remaining streams internal to the
cascade (Li and Vi+1 for i = 1, 2 ... N-1), we need a convenient method for representing mass
balances graphically. The “mixing-point method” we have used above to handle mass balances
cannot be easily extended for this purpose: since we do not know both input flowrates to Stage
#1 or both output flowrates, the mixing point for Stage #1 cannot be determined using the lever-
arm rule. A trial-and-error would be required to determine the mixing point for each separate
stage (very tedious).
An easier graphical method for handling the mass balances is to use “operating lines”.
Operating lines on a triangle diagram connect two points representing adjacent streams. (e.g. Li
and Vi+1). It turns out that all such operating lines (when extended in either direction) share a
common point, called the difference point, which can be calculated in advance.
operating lines: Li Vi+1 for i = 0, 1, 2 ... N
tie lines: Li Vi for i = 1, 2 ... N

06-361 page 244 Spring, 2017
Difference-Point Method
Consider a total mass balance around some arbitrary subset of stages (denoted by the blue
line in the figure above) in the cascade. At steady state, the rate of total mass entering the system
equals the rate leaving, or:
L0 + Vi+1 = Li + V1
which can be rearranged to obtain difference between the flowrates of two “adjacent” streams in
the cascade
Li – Vi+1 = L0 – V1 for i = 1,2 ... N
This can be repeated for any subset of stages in the cascade (i.e. i = 1,2 ... N) and the difference
in the flowrates of adjacent streams has exactly the same value:
L0 – V1 = L1 – V2 = … = LN – VN+1 ≡ ∆ (220)
We denote this difference in flowrates by ∆.
Performing a similar mass balance on component A:
L0xA0 + Vi+1yAi+1 = LixAi + V1yA1
which can be rearranged to obtain
LixAi – Vi+1yAi+1 = L0xA0 – V1yA1 for i = 1,2 ... N
Once again the difference the mass flowrates of component A in any two adjacent streams is
exactly the same, regardless of how many stages are included in the subset. For i = 1,2 ... N:
L0xA0 – V1yA1 = L1xA1 – V2yA2 = … = LNxAN – VN+1yAN+1 ≡ ∆xA∆ (221)

06-361 page 245 Spring, 2017
We denote this difference as ∆xA∆: the product of ∆ and a fictitious mass fraction which we call
xA∆.♦ We say “fictitious” because this mass fraction might be a negative number: it does not
have to lie between zero and unity.
The value of this fictitious mass fraction can be calculated by dividing (221) by (220) for n=1
(chosen because we know the flowrates and compositions of the two adjacent streams at this end
of the cascade from Example #3):
L x −V y L x −V y
x A∆ = 0 A0 1 A1 and xC ∆ = 0 C 0 1 C1 (222)
L0 − V1 L0 − V1
An equation similar to this can be obtained for component C, which leads to the second equation
above. The point (xA∆, xC∆) is called the difference point and is denoted by ∆.♠ All the
operating lines pass through this common point. In the example below, we see that the location
of this point can also be determined graphically without any calculations.
EXAMPLE #4: How many equilibrium stages are requir-

ed to perform the separation described in Example #3?
In other words, what is the value of N?
Solution: recipe
Step 1: Locate given points L0 and VN+1
Step 2: Locate points LN and V1 (see Example #3).
Step 3: Connect LN and VN+1 with a straight line.
Step 4: Connect L0 and V1 with a straight line.
Step 5: Extend these two straight lines until they

intersect. The intersection point is ∆.
Step 6: Determine the tie line passing through V1. Locate L1 as the other end of this tie line.
Step 7: Connect L1 and ∆ with a straight line.
Step 8: Locate V2 as the point where this operating line crosses the (upper part of) the
equilibrium curve
♦ We could just as easily have called this y . The main idea is that (x , x ) [or (y , y )] denotes a unique
A∆ A∆ C∆ A∆ C∆
point on the rectangle diagram.
♠ Geankoplis instead calls this the “operating point.”

06-361 page 246 Spring, 2017
Step 9: Determine the tie line passing through V2. Locate L2 as the other end of this tie line.
Step 10: Repeat steps 7-9 until the desired composition is obtained for LN.
Proof (of difference point method): We need to show that the slope of the line drawn from
L0 to V1 has the same slope as a line drawn from V1 to ∆. Only then can the points L0, V1 and ∆
lie on the same straight line. The equality of these two slopes requires
x A0 − y A1 x A0 − x A∆
=
xC 0 − yC1 xC 0 − xC ∆
Substituting xA∆ and xC∆ from (222):
 L x −V y 
x A0 −  0 A0 1 A1 
x A0 − y A1  L0 − V1 
=
xC 0 − yC1  L x −V y 
xC 0 −  0 C 0 1 C1 
 L0 − V1 
Multiplying numerator and denominator of the right-hand side by L0-V1:
x A0 − y A1 L0 x A0 − V1 x A0 − L0 x A0 + V1 y A1
=
xC 0 − yC1 L0 xC 0 − V1 xC 0 − L0 xC 0 + V1 yC1
− V1 x A0 + V1 y A1
=
− V 1 xC 0 + V1 yC1
− x + y A1
= A0
− xC 0 + yC1
x − y A1
= A0
xC 0 − yC1
After cancellation, we end up with an identity. For two line segments having a common point to
have the same slope, these two segments must be part of the same straight line. Thus L0, V1 and
∆ lie on the same straight line. Recalling that (222) was
obtained by dividing (221) by (220), we could easily
generalize this proof to show that Ln, Vn+1 and ∆ lie on
the same straight line.
Minimum Solvent Flowrate
In stepping off the stages above, we alternated

between an “operating line” and a “tie line” (analogous to
stepping between the operating line and equilibrium

06-361 page 247 Spring, 2017
curve in the McCabe-Thiele method). As long as the slope of the tie line (dashed in the figure at
right) is more negative than the slope of the operating line for that step (OLs are solid lines), we
make progress in reducing the concentration from xAn-1 to xAn.
Should the slope of the operating line ever become equal to that of the tie line, we would step
back and forth endlessly without making progress. In effect, we would have reached a “pinch”.
Then even an infinite number of stages would not allow us to achieve our goal of reducing xA to
xAN.
Consider what happens in this graphical method if we

reduce the solvent flowrate VN+1. According the
Example #3, this will cause the mixing point M for the
cascade to move along line L0VN+1 closer to point L0.
In turn according to Example #4, this causes the point V1
to move downward along the phase boundary towards the
plait point (see red arrow in figure at right). Finally, if
we extend the new line L0V1 (shown as blue dashed line),
it will intersect the extension of old line LNVN+1 (also
blue dashed line) higher up. This new point of
intersection is the new ∆ point.
This new ∆ point will make all the operating lines steeper
than in Example #4. If any of them becomes parallel to a
tie line, we have a pinch. This happens whenever the
solvent flowrate is at or below the minimum.
EXAMPLE #5: find the minimum solvent flowrate for the

conditions given in Example #3.
Recipe:
Step 1: Locate points L0, LN and VN+1.
Step 2: Connect LN and VN+1 with a straight line and

extend this line in both directions. The ∆-point lies
somewhere along this line.
Step 3: Locate (by interpolation) the tie line whose

extension passes through L0 and extend this line until
it intersects the line in Step 2.
Step 4: Locate other tie lines whose extensions pass

between points L0 and LN and extend each until it
intersects the line in Step 2. Depending on the slope
of the tie lines relative to the slope of the operating

06-361 page 248 Spring, 2017
line connecting LN and VN+1, this intersection might occur either above VN+1 or below LN.
Step 5: The lowest point of intersection of these tie lines with the line in Step 2 is the ∆ point
corresponding to the minimum solvent flowrate.

06-361 page 249 Spring, 2017
Step 6: Connect ∆min with L0. Intersection with the extract portion
of the equilibrium curve is V1.
Step 7: Draw a straight line connecting L0 with VN+1 and a second

line connecting LN with V1. Their common intersection is the
mixing point M for the cascade.
Step 8: Determine the minimum solvent flowrate using the lever

arm rule:
L M 
VN + 1 =  0
VN+1M  0
L

EXAMPLE #6: determine the minimum value of VN+1 in Example #3

(aqueous feed is 30 wt%; raffinate product is 10 wt%).
Solution: drawing elements mentioned in the description below refer to the full page graph on
the next page.
Steps 1&2: The tie lines given as part of the equilibrium phase behavior are shown in cyan.
Other tie lines were interpolated using Mathcad. The heavy black line is the operating line
drawn through LN and VN+1.
Step 3: the heavy green line is the tie line which passes through L0.
Step 4: about ten other tie lines, equally spaced between L0 and LN are drawn in red.

06-361 page 250 Spring, 2017
240
230 0
∆ min
220
210
200
190
180
170
160
150
140
130
Ether wt%
120
110
VN + 1 100
100
90
V1
80
70
60
M
50
40
30
20
10
0 LN L0
10 0 10 20 30
Acetic acid wt%

06-361 page 251 Spring, 2017
Step 5: of the ten red tie lines and the green one, it is the green tie line which intersects the black
operating line closest to VN+1. This closest intersection point represents the critical ∆.
Typically (as in this problem) it is the tie line through L0 which determines the minimum solvent
flow rate.
Step 6: The intersection of the green line with equilibrium curve is V1 corresponding to
minimum solvent flowrate.
Step 7: Next we draw straight lines connecting the two feed points L0 and VN+1 and the two
product points LN and V1. Their intersection is the mixing point for either pair of streams.
Step 8: using the lever-arm rule, we can now calculate the minimum solvent flowrate by
measuring the lengths of the two line segments:
VN + 1 L0 M lb of ether
= = 1.283
L0 VN + 1 M lb of aqueous solution
View publication stats

Unit Operations of Chemical Engineering: January 2018

Uploaded by

Copyright:

Available Formats

Unit Operations of Chemical Engineering: January 2018

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Unit Operations of Chemical Engineering: January 2018

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Unit Operations of Chemical Engineering

Book · January 2018

Total Internal Reflection Microscopy View project

Teaching View project

The user has requested enhancement of the downloaded file.

Copyright © 2017 by Dennis C. Prieve

CHAPTER 4. SHELL-AND-TUBE HEAT EXCHANGERS ....................................... 24

CHAPTER 5. EVAPORATION ................................................................................. 39

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

Example: Simple Multiple-Effect Evaporator Problem...........................................................59

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

Minimum Reflux Ratio ..............................................................................................................155

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

Chapter 1. Course Introduction

C. J. Geankoplis, Transport Processes and

I will be making extensive use of BlackBoard

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

WHAT ARE “UNIT OPERATIONS”?

process = sequence of “unit operations”

• shell-and-tube heat exchangers (change the temperature of a stream)

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

Overview of ChE Curriculum

rates of approach to thermal equilibrium (heat

unit operations design equipment for physical transformations

1) rate of approach to chemical equilibrium

divide process into sequence of steps

Chapter 2. Heat Exchange Equipment

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

Below is a third design, called a 2-4 hxer.

Chapter 3. Fundamentals of Heat Transfer: Review

LOG-MEAN DRIVING FORCE

where the proportionality constant U is some heat-transfer coefficient.

In any real heat exchanger, the local driving

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

60 Nonzero values must be supplied for the following heat capacities:

dq = q(x+dx) – q(x) [=] J/s = W

x=0 x=x x=L

dA is the heat exchanger area in this slice between x

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

cylinder of diameter D, then the heat exchange area of this slice is

dA = πD dx and A(x) = πDx

Consider steady-state countercurrent flow through a pipe wall

• unprimed refers to the cold stream,

A similar balance on the cold fluid (shown in blue color) gives

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

rate in = rate out

Solving (4) and (5) for q:

• dH is the change in enthalpy (energy/mass)

Assumption #1: no phase changes occur in either fluid (dy = 0)

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

Assumption #2: constant specific heat for both streams

Copyright© 2017 by Dennis C. Prieve PDF generated May 5, 2017

is the same at every point:

Assumption #3: U is constant

Now we are in a position to evaluate the integral in (3) which becomes

Next, we multiply and divide the integrand by d(∆T):