ORNL-2477

Chemistry —Separotion Processes for

Plutonium and Uranium

Contract No. W-7405-eng-26

EDUCATION DIVISION

A SUMMARY OF THE SEPARATION OPERATIONS

M. E. Whatley

DATE ISSUED

MAR 2 61958

OAK RIDGE NATIONAL LABORATORY

Oak Ridge, Tennessee
operated by
UNION CARBIDE CORPORATION
(or the
U.S. ATOMIC ENERGY COMMISSION
WlKSIJfHFP1 en™* systems libraries

3 4145b, DDSDSMb, D
 FOREWORD

In many of the economic studies of future power reactors, one will find the estimated
cost of processing spent fuel given as about ten per cent of the total power cost. This
is, perhaps, a very optimistic guess. Cost figures from actual operating data have
never approached those routinely assumed. In spite of the inevitable significance of
chemical reprocessing, reactor engineers are tempted to turn away from the problem in
hope that, if thoroughly ignored, it might go away.
Any rational evaluation of this problem and of its possible solutions requires an
understanding of some chemistry, some chemical engineering, and the present technology
in the reprocessing of irradiated materials. This short text deals with chemical engi
neering fundamentals which are thought essential to the understanding of the technology.
It does not treat the necessary chemistry nor the technology itself.
This text is primarily intended for use in the Chemical Technology Course of the
Oak Ridge School of Reactor Technology, and the level of presentation was determined
by the averaged academic backgrounds of the ORSORT students and by the time which
could be alloted to these subjects in the ORSORT curriculum. It is primarily intended
for those who are not chemical engineers.
The main object of the text is to give to the student some understanding of and a
feel for multistage countercurrent operations. The operations discussed in detail are
distillation, solvent extraction, gaseous diffusion, and ion exchange. An effort is made
to unify these by use of the theoretical stage concept throughout. In spite of the fact
that some operations can be more accurately described by a transfer unit concept, the
length of time required to develop it was not thought justified. Similarity, simple X-Y
operating diagrams were used throughout. The use of a triangular diagram for solvent
extraction might have been more elegant (though no more accurate in radiochemical
applications), and distillation could have been treated by the enthalpy balance method.
Were this a text for chemical engineers, these omissions would be a failing, but use
of a standard set of concepts provides a common basis for discussion of the operations
and presents a clear picture which is not, in a qualitative sense, inaccurate. In this
context I think this approach is well justified.
In the early chapters the concepts of equilibrium, steady state, and system tran
sients are treated to the degree required in the later chapters. The material balances
in cascade operations are in many ways similar to those in reactor systems. This
prompted the use of continuous reactor processing problems as examples, which evolved
into a large minor topic of the text.
It is hoped that this text is sufficiently clear and concise that those who have not
the time or drive to make exhaustive studies of the chemical engineering operations but
do have the incentive to inquire "What is it good for?" or "How can you do such and
such?" will be able to study it and gain a degree of understanding.
 CONTENTS

Foreword ii

Chapter 1. Chemical Engineering Systems 1

Chapter 2. Common Concepts Used in Separation Operations 19

Chapter 3. Distillation 26

Chapter 4. Gaseous Diffusion 49

Chapter 5. Solvent Extraction 69

Chapter 6. Ion Exchange 87

 Chapter 1

CHEMICAL ENGINEERING SYSTEMS

Think of an operation, or process, or device, or group of devices which receives

material, perhaps from several streams, and effects some change, or combination, or
separation of this material; then this is a system which can be analyzed by standard
chemical engineering methods. The rates of introduction of material, the rate of change,
the rate of the production of new materials and by-products, and the rate of removal of
these lend themselves to analysis by techniques that are common to petroleum refining,
the production of fertilizer, the purification of uranium, the processing of irradiated
reactor fuels, and the operation of nuclear reactors.

The three concepts which form the basis for all the analyses are the material
balance, the equilibrium conditions, and the kinetics. The material balances simply
state that what is put in must either be consumed, accumulate, or come out. The equi
librium conditions have thermodynamic bases and tell us either the performance of an
ideal operation or the direction in which change takes place at any given point in an
operation. The kinetics of the system have to do with the rate at which equilibrium is
approached; they tell us how much time should be allowed for a given operation or how
big a piece of equipment will be needed in order to do a given job.

Kinetics. —Since only nominal attention will be devoted to the subject of kinetics,
it might be well to discuss it first. A steel bridge crossing a river might be considered
as a process to produce ferric oxide from metallic iron and air. All the ingredients are
present, and the equilibrium is certainly favorable for the reaction. Fortunately for the
traffic using the bridge, the kinetics of the reaction are not favorable, and, by judicious
use of various inhibitors, the reaction may be made to take a large number of years to
come to equilibrium. This is entirely a problem in kinetics.

On the other hand, consider the neutralization of a tank of dilute hydrochloric acid
by the addition of caustic. The rapidity of this reaction is limited only by the degree
of agitation of the tank.

Between these extremes there are the reaction rates of many processes. Esteri-
fication, polymerization, and the cooking of meat are examples. These occur primarily
in one-phase systems. Of more interest to us are two-phase systems such as vapor-
liquid, liquid-liquid, and solid-liquid systems. If a bucket of water is set in a closed
room which is kept at constant temperature, eventually the humidity of the air in the
room will reach saturation. The rate at which this occurs (that is, the kinetics) is
limited by the diffusion process. Saturation could be achieved at a much faster rate
if the water were dispersed in the room as a fine spray. Similar to this are cases of
solid phases precipitating from ionic solutions, boiling liquid mixtures, and many others.
 In very slow processes the kinetics are all-important; in intermediate-rate processes
the kinetics determine the size of reaction chambers, mixers, etc. (although this is
frequently done experimentally without going through formal calculation); and finally
there are processes which are so fast that immediate equilibrium can be assumed.
From here on, all operations and processes will be assumed to be ideal, which
means that they will be assumed to bring the system to equilibrium. This implies either
that the kinetics are so rapid that they are not important or that the determination of
the adequate size of equipment has already been made and is available. In counter-
current multistage flow processes this assumption will enter in the form of the size of
an equivalent theoretical stage, a concept which, as will be seen, provides an excellent
representation of the kinetics and is quite accurate in its application. In concurrent
flow processes the assumption of an ideal operation always involves an error (sometimes
quite small), and the fact that the rather complex aspects of kinetics have been omitted
should always be remembered.
Transients in Engineering Systems. - In any system, regardless of the number of
components, if the inlet streams are held uniform in composition and rate, after a long
time of operation all the streams in the system will come to a constant rate and compo
sition and all the inventories in all parts of the system will remain constant (at least
in an average sense). This is the condition of "steady state." This does not imply
that equilibrium has been reached in any part of the system. In fact, this has nothing
to do with equilibrium. As will be seen, it is possible to have equilibrium without
steady state and steady state without equilibrium. The analysis of systems at steady
state is very much simplified, not only by the elimination of time as a variable, but also
by the elimination of all inventories as variables.
When an operation is started after some shutdown or perturbation in its behavior, it
goes through a transient period before steady state is again established. The relative
importance of the transient period (aside from the possible taxing of the capacity of
some of the pieces of equipment) is dependent upon the fraction of the total operating
time that is spent in the transient. Some operations are cycled and never allowed to
reach steady state, while others go through a brief start-up period and operate con
tinuously at steady state for months, or in some cases for years. Analysis of the
transients in a few simple systems will be dealt with before the steady-state analysis
of. more complex systems is considered.
The simplest transient problem that might have significance is a step change in
concentration in the feed to a well-mixed tank of constant volume. Let the holdup
capacity in the tank be A, and let the flow rate into the tank (and out of the tank) be
F capacity units per unit time. The concentration of a solute in the feed stream is
x» for all times before t = 0 (and therefore the concentration of solute in the tank is
xD at t = 0), and the concentration in the feed stream is x. for all times after t = 0.
The dimensions of x are quantity units of solute per capacity unit, so that Fx is a
quantity of solute per unit time and the instantaneous inventory of solute in the tank
is Ax.

UNCLASSIFIED
ORNL-LR-DWG 22788

A material balance around the tank says that the change in inventory of solute must
be equal to the rate at which solute is introduced minus the rate at which it is removed;
or

d(Ax) dx
(1.1) = A = Fx, - Fx = F(x, - x)
dt dt ' '
Equation 1.1 is amenable to solution by separation of variables, and, when integrated
from t = 0 to t = t and from x = xn to x = x, it gives:

(1.2) x =(xQ - x,) exp {- —j +x,

The transient equation 1.2 expresses the tank concentration as a function of time. It
is very similar to equations for radioactive decay.
It should be noticed that there was no use of an equilibrium condition here; however,
the assumption of perfect mixing was analogous to the assumption of an ideal stage.
A model which does use an equilibrium condition is a single-stage solvent extraction
operation. The system consists of three components, two solvents which are immiscible
(like kerosene and water) and a solute which is soluble in both of them. This system
separates into two phases, which will be called the heavy phase and the light phase.
With corresponding quantities designated by the symbols used for the first illustration,
the nomenclature is:
Heavy Phase Light Phase

Flow rate into and out of the stage F Q

Holdup capacity in the system A S

Concentration of solute in system x y

The capacities A and S are constant, and the rates F and Q uniform in time. The
concentrations in the inlet streams are x, and y,, but the concentrations in the system
at time zero are xQ and y.:
 UNCLASSIFIED
ORNL-LR-DWG22789

Qy,
frXP~\:\ <?/
_ \ ° ^ ° °~
oV/o 5
°S/~^ ° °
n 0 O C-*"o 0

Fx, »» o o °
f*

Again the system will be assumed to be well mixed, which now means that an equilibrium
distribution of the solute between the two phases must be satisfied. This relation will
be generally expressed as y = 4>(x), indicating that the concentration in one phase is
a single-valued function of the concentration in the other. (Therefore x„ and y0 must
satisfy this relation.)
The material balances for the solvents are trivial. The material balance for the
solute says that the rate of change in inventory in both phases must be equal to the rate
at which solute is being introduced in both streams minus the rate at which it is being
removed in both exit streams:

d{Ax) d{Sy) A dx S dy
(1.3) + = Fx, + Qy, - Fx - Qy
dt dt ~dt~ dt 7 7

Now we may substitute

dy
(1.4) y = <f>(x) dy —
— dx = a.'i
0 (x)\ —
dx ,
dt dx dt dt

producing the equation

dx
(1.5) [S</>'{x) + A] F[x - x] + Q[y, - <p(x)] .
dt f '

Here again the variables are separable. The integral expression can be written:

* [S<f>'(x) + A]dx
(1.6)
/'-/• Fbcf - x] + Q[yf - <£(x)]
Whether the integration can be analytically performed depends, of course, upon the
nature of the equilibrium relationship. Frequently this is available only as experimental
data.

After a sufficient passage of time the system approaches steady state, and the
derivatives of x and y go to zero. From the original differential equation, it is possible
then to write

(1-7) F(x/ - x) = Q(y - yf) ,

which is conveniently independent of all capacities and initial concentrations.
This development holds in a simplified form for distillation. Because of the
 relatively low density of the vapor in almost all distillation systems, the inventory in
the vapor phase can be neglected. In this case the integrated form becomes:

(1.8)
r* _ rx Ajx
A *~JXo Fbcf - x] +Q[yf - #*)]
With the simplification that the inventory of the vapor phase is neglected, somewhat
more complex distillation systems can be considered. The feed, instead of comprising
a vapor and a liquid stream, is usually introduced as a liquid, and heat is added to
convert some fraction of it to a vapor. Consider an operation where two such stages
are used in sequence, the liquid from the first being used as feed to the second:
UNCLASSIFIED
ORNL-LR-DWG22790

pM Pziy2

F _. *-. L

"f *i
*

Let Ay and A2 be the liquid holdups in stages 1 and 2, P. and P, the vapor removal
rates, and L^ and L2 the liquid removal rates. The over-all material balances are
F = Py + L, ,
^ = P2 + L2 ,
and the equilibrium relationship is
y = <f>M .
The differential equation describing the first stage is very much like that for the illus
tration above:

dx,
(1.9)
A} ~dt " FXf " P,y' " L,Xl '
which integrates to the form

J»t /»*i A} dx.

(1.10) dt = /

This may be solved for x, = x^U), which gives us a variable feed to the second stage.
The differential equation describing this stage is then:
dx~
(1.11) A2 —— = L,x,(/) - P2<f>(x2) - L2x2
At this point life ceases to be simple. In its general form, this equation cannot be
solved by separation of variables; indeed, the possibility of an analytical solution
depends upon the nature of the equilibrium relation.
 Leaving this example without further discussion, we move on now to a more realistic
distillation operation. Consider the two stages connected in cascade so that the vapor
from stage 1 goes into stage 2 and the liquid from stage 2 into stage 1. Feed is in
troduced into stage 2, heat is added to stage 1, and liquid is removed from stage 1:

UNCLASSIFIED
ORNL-LR-DWG 22791

Fxf p2yz

WxL

HEAT '

Given that P, = P2 = P, that L2 = F, and that L^ =W. The material-balance differential

equations for stage 2 and stage 1 are:
dx2
(1.12) A2 — = Fxf + Py} - Fx2 - Py2 = F[xf - x2] + P[<£(x,) - <£(x2)] ,
dt

dx.
(1.13) l, — = Fx2 - Py^ - Wx, = F[x2 - x,] - P[^(x,) - x,]
These equations are coupled in such a manner that they cannot be solved one at a
time but must be solved simultaneously. Again, if the equilibrium expression is not
simple, or if it can be expressed only graphically, an analytical solution may not be
possible.
Suppose that the equilibrium relation be simply y = Kx (a fair approximation for some
particular systems) and that at time zero x1 = x2 = x,. The solution is now possible
by use of Laplace transforms.
On substitution of

Z=£(x) , pZ - X^ =£f—j
d') ., —=£(Xf)
"P ,
the transformed equations become
(1.14) A,(p2, - Xf) = F(z2 - z,) - P(K - l)z, ,
(1.15) A2(p2z2 - pXf) = F(x/ - pz2) + PKp(zy - z2)
Equations 1.14 and 1.15 can be solved simultaneously for z. and z~, by the use of
determinants, to give

(1.16) :, =—\A}A2p2 + \A}(F + PK) + FA2]p + F2 I ,

xf r
(1.17) z2 =— \A^A2p2 + [A2(F + P\K - 1{) + PKA, + FA,]p

+F(F +P{K - 1})j ,

/here

(1.18) A= pU}A2p2 + [(F + PK)U, + A2) - PA2]p +

+ [(F - P)(F + PK) + P2K2]\ .

The numerators of both z, and z2 are quadratic in p, and A itself is factored top
times a quadratic. By solution of these quadratics, Eqs. 1.16 and 1.17 may be written
in the form

(p + a,)(p + b})
(1.19)
'' p{p + a3)(p + b3)

(p + a2)(p + b2)
(1.20)
p(p + a3)(p + b3)
where the a's and b's are the solutions of the quadratic expressions. Finally, Eqs. 1.19
and 1.20 can be split up by the method of partial fractions to
B, C, D,
(1.21)
p p + «, p + b.

B2 C D2
(1.22)
p p + a3 p + b3
where the B's, C's, and D's are determinable constants. In this form they can be found
in tables of inverse transforms. The solution, then, is

(1.23) x, = B, +C^e'"3' + D,e~*3' ,

(1.24) x2 = B2 + C2e~a3' + D2e~ 3' .
The complexity of the problem increases with the number of stages involved, the
extent of coupling between them, and, of course, the degree of complexity of the equi
librium relation. The previous distillation example would not have been as involved
as the problem just solved, yet both of these are very simple examples of chemical
engineering operations.

The analysis of the steady-state operation is, however, another matter entirely.
Steady State. - The analysis of systems at steady state consists simply in taking
the system with imposed concentrations, conditions, flow rates, and efficiencies;
identifying the unknowns (which are usually concentrations); and writing sufficient
balance and equilibrium equations to allow solution. If insufficient equations can be
written, a solution is not possible; if more than sufficient independent equations can
be written, there may be two incompatible solutions. The latter case usually arises
because more data are available than the minimum required, and the most nearly correct
solution to the problem is the one that uses the most reliable information. It might be
inferred that the proper procedure is to write down the n unknowns and the n equations
and to solve them simultaneously, but the problem is rarely this difficult. We will later
see that a system with 30 or 40 unknowns may be solved simply. In almost all cases
the connectivity of the system is such that the problem unwinds easily from one end,
and if the start is made at the proper place, the solution is relatively easy. Because
of the wide variety of problems that might be encountered there are few rules of approach
which hold generally, and there are exceptions to these. One such rule would be to
include in the first material balances as much of the system as possible (preferably the
whole system); then, after as many as possible of the terminal streams have been
determined, to start into the system from the end about which the most is known (this
will often be the product end).
As a simple example, take the two-stage distillation cascade just considered. If
all the flow rates are given, material balances can be written around the entire system
and around each stage (one of these will not be independent), and there are two equi
librium relations. There are, then, four independent equations, and the possible concen
tration variables number five. This means that if one concentration is specified the
problem is workable. If the waste concentration is specified, the problem can be solved
one equation at a time, with the successive use of an equilibrium relation, then a
material balance, again an equilibrium relation, and a final material balance. If either
the feed or the product concentration is given, the problem is not simple but may be
solved by iteration (this is the most convenient method when the number of stages is
large). This method will be discussed in more detail when the individual operations
are treated.

One of the deviations from perfect performance of a unit might be the incomplete
separation of phases, as a result of which the efficiency of the actual stages is less
than it should be. In multistage cascades this is accounted for by the use of more
actual stages per theoretical (or ideal) stage. Where the number of stages is small this
cannot be easily done, and the efficiency of the phase separation must be taken into
account. It may be possible to consider the performance of a unit as constant, or it
may be necessary to allow for changes in performance with operating conditions.
We will now consider some examples of steady-state analysis of systems with
simple equilibrium conditions and performance characteristics, where the emphasis is
on the material balance.

Steady-State Material Balance: Example 1. — The core system of a homogeneous

reactor has a volume of 4000 liters. The power level is such as to produce 30
mg/day/liter of fission products that will precipitate. The over-all effective solubility
of these fission products is 10 mg/liter. The precipitated solids are removed by two
hydroclones in series, the underflow from the first being used as feed to the second.
The feed rate to the first (F2) is 1000 liters/day; 100 liters/day is taken as underflow
and fed to the second, and 10 liters/day is removed from the second and bled out of the
system (see Fig. 1.1). The hydroclones have an efficiency of 80%, which means that
80% of the solids entering with the feed report to the underflow stream.
UNCLASSIFIED
ORNL-LR-DWG 22767

T7
T6

$ FZ
V
5,
\

BA

Fig. 1.1. Flow Sheet for Example 1.

Determine the steady-state concentrations of solids in all streams.

Solution. - Flow rates are expressed in liters per day, concentrations are in milli
grams per liter, x refers to the concentration of solids, and s refers to the concentration
of dissolved fission products capable of becoming solids.
The over-all material balance is

B4(x4 + s) = 4000 x 30 = 12 x 104 mg/day ,

where

B4 = 10 liters/day ,
so that
xA = 12,000 - 10 = 11,990 mg/liter .
At steady state there will be s = 10 mg/liter of dissolved fission products capable
of precipitating in every stream with the possible exception of T?.
The performance relationship for the second hydroclone is
B4x4 = O.8OB3X3 ,
from which
10x4
= 1498.7 mg/liter
*3 100 x 0.80
The material balance around the second hydroclone is

B3*3 " B4X4 + T5*5 '

so that

1 B4*4
=7" <B3*3 " B4*4> (1.25 - 1)

= — x 11,990 x 0.25 = 333.05 mg/liter

The performance relationship for the first hydroclone is

B3X3 = 0.80F2x2 ,
whence

B4*4
F2 x 0.802
10 x 11,990
187.3 mg/liter
1000 x 0.64

The material balance around the first hydroclone is

F2x2 = T6x6 + B3X3 ,

that

B4x4
x. = [F~x~
2*2
- B,x,]
'3A3J =
0.80 0.80

10 x 11,990 x 0.25
= 41.63 mg/liter
900 x 0.80

The material balances at the stream junctions are: at the junction of streams 1, 2,
and 5,
F2x2 = F,x, + T5x5 ,
F, = F2 - T5 = 1000 - 90 = 910 liters/day ,

10
from which

F2x2 - T5xs 187>3 x 1000 - 90 x 333.05

= 172.9 mg/liter ;
F, 910
at the junction of streams 1 and 7,
T7 = F,
at the junction of streams 4 and M,
m = e4
(note that the make-up contains no dissolved fission products); at the junction of streams
6 and 7,
T7(x7 + s) = T6(x6 + s) ,
from which

T6 900
x7 = (xA + s) - s = x 51.63 - 10 = 51.11 - 10 = 41.11 mg/liter .
7 T7 6 910
Summary of Answers. —

Point Flow (liters/day) Solids Concentration (mg/liter)

1 910 172.9

2 1000 187.3

3 100 1,498.7

4 10 11,990

5 90 333.05

6 900 41.63

7 910 41.11

Steady-State Material Balance: Example 2. —A homogeneous reactor operating at

300°C and 2000 psig, generating 5 megawatts of heat, makes use of a pressure-let-down
system to remove the fission-product xenon (see Fig. 1.2). The main loop of the reactor
circulates at 400 gpm from the core through a gas separator, through a heat exchanger,
through a pump, and back to the core. As a result of fission-fragment action, water is
decomposed into hydrogen and oxygen at a rate which can be controlled by catalytic
additives. There is also some oxygen added to the system to maintain the proper
oxidizing conditions to minimize corrosion. These gases form a vapor phase which
purges the xenon from the system through the gas separator. The total extent of the
vapor phase is small, and its separation is difficult. Hence, not all of the vapor passing
through the separator is caught, and some of the liquid phase leaves the separator with
the vapor. The efficiency of the separator for removing the vapor phase can be related
to the fraction of the passing liquid phase accompanying it by the relation:
Efficiency (%) = 100(1 - e-1°°0Z) (
where Z is the fraction of the liquid in the main stream let down.

11
 UNCLASSIFIED
ORNL-LR-DWG 22768

OFF GAS

POWER-. 5Mw
j®
TEMPERATURE OF CORE SYSTEM: 300°C CONDENSER

PRESSURE OF CORE SYSTEM: 2000 psig

RECOMBINER

GAS

GL COOLER SEPARATOR 2

LIQUID

DUMP
TANK
HEAT
© SEPARATOR* EXCHANGER

© PUMP 2

400 gpm OXYGEN

PUMP I

Fig. 1.2. Simplified HRT Flow Sheet for Example 2.

The let-down system consists of a cooler, a let-down valve, a second disengager

which completely separates the liquid and gas phases, a recombiner which can be
considered to completely burn all the hydrogen, a condenser and associated cold trap
which can be considered to remove all water vapor, a dump tank which continuously
boils to remove any dissolved gases, and a pump which returns the let-down liquid to
the high-pressure system. The let-down system operates at or near atmospheric
pressure.

For analytical purposes the system can be considered to consist only of water,
oxygen, hydrogen, and xenon.
The following data are given:
Henry's Law Constant at 300°C
Constituent Production (or Addition) Rate
(psi per mole fraction)

Xe 1.2 X 105 0.3 atoms per fission (effective)

H„ 2.0 x 105

2.0 X 10S 0.01 g mole/min (added)

H20 4.7 g moles/min decomposed

12
 (a) If 0.1% of the main-stream liquid phase is let down, what are the flow rates of
each phase and the compositions of each phase at the four points indicated on the
diagram?

{b) Calculate and plot the over-all concentration of xenon in the main stream as a
function of the fraction of the liquid phase let down. How sensitive is this to the
Henry's law constant for xenon?

(c) What would be the effect of allowing the net decomposition rate of water to
increase by changing the amount of catalytic additive?

Solution. —

Nomenclature

X Mole fraction of component in liquid phase

y Mole fraction of component in vapor phase
L Total liquid-phase flow rate (moles/min)
V Total vapor-phase flow rate (moles/min)
p Partial pressure
pT Total pressure
H Henry's law constant (psi per mole fraction)
k Total (two-phase) flow of a component at point 3
z Fraction of liquid let down through the separator
E Efficiency of the gas separator
Z Average (two-phase) mole fraction of Xe

Subscripts
0 Oxygen (02)
H Hydrogen (H2)
Xe Xenon (Xe)
C Water (H20)
A
°2 + H2
Number subscripts and sub-subscripts refer to points in the system.

(a) The rate of production of xenon is

-* . ..in fissions sec 1 mole

5 x 106 watts x 3 x 1010 x 60 x x 0.30 yield
watt-sec min 6>Q2 x 1023 atoms

= 4.48 x 10"6 moles/min .

It follows from the over-all material balance that

flow rate of Xe = 4.48 x 10~° moles/min ,

flow rate of 02 = 0.01 moles/min ,

13
 4.48 x 10-6 x 100 _
per cent Xe = 100Y¥ = — = 0.0448% ,
x»4 0.01
per cent 02 = 100yQ = 99.9552% .
At point 3, the flow rates of the individual constituents are:
*03L3 + y03^ =k0 '
^.3L3 + yXe3^
XH3L3 + yH3V3 - ^H
From Henry's law,
Po = Ptyo3 = woxo3 '
P*. - Ptyx.3 = //x.xx.3 '
^H = ^TyH3 " WHXH3 '
By the definition of mole fraction,
y + y + y + y_ = 1 ,
°3 H3 Xe3 C3
but yXe is insignificant in this relation. Since pc is the vapor pressure of water,
which is known, we also know

Yc 3 = pT •

It is given that HQ = H„. It is convenient, then, to consider the sum of the oxygen and
hydrogen as a new component, A. Then,

H0 = HH = HA '
and

X0 + XH = XA ' Y0 + YH = YA > kO + kH = kA '

From this it follows that

XAjL3 + yA3^3 - *A •
By combining these equations, an expression for the vapor-phase flow rate may be
obtained:

kA - (pT/HA){1 - (pc/pT)]L3
V
3" 1 ~ (PC/PT)
By material balance,
kQ = 2.36 moles/min ,
k„ = 4.7 moles/min ,

k. = 7.06 moles/min ,

ky. = 4.48 x 10 moles/min

14
From the steam tables,
pc = 1250 psi ,
whence

Yr = 0.625 .
c3
The fraction of the liquid stream letdown = z = L3/L^ = 0.001. Therefore
L3 = 84.0 moles/min .
On substitution into the equation for V3,
V3 = 18.0 moles/min .
The equations for the flow rates and for Henry's law may now be used to give:

X0 = 1.25 x IP"3 , YQ = 0.125 ,

XH = 2.50 x 10"3 , yH = 0.250 ,

H3 H3

X¥X«3 = 3.83 x 10-' , yvXe3 = 2.3 x 10-7 .

At point 1, the compositions of the two phases are the same as at point 3. The
flow rate of the vapor phase is given by EV. = V3, from which
18.0
V, = = 28.4 moles/min .
' 0.034
At point 2, the stream is pure liquid water (by virtue of assumptions), and its rate
must equal the decomposition rate plus the rate at which water is carried past point 3:

L2 =*H +L3(' - X03 ~ XH3) +Myc3)

= 97.3 moles/min .

(b) In part (a) the equations for the sum of oxygen and hydrogen were used to
produce the expression
kA - (pT/HA)zL}YA
V, =
3 YAA3
A similar procedure using the equations for xenon gives
n„Xe XvX Bn
V-3
Xv zL, + = kv .
Xe3 1 p Xe

Xe
XYX
•3
e-
zL} + HXaV3/pT
It is desired to evaluate the average xenon concentration Z at point 1, which can be
written in the form

15
 l,xy
1 Xe.
+ v.1 yYXe.
z =
L, + V,
On elimination of Yv (by use of the Henry's law relation) and XY , the expression
"3 *e3
becomes

Xe
L, + (V,//Xe/pT)
zLy + (Hx,VfT)
Z =
L, + V,
Now, using V^E - V3 and substituting the derived expression for V3 give

zL, y
H
1 'A.
Xe
tf„
//
Xe
PTzLiyA. L, +
zLy + EpT
Pry H,
TlA,
Z =

PTzLiyA.
EYA kA ~ HA

By algebraic manipulation, the following somewhat simpler form is obtained:

HAHX,kA
Xe
EHA - zHXf
?TLiy/
Z =
HAHXe^A
z(HA - Hx.) + F +
L,y, H,
PrLiyA,

(c) The effect of the magnitude of the Henry's law constant is pointed out in the
plotted values (calculated from this equation) shown in Fig. 1.3.

PROBLEMS

Problem 1.1. - A helium stream containing 10~5 mole fraction of Xe135 is con
tinuously fed at a rate of 10 g/hr to a compressor; it is brought to 1000 psia, cooled
to 15°C, and detained in a 20-liter cylinder. Gas is continuously bled from the cylinder
and passed on to other operations. Assume that at time zero the cylinder contained
pure helium and that the gas in the cylinder is well mixed. Write the equation that
gives the build-up of xenon in the bleed stream as a function of time.
Problem 1.2. — A homogeneous reactor core system has a volume of 4000 liters.
The power level is such as to produce 30 mg/liter/day of fission products that will
precipitate (consider them completely insoluble). The precipitated solids are removed

16
 UNCLASSIFIED
ORNL-LR-DWG 22769
12 1

1
1

10

.9 9
\ x/j, =2X 105, Hg =\.2X 10S psi PER MOLE FRACTION
a> 8

O
k-V^"4 =2X<05, /r^ = 3 X1C 5

I-
<

•t-

uj 4

<
a. , /
\/

10 10" 10" 10 10 10 10u

FRACTION OF MAIN-STREAM LIQUID LET DOWN

Fig. 1.3. Average Xenon Concentration as a Function of Let-down (Part b of Example 2).

by two hydroclones in series, the underflow from the first being used as feed to the
second, and the solids being detained in a 4-liter pot in a recycle line from the underflow
of the second back to the feed port of the second (see Fig. 1.4). The overflow from the
second is cycled to the feed of the first, and the overflow from the first is sent back to
the reactor. The stream from the reactor is 1000 liters/day. Ten per cent of the liquid
entering each hydroclone reports to the underflow, but eighty per cent of the solids
entering each hydroclone reports to the underflow. Determine the flow rates in all parts
of the system, and write the differential equations that describe the solid concentrations
in the reactor core and in the underflow pot in terms of the generation, appropriate
volumes, the two concentrations of interest, and time. What is the limiting ratio of pot
concentration to core concentration approached by long operation? Solve these equations
on the assumption that the concentration is everywhere zero at time zero (last part
optional).

17
 UNCLASSIFIED
ORNL-LR-DWG 22770

Fig. 1.4. Flow Sheet for Problem 1.2.

18
 Chapter 2

COMMON CONCEPTS USED IN SEPARATION OPERATIONS

The overwhelming majority of the demands made on the chemical engineer in reactor
chemical technology are concerned with separation. The separation of impurities from
reactor materials set new standards in commercial materials, the separation of isotopes
for nuclear purposes is an unparalleled accomplishment, and the decontamination of irra
diated reactor fuels by factors of 10° and higher when impurities were originally present
in fractions of a per cent is an undertaking of fantastic implications. Some understanding
of this technology is important to all who are seriously engrossed in the field of nuclear
energy.

The operations involved include precipitation, distillation, solvent extraction, ion

exchange, gaseous diffusion, and several others. With the possible exception of precip
itation, these operations have many things in common. All are of the multiple-stage coun-
tercurrent type of operation, which involves the manipulation of equilibrium conditions
and material balances to repeatedly make use of a small effect until the desired separation
is accomplished. It is the present purpose to discuss those things common to all these
operations so that, as each operation is later covered in detail, its relation to the others
may be understood.

The word "separation" implies some method of dividing the system to produce two
exit streams of different concentrations. For this to be effected, there must be two dis
tinct phases: liquid and solid, liquid and vapor, liquid and liquid, etc. Separations which
are considered strictly chemical in character finally must depend upon a physical phase
separation. When the separation depends essentially upon chemical changes, the sepa
ration is known as a "unit process." When the separation exploits essentially physical
characteristics such as solubilities or vapor pressures, the separation is known as a
'unit operation." Precipitation is the only separation listed above which is a unit proc
ess; all the rest are unit operations.

The term "equilibrium," as used here, is to be taken in its thermodynamic sense.

When applied to a two-phase system, it corresponds to the condition of lowest free energy,
or that state which is reached after long contact, or the condition when no further mass
transfer between the phases is possible.

In a continuous operation there is usually a transient period after start-up during which
concentrations change rather rapidly, but after a while, ifthe feed rates and other operating
variables are held constant, everything levels out, and the concentrations of the exit
streams as well as all intermediate streams in the operation become constant. This is
known as "steady state." At steady state the inventory of the system must remain con
stant, and material balances around any section of the operation can be made.

19
 An analysis of all steady-state separations can be effected using four concepts:
1. Equilibrium. This provides a functional relation which allows the determination of
the concentrations in one phase if the concentrations in the other are known.
2. The theoretical stage. This divides the operation into increments for which the
equilibrium relation can be used and which can be used in turn as design criteria.
3. The material balance. This provides a functional relation which allows the calcu
lation of what is leaving a group of stages from what is going into it.
4. An interstage material balance condition. This is a characteristic of the particular
operation, and, in general, provides a form for the material balance which allows it to be
extended from stage to stage. These concepts will be considered below in this order.
Of particular interest to separation operations are systems with one degree of freedom.
For such a system it is possible, at equilibrium, to express any intensive property of the
system as a function of any other intensive property; particularly, to express a concen
tration in one phase (or any index thereof) as a function of a concentration in the other
phase.
Two-Phase Systems with One Degree of Freedom. - The specification that the system
have a single degree of freedom is not an inherent requirement for a successful separation
operation, and many operations have more. But when there is more than a single degree of
freedom, the analysis of cascaded systems becomes tremendously complex. It behooves
us, in this simple treatment, to concentrate on systems with a single degree of freedom, or
those reducible to a single degree of freedom by reasonable assumptions.
The Gibbs phase rule is stated:

F «= C - P + 2 ,

where F is the number of degrees of freedom (the number of variables necessary to fix the
system), C is the number of components (number of substances required to construct the
system), and P is the number of phases.
The two-phase single-component system is trivial to separation processes, since con
centration is not a variable. The one degree of freedom would be either the temperature
or the pressure, and the resulting relations would be of the nature of the saturated steam
tables.

In a two-phase two-component system, the variables include the concentration in the

heavy phase (xj, the concentration in the light phase (yj, the temperature, and the
pressure. The phase rule gives two degrees of freedom for this system, but, if the oper
ation proceeds at constant pressure (as is usually the case), it reduces to the desired
single degree of freedom. An obvious illustration of such a system is a binary distillation
system. The boiling point of the mixture is a simple function of its composition; also, the
composition of the vapor is simply related to the composition of the liquid. In fact, any
of the three variables may be simply plotted against either of the others.

20
 When there are three components, the variables include two concentrations in the
heavy phase, two concentrations in the light phase, the temperature, and the pressure.
The phase rule allows three degrees of freedom. This system is reduced to one having a
single degree of freedom if both the temperature and the pressure are fixed. Distillation is
then eliminated from consideration. However, liquid-liquid solvent extraction is exactly
such a system when there are but one solute and two solvents. In this system it should
be possible torelate all concentrations in either phase to any concentration in either phase.
While thermodynamically this is strictly true, it is not practical to relate them to the con
centration of light solvent dissolved in the heavy phase or to the concentration of heavy
solvent dissolved in the light phase, because these concentrations are usually small.
When there are more than three components, there are more than three degrees of
freedom, and specification of both the temperature and the pressure will not reduce the
number of degrees of freedom to 1.

Systems Reducible to One Degree of Freedom. - It is logical that if the concentration

of a component in a system is to be considered a variable of the system, there should be
some sensible response of the system to changes in this concentration. When components
are added to a system in very small quantities, it may be that nothing but their own con
centrations is sensitive to their presence, and, as far as the gross constituents are con
cerned, they are not significant components of the system. Thus, if there is a system of
two solvents, 5(1) and S(2), and two solutes, A and B, and if B is present in very small
quantities, the system 5(1), S(2), A might be analyzed regardless of B, and, as such,
would be a system of one degree of freedom. If A is a major component, its presence can
not be disregarded in the analysis of the behavior of B. If A also is a minor component,
then analysis of the behavior of each solute can be made without regard for the presence
of the other, and the two solutes are "independent."

Another common case is the addition of a solute which does not distribute itself
between the two phases but remains with a fixed concentration in one phase. Actually,
the specification of the fixed concentration of the added component absorbs the degree of
freedom introduced by its addition. Or, alternatively, one might redefine the solvent in
question to include this additional component. The result is the same: a system of one
degree of freedom before the addition of such a component remains a system of one degree
of freedom. Examples of such components are diluents of organic phases and certain
salting agents in aqueous phases in solvent extraction.

A more complex case of a four-component system reducible to one of a single degree

of freedom is the case of a two-solute solvent extraction system, 5(1), 5(2), A, B, where
the two solutes are so similar in properties that they behave nearly as a single component
in the system. With the use of these symbols to represent the amounts of the respective
constituents, the concentrations may be defined as follows:

21
 A . 5(1)
X,
1 A + B + 5(1) + 5(2) ' 3 A + B +5(1) +5(2) '
B 5(2)
X- = -— —— , A. -
2 A + B + 5(1) + 5(2) ' * A + B + 5(1) + 5(2) '
A'
1 A' + B'+ 5(1)'+5(2)' '
etc., where primes denote the light phase. Let
ot«=x,+x2, «=y1 + y2.
The statement that the two solutes are sufficiently similar to act as though they were a
single solute is to say that m is a simple function of n in a system of one degree of
freedom. An illustration of this occurs where A and B are isotopes of a solute. More
realistic an example is the case where A and B are adjacent rare earths. (Without violation
of this condition it may be true that there is a subtle separation between A and B.) If
r = X./m and v «= Y,/«, it may be possible to treat r and v as concentrations in a binary
system of one degree of freedom. It more frequently turns out, however, that r and v are
functions of the concentrations m and n, as well as of each other.
Systems Not Reducible to One Degree of Freedom. —When the system does not yield
justifiably to reduction to a single degree of freedom, it is necessary to express the
equilibrium relationships in as many variables as may be required. Sometimes this can be
done through simple relations when the system is "ideal," sometimes empirical equations
are available, and sometimes it is necessary to obtain data on a multiplicity of points
and to use the data plotted on three-dimensional plots or worse.
There is, however, a frequently encountered case for which calculations are not too
difficult. When a system can be broken down in such a manner that a one-degree-of-
freedom calculation can be made by exclusion of one (or more) components from con
sideration, calculations on the excluded component can be made directly if the effect of
the major components on the distribution of the minor component has been determined.
Specifically, the effect of the uranium concentration on the distribution of the rare earths
in a solvent extraction system is known. The calculation of the behavior of the uranium
can proceed with neglect of the effect of trace amounts of rare earths. This establishes
the concentrations of uranium in the extractor and allows the calculation of the behavior
of the rare earths.

Another illustration is the system involving A and B, described above. If it is possible

to express v as a function of r and m, and if it is possible to calculate m for a system
having one degree of freedom, then the separation of A and B should yield to direct
calculation.
Forms of Equilibrium Relationships. - It has been repeatedly emphasized that the
equilibrium relationships may be presented in any fashion. The important point is that

22
the equilibrium relationship is a simple single-valued function only when there is but one
degree of freedom. It will be found convenient in the discussion of the various cascades
to use different bases and units of concentration. There is nothing sacred about any one
basis or set of units, and the choice will be dependent upon the material balance (or
interstage material balance) condition being used.

The Theoretical Stage. —Countercurrent cascades can be considered as long devices

being fed with two different phases, one into each end, which move generally opposite to
each other, transferring mass as they move, and which finally leave at opposite ends.
Countercurrent cascades come in many styles. In one extreme, the cascade consists of
a sequence of distinct stages, each of which brings the two phases very near to equilib
rium, separates them, and passes them on. In the other extreme, the phases are maintained
in contact and in countercurrent motion uniformly throughout the length of the cascade,
transferring mass all the way, yet achieving equilibrium at no one point. Between these
extremes there are many examples of nearly discrete stages, sometimes in the form of
plates or trays, where the phases are momentarily held but where the approach to equilib
rium is far from complete.

Regardless of what the actual cascade may be, it can for analysis be approximated
by its equivalent in a sequence of stages each of which is perfect, bringing the phases to
equilibrium and separating them. When a stage does this it performs the function of a
theoretical stage. The theoretical stage is defined conceptually as a device which dis
charges separated phases that satisfy the equilibrium relationship.

The General Cascade Problem. - Now that the cascade has been established as a
sequence of theoretical stages and it is known that the streams leaving each of these
stages satisfy the equilibrium condition, how is this applied to cascade calculations?
If the whole cascade is considered as a unit, the variables include the concentrations of
both entering streams, the concentrations of both leaving streams, the rates of these
streams, and the number of stages in the cascade. Various combinations of these may be
specified or required, but, if adequate information is given, use of the over-all material
balance equations will reduce the problem to one of two general types: (1) all the entering
and exit concentrations and rates are known, and the number of stages remains to be
determined; or (2) the entering and exit concentrations and rates at one end are known,
the number of stages is known, and the problem is to determine the conditions at the other
end. In either case the method of operation is the same. Calculations are performed one
stage at a time until the other end of the cascade is reached.

If it is possible to start at an arbitrary point in the cascade, between two stages, and
from the known values of all the variables at that point to calculate all the variables at a

23
point one stage away, then it is possible to calculate the whole cascade. Such an arbi
trary point is shown in the diagram below. The symbols L and V represent flow rates of
the heavy and light phases, respectively, in such units that the products Lx and Vy are
the mass rates of the same solute in the respective phases, where x and y are the concen
trations of the solute in units convenient for the particular cascade in question. The
UNCLASSIFIED
ORNL-LR-DWG 22792

fc-2 fc-1 vn
//i-2 fn

rt-2 n-< n

L„-? l-n-< <-„ <-„+,

heavy phase enters stage n at a rate Lw+] and a concentration x ,,, and the light phase
leaves at a rate Vn and a concentration yn, all of which are known. To be determined are
the four variables Ln, xn, V„_1# and yn_y The relations available for doing this are:
1. the equilibrium relation, x = f{y ),
2. the total material balance, L + V = L ., + V ,,
3. the material balance for the component in question,

L„x„
n n
+ V„y„
n' n
= L„+1x ., + V„n - K
rc + 1 w + 1
iV„n - l, .

There are three equations and four unknowns, and therefore the system is indeterminate.
It is necessary, then, to specify some other relation, which will be called "the inter
stage material balance condition" in its most general sense. It may be a relation of L to
V„_yi or Ln +y to Ln, or Vn to Vn_y The principal difference between thecascades which
will be studied is in the interstage material balance conditions.
A basis for an interstage material balance condition can be a quantity which passes
through the cascade unaltered and constant. In distillation it is the heat which passes
from the still pot to the condenser. In solvent extraction it can be the masses of solvent
carriers. Where there is varying mutual solubility between the solvents in solvent ex
traction, or where heat is lost from the tower in distillation, the interstage material bal
ance condition can become complex.
Working Form of Material Balances. — Although the above discussion was based on the
material balance taken around a single stage, it is more convenient in practice to include
all the stages from one end of the cascade to the stage in question. This in effect says
that the difference between what is going into an end of the cascade and what is coming
out of that end must be the net movement of material through the cascade, which is con
stant and is equal to the difference between what is going out and what is coming in at
the arbitrary point in the cascade. This is true for the total mass flows and for the flows
of each individual component. Stated mathematically, the material balance taken- around
the top end of a cascade is:

24
 Lnxn ~ Vn-\yn-\ = Lff ~ Vp
and

K ~ Vn,} =L7- Vp ,
where L, and x, are the rate and concentration of the feed and V and y are the rate and
concentration of the product. A similar material balance could be taken around the bot
tom of the cascade.

At the ends of some cascades there are auxiliaries which are closely tied to the
cascade and may serve such functions as changing the exit phase to the entering phase and
returning part of it to the system. In such cases it often becomes expedient to include
such auxiliaries in the material balances.
What is covered here is applicable to all the separation operations to be considered.
More generalization would add little clarification, and the application of these principles
will become more clear as the individual operations are discussed.

25
 Chapter 3

DISTILLATION

As a separation operation applied to radiochemical processing, distillation has yet

to come into its own. The structural simplicity of a distillation cascade and the fact
that there is a chemically stable, volatile compound of uranium, UF,, promise that a
radiochemical plant based on distillation may be feasible. Such a plant was considered
from the beginning of the Manhattan project, but the technical problems held back de
velopment. Work is continuing, and the promise is still real.

As a means for separating isotopes, distillation has certainly proved its worth. The
heavy water used during the early days of the project was, for the most extensive part
of the enrichment, enriched by distillation. In recent years alternative processes have
grown, and among them is the low-temperature distillation of elemental hydrogen.

An additional incentive for the study of distillation is that it embodies all the
principles of countercurrent multistage operations and is considerably better treated
in the literature than the other separation operations which we will later find to be
important. Distillation is the oldest of the cascade operations and perhaps the best
understood.

Distillation is an operation for the separation of two volatile substances by em

ployment of the countercurrent contacting of a vapor phase by a liquid phase. This
distinguishes it from the operation of evaporation, in which one of the components is
nonvolatile, and from sublimation, in which the vapor is in equilibrium with a solid
phase.

The heart of a distillation facility is a column, which is usually either a packed

tower or a bubble-plate tower. The packed tower is simply filled with some material
of high surface area (Raschig rings or Berl saddles) to enhance the contact of liquid
with vapor. In the bubble-plate tower, spaced a foot or two apart, are horizontal plates
or trays connected by downcomers and covered with holes or bubble caps. Small dams
back up the liquid to a depth of a few inches on each plate, and the vapor, passing
through the bubble caps, is brought into intimate contact with it. Above the tower is
a condenser, which condenses the vapor to liquid, most of which is returned to the
tower as reflux, while part is drawn off as product. Below the tower is a reboiler,
which collects the liquid and returns most of it to the tower as vapor, while some is
withdrawn as waste. The feed to the system is usually introduced at some point along
the height of the tower (see Fig. 3.1).

A theoretical ideal stage in a distillation system is that length of tower which

receives liquid and vapor at some concentrations and discharges them in equilibrium
with each other. In a bubble-plate tower, a single tray may approach an ideal stage.

26
 UNCLASSIFIED
ORNL-LR-DWG 23358

VAPOR

PRODUCT
LIQUID

DISTILLATION
TOWER

FEED

WASTE

Fig. 3.1. Schematic Diagram of a Distillation Operation.

As is shown in Fig. 3.2, a given plate, n, receives liquid from plate n + 1 and retains
it by a weir at a level that covers the bubble caps. Before it can flow down to plate
n - 1, the vapor from plate n - 1 bubbles up through the bubble caps and mixes in
timately with the liquid. A small part of the less volatile component is transferred from
the vapor to the liquid, while a small part of the more volatile component is transferred
from the liquid to the vapor. All this, of course, takes place continuously. If the mixing
has been perfect, then the vapor leaving the plate on its journey up will be in equi
librium with the liquid on its way down, and this plate will function as a theoretical
stage. In practice, due to poor contact, or entrainment, or other causes, it may be found
that it takes more plates to effect a separation than the calculated number of theoretical
stages. In this case, the ratio of the number of theoretical stages to the actual number
of stages becomes a stage efficiency, which can be used in subsequent design cal
culations. When packed towers are used, they can be characterized by the height of
tower required to develop a theoretical stage. This is commonly referred to as the
"height equivalent to a theoretical plate" and abbreviated HETP.

27
 UNCLASSIFIED
ORNL-LR-DWG 22771

^V^^F-5

Yn

STAGE n

*
WC~7feff~^S

/7-1

Fig. 3.2. Stages (or Plates) in a Distillation Tower.

Equilibrium in Distillation Systems. - Distillation systems, involving a two-phase

constant-pressure operation with temperature as a variable, will have a number of
degrees of freedom equal to the number of components minus 1. The number of volatile
components in the feed can be very numerous, as in the case of straight-run petroleum,
but a single cascade divides the feed into only two parts. In a highly efficient cascade
separating a mixture of a large number of components, the division is made between two
components adjacent in characteristics, and everything more volatile than the more
volatile of these two goes to the product, while everything less volatile than the less
volatile of these two goes to the waste (note that "waste" is the name of a stream and

28
is not necessarily that which is thrown away). There are precise methods for cal
culating multicomponent systems which are useful for exacting work, particularity for
cascades of poor effectiveness. For purposes of illustrating the principles of dis
tillation, this discussion can be restricted to two-component systems without serious
loss. Multicomponent systems can be reduced to two-component systems by consider
ation of those two components between which the separation is to be made. There is
error in doing this, but qualitatively the answers are reasonable.

For a two-component constant-pressure distillation system, there is only one degree

of freedom. Any intensive physical property of the system can then be expressed as
some function of any other intensive physical property of the system. For instance,
the boiling point of a liquid mixture of these two components can be plotted against
the composition. Similarly, the condensing temperature, or dew point, of a vapor can
be plotted against its composition. Such a plot is shown in Fig. 3.3. It is noticed that
the dew point and the boiling point are the same only for the pure components A and B,
while for any given mixture these temperatures are different. In the region between the
two curves, at a constant temperature, two phases exist, with concentrations indicated
by the points on the two curves at opposite ends of a horizontal line. Thus, for equi
librium at any temperature within this temperature range, there is a liquid-phase compo
sition and a vapor-phase composition. Instead of this plot, the concentration of the
vapor can be plotted against the concentration of the liquid with which it is in equi
librium (see Fig. 3.4). When the vapor composition is expressed in terms of the mole
fraction of the more volatile component (Y. or simply Y) and the liquid composition is
expressed in the same way (X. or simply X), this is known as an X-Y diagram, or an

UNCLASSIFIED UNCLASSIFIED
ORNL-LR-DWG 22772 ORNL-LR-DWG 25571

X
A'(mole7oOF/l IN LIQUID)
mole 7o OF ,4

Fig. 3.3. Constant-Pressure Phase Diagram. Fig. 3.4. An X-Y Operating Diagram.

29
operating diagram, qnd is very useful for the graphical solution of cascade problems.
This is the way equilibrium data are usually used.
Ideal Systems. - An ideal system is one for which both Raoult's law and Dalton's
law hold. Only at very high pressures do variations from Dalton's law become ap
preciable, while it is an exceptional system for which Raoult's law is valid over the
whole range of concentrations. Isotopic systems and mixtures of homologous hydro
carbons are examples of ideal systems. These are easy to study and constitute a point
of departure for less well-behaved systems.
Raoult's law states that the partial pressure of a component over a liquid mixture
is the vapor pressure of the pure component at the temperature of the mixture times the
mole fraction of the component in the liquid:

(3-D PA = PAXA , PB = PBXB .

Dalton's law states that the sum of the partial pressures is the total pressure, or
that the mole fraction of a component in a vapor phase is its partial pressure divided by
the total pressure:

Pa Pb
(3.2) YA = , YR =
A Pa + Pb ' B Pa + Pb
On substitution of Eqs. 3.1 into Eqs. 3.2 and division of YA by YB,

YA Pa paxa
YB Pb pbxb
from which

PA ya/yb
(3.3)
PB XA/XB
At constant pressure, the temperature decreases as the concentration of the liquid phase
is increased from pure B to pure A, so that P„ is equal to the total pressure when B
is the only constituent, and P. is equal to the total pressure when A is the only con
stituent. If the temperature difference is not too large, the ratio of these vapor pressures
will not change appreciably, and the quantity a will be a constant. This quantity is
known as the "relative volatility" in distillation and has a counterpart called the
"separation factor" in other types of cascades. It can, of course, be defined whether
or not it is constant.

Now, note that the sum of the mole fractions in either phase is unity, so that compo
sition can be specified by means of the concentration of either component. The nomen
clature is somewhat simplified if the convention is adopted that concentration with no
subscript refers to the more volatile component. Thus,

X = X.A ,
'
and Y ^ Y.A .

30
Then Eq. 3.3 becomes

Y/(l - Y)
(3.4) — = a .
V ' X/(1 - X)

Equation 3.4 can be solved for the composition of the vapor as a function of the compo
sition of the liquid:

aX
(3.5) Y =
1 + (a - 1)X

When plotted on an X-Y diagram, Eq. 3.5 gives a regular bow-shaped curve. The equi
librium line becomes a 45-deg line when the relative volatility goes to unity, and then
no separation is possible. As the relative volatility increases above unity, separation
becomes increasingly easy, and the equilibrium line moves out away from the 45-deg
line (see Fig. 3.4).

Nonideal Systems. — Simply, all systems which are not ideal are nonideal. This
means that the relative volatility changes with concentration. It may change from a
value much larger than 1 to a value very close to 1, or it might even become less than 1.
When the relative volatility goes from larger than 1 to less than 1, we say that the
system is "azeotropic." Or, an azeotropic system can be defined as a distillation
system in which the relative volatility changes with concentration to the extent that
one component is the more volatile over part of the concentration range and another is
the more volatile over the rest. In a system like this, there is a composition such that
the composition of the vapor is the same as the composition of the liquid, and no sepa
ration by distillation is possible. This concentration is called the "azeotrope." The
azeotrope may be either the most volatile composition possible for the system or the
least. In the first case, it is called a minimum-boiling mixture, and the product concen
tration always approaches the azeotrope concentration. In this case either component
can be considered the more volatile when it is dilute. In the second case, it is called
a maximum-boiling mixture, and the product can approach a pure component, while the
waste approaches the azeotrope concentration. In this case neither component can be
considered the more volatile when it is dilute, but either can be considered the more
volatile when its concentration is greater than the azeotrope concentration. These
systems are shown in Figs. 3.5, 3.6, and 3.7.

Generally, the equilibrium data for nonideal systems must be determined experi
mentally. The literature does have some rules based on thermodynamic considerations
by which these data may be determined or expanded. Perhaps the most applicable of
the simple rules is that Raoult's law always holds as X approaches unity, and that for
very low concentrations a similar law, Henry's law, says that the partial pressure of a

31
 UNCLASSIFIED UNCLASSIFIED
ORNL-LR-DWG 25572 ORNL-LR-DWG 22773

L±J
rr
=>
r-
<
cr
Ll)
o_

Ld

mole 7o OF A mole %OF/4

Fig. 3.5. Phase Diagram Showing a Minimum- Fig. 3.6. Phase Diagram Showing a Maximum-
boiling Azeotrope. boiling Azeotrope.

UNCLASSIFIED
ORNL-LR-DWG 25573

Fig. 3.7. Operating Diagram for an Azeotropic Systen

Having a Minimum Boiling Point.

32
volatile component above a liquid is proportional to its concentration, that is,
(3-6) pA = HXA ,
where H is the Henry's law constant and must be determined experimentally. Since
these two laws apply at opposite ends of the concentration range, H is equal to P.
only for ideal systems.

Material Balances. - The general material balances for distillation cascades are
made on the basis of total molal flow rates (moles of A plus moles of B per unit time)
and concentrations expressed in mole fractions of the more volatile component. The
symbols for total molal flow rates will include L, V, F, P, and Wfor the flow rates of
liquid, vapor, feed, product, and waste, respectively. Liquid-phase concentrations will
generally be denoted by X and vapor-phase concentrations by Y. Appropriate subscripts
will be used to indicate the location in the cascade.

The distillation cascade makes use of a device common to many cascades, that of
using reflux at its ends. The vapor leaving the cascade passes to a condenser, where
all of it is condensed to liquid at its boiling point. This liquid stream is split into a
product, which is withdrawn from the system, and a reflux stream, which is returned to
the cascade. A material balance around the condenser, then, is
(3.7) v - P = r .
P P

When a material balance is taken around an arbitrary number of stages at the top of
the cascade, it is convenient to include the condenser (see Fig. 3.8), so that the general
material balance equations are

(3.8) V - L - P
«— I n

and

y»-lV«-l = PYp + LnXn ' PY» + (VB_, - P)X„

(3.9) Yn , =-^ *H+—.

In the analysis of the part of the distillation column below the feed point, the boiler
should be treated as the condenser was above. Liquid falls from the bottom of the
cascade into the boiler, where heat is added, and part of it is returned to the vapor
state, while part of it is removed as waste:

(3.10) L - W= V .

33
 UNCLASSIFIED
ORNL-LR-DWG 22774

"~l

CONDENSER

REGION FOR UPPER

MATERIAL BALANCE EQUATION
K

L„\ Vn-y
Yn-K

') -n-\

K77-1

F
Xf Ym

•-m t^-1
Xm Ym-\

"] Lm-\
Xm-\
REGION FOR LOWER
MATERIAL BALANCE EQUATION

\** W
BOILER Xw

Fig. 3.8. Material Balances on a Distillation System.

The material balances around the boiler and some arbitrary number of stages above the
boiler are

(3.11) L„ - Vm i - w

and

WX
v„-i + w
(3.12) m-1
X
m y
m-1 m-1

(the subscript n is used above the feed point, and m below). It is thus noted that, in the
part of the cascade above the feed point, the vapor rate exceeds the liquid rate, and that,

34
in the part of the cascade below the feed point, the liquid rate exceeds the vapor rate.
The difference is the feed rate, and the two material balance equations are tied together
by the over-all material balance, which states that the feed rate is equal to the waste
rate plus the product rate:

(3.13) F = P + W ,
and

(3.14) FXf = PYp + WXw .

A single difference between the treatment of the top part of the cascade and that of the
bottom part is that the condenser is considered to condense all of the vapor, making
the composition of the vapor entering the condenser, the product concentration, and
the concentration of the returning liquid equal to each other. This is not the case with
the boiler, which acts like a stage in the cascade, returning vapor to the cascade which
is in equilibrium with the discharged waste (rather than of the same concentration). The
ratio of the liquid returned to the cascade by the condenser to that withdrawn as product
is called the "reflux ratio" and given the symbol R. The ratio of the vapor returned to
the cascade by the boiler to that withdrawn as waste is called the "boilup ratio." The
equation relating Y j to Xn when plotted on an X-Y equilibrium diagram becomes an
"upper operating line," while a similar plot for the part of the cascade below the feed
point is the "lower operating line."
Interstage Material Balance Condition. - The equations developed to this point are
general and rigorous; however, they are insufficient to solve the cascade problems.
The interstage material balance condition is still to be specified.
Perhaps the most nearly correct condition which could be imposed is a general heat
balance, taken in the same manner as the material balances above. This would state
for the top of the cascade that the heat contained in the vapor rising into stage n minus
the heat carried by the liquid descending from stage n must equal the heat removed in
the condenser plus the heat carried off by the product. (This just assumes steady state
with no heat sink in the system other than the condenser.) These heat rates would
depend upon the total molal flow rates, Ln and V v and the molal enthalpies, which,
in our two-phase constant-pressure system, are functions of composition alone. This
would give the relation needed to allow cascade calculations. A very workable graphical
technique based on this enthalpy method has been contrived by the workers Ponchon
and Savarit and is presented in several standard chemical engineering texts (see bibli
ography). It is still awkward for purposes of illustration and lends itself poorly to
extrapolation to other separation operations which are to be studied. A less accurate
assumption, but one which makes things simpler, is that the molal heat of vaporization
of the liquid mixture is constant (independent of concentration). This is to say that
V moles of vapor leaving the boiler will contain exactly enough heat to vaporize V

35
moles of liquid at any concentration, or that the vapor rates leaving all stages above
the feed point are equal. The material balances then tell us that the liquid rates over
these regions are also equal. This condition is described as "constant molal overflow"
and produces what is known as a "square cascade." It will also be noted that the
operating lines (the plot of the material balance equations on the equilibrium diagram)
are straight.
There is one remaining point of simplification necessary before cascade calculations
can be made. The cascade must be sensitive to the heat content of the feed. The feed
might be superheated vapor, saturated vapor, a mixture of liquid and vapor, liquid at
its boiling point, or liquid below its boiling point, and accordingly will join either the
vapor stream or the liquid stream, causing additional condensation or vaporization as
the heat balance demands. Although the treatment of the general case is not overly
difficult, little is lost if this treatment is restricted to liquid feeds at their boiling point.
Under this restriction, the vapor rate in the top section of the cascade is equal to the
vapor rate in the bottom section, and the liquid rate in the bottom section is larger than
the liquid rate in the top section by the rate at which feed is introduced.
The cascade problem as it might be posed would, then, specify the compositions of
the feed, product, and waste and a reflux ratio, and it would require the determination
of the number of stages needed in each section. It is not necessary to specify the rates,
since the operation can be referred to a unit feed rate, the relative product and waste
rates being determined by the over-all material balances. Similarly, the boilup ratio is
determined by the reflux ratio and the given concentrations. (The student should satisfy
himself that these things are true.)
McCabe-Thiele Graphical Method. — The solution of a cascade problem when the
end conditions and reflux ratio are specified includes a stage-by-stage calculation of
concentrations from one end of the cascade, say the top end, to the feed point, in which
the upper material balance equation is used; then a stage-by-stage calculation from the
feed point to the other end, in which the lower material balance equation is used; until
finally the required number of stages is determined.
Suppose that the calculation is started at the top end, where the vapor leaving the
cascade and the liquid returning are known to be of the product concentration. The
liquid leaving the top stage is known to be in equilibrium with the vapor leaving it.
The concentration of this liquid, when used in the upper material balance equation,
allows the calculation of the composition of the vapor entering the top stage. The vapor
entering the top stage is the vapor leaving the next stage down, and must be in equi
librium with the liquid leaving that stage.
And so the procedure goes. First the equilibrium condition is used, then the material
balance condition, and so on down the cascade. This involves many numerical calcu
lations which can be eliminated by use of a simple graphical method. Both the equi
librium line and the operating lines are plotted in terms of mole fraction of the more

36
volatile constituent in the vapor against the mole fraction of the more volatile con
stituent in the liquid. The operating line (the material balance equation) will be straight,
will lie between the equilibrium line and the 45-deg line, and will intersect the 45-deg
line at the product concentration. The solution of the two relations alternately consists
merely in stepping off the stages by drawing horizontal lines from the operating line to
the equilibrium line, then dropping vertical lines from the equilibrium line back to the
operating line. This procedure, when for instance one is working down from the product
concentration, is exactly the same as using the Y value for the product to get an X value
for the liquid leaving the top stage, putting this X into the material balance equation to
get a Y, putting this Y into the equilibrium relation to get an X, and so on. Each point
thus made on the equilibrium line represents a stage. This graphical procedure is shown
in Fig. 3.9. Note that, after the feed concentration is reached, the lower operating line
is used instead of the upper.

UNCLASSIFIED
ORNL-LR-DWG 22775

STAGES REQUIRED = 10.3

^s^y
9
>9 //
7

y^y7/
6ZK/
5Zir X
4/_/ y \
/ ' /\ / i L
JZ y f\ oLUrL — .,

/ I / / \
ZI~Y/ \
1/ / / r-\ r\nrr — ™ \
rw oLUrL - i, \
i

x„w xf xa X\0 XP~YP

Fig. 3.9. Typical McCabe-Thiele Graphical Solution to a Distillation Problem.

37
 Limiting Conditions. —After the diagram is studied for a while, it becomes obvious
that the'number of stages necessary to make a separation and the reflux ratio used are
interdependent. The number of stages goes down as the reflux ratio goes up. The
limiting conditions occur when the reflux ratio goes to infinity, that is, when no product
is removed; and when the reflux ratio is a minimum, in which case an infinite number
of steps is needed in order to reach the equilibrium line at the feed point. All real
operations must operate somewhere between these limiting cases.

The case of an infinite reflux ratio is called "total reflux," and the operating lines
merge together, coinciding with the 45-deg line. This of course results in the minimum
possible number of stages needed to make the desired separation. Since at total reflux
there is no product or waste removed, no feed is introduced, and the feed composition
becomes unimportant; only the product and waste concentrations need be considered.
From the material balance equations it is also found that

(3-15) Ln = VB_, and X„ - Y„_, .

As a further consequence of total reflux, the interstage material balance becomes

meaningless, and it is found that all cascades operating at total reflux are equivalent.
Here, however, the major significance of total reflux is that it corresponds to the con
dition requiring the minimum number of stages, and that any real installation must
provide more theoretical stages than this number. (Twice this number is sometimes
considered a good rule of thumb.)

Since practical installations withdraw product, something less than total reflux is
used. There is, however, a point beyond which the reflux ratio cannot be further reduced
unless the end conditions are changed, even though an infinite number of stages be
available. This is the condition where the intersection of the operating lines touches
the equilibrium line. Any number of excess stages could operate at this point of contact
at the same concentration. This is called "pinching-in." The value of the minimum
reflux ratio can be obtained by taking a material balance around the top of the cascade
to a point just above the feed point. Because of the pinching-in, all stages in this
region will have the same liquid concentration as the feed, X,, and the vapor will be in
equilibrium with it, having a concentration which we will call Y,. Then

LX, PX„
Yf/ - -J'
v
+-t
V
,
or

(3.16) (L + P)Yf = LXf + PXp ,

38
from which it follows that

(3.17)
L
— = R
XP - ^
P
m in
Yf-Xf
The effect of the reflux ratio is shown in Fig. 3.10.
Significance of Reflux. - One of the most difficult things for a student to grasp
when considering multistage cascades is the necessity for reflux. On more than one
occasion he asks himself, "Why is it necessary to return to the system so much enriched
material once it has been separated?" The answer comes in many forms as soon as
some familiarity has been established, but, for one who has recently been introduced
to such cascades, perhaps a look at the effectiveness of a given stage is a good place
to start.

If a material is to be purified, its impurities must be carried back down the cascade.
The stream that carries the impurities cannot, however, have a concentration of im
purities higher than the equilibrium conditions permit. When the relative volatility is

UNCLASSIFIED
ORNL-LR-DWG 22776

Yf

X», X,
XP

Fig. 3.10. Effect of Reflux Ratio on Operating Lines when End Conditions Are Specified.

39
small, the rate of movement up the cascade cannot be much larger than the rate of
movement down, and the product rate, which is the difference of these, is comparatively
small. The mathematical statement of this follows. If the upper material balance
equation 3.9 is written in the form

(3.18) -VYn_} = -LXn - PYp = -VXn - P(Yp - Xj ,

and if the quantity VY is added to both sides, it becomes

V(Y« ~ Yn-J = V<Yn " *«> ~ P^Yp " Xr) <

(3-19) Y„ - Y„_, - Yn - Xn - -J^- (Yp - Xn) .

In words, Eq. 3.19 says that a stage achieves less enrichment than a single equilibrium
contact by an amount proportional to the distance (concentrationwise) from the product
end of the cascade and inversely proportional to the reflux ratio. The difference in
concentration between the liquid and the vapor in equilibrium with it is usually quite
small compared with the difference between the product concentration and the concen
tration on most stages. This can be compensated only by adequate values of the reflux
ratio. In a distillation cascade having constant molal overflow the reflux ratio is
constant from stage to stage, so its value must be such as to allow some finite en
richment on the stage farthest from the product end. Since the expression derived above
is valid only for stages above the feed point, the maximum distance from the product
end occurs at the feed point. It may also be noted that, when the reflux ratio increases
without limit, the last term drops out, allowing each stage to achieve its maximum
effectiveness, that of equilibrium contact, and X = Y j, which is a starting pointfor
total-reflux calculations.

Total Reflux in Cascades with Ideal Systems. - When the relative volatility is
constant and there is an expression relating Y and X at equilibrium, the calculation of
the number of stages in a cascade at total reflux reduces to a simple form.
First, note that since there is no product removal L = V _.. Then, necessarily,
X- - Y .. Recall that the definition of relative volatility is:

YJV ~ YJ
(3.4) —t = a .
X n / 1 - X n')

This is also true for stage n — 1:

Y n— /(l
I v
- Yn—V
,)
xn-1,/(i
v
_ xn—V~) = a

40
 The product of these expressions, after cancellation of corresponding factors in X and
Yn-V becomes
Yn /(I
x - Yn') ,
a

Next, the definition for a can be written for stage 72-2 and multiplied into this
pression; the product reduces to:

= a3

Repetition of this procedure « times gives:

y/n - YJ
(3.20) JL 1 = a« ,
X/(l - X,)
where X1 and Y^ are the concentrations at opposite ends of the cascade. Equation 3.20
is known as the Fenske equation by some and as the Underwood equation by others.
It determines the minimum number of stages needed to make a given separation when
the system is ideal, and can be solved by taking the logarithm of both sides.
Analytical solutions for the case of finite reflux ratios have been worked out for
ideal systems by Smoker, but this case is more involved, and since it is restricted to
ideal systems it will not be discussed here.
There are, of course, equations that give the minimum reflux directly for ideal
systems, but these are derived so simply from the general equations that they are not
worth space here.
Optimum Reflux Ratio. - The achievement of the separation of a given feed into a
product and a waste of specified concentrations is possible with any one of an infinite
number of combinations of reflux ratios and numbers of stages. In order to settle on a
particular combination, it is necessary to consider some new aspect of the operation,
such as economics. The rate of withdrawal of product being fixed, the liquid and vapor
rates in the tower must increase as the reflux ratio goes up; thus the diameter of th e

tower must be increased. The costs of the boiler and condenser go up with increasing
reflux ratio, as do the steam costs and the condenser water costs. The number of
stages goes down with increasing reflux ratio. There is, therefore, an optimum corre
sponding to minimum costs. In special applications, such as radiochemical processing
plants, such factors as a maximum diameter due to criticality considerations or the high
costs of shielding a tall tower might be influential in determining the optimum.
Example Problem. - A process uses a mixture of two solvents, A and B, in a con
centration of 90 mole % of A. During the process the mixture is diluted to 20 mole %

'e. H. Smoker, Trans. Am. Inst. Chem. Engrs. 34, 165-172 (1938).

41
of A. It was decided to recover 75% of the A from this stream and to bring it back to
its original concentration by distillation. It is fortunate that this system is ideal with
a relative volatility of 2.10, which is high enough to make distillation attractive. It is
desired to know the number of theoretical stages as a function of reflux ratio, so that
the tower design can be optimized.
Solution. — Before any distillation calculations are possible, the waste concen
tration must be determined by a material balance. This is shown in Table 3.1.

Table 3.1. Material Balance over Tower

Moles per 100 Moles of Feed

X, Mole Fraction of A
Total A B

Feed 100 20 80 0.20

Product 16.66 15.00 1.66 0.900

Waste 83.33 5.00 78.33 0.060

The minimum number of stages can be calculated from the total reflux conditions
by means of the Fenske equation,
X /(l - X )
(3.20) a" =
X /(l - X ) '
w w

which gives

0.9/0.1
In
0.06/0.94
n = 6.68 .
In 2.10

The minimum reflux ratio can be calculated from the expression

xP - yf
(3-17) Rmln = -Yf - Xf .

where Y, is the composition of the vapor in equilibrium with the feed. Using the equi
librium relation

aX
(3.5) Y =
1 + (a _ l)x '
and putting in Xf =0.20, we get Y, =0.3442. Using these values with X =0.90, we
find that R min
, = 3.85.

42
 The procedure from here is to select several reflux ratios greater than the minimum
and to determine graphically the number of stages required to make the specified sepa
ration when each reflux ratio is used. It is first necessary to make an operating diagram
from the equilibrium expression. Calculated equilibrium points are given in Table 3.2.
The second step is to plot the operating lines on the diagram. The upper operating line
has the slope LjV = R/(R + 1), and since it must pass through the point (0.9, 0.9), it
can be plotted. The lower operating line is obtained simply by drawing a line from the
point where the upper operating line crosses X = X, = 0.20 to the point (0.060, 0.060).
The third step is to step off the stages graphically and count them. The operating
diagram for the case of R = 2P . min
is shown in Fig. 3.11. **
The results are qiven in
°

Table 3.3 and the plot of n vs R is shown in Fig. 3.12.

Table 3.2. Equilibrium Data for Example Problem

a » 2.1

X ax (a - 1) x Y

0.05 0.105 0.055 0.09952

0.10 0.21 0.11 0.1892

0.2 0.42 0.22 0.3442

0.3 0.63 0.33 0.4737

0.4 0.84 0.44 0.58333

0.5 1.05 0.55 0.6774

0.6 1.26 0.66 0.7590

0.7 1.47 0.77 0.8305

0.8 1.68 0.88 0.8936

0.9 1.89 0.99 0.9497

0.95 1.995 1.045 0.9755

Table 3.3. Work Sheet for Example Problem

R . = 3.85
mtn

Stages
R/R min
.. R R/(R + 1)
Up Down Total

1 3.85 0.794

1.5 5.78 0.853 7.0 4.6 11.6

2.0 7.71 0.885 6.2 3.5 9.7

9.10 0.901 5.9 3.1 9.0

3.0 11.56 0.920 5.6 2.8 8.4

4.0 15.42 0.939 5.4 2.4 7.8

oo OO 1.00 6.7

43
 UNCLASSIFIED
ORNL-LR-DWG. 26009

70

fiBlfl

10

10 20 30 40 SO 60 70 90 100
X

Fig. 3.11. Operating Diagram for Example Problem.

44
 UNCLASSIFIED
ORNL-LR-DWG 22777
20 I I I I I I I I I I I I I I I I

I
15

Q —

gc

O
LU
DC 10
CD
LU
<£>
< "•—'••

I I I
0
0 10 15 20
REFLUX RATIO
Fig. 3.12. Number of Stages Required vs Reflux Ratio for Example Problen

45
 PROBLEMS

Problem 3.1. —In a certain high-temperature water system, iodine in extremely small
concentrations exhibits a Henry's law constant of 1.6 x 10 psi per mole fraction. If
water and iodine were the only components present, what would be the relative volatility
in a vapor-liquid system at 2000 psi total pressure?

Problem 3.2. - If the relative volatility in Problem 3.1 were 10 (it isn't), and if the
system were a reactor system from which it was desirable to remove iodine, distillation
might be used. If a five-stage cascade (shown below) is employed, with a feed rate of

UNCLASSIFIED
ORNL-LR-DWG22793

FEED
50 moles/min
PRODUCT
5 moles/min
(TO FURTHER
PROCESSING)

4 STAGES

WASTE
BOILER - 45 moles/min
(TO REACTOR)

50 moles/min and a vapor removal rate of 5 moles/min; if the feed is a liquid at its
boiling point; if steady state has been established; and if the concentration of iodine
is very low:

(a) What fraction of the iodine is removed?

(b) What is the ratio of Yp to X^?

46
 LIST OF SYMBOLS FOR DISTILLATION

F Feed rate (moles or mass per unit time)

H Henry's law constant (pressure per concentration unit)
HETP Height equivalent to a theoretical plate
HETS Height equivalent to a theoretical stage
L Liquid-phase flow rate (moles or mass per unit time)
m Stage number (below the feed point)
n Stage number (above the feed point)
P Vapor pressure of pure component
p Partial pressure of a component in a mixture
P Product withdrawal rate (moles or mass per unit time)
R Reflux ratio, L/P
V Vapor-phase flow rate (moles or mass per unit time)
W Waste withdrawal rate (refers to material taken off the bottom of a still) (moles
or mass per unit time)
X Mole fraction, or mole fraction of the more volatile component, in the liquid
phase
Y Mole fraction, or mole fraction of the more volatile component, in the vapor
phase
a Relative volatility or separation factor

Subscripts

A, B Components of a mixture
p, f, w Refer to product, feed, and waste streams, respectively

47
 BIBLIOGRAPHY

References for further reading in distillation

McCabe-Thiele Method

Coulson, J. M., and J. F. Richardson, Chemical Engineering, vol II, p 602—630,

McGraw-Hill, New York, 1955.
Robinson, C. S., and E. R. Gilliland, Elements of Fractional Distillation, 4th ed.,
chaps. 2 and 7, McGraw-Hill, 1950.
Badger, W. L., and W. L. McCabe, Elements of Chemical Engineering, 2d ed., p 323—
366, McGraw-Hill, New York, 1936.
Walker, W. H., W. K. Lewis, W. H. McAdams, and E. R. Gilliland, Principles of
Chemical Engineering, 3d ed., McGraw-Hill, New York, 1937.
Perry, John H. (ed.), Chemical Engineers' Handbook, 3d ed., sec 9, particularly
p 588-594, McGraw-Hill, New York, 1950.
The Enthalpy Method
Badger, W. L., and J. T. Banchero, Introduction to Chemical Engineering, McGraw-
Hill, New York, 1955.
Brown, G. G., Unit Operations, Wiley, New York, 1950.

48
 Chapter 4

GASEOUS DIFFUSION

The separation operation of gaseous diffusion is a multistage continuous counter-

current operation, not too dissimilar to the more familiar operations of distillation,
solvent extraction, and ion exchange. There is a subtle difference in that the "equi
librium relation" used does not come from a true equilibrium condition. Further, in
gaseous diffusion there is no inherent interstage material balance condition, and the
solution of cascade problems requires a selection of one of several convenient inter
stage material balance conditions, which, presumably, may be maintained in practice
by the manipulation of valves and compressors. Finally, there is a quantitative
difference, arising from the relatively small separation factors found in gaseous
diffusion, which necessitates large reflux ratios and cascades with large numbers of
stages. This quantitative difference allows the use of several simplifying assumptions
which convert the incremental analysis into a differential analysis and allow the
enrichment to be expressed as a continuous function of the number of stages (and other
variables).

Mechanism. — If a gas having two components of different molecular weights is

placed in a container with a leak of about the same size as the mean free path, the
lighter component will leak more rapidly. When considered in relation to surfaces,
ducts, or holes that are much larger than the mean distance a molecule travels between
collisions, the gas can be treated as a homogeneous, continuous fluid, exhibiting gross
properties. But when the dimensions of a hole in the container approach this mean free
path, the properties of the individual molecules become distinct. The hole can be
considered as an area experiencing collisions from the gas at a certain rate proportional
both to the number of molecules in the medium and to their velocities. This is in many
ways analogous to the absorption of neutrons in a nuclear reactor being proportional to
a cross section and a neutron flux (or track length). The kinetic theory of gases tells
us that the kinetic energies of the molecules in a gas are distributed according to the
Maxwell-Boltzmann distribution, independent of the masses of the molecules. This
means that the average velocity of the lighter molecules must be higher than the average
velocity of the heavier molecules, and, in turn, the flux of the lighter molecules must
be higher per unit concentration. Gaseous diffusion in application uses a barrier with
an extremely large number of holes so that the rate of leakage is reasonably high.

It should also be noted that, for this phenomenon to work ideally, a small pressure
drop across the barrier is not sufficient, since, on the scale of individual molecular
movement, random motion from the downstream side could cause the diffusion of some
material back to the high-pressure chamber, and, of course, this would also be selective
for the lighter material. Ideally, the downstream side would be a perfect vacuum.

49
Practically, considerable compromise is made; nevertheless, it is still necessary to
provide a substantial pressure drop across the barrier.
Certainly there is no true equilibrium existing between the high-pressure and the
low-pressure side of the barrier. There is, however, a relation which can be derived
from the diffusion rates. If the average energy of a gas molecule is E, the corre
sponding velocity is given by

/ 2E / 2E
„M-y- and vB.J—.
where A and B are the two constituents in question. Their rates of diffusion are
proportional to their concentrations (CA and Cfi) times their respective mean velocities:

rA « CAVA and rB " CBVB •

If standard distillation nomenclature is now adopted, where X and Y are moles of more
volatile (lighter) component divided by the total moles of the phase, and if the high-
pressure side of the barrier is identified with the heavy phase and the downstream side
with the light phase, then

CA
X =
CA + CB
and

CAVA
Y =
TA + rB CAVA + CBVB
By simple manipulation,
X CA Y CAVA
and
1 - X CB 1 - Y cBvB
From this form it is again a simple step to the relation

Y/(l - Y) VA fMB
(4.1) —'- =— = / . a ,
X/(l - X) vB V MA
which is recognized as the equilibrium relationship found in ideal distillation systems.
The high-pressure gas and low-pressure gas can be treated as two distinct phases;
there is an expression for the composition on one side of a barrier in terms of the
composition on the other; and techniques developed for true two-phase systems can
be applied.
The derivation of the X-Y relationship above assumed a zero downstream pressure
and a perfect barrier. It is surprising how well the ideal separation factor a is
approached in practice. This is explained in part by the more or less countercurrent

50
flow pattern used in the stages (or converters as they are called). For accurate work,
experimental values for a should be used.

The Cascade. — Diagramatically, the gaseous diffusion cascade looks very much
like a distillation cascade, except that in distillation the energy added in the form of
heat at the bottom of the cascade progresses through the cascade to be removed at the
top, while in the gaseous diffusion cascade it is necessary to supply mechanical energy
to each individual stage to make up for the energy lost in diffusion through the barrier.
A diagram of a section of a gaseous diffusion cascade is shown in Fig. 4.1.

UNCLASSIFIED

ORNL-LR-DWG22778

XP

Yn
COMPRESSOR

Vn^ = L„+P

Fig. 4.1. Diagram of a Portion of a Gaseous Diffusion Cascade.

A compressor feeds the gas to a stage, where it is passed along the barrier. The
pressure difference across the barrier causes some of the gas to diffuse, leaving a
slightly depleted stream to be drawn off and sent to the compressor for the next stage.
The control of the cascade is effected by valves in these drawoff lines. Since the only
pressure loss of this stream is the drop through the valve, less compression is required

51
to restore its inlet pressure than is required for gas passed through the barrier. Hence
the compressor has two inlet ports to accommodate these different streams. Gas that
has passed through the barrier of a stage is slightly enriched and moves up the cascade
to the next higher stage, and gas that has not passed the barrier is slightly depleted
and moves down the cascade to the next lower stage, the streams entering the ap
propriate ports of the compressors and being mixed therein. The compressed gases
are, of course, cooled to maintain a desirable temperature.
A complete description of a gaseous diffusion plant is beyond the scope of this
work, which is concerned more with the analysis than with the hardware. It is perhaps
pertinent to mention that the only significant application of gaseous diffusion so far
is for the separation of U 3S from U 38 by use of the volatile hexafluoride. The
separation factor in this case is 1.0043, which, although small, offers the best mode
of separation for these isotopes. Literally thousands of stages are used in these
cascades, and the stages vary in size by orders of magnitude. By early 1957 the
United States had about 2.3 billion dollars invested in these plants, which spread over
the landscape of Oak Ridge, Tennessee, Paducah, Kentucky, and Portsmouth, Ohio.
The Material Balance. —Although all the material is present in the gas phase, there
are two distinct regions separated by a barrier, and it is convenient to think of the
high-pressure gas as a heavy phase and the low-pressure gas as a light phase. The
heavy-phase rate leaving any stage n will then be L moles of gas per unit time, and
the light-phase rate will be Vn moles of gas per unit time. Out of the top of the
cascade, product will be removed at a rate P, feed will be introduced at some inter
mediate point at a rate F, and at the bottom of the cascade, waste will be withdrawn
at a rate W. The feed point divides the cascade into two parts which are in all ways
equivalent. Equations will be derived for the top of the cascade, but by the obvious
change in basis could apply to the bottom.
A material balance taken around the top of the cascade, including all stages above
stage n, gives for the total gas flow

(4.2)
v '
V n = L n +.,1 + P ',

and for the more "volatile" component

(4.3)'
^
Vn Yn = L fj +..X .. + PX„p .
1 rc + 1

If V X is subtracted from both sides of Eq. 4.3, it becomes

n n ^ '

(4.4)
x '
Vnv(Y n - Xn ) = L n +..(X +, - Xn') + P(X„
1* rc+ l x p
- Xn ) .
At this point, assumptions important to simple analysis of cascades with low separation
factors are employed.
The first assumption is

dX
(4.5) X„ +1 - X„ =
dn

52
This depends on there being a large number of stages and only a small change in
concentration between adjacent stages. It asserts that the previous step-by-step
calculation can be replaced by one involving continuous functions. This approximati on

is valid to a high degree of accuracy in these systems. The application of this a s-

sumption to Eq. 4.4 replaces dependence on Xfi +1 with dependence on n. Now there
are still Ln +] and Yn to be reduced to dependence on either n or X before the equation
can be used.

The quantity Yn can, of course, be eliminated by use of the equilibrium relationship,

Eq. 4.1, which will be rewritten in the form

(4.1«) y = — .
1 + (a - 1)X
If X is subtracted from both sides,

„ aX aX - X - (a - 1)X2
Y - X = X =
1 + (a - 1)X 1 + (a - 1)X
or

(4.6) y - x = (a ~ 1)X0 - X)
1 + (a - 1)X '
The second simplifying assumption is that the term (a- 1)X is small compared with
unity and can be neglected. For the separation of U235 from U238, the maximum value
of this term (when X is unity) is 0.0043, which makes this approximation reasonable.
Then

(4.7) Y - X = (a - 1)X(1 - X) .
Putting Eqs. 4.2, 4.5, and 4.7 into the material balance equation 4.4 gives

(4.8) dx = L"+1 + P (a - 1)X(1 - X)

— P (X„ - X) .

For the bottom section of the cascade, a similar derivation gives

,„ o x
(4.8a)
dX
— =
Lm +1 " W,(a - 1)X(1 - X)
V (X - X ) .
dn L to +.,1 L m +A11 w

The only remaining quantity which must be reduced to dependence on X or n is

Ln +1. Herein lies one of the distinguishing features of gaseous diffusion. There is
no intrinsic relation to be applied, and Eq. 4.8 is as far as general analysis can be
made.

Interstage Material Balance Condition. - Consider again the material balance,

<4'2> Vn = L« +l + P .

53
which says that the flow rate down the cascade is less than the flow rate up the
cascade by an amount equal to the product rate. The relation for the more volatile
component (Eq. 4.3) says no more. This gives no relation between equivalent rates
for adjacent stages. As far as these equations are concerned, Lfj +1 might approach
infinity while L +_ was zero. This could not be so in distillation unless, in effect,
a boiler or condenser were put into the middle of the cascade, nor in solvent extraction
unless solvent were added or removed at points along the cascade. For these operations
there are inherent interstage material balance conditions which provide the necessary
relations. The necessary relations for gaseous diffusion must come from artificially
imposed conditions which optimize operation, and these conditions are imposed in
practice by the speeds of compressors or the settings of valves. In short, the relation
of L to X is at the discretion of the operator.

There are three cases of interest which will be treated here. There is the case of

total reflux, where P goes to zero and Eq. 4.8 is independent of L; there is the case
of constant molal overflow, where L is maintained constant and independent of X over
some range of concentrations; and finally there is the case of the ideal cascade, where
the total interstage flow is minimized.

Total Reflux. —When P goes to zero, Eq. 4.8 becomes

dX
(4.9) — - (a - 1)X(1 - X) ,
dn

which can be simply solved for the minimum number of stages needed to go between a
waste concentration X and a product concentration X^:
w r p

4.10 n mI
rxp dx =
i In —
V1 - xv) .
Jx (a - 1)X(1 - X)
w
a - 1 Xjl - X) *

A somewhat different form of this equation is obtained by expansion of the logarithm

of a and neglect of higher order terms:

(a - l)2 (a - l)3
In a - ln[l + (a - 1)] = (a - 1) - + -

In a = a - 1

With this approximation, Eq. 4.10 becomes

X(] - X )
4.11 a" = —
Xw*(1 - X)
p'

54
 Equation 4.11 is recognized as the Fenske equation from distillation. As in the
case of distillation, it is possible to derive this equation exactly from the equilibrium
relationship (Eq. 4.1) and the total-reflux material balance, L .. = V . The fact that
it was possible to arrive at it through Eq. 4.9 means that the assumptions used to this
point are as good as the approximation of In a by a —1. When the separation factor is
1.0043, the error in the number of stages is about two parts per thousand.
The significance of Eq. 4.11 is that it gives the least possible number of stages
to perform a separation. Since there is no product or waste removed, there is no feed
entering, and it is independent of the feed concentration. Obviously any real cascade
would require the removal of product and waste and would require a larger number of
stages.

Constant Molal Overflow. — The removal of product re-establishes the necessity

of a functional relationship between L and X. The simplest such relationship might
be that L is constant over some range of X. This is the very practical case of constant
molal overflow, or the square-cascade condition. In practice the same value of L need
not be kept over the entire cascade, but the cascade may be broken into any number
of segments each having a different, but constant, value of L. Each segment can be
treated separately, and the total number of stages required is the sum of the require
ments for all segments. Discontinuities in L will occur between all adjacent segments
and at the feed point.

Consider the problem of determining the number of stages in a segment, in the

enriching section of the cascade, which covers the range Xfl to X,, having an interstage
flow L. The cascade produces product at a rate P and of a concentration X . With
the conditions outlined here, Eq. 4.8 is applicable:

dX +1 "*" ' P
(4.8) — = -^- (a - 1)X(1 - X) - (X - X) .
dn L„+l L«H
Equation 4.8 can, of course, be integrated in its present form; however, some simplifying
modifications can be made. First, define

L«+l
(4.12) R = —— = reflux ratio
P

and

L +i
(4.13)
(Ln +] + P)(a - 1) (R + l)(a - 1)

Then the right-hand side of Eq. 4.8 takes on the form of a quadratic in X, the roots of
which can be determined:

55
 dX 1 1\ XP
(4.14) _ + _ X + —
~dn~ A R R

The roots are:

(4.15)

A\ A
X„ = 1/2 ( 1 + — 1/2 1 +— - 4X„ —
R P R

Equations 4.15 can be used to rewrite Eq. 4.8 in the form

dX 1
(4.16)
- -r
dn
' TA <x ~ x^x ~ x2>
The value of X. may or may not be negative, and the value of X_ may or may not be
greater than 1. The order of relative values which must be preserved is
X, < X < X < X, < X, .
\ a b 2

If this order does not hold, then the parameters were poorly chosen; the operating line
probably crosses the equilibrium line, and the problem is absurd.

Since this calculation is restricted to the enriching section of the cascade, it should
also be true that

XflXa and X„t Xp •

The required number of stages can now be obtained from integration of Eq. 4.16:
X=Xj
-dX
(4.17) n = AC
x=x (X - x,)(x - x2)

px=xt dX
x=xt
J.x=>
dX

x2 - X, Jx=x_ (x - x.) x=x. (X, - X)

from which

A lXb ~ X,X*2 - XJ
(4.18) In
x2 - X, (xa - x,)(x2 - xb)

Equation 4.18 is the workLng equation for the determination of the number of stages
required in order to enrich material from X to X in a square cascade when R, a, and
X are given to fix the constants A, X^, and Xj.

56
 Ideal Cascade. — A little reflection on constant-mo Ial-overflow reflux will reveal
that this is not the most efficient way to operate a cascade. When the range X to X,
is large, it is particularly noticeable that there is more reflux than is desirable over
part of the range, while there may be less than is desirable over another part. When
the range is decreased the effect is minimized, but quantitatively rather than quali
tatively. Since there is nothing inherent in the cascade which restricts it to constant
values of L, the question is raised: What is the functional relation between L and X
that would give the smallest total interstage flow?

An intuitive approach is to note the principal cause of nonideal flow and eliminate
it. Each stage receives two streams, which are mixed in the compressor. If these
streams have different concentrations, separative work is wasted. Therefore, the
"ideal cascade" is the one for which the concentrations of these two streams are equal:

(4.19) V, - V , •
Equation 4.19 will, then, be used as the defining relation for the ideal cascade. In
order to use this, recall the assumption of a large number of stages:

dX
(4.5) Xn+, - X
n
dn

Equation 4.5 may also be written:

dX
Xn+2 - Xn ++y1 ~ ~dn~

The sum of these equations gives:

2dX
X ., - X =
"+2 * dn

But Eq. 4.19 gives X ., = Y , so

2dX
= y - x .
dn n "

It is now possible to employ the equilibrium relationship 4.7 to give

dX 1
(4.20) — = -(a - 1)X(1 - X) .
dn 2

Except for the factor of 1/2, Eq. 4.20 is exactly the same as the total-reflux equation
4.9, and it can be integrated in the same way either from X. to X or from X^ to X :

9z XJl
p*
- X w')
(4.21) n= In -!-— = 2« . .
a - X - XJ min

57
Equation 4.21 is the working equation for the determination of the number of stages
required in order to enrich material from any concentration Xfl to X, in an ideal cascade.

In order to find the functional relationship of L to X from this, subtract Eq. 4.20
from the general equation 4.8, obtaining:

0 = _(a - 1)X(1 - X) -- (X - X) +- (a - lTx(l - X) ,

d. Li Li

XP-X
(4.22) L, = 2P 1
(a - 1)X(1 - X)

Since a - 1 is very much less than 1, the term -1 is neglected, and Eq. 4.22 is
written:

2P(X - X)
(4.23)
(a - 1)X(1 -X)

The equivalent expression for the bottom section of the cascade is:
2W(X - Xj
(4.23a) L =
(a - 1)X(1 - X)

Instead of the intuitive approach based on the requirement that X +. = Y ., a

more fundamental approach is based on the total interstage flow. Work is being put
into the cascade at every stage, so the total interstage flow is proportional to the total
work used for the separation. The separative work, S, is proportional to the actual work
and is defined by the relation

x=x,
(4.24) L dn
w .x=x

Both L and n are functions of X, dn being given by the general equation 4.8 and L being
an unknown function. The problem becomes the determination of L as that function of
X which will minimize S. Use of Eq. 4.8 produces:

L2 dX
(4.25)
(L + P)(a - 1)X(1 - X) - P(X„ - X)

If the integral is to be a minimum over an arbitrary interval, then the derivative of the
integrand with respect to L must be zero. This gives:

0 = LJ2(L + P)(a - 1)X(1 - X) - 2P(X - X) - L(a - 1)X(1 - X)] .

58
Neglecting the trivial solution L = 0 and solving for L, we get

Xp -X
L = 2P - 1
(a - 1)X(1 - X)

which is identical with Eq. 4.22.

Separative Work. —It is now valuable to examine more closely the work associated
with the enrichment of isotopic mixtures and to evaluate the integral 4.24 for the ideal
cascd"de. Recall that this is an index of the minimum work necessary to effect the
enrichment:

(4.24) S = f b L dn
Jx

The evaluation of this quantity for an arbitrary interval in the enriching section when
the product concentration and rate are known is possible and not overly difficult.
However, for the sake of conserving space, the general solution will be bypassed, and
only the application to a whole cascade will be presented. That is, a feed of concen
tration X. will be separated into a product X and a waste X in an ideal cascade.
The separative work for this separation is

X
A . A

S = / ' L, dn + / * L2dn ,
X

where, from Eq. 4.20,

dn 2 dX
(4.26) dn = dX =
dX (a - 1)X(1 - X)

In order to put this expression in a more desirable form, an arbitrary concentration,

a, is chosen for the lower limit of integration, which, as will be shown, is eliminated
from the final expression:

(4.27) S = fr Xi* Lydn - frX wL} dn - J,Xf' L2 dn + frXt>P L2dn

X X X
f PL2dn + f f (L, - L2) dn - / wL, dn .
a a a

On substitution for L2 and L by means of Eqs. 4.23 and 4.23a,

59
 x (x - x) dx xp (Xp - X) dX
(4.28) S
(a - l)2 '/. X2(l - X)'
+ p
Xa X2(l - X)2

x, W(X - XJ - P(X - X)
I. x2(i - xy
JX

But

W(X - Xj - P(X - X) = (P + W)X - PX - WX^ = FX - FX = -F(X. - X) ,

so that Eq. 4.28 becomes

S(a -- I)2 xw (xw - X)dX xp (X - X)dX

ir r
(4.29) = W I + p '
4 Ja X2(l - X)2 •/ X^(l - X)^

x, (X - X) a-X
- F I -L
J a Y2/
xz(] - xy

In evaluating this expression, it will be valuable to note that W + P — F = 0 and

WXw + PX —FX, = 0. From these facts, it follows that any term which appears in all
three of these integrals and is independent of the upper limit, or out of which the upper
limit can be removed as a factor, will cancel itself from the expression. The solution
of the integrals is best effected by partial fractions:

x0 - X ex DX
+ +

X2(l - X)2 X2 (1 - X)2 X2 (1 - X)'

x„ 3XQ - 2 1
+ (2XQ - 1)
X
X' (i - xy (1 - X)2

x0 (x0 - X) dX 1 1 1
+ (3XQ - 2)
Xa X^(l - xy
= x„
a X„ 1 - XQ 1 - a

x„
+ (2XQ - 1) In X - In a + ln(1 - a) - ln(l - X ) +
1 - X„

On elimination of terms which are either constant and independent of XQ or constant

coefficients times XQ, this reduces to
X„
(4.30) (2Xn - 1) In v(V
0 (1 - XJ

60
Equation 4.30 defines the "value function" of X.. Equation 4.29 then reduces to:

(4.31) i^-I =WV(XJ +PV(Xp) - FV(Xf) .

Equation 4.31, then, is the working equation for determining the relative difficulty of
separating a feed of concentration X. into a product of concentration X and a waste
of concentration X by the process which optimizes interstage flow. It can be converted
to an absolute basis by factors depending on the costs and on the efficiency of pumps
and stages. It is very useful in determining the optimum economic recovery of the
more desirable component when the cost of the raw material is known.
This expression is also useful for comparing the work required in the ideal cascade
with that in a practical cascade.
The Practical Cascade. — While it appears quite desirable to minimize the total
flow in a cascade, the construction and operation of a cascade in which the flow is
unique for every stage is an extremely difficult engineering task. The operation of a
single square cascade from feed to product is, on the other hand, an extremely in
efficient operation. Fortunately it is not necessary to choose between these extreme
alternatives, for it is possible to use a number of square cascades of different sizes,
large at the feed point and small at the product end, so that the ideal cascade is ap
proximated to some degree. Such a cascade is known as a "practical cascade." The
number of stages required in this cascade is the sum of the stages required in the
individual segments of it, and the total separative work is given by

(4.32) S = £-,«, + L2«2 + Lj/23 + L4«4 + . . . ,

where the subscripts indicate the sections.

General Application of Gaseous Diffusion Cascade Concepts. — In reviewing the
development of the equations presented here, it is found that once the separation factor
was established no further reference was made to the mechanism of gaseous diffusion.
The inherent requirements were: that the separation factor be close to unity, that there
be a large number of stages, and either that the interstage material balance condition
be given or that it be at the manipulation of the operator. The assumptions employed
included the replacement of an inherently stepwise enrichment with a continuous
function, the discarding of terms of the order of (a - 1) to reduce the equilibrium
expression to a simple quadratic, and, in the development of the ideal cascade, one
or two liberties with second-order effects of a less significant nature than those above.
In addition to gaseous diffusion, there are many separation operations for which
the necessary requirements are met and for which the employed assumptions are good.
These include the separation of heavy water by distillation, the separation of heavy

61
water by chemical exchange, the separation of nitrogen isotopes by chemical exchange,
and, in short, almost any difficult separation.

Example Problem. — Among the problems connected with aqueous homogeneous

breeder reactors is that of obtaining sufficiently high concentrations of thorium in the
blanket. One of the few soluble salts of thorium is the nitrate, which has among its
disadvantages the high capture cross section of natural nitrogen. There is, therefore,
a potential use for N15. A TBR-type reactor would have an inventory of about 3600 kg
of N1S.
The Escambia chemical plant in Florida produces 5 x 10 g moles of ammonia per
day. Consider a facility which will extract from this entire stream 1000 g moles of
99% N H3 per day with a process having a separation factor per stage of 1,029. (This
would be the separation factor for gaseous diffusion were it not for the presence of
deuterium.) The abundance of N in natural nitrogen is 0.37%.

(a) What is the composition of the waste?

(b) How many stages would be required in an ideal cascade (1) in the stripping
section, (2) in the enriching section?
(c) Plot interstage flow rate (of ammonia) vs stage number for the enriching section.
(d) What is the total interstage flow for the entire cascade?
(e) How many stages would it take to go from 1% to 2% of N (1) in a square
section with L = 2.5 x 10 g moles/hr, (2) in a square section with L = 1.6 x 10
g moles/hr, (3) in an ideal cascade?
Solution. —

(a) The over-all material-balance condition is:

WX = FX. - PXA ,
w I p

from which

x„ px„
F - P-l 1 __£
Xw Xf FXf
X, F - P P
/ 1 __
F

PX
P' = (1 + 2 x 10~4)(1 - 5.35 x 10-2)

It is given that

Xf = 0.0037

62
Therefore

Xw = 0.0035028 .

(b) (1) According to Eq. 4.21, the number of stages in the strip section of an ideal
cascade is

xn - X )
In
a - 1 X (1 - X,) '

whence

X / X - X
In —L + In 1 + —
0.029 X \ 1 - X
/

2 1
In + In (1 + 0.000198)
0.029 1 - 0.0533

2
(0.0546 + 0.000198) = 3.78 stages .
0.029

(Note that the assumptions are very weak for this few stages.)
(2) The number of stages in the enriching section is

yi -X/)
In
a- 1 X/l - Xp) '
whence

2 0.99 x 0.9963
n = In = 704 stages .
0.029 0.0037 x 0.01 —

The total number of stages required is 708.

(c) In order to plot the interstage flow L as a function of stage number n, it is
probably best to calculate n for several values of X, then calculate L for the same
values of X, and thereby get the desired relation.
For values of X close to the feed concentration it is convenient to use the very
accurate approximation

2 X(l - Xf) X - X. x
(4.33) n = In + In
1 - a X (1 - X) 1 - a 1 - X X,

and for values of X significantly removed from the feed concentration it is convenient
to use the form

(1 - Xf
(4.34) In + In
1 - a 1 - X
Xl

63
Throughout the entire range it is possible to use the relation

2P{Xp - X)
(a - 1)X(1 - X)

The results are shown in Table 4.1, with two columns for n corresponding to the two
forms of the formula (Eqs. 4.33 and 4.34). These are also plotted in Fig. 4.2.

(d) The total interstage flow for the ideal cascade is obtained from Eq. 4.31:

/ L dn = S =
(a - I)2
[WV(X ) + PV(X) - FV(X,)]

Table 4.1. Stages, Composition, and Interstage Flow Rates within Isotope Enrichment Plant

Stage Number, n
je Interstage Flow Rate, L
Mole Fraction of N , X Calculated Calculated
(g moles of NH, per day)
by Eq. 4.33 by Eq. 4.34

0.0037 0 1851

0.004 5.39 1713

0.005 20.8 1370

0.007 44.1 977

0.009 61.7 761

0.010 69.0 69.1 699

0.012 81.7 571

0.020 117.6 117.4 342.3

0.050 182.8 183.1 137.0

0.100 243.5 234.4 68.4

0.200 290.3 34.17

0.300 327.6 22.72

0.500 386.0 13.56

0.700 444.4 5.49

0.800 481.7 8.22

0.900 537.6 6.92

0.950 588.9 5.83

0.99 702.9 0.10

64
Because of the proximity of X. and Xw, modification of the form of this relation is in
order:

X. X
FV(X ) - WV(XJ = F(2X - 1) In —L_ - W(2X„ - 1) In
I — X / I — X
"
/ tu

x.
- [P(2X - 1) +W(2X - l)]|n —J- W(2X - 1)ln
1 - X 1 - X
/

xf xp - XJ
= P(2X. - 1) In + W(2X - 1) In —
p 1 - X, w X (1 - X,)
I "/'

UNCLASSIFIED
ORNL-LR-DWG 22779

NUMBER OF STAGES FROM FEED POINT

10 20 30 40 50 60 70
2000 1.00

1800 0.90

1600 ^v~" 0.80

1400 0.70

o
T3
1200 0.60 -
0)
O O
E
z
o
~r 1000 0.50 I-
o
o <
_i or
u.
LlI
o 800 0.40
r?
CO
CC

600 0.30

* 1 / m

400 \\ ^-11 0.20

200 0.10

0
0 100 200 300 400 500 600 700
NUMBER OF STAGES FROM FEED POINT

Fig. 4.2. Interstage Flow in Enriching Section of Ammonia Gaseous Diffusion Cascade.

65
But

1 - X X. - X
w f w
= 1 +
1 - X 1 - X
/ /

So, approximately,

4
Xp0 - X) / X - Xw X
S = P(2X„ - 1) In — - W(2Xtll - 1) — + In —
(a - 1)' P XP ~ Xp) •" \ 1 ~ Xf Xw,
99 x 99.63
—y
0.029/
1000 x 0.98 In
0.37 x 1

,'0.000197 1
- 5 x 106(-0.993) ( + In
0.9963 1 - 0.0533

= 1.38 x 109 g moles/day .

Note that this gives the flow in one direction only, and that the total movement would
be twice this.
(e) (1) The reflux ratio in a square section with L = 2.5 x 105 g moles/hr is

L 24 x 2.5 x 10s
R = T= iooo = 6000
From Eq. 4.13,

1 1
= 0.00575 .
R (R + l)(a - 1) 6001 x 0.029

Then, from Eq. 4.15,

X. =- x 1.00575 J 1.005752 - 4 x 0.00575 x 0.99

l 2 2

1 1 / ;
= — x 1.00575 - - yf 1 + 2 x 0.00575 + 0.005752 - 4 x 0.00575 x 0.99

=—x 1.00575 v/l - 2 x 0.00564 + 0.005752

2 2

1 1
=- x 1.00575 (1 - 0.00564) = 0.005695 .
2 2

Similarly,

X, =— x 1.00575 + — (1 - 0.00564) = 1.000055 .

2 2 2

66
On substitution into Eq. 4.18,

A <X* - Xl)<X2 - XJ
In
x2 - X, (Xfl - x,)(x2 - xb)
(0.02 - 0.005695)(1.000055 - 0.01)
= 34.7 In
(0.01 - 0.005695)(1.000055 - 0.02)

0.0143 / 0.01
= 34.7 In 1 + = 42.0 stages
0.004305 \ 1.00,

(2) The same calculation as above, with L = 1.6 x 10 g moles/hr, gives

n = 51.6 stages .

(3) The number of stages of an ideal cascade which are needed to do the same job
can be obtained directly from the values calculated in (c) and listed in Table 4.1:
n = 48.5 stages .

PROBLEM

Problem 4.1. —A neon gas purge is being considered for the detection of leakage
of fission-product gas in a certain nuclear experiment. The isotope Ne is deemed
undesirable because it is activated by neutron capture to radioactive Ne . It is
necessary, then, to consider the practicality of removing Ne from natural Ne. The
suggested method is gaseous diffusion.

The anticipated requirement is 10 liters (STP) per day containing less than 1%
of Ne22.

The isotopic content of natural neon can be assumed to be Ne , 91%; Ne , nil; and
Ne22, 9%.
(a) What will be the theoretical separation factor?
(b) If the waste-stream concentration is set at 30% of Ne , what will be the feed
rate and the waste-stream rate? What is the recovery of Ne ?
(c) How many stages would be required to make this separation at total reflux?
(d) How many stages would be required if an ideal cascade were used?
(e) If a square cascade were used, what are the minimum reflux ratios at the product
and waste ends?

(/) Approximately how many stages would be required in the enriching section of a
square cascade if twice the minimum reflux ratio were used?
(g) What would be the minimum separative work if simple square cascades were
used for both the enriching and rectifying sections, and how would this compare with
the minimum separative work required for an ideal cascade?

67
 LIST OF SYMBOLS FOR GASEOUS DIFFUSION

A, B, C, D Constants
A Parameter defined by Eq. 4.13
a Arbitrary concentration
C Concentration (parts per unit volume)
E Molecular translational energy
F Feed rate (moles or mass per unit time)
L Interstage high-pressure gas flow (moles or mass per unit time)
M Molecular mass

m Stage number (below the feed point)

n Stage number (above the feed point)
P Product removal rate (moles or mass per unit time)
R Reflux ratio, L/P (function of stage number)
r Diffusion rate

S Total interstage (unidirectional) flow for a section of a cascade; separative

work

V Interstage low-pressure gas flow (moles or mass per unit time); value
function defined by Eq. 4.30
v Mean molecular gas velocity
W Waste removal rate (moles or mass per unit time)
X Mole fraction of lighter isotope in the high-pressure "phase"
Y Mole fraction of lighter isotope in the low-pressure "phase"
a Separation factor

Subscripts
A, B Isotopic components of system
/, p, w Refer to feed, product, and waste streams, respectively
1, 2 Roots of quadratic expression; sections of a cascade
a, b End conditions of a square cascade

BIBLIOGRAPHY

References for further reading in gaseous diffusion

Glasstone, S., Principles ofNuclear Reactor Engineering, sees 8.60-8.90, Van Nostrand,
New York, 1955.
Shacter, J., and G. A. Garrett, Analogies Between Gaseous Diffusion and Fractional
Distillation, AECD-1940 (April 7, 1948).
For an alternative method, see:
Kaplan, I., Nuclear Physics, chap. 22, Addison-Wesley Pub. Co., Cambridge, Mass.,
1955.

Cohen, K., The Theory of Isotope Separation as Applied to the Large-Scale Production
of U23', NNES, vol Ill-IB, McGraw-Hill, New York, 1951.

68
 Chapter 5

SOLVENT EXTRACTION

The unit operation of solvent extraction constitutes the heart of most of the
processes currently being used for the processing of reactor fuels and is also the
principal operation in the purification of uranium and zirconium for reactor use. It is
a countercurrent operation which can be described in terms of a number of theoretical
stages and can be easily analyzed by the alternate use of equilibrium and material
balance conditions. As in the case of distillation, solvent extraction problems are most
easily analyzed graphically, the method being very similar to the McCabe-Thiele method
used in distillation. Another resemblance to distillation is that the most common form

of problem is the determination of the number of stages needed to produce specified

end conditions when flow rates, etc., have been fixed. When the number of stages is
given, the problem often must be inverted, answers being assumed and the determination
of the number of stages being used to effect a solution by successive approximation.

In solvent extraction, it is desirable to study the operation by considering first a

simple one-section model and working from this to the more involved two-section and
two-solute systems. This is a-little different from the treatment of distillation, where
it was expedient to start with the two-section cascade.

The first model to be studied is a simple countercurrent multistage operation with

immiscible solvents and a single solute. The operation is depicted diagramatically in
Fig. 5.1. The feed material enters the "top" of the cascade; it consists of the solute
carried in solution in the heavy liquid phase. The product leaves the stage which
receives the feed; it consists of a solution of the solute in the light liquid phase. The
waste stream leaves the "bottom" of the cascade; it is the heavy liquid phase from
which the solute has been removed. The fresh solvent, or extractant, enters the stage
from which the waste leaves; it consists of light liquid phase with very little or no
solute. In the model, all operations are assumed to be perfect; that is, the two dispersed
phases in the stream leaving the mixer of each stage are assumed to be at equilibrium,
and the two separated streams leaving the settler are assumed to be completely free
from entrainment. In other words, the model is assumed to consist of ideal, or
"theoretical," stages.

Since the twc solvents are immiscible, and since there is no perturbation by en
trainment, their flow rates from stage to stage are constant. The heavy phase, which
will be identified as aqueous, will have a mass flow rate of A mass units per time unit,
and the light phase, which will be identified as organic, will have a mass rate of Q

69
o

UNCLASSIFIED
ORNL-LR-DWG 22780
SOLVENT FEED
0 A
y0

STAGE NUMBER I /7+1 n+2. FINAL

MIXERS V >2 yn "n+2 yn+2

SETTLERS

MIXERS J V
yn-\ *n + \ *n+\ "V7+3

PRODUCT
0
yP

Fig. 5.1. Schematic Drawing of a Simple Countercurrent Multistage Solvent Extraction Cascade.
mass units per time unit. For convenience, then, the concentrations of solute should
be based on these units:

x is the mass ratio of solute in the aqueous phase,

y is the mass ratio of solute in the organic phase.

The mass ratio is defined as the mass of solute per unit mass of solute-free solvent.
This is different from the mass fraction (X or Y) used in distillation, which was the
mass of solute per unit mass of total phase. The conversion of equilibrium data from
mass-fraction to mass-ratio units involves only simple algebra, since

X Y
and y =
1 - X 1 - Y

From these definitions, it is clear that a plot of equilibrium data in terms of mass
fraction is "closed," going from 0 to 1, while an equilibrium diagram based on mass
ratio is "open," going from 0 to infinity.
If concentrations are given in grams of solute per liter of aqueous phase, it is
necessary to know the density of the phase in order to compute the mass ratio or the
mass fraction.

The virtue in the use of mass ratio is that Ax and Qy represent the mass flow rates
of solute.

With the defined terms, it is possible to write a solute material balance around the
entire operation:

Axf + Qya - Axw + Qyp •

or

(5.D A{xf - xj - Q(yp - yo) .

It is also possible to divide the cascade at an arbitrary point between stages and
to write a material balance around each of the two parts:

(5.2a) A{Xf - xB +1) = Q(yp - y„) ,

and

Mx +1 - xw ) = Q(y
y n + l *- J n
- yJ o') ,

or

A
(5.2b)
x '
y
•'n
=—
n
(xn +1
+l
- x w ) + 7o
y .

This is the relation between the compositions of the two streams as they pass each
other in the operation. It relates the light-phase composition in one stage to the heavy-
phase composition in the next.

71
 There are, then, two relations which are necessary to the solution of the solvent
extraction problem. One is the equilibrium relation, y = cf>(x ), which is usually
available only in the form of plotted data; the other is the material balance relation
given above. As has already been mentioned, the problem as presented would be to
determine the number of theoretical stages required for the extraction of solute from a
feed of concentration x, in order to produce product of concentration y. and waste of
concentration xw when an organic solvent of concentration y was used at a given flow
ratio of A/Q. One might start at the bottom of the contactor, first using the equilibrium
relation and the concentration of the waste stream to determine the concentration of
the organic phase going from the first to the second stage. This concentration, y.,
is then used in the material balance condition to determine the concentration of the
aqueous phase leaving the second stage. This concentration, x», is then used with
the equilibrium condition to determine the concentration y, of the organic phase leaving
the second stage. By repetition of this procedure, all the concentrations are calculated
until the aqueous-phase concentration is found to exceed the feed concentration. The
number of stages required is simply counted, allowance being made for the fact that
the last stage is only partially utilized and that a fraction of a stage would suffice.
Usually simpler than the analytical step-by-step calculation is the graphical method
of solution. When the material balance equation relating x +. to y is plotted in the
same graph with the equilibrium data relating x to y , it is seen to give a straight line
called an operating line. This plot is called an operating diagram (see Fig. 5.2). The
operating line joins the point corresponding to the concentration conditions at the
bottom of the cascade with the point corresponding to the concentration conditions at
the top of the cascade, with a slope of A/Q. The calculation by the graphical method
proceeds in the same manner as the analytical calculation; here, however, each manipu
lation consists simply in drawing either a horizontal or a vertical line connecting the
operating and the equilibrium lines. This process is known as stepping off stages.
The calculation of y1 is made by reading from the equilibrium line from the point x .
Then, x2 is obtained from the operating line and y^; y2, from the equilibrium line and
x2; etc.
The use of an operating diagram of this type makes the understanding of the pa
rameters of the solvent extraction system much easier, provided the diagram itself is
understood. A few of the characteristics of the diagram will now be discussed.
First, it is clear that the operating line and equilibrium line cannot cross at any
point in a simple cascade, or even touch, unless the number of stages goes to infinity.
Complete removal of a solute from an aqueous phase by means of a pure organic solvent
requires that the operating and equilibrium lines touch at the origin; theoretically, this
can never be achieved. In actual practice, concentrations can be reduced to a negligible
level by the proper choice of solvent rates and number of stages. If there is some solute

72
 UNCLASSIFIED
ORNL-LR-DWG 22781

EQUILIBRIUM LINE
y„ = 4>(x„)

OPERATING LINE
'2
n-\ ~ 75" ( xn~ xw> +>o

y0

Fig. 5.2. Graphical Solution to a Simple Countercurrent Multistage Solvent Extraction Cascade.

in the "fresh" organic solvent used for the extraction, the aqueous-phase concentration
can never be reduced below the concentration in equilibrium with this organic-phase
concentration.

Whenever the slopes of the equilibrium and operating lines differ (and this is
virtually always true), the use of a very large number of stages results in pinching-in,
the condition where the two lines tend to touch, accumulating a large percentage of the
stages at the point of near contact. If the slope of the operating line is greater than
the slope of the equilibrium line, the pinching-in will occur at the feed and product end
of the cascade, and the concentration of solute in all the stages in this pinched-in
region will be essentially the same (see Fig. 5.3). The waste end of a cascade oper
ating in this manner will always discharge a finite concentration of solute, despite the
large number of stages used.

If, on the other hand, the slope of the operating line is less than the slope at any
point in the equilibrium line, the pinching-in will occur at the solute concentration of

73
the fresh organic solvent, which is usually zero. When this is the case, each stage
reduces the concentration of the solute in the aqueous phase by a significant factor,
and the use of a large number of stages can reduce the concentration in the waste to
as small a quantity as desired (see Fig. 5.4).

UNCLASSIFIED
ORNL-LR-DWG 25570
UNCLASSIFIED
ORNL-LR-DWG 22782

*f

Fig. 5.3. Pinching-in at the Feed Point. Fig. 5.4. Pinching-in at the Origin.

Analogous to the simple extraction operation described above is the simple stripping
operation. When the organic solvent entering the cascade is the primary carrier of
solute, and the aqueous stream enters containing no (or little) solute, the solute can
be transferred from the organic phase to the aqueous phase. This operation is known
as stripping. The only difference is that in stripping the operating line is above the
equilibrium line, while in extraction the operating line is below the equilibrium line.
Similarly, the operating line intersects the y axis at the concentration of the depleted
organic phase corresponding to conditions at the top of the cascade, and the upper end
of the operating line corresponds to the bottom of the cascade, where the solute-carrying
organic phase is entering and the solute-laden aqueous phase is leaving.
Before leaving the simple single-solute model, it would be well to mention several
terms which are sometimes used in the literature. Because the determination of the
entire equilibrium line is laborious, and since the line can be shifted by the addition
of various reagents, it will not always be found available in systems not completely
developed. An indication of the extractability of a solute with a given solvent can be
obtained from a single batch contact from which the concentrations of the solute in
both phases are determined. The ratio of these concentrations (organic-phase over
aqueous-phase) is called the distribution coefficient, D. If the equilibrium line is
curved, as is commonly the case, the distribution coefficient is a function of concen
tration and is of little use in calculation except as a crude approximation. A few
systems have straight equilibrium lines, in which case the distribution coefficient

74
represents the slope of the equilibrium line. Where this is the case, another useful
term, called the extraction factor, E, can be defined:

This is essentially the ratio of the slopes of the equilibrium and operating lines. Where
it is applicable, the amount by which n stages will reduce the concentration of solute
in a feed can be determined from the relation

E - 1
(5.3)
xf En+1 - 1
where x is the concentration leaving the nth stage and x, is the concentration of the
feed. Although commonly used to obtain approximate answers when a single distribution
coefficient is the only datum available, this formula can lead to serious error when
the equilibrium line is not straight.

The second model to be studied is a simple countercurrent multistage operation,

with immiscible solvents but with two independent solutes. It is the object of such an
operation to effect a separation between the solutes. The assumption that they are
independent implies that each will have an equilibrium line which is not altered by the
presence of the other. This is not usually true unless the solutes are present in very
low concentrations. In the range of concentrations used in radiochemical processing,
there is little error here.

The schematic drawing of the operation is the same as that shown in Fig. 5.1;
however, now the feed has two concentrations, x(l), and x(2),, and so, correspondingly,
does each stream. The two equilibrium lines are, if the separation is to proceed easily,
quite different, and might be as shown in Fig. 5.5. The operating lines, while having
different terminal conditions, must have the same slope, since this depends only on the
ratio of the flow rates of the solute-free solvents. The separation is effected by choice
of a slope of the operating lines such that one solute pinches in at the origin and the
other pinches in at the feed concentration. It is then possible to discharge a waste
stream which is completely depleted in the more easily extractable solute 1 and contains
pure solute 2. But the organic stream cannot avoid the extraction of some minimum of
the more difficultly extractable solute 2, and so the degree of purification of solute 1
in the product is limited. This is also shown in Fig. 5.5.

The limited effectiveness of the simple operation is overcome in the compound

operation, which is the third model to be studied. A compound operation comprises
two adjacent simple operations. The upper operation, called the scrub section, acts
as a stripping or partial stripping operation on the product of the lower operation, and
the feed to the lower operation, called the extraction section, consists of the aqueous

75
 UNCLASSIFIED
ORNL-LR-DWG 22783

EQUILIBRIUM LINE \

y(\)

EQUILIBRIUM LINE 2

/(2),

x{\)w x(2)w x(2)f x{\)f

Fig. 5.5. Two Independent Solutes in a Simple Solvent Extraction Cascade.

stream from the scrub section plus the fresh feed to the system. This is shown sche
matically in Fig. 5.6.

The purpose of the compound operation is to separate two solutes, but, since the
solutes are assumed to be independent, they can be considered one at a time. Solute 1
will be taken as that desired in the cascade product and solute 2 as that desired in the
waste. The slope of the operating line for the scrub section will be determined primarily
by the characteristics of solute 2, and the slope of the operating line in the extraction
section by the characteristics of solute 1.

Consider first solute 1. It will again be presumed that the over-all material balance
is available and that the problem is to determine the number of stages. The extraction-
section operating line has the slope (F + S)/Q, and it originates at the point x = x ,
y = yo = 0. Stages may be stepped off as indicated before, but there is the ambiguity
of the top end of this operating line (it obviously does not go to the feed concentration).
The same dilemma is reached by starting from the product end of the cascade and
working down through the scrub section, where the slope of the operating line is S/Q.
There is clearly a discontinuity in the material balance at the point where the feed

76
 UNCLASSIFIED
ORNL-LR-DWG 22784

P = Q

yp

A = S

A = F + S

0 — W = F + S
/0=0

Fig. 5.6. Schematic Diagram of Compound Counter-

current Multistage Solvent Extraction Cascade.

77
enters the cascade, but, since the concentration of the stream mixing with the feed is
dependent on the number of stages in the scrub section and, through the dependence
upon the concentration in the solvent stream entering the scrub section, on the number
of stages in the extraction section, additional information is required. To be adequate,
the information given must include one of the following:
1. the ratio of the number of stages in the scrub section to the number of stages in the
extraction section,
2. the number of stages in the scrub section,
3. the number of stages in the extraction section,
4. the concentration of the organic solvent leaving the extraction section (entering the
scrub section),
5. the concentration of the aqueous stream leaving the scrub section (mixing with the
feed),
6. an over-all material balance for solute 2.

If the number of stages in the scrub section is the available information (perhaps
assumed), and it is desired to know the number of stages in the extraction section, the
calculation proceeds from the top of the cascade through the specified number of stages
and yields the concentrations of both the aqueous stream leaving the scrub section and
the organic solvent entering it. This fixes the upper point of the extraction-section
operating line, the solvent concentration being the same as that entering the scrub
section (yfh), and the aqueous-phase concentration being easily determined from those
of the feed stream and the exit aqueous stream from the scrub section. The calculation
of the entering aqueous-phase concentration is automatically determined in the McCabe-
Thiele graphical calculation by use of the proper solvent concentration on the extraction-
section operating line, which has already been established.
This graphical calculation is shown in Fig. 5.7 for the case of two stages in the
scrub section. Notice that for solute 1 the concentration in the solvent increases above
the product concentration as we move toward the feed point. This represents a reflux
which does not result in a loss of solute but is caught in the extraction section, is
returned again to the organic solvent, and culminates with the product.
Meanwhile, these operating lines work differently on solute 2. Because its equi
librium line is lower, the scrub section very effectively removes it from the solvent, and
the extraction section cannot overcome this effect. Therefore, solute 2 leaves with
the waste. This is shown in Fig. 5.8.
When there are two solutes going through the cascade, the material balance for both
being given, and when they experience the same number of stages, the ambiguity as to
the number of stages in each section is resolved, and there is only one possible solution
(though trial and error is usually necessary). In this same vein, one will note that it
is not possible to specify arbitrary end conditions for a third solute.

78
 In radiochemical separations, the contaminant is usually present in very small
amounts, and the effectiveness of the separation is measured by what is known as a
decontamination factor (abbreviated DF, df, D.F., or d.f.). The decontamination factor
is defined as the ratio of the amount of contamination (or activity) associated with a

UNCLASSIFIED
ORNL-LR-DWG 22785

xf

Fig. 5.7. Calculation for Compound Cascade, Solute 1.

UNCLASSIFIED
ORNL-LR-DWG 23265

EXTRACTION

EQUILIBRIUM

Fig. 5.8. Calculation for Compound Cascade, Solute 2.

79
unit mass of valuable solute in the feed to the amount of contamination associated with
a unit mass of valuable solute in the product. Hence

xWiy\*)p Percentage of solute 1 recovered in the product

(5.4) DF = = • .
x(]),y(2)t Percentage of solute 2 recovered in the product
If, instead of specified end concentrations, one is given a cascade having a fixed
number of stages, fixed flow rates, and fixed feed, and is asked to determine the end
conditions, a trial-and-error solution is required. A fact important in the mechanics of
obtaining a solution is that the extraction-section operating line, the scrub-section
operating line, and the over-all material balance combine to require that the two oper
ating lines intersect at the feed concentration, x,. This is shown mathematically below.
The material balance for the scrub section (upper operating line) is

S
<5-5) ym =-q <*m-i - *,> +yP <
and the material balance for the extraction section (lower operating line) is

S + F
<5-6> yn ~—q— (*«-i ~*J + y0 •
At the intersection of the operating lines,

yn = ym - y • and
With this condition, subtraction of Eq. 5.5 from Eq. 5.6 gives

F S
— (x — x ) (x — x) — y^+y =0,
q y w' q w s P °

F(x - x ) - S(x - x ) = Q(yh - y ) .

The over-alt material balance states:

(5-7) Fxf + Sxs + Qyo = Qyp + (F + S)xw .

Therefore

x = x, .

The solution to the problem, then, can be conceived as being one pair of a family
of operating lines having the specified slopes and intersecting at x = %,, as shown in
Fig. 5.9. A pair of operating lines is chosen, the proper number of stages is stepped
off in the scrub section, and the number of stages required in the extraction section is
determined. If this is less than the number specified, the next set of operating lines
is chosen with a smaller x . Reasonable accuracy can usually be attained in three or
four tries.

80
 UNCLASSIFIED

ORNL-LR-DWG 22786

Fig. 5.9. Family of Operating Lines.

Where there is a very large difference between the equilibrium lines in a two-solute
system, flow rates can be chosen so that the operating lines pinch in: solute 1 in the
scrub section and solute 2 in the extraction section. As the equilibrium lines approach
each other, the choice of flow rates becomes more critical and pinching-in less definite.
Finally, in the limiting condition where the difference between the equilibrium lines is
very slight, the upper and lower operating lines approach being parallel. This condition
corresponds to a vanishing feed rate; the slope of the operating lines is intermediate
between those of the two equilibrium lines, and the number of stages is very large.
The model which has been treated involves three assumptions which should be
discussed further: the assumption that the solvents used (water and some organic
solvent) are completely immiscible, the assumption that the solutes are independent,
and the assumption that the operation can be described by discrete theoretical stages.

If the solvents used are not completely immiscible, that is, if some water dissolves
in the organic solvent and some organic solvent dissolves in the water, these two
substances to which all the mass balances have been tied will transfer between phases
throughout the cascade, and their rates will change from stage to stage. This results
in a curved operating line, and the McCabe-Thiele graphical method becomes unsatis
factory. This is particularly true if the amount of their mutual solubility is sensitive

81
to the solute concentration. For a single-solute (three-component) system, it is practical
to use a triangular plot to obtain a graphical solution. In more complex systems, if the
equilibrium data are available, a stage-by-stage numerical calculation can be used.
The necessity for the use of these methods is obviated by the fact that systems ex
hibiting marked mutual solubility are inherently undesirable for radiochemical processing
work. If appreciable amounts of organic solvent reported to the raffinate of a radio
chemical processing plant, it would be necessary either to sustain a costly chemical
loss or to perform a costly recovery operation from a highly radioactive stream.
Solubility of water in the organic solvent is less undesirable from an operational point
of view. Tributyl phosphate mixed with a kerosene diluent, a common organic solvent,
will dissolve several per cent of water. This water content is not sensitive to the
concentration of uranium or acid and can be treated as a constant constituent of the
solvent.

The assumption of the independence of two solutes is more rash. In fact, the use
of a single equilibrium line for any constituent in a multisolute system is of questionable
accuracy, since it is common practice to use agents (which are called "salting agents")
to change the form of the equilibrium curve. These agents frequently do not themselves
extract into the solvent but work only in the aqueous phase, usually contributing a
common anion which raises the equilibrium curve (or increases the distribution coef
ficient). When this is the case, an equilibrium line can be drawn corresponding to the
particular concentration of salting agent used, and the method described can still be
employed. When nitric acid is used as a salting agent, things become more difficult,
because the nitric acid is extracted. In this case, depending on the operating conditions,
the nitric acid operating diagram may "pinch in." Then the concentration of nitric
acid is constant over most of the stages, and once more it is possible to apply the
calculation method described. When the concentration of uranium (or any valuable
solute) is so high as to affect the nitric acid equilibrium line, the calculation is again
complicated. A significant difficulty is that the uranium concentration (or the concen
tration of any major solute) will affect the equilibrium line for the contaminants, par
ticularly when the uranium concentration is high. This usually reduces the calculation
for the contaminant to the level of an approximation (the data rarely require more). The
contaminants, however, are very rarely present in concentration sufficient to influence
the uranium equilibrium.

The third assumption, that the operation can be treated as a combination of discrete
stages, has obvious limitation when applied to equipment which operates differentially.
The misrepresentation is not, however, generally gross, and it decreases as the number
of stages increases. There is an alternative method of solving solvent extraction
problems by means of mass transfer coefficients and transfer units. This method is
built on a differential mathematical treatment and is thought by some to be superior to

82
the theoretical stage method. It was not covered in this discussion, because it was
thought to be less conducive to illustration of the various solvent extraction parameters
than is the theoretical stage method. A discussion of transfer units may be found in
most texts on the unit operations.

PROBLEMS

Problem 5.1. — For the solvent extraction system whose equilibrium data are repre
sented by Fig. 5.10, what is the minimum ratio of solvent-phase rate to aqueous-phase
rate (Q/F) that will allow complete transfer of solute A to the solvent phase in a
simple countercurrent extraction with a very large number of stages if the feed concen
tration is 50 g of A per kg of H,0?
Problem 5.2. — If, in addition to solute A described in Fig. 5.10, there is another
solute B which is a contaminant with a constant distribution coefficient of 0.16 (slope
of the equilibrium line), what is the decontamination factor for the Q/F used in Problem
5.1?

UNCLASSIFIED
ORNL-LR-DWG 22787

40

30
o
cr

a>
a. —

20

o
en
— /

10

/I I I I I I I I Mil
10 20 30 40 50 60
g of solute Aper kg of H20

Fig. 5.10. Equilibrium Diagram for Solvent Extraction of Solute A.

83
 Problem 5.3. - Would the decontamination factor in Problem 5.2 be improved by:
(a) increasing Q/F?
(b) drastically decreasing the number of stages?

Problem 5.4. - It was finally decided to use a compound solvent extraction cascade
instead of a simple one to separate A from B (same chemical system as in Problem 5.2).
The extraction section uses a very large number of stages, but the scrub section only
two. The operating conditions are as follows:
Solvent rate = 1 kg/min.
Feed rate = 0.4 kg/min (solute-free basis).
Scrub rate = 0.5 kg/min.
Feed concentration = 50 g of A per kg of HjO.
What is the concentration of A in the solvent leaving the extraction section?

Problem 5.5. —Now the equipment in Problem 5.4 is modified to provide a very large
number of stages in the scrub section in order to meet the specifications of virtually
complete recovery and complete decontamination.
(a) If only the feed rate in the operating conditions of Problem 5.4 is changed, what
is the minimum value which would meet the specifications?
(b) What is the maximum value?
(<r) If only the scrub rate is changed, what is the minimum value which would meet
the specifications?
(d) What is the maximum value?

Problem 5.6. — A pulse column 14 feet tall is used in the simple countercurrent
solvent extraction of a solute, M, from an aqueous phase. The concentration of the
feed is x, = 17 g per kg of hLO, and, when the solvent flow rate is equal to the aqueous
flow rate, 93% of M is recovered. The system has an equilibrium line given by the
expression

y = 20(1 - e-0-1*) ,
where x and y are in grams of Mper kilogram of water and solvent, respectively.
(a) How many theoretical stages does the pulse column develop?
(b) What is theHETS?

Problem 5.7. — A valuable solute, Z, is purified by solvent extraction using a

compound countercurrent solvent extraction column. The feed concentration is
x, = 10 g per kg of water. The feed rate is 100 pounds per minute, the aqueous scrub
rate is 20 pounds per minute, and the solvent rate is 200 pounds per minute, all on a
solute-free basis. The equilibrium line is approximated by the expression

lOx = y(y + 1) .

84
The column has eight stages in the scrub section and an adequate number of stages
in the extraction section to assure virtually 100% recovery (x = 0)7
(a) What is the concentration of Z in the aqueous phase six stages from the product
end of the extractor?

(b) What is the concentration of Z in the stream entering the extraction section?
(c) On what stage below the feed point has the concentration of Z been reduced to
0.01 g per kg of water?
Problem 5.8. — It is desired to decontaminate a material, M, which contains 0.001
per cent of a contaminant, C. For this, a solvent is to be used under conditions where
the equilibrium line for M is closely approximated by the equation

y = 2000 sin , 0 < x < 250 g/liter ,

200

where x and y are in grams per liter and x/200 is in radians. A seven-stage extractor
(100% stage efficiency) is to be used for simple countercurrent extraction of M from an
aqueous feed solution of 200 g of M per liter.
(a) If a straight operating line is assumed, what is the minimum ratio of solvent
flow rate to aqueous-phase flow rate that can be used to recover 99.90% of the M?
(b) If the contaminant has an equilibrium line approximated by y = O.Olx, what is the
over-all decontamination factor?

(c) For further processing, the extracted M is stripped from the solvent with water.
If there were available a contactor with an infinite number of stages, and if the same
equilibrium line for M were used, what is the maximum concentration of aqueous product
that could be obtained while still getting 100% stripping?
Note that an operating diagram plotted on Cartesian coordinates will prove in
adequate for stages of low concentrations. It is recommended that either a logarithmic
plot of the operating diagram or a numerical stage-by-stage calculation be used.

LIST OF SYMBOLS FOR SOLVENT EXTRACTION

A Aqueous-phase flow rate in cascade (mass of solute-free solvent per unit

time)
D Distribution coefficient, y/x
E Extraction factor, DQ/A
F Aqueous feed rate (mass of solute-free solvent per unit time)
Q Organic-solvent flow rate in cascade (mass of solute-free solvent per unit
time)
S Aqueous scrub rate (mass of solute-free solvent per unit time)
x Solute concentration in the aqueous phase (mass of solute per unit mass of
solute-free solvent)
xf Concentration in feed

85
 <, y Concentrations leaving stage n
x Concentration in scrub
s

x Concentration in waste
w

y Solute concentration in the organic phase (mass of solute per unit mass of
solute-free solvent)
y. Concentration in organic solvent leaving the extraction section
y o
Concentration in entering "fresh" organic solvent
Concentration in product
Concentrations leaving stage m (in scrub section)
a. , Separation coefficient, D(l)/D(2)
(1), (2) Indices designating solute 1 and solute 2

BIBLIOGRAPHY

References for further reading in solvent extraction

Badger, W. L., and J. T. Banchero, Introduction to Chemical Engineering, p 360,

McGraw-Hill, New York, 1955.
Coulson, J. M., and J. F. Richardson, Chemical Engineering, vol II, p 754, McGraw-
Hill, New York, 1955.
Treybal, R. E., Liquid Extraction, p 172, McGraw-Hill, New York, 1951.

86
 Chapter 6

ION EXCHANGE1

For the purposes of this discussion, "ion exchange" may be thought of as a chemi
cal engineering unit operation by which ions in an electrolyte solution are exchanged
for other ions of the same sign in an "ion-exchanger" phase by equivalent counter-
diffusion. The electrolyte solution is usually, though not necessarily, aqueous. The
ion-exchanger phase is usually physically a solid. The ion exchange materials most
commonly used are synthetic, organic-polymer-based resins, though the first exchangers
used were the inorganic aluminosilicates and these still have application.
If after the exchange took place the exchanger phase were treated as a reaction
product, ion exchange would properly be described as a unit process rather than a unit
operation. In the great majority of cases, however, the ion exchanger is an intermediate
in a purification or concentration process and is recycled repeatedly, as is the solvent
in the solvent extraction unit operation. Ion exchange belongs to the broad category of
diffusional operations, in which mass or heat is transferred between two phases across
an interface. Ion exchange calculations can be generalized in terms of either theoretical
stages or transfer units and of either distribution coefficients or separation factors.
These concepts have already been discussed for the continuous-countercurrent steady-
state case in the chapters on "Common Concepts Used in Separation Operations,"
"Distillation," "Solvent Extraction," and "Gaseous Diffusion." In this respect it is
of interest to note that ion exchange HETS values are often measured in centimeters
or inches, not feet, and may be as small as a few particle diameters. One reason for
this is that in a cubic foot of 30-mesh resin the surface area of the particles totals
approximately 1300 square feet.
The big difference between ion exchange and the other unit operations just mentioned
is that since the resin phase is mechanically a solid it is usually employed in fixed
beds instead of in continuous-countercurrent columns. When carried out in this manner,
the operation is not at steady state and belongs to the "percolation" operations, in
which mass or heat is transferred between a flowing fluid and a stationary bed of granular
solids. In general these are considerably more difficult to treat mathematically than
continuous countercurrent steady-state operations. Continuous countercurrent ion ex
change is, however, now beginning to come into use.
Ion exchange is used in the AEC program to concentrate and purify uranium from
ore leach liquors, to purify light and heavy water for use in reactors and steam boilers,
to purify materials for use in reactors, to separate radioactive elements from irradiated
materials and from each other, to decontaminate or concentrate radioactive wastes prior

This chapter was originally written by J. T. Roberts as a supplement to a lecture given by

him to several ORSORT classes.

87
to disposal, to purify and concentrate ("isolate") fissile materials, as a broad-spectrum
analytical tool, etc.

Mathematical Description. — In terms of the theoretical stage concept and the nota
tion employed in previous lectures in this course, a material balance for the unsteady-
state system can be written:

, .
dx n dyJn
(6.1) L .x . + V„ +1y ., = L x + V v + I + v ,
« —In— 1 b +1' » +1 n n nJ n n dt n At

where L and x refer to the solution phase and V and y to the resin, / is the holdup
of solute by the solution in stage n, and v is the holdup of solute by the resin in
stage n. There is also the equilibrium condition,

(6.2) yn = /<*„> <

from which

dy
'n
dyJn dx n df(x ) dx n
' * n' dxn
(6.3)
dt dx dt dx dt
n

At steady state, dx/dt and dy/dt are zero, and one has the system considered previously.
For a fixed bed, V = 0, and for flows and holdups constant, the result is
dx_
(6.4) Lx . = Lx + [I + vf'(x )] .
n— 1 n n dt

Equation 6.4 makes possible the determination of x as a function of t from (1) the value
of x _1 as a function of t, (2) the value of x at one value of t, and (3) the equilibrium
relationship 6.3. This, in turn, determines x .. as a function of t. Thus, if the concen
tration history at the inlet to a bed is given, the concentration history at the exit can
be found by iteration or can be expressed analytically if the equilibrium relationship
is simple enough.

UNCLASSIFIED
ORNL-LR-DWG 22794

l-n-2 *•„-( Ln */,+'

i 1 t 1 ~t r

n-\ n />+<

1 ♦ 1 I J L

This mathematical approach is only an approximation to the actual behavior of a

fixed bed, which should be described in terms of a material balance on a differential
length of the bed and a mass transfer rate equation:

♦"i*). ♦-&;.-■
/ dx
(6.5) M

88
(6.6) [^jLj =F(x,y) ,
where Z is the bed length, measured in the direction of flow, and I* and v* are solution
and resin holdups per unit length of bed. This more sophisticated analysis is beyond
the scope of this treatment.
Physical Chemistry Aspects. —A brief description of the preparation and properties
of a typical modern strongly acidic cation exchange resin, such as Dowex 50, is in order.
If a copolymer of styrene and divinylbenzene is sulfonated, the resulting resin, usually
in the form of uniform small spheres (10 to -400 mesh, depending on application), is
similar to a thermosetting plastic in its properties of temperature stability and essen
tially complete insolubility in almost all solvents. When placed in water the resin will
swell somewhat, taking up water to the extent of 30 to 50% by weight, forming a homo
geneous gel. The resin should not be thought of as a two-phase porous solid containing
water in rigid channels. As a first approximation the interior of such a hydrated resin
particle may be visualized as a highly ionized, concentrated aqueous solution of a
high-molecular-weight sulfonic acid surrounded by a semipermeable membrane which
will permit the passage of any molecules and ions except the sulfonate anions bound
to the resin matrix. This resin-phase solution typically has a concentration of about
7 equivalents per 1000 g of water, though this figure is sensitive to the divinylbenzene
content of the copolymer. The hydrogen ions are completely dissociated from the sul
fonate groups and are free to migrate in the particle as in an ordinary aqueous acid.
If the resin is placed in a solution, mass transfer will take place in the direction
of achieving two-phase equilibrium. The total ionic concentration of each phase will
remain nearly constant, since the fact that the anionic resin groups cannot migrate out
leads to a "Gibbs-Donnan" equilibrium, with relatively few free anions transferred
from the solution phase to the resin except when the solution is comparatively highly
concentrated. Thus, transfer of cations takes place primarily by "exchange," that is,
equivalent counterdiffusion. Figure 6.1 (drawn for a cation resin in "membrane,"
that is, sheet, form but equally applicable to a spherical particle) illustrates the re
lationship between the resin and the solution. In general, the ratio of the concentration
of one cation to that of another will be different in the resin phase from the same ratio
in the solution phase, and this is the basis for the use of ion exchange as a separation
process. The relative distribution of cations will be influenced by such things as their
hydrated radii, their valences, and the presence of a complexing agent in the solution.
As a rule, for ions of the same valence the one with smaller hydrated radius will be
concentrated in the resin phase, for example, K is sorbed in preference to Na by
Dowex 50. In dilute solutions the ion of higher valence is preferentially sorbed, while
in concentrated solutions the mass action effect tends to reverse this order of preference.
(In water softening, Ca is efficiently removed from dilute solutions by exchange with

89
 UNCLASSIFIED
ORNL-LR-DWG 20856A

l_©0j ^.

©0 ©0 |

©0
I MATRIX

ACTIVE GROUPS

© © ©0
S03H,OH,COOH)
©0

©0
DONNAN DIFFUSED IONS
©Q
IN MEMBRANE PHASE

© ©
© 0
©©
©©
SOLUTION PHASE MEMBRANE PHASE

Fig. 6.1. Schematic Drawing of Permselective Cation Membrane.

a resin in the sodium form, but the Ca is readily desorbed by a concentrated NaCI
solution.) Even in dilute solutions a polyvalent ion may be prevented from competing
with a monovalent ion by the presence of a complexing agent which converts the poly
valent cations to neutral or even anionic species. Nonelectrolytes tend to distribute
themselves between the two phases equally or to prefer the resin phase slightly, and
this behavior is the basis for separating electrolytes from nonelectrolytes by "ion
exclusion."
If the resin is prepared with a quaternary amine group instead of the sulfonic group
it will have properties similar to those just described except that its behavior toward
anions and cations will be reversed, that is, it will act as a strongly basic anion ex
changer. Attachment of weakly basic or weakly acidic or complexing groups to the
resin framework will affect its behavior toward hydrogen or hydroxyl ions or give it a
special affinity for certain types of ions. Figure 6.2 illustrates typical equilibrium
relationships, specifically for three anion exchangers of different basicities.
These properties, then, delineate the area of usefulness of ion exchange resins.
They can be used to exchange an ion in solution for another of like sign (for example,
jj233 0_++ or Pu+3 can be sorbed on Dowex 50 from a dilute nitrate solution produced
by solvent extraction and subsequently desorbed with a concentrated reagent, both
purification and concentration being achieved). They can be used to deionize an aqueous
solution by exchange of the cations for H ions and the anions for OH" ions (to produce
high-quality water for reactors or boilers). They can be used to separate two ions of
like sign by selective exchange with a third (for example, adjacent rare earths can be

90
 UNCLASSIFIED
ORNL-LR-DWG 4755A
1.0

WEAK-BASE
EXCHANGER

0.8 —

LU
or
0.6 —
z
o
z
o
I-
o
<
or 0.4

0.2 0.4 0.6 0.8 1.0

0H~ FRACTION IN SOLUTION

Fig. 6.2. Chloride-Hydroxide Equilibrium of Dowex 1 and Dowex 2.

separated by multistage exchange with NH4 in citrate solutions). They can be used
to separate electrolytes from nonelectrolytes (for example, CjHgOH from NaCI in water).
They can be used to sorb neutral salts by complex ion formation, for example, UOjSO^
by a sulfate-form anion exchanger which is thereby converted to the U02(S04)2 form
(as in recovery of uranium from sulfuric acid ore leach liquors). For many reactions
catalyzed by acids or bases the resins are ideal, since, being solids, they may be
separated easily from the reactants or products. In sheet form, ion exchange resins
act as membranes permeable to cations but only slightly so to anions ("permselective")
or vice versa, and hence can be used in continuous electrolytic cells to separate ions,
demineralize solutions, etc., with no chemical consumption.

Radiation Damage. - Like other organic materials, ion exchange resins are subject
to radiation damage. The order of magnitude of the problem is about the same as with
tributyl phosphate in solvent extraction, that is, power reactor fuels have to be "cooled"
at least several days before processing, and preferably considerably longer. Inorganic
ion exchangers are not so sensitive in this respect and may eventually find application
in continuous processing of fluid-fuel reactors.

91
 Operational Aspects. —Cyclic, fixed-bed operation involves alternate sorption and
desorption (also called "loading" and "elution") periods. Two or more beds can be
operated on a staggered cycle to permit continuous feed and product withdrawal, but,
even so, the composition of the exit streams will vary during the cycle* Figure 6.3
shows a cutaway view of a typical water-softening installation. This type of unsteady-
state operation, with its operational problems, varying stream composition, inefficient
utilization of resin and reagents, and complexity of mathematical analysis might seem
to be out of place in today's world of continuous chemical processing, and, indeed,
this has prevented wider use of ion exchange for large-scale applications. The design
and operation of continuous-countercurrentsolid-liquid contactors are difficult, especially

UNCLASSIFIED
ORNL-LR-DWG. 26648

Upper
manifold

Nozzles

Resin

Hard water
inlet Regenerant
Meter
Solo valve
Backwash controller.

Sight glass

Graded quartz

Backwash outlet

Strainer nozzles
Floor drain
Lower manifold

Fig. 6.3. Typical Ion Exchange Water Softener. (Drawing courtesy of Rohm & Haas Co.)

92
if the high throughput capacity and low HETS of the fixed bed are to be retained. Such
contactors are, however, coming along rapidly now.

Figure 6.4 shows fixed-bed and continuous-countercurrent methods of carrying out

simple sorption-desorption processes by ion exchange. A fixed bed is fed with the
solution to be processed until the concentration of the ion of interest rises to an objec
tionable level. The resin is then regenerated by elution with an electrolyte in order to
desorb the ion.

UNCLASSIFIED
Dwg No 19242

Feed •Elutant Elutant

Desorption

Alternate Sorption
and Desorption

Concentrate —-
Regenerated
Feed — Sorbent

Sorption

Processed _ Processed
•Concentrate
Solution Solution

FIXED BED CONTINUOUS COUNTERCURRENT

Fig. 6.4. Simple Sorption-Desorption Process.

The same type of sorption-desorption process could be carried out continuously

in a two-section countercurrent contactor. In the sorption section the regenerated resin
flows counter to the process feed, and in the desorption section the resin flows counter
to the regenerating solution. At steady state the processed stream and the concentrate
stream are of constant composition. In general a continuous contactor will have at
least as many sections as there are influent streams, while a fixed-bed system may
consist of a single bed.

93
 Figure 6.5 indicates how the ion concentration in solution varies with bed length
and with total volume put through the bed. The S-shaped "exchange waves"are charac
teristic of percolation processes. In a fixed-bed sorption step, if the exchange equilib
rium is favorable, the exchange wave (or "zone ') that moves down the bed finally comes
to a constant concentration-versus-length relationship which is the same as if the bed
were moving up at the same rate that the exchange wave is moving down. Essentially
all the useful mass transfer takes place within this zone, since above it the resin is
in equilibrium with the influent solution and below it the resin is in its unused re
generated form. It follows that the sorption section of a continuous-countercurrent
column need be only as long as the exchange zone in order to effectively saturate the
resin, while a fixed bed must be several times this long to even approach saturation,

UNCLASSIFIED
ORNL-LR-DWG 4756A

CONCENTRATION IN FLUID

Q
UJ
ao
Z

LU
O
z
<
r-
co
Q

Fig. 6.5. Ion Exchange Bed Concentration Profiles at Various Volume Throughputs.

94
since the resin in the exchange zone is only half-saturated. Extension of this reasoning
to the succeeding desorption operation shows that continuous countercurrent operation
makes possible lower reagent consumption and higher product concentration, in addition
to lower resin inventory.

Separation operations, a second general category of ion exchange problems, are

indicated in Fig. 6.6. A simple separation might be carried out by feeding a fixed bed

UNCLASSIFIED
Dwg No 19243

Feed Recovered
Elutant
Elutant
A + B
L_r
Sorption

Alternate Feed Separation

and Elution

A+B-
Regenerated
Sorbent

Separation

Desorption

oduct 1 ^ Product Elutant -

A B

FIXE D BED CONTINUOUS COUNTERCURRENT

Fig. 6.6. Simple Separation Process.

with a mixture of the two ions to be separated, then selectively eluting them. The
less strongly sorbed ion moves down the bed more rapidly, and the effluent is collected
in two fractions. The same separation could be carried out in a four-section continuous-
countercurrent contactor. In Fig. 6.6 the resin is shown moving down through the bed,
as it would in a gravity-operated contactor. The mixed feed enters at the middle of the
column, the eluting solution enters at the bottom, and regenerated resin is recycled
from bottom to top. The separation takes place in a manner closely analogous to dis
tillation, the separation sections above and below the feed corresponding to the en
riching and stripping sections, the top and bottom sections corresponding to the con
denser and reboiler, and the eluting solution corresponding to the heat input (part of
which is recovered and part of which comes out in the top and bottom product streams).

95
Figure 6.7 is a McCabe-Thiele analysis of a moving ion exchange bed continuously
recovering uranium from uranium ore leach liquors. It is fundamentally no different
from the analyses used in solvent extraction.

UNCLASSIFIED
ORNL-LR-Dwg. 9833

.01 .02 0 .1
EQUIVALENT FRACTION URANIUM IN SOLUTION

Fig. 6.7. McCabe-Thiele Diagram for Run 6-14.

Continuous Ion Exchange Equipment. —Groups at Dorr-Oliver, Inc., Allis-Chalmers

Mfg. Co., Stanford Research Institute, Yale University, University of Tennessee, and
ORNL, to mention a few, have worked on the development of continuous ion exchange
contactors. None of the units produced to date have found unqualified acceptance, but
the future appears promising.

Figure 6.8 indicates the mode of operation of the Higgins contactor developed at
ORNL. Strictly speaking, its operation is intermittent rather than truly continuous,
but in practice the distinction is unimportant. The solution flow is down against a
packed bed of resin; hence there is no danger of fluidization (which is a limitation on
the capacity of some designs). Intermittently, the bed is moved up a short distance
by momentary interruption of the solution flow and application of the proper hydraulic
pressure differential across the bed. The column of resin slides as a piston in a cyl
inder, lubricated by the liquid, while the individual particles continue to maintain
their relative positions. The result is "effectively-continuous-countercurrent" flow
retaining the high throughput and low-HETS features of the fixed bed.

96
 UNCLASSIFIED
Dwg No 217 38

u in

PUMPING PERIOD RESIN MOVEMENT PERIOD

•Several Minutes — 10 Seconds

^1
Fig. 6.8. Higgins Contactor — Flow Detail.
 The previous approach to resin-in-pulp ore processing was the use of "basket"
contactors. The large-mesh resin is retained loosely in screen baskets in a trough
through which the pulp flows. The baskets are jiggled to prevent plugging. Basket
contactors are comparatively inefficient, with poor resin and chemical economy and
bulky equipment, but they have been accepted as superior to fixed beds for the difficult-
to-filter ore leach liquors. Most of the ore mills still used fixed beds with filteredfeeds.
At least one large recently built mill uses baskets. Whether the baskets will be super
seded by the Higgins contactor remains to be seen.

BIBLOGRAPHY

References for further reading in ion exchange

Glasstone, S., Principles of Nuclear Reactor Engineering, sees 7.114-7.153, D. Van
Nostrand Co., New York, 1955.
Kunin, R., and A. F. Preuss, "Ion Exchange in the Atomic Energy Program," Ind. Eng.
Chem. 48(8), 30 A (August 1956).
Nachod, F. C, and J. Schubert (eds.), Ion Exchange Technology (including chap. 16,
"Processing Radioisotopes by Ion Exchange," and chap. 17, "Treatment of Radio
active Wastes"), Academic Press, New York, 1956.
Marshall, W. R., Jr., and R. L. Pigford, The Application of Differential Equations to
Chemical Engineering Problems (especially chap. XI, "The Transient Behavior
of Diffusional Operations Equipment"), University of Delaware, Newark, Del., 1947.

99
 ORNL-2477
Chemistry — Separation Processes for
Plutonium and Uranium
TID-4500 (13th ed., Rev.)

INTERNAL DISTRIBUTION

1. C. E. Center 81. J. A. Lane

2. Biology Library 82. M. J. Skinner
3. Health Physics Library 83. W. H. Jordan
4-6. Central Research Li brary 84. R. A. Charpie
7. Reactor Experimental 85. R. R. Dickison
Engineering Library 86. J. B. Adams
8-56. Laboratory Records Department 87. L. G. Alexander
57. Laboratory Records, ORNL R.C. 88. E. D. Arnold
58. A. M. Weinberg 89. J. C. Bresee
59. L. B. Emlet (K-25) 90. W. L. Carter
60. J. P. Murray (Y-12) 91. C. V. Chester
61. J. A. Swartout 92. G. K. Ellis
62. E. H. Taylor 93. A. T. Gresky
63. E. D. Shipley 94. C. E. Guthrie
64. S. C. Lind 95. P. A. Haas
65. M. L. Nelson 96. R. W. Horton
66. C. P. Keim 97. H. F. Johnson
67. J. H. Frye, Jr. 98. S. H. Jury
68. R. S. Livingston 99. P. G. Lafyatis
69. F. L. Culler 100. L. E. McNeese
70. A. H. Snell 101. J. J. Perona
71. A. Hollaender 102. J. T. Roberts
72. M. T. Kelley 103. A. M. Rom
73. K. Z. Morgan 104. C. D. Scott
74. T. A. Lincoln 105. J. C. Suddath
75. A. S. Householder 106. J. W. Ullmann
76. C. S. Harrill 107. C. D. Watson
77. C. E. Winters 108-113. M. E. Whatley
78. D. W. Cardwell 114-163. ORSORT
79. D. Phillips 164. ORNL - Y-12 Technical Library,
80. D. S. Billington Document Reference Section

EXTERNAL DISTRIBUTION

165. Division of Research and Development, AEC, ORO

166-662. Given distribution as shown in TID-4500 (13th ed., Rev.) under Chemistry - Separation
Processes for Plutonium and Uranium category (75 copies OTS)

101