NOTE TO THE EDITOR
Some Fundamental Aspects of the Behaviour
of Ion Exchange Equipment+
EDWARD J . FARKAS
University of Waterloo, Waterloo, Ontario
and
ALEXANDER HIMSLEY
Himsley Engineering Limited, Toronto, Ontario
on exchange in fixed beds i s a wellestablished
technology. A t t h e same time there is increasing
interest in carrying out ion exchange processes i n
agitated vessels. The advantages have been discussed
frequently"". Furthermore, a series of agitated vessels of proper design would o f f e r the possibility of
continuous operation"), as against t h e cyclic operation
necessary with fixed beds.
I n an agitated vessel, t h e beads of ion exchange
resin a r e suspended in the liquid and can move about
freely. However in many respects the process which is
occurring in an agitated o r wellmixed bed a n d in a
fixed bed is the same. The feed i s a solution containing concentration C, of the ion to be transferred.
Consider t h a t a t time t = 0 we begin t o feed this
solution to a mass of resin which is initially free of
the ion of interest. It will then be found that the
concentration of this ion in the effluent f r o m either
the agitated bed o r the fixed bed sooner o r later
begins to increase with time and ultimately approaches the feed value C,. Simultaneously the resin
is becoming loaded with the ion and the resin loading
will approach the maximum possible or saturation
loading as the concentration in the effluent approaches ('<,.
The present note is based on results of efforts t o
develop procedures for design of ion exchange equipment consisting of series of agitated beds. These
efforts began with analysis of previous work in this
field"". I t was found t h a t there were fundamental
errors i n some of t h e existing work, The object of
the present note is t o discuss this situation, to present
a corrected analysis, and to present some conclusions
which follow therefrom.
The basis for any analysis of the agitated ion exchange bed, whether f o r purposes of design or f o r
any other purposes, is the mass or more properly
component balance on the ion being exchanged. None
of the existing papers known to the present authors
appears to have included the f a c t that the presence
of two phases allows and even requires t h a t the component balance be written twice, once f o r each phase.
The transient behaviour described above indicates
tCorrespondenrp concerning this paper sholilcl Ire arl<lieh\e<l to E. J.
Farkaa. Department of ManEnvilonrnent Studies, University of
Waterlno. Wnierlon. Ontario, Canada
t h a t these component balances will take t h e form of
differential equations.
A f u r t h e r error has been made'""' in t h a t the
transient or differential t e r m was apparently inadvertently omitted from the differential equations.
It may be t h a t in some systems t h e transient term
would be numerically small compared to the other
terms in the equation. However t h e papers cited did
not attempt to use this argument. Furthermore, as
will be shown below, t h e characteristics of t h e system
of equations a r e such t h a t omission of t h e transient
term in one of t h e two differential equations results
i n a system of algebraic equations which is no longer
applicable to t h e ion exchange process.
A t least two early indications"","' of t h e correct
form of the equations exist, b u t t h e accompanying
discussion was fragmentary.
Component balances
Except in nuclear reactions, mass can neither be
created nor destroyed, but a particular chemical
species can be destroyed through conversion to another species. Therefore a species o r component
balance has a n additional t e r m as compared to a total
mass balance. The component balance must be
written
input by flow = output by flow
accumulation
disappearance due t o chemical reaction o r other nonflow causes
If t h e control volume f o r a n unsteady statte component
balance on t h e ion which i s to be transferred is the
vessel containing t h e suspension of resin particles in
liquid, then t h e component balance becomes
FC,
FC
+ d( V C ) / d t + d( WL)/dl + 0 .
. (1)
The last term on t h e r i g h t hand side is zero because
no ions a r e destroyed o r otherwise disappear from
within the tank. It should be noted t h a t there a r e
two accumulation terms, one f o r accumulation in the
liquid and one f o r accumulation on the solid. These
a r e t h e transient terms.
Equation (1) is one differential equation in two
unknown dependent variables, and therefore a solution is n o t possible. T h i s demonstrates why two component balances must be written.
The Canadian Iortrnal of Chemical Engineering, Vol. 53, October, 1975
575
If the control volume is selected to be the liquid
only, then the component balance becomes
FC,
FC
+ d(VC)/dl + R
(2)
Here the last term on the right hand side is not zero.
There is disappearance of the component for which
the balance is written from the control volume by a
process other than bulk flow. Each term in the above
equation has units of amount of material per unit
of time. The last term on the right hand side is the
rate of ion exchange.
The rate of ion exchange, as is well known, is
determined by resin particle size and pore structure,
size of the diffusing ion, electrical characteristics,
etc., and in general by the instantaneous concentration and gradient of concentration, of ion in the
liquid and in the solid (resin) phases. Despite the
existence of many hundreds of papers on the subject,
it is not yet possible to predict reliably from these
fundamental factors the numerical value of the rate
of ion exchange. In any case, i t is entirely adequate
at the present point in the discussion to represent
this complex functionality by R. Several workers in
the ion exchange fieldc6)have referred to R as the
function representing the kinetics of the ion exchange process. I t will be shown below however that
while there is of course a t all times a numerical
value of the rate of ion exchange, there is in effect
no such thing as kinetics in the ion exchange field
in the usual sense of the term.
Now writing the component balance for the control
volume consisting of the solid phase only, we have
0
+ d(WL)ldl + ( R)
(31
Here the disappearance term is negative because
there is actually appearance of ions within the control
volume by processes other than bulk convective flow.
Any two of the three Equations (13) a r e independent. The analysis presented in previous
involved selection of Equations ( 1 ) and ( 3 ) . This
selection is valid but less convenient than selection
of Equations (2) and ( 3 ) . However Equation (1)
was written incorrectly with the transient term for
the liquid omitted. The equation is then
FC.
FC
+ d(WL)/dt..
(4)
The next step in the previous work was to eliminate
the transient term for the solid phase between Equations ( 3 ) and ( 4 ) . The result is
FC, = FC
R
(5)
which is a steady state equation. Thus, the peculiarities of the applicable system of simultaneous differential equations lead to an inadvertant dropping
of all transient effects if the transient term for the
liquid is omitted. In one ~ a s e ~ ~
the
, next step was
differentiation, which retains the appearance of a
differential equation but does not change the fact
that the equations which a r e being handled describe
a steady state situation only, and are thus not applicable to ion exchange.
Numerous sets of experimental data on concentration in the effluent as a function of time for agitated
ion exchange beds have been published and these
data of course demonstrate clearly t h a t the situation
is one of unsteady state. Examples are Figure 10 of
Belter et al, Figure 4 of Marchello and Davis), and
Figure 6 of Slater.
Belter et
correctly wrote Equation (1) includi n g the transient term for the liquid phase. However
576
they did not write a second equation and were only
able to solve Equation ( 1 ) because they assumed that
the rate of ion exchange R depends only on C. In
the general case, as already noted, the rate depends
on both C and L . There must thus be two equations
which a r e linked through this dependence and must
be solved simultaneously.
Rodriguesc9also began his analysis using an equation identical to Equation (1). He also did not write
a second differential equation but was able to obtain
a mathematical solution by assuming equilibrium between the liquid and solid phases a t each instant of
time during the transient process.
Equations (2) and ( 3 ) completely describe any
situation where the ion exchange equilibrium is
favourable. For other cases, the buildup in the liquid
of the ion being displaced from the resin can affect
the magnitude of R. In this type of system, a third
differential equation would be needed. This equation
would express the component balance of the displaced
ion and would have a term containing R multiplied
by a factor expressing stoichiometry. R would now
depend on three quantities, C, L , and concentration
of displaced ion, and the three differential equations
would have to be solved simultaneously.
Solution of the component balance equations
In connection with a series of agitated ion excharigc
beds, Equations (2) and ( 3 ) can be written for
each bed in the series. For the n t l i bed in the series,
these equations become
FC,I = FC,
d(V,C,,)/dl
0 = d(W,L,)/dt  R . .
+R
(h)
I?J
The obvious next step is to solve these equations to
yield the behaviour of the concentration of the ion
in the liquid and solid phases with time. There is
nothing difficult in principle in this solution, with
the output from one bed in the series becoming the
input for the next bed. Explicit information on R
as a function of c and L is needed in carrying out
the solution. In practice the difficulty is that this
information will rarely be available.
For this reason and also because of the extensive
computational effort involved, solution of Equations
( 6 ) and ( 7 ) will not be useful f o r general design
purposes. However, as part of the present work, these
equations were solved for illustrative purposes, using
various empirical or assumed expressions for R as a
function of C and L . The caIculations were carried
out on an IBM 36075 computer using a fourth order
RungeKutta procedurel2. The calculations were
carried out for a set of eight stages in series, with
an equal amount of resin in each stage, with an
equal amount of liquid in each stage, and with the
resin i n each stage initially free of the ion of interest, the initial condition mentioned above.
For any particular assumed rate function, the calculation yields values of concentration C in the liquid,
loading L on the resin, and rate of transfer R as
functions of time. Values of concentration in the
liquid leaving each of the f i r s t f o u r stages in the
series a r e shown in Figure 1, for one such assumed
rate function and set of parameters C,, F, V , and W .
In commercial application of the ion exchange
equipment developed by Him~ley.~,the operation
would not be carried out as shown in Figure 1. Instead, a t somewhat over 100 minutes, when the resin
in stage 1 has become essentially saturated, this
The Canadian Journal of Chemical Engineering, Vol. 5 3 , October, 197.5
t , minutes
Figure 1  Concentration as a function of time in the effluent
from stages in series. C, = 1 gram/Iiter, F = 75 liter/minute,
V = W = 56.6 liters.
batch of resin would be removed. Then the resin in
stage 2 would be transferred into stage 1, followed
by transfer of resin from stage 3 to stage 2, etc.
These transfers a r e carried out without interrupting
the treatment of the feed solution.
Thus the operation is continuous.
The calculations described here are readily extended to this continuous mode of operation and in
fact form the basis for a practical design procedure
which will be the subject of a later paper. However
the extension to continuous operation is beyond the
scope of the present note and Figure 1 and succeeding Figures apply only to the case where each batch
of resin remains in the stage in which it was initially
located.
In pilot plant work using a Himsley unit with a
diameter of 2 feet, one set of data was obtained just
a t the time when the resin in the f i r s t stage in the
column became saturated. This occurred a t 135 minutes, before any resin was transferred. Therefore
this set of data falls within the scope of the present
discussion.
This set of data was not in t h e form of concentration as a function of time but rather was a set of
values of concentration in the liquid leaving each
stage all a t the same instant of time, namely 135
minutes. These values can be visualized as the intersections of the curves in Figure 1 with a vertical line
drawn at t = 135 minutes.
By trial and error a r a t e function was found which
allowed an approximate simulation of this set of experimental data. The comparison between experiment
and calculation in this format is shown in Figure 2.
The same r a t e function was used in obtaining Figure 1.
The trial and error procedure involved assuming a
rate function, solving the differential equations to
yield concentration in the liquid a s a function of
time, and comparing the calculated values at t = 135
minutes with the experimental values. This trial and
error procedure involved not only adjustable parameters within the assumed function, but also the
form of the function itself. The rate function which,
among those tried, yielded the best agreement between experimental and calculated values is
R
6/(R5
+ Ro) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
(8)
where
R.5
1/40C.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Re
3L/(3600L). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
(9)
I t must be emphasized that this function was used
for illustrative purposes only, and no claim is made
that this function has general theoretical significance
or yields a better f i t than any other function which
could be tried, for the particular ion and type of
resin used in the experiments. However, the functionality (810) can be said to have very approximate
theoretical justification, the description of which is
somewhat lengthy and beyond the scope of the
present note.
The general agreement between experimental and
calculated values as seen in Figure 2 allows some
confidence to be placed in the general conclusions
which were drawn from the calculation and which
are now described.
Significance for fundamentals of ion exchange
0999
>
2 0990
Lo
I
I
r
%
0900
:
m
3
*
e

0600
0300
It is clear from Figure 1 that the behaviour of
concentration with time in the effluent from the
f i r s t stage in the series i s quite different from the
behaviour seen in succeeding stages. The behaviour
in stage 2 is also somewhat different. This of course
is predictable without doing any calculations but
the data in Figure 1 supply some quantitative details. The significance of the fact t h a t resin in stage
1 experiences a different liquid concentration history
than does resin in other stages follows from an important statement made by Helffericho3:
60
/ I
cn
0100
l
n
0010
c
0
/ I
u
0 001
2
3 4
6 8
stage number ( log scale1
Figure 2  Comparison between experimental and calculated
results. a experimental; 0 calculated.
The Canadian Journal of Chemical Engineering, Vol. 53, October, 1975
0 5
10
concentration in liguid, grorn/liler
Figure 3 L, C Profiles.
577
The momentary rate of I)articlediffusion controlled
exchange depends on the shape of the nonsteady concentration profiles within t h e particle and thus on
t h e particle h i s t o r y . . . Thus i t is inherently impossible to give a universal relation between t h e moment a r y rate and t h e average concentrations in t h e particle.
To investigate this aspect further, we utilized the
viilues of L that were obtained in the same calculation which led to Figure 1. Rather than plotting
against time, we plotted L versus C, a plot we termed
the L , C profile, as a n expression of t h e total historical environment and experience of the resin. Plots
of the L , C: profile for each stage using t h e calculated
results a r e shown in Figure 3. I t is again seen t h a t
the profile for the f i r s t stage is significantly different than the profile for succeeding stages.
I t is seen t h a t t h e profiles coincide at higher
values o f L and C. Consider any point X where all
profiles have already coincided. The numerical value
of rate of ion exchange will not be t h e same at t h i s
point in each stage because t h e history up t o t h a t
point in each s t a g e h a s n o t been t h e same. Specifically, the numerical value of r a t e of ion exchange
at point X in stage 1 will be substantially different
than t h e numerical value of r a t e a t t h i s same point
in stages 3 through 8.
Figure 3 merely emphasizes t h e second portion of
t h e above quotation from Helfferich. T h e r e is t h u s
ii further. indication of the difficulty of obtaining
a completely valid rate expression R.
The concept of a kinetic expression intrinsic to
a given set of components originates f r o m t h e study
of chemical reactions f o r which it is of course valid.
Unfortunately, a s should now be clear, t h i s concept
does not carry over to ion exchange. I n particular
it is important to note t h a t r a t e d a t a obtained
in it single stage during pilot plant work will not
be representative of all stages i n a series a n d cannot
be used f o r designing a commercial unit consisting
of such ;i series.*
Some related conclusions can be obtained by examining the fundamental Equations ( 2 ) and ( 3 ) even
without solving them. It has not been possible for.
previous workers to draw these conclusions since they
did not write these equations correctly o r completely.
It is first convenient to define a specific r a t e or
rate per unit volume r:
R/W
(11)
Equations 2 and 3 can then be written
FC,
FC
lidLjdl
+ VdC/dl + r M
rll
1131
12)
Dividing through by F , and using the definitions
ru
7,<=
V/F
WjF
(15)
11
the equations become
C,
+ r VdC/dI + 7 , , r
(ISJ
and
clLldt
The set of simultaeous differential equations w n t t r i i
in the form (16) and (17) shows t h a t there a r e only
three operative parameters, C,, T ~ a, n d 7,(. The appearance of C,, in this list is another indication of
the history o r experience effect. The feed concentration f o r vessel 1 in F i g u r e 1 is substantially different
than the feed concentration f o r succeeding vessels.
If for any given situation the numerical values of
the three basic parameters change, then the L, C
profile can be expected t o change and the numerical
values of r a t e of ion exchange will be different. A s a
result the behaviour of effluent concentration with
time will be different. The data of Table 1 of Marchello and Davis (see also5) bear out this conclusion t h a t the behaviour of effluent concentration
with time will be different for different runs unless
all three basic parameters have t h e same values.
The indication is t h a t these three parameters a r ~
critical for scaleup of agitated ion exchange processes.
*The fourth run in Table 1 of Marchello and Davis does not
this conclusion.
Acknowledgment
Maurice Grimes and other members o f the renearcb and ile\v!loimic,nt.
staff of Rio Algom Mines Ltd.. Elliot T.ake, Ontaria, were of I(iciit
help in obtaining the experimental data. Hans Feige wrote the rginiiiiiter program used in the mlrulatiuns. Financial siipwrt foi. the h l i i i l y
W ~ Hrecwived in gart from the industikl researvh Proaram of
the.
National Research Council of Canada.
Nomenclature
c = concentration of ion of interest in the liquid, granis % t w /
c,
F
L
r
concentration of ion of interest in feed solution. grams
liter
= volumetric liquid fced rate, liters/minute
= concentration or loading of ion of interest on the i o t i
exchange resin, grams per liter of wet settled resin
= rate of exchange per unit volume of resin, grams/minutr.
liter
=
= rate of ion exchange in an agitated vessel, gramshiwitr.
= time, minutes
7,. = ratio of volume of liquid in one stage in ii series oC .;t;!Rc.
to volumetric liquid feed rate, minutes
1,
= ratio of volumc of resin in one stage to volumetric l i q t i i t l
feed rate, minutes
1,
volume ol liquid on one stage, liters
12 = Wet settled volume of resin on one stage
R
t
I<ef erences
Himsley, A,, Paper piesenled :it Canadian Uranium I r w l i i c ( ~ r ~
Association meeting, Elliot I,ake, Ontariv, May 1!)73.
Slater, M. J.. Canadian Mining a n d Metallurgical I%ulletir~.
67
( 7 4 2 ) , 93 (February 1974).
Stevenson. D. G., Proceedings, Conference on lon Exchaiiw. i i i
the Process Induatries 1969, pages 114120 (puhl. 1!17Ii hy Srwii*tg
of Chemiral Industry, London, England)
Himsley, A., Paper presented at Canadian Mineral PI,(II.~+.LI(II..,
conference, Ottawa, January 1975.
Belter, 1. A,, Cunningham, F. I,. and Chen, J . W.. l%iotrh.aml
Bioeng.. 15, h33 (1973).
Marehello, J. M. and Davis, M. W., Jr., IEC Fundamentals. 2 I I I ,
27 ( 1 9 6 3 ) .
Slater. M. J. and PrudhiJmme, l., MineR Branch Terhnir.al
Bulletin TB 158, DeiJartnient of Eneray, Mines and Hesi,urces.
Ottawa, September 1872 (Infurmation Canada ratalrrlr N I I . M X i 20/158).
Slater, M. J., Can. J. Chem. Enc.. 52. 43 (1974).
Rodrisues, A. E.. J. Chromatog., 102, 437 (1974).
Hancher, C. W.. Eng. and Min. J., 160 ( Y ) , 80 ( 1 9 5 9 ) .
Mason, D. R. and Piret, E. L.. Ind. Eng. Cheni., 4 2 , 817 ( 1 9 6 0 1 .
Fox, L. and Mayers, D. F., Computing Methoils for Scientists
and Engineers, Oxford, Clarenrlon Preaa, 10B8, iiage 202.
Helfferich. F.. Ion Exchange, Toronto, McGrawHill I%whC o . ,
Tne.. 1962, page 287.
Selke. W. A., Bard. Y. and Pasternak. D. A., AIChE .Journal. 2 11I ,
468 (195GI.
Davis. M. W., Jr.. K l i y l e r , M. It. and W a i d . .I. F., J r . . 11,s.
Atomic Energy C o m m i s s ~ m Rescarvh arid 1)evelopment tttwsrt
UP850. October l!K%
117)
Mnnusrript ieceived F e b ~ ~ i i i r2y0 , 1!)75 : arceljted f o r Iiublieation. .Iiine
1 7 , 1975.
Selkt. et a11111 made a somewhat fiimilar point roneerning the various
layers o f resin in n fixed bed.
578
The Canadian Journal
Y
of
Chemical Engineering, Vol. 5 3 , October, I Y 7 5