You are on page 1of 42

Ion Exchange in Molecular Sieves

by Conventional Techniques
Rodney P. Townsend 1, Risto Harjula 2
1 Scientific Affairs, Royal Society of Chemistry, Burlington House, Piccadilly,
London W1J 0BA, UK; e-mail: townsendr@rsc.org
2 Laboratory of Radiochemistry, PO Box 55, 00014 University of Helsinki, Finland;
e-mail: risto.harjula@helsinki.fi

Dedicated to Professor Gerhard Ertl on the occasion of his 65th birthday

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1 The Importance of Ion Exchange Phenomena in Molecular Sieves 2
1.2 Origin and Nature of Ion Exchange Behaviour in Molecular Sieves 5

2 The Importance and Utility of Theoretical Approaches . . . . . . . 9


2.1 Preference, Uptake and Selectivity . . . . . . . . . . . . . . . . . . . 9
2.2 Batch and Column Exchange Operations . . . . . . . . . . . . . . . . 13
2.3 Thermodynamic Parameters, Non-Ideality and the Prediction
of Exchange Compositions . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Kinetic Processes and the Prediction of Rates of Exchange . . . . . . 20
2.4.1 Hierarchical Model of Zeolite Particle or Pellet . . . . . . . . . . . . 21
2.4.2 Intraparticular Exchange Rate Processes . . . . . . . . . . . . . . . . 22
2.5 Trace Ion Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6 Column Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3 Experimental Approaches . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1 Practical Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Pitfalls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2.1 Selectivity Reversal and Ion Sieving . . . . . . . . . . . . . . . . . . 31
3.2.2 Zeolite Hydrolysis Effects . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2.3 Colloidal Solids in Suspension . . . . . . . . . . . . . . . . . . . . . 36

4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Molecular Sieves, Vol. 3


© Springer-Verlag Berlin Heidelberg 2002
2 R.P. Townsend · R. Harjula

1
Introduction

1.1
The Importance of Ion Exchange Phenomena in Molecular Sieves

Throughout the 1990s there was a decline in the number of fundamental studies
carried out on the ion exchange properties of zeolites and related materials. One
has only to examine the content of published conference proceedings on the
subject over the last 20 years to observe this trend: the situation has moved from
one where whole sessions were devoted to ion exchange studies, to one where
the subject is subsumed into sessions covering other areas. Part of this decline is
to be expected, as increased attention has been rightly paid to the intriguing
possibilities that can arise through the exploitation of newer alternative post-
synthesis methodologies, many of which are discussed elsewhere in this volume.
Nevertheless, the fact remains that conventional ion exchange techniques
continue to be used routinely for post-synthesis modification during the prepa-
ration of molecular sieves for major industrial applications. Also, there are now
areas where molecular sieves find major application directly as ion exchangers
per se. In this respect the situation has changed markedly since the early 1960s,
when Helfferich, in his classic book on ion exchange, could justifiably describe
zeolites “as ion exchangers they are of little practical importance” [1]. These
direct applications are especially detergency [2–7] and also the removal of
nuclear waste [8–13] or other environmental pollutants [3]. However, it is
generally a combination of properties of a particular zeolite in addition to its
ion exchange capability that has tipped the balance in favour of its use, rather
than any intrinsic superiority per se, which the zeolite may possess as an ion
exchanger.
If, therefore, conventional ion exchange remains an important post-synthesis
preparative technique, and the materials have in addition major direct applica-
tions as ion exchangers, why have the number of fundamental studies decreased?
It is certainly not because ion exchange behaviour of molecular sieves is suffi-
ciently well understood and predictable to render further fundamental research
studies unnecessary. Two causes are suggested to explain this decline:
1. Many theoretical treatments of the ion exchange reaction within zeolites
(both equilibrium and kinetic) are obscure and complicated. This has with-
out doubt rendered inaccessible the real value of the work to those many
workers who have a practical need to predict and control ion exchange be-
haviour during the industrial exploitation of molecular sieves. Although
theoretical understanding is important, it is easy to forget that the end pur-
pose of such work should be to provide information and tools that the
chemical engineer or other user of the molecular sieve can apply simply and
effectively. Obscurities in theoretical treatments mean that users often do not
appreciate how basic theory can be used, not just to simplify the number of
measurements which need to be made, but also to predict and control be-
haviour during application. The theory should not be an end in itself!
Ion Exchange in Molecular Sieves by Conventional Techniques 3

2. The second cause is related to the first. Even where the value of theory for the
prediction and control of the behaviour of these materials has been recog-
nised, the utility of these approaches has often been greatly reduced because
of the experimental methods which have been employed or by the poor
experimental data which have been available, or both. Indeed, it is only com-
paratively recently that a proper recognition has arisen concerning the num-
ber of potential pitfalls and difficulties that can militate against the acquisi-
tion of meaningful and accurate experimental data.
A good example of this is the frequently studied Na/Ca-zeolite A system, which
has received much attention because of its importance in detergency applica-
tions. Careful and detailed experimental studies over a period spanning some
20 years by different sets of workers [14–20] resulted in calculated values of the
standard free energy of exchange (kJ equiv–1*) which ranged from –0.59 [14] to
–3.09 [17]. Plots of the corrected selectivity coefficient (defined below; see

E
Fig. 1. Plots of the logarithm of the corrected selectivity coefficient ln KG [cf. KA/B in

Eq. (7b)] as determined by different workers for the Na/Ca exchange in zeolite A. ECa is the
equivalent fraction of calcium in the zeolite [(Eq. (3b)]. BRW Barrer, Rees and Ward [14];
A Ames [15]; WF Wolf and Furtig [16]; SW Sherry and Walton [17]; BR Barri and Rees [18];
WGC Wiers, Grosse and Cilley [19]; FT Franklin and Townsend [20]. Taken from [8]

* Throughout this paper the term “equiv” denotes 1 mol of unit negative or positive charges.
4 R.P. Townsend · R. Harjula

Eq. 7b) naturally show a similar diversity but also differ from each other in curve
shape and trends (Fig. 1). These marked differences (particularly at the extrema
of the plots) were variously ascribed to experimental error [20], to variable
quantities of non-exchangeable sodium in the materials employed [20] (the
materials differed in their source and in their method of preparation [14–20])
or to variable levels of hydronium exchange depending on the pH and other con-
ditions used [20, 21].
Thus, even for this very important example, not only is some of the published
theoretical work difficult to interpret, but also experimental data from different
studies are frequently incompatible and incomplete.
It is essential therefore that a critical review of advances over the last decade
should look at the developments in the context of the field as a whole. This is our
intention here. After a discussion of the origin, ubiquity and nature of ion
exchange behaviour in molecular sieves, recent advances in the application of
thermodynamic and kinetic descriptions of the ion exchange process will
be described. This will demonstrate some of the shortcomings of current
approaches, together with the relative paucity of reliable literature data that
can be applied easily and practically. This whole topic has particular rele-
vance to those industrial applications where zeolites are used directly as ion
exchange materials and this will be exemplified throughout the chapter using
two main examples. The first of these is the application of A- and P-type zeolites
as detergent builders, where the approach is to use a batch exchange approach
to remove hardness ions (especially calcium) as fast as is practicable before
the indigenous water hardness harms the wash performance of the detergent
product. The second concerns the treatment of nuclear waste, where a variety
of higher silica zeolites have been employed using a continuous (column)
process to remove, and subsequently store, high concentrations of monovalent
and divalent radionuclides such as caesium and strontium. For both these
major applications, in addition to selectivity, it is noteworthy that the systems
are normally multicomponent, that the kinetics of exchange are all im-
portant and that the morphology of the exchanger material must be controlled
carefully.
Post-synthesis modification comes into its own when preparing molecular
sieves with desirable and exploitable properties other than those of ion ex-
change, be they optical, magnetic, catalytic or adsorptive. Here it is not directly
the thermodynamic and kinetic ion exchange properties that are of prime
importance but rather which experimental, preparative methods are most com-
monly used. Thus it is important to assess what are the most appropriate exper-
imental methods of preparation, as well as to review the many pitfalls one can
fall into which can subsequently give rise to very inaccurate and inadequate
experimental data. These experimental problems can include framework
hydrolysis, hydronium exchange, dealumination, the presence of key trace impu-
rities, dissolution phenomena, carbonate and bicarbonate interference, colloidal
phenomena, metal ion complex formation and cation hydrolysis.
Having thus reviewed developments and advances over the last decade, the
chapter concludes with some recommendations on directions and topics for this
area of research in the future.
Ion Exchange in Molecular Sieves by Conventional Techniques 5

1.2
Origin and Nature of Ion Exchange Behaviour in Molecular Sieves

Ion exchange is a characteristic property manifested by most molecular sieves.


In essence, whenever isomorphous replacement of one cation by another of dif-
ferent charge occurs within an initially neutral crystalline framework such as a
pure silica molecular sieve, then a net electrical charge remains dispersed over
that framework. This is neutralised through the presence, within the micropo-
rous channels, of cations of opposite charge (often referred to as counterions).
An example of this is seen in the introduction by direct synthesis of small quan-
tities of aluminium into the silicalite framework to give the material ZSM-5.
Silicalite, the pure silica analogue of ZSM-5, is then seen to be just the end-mem-
ber of a set of isomorphous microporous molecular sieves that exhibit ion
exchange properties which are a function of the quantity and distribution of
aluminium atoms within the structurally similar frameworks. In addition, since
one can prepare, through post-synthesis modification of the framework com-
position, a variety of other isomorphous metallosilicates and metal alumino-
silicates, it is obvious that zeolites possessing ion exchange capabilities are a
common occurrence.
Pure aluminium phosphate molecular sieves are probably more common
than are pure silica analogues of zeolites. They resemble pure silica zeolites in
that they possess frameworks that are electrically neutral, but there is a signifi-
cant difference between these two classes of inorganic solids. In topological
terms both are 4:2 connected nets of T:O atoms (“T” denoting tetrahedral
framework and “O” denoting oxygen). From this it is obvious that it is only
required for the T ion to have a charge of +4 for the connectivity of the net to
give rise naturally to a neutral framework in concert with the oxide anions. This
is fulfilled for pure silicalite. In the case of ALPO molecular sieves the require-
ment is also fulfilled, but the 4:2 T:O net now comprises two types of strictly
alternating T-cations (aluminium and phosphorus, possessing respectively for-
mal positive charges of 3 and 5). Providing the cations alternate strictly through-
out the framework, the 4:2 Al,P:O net holds no overall charge; however, in
contrast to a pure silica zeolite, where the formal charge at every atomic centre
is zero, within a pure AlPO the formal charge is not dispersed homogeneously,
but changes from –1 at each aluminium to +1 at each phosphorus. This greater
heterogeneity of charge distribution may in part explain the experimental
observation that ALPOs frequently exhibit poorer thermal stability than do pure
silica zeolites.
For a particular ALPO molecular sieve to possess an ion exchange capacity as
an intrinsic property, it is necessary to prepare a material where some of the alu-
minium and/or phosphorus framework atoms have been replaced by other
atoms of different charge. This can occur using for example silicon, to form the
so-called SAPO materials, or with metals in addition or not to silicon, to form
respectively the so-called MeAPSO and MeAPO analogues. However, it is impor-
tant to note that although silicon could in principle replace either aluminium or
phosphorus to give rise to positively or negatively charged SAPO molecular
sieves, respectively, in practice only the latter process seems to occur, or another
6 R.P. Townsend · R. Harjula

process in which two silicons replace one of each of aluminium and phosphorus,
which gives rise to no net change in framework charge [22]. In MeAPSOs, diva-
lent or trivalent metal ions replace the aluminiums in the framework. In this way
the charge imbalance is minimised as these isomorphous substitutions either
make no difference to the overall framework charge (T3+ for Al3+) or only
increase it by one negative charge per substitution (e.g. Mg2+ for Al3+), a process
analogous to when aluminium replaces silicon in aluminosilicates [22].
Overall therefore, and in common with aluminosilicate zeolites, the norm is
for MeAPSOs and MeAPOs to possess cation exchange properties rather than
the reverse. In this respect, zeolites and ALPOs resemble many other classes of
ion exchangers that are mineralogical in origin, such as the clay minerals. These
are layered materials where a cation exchange property can arise primarily from
isomorphous replacement of trivalent cations by divalent, or tetravalent cations
by trivalent ones, within the layers [23]. However, there is a major exception:
these anionic exchangers are the double metal hydroxides, which are also lay-
ered structures but which exhibit a net positive charge across the lattice. The
“parent” material here is the mixed Mg,Al hydroxide, commonly referred to as
hydrotalcite. It would be intriguing to understand better the conditions (if any)
under which one might expect to synthesise microporous three-dimensional
framework structures which similarly have a net positive charge dispersed over
the lattice and hence an anion exchange capacity coupled with a molecular sieve
capability.
It is important to note that, up to this point, we have been considering the zeo-
lite, ALPO, SAPO, etc., as being described adequately as a 4:2 T:O net. This topo-
logical description, which in general terms is, as Smith points out [24], nothing
more than a mathematical construct of the human brain, does nevertheless
allow us to appreciate both the origin and magnitude of an ion exchange capac-
ity arising from T-atoms being replaced by others of different charge. However,
this description is not sufficient to cover the observed differences in ion
exchange properties (i.e. selectivity, kinetic rate, level of exchange) that may be
seen between various molecular sieves having similar exchange capacities. To
understand these differences, one must not only examine more closely the topo-
logical properties of the nets but also bring to bear structural considerations.
Considering these topological properties in more detail, it is adequate at this
point to take as read that all the T-atoms within the microporous net are joined
to each other by bridging oxygens. One can therefore concentrate on the T-
atoms only and describe molecular sieves in terms of four-connected three-
dimensional (4-conn.3D) nets of T-atoms [25] that, in turn, can be derived from
appropriate 3-conn.2D nets [26]. Considering the latter nets first, these differ
from one another in the ways the nodes (T-atoms) link to each other via net-
works of polygons. Any node can then be described by its “vertex symbol”, viz.
by its surrounding polygons with the number of each type of similar polygon
surrounding the node being denoted by a superscript [26]. Thus the simplest
example of a 3-conn.2D network (the hexagonal net) becomes a 63-net; a more
complicated example could be the 4.6.12-net which forms the basis for the
gmelinite structure [26]. Note that all the nodes within each of these two sepa-
rate examples are topologically equivalent. This need not be the case. For exam-
Ion Exchange in Molecular Sieves by Conventional Techniques 7

Fig. 2. Structure of mordenite viewed along the main 8-ring and 12-ring channels parallel to
the c-axis. Four topologically distinct types of T-atoms are observed within the 3-conn.2D
(4.5.8)1 (4.5.12)1 (5212)1 (5.8.12)1 net

ple, consider the case of mordenite (Fig. 2), which is derived from a (4.5.8)1
(4.5.12)1 (5212)1 (5.8.12)1-net containing four topologically distinct types of T-
atoms [24].
Similar considerations apply when one considers the 4-conn.3D nets that
constitute molecular sieves. Here it is often convenient to describe the structure
in terms of polyhedral units or cages, with the polyhedra described topological-
ly in terms of face symbols [25] (not to be confused with vertex symbols defined
above). Thus the face symbol for the familiar sodalite unit, which is geometri-
cally a truncated octahedron, is 4668 with all vertices geometrically and topolog-
ically equivalent. If these units are then linked together, for example either
through their 4-windows or half their 6-windows, one forms respectively the
zeolite A and faujasitic structures. Both these structures possess cubic symme-
try, with each structure comprising 26-hedral cages connected to each other
throughout the microporous zeolite framework, but the vertices of the sodalite
units are no longer all topologically equivalent. For zeolite A the sodalite units
enclose a cage which is the great rhombicuboctahedron (4126886) [25] whereas
for faujasite the cage is the so-called 26-hedron type II, denoted by the face sym-
bol 4641264124 [25].
Why are these matters significant when one considers the ion exchange prop-
erties of molecular sieves? The answer is that these topologically non-equivalent
T-atoms combined with the overall structural properties of the three-dimen-
sional microporous framework often give rise to several very different types of
local environments which repeat themselves regularly throughout the crys-
talline structure. These different local environments, evidenced by solid state
NMR combined with X-ray crystallography [27], are distinct in themselves, dif-
fering from each other sterically and electronically, and these differences will be
8 R.P. Townsend · R. Harjula

manifested not only through their characteristic adsorptive and catalytic behav-
iour, but also through their ion exchange properties. Formally, therefore, zeolites
may be regarded as comprising a set of crystallographically distinct sublattices,
each having characteristic selectivities for different exchanging cations, depend-
ing on these local environments [28]. The overall ion exchange behaviour of a
molecular sieve can therefore be a subtle function of the structural and topo-
logical properties combined. An important combination of structural and topo-
logical properties concerns the ordering of isomorphously substituted frame-
work atoms [29]: this determines what fraction of the overall framework charge
is found on each sublattice. Other significant structural properties can be losses
in symmetry through restricted rotation [27], and whether the sites are accessi-
ble to exchanging cations (i.e. the sizes of the micropore channels allowing
ingress and egress of exchanging cations plus water).
A further point is worth emphasising: since site heterogeneity in a particular
zeolite is manifested through such a set of crystallographically distinct sublat-
tices, zeolites differ in this respect significantly from some other common class-
es of ion exchangers, such as the clay minerals or the resins. Whereas in zeolites
well-defined sites are repeated regularly through the crystalline matrix, in clay
minerals and resins site heterogeneity is often manifested in terms of patches, or
regions of the surface where the sorption energies are approximately constant
[30]. Thus a statistical thermodynamic model of ion exchange for clay minerals
and resins [30] can differ markedly in character from ones developed for zeolites
[31, 32].
As a consequence of all these factors combined, both the equilibrium and
kinetic aspects of selectivity and uptake of ions within molecular sieves can
rarely be understood in a straightforward manner. Phenomena which have
received either considerable attention in recent years or deserve further study
include the so-called “ion sieve effect”, behaviour of high silica materials, the
effects that framework flexibility can have on selectivity and rates of exchange,
multicomponent ion exchange, prediction of exchange equilibria, and the possi-
bility of inducing phase transitions within zeolites through ion exchange. Many
of these are considered further below.
So far we have considered topological and internal structural factors which
give the molecular sieve particular ion exchange properties. However, an ion
exchange capacity can also be manifested which is not an intrinsic property of
the material. The source of this property is unsatisfied valencies occurring at the
termination of the crystal edges and faces, or at faults within the crystalline
structure. In formal terms, the origin of this is topological, in that this inciden-
tal and secondary property arises from disruptions in the net at interfaces, sur-
faces and faults, but the nature and extent of this incidental property depends
essentially on structural and morphological characteristics. For the former, we
can take as an example an ion exchange capacity arising either from the pres-
ence of silanol groups [33, 34], or from hydroxyl groups attached to aluminium
atoms situated at the surface [35]. In clay minerals, as much as a fifth of the total
exchange capacity may arise from such sources whereas in the case of zeolites
the contribution of such incidental (or secondary) ion exchange properties is
usually small compared to the intrinsic, or primary source. The exception here
Ion Exchange in Molecular Sieves by Conventional Techniques 9

can be high silica zeolites [35, 36], whose overall ion exchange properties have
received considerable attention over the last decade [37–42].
Interestingly, the external morphology can also be an important factor in
determining ion exchange behaviour of molecular sieves. The crystal habit, the
average crystallite size, the distribution of crystallite sizes and the properties of
aggregates of crystallites can all affect the magnitude of secondary ion exchange
characteristics, since these can alter significantly the surface to volume aspect
ratio and hence the number of external surface sites available [35]. Also, the
kinetic properties may depend on these morphological characteristics, as
instanced by recent studies on a highly aluminous form of zeolite P [6, 7].

2
The Importance and Utility of Theoretical Approaches

When a zeolite in (say) the sodium-exchanged form is suspended in a solution


comprising a mixture of different cations and anions, two properties of the
material are brought into sharp focus. The first of these concerns which types of
cations are “preferred” over sodium or each other by the zeolite. This property is
commonly referred to as the selectivity of a given form of zeolite for another
cation, but there are so many definitions of “selectivity” that the term “prefer-
ence” may be better used for the present. The second key property to which one’s
attention is drawn, and which is separate from selectivity (however defined), is
the rate at which the mixture of cations achieves its equilibrium distribution
between the exchanging phases (viz., the electrolyte solution and the sublattices
within the zeolite).

2.1
Preference, Uptake and Selectivity

The preference manifested by a molecular sieve for a particular cation is strong-


ly dependent not only on the character of the material under examination, but
also on the conditions of the system as a whole (viz., temperature, perhaps pres-
sure, composition of exchanger and solution phases, pH, nature of solvent, etc.).
Given a comprehensive definition of these conditions, the preference of a given
form of zeolite for a given cation will then be invariant for that set of conditions
because it is essentially an equilibrium property of the system. However, it is
important to define clearly what is meant by “preference”. There are numerous
selectivity coefficients defined in the literature and, on occasion, “selectivity
coefficient” is confused with “separation factor”, a function whose value does
depend strongly on the total ion concentration in solution. Similarly,“uptake” or
“loading” is often confused with “capacity”. To distinguish these terms, a few
basic definitions are required.
Considering as an example a binary exchange involving cations A (valency
zA) and B (valency zB), the reaction equation is usually written as:
– –
zA B zB + zB AzA = zA BzB + zB AzA (1)
10 R.P. Townsend · R. Harjula

where the overbars denote the exchanger phase. The preference displayed by the
zeolite for one ion over another is then described by a selectivity coefficient,
which is just a mass action quotient. According to the choice of concentration
units, a series of these selectivity coefficients may be defined which differ
numerically from one another:

c–AzB cBzA x
–x zB c zA
A B E
EAzB cBzA
kA/B = 0 ≠ kA/B = 01
–x zAc zB ≠ k A/B = 01
– (2)
cAzB c–BzA B A EBzA cAzB
where cA , cB are the cation concentrations in solution (mol dm–3) and the corre-
sponding concentrations in the molecular sieve are indicated with an overbar
(equiv kg–1 dry exchanger). The definition of kA/B is consistent with IUPAC rec-
ommendations [43] but is not very convenient for zeolites because of the signif-
icant water content. kA/B X and k E are selectivity coefficients in which the zeolite
A/B
phase cation concentrations are defined in terms of the mole fraction and equi-
valent fraction E, respectively:
––
X A = c–A / Si c–i . (3a)

E = z c– / S z c– .
A A A i i i (3b)
When zA = zB = zi , then equivalent and mole fractions are numerically identical
and k XA/B = kA/B
E . Otherwise, these functions are not numerically identical. In prac-
E
tice, kA/B has been used most extensively for studies on zeolites.
The selectivity coefficients given in Eq. (2) can be used to derive more funda-
mental equilibrium properties of the system, such as the standard thermody-
namic functions describing the exchange reaction (viz. DGq, DHq, DSq), provided
one has information on the nature and extent of all activity corrections for non-
ideality. However, the key point to note is that having defined the reference
states, by contrast with a selectivity coefficient, these standard thermodynamic
functions are independent of exchanger composition since they refer by defini-
tion to a reaction between components which move from one set of specific,
defined standard states to another. The magnitudes and signs of these standard
functions therefore give no immediate information whatsoever on the actual
preference which a zeolite may display for a particular ion under a given set of
experimental conditions. This point, obvious to the thermodynamicist, has
often been missed, and effort has been invested uselessly in attempting to relate
calculated values of standard thermodynamic functions to mechanistic theories
of exchange under real conditions. This has resulted in work being published
that is of little practical utility, if not plainly wrong. The issue of misunder-
standing and consequently misusing thermodynamic data in this manner is
expanded elegantly by McGlashan [44].
The selectivity of a particular molecular sieve for a given ion as a function of
exchanger composition is normally measured from an ion exchange isotherm,
which is an isonormal [45], isothermal and reversible plot of equilibrium distri-
butions of ions between the solution and zeolite phases. It is emphasised that it
is only valid to calculate selectivity coefficients, and derived thermodynamic
data, from isotherms which are reversible (that is, the forward and reverse
isotherms coincide within experimental uncertainty). The types of isotherms,
Ion Exchange in Molecular Sieves by Conventional Techniques 11

Fig. 3. Examples of ion exchange isotherms exhibiting both unselective and selective behav-
iour towards the incoming ion A (curves are respectively convex and concave with respect to
the ordinate). Clear limits to exchange are also observed which are lower than those expected
on the basis of the theoretical exchange capacity of the zeolite. The arrows depict reversible
behaviour

Fig. 4. Example of an ion exchange isotherm showing non-reversibility of exchange within a


plateau region, characteristic of phase separation and coexistence of two phases over the com-
position range corresponding to hysteretic behaviour

and the causes for the shapes observed, are discussed elsewhere [45]. However,
two isotherm types, which are particularly characteristic of molecular sieves
(although not uniquely so), are shown in Figs. 3 and 4.
Figure 3 shows isotherms for which only partial exchange for the incoming
cation occurs. The isotherm plots enable one to distinguish clearly various
basic definitions. Taking, for example, a constant level of exchange or uptake for

an incoming ion (e.g. EA = 0.5, then for this given uptake, the selectivity coeffi-
cient can vary from low to high values (cf. the two depicted curves). The abscis-

sa of the isotherm ranges from EA = 0 to EA = 1; values of EA are determined
by dividing the uptake by the ion exchange capacity, which is the number of
exchange sites of unit charge per unit quantity of exchanger (defined as con-
12 R.P. Townsend · R. Harjula

venient – see comments on this above). However, Fig. 3 shows curves which are

asymptotic to values of EA < 1, demonstrating that the maximum uptake (or
loading) under specified experimental conditions for the incoming cation can
be less than what would be expected from the value of the ion exchange capac-
ity. The cause of this may be due to inadequate experimental rigour, especially
during batch exchange experiments (see Sects. 2.2 and 3.2.1 for further discus-
sion); however, genuine “ion sieve” or “volume steric” effects can also operate as
a consequence of the crystalline and microporous nature of molecular sieves.
Ion sieving, known for a long time and commonly observed in zeolites, arises
when part of the microporous channel network within the molecular sieve is
inaccessible to the incoming exchanging cations simply because their ionic
diameters exceed the free diameters of the windows through which they must
pass [46]. The “volume steric” effect is less common, and arises when the
cations have free access to the microporous voids and channels within the
crystal but nevertheless the size of the incoming ions is such that the channels
are completely filled before 100% exchange for the incoming ion can be
achieved [47].
Over the last decade, during a series of studies on high silica zeolites includ-
ing ZSM-5, ZSM-11 and EU1-1, another possible cause for partial exchange has
been identified. Although full exchange of hydronium ion for sodium was
observed by Chu and Dwyer for a range of high silica zeolites [37], and ion sieve
effects were identified by the same workers to explain partial exchange with
some organic-substituted ammonium cations in ZSM-5 [39], Matthews and Rees
found more complex behaviour with alkaline earth and rare earth cations in
ZSM-5 [38]. Univalent cations exchanged to 100% but this was not the case for
multivalent cations. Part of the explanation for the significantly lower maximum
loadings found with multivalent cations (especially Ca2+ and La3+) was ascribed
to the distribution of the relatively low number of aluminium atoms in the
framework, which could make it difficult for multivalent cations to neutralise
effectively widely spaced negative charges on the framework [38]. To test this
hypothesis, McAleer, Rees and Nowak [40] carried out a series of Monte-Carlo
simulations which implied that the charge on divalent cations could only be sat-
isfied adequately by aluminium atoms within the framework which were spaced
apart by < 0.12 nm. More recently, similar experimental and theoretical studies
were carried out on zeolite EU-1, where analogous behaviour to ZSM-5 was
observed, although cut-off values for exchange were much higher in EU-1 [41].
Topological and structural differences between ZSM-5 and EU-1 were proposed
as explanations for this different behaviour [41] (see the earlier discussion in
Sect. 1.2).
Figure 4 shows a type of isotherm shape that is seen with crystalline ion
exchangers such as molecular sieves and clay minerals, but is nevertheless rela-
tively uncommon. The shape resembles the type II vapour adsorption isotherm
of the Brunauer classification, having a clear “plateau” region and inflexion
point. An example is the Na/K exchange in zeolite P [48] that was found to be
reversible over the whole range of equivalent fraction of potassium in the crys-

tal (EK ). Zeolite P has the gismondine-type structure (GIS [49]). More common-
ly, isotherms of this type are found to be partially irreversible in the plateau
Ion Exchange in Molecular Sieves by Conventional Techniques 13

region, resulting in a hysteresis loop between the forward and reverse isotherms
(Fig. 4). Examples of such hysteretic behaviour include the Na/K and Na/Li
exchanges in zeolite K-F [50], which is a framework structure isotype of eding-
tonite EDI [49], and the Sr/Na exchange in zeolite X [51]. Isotherms of this type
(whether fully reversible or not) are characteristic of systems where the process
of exchanging one cation for another induces structural distortions and changes
in the molecular sieve framework, resulting in the end-members of the exchange
– –
(EA = 0 and EA = 1, respectively) being different phases. If the framework is flex-
ible and consequently the required structural transformation can occur readily,
the plateau region (where the two phases coexist) will be reversible. This is the
situation observed for the Na/K exchange in zeolite P [48] which has long been
recognised as a material which exists as several structural varieties [52] depend-
ing on ion exchange form and level of hydration [53] and which is recognised as
having an unusually flexible framework [49, 52].
When a hysteresis loop occurs, this corresponds to a situation where the end-
members of the exchange exhibit limited mutual solid solubility; in other words,
over this region of the isotherm two separate phases coexist. Barrer and Kli-
nowski considered the conditions under which phase separation may be expect-
ed to occur in a statistical thermodynamic treatment involving an interaction
energy for entering ions wAA/kT [31]. When this term is sufficiently negative, so
that the cations segregate rather than form a homogeneous phase, they showed
that conditions could arise under which a physical mixture of two A- and B-type
crystals has a lower free energy than the homogeneous A/B phase [31]. If in
addition the nuclei of the A-rich phase grow within the B-rich “parent” phase
matrix then two positive free energy terms are involved in the exchange process.
These are a strain free energy resulting from the misfit between the new grow-
ing phase within the old, and an interfacial free energy. These tend “to delay the
appearance of the new phase beyond the true equilibrium points for forward
and reverse reactions” [31]. This is the proposed explanation for the hysteretic
behaviour seen in systems such as the Na/K and Na/Li exchanges in K-F [50] or
the Sr/Na exchange in X [51], and contrasts with P [48, 53]. This has significance
for the use of a high aluminium analogue of P in detergency [6, 7]. This materi-
al, named “maximum aluminium P” (MAP), has the gismondine framework
structure of zeolite P but with a Si/Al ratio of unity [6]. The unusually flexible
framework [49, 52] is reported to lead to cooperative calcium binding, as well as
to unusual water adsorption/desorption properties that enhance bleach stability
[6, 7]. These properties, combined with superior kinetic behaviour, result in a
material that reduces water hardness much more effectively than zeolite A (sic,
[6, 7, 45]).

2.2
Batch and Column Exchange Operations

Practically all industrial ion exchange applications, except the use of zeolites in
detergency, involve column operations (e.g. the removal of radionuclides from
nuclear waste effluents). However, basic studies of ion exchange equilibria are
usually carried out using the batch method.
14 R.P. Townsend · R. Harjula

It is instructive at this point to compare these two techniques by considering


the conversion of a zeolite from one ionic form (B) to another (A) as shown in
Eq. (1) and using the selectivity coefficient kA/B defined in Eq. (2).
In batch ion exchange, a given amount m of zeolite in the B-form is contact-
ed with a given volume v of a salt solution of ion A. At equilibrium, the ions are
distributed between the solid and solution phase according to:
c–AzB cAzB
= k . (4)
c–BzA
A/B 5
cBzA
5

The progress of the reaction is illustrated in Fig. 5 for two univalent cations
(zA = zB = 1) assuming a constant selectivity coefficient kA/B = 10 and an ion
exchange capacity of 4 mequiv g–1. It is clear that it is difficult to obtain a high
degree of conversion by a single batch equilibration. In this example, 430 cm3 of
0.1 equiv dm–3 solution of ion A is required for 99% conversion. This is almost
an 11-fold excess even though the exchange equilibrium operates in favour of
ions A.
In zeolites strong selectivity reversals are often observed and this makes it
very difficult to obtain a high conversion to the required ionic form. This prob-
lem is discussed in more detail in Sect. 3.2.1. Here, conversion will be discussed
in qualitative terms. The solution concentrations of A and B [Eq. (4)] can be
written as:

cB = c–A /(V/m) = EAQ/(V/m) (5)
and

cA = cA(o) – cB = cA(o) – EAQ/(V/m) (6)
where Q is the ion exchange capacity (equiv kg–1), V/m is the solution volume
(dm3) to zeolite mass (kg) ratio in the batch equilibration and cA(o) is the initial
Solution concentration (N)
Loading (meq/g)

Solution volume (ml)


Fig. 5. Batch exchange: loading of ion A in zeolite (solid curve) and concentration of A in solu-
tion (broken curve) as a function of solution volume when contacting 1 g of zeolite in B-form
batchwise with 0.1 g equiv–1 solution of A. Selectivity coefficient kA/B and exchange capacity Q
have been given values of 10 and 4.0 mequiv g–1, respectively
Ion Exchange in Molecular Sieves by Conventional Techniques 15

Average loading (meq/g)

Outlet concentration (N)


Effluent volume (ml)
Fig. 6. Column exchange: average loading of ion A in zeolite (solid curve) and concentration
of A in outlet solution (broken curve) as a function of solution volume passed through the
column. Mass of zeolite bed 1 g, inlet solution pure A at 0.1 equiv dm–3 concentration. kA/B and
Q as in Fig. 5

concentration of A (equiv dm–3) in the solution. To obtain a high conversion to


the A-form in a single equilibration, kA/B and cA must be high and cB must be low.
cB can be made low by using a large volume of solution per unit mass of zeolite
(maximum value of cB = Q/(V/m)) (Eq. 5). cA can be made large by using a high
initial concentration of A and large V/m ratios (Eq. 6).
In column exchange, a solution of ion A is passed through a column that con-
tains a given quantity (m) of zeolite. This process is illustrated in Fig. 6 using the
same parameters as in Fig. 5 for the batch exchange. In column exchange, the
conversion to the A-form proceeds much more easily, as ion B is constantly
removed from the system. However, ion A is not homogeneously distributed in
the bed, but is first taken up by material near the column inlet and the conver-
sion proceeds in the direction of solution flow. When most of the zeolite has
been converted to the A-form, ion A starts to emerge from the column and cA
tends to the value of the feed concentration, when the column has become com-
pletely exhausted.
The important point to note is that by contrast with batch exchange, far less
solution is needed for full conversion. In the example of Fig. 6, only 50 cm3 of
0.1 equiv dm–3 solution is required for every gram of zeolite to achieve 99% con-
version. This is only a 25% excess.
Figures 5 and 6 represent highly idealised cases and serve here only to
describe qualitatively the differences between batch and column exchanges. In
realistic situations, the selectivity coefficient decreases with increasing loading
of A in the zeolite (see Fig. 1). This means that an even higher excess of A must
be used under real conditions. In addition, in column exchange, the rate of
exchange reaction often tends to decrease at high loadings, which lowers the
gradients of the loading and concentration curves (Fig. 5) and increases the
solution volume needed for full conversion.
Pure synthetic zeolites are fine powders that are usually unsuitable for col-
umn operation. Therefore, batch methods are used for the study of ion exchange
16 R.P. Townsend · R. Harjula

equilibria. Granular zeolite exchangers that are suitable for column work are
manufactured by using suitable binders (e.g. clay, silica, alumina) and care must
be taken in extrapolating data obtained from batch experiments to column
operation.

2.3
Thermodynamic Parameters, Non-Ideality and the Prediction
of Exchange Compositions

To derive thermodynamic parameters of ion exchange, the normal procedure is


to correct for solution phase non-ideality first by deriving a corrected selectivity
coefficient in which concentrations within the external solution are replaced by
activities. The means by which this may be done, for binary or multicomponent
systems, is described elsewhere [54, 55]. The corrected selectivity coefficients
corresponding to k XA/B and kA/B
E are then:

––
XAzB aBzA
KXA/B = 92
–– , (7a)
XBzA aAzB

E zB a zA
E = A B .
KA/B – (7b)
EBzA aAzB
92

E is identical to the function K shown in Fig. 1 and taken from [20].


K A/B G
The thermodynamic equilibrium constant Ka is then obtained by integrating
the appropriate form of the Gibbs-Duhem equation to give as corresponding
expressions for Eqs. (7a) and (7b), respectively, the following:
1

lnKa – D = Ú lnKA/B
X dE
A, (8a)
0
1

lnKa – D = (zB – zA) + Ú lnKA/B
E dE
A, (8b)
0

where D is the water activity term [56, 57]. D is normally ignored on the assump-
tion its magnitude is small; however, it should be noted that for the most com-
monly employed formulation, corresponding to Eq. (8b) and after Gaines and
Thomas [58], D π 0 when the system is behaving ideally if zA π zB but rather
equates to (zA – zB) [56, 59]. This must follow since, when the system is behaving
ideally, the values of all the activity coefficients are by definition unity for all
compositions and hence Ka = K XA/B = KA/B E = constant [56, 57] since


x
f AzB E
g–AzB
Ka = KA/B – = K (9)
g–BzA
A/B 5
f BzA
5

where fi , gi are the appropriate rational activity coefficients for cations in the
exchanger phase in association with their equivalents of anionic charge.
Equations (8a) and (8b) provide the starting point for the prediction of ion
exchange equilibria in molecular sieves, an activity which has received a signif-
icant level of attention over the last decade or so. The basis for prediction comes
Ion Exchange in Molecular Sieves by Conventional Techniques 17

from a principle put forward some time ago [60], viz., that because D is small and
changes little with zeolite composition, and providing salt imbibition is negligi-
ble (which is true for relatively dilute electrolyte solutions [61]), then for a giv-
en zeolite composition, the ratios of activity coefficients fi ,gi will hardly change
in value as the total concentration of electrolyte in the external solution is
changed [56, 60, 62]. Providing these assumptions hold, then taking as an exam-
ple a binary exchange process, from Eqs. (8b) and (9), it follows that [62]:
1

E – = (z – z ) + lnK E dE – zB – – zA –
lnKA/B(EA) B A Ú A/B A – ln (g A(EA)/g B(EA)) (10)
0

where the subscripted EA in parentheses indicates that the values of the correct-
ed selectivity coefficient and the rational activity coefficients refer to a particu-
– –
lar composition EB , EA and must be invariant since all the terms of the right-
hand side are constant or hardly change when the total concentration of the
external electrolyte solution is changed. The details of the methods which must
be employed to predict selectivity trends are described elsewhere [62]; the
important point to note is that if the above assumptions hold then for successful
predictions it is only required to evaluate the appropriate corrected selectivity
coefficient as a function of zeolite phase composition and to have an accurate
knowledge of the solution phase activity coefficient g [54, 55, 62]. For binary
exchanges, this approach has been used to test a variety of systems over the last
decade, including exchanges involving Pb/Na, Pb/NH4 , Cd/Na and Cd/NH4
equilibria in clinoptilolite, ferrierite and mordenite [63–65] using different co-
anions (chloride, nitrate and perchlorate [62, 66]) as well as the Cd/Na-X and
Cd/K-X systems [67], with a high level of predictive success [62, 67]. Recently, a
related model has been used with good accuracy for the prediction of K/Na and
Ca/Na equilibria over a wide range of total ionic concentrations in solution for
natural clinoptilolite [68]. Successful predictions were also achieved for the
Ca/Na, Ca/Mg and Mg/Na systems in zeolite A [18, 20, 69]; however, for Mg/Na
and Mg/NH4 exchanges in a range of faujasites [70], predictions failed badly in
some cases. The failures were attributed at the time to salt imbibition, but fur-
ther detailed experimental studies involving hydronium exchange in the Ca/Na-
X, Ca/Na-Y, Cs/Na-MOR and Cs/K-MOR systems [71–75] have shown that the
situation is in reality much less straightforward. Failures in predictive methods,
particularly at trace levels of exchange, cannot be attributed simply to hydroly-
sis, hydronium exchange or salt imbibition despite earlier suggestions to this
effect [70, 76]. An important factor appears to be the presence of colloid-size
zeolite particles [74]. These matters are discussed further in Sect. 3.2.3.
To apply the same prediction procedure as that described above for ternary
or multicomponent exchanges, it is helpful to derive analogous equations to
those shown in Eqs. (8), (9) and (10) for binary exchange. For ternary exchange,
this was done by Fletcher and Townsend [77] and this approach was used to
predict compositions for Na/Ca/Mg-A [20, 69], Na/K/Cd-X [67] and Na/NH4/
Mg-X,Y ternary equilibria [70]. For the first two of these systems, ternary
exchange equilibria were predicted successfully but for the Na/NH4/Mg-X,Y sys-
tems, the procedure failed for the higher silica Y materials, as for the corre-
sponding conjugate binary exchanges [70]. In parallel with these studies, the
18 R.P. Townsend · R. Harjula

model of ternary ion exchange in zeolites [77] was compared with other models
published in the literature for clay minerals and resins [78, 79] and a further
detailed study [80] came to the conclusion that these other approaches were
appropriate under certain specified conditions [80] for the prediction of ex-
change equilibria in zeolites.
A recent criticism of the ternary exchange model [81], on the basis that the
equations could have simply been built up from the conjugate binary systems
(obviously true), overlooks the main point. If one uses the conjugate binary sys-
tems it is necessary to use a model-based approach to predict activity coeffi-
cients for the multicomponent exchange equilibrium in the zeolite and the pres-
ence of sublattices within the zeolite framework can make this more difficult to
do than for clay minerals and resins (Sect. 1.2) [80]. The ternary exchange mod-
el of Fletcher and Townsend [77] does not require one to measure at all the activ-
ity coefficients, let alone predict them for multicomponent systems from binary
data, using some model. All that is required is knowledge of the ternary correct-
ed selectivity coefficients that are obtained by integrating the appropriate
Gibbs-Duhem equations over the ternary composition surface [77] in analogy
with the binary approach pioneered by Gaines and Thomas [58]. However,
acquiring sufficient data for a ternary system is a difficult and time-consuming
exercise [20, 67, 70, 82] and simpler approaches can prove quite adequate pro-
vided one validates some of the predictions made [83]. Thus, another model,
developed originally for clay minerals [84], has been shown after minor revision
to work well for ternary anion [85] and cation [86] exchanges in organic resins
and has even been extended successfully to a five-component zeolitic system
(Sr/Cs/Ca/Mg/Na equilibria in chabazite) [87]. This system is very important in
the field of nuclear waste treatment [87].
Accurate prediction is similarly much needed for detergent applications [2, 7,
18, 69]. The level and nature of “hardness” in household water varies extensive-
ly from one location to another, as do the conditions under which consumers
expect effective laundering to occur (e.g. temperature). Thus accurate selectivity
data (i.e. isotherms and selectivity plots as a function of loading), and reliable
predictive models that are simple to use, are important, since it would clearly be
impossible to measure directly the performance of a given “builder” zeolite for
all conceivable situations. Successful predictions have been achieved for the
binary Na/Ca-A, Na/Mg-A and Ca/Mg-A systems [2, 18, 69] as well as for the cor-
responding ternary system [2, 69]. Similar successful predictions were recently
achieved also for zeolite MAP [7] once the original iterative procedures of
Franklin and Townsend [69] had been modified appropriately. Figures 7 and 8
show examples of such successful predictions in A, for both the binary and
ternary cases.
Occasionally, isotherms of binary and multicomponent exchanges are
described using various empirical adsorption equations. These cannot be used
for the prediction of multicomponent equilibria [88]. In fact, a closer inspection
of these equations reveals that they have no in-built facility for true prediction
(i.e. for the calculation of equilibria over ranges of different total solution
concentrations for heterovalent exchanges). Thus these equations are useful
in describing the observed isotherm in a mathematical form but the only pre-
Ion Exchange in Molecular Sieves by Conventional Techniques 19

Fig. 7a, b. Predicted isotherms and experimental points for a the Na/Ca-A system and b the
Na/Mg-A system. Solid lines are predicted isotherms; experimental points are measured at
normalities of 0.025, 0.10 and 0.4 equiv dm–3, shown respectively as solid triangles, circles and
squares. Taken from [69]
20 R.P. Townsend · R. Harjula

Fig. 8. Ternary experimental and predicted points for the Na/Ca/Mg-A system at a normality
of 0.4 equiv dm–3. Measured solution and zeolite phase equilibrium compositions are shown
as unfilled stars and filled squares, respectively. The predicted zeolite phase at 0.4 equiv dm–3
is shown as an unfilled circle while the filled circle represents experimental validations at
0.4 equiv dm–3. Taken from [69]

diction these equations can give is the interpolation of the isotherm under
one given set of experimental conditions. With such limited utility, these empir-
ical approaches are not recommended for the “prediction” of ion exchange equi-
libria.

2.4
Kinetic Processes and the Prediction of Rates of Exchange

In direct applications involving zeolites as ion exchangers, it is not normally the


case that the system is allowed to reach equilibrium. In batch operations (e.g. in
detergency) the time available may be such that the exchange process is inter-
rupted long before equilibrium is reached. Similarly, in column operations (e.g.
effluent purification), when the system is operating under steady-state condi-
tions, the balance between throughput of liquid and time of exchange means
Ion Exchange in Molecular Sieves by Conventional Techniques 21

that the system is frequently operating under non-equilibrium conditions.


Knowledge of the kinetics of the multicomponent exchange processes (i.e. all of
the reaction rates, diffusive mechanisms and hydrodynamic processes which
contribute to the overall rates of exchange of all of the different types of ions
involved) is therefore of key importance if one is to be able to predict and con-
trol behaviour. Unfortunately, this is easier said than done. The kinetics of ion
exchange processes in zeolites are extremely complicated even when one focus-
es on just one mechanistic process [45]; only recently, it was rightly stated that
the “picture presented in the literature for diffusion in zeolites is confusing, con-
flicting and/or inconsistent with theory” [89]. Space permits only a brief
overview of the current state of affairs and this is presented here using a hierar-
chical model [90] for the zeolite particle or pellet. Much more detail is given else-
where [45].

2.4.1
Hierarchical Model of Zeolite Particle or Pellet

Whether one is considering an agglomerate of aggregated zeolite crystallites, or


a pellet, a hierarchical model [89, 90] allows one to distinguish the different
transport and/or rate processes which operate at different length scales.
The highest level is concerned with the macroparticle or pellet itself; and the
key issue here is whether transport of ions through the fluid film which encom-
passes the macroparticle is rate-controlling or not. That this process can be rate-
controlling has been recognised for a long time, being favoured by a low con-
centration of exchanging ions in solution and a small mean particle size; how-
ever, it is known that the hydrodynamic regime pertaining can affect its influ-
ence markedly, with high levels of agitation (such as are achieved at high
impeller speeds in a batch reactor [89]) rendering relatively insignificant any
mass transfer resistance through the boundary film. The mechanical integrity of
the macroparticle can also be very important. Taking detergent powder particles
as an example [which can comprise agglomerates of (primary) zeolite crys-
talline particles held together by means of adhesive, viscoelastic surfactant
bridges], these are designed to break up under shear and/or other hydrodynam-
ic regimes that are imposed as part of the wash cycle. On breaking up and dis-
persing, some of these dispersed smaller particles may find themselves in
regions of low agitation and consequently the rate of removal of hardness ions
from the wash liquor can be slower than desired due to the onset of film diffu-
sion control.
Generally, however, the aim is to avoid conditions leading to film diffusion
control. This means that the focus is shifted towards transport processes that
occur at the intermediate level (that is, in the mesopores and macropores with-
in the macroparticle or pellet itself) and those which occur at the smallest
dimensional level (viz., in the very micropores of the molecular sieve) [45,
89]. Within the mesopores and macropores between the primary zeolite crystal-
lites transport will be dominated by molecular and ionic intercrystalline dif-
fusion possibly coupled to surface diffusion processes, while, in the zeolite
micropores themselves, intracrystalline diffusion occurs, also possibly coupled
22 R.P. Townsend · R. Harjula

with specific exchange rates associated with the different zeolite sublattices
[91, 92].
The overall observed kinetics of exchange is of course the result of all the
above-described mechanisms working in concert [45, 89]. To cope with the com-
plexities of the system, a simple approach one may adopt is the homogeneous
diffusion model, which assumes that the behaviour of each distinct diffusing
species within the macroparticle may be described in terms of a single solid-
phase “effective diffusivity” [89]. More sophisticated approaches include the
heterogeneous diffusion models, where the macropore and micropore diffusion
processes are described separately and are then assumed in different mathe-
matical treatments either to occur in series or in parallel [45, 89].
In practice, to date, most research activity has focused on the intraparticular
diffusion which takes place in the zeolite micropores themselves, on the ques-
tionable assumption that these processes are normally the rate-controlling ones.

2.4.2
Intraparticular Exchange Rate Processes

Our understanding of the processes which govern the rates of ion exchange
within the micropores of molecular sieves has advanced little over the last
decade, yet the imperative to be able to control and manipulate these rates
remains as strong as ever. To summarise the current situation it is necessary first
to emphasise some basic principles and then to define certain terms and coeffi-
cients.
To begin, it is important to distinguish the intrinsic dynamic nature of the
system from the kinetic processes we actually observe during an ion exchange
reaction. An obvious yet important point to remember is that even after
exchange equilibrium has been attained, the equilibrium is a dynamic one. Thus
transport of all exchangeable cations and of the solvent molecules continues but
after equilibrium has been reached there are no net changes in the relative dis-
tribution of species between, and hence concentrations in, phases with time.
This dynamic character is readily verified by adding to the equilibrated system
a trace amount of a radioactive isotope of one of the cation types into (say) the
solution phase of the system and then observing the rate at which isotopic
exchange between the two phases takes place. The isotopic exchange process
may include as a rate-determining step an intracrystalline exchange process [91,
92] but it is also certainly a transport process, which is described in terms of a
self-diffusion coefficient D*AA [93]. Self-diffusion coefficients D*AA and D*BB ,
which can change markedly with temperature [45] or as the equilibrium con-
centrations of different cations within the zeolite are altered [45, 94], should be
sharply distinguished from the exchange diffusion coefficient DAB [95]. DAB
describes the kinetics of the A/B exchange process, that is, the observed rates of
change of concentrations of ions A and B within each phase as a function of time
and as the system moves to equilibrium.
Consider therefore a binary A/B exchange between the zeolite and external
solution, which is not initially at equilibrium. On mixing the two phases, the A
and B cations, which will almost certainly possess different ionic radii and pos-
Ion Exchange in Molecular Sieves by Conventional Techniques 23

sibly charge, will begin to move in their respective directions of negative chem-
ical potential gradient in order to equalise their respective chemical potentials
within all phases in the system. However, the mobilities of the two cation types
A and B are likely to be different, which means that the more mobile cation type
will tend to build its concentration, and hence lower its concentration gradient,
faster than the other. If this process were to continue unchecked, charge separa-
tion within each phase and between the phases would occur, with a concomitant
electrical potential gradient. In practice, of course, the electrical potential gradi-
ent that forms as charge separation takes place does not build, but rather acts to
slow the faster moving cations and speed the slower ones. Thus it is not adequate
to consider only the chemical potential gradients. The net flux JA of (say) the A-
exchanging species is actually described by:
J = – D [grad c– – (z c– F/RT) grad V]
A AB A A A (11)
where F is the Faraday constant and V the electrical potential. An expression for
DAB has been derived by Barrer and Rees using an irreversible thermodynamic
approach. The form of this is complicated but, if cross-coefficients other
than those due to the electrical potential gradient are assumed to be negligible,
then [96]:
D* – 2 – – – 2 – –
AA D*
BB [c Az A(∂ ln a B / ∂ ln c B) + c B z B (∂ ln a A/ ∂ ln c A)]
DAB = 00000000 – 2 – 2 0. (12)
D*AAc AzA + D* BBc BzB

Two points should be noted from Eq. (12). First, the magnitude of DAB depends
strongly on the composition of the exchanger not only because it is a direct func-
tion of ionic concentrations, but also because it is a function of both D*AA and
D*BB, which we have already noted vary with exchanger composition [45]. Sec-
ondly, DAB is a function of the non-ideality of the zeolite [data for which can be
obtained, as we saw earlier, from the activity coefficients described in Eq. (10)].
One may expect therefore that to describe adequately the kinetic behaviour of
even a binary exchange process in a molecular sieve would be a very complicat-
ed task.
To validate this and other similar models, it is necessary to solve, using appro-
priate boundary conditions, the differential equations describing overall the
transient diffusion process for each ion, of the general form:
(∂ c– /∂ t) = div D gradc–
i AB i (13)
which for spherical symmetry (a good approximation for most primary zeolite
particles) becomes [95]:
∂ c–i 1 ∂ 2 ∂ c–i
6 42 5
∂t
=
r ∂r
r D 
AB 6 .
∂r  (14)

As an example of the above approach, Brooke and Rees [95] studied the Sr/Ca-
chabazite system. Figure 9 shows their computed time-dependent concentration
profiles within the zeolite particles both before and after non-ideal behaviour
was taken into account. The effect on DAB of taking non-ideality into account was
even more dramatic, with a discontinuity appearing in the plot of ∂DAB / ∂ c–A
24 R.P. Townsend · R. Harjula

Fig. 9a, b. Radial concentration distributions at various fractional attainments of equilibrium


for the Ca/Sr exchange in chabazite for a the ideal exchanger and b the non-ideal exchanger.
The continuous lines represent the Ca/Sr exchange and the broken lines the reverse process.
Taken with permission from [95]

(Fig. 10). Even allowing for this non-ideal behaviour, prediction of exchange
rates was still poor [95]. Other similar studies, including the measurement of
∂D*ii / ∂ c–i functions, are described elsewhere [45, 94].
It is unfortunate that we are still not able to rationalise adequately the kinet-
ics of ion exchange in zeolites, let alone manipulate rate processes. For example,
in realistic detergency applications, the issue can be of prime importance, since
the contact time of zeolite suspended in the wash solution is usually shorter than
the time required to attain the equilibrium state. Elsewhere in this chapter the
strong effects that crystallite size and mesoporosity can have on kinetic rates has
been emphasised (Sects. 1.2 and 2.4.1); it is precisely these properties which are
identified as being key (in addition to cooperative calcium binding) for the
superior performance of zeolite MAP as a builder [6, 7].
For zeolite A, binary and ternary kinetic measurements of the Na/Ca/Mg
exchange have been undertaken [21] in addition to equilibrium studies. For the
ternary system, the inhibiting effect of magnesium on the uptake of calcium ions
was clearly demonstrated (Fig. 11) [21].

2.5
Trace Ion Exchange

In the preceding sections ion exchange processes involving large changes in the
chemical composition of the solution and zeolite phase have been discussed.
Under these circumstances, attention has to be paid to the changes in the value
of the selectivity coefficient with composition. In the case of exchange of trace
ions for the ion present at much higher concentrations (described henceforth as
the “bulk ion”), the situation is somewhat different. This brings us to the other
important area of zeolite application, viz., the purification of nuclear waste efflu-
Ion Exchange in Molecular Sieves by Conventional Techniques 25

DAB (¥1014) cm2 sec–1

CSr
Fig.10. Variation of the exchange diffusion coefficient DAB as a function of equivalent function
of strontium (CSr) for the Ca/Sr exchange in chabazite. Taken with permission from [95]

Fig. 11. Examples of ternary ion exchange kinetic measurements within the Na/Ca/Mg-A sys-
tem. The dashed lines which bound the experimental data represent the simple binary
exchange rates (viz, Na/Ca and Na/Mg). When magnesium is added to the system in progres-
sively larger amounts (i.e. O > ■) a progressive slowing of the Na exchange rate is seen. Taken
with permission from [21]
26 R.P. Townsend · R. Harjula

ents. This always involves column ion exchange. What is of interest in this appli-
cation is the capacity (Qv) of the ion exchanger, in terms of the solution volume
(V) that can be treated with a given amount (m) of ion exchanger. The maximum
(saturation) value of this capacity (Qv,max) is unambiguously given by the distri-
bution coefficient KD of the radionuclide. In general, the distribution coefficient
KD of ion A is defined by the equilibrium ratio:
K = c– /c .
D A A (15)
The distribution coefficient is determined by two factors, selectivity and ion
exchange capacity. Let us consider here uni-univalent exchange for clarity.
Inserting c–B = Q – c–A (where Q = ion exchange capacity) into kA/B in Eq. (2) and
combining the resulting expression with Eq. (15) gives for KD
KD = Q/((CB/kA/B) + CA) . (16)
It can be seen that when cB /kA/B  cA (i.e. when the ion A is a radioactive trace
ion), then the capacity of the ion exchanger is independent of the concentration
of the trace ion A in solution but depends only on Q and CB .
Let us consider removal of radioactive Cs ions (e.g. 137Cs) from a waste solu-
tion containing sodium salts.As an example, the chemical concentration of 137Cs
in solution corresponding to an activity concentration of 1 µCi dm–3 (which is
typical in low-active waste) is 8 ¥ 10–11 mol dm–3. Selectivity coefficients kCs/Na
are typically in the order of 10–100 in zeolites. Thus, unless the Na concentra-
tion in solution is very low ([Na]  10–8 mol dm–3), the KD (and volumetric
capacity) of the exchanger is independent of the concentration of caesium in the
solution and the familiar relationship is obtained from Eq. (16) (B = Na, A = Cs)
for KD , in the logarithmic form
log KD = log (kCs/NaQ) – log cNa . (17)
In other words, the KD of the trace caesium ion is inversely proportional to the
concentration of the macro-ion (sodium) in the solution. The selectivity coeffi-
cient can be assumed to be constant in this case as the loading of caesium in the
exchanger is very low. In general, the logarithmic equation for KD is [75]
log KD = (1/zB) log (kA/BQzA) – (zA/zB)log cT (18)
where CT is the total concentration of exchanging ions in solution (mol dm–3).
Thus, plotting log KD against the logarithm of the bulk ion concentration yields
a straight line with a slope of –(zA/zB). In experiments this equation is used to
determine the selectivity coefficient kA/B , which is obtained from the intercept of
the linear plot. A linear plot also confirms the stoichiometry of the exchange
reaction over the concentration range of interest. However, quite often the log KD
plots are linear in the more concentrated solution of B only. In dilute solutions,
leveling-off of the log KD plot is often observed. This can be rationalised when it
is kept in mind that Eq. (18) is valid for the free cations with the charges indi-
cated at the equilibrium concentrations of the cations. Many meaningless data
have been produced when this point has been forgotten. This issue is discussed
further in Sect. 3.2.2.
Ion Exchange in Molecular Sieves by Conventional Techniques 27

Very few data can be found in the literature for the exchange of trace ions in
the presence of more than one type of bulk ion. Harjula et al. [75] have studied
exchange of 134Cs in mordenite and in mixed salt solutions of sodium and potas-
sium. It was found that at a given constant total concentration CT (cNa + ck) of
solution, KD of 134Cs was a linear function of the potassium loading in the zeo-
lite. The equilibria were treated using appropriate pseudobinary selectivity co-
efficients. From the linear dependence of log KD a more simple treatment can
also be obtained, i.e.:
––– –––
log KD = ENa log (kCs/Na Q/CT) + EK log (kCs/K Q/CT) (19)
where kCs/Na and kCs/K are the (limiting) binary selectivity coefficients for trace
Cs exchange in pure Na- and K-forms of the zeolite. There are too few data to
conclude whether the form of Eq. (19) is generally valid for the calculation of
trace ion distribution coefficients in the presence of two or more types of bulk
ions. If Eq. (19) were always valid the following equation would apply for trace
ion A in the presence of M other different ions present in much higher concen-
trations:
M
log KD (A) = Â (1/zi) Ei log (kA/i [Q/CT]zA) (20)
i=1

where the kA/i are the selectivity coefficients of the trace ion A in the pure i-forms
of the zeolite. In order to use Eqs. (19) and (20), additional models for the pre-
diction of bulk ion concentrations need to be used (see Sect. 2.4).

2.6
Column Models

With a few exceptions, industrial applications of zeolites involve column opera-


tion in feeds containing more than two counterions. In general, therefore, the
prediction of column performance involves the prediction of multicomponent
equilibria and kinetics under dynamic flow conditions. Considering the com-
plexity and diversity of these models (see Sects. 2.3 and 2.4), it is obvious
that simplifications and approximations need to be made for practical column
modelling. For engineering purposes, the most popular approach for column
modelling is the “linear driving force – effective plate” concept [97].
Consider again the removal of a radionuclide from a solution as an example
of column exchange. By definition (Eq. 15), KD describes the equilibrium distri-
bution of ions between the zeolite and the solution. However, at the same time,
it is a measure of the equilibrium distribution of the solution volume and zeo-
lite mass. Thus the total column capacity can be calculated from Eq. (16) or (18)
for a simple binary system. The volume that can be treated with the column con-
taining m kg of zeolite is equal to the area above the breakthrough curve
(Fig. 12), and can then be calculated from V = mKD . However, in the purification
of radioactive effluents, it is necessary to discontinue operation, and change to a
fresh column, immediately when the radioactive ion first starts to emerge from
the column. The volume at which the breakthrough of the radionuclide com-
28 R.P. Townsend · R. Harjula

Fig. 12. Schematic representation of the effect of plate number N on the breakthrough curve
in column exchange

mences is called the breakthrough capacity (QBT) of the column. The break-
through capacity is usually defined in terms of some chosen level of break-
through (e.g. at 1%). The efficiency of the column can then be measured by the
degree of column utilisation, which is the ratio of breakthrough capacity to the
maximum capacity (QBT /KD).
According to the plate concept, the number of “transfer units” or “effective
plates” (N) is a measure of the column efficiency. Increasing N makes the break-
through curve steeper and thus improves the degree of column utilisation
(Fig. 12). The number of effective plates can be calculated from fundamental
data. Thus for film-diffusion controlled exchange:
Nf = A(Df)1/2d –3/2 suo–1/2 (21)
where s is the column length, d is the particle diameter, uo is the linear flow rate
and Df is the diffusion coefficient. For particle diffusion controlled exchange:
Np = BKDDpsd –2uo –1 . (22)
A and B in Eqs. (21) and (22) are empirical factors. Both mechanisms may con-
tribute simultaneously to the overall exchange kinetics but can be taken into
account in an appropriate model [98].
It can be seen from Eqs. (21) and (22) that the degree of column utilisation
and the breakthrough capacity increase when the zeolite grain size is decreased
and the solution flow rate is decreased (Fig. 12). In the limit, the kinetic perfor-
mance is determined by the magnitude of the diffusion coefficients.
The plate approach has been used in the development and operation of the
process for the purification of the highly radioactive solutions that arose in the
accident in the Three Mile Island nuclear power station [99]. In this process,
a mixed zeolite bed (Linde IE-96 and A51) was used for the removal of 137Cs
and 90Sr.
Ion Exchange in Molecular Sieves by Conventional Techniques 29

3
Experimental Approaches
It has been emphasised already that accurate and reliable data are essential in
the construction of adequate ion exchange models for the industrial applica-
tions of zeolite and other ion exchangers. In this section we will discuss ion
exchange experimentation and its utility for industrial applications. We also
discuss major pitfalls that may lead to unreliable results. Although industrial
applications always involve more than two exchanging ions, seeing trends in the
overall equilibria under these circumstances may be difficult. Therefore, in this
section, only binary ion exchange equilibria are considered in order to keep the
major issues in focus.
There are basically two major reasons for studying ion exchange equilibria.
The first of these is concerned with understanding selectivity and its causes. For
these studies, correlations are sought between the properties of exchanging
cations (e.g. cation size, charge, acidity, etc.) and the ion exchanger (charge den-
sity, pore diameter, acidity, etc.) with the aim of predicting the magnitude of the
selectivity. In the case of zeolites, where strong decreases in selectivity are often
observed (see, for example, Fig. 1), it is also of great interest to predict how selec-
tivity changes with loading. To date, no useful and general theory has been
developed for these predictions, in zeolites or indeed in any other ion exchange
materials.
Secondly, for application-oriented studies, selectivity data are measured in
order to predict the performance of the zeolite under given operational condi-
tions (ion concentration, temperature, contact time, etc.) using appropriate ther-
modynamic or kinetic approaches and hence to choose and optimise the oper-
ating conditions for the application in question.

3.1
Practical Experiments

It should be obvious by now that one of the key tasks in ion exchange experi-
ments is the accurate determination of the selectivity coefficient. In principle,
this is straightforward: the zeolite, initially in the B-form, must be equilibrated
in solution using an increasing ratio cA /cB and, after equilibrium has been
attained, the concentrations of A and B in the zeolite and also in the solution are
measured.
This can be done either batch-wise, or, in the case of the granular zeolites, col-
umn-wise. Both techniques should give in principle an identical result. Howev-
er, there is one important difference between the two techniques. In column
exchange, the equilibrium concentrations of A and B in the solution at equilib-
rium will be known beforehand since these will be equal to the initial concen-
trations of A and B in the feed solution. Because of this, it is a relatively simple
matter to decide the initial conditions for the experiments in order to determine
the selectivity coefficient as a function of the loading. For instance, isonormal
(e.g. 0.1 g equiv –1) solution mixtures of A and B may be prepared, containing
progressively increasing amounts of A (1%, 5%, 10%, 20%, …, 90%). Each of
30 R.P. Townsend · R. Harjula

these solutions is pumped through the column until the outlet concentrations of
A and B are equal to those in the inlet. Equilibrium concentrations of A and B in
the zeolite are then determined by direct analysis of the solid, or by analysing a
solution containing the dissolved zeolite.
In batch ion exchange, the equilibrium concentrations of A and B in the solu-
tion will not be known beforehand. These will depend on the experimental con-
ditions (total solution concentration, ion exchange capacity, and the solution
volume to zeolite mass ratio V/m). This makes it difficult to decide on the best
initial conditions for the experiments. Commonly, the zeolite is converted step-
wise from the B-form to the A-form by successive equilibrations in solution mix-
tures of A and B (low conversion) or in solutions of pure A (high conversion).
Only some rough guidelines are available for choosing the initial conditions for
the experiment. Thus, after the first measurements have been evaluated, it is
often necessary to carry out further experiments in order to fill in gaps in the
distribution of the data points across the isotherm.
In general, the advantage of batch equilibration is that the experimental appa-
ratus is simple so that a large number of experiments can be carried out in par-
allel using a minimal amount of solution and zeolite. In column experiments,
one column “run” is required for each selectivity measurement and a large num-
ber of solution concentration measurements have to be carried out to check that
equilibrium has been finally attained. Run times can be very long, especially
when the equilibrium is unfavourable at higher loadings or when the uptake of
trace ions is being studied. For instance, determination of a KD value of
20,000 cm3 g–1 requires that at least 40,000 cm3 of solution is passed through a
1-g zeolite bed. Such an experiment may take several months. By the batch
method, the same information can be obtained by carrying out the experiment
in a 20-cm3 plastic bottle in just 1 week. It is clear that the batch method is the
preferred option when large quantities of materials are to be assessed in paral-
lel, or when multicomponent equilibria are to be studied. In finely divided zeo-
lites the batch method may be anyway the only alternative.

3.2
Pitfalls

One might infer from the above that the measurement of zeolite ion exchange
selectivities is simple. In practice, several factors may interfere which may
distort the result. As a consequence, in theoretical work, understanding the
selectivity data may become impossible as one tries to rationalise these distor-
tions without knowing their origin. In application-oriented work, a com-
pletely wrong picture may be obtained about the performance and utility of a
given zeolite due to these problems. In the following these problems are exam-
ined.
Ion Exchange in Molecular Sieves by Conventional Techniques 31

3.2.1
Selectivity Reversal and Ion Sieving

Many ion exchanges in zeolites involve incomplete exchange so that some of the
ions (usually Na) originally present in the zeolite are not exchangeable for the
incoming cations. This may arise from ion sieving, volume steric effects or from
very low framework charge densities (see Sect. 2.1). Often divalent cations or
large cations are partially excluded. In these cases it is also common to observe
very strong selectivity decreases with increasing loading of the incoming cation.
A strong selectivity reversal is also common in exchanges that have nearly gone
to completion. Due to this selectivity decrease it is very difficult to convert a
zeolite from one ionic form to another even though there should be no steric or
other hindrance to 100% exchange. In addition, it may be very difficult to detect
whether a genuine saturation, or maximum loading, has been achieved, since
when the selectivity becomes low, very small changes take place in the ion con-
centrations even when the exchange is pushed forward by large increases in the
amount of incoming cation.
Accurate determination of maximum exchange level is very important,
since in the determination of the selectivity coefficient one should consider
only those ions that are exchangeable [100]. Non-exchangeable cations are
obviously not formally involved in the equilibrium, so their presence need not
be taken into account directly, although of course their effects may be made
manifest indirectly in the values of the activity coefficients of the exchanging
ions [100]. The use of the correct value for the maximum exchange limit is
especially vital in the determination of the thermodynamic quantities of the
exchange reaction (i.e. the thermodynamic equilibrium constant or the ionic
activity coefficients in the zeolite), since the determination of these quantities

involves integrating the appropriate selectivity coefficient from EA = O to 1
(Eq. 8) and this scale and the magnitude of the selectivity coefficient strongly
depend on the choice of the maximum exchange level fmax . Barrer, Davies
and Rees [101] demonstrated the great effect of the choice of the fmax on the

magnitude and variation of the selectivity coefficient with EA . A major aim
of many fundamental ion exchange studies in the zeolites has been the ratio-
nalisation of the selectivity gradient, since this reflects the non-ideality of
the zeolite phase. It is clear that any attempts to do this require a very reliable
value for fmax . In addition one should not compare systems from which ther-
modynamic parameters have been derived using different values of fmax , since
the reference states of the systems are different and therefore are not directly
comparable.
As a general rule, conversion of a zeolite from one ionic form to another in
one single batch equilibration is difficult, even when the zeolite is selective for
the incoming cation. Considering also the common selectivity reversal exhibit-
ed by most zeolites, conversion by a single equilibration becomes a practical
impossibility in most cases.
Despite this, in many studies in the past, only single equilibrations or at best
a few successive equilibrations have been carried out to measure maximal
exchange levels in zeolites. It is doubtful whether these results and the selectivity
32 R.P. Townsend · R. Harjula

plots derived from these maximum exchanges are correct. It is also obvious that
the inaccuracies in determining the maximum exchange contribute strongly to
the frequently observed high levels of scatter in zeolite selectivity data. For
instance, zeolites NaX and NaY appear to have very different selectivities for cal-
cium ions at room temperature, when the maximum exchange level of 68%,
determined by a single equilibration, is used for CaNaY [102]. However, when
the exchange is pushed to a higher level by using 8–12 successive equilibrations,
an 85% exchange level is obtained and the pattern of selectivity in X and Y for
Ca starts to appear very similar [72], as one would expect intuitively for the two
isomorphous zeolites.
Another obvious point, easily overlooked, is that in some cases impurities
in the salt solutions may cause the exchange to appear not to go to completion.
If at high loadings of the incoming ion A the selectivity coefficient leads to a
value of (say) 0.01 and the impurity level of B in A within the salt used is (say)
0.1%, it can easily be shown that no matter how many successive equilibrations
are carried out, only about 90% conversion to the A-form will be obtained. It
is therefore very important to use reagents of very high purity for the experi-
ments.

3.2.2
Zeolite Hydrolysis Effects

Hydrolysis of zeolites gives rise to a range of “impurity” species in both the solu-
tion and zeolite phases, which may interfere with the study of binary metal
cation exchange. Up to now, most zeolite ion exchange research has been carried
out using initially the sodium forms of the zeolites. In this form, zeolites hydro-
lyse by taking up hydronium ions from the water, viz.,
–––+ –––––
Na + 2H2O = H3O + + Na+ + OH – . (23)
The tendency to hydrolyse increases with an increasing aluminium content in
the zeolite [103]. An example of this is shown in Fig. 13, which shows the exten-
sive level of hydronium exchange which incidentally occurred during a series of
studies on Na/NH4 exchange equilibria in faujasitic zeolites [103]. This can also
be seen in the increase of the selectivity coefficients for H3O+/Na+ exchange with
the increasing aluminium content of faujasite zeolites [73–75]. Zeolite hydroly-
sis also leads to several secondary phenomena. First, since the zeolite imparts an
alkaline reaction to the water imbibed in the pores, carbon dioxide is picked up
from the air. When the zeolite is then immersed in water, carbonate and bicar-
bonate ions are released into the solution. Secondly, hydroxyl ions released into
the solution enhance the dissolution of silica and alumina from the zeolite
framework into the solution [73]. As a consequence, the following electroneu-
trality condition can be found to hold in pure water after it has been contacted
with a zeolite such as NaX [73]:
[H3O+] + [Na+] = [OH–] + [HCO3–] +2 [CO32–] + [Al(OH)–4]
+ [SiO(OH)–3] +2 [SiO2(OH)2–
2 ]. (24)
Ion Exchange in Molecular Sieves by Conventional Techniques 33

Fig. 13. Diagrammatic representation of F [F = (mNa + mNH4)/mAl] as a function of equivalent



fraction of sodium (ENa in four faujasitic zeolites with an Si/Al ratio which increased in the

order X < Y2 < Y3 < Y4. The fractional level of hydronium exchange at any composition ENa ,

ENH4 is given by (1–F). Taken from [100]
34 R.P. Townsend · R. Harjula

These reactions may have several effects on the binary metal cation exchange
process that is primarily under study. These may be summarised as:
1. mass-action effect of hydronium ion exchange on the binary metal cation
equilibrium;
2. association of metal cations with bicarbonate, carbonate, silicate and alumi-
nate ions in the solution; and
3. cation hydrolysis (i.e. association of metal cations with the OH– produced by
zeolite hydrolysis and/or precipitation of metal hydroxides).
Each of these effects will now be examined in more detail, together with a con-
sideration as to how they can be avoided or taken into account.

Hydronium Ion Exchange. Considerable quantities of hydronium ions can be


exchanged from water into the zeolite, when for example the zeolite is washed
after synthesis or even prior to ion exchange experiments. Thus, before the
experiments are begun, the zeolite is likely already to be partially exchanged into
the hydronium form. Preparing the zeolite in the pure sodium form may be dif-
ficult, since hydronium ions can be picked up even from concentrated salt solu-
tions of sodium [73]. When the zeolite is in the process of being converted to
another ionic form, further hydronium ion exchange can take place in one direc-
tion or the other [103]. For instance, NH4+/Na+ exchange in zeolites X and Y is
accompanied by significant hydronium exchange. Initially, almost 20% of the
exchange capacity of NaX (Si/Al = 1.26) was taken up by H3O+, and this amount
decreased steadily to about 12% upon conversion to the NH4+ form (Fig. 13). In
contrast, zeolite NaY (Si/Al = 2.47) contained no H3O+ initially, but conversion
to the NH4+ form was accompanied by H3O+ exchange so that in the NH4+ form,
about 7% of the exchange capacity was taken up by the hydronium ions [103].
The consequences of this can be profound. In the past, it was common prac-
tice for metal ion concentrations in the zeolite phase to be determined from the
changes in the corresponding concentrations in the solution phase. It is obvious
that this can lead to significant error if significant hydronium exchange also
takes place in parallel. However, even if the metal ion concentrations are mea-
sured in both phases, the calculated selectivity coefficient will not be that of the
pure binary metal exchange if concomitant hydronium exchange occurs. There-
fore, kA/B values will not reflect the relative preference of the zeolite framework
for the two metal cations. Because a three-component system is actually
involved, relative preferences between the metal cations and hydronium ions
would be intrinsic in the selectivity coefficient and it is doubtful if this selectiv-
ity coefficient could then be used for accurate prediction of the binary ion equi-
libria. For an accurate description of the binary and overall equilibria one may
be forced to use an appropriate ternary model (see Sect. 2.3).
In general, one can detect hydronium ion exchange by measuring the balance
of the contents of aluminium and exchangeable metal cations in the zeolite
[103]. If all the aluminium in the zeolite is present in the framework in tetrahe-
dral coordination, then the degree of hydronium exchange (DH) is given by
DH = 1 – Âi zic–i/c–Al  . (25)
Ion Exchange in Molecular Sieves by Conventional Techniques 35

However, a minor fraction of non-framework aluminium may be present in the


zeolite (e.g. in the cation exchange sites). Framework aluminium (tetrahedral)
and non-framework aluminium (octahedral) can be distinguished using 27Al
NMR although it is very difficult to be quantitative [104].
It is obvious that often the interfering effect of zeolite hydrolysis and hydro-
nium ion exchange cannot be avoided. In general, hydronium ion exchange is
favoured in solutions of low salt concentration so one can try to minimise it by
carrying out the experiments in moderate salt concentrations.
The acid/base nature of high alumina zeolites is in fact very similar to that of
weak-acid organic resins. In these materials metal ion uptake depends strongly
on the solution pH [105]. This can be seen for zeolites, too. For instance, uptake
of caesium and strontium by chabazite or sodium A zeolite depends strongly on
solution pH (pH 2–10) at a constant sodium background of 0.1 equiv dm–3
[106]. The effects of hydronium ion exchange and solution pH on metal cation
exchange in zeolites have been almost completely overlooked in past studies. As
a consequence, there may be considerable systematic error in many published
zeolite selectivity data, especially for high aluminium zeolites.

Ion Association in the Solution Phase. Ion exchange experiments are usually
carried out using anions (chloride, nitrate, perchlorate) that do not interact
appreciably with the metal cations under study. However, zeolite hydrolysis
produces many anionic species that tend to associate with metal ions (see Eq.
24). For instance, when pure water is contacted with zeolite NaX, between
5 ¥ 10–4 –5 ¥ 10–5 mol dm–3 soluble silica and alumina and between 1 ¥ 10–3–1 ¥
10–4 mol dm–3 of total carbonates can be found in solution [74], depending on
the value of V/m. The pH of the solution contacted with NaX can become mark-
edly alkaline (pH 11.5–9.5) so precipitation of metal hydroxides (or carbonates,
aluminosilicates) is also possible. Even if no precipitation of metals takes place,
ion association can have a marked effect on the observed selectivities of the zeo-
lites. For example, consider the exchange of a divalent metal ion M 2+ for sodium
in a zeolite. If it is assumed that the metal cation associates with a univalent
ligand L– to form the complex ion ML+, then this ion association can be charac-
terised by an association constant k1 of the form
[ML+]
kl = 06 . (26)
[M2+][L–]
Most analytical techniques employed in the measurements of metal ion concen-
trations in solution yield the total concentration of the metal [M]T , viz.,
[M]T = [M2+] + [ML+] . (27)
Assuming that only the free metal cations are exchanged into the zeolite, the
observed selectivity coefficient [kM/Na(obs)] that one obtains from the measure-
ment for the exchange is thus:
[M
6 2+][Na+]
kM/Na (obs) = 00 6+] (28)
[M] [Na T
36 R.P. Townsend · R. Harjula

but since [M]T = [M2+] (1 + k1[L–]), it follows that


kM/Na
kM/Na (obs) = 06 . (29)
1 + k1[L–]
It can be seen, therefore, that ion association can apparently decrease the selec-
tivity coefficient of the ion exchange reaction. This apparent decrease in selec-
tivity can take place when k1[L–]>1, that is when k1 is large (strong complexing
ligand), and/or when there is a large excess of ligand present in the solution. In
the selective exchange of multivalent metal ions the equilibrium concentrations
can be very low, well below the concentrations of potential complexing ligands
(carbonates, silicates, aluminates, hydroxyl ions), so that appreciable amounts of
free ligands can be present in the solution at equilibrium.
Carbonate, bicarbonate and silicate ions form moderately strong complexes
with most metal ions. For instance, the association constants of alkaline earth
cations for carbonate and bicarbonate are in the range of 10–1000 [107, 108].
Similarly, silicate ions complex readily with many metal ions, e.g. with calcium
(k1 = 1230) and magnesium (k1 = 1.5 ¥ 104) [108]. There seems to be no data on
metal ion association with aluminate ions but it is likely that ion association is
moderately strong here also.

Cation Hydrolysis. Many metal hydroxides have a low solubility in moderately


alkaline solutions arising from the hydrolysis of high alumina zeolites. For
instance, most transition metals and magnesium precipitate at pH 9–10 and at
this pH range carbonates are likely to precipitate other metals such as calcium,
strontium and cadmium [109]. Such precipitation phenomena can seriously dis-
tort the measurements of ion exchange selectivities. In addition, even when the
metal concentrations are below the limits of hydroxide precipitation, hydrolysed
species, such as MOH+, M(OH)2 (aq) and M(OH)–3 , often form the majority of the
metal species in solution. For the determination of the ion exchange selectivity
coefficient, the concentration of free, non-hydrolysed metal cation should be
known. If the concentrations of hydrolysed species are used, a too low value may
again be obtained for the selectivity coefficient.

3.2.3
Colloidal Solids in Suspension

Very fine colloidal particles in the nanometre size range may be left suspended
in solution when centrifugation or filtration is used for the phase separation
operation during the measurement of ion exchange equilibria in zeolites and
other inorganic materials [75]. Especially in the study of the ion exchange of
radionuclides, which can be present in very low concentrations, the presence of
colloidal particles carrying the metal cation under study can bring large errors
in the determination of discrete metal cation concentrations in solution.
Depending on the analytical technique used, metal ions associated with col-
loidal particles may be indistinguishable from free metal ions. For instance, in
the determination of the distribution coefficients of radionuclides (Eq. 15), large
errors may take place [75]. In cases where the metal ion concentration in the col-
Ion Exchange in Molecular Sieves by Conventional Techniques 37

loidally suspended zeolite is much larger than that of the free metal ions, the
measurement of KD (cm3 g–1) yields just the reciprocal of the concentration
(g cm–3) of suspended zeolite in the solution, instead of the ratio of metal con-
centrations in the zeolite and solution phases [75].
The presence of colloidal zeolite particles in solution may also apparently
decrease the selectivity of the zeolite for a given metal ion. This problem is
encountered especially when very low elemental concentrations, corresponding
to low degrees of loading, are measured by highly sensitive methods such as by
the use of radioactive tracers [75] or by atomic emission or absorption spec-
trophotometry with plasma or graphite furnace atomisation. Considering again
the exchange of a divalent metal cation M2+ for sodium as an example, the
observed selectivity coefficient kM/Na (obs) would be given by

[M3 2+][Na+]2
kM/Na(obs) = 0004
2+ 4 +]2 (30)
([M ]+[M]c)[Na
where [M]c is the amount of metal M in the suspended colloids per unit volume
of solution. Recent studies indicate that as much as 3 mg dm–3 of suspended zeo-
lite may be present in solution after centrifugation with a low-speed centrifuge
(G = 2000) [75].
The results of several studies in the past have shown very low selectivities and
“strange” selectivity gradients compared to more recent studies. For instance, in
the study of calcium, strontium and barium exchanges in zeolites NaX and NaY,
selectivities were low for these ions at low degrees of loading and then increased,
finally exhibiting maxima at high loadings [101, 110]. In these experiments ini-
tial metal concentrations were very low (and the corresponding equilibrium
concentrations even much lower) and cation loadings in the zeolite were
increased by increasing the initial metal ion concentrations.Appearance of max-
ima in the selectivity plots is difficult to rationalise in the absence of a phase
change (Sect. 2.1) since one would expect that the most selective cation sites
would be occupied first so that the selectivity would steadily decrease with
cation loading. This common pattern has been observed in other studies,
carried out in isonormal solutions at considerably higher concentrations
(0.1–1.2 g equiv–1) for most alkali and alkaline earth cations in NaX and NaY
[72, 102, 111]. It is therefore likely that the observed low selectivities at low met-
al loadings and very low solution concentrations for NaX and NaY are due to ion
association or suspended colloidal zeolite in the solution phase, since, at these
very low solution concentrations, most of the metal ions in the solution may
have been present as other species rather than as free metal cations. Similar
decreases of selectivity were observed for the calcium exchange in NaX and NaY
in dilute isonormal solutions (NT <0.1–0.001 equiv dm–3), but, in more concen-
trated solutions, the selectivity was independent of total ion concentration in
solution (NT = 0.1–1.2 g equiv dm–3) [72].
Theoretically, after a correction is made to the selectivity coefficients (Eq. 2)
for solution phase non-ideality (Eq. 7), the obtained corrected selectivity coeffi-
cients are then independent of the total ion concentration in the solution at any
given degree of ion loading in the zeolite, provided that no other parallel reac-
38 R.P. Townsend · R. Harjula

tions (salt imbibition, ion association) affecting the metal ion distribution
between the zeolite and solution phase take place. This means that one can in
principle check whether the selectivity coefficient is unaffected by the above fac-
tors by carrying out the determinations of selectivity in isonormal solutions of
different total ion concentrations. In general, these interfering phenomena are
strongest at low metal concentrations, when the metal ion concentrations
become comparable to or lower than the concentration of interfering species
(e.g. complexing ligands, suspended particles). In many cases these interfer-
ences can be avoided by using sufficiently high total ion concentrations in solu-
tion; however, in some cases, interferences may operate even in rather high con-
centrations (NT = 0.1–0.4 equiv dm–3). Additional interference in the determi-
nation of selectivity may be caused in concentrated solutions (NT >1 equiv dm–3)
by salt imbibition [61, 112].

4
Concluding Remarks
In this chapter an overview has been given of some recent developments in our
understanding of ion exchange in molecular sieves, with particular reference to
experimental implications and methods. Research activity has declined from
what was a quite high level in the 1960s and 1970s; the case for more fundamen-
tal studies on this topic seems compelling, especially the kinetic aspects.
Specifically, three final remarks are made. The first of these arises out of the
complexities one encounters when attempting to compare data from various
sources or when predicting exchange behaviour. The need for a simplification
and rationalisation of the diverse theoretical approaches and descriptions of
exchange behaviour is obvious; it is hoped that current attempts to achieve this
[113, 114] will continue and bear fruit.
Secondly, the relative neglect of detailed kinetic studies in recent years has
been noted. Although our attempts to describe theoretically the rates of ion
exchange (let alone predict them) remain simplistic (Sect. 2.4), nevertheless it
should be obvious to the reader that most applications involving ion exchange
processes in molecular sieves are likely to be kinetically controlled. There
remain major theoretical and computational challenges in this area which will
entail the utilisation of the current rapidly growing computational power
together with increasingly sophisticated models to throw light on exchange rate
processes, both at the atomistic and mesoscopic scales.
Finally, computational approaches are obviously only as good as the experi-
mental studies which underpin and validate them. Consequently, we hope that
the growing awareness of the experimental pitfalls and complexities which can
hinder the acquisition of reliable data [115] will encourage further fundamental
studies on ion exchange processes, not only in zeolites, but especially in the
aluminophosphate families of molecular sieves, where so much unexplored ter-
ritory remains.
Ion Exchange in Molecular Sieves by Conventional Techniques 39

References
1. Helfferich F (1962) Ion exchange. McGraw-Hill, London, p 12
2. Costa E, de Lucas A, Zarca J, Sanz FJ (1987) Lat Am J Chem Eng Appl Chem 17:135
3. Dyer A (1988) An introduction to zeolite molecular sieves, 1st edn. Wiley, Chichester,
pp 76, 80, 83
4. Schwuger MJ, Liphard M (1989) Fundamentals of phosphate substitution in detergents
by zeolites. In: Karge HG, Weitkamp J (eds) Zeolites as catalysts, sorbents and detergent
builders. Elsevier, Amsterdam. Stud Surf Sci Catal 46:673
5. Denkewicz RP Jr, Monino AG, Russ DE, Sherry HA (1995) J AOCS 72:11
6. Adams CJ, Araya A, Carr SW, Chapple AP, Franklin KR, Graham P, Minihan AR, Osinga
TJ, Stuart JA (1997) Stud Surf Sci 105:1667
7. Adams CJ, Araya A, Cunningham KJ, Franklin KR, White IF (1997) J Chem Soc Faraday
Trans 93:499
8. Sorlie AA, Bowerman BS, Czajkovski C, Dyer RS (1998) Low-level liquid radioactive waste
treatment at Murmansk, Russia: facility upgrade and expansion. Proceedings of the Sym-
posium on Waste Management at Tucson, Arizona, paper 56–07
9. Robinson SM, Arnold WD, Byers CH (1990) Design of fixed-bed ion exchange columns
for wastewater treatment. Proceedings of the Symposium on Waste Management at Tuc-
son, Arizona, vol 2, p 1635
10. Horsley DMC, Howden M (1990) Trans I Chem E 68(B):140
11. Cauthen BE, Taylor JC (1990) Liquid radwaste process optimisation at Catawba nuclear
station. Proceedings of the Symposium on Waste Management at Tucson, Arizona, vol 2,
p 305
12. Ekectukwu OE, Loucks LE (1992) Reduction of caesium and cobalt activity in liquid rad-
waste processing using clinoptilolite zeolite at Duke power company. Proceedings of the
Symposium on Waste Management at Tucson, Arizona, vol 2, p 1635
13. James KL, Miller CC (1992) The impact of ion exchange media and filters on LLW pro-
cessing. Proceedings of the Symposium on Waste Management at Tucson, Arizona, vol 2,
p 1575
14. Barrer RM, Rees LVC, Ward DJ (1964) Proc R Soc London Ser A 237:180
15. Ames LL (1964) Am Miner 49:1099
16. Wolf F, Furtig H (1965) Kolloid Z Z Polymer 206:48
17. Sherry HS, Walton HF (1967) J Phys Chem 71:1457
18. Barri SAI, Rees LVC (1980) J Chromatogr 201:21
19. Wiers BH, Grosse RJ, Cilley WA (1982) Environ Sci Technol 16:617
20. Franklin KR, Townsend RP (1985) J Chem Soc Faraday Trans 1 81:1071
21. Drummond D, De Jonge A, Rees LVC (1983) J Phys Chem 87:1967
22. Wilson ST (1991) Synthesis of AlPO4-based molecular sieves. In: Van Bekkum H, Flani-
gen EM, Jansen JC (eds) Introduction to zeolite science and practice. Elsevier, Amster-
dam, Stud Surf Sci Catal 58:137
23. Schoonheydt RA (1991) Clays from two to three dimensions. In: Van Bekkum H, Flani-
gen EM, Jansen JC (eds) Introduction to zeolite science and practice. Elsevier, Amster-
dam, Stud Surf Sci Catal 58:201
24. Smith JV (1989) Towards a comprehensive mathematical theory for the topology and
geometry of microporous materials. In: Jacobs PA, van Santen RA (eds) Zeolites: facts,
figures, future, part A. Elsevier, Amsterdam, Stud Surf Sci Catal 49:29
25. Smith JV (1988) Chem Rev 88:149
26. Van Koningsveld H (1991) Structural subunits in silicate and phosphate structures. In:
Van Bekkum H, Flanigen EM, Jansen JC (eds) Introduction to zeolite science and prac-
tice. Elsevier, Amsterdam, Stud Surf Sci Catal 58:35
27. Kokotailo GT, Fyfe CA, Kennedy GJ, Gobbi GC, Strobl H, Pasztor CT, Barlow GK, Bradley
S (1986) Zeolite structural investigations by high resolution solid state mas nmr. In
Murakami Y, Iijima A, Ward JW, New developments in zeolite science and technology.
Kodansha, Tokyo, Elsevier, Amsterdam, p 361
40 R.P. Townsend · R. Harjula

28. Barrer RM, Klinowski J (1972) J Chem Soc Faraday Trans 1 68:73
29. Engelhardt G (1991) Solid state nmr spectroscopy applied to zeolites. In: Van Bekkum H,
Flanigen EM, Jansen JC (eds) Introduction to zeolite science and practice. Elsevier,
Amsterdam, Stud Surf Sci Catal 58:285
30. De Kock FP, van Deventer (1995) Chem Eng Comm 135:21
31. Barrer RM, Klinowski J (1977) Phil Trans Roy Soc 285:637
32. Smolders E, van Dun JJ, Mortier WJ (1991) J Phys Chem 95:9908
33. Woolery GL, Alemany LB, Dessau RM, Chester AW (1986) Zeolites 6:14
34. Dessau RM, Schmitt KD, Kerr GT, Woolery GL, Alemany LB (1987) J Catal 104 :
484
35. Handreck GP, Smith TD (1989) J Chem Soc Faraday Trans 1 85:645
36. Chester AW, Chu YF, Dessau RM, Kerr GT, Kresge CT (1985) J Chem Soc Chem Commun
1985:289
37. Chu P, Dwyer FG (1983) Zeolites 3:72
38. Matthews DP, Rees LVC (1986) Chem Age India 37:353
39. Chu P, Dwyer FG (1988) Zeolites 8:423
40. McAleer AM, Rees LVC, Nowak AK (1991) Zeolites 11:329
41. Watling TC, Rees LVC (1994) Zeolites 14:687
42. Watling TC, Rees LVC (1994) Zeolites 14:693
43. Recommendations on ion exchange nomenclature (1972) Pure Appl Chem 29:619
44. McGlashan ML (1979) Chemical thermodynamics, 1st edn.Academic Press Inc., London,
p 111
45. Townsend RP (1991) Ion exchange in zeolites. In: Van Bekkum H, Flanigen EM, Jansen
JC (eds) Introduction to zeolite science and practice. Elsevier, Amsterdam, Stud Surf
Sci. Catal 58:359. See also the revision of this paper by Townsend RP, Coker EN, in
print
46. Barrer RM, Townsend RP (1976) J Chem Soc Faraday Trans 1 72:2650
47. Barrer RM, Townsend RP (1978) J Chem Soc Faraday Trans 1 74:745
48. Barrer RM, Munday BM (1971) J Chem Soc A 2909
49. Meier WM, Olson DH (1992) Atlas of zeolite structure types, 3rd edn. Butterworth-
Heinemann, London
50. Barrer RM, Munday BM (1971) J Chem Soc A 2914
51. Olson DH, Sherry HS (1968) J Phys Chem 72:4095
52. Hansen S, Häkansson U, Landa-Canovas AR, Fälth L (1993) Zeolites 13:276
53. Taylor AM, Roy R (1964) Am Miner 49:656
54. Fletcher P, Townsend RP (1981) J Chem Soc Faraday Trans 2 77:2077
55. Fletcher P, Townsend RP (1983) J Chem Soc Faraday Trans 2 79:419
56. Townsend RP (1986) Pure Appl Chem 58:1359
57. Barrer RM, Townsend RP (1984) J Chem Soc Faraday Trans 2 80:629
58. Gaines GL, Thomas HC (1953) J Chem Phys 21:714
59. Barrer RM, Townsend RP (1985) Zeolites 5:287
60. Barrer RM, Klinowski J (1974) J Chem Soc Faraday Trans 1 70:2080
61. Barrer RM, Walker AJ (1964) Trans Faraday Soc 60:171
62. Townsend RP, Fletcher P, Loizidou M (1984) Studies on the prediction of multi-
component ion-exchange equilibria in natural and synthetic zeolites. In: Olson D, Bisio
A (eds) Proceedings of the Sixth International Zeolite Conference. Butterworths, UK,
p 110
63. O’Connor JE, Townsend RP (1985) Zeolites 5:158
64. Loizidou M, Townsend RP (1987) Zeolites 7:153
65. Loizidou M, Townsend RP (1987) J Chem Soc Dalton Trans 1911
66. Fletcher P, Townsend RP (1985) J Chem Soc Faraday Trans 1 81:1731
67. Franklin KR, Townsend RP (1988) J Chem Soc Faraday Trans 1 84:687
68. Pabalan RT (1994) Geochim Cosmochim Acta 58:4573
69. Franklin KR, Townsend RP (1985) J Chem Soc Faraday Trans 1 81:3127
70. Franklin KR, Townsend RP (1988) J Chem Soc Faraday Trans 1 84:2755
Ion Exchange in Molecular Sieves by Conventional Techniques 41

71. Harjula R, Lehto J, Pothuis JH, Dyer A, Townsend RP (1991) Hydrolysis and trace Ca2+
exchange in zeolites NaX and NaY. In: Proceedings International Conference Ion
Exchange, ICIE ‘91. Kodansha, Tokyo, p 157
72. Harjula R, Dyer A, Pearson SD, Townsend RP (1992) J Chem Soc Faraday Trans 88:
1591
73. Harjula R, Lehto J, Pothuis JH, Dyer A, Townsend RP (1993) J Chem Soc Faraday Trans
89:971
74. Harjula R, Dyer A, Townsend RP (1993) J Chem Soc Faraday Trans 89:977
75. Harjula R, Lehto J, Pothuis JH, Dyer A, Townsend RP (1993) J Chem Soc Faraday Trans
89:1877
76. Franklin KR, Townsend RP, Whelan SJ, Adams CJ (1986) Ternary exchange equilibria
involving H3O+, NH +4 and Na+ ions in synthetic zeolites of the faujasite structure. In:
Murakami Y, Iijima A, Ward JW (eds) New developments in zeolite science and technol-
ogy. Kodansha, Tokyo, Elsevier, Amsterdam, p 289
77. Fletcher P, Townsend RP (1981) J Chem Soc Faraday Trans 2 77:965
78. Bajpai RK, Gupta AK, Gopala-Rao M (1973) J Phys Chem 77:1288
79. Brignal WJ, Gupta AK, Streat M (1976) Theory and practice in ion exchange, Soc Chem
Ind, London, paper 11
80. Franklin KR, Townsend RP (1988) Zeolites 8:367
81. Zuyi T, Gengliang Y (1995) React Funct Polymers 27:117
82. Fletcher P, Franklin KR, Townsend RP (1984) Phil Trans R Soc London A 312:141
83. Gopala Rao M (1995) Sep Sci Technol 30:1385
84. Elprince AM, Babcock KL (1975) Soil Sci 120:332
85. Smith RP, Woodburn ET (1978) AIChEJ 24:577
86. Shallcross DC, Hermann CC, McCoy BJ (1988) Chem Eng Sci 43:279
87. Perona JJ (1993) AIChEJ 39:1716
88. Robinson SM, Arnold DW, Byers CH (1991) ACS Symp Ser 468:133
89. Robinson SM, Arnold WD, Byers CH (1994) AlChEJ 40:2045
90. Ruthven DM (1994) Principles of adsorption and adsorption processes. Wiley, New
York
91. Brown LM, Sherry HS, Krambeck FJ (1971) J Phys Chem 75:3846
92. Brown LM, Sherry HS (1971) J Phys Chem 75:3855
93. Dyer A, Townsend RP (1973) J Inorg Nucl Chem 35:3001
94. Duffy SC, Rees LVC (1974) J Chromatogr 102:149
95. Brooke NM, Rees LVC (1968) Trans Faraday Soc 64:3383
96. Barrer RM, Rees LVC (1964) J Phys Chem Solids 25:1035
97. Hiester NK, Vermuelen T, Klein G (1963) Adsorption and ion exchange. In: Perry JH,
Chilton CH (eds) Chemical engineer’s handbook. McGraw-Hill, New York, p 16-1
98. Klein G (1985) AIChE Symp Ser No 242, 81:28
99. Collins ED, Campbell DO, King LJ, Knauer JB, Wallace RM (1985) Evaluation of zeolite
mixtures for decontaminating high-activity-level water at the Three Mile Island Unit 2
nuclear power station. Technical Document 337, International Atomic Energy Agency,
Vienna, p 43
100. Barrer RM, Klinowski J, Sherry HS (1973) J Chem Soc Faraday Trans 2 69:1669
101. Barrer RM, Davies JA, Rees LVC (1968) J Inorg Nucl Chem 30:3333
102. Sherry HS (1968) J Phys Chem 72:4086
103. Townsend RP, Franklin KR, O’Connor JF (1984) Adsorption Sci Technol 1:269
104. Engelhardt G, Michel D (1987) High resolution solid-state nmr of silicates and zeolites.
Wiley, Chichester, p 213
105. Helfferich F (1962) Ion exchange. McGraw-Hill, London, p 16
106. Mimura H, Kanno T (1987) J Nucl Sci Technol 22:284
107. Martell AE, Smith RM (1982) Critical stability constants, vol 5. Plenum Press, New
York
108. Hogfeldt E (1982) Stability of metal-ion complexes. IUPAC chemical data series no 21.
Pergamon Press, Oxford
42 R.P. Townsend · R. Harjula: Ion Exchange in Molecular Sieves

109. Baes CF, Mesmer RE (1976) Hydrolysis of cations. Wiley, New York
110. Barrer RM, Rees LVC, Shamsuzzoha M (1966) J Inorg Nucl Chem 28:629
111. Sherry HS (1966) J Phys Chem 70:1158
112. Lowe BM, Pope CG (1989) J Chem Soc Faraday Trans 1 85:945
113. Harjula R, Lehto J (1997) Harmonisation of ion exchange formulations and nomen-
clature: what could be done? In: Dyer A, Hudson MJ, Williams PA (eds) Royal Society
of Chemistry Spec. Pub. No. 196
114. Soldatov VS (1995) React Funct Polymers 27:95
115. Lehto J, Harjula R (1995) React Funct Polymers 27:121

You might also like