The hydrothermal synthesis of zeolites-Precursors, intermediates and reation mechanism

bond-The process of hydrothermal zeolite synthesis can be most adequately explained by a mechanism based upon the ation model, whether or not there is a visible liquid phase.. the induct

The hydrothermal synthesis of zeolites: Precursors,

intermediates and reaction mechanism

a Centre for Microporous Materials, School of Chemistry, University of Manchester, P.O Box 88, Sackville Street,

Manchester M60 1QD, United Kingdom

b School of Pharmacy and Biomedical Sciences, University of Portsmouth, St Michael’s Building, White Swan Road,

Portsmouth PO12DT, United Kingdom Received 5November 2004; accepted 11 February 2005

Available online 8 April 2005

Abstract

An account is presented of the mechanistic aspects of hydrothermal zeolite synthesis The introduction provides a historical andexperimental perspective and is followed by a summary of proposed mechanisms and associated modelling studies The central sec-tion of the review contains a description of the most probable mechanistic pathways in zeolite formation In this, the reaction stages

of the induction period, nucleation and crystal growth are examined in chronological sequence Finally, particular aspects of thesynthesis process such as the constitution of growth species, template–framework interactions and the nature of zeolite solubilityare treated in more detail

Emphasis is placed upon the chemical basis of zeolite synthesis Fundamental to this are the TAOAT making and breaking reactions which establish the equilibration between solid and solution components The consequent generation of order,driven by energy differences and strongly moderated by kinetic limitations, is essentially one of continuous evolution However, thediscreet step of nucleation provides a discontinuity in which isolated regions of local order are superceded by the establishment of aperiodic crystal lattice, capable of propagation Crystal growth occurs through an in-situ, localised construction process from small,mobile species ordered by the participating cations

bond-The process of hydrothermal zeolite synthesis can be most adequately explained by a mechanism based upon the ation model, whether or not there is a visible liquid phase The common presence of mobile species emphasises the overall similarity

solution–medi-of zeolite synthesis reactions so that the need to distinguish any separate ‘‘gel rearrangement’’ or ‘‘solid-phase transformation’’mechanism becomes unnecessary

2005 Elsevier Inc All rights reserved

Keywords: Hydrothermal synthesis; Nucleation; Crystal growth; Modelling; Mechanism

Contents

Part I: Background 4

1 Introduction 41.1 History of hydrothermal zeolite synthesis 4

1387-1811/$ - see front matter
2005Elsevier Inc All rights reserved.

doi:10.1016/j.micromeso.2005.02.016

* Corresponding author Tel.: +44 161 200 4512; fax: +44 161 200 4559.

E-mail address: colin.cundy@manchester.ac.uk (C.S Cundy).

www.elsevier.com/locate/micromeso

1.2 Scope and structure of this review 4

2 Experimental observations 5

3 Summary of proposed mechanisms 6

3.1 Richard Barrer 6

3.2 The early work of Breck and Flanigen 6

3.3 Kerr
s recirculation experiment and the work of Ciric 6

3.4 Studies at Leningrad 8

3.5 Overview—1959 to 1971 and beyond 8

3.6 Introduction of organic templates 8

3.7 Chang and Bell and after 9

4 Modelling the processes of zeolite synthesis 10

4.1 Mathematical models of synthesis reactions 10

4.2 Molecular modelling 11

4.2.1 Modelling zeolite-template pairs 11

4.2.2 Cluster calculations 11

Part II: Synthesis mechanism 12

5 The induction period 12

6 The evolution of order 12

6.1 The nature of the amorphous material 12

6.2 Primary and secondary amorphous phases 12

6.2.1 The Montpellier study 14

6.2.2 Related investigations 14

6.3 Further evidence for pre-crystalline order from synthesis studies 14

6.4 Summary 15

7 Nucleation 15

7.1 Introduction 15

7.2 General considerations 16

7.3 Determination of zeolite nucleation patterns from measurements on the resulting crystals 16

7.3.1 Studies under isothermal conditions 17

7.3.2 Ageing studies 17

7.4 The use of seed crystals 17

7.5 Autocatalytic nucleation 19

7.6 Nucleation in zeolite systems—The nature of the reaction sol 20

7.7 Nucleation in zeolite systems—Homogeneous or heterogeneous? 20

7.8 Nucleation in zeolite systems—Mechanism 22

7.9 Summary 24

8 Crystal growth 24

8.1 Experimental methods 24

8.2 Experimental observations—Introduction 25

8.3 Experimental observations—Studies of macrocrystalline systems 26

8.4 Experimental observations—Studies of nanocrystalline systems 27

8.5 Size-dependent growth of nanocrystals 28

8.6 Growth models 29

8.7 Mechanism 31

8.8 Summary 31

Part III: Key topics 32

9 The nature of growth species and the role of aggregation processes 33

9.1 Growth from soluble, pre-fabricated units 33

9.2 Growth from simple species 35

9.3 Mineralising agents other than hydroxide 35

9.4 Growth from particles 36

9.4.1 Particle aggregation 36

9.4.2 Chemical and physical consequences of an aggregation mechanism 37

9.5 Summary 38

10 Solid state transformations 38

10.1 Hydrothermal synthesis in the presence of a liquid phase 39

10.2 Hydrothermal synthesis in the apparent absence of a liquid phase 40

10.3 Non-aqueous syntheses 41

10.4 The role of water in apparent solid state transformations 41

10.5 Solid state transformations at high temperatures and pressures 42

10.6 Summary and conclusions 42

11 Ageing effects in zeolite synthesis 43

11.1 Ageing as a means to control product phase purity and crystal size 43

11.2 Rationalisation of ageing effects 44

11.3 Detailed analyses of ageing-related effects in silicalite synthesis 44

11.4 Summary 46

12 X-ray amorphous zeolites 47

12.1 XRD evidence for ‘‘X-ray amorphous zeolites’’ 47

12.2 IR evidence 47

12.3 Evidence from other physical measurements 48

12.4 Evidence from catalysis 48

12.5 Evidence from the synthesis process 48

12.6 Other amorphous materials related to zeolites—zeolite degradation 49

12.7 Conclusions 49

13 Template–framework interactions 50

13.1 Geometric matching 50

13.2 Template classification and versatility 50

13.3 Structure blocking 51

13.4 Variations induced by heteroatoms 51

13.5 Conclusions 52

14 Solubility and supersaturation 52

14.1 Zeolite solubility 52

14.2 Zeolite solubility as a function of base concentration 53

14.3 Supersaturation in relation to zeolite crystal growth 53

14.4 Thermodynamic vs kinetic factors in zeolite synthesis 54

14.5 Summary 54

15 Zeolite dissolution 54

15.1 The kinetics of dissolution 55

15.2 Morphological and compositional changes on dissolution 56

15.3 Summary 57

16 Metastability 57

16.1 Precursors, intermediates and co-products 57

16.2 Layer structures as transients and precursors 58

16.3 Conversion of one zeolite into another 59

16.4 Ostwald ripening 60

16.5 Summary 61

17 Optimisation of zeolite syntheses 61

17.1 Comparisons of zeolite synthesis reaction rates 61

17.2 Procedures for improvement 63

17.2.1 Reaction optimisation (and its limitations) 63

17.2.2 Addition of seed crystals 63

17.2.3 Additives and ‘‘promoters’’ 63

17.2.4 Microwave synthesis 63

17.3 Summary 64

18 Relationship of zeolite synthesis mechanism to that of other porous materials 64

18.1 Zeolites and clathrate hydrates 64

18.2 Zeotypes 65

18.3 Microporous vs mesoporous structures 66

18.4 Summary 67

19 Summary and conclusions 67

Acknowledgements 69

Appendix A A chemical model for the crystal growth of zeolite molecular sieves 69

References 70

Part I: Background

1 Introduction

1.1 History of hydrothermal zeolite synthesis

The history of man-made zeolites can be traced back

to the claimed laboratory preparation of levynite by St

Claire Deville in 1862[1] However, zeolite synthesis as

we know it today had its origins in the work of Richard

Barrer and Robert Milton, commencing in the late

1940s Barrer (principally at Imperial College, London)

began his work by investigating the conversion of

known mineral phases under the action of strong salt

solutions at fairly high temperatures (
170–270 C)

Among the products, species P and Q [2–4]

(isostruc-tural variants) displayed unique characteristics and

rep-resented the first synthetic zeolite unknown as a natural

mineral These materials were later found to have the

KFI structure [5] determined subsequently for zeolite

the Union Carbide Corporation, Tonawanda, New

York) pioneered the use of more reactive starting

mate-rials (freshly precipitated aluminosilicate gels), enabling

reactions to be carried out under milder conditions and

leading to the discovery of zeolites A [8]and X[9] By

1953, Milton and his colleagues had synthesised 20

zeo-lites, including 14 unknown as natural minerals[10]

Following the foundations laid in the 1950s, the next

decade saw many significant developments Earlier work

on zeolite synthesis had utilised only inorganic reaction

components but in 1961 the range of reactants was

ex-panded to include quaternary ammonium cations [11–

13] The introduction of organic constituents was to

have a major impact upon zeolite synthesis and the

key step followed quite rapidly with the disclosure in

1967 of the first high-silica phase, zeolite beta[14], whilst

the archetypal high-silica zeolite, ZSM-5, was

discov-ered in 1972[15]

There has subsequently been a large rise in the

num-ber of known synthetic zeolites[16]and also the

discov-ery of new families of zeolite-like or zeolite-related

materials[17] The latter ‘‘zeotypes’’ may be represented

by the microporous alumino- and gallo-phosphates

(Al-POs and Ga(Al-POs) [18–20] and titanosilicates (such as

ETS-10)[21–23] Such materials display great

composi-tional diversity and frequently have frameworks

un-known for zeolites This increased structural flexibility

has its origins in the available spectrum of heteroelement

atomic radii, bond lengths and bond angles, and in the

emergence of coordination numbers greater than four

Even greater divergence from the norm of microporous

aluminosilicates is seen in a major new class of

zeolite-related phases discovered in the early 1990s

Mesopor-ous materials, synthesised with the aid of surfactant

molecules and typified by the M41S [24,25] and SBA

[26,27] series, have periodic structures with far largerpore sizes (up to
200 A˚ ) but are not conventionallycrystalline [28–30]

Investigative work aimed at gaining an understanding

of the synthesis process has its origins in the 1960s.These studies have continued up to the present day,spurred on at various points by discoveries of new mate-rials, advances in synthetic techniques, innovations intheoretical modelling methods and, especially, by thedevelopment of new techniques for the investigation ofreaction mechanisms and the characterisation of prod-ucts It is the purpose of the present Review to offer,for the case of zeolites, an account of such exploratoryand background work

1.2 Scope and structure of this review

In an earlier survey[31], a summary was given of themain discoveries and advances in thinking in the field ofzeolite synthesis from the 1940s up to 2002 That ac-count was principally concerned with the pattern ofdiscovery and the consequent progression of ideas.Discussion of the mechanism of zeolite synthesis waslimited to this evolutionary context This present reviewattempts to expand this critical argument and to de-scribe in detail the most probable steps by which amor-phous aluminosilicate reagents are converted tocrystalline molecular sieves In addition to summarisingearlier proposals, it will be necessary to bring forwardsome further ideas which are perhaps new in the currentcontext This should clarify the link between nucleationand crystal growth by considering the chemical stepswhich are common to both Fortunately, when this isdone the picture becomes simpler rather than more com-plex and the need to make any differentiation between,for example, ‘‘solution-mediated growth’’ and ‘‘gel rear-rangement’’ finally disappears The text is confined tohydrothermal methods of synthesis and concentrates

on aluminosilicate zeolites, mentioning alternative types or other porous materials only when this is neces-sary to illustrate or broaden the main argument.However, it seems very probable that the main featuresobserved for zeolites will also be found in the synthesis

zeo-of closely related materials, modified in some casesthrough the differences in composition, structure, polar-ity and solution chemistry

The sections of this survey fall into three groups InPart I and following the above brief historical introduc-tion, the experimental observations associated with atypical hydrothermal zeolite synthesis are outlined andthe various interpretations which have been advanced

to explain them summarised (Sections 1–3) Section 4 views work on modelling the processes of zeolite synthe-sis Part II represents an attempt to put forward adetailed and self-consistent view of the most probablemechanistic pathways in zeolite formation, presented

re-within the overall chronology of the hydrothermal

syn-thesis reaction, i.e the induction period and the nature

of the amorphous material (Sections 5and 6), the

mech-anism by which zeolite crystals are nucleated (Section 7)

and the mechanism of zeolite crystal growth (Section 8)

In Part III (Sections 9–18), some key problems and

questions associated with zeolite synthesis are addressed

in more detail, leading, finally (Section 19), to some

overall conclusions

For further general information on the subject of

zeo-lite synthesis, the reader is referred to the standard

text-books[17,32–35]and recent reviews[31,36–43]

2 Experimental observations

A typical hydrothermal zeolite synthesis can be

de-scribed in briefest terms as follows:

1 Amorphous reactants containing silica and alumina

are mixed together with a cation source, usually in

a basic (high pH) medium

2 The aqueous reaction mixture is heated, often (for

reaction temperatures above 100C) in a sealed

autoclave

3 For some time after raising to synthesis temperature,

the reactants remain amorphous

4 After the above ‘‘induction period’’, crystalline

zeo-lite product can be detected

5 Gradually, essentially all amorphous material is

replaced by an approximately equal mass of zeolite

crystals (which are recovered by filtration, washing

and drying)

This is illustrated schematically in Fig 1 The

ele-ments (Si, Al) which will make up the microporous

framework are imported in an oxide form These oxidicand usually amorphous precursors contain SiAO andAlAO bonds During the hydrothermal reaction in thepresence of a ‘‘mineralising’’ agent (most commonly analkali metal hydroxide), the crystalline zeolite product(e.g zeolite A) containing SiAOAAl linkages is created.Since the bond type of the product is very similar to thatpresent in the precursor oxides, no great enthalpychange would be anticipated In fact, the overall free en-ergy change for a zeolite synthesis reaction is usuallyquite small, so that the outcome is most frequentlykinetically controlled[31,43–46]

Kinetic control is a pervading influence throughoutzeolite synthesis, where the desired product is frequentlymetastable Much of the know-how in this industriallyimportant area centres around choice of the exact condi-tions for product optimisation, so that the requiredmaterial can be prepared reproducibly and to the samespecification [47] Such considerations will often influ-ence the choice of starting reagents Whilst these may in-clude the simple oxides or hydroxides mentioned above(e.g precipitated silica or alumina trihydrate), it is alsovery common for the reagents to represent some degree

of pre-combination, as for example in sodium silicatesolution or solid sodium aluminate These materialsmay represent advantages in cost or ease of processingbut may also offer optimum routes to particular materi-als, since flexibility in the choice of reagents enablesequilibria to be approached from different directions.This may offer kinetic benefits, such as the preferrednucleation of one phase over another in situations wheremixtures may otherwise co-crystallise A good generalappreciation of the field of practical zeolite synthesiscan be obtained from the handbook issued by the Syn-thesis Commission of the International Zeolite Associa-tion (IZA)[48]

Fig 1 Hydrothermal zeolite synthesis The starting materials (Si AO and AlAO bonds) are converted by an aqueous mineralising medium (OH  and/or F  ) into the crystalline product (Si AOAAl bonds) whose microporosity is defined by the crystal structure.

3 Summaryof proposed mechanisms

The brief summary given above (Section 2) outlines

the transformation of an amorphous, aqueous

aluminosilicate gel under the action of heat into a

crys-talline zeolite product In the present section, an

over-view is presented of the main suggestions which have

been put forward to explain these experimental

observa-tions The historical aspects of this topic have been

dis-cussed more fully in our earlier account[31]and some of

the main mechanistic proposals are summarised in

subject of Part II (Sections 5–8)

3.1 Richard Barrer

The first consideration of synthesis mechanism was

that given by Barrer, Baynham, Bultitude and Meier

in 1959 The discussion section of a paper[49]in which

a wide variety of alumino-, gallo- and germano-silicates

were synthesised begins as follows:

‘‘The formation of diverse kinds of structural

frame-work leads to questions as to the mechanism of growth

The phases are often obtained reproducibly in yields

nearing 100% and the free-energy balance between the

many possible aluminosilicate nuclei must be delicate

The development of elaborate and continued space

pat-terns by progressive additions of single (Al,Si)O4

tetrahe-dra is difficult to imagine, particularly in the case of very

open zeolite structures The formation of these

frame-works is, however, much more easily visualised if in the

aqueous crystallising magma there are secondary

build-ing units in the form of rbuild-ings of tetrahedra or polyhedra

These may pack in various simple coordinations to yield

different aluminosilicates.’’

Examples of some possible ions were then tabulated:

rings of 3–6 tetrahedra, the double-4-ring, the

double-6-ring (or 3 4-double-6-rings) It was further pointed out that such

units could give rise to more complex ones, such as the

linking of six 4-ring anions to give the nosean-sodalite

cubo-octahedral unit Barrer returned to this theme in

a later review, considering that the growth of

alumino-silicate crystals from alkaline media was unlikely to

pro-ceed by the capture of single monomeric silicate and

aluminate tetrahedral ions TOn4 since ‘‘in the elaborate

porous crystalline structures of the zeolites, for instance,

it would seem difficult for the lattice to persist in its very

open pattern when rapidly adding such small units’’[50]

He felt that ‘‘a plausible process would be the accretion

in simple coordination of polygonal or polyhedral

an-ions by condensation polymerisation’’, giving as

exam-ples the 4-ring, 6-ring, cube and hexagonal prism and

the formation of the ‘‘crankshaft’’ double chain (found

in feldspars and the phillipsite-harmotome zeolites) by

linkage of 4-rings

3.2 The early work of Breck and Flanigen

In 1960, Flanigen and Breck reported[51,52]a study

in which XRD measurements were employed to followthe crystallisation with time of zeolite Na-A (at

100C) and Na-X (at 50 and 100 C) They showedthe now-familiar S-shaped growth curves and described

an induction period followed by a sudden rapid growth.The morphological changes observed [53] were inter-preted as a successive ordering of the gel as crystallisa-tion proceeds, leading to a conclusion that crystalgrowth takes place predominantly in the solid phase.Their conclusions may be summarised as follows

by Barrer et al.[49]

3 During the induction period, the nuclei develop to acritical size and then grow rapidly to small and uni-form sized crystals

4 Growth of the crystal proceeds through a type ofpolymerisation and depolymerisation process (break-ing and remaking Si,AlAOASi,Al bonds), catalysed

by excess hydroxyl ion and involving both the solidand liquid phases (although the solid phase appears

to play the predominant role)

In a subsequent review [54], Breck described zeoliteformation in the following terms: the gel structure isdepolymerised by hydroxide ions; rearrangement ofthe aluminosilicate and silicate anions present in thehydrous gel is brought about by the hydrated cationspecies present; tetrahedra re-group about hydrated so-dium ions to form the basic polyhedral units (24-hedra);these then link to form the massive, ordered crystalstructure of the zeolite

3.3 Kerr’s recirculation experiment and the work of Ciric

A paper published by George Kerr in 1966 describes

[55]an experiment carried out to test the hypothesis that

[56]‘‘a zeolite could be formed via dissolution of gel bysodium hydroxide solvent followed by deposition of zeo-lite crystals from gel-derived species in solution.’’ In theexperiment, a sodium hydroxide solution at 100C wascirculated through two filters, the first of which con-tained a specially prepared amorphous sodium alumino-silicate, whilst the second held crystals of zeolite Na-A.When the experiment was terminated after about 4 h,nearly all of the amorphous solid had been dissolvedand the zeolite sample (estimated to be essentially100% zeolite A by water sorption) had approximately

Summary of principal proposals for zeolite synthesis mechanism, 1959–2004

phases

Condensation polymerisation

of polygonal and polyhedral anions

M+-assisted arrangement of anions):

crystal growth mainly in the solid phase

Kerr [55,56] Na-A Crystal growth from solution species Amorphous solid !fastsoluble speciesðSÞ

ðSÞ þ nucleiðor zeolite crystalsÞ !slowzeolite A

nuclei from condensation reactions, crystal growth from solution

Accumulation of zeolite crystals Formation of nuclei Derouane, Detremmerie,

Gabelica and

Blom [58–62]

Na,TPA-ZSM-5Synthesis ‘‘A’’: liquid phase ion

transportation Synthesis ‘‘B’’:

solid hydrogel phase transformation

ordered into nuclei through OH-mediated

Si AOASi cleavage/recombination Burkett and Davis [64–66] TPA-Si-ZSM-5Pre-organised inorganic–organic composites,

nucleation through aggregation, crystal growth layer-by-layer

! ·12 ! ‘‘nanoslabs’’, growth by aggregation

doubled in mass From these (and other[55])

observa-tions, the mechanism was perceived to be that of rapid

dissolution of the amorphous solid to yield soluble

spe-cies The rate-determining step was then the

combina-tion of these soluble species with nuclei or zeolite

crystals to yield the zeolitic product

Working for the same company (Mobil) but not in

the same laboratories, Julius Ciric presented in 1968

the most detailed study of zeolite synthesis published

at that date [74] Kinetic curves were determined from

water sorption and chemical analyses were carried out

on reaction filtrates In addition, data were obtained

by particle counter, optical microscopy and BET surface

area methods The work adds much to the ideas set out

in the Kerr report[55], pointing to a solution-mediated

growth mechanism modified by the presence of the gel

phase (so that transport of growth species to crystals

embedded in gel is restricted by diffusion through the

gel) In addition, Ciric pointed out that his kinetic

re-sults were consistent with Barrer
s ideas of anionic

blocks [49,50] as well as with the Flanigen–Breck view

[51]on the catalysis by OHions

3.4 Studies at Leningrad

Some striking advances in thinking and technique

were reported by Zhdanov at the Second International

Zeolite Conference in 1970[57] Measurements on

crys-tal linear growth rates for zeolite A showed directly for

the first time the effect of temperature in increasing

growth rate and that the crystals grew at a near-constant

rate over the majority of the synthesis period From this

latter observation and the product crystal size

distribu-tion, Zhdanov was also able to deduce the nucleation

rate profile over the course of the reaction These

con-siderations, together with measurements of chemical

changes in the solution phase of the reaction mixture

and detailed consideration of such phenomena as the

induction period and seeding effects, led to a more

chem-ically detailed picture of zeolite crystallisation In this

view, the solid and liquid phases are connected by the

solubility equilibrium Condensation reactions give rise

to ‘‘primary aluminosilicate blocks (4- and 6-membered

rings)’’ and crystal nuclei Crystal growth occurs from

solution until dissolution of the amorphous phase is

complete Analytical data supported the proposition

that the composition of the crystals depended on that

of the liquid phase from which they crystallised

3.5 Overview—1959 to 1971 and beyond

It may be useful at this point to summarise the main

opinions expressed up to the year 1971 Barrer had

con-cluded that zeolite crystallisation was a

solution-medi-ated process, the structure being formed by the

condensation polymerisation of anionic polygonal or

polyhedral building units Similar precursor units wereenvisaged by Flanigen and Breck, although their think-ing was focused largely upon the solid phase In thisview, the initial, random aluminosilicate gel structurewas dis-assembled into its constituent tetrahedra bythe action of OHions and new, oligomeric polyhedralunits were formed through the ordering influence of thecations Crystal growth proceeded by an OH-catalysedpolymerisation and depolymerisation process, involvingpredominantly the solid phase but with some contribu-tion from solution species The studies of Zhdanovand of Kerr provided a more solution-oriented perspec-tive The original amorphous gel was seen as a dynamicentity, in equilibrium (or coming to equilibrium) withthe liquid phase Dissolving under the action of heatand base, the gel released active soluble species intothe solution from which nuclei formed and grew, fromsolution, into crystals, although the detailed nature ofthe migrating units was unspecified

During the later 1970s, the significance of the solutionphase in zeolite synthesis was to become increasinglyapparent, as demonstrated by two Raman spectroscopicstudies on the formation of zeolite A Observing nochanges with time other than the appearance of crystal-line product, McNicol et al concluded that crystallisa-tion occurred within the solid phase of the gel [75,76].However, by using a combination of chemical analyses,Raman spectroscopy, XRD, sorption and particle sizemeasurements, Angell and Flank[77]reached the oppo-site conclusion They demonstrated that the mechanisminvolved formation and subsequent dissolution of anamorphous aluminosilicate intermediate, with solutiontransport from the gel to the growth surface of the crys-tallite This view was reinforced by two further syntheticstudies Culfaz and Sand examined crystallisation ratesfor mordenite, zeolite X and zeolite A[78] From consid-erations of rate limitations by diffusion and seed crystalsurface area, they deduced that crystal growth in thesecases occurred from solution Kacirek and Lechert [79]

used detailed kinetic studies on seeded faujasite ses to develop further the solution growth model, con-cluding that the rate-determining step was theconnection of silicate species to the surface of the crys-tal They also pointed out that, under their conditions,the solution phase would contain essentially only mono-mers and dimers during the crystallisation of zeolite X,with higher oligomers (perhaps up to Si20) present inthe synthesis of the more siliceous Y-types

synthe-3.6 Introduction of organic templates

In 1961, two groups of workers disclosed the effect ofintroducing quaternary ammonium cations into zeolitesynthesis Barrer and Denny described amine-associatedroutes to zeolites A and X [11]whilst Kerr and Koko-tailo published[12,13]data on a tetramethylammonium

(TMA) silica-rich version of zeolite A named ZK-4 (Si/

Al up to 1.7) At first, no new structures resulted from

this pioneering work, but in 1967 Wadlinger, Kerr and

Rosinski reported[14]the discovery of the first

high-sil-ica zeolite, zeolite beta (5< Si/Al < 100), made using the

tetraethylammonium cation ZSM-5followed in 1972

[15], the original syntheses being based on a

tetrapropyl-ammonium (TPA)–sodium mixture All these materials

were formed as crystalline products containing the

encapsulated organic cations, leading to the idea of

‘‘templated’’ synthesis with the organics acting as

struc-ture-directing agents (SDAs) In terms of the

mechanis-tic alternatives discussed above (Section 3.5), the

introduction of these organic reactants provided new

possibilities for probing the chemistry of the synthesis

reaction Investigating the synthesis of zeolite A and

other aluminous zeolites, McNicol et al were able to

de-tect the clathration of TMA units by a shift in the

754 cm1 Raman band, supported by results from

Eu3+ phosphorescence spectroscopy [76] They found

no evidence for cage-like building blocks in either

solu-tion or solid before the onset of crystallisasolu-tion

In time, the attractive idea of a strong ‘‘lock and key’’

relationship between framework and template would

come to dominate much of the thinking on synthesis

mechanism and two reviews published in the early

1980s focused attention on this area of host-guest

sci-ence[80,81] However, a scheme[58]introduced by

Der-ouane and co-workers at about the same time was aimed

mainly at explaining experimental observations and

con-centrated on the inorganic gel chemistry[58–62] Based

on investigations using a wide variety of techniques,

they proposed two pathways for ZSM-5formation

The use of Al-rich ingredients and polymeric silica was

pictured as generating a small number of nuclei which

grew by a liquid phase ion transportation process to

yield large ZSM-5single crystals (synthesis A) This

as-pect of the suggested reaction mechanism therefore

bears many resemblances to the solution-mediated

scheme of Zhdanov (Section 3.4) For syntheses of type

B (typified by high Si/Al ratios and the use of

‘‘mono-meric’’ Na silicate), the results were interpreted in terms

of numerous nuclei which rapidly yielded very small

ZSM-5microcrystallites directly within the hydrogel

in a process described as a solid hydrogel phase

transformation

The first suggestion as to how the presence of an

or-ganic template molecule might modify the physical

chemistry of the synthesis medium was put forward by

Flanigen and co-workers [82,83] It was proposed that

the crystallisation mechanism of siliceous zeotypes

in-volves clathration of the hydrophobic organic cation

in a manner analogous to the formation of crystalline

water clathrates of alkylammonium salts Thus, under

synthesis conditions, the silica tetrahedra assemble into

a framework in place of the hydrogen-bonded water

‘‘lattice’’ of the water clathrate and surround the phobic organic guest molecules In this way, the struc-tural chemistry of water below room temperature istranslated to that of silica near 200C This conceptwas developed and extended in a landmark paper byChang and Bell[63]

hydro-3.7 Chang and Bell and afterThe work of Chang and Bell [63] was based uponstudies of the formation of ZSM-5from Al-free precur-sor gels at 90–95C using XRD,29Si MAS NMR spec-troscopy and ion exchange The NMR results suggestedthat major changes in gel structure occur during theearly stages of reaction This was confirmed by the dem-onstration of ion sieve effects indicating that, in thetetrapropylammonium (TPA) system, embryonic struc-tures with Si/TPA = 20–24 are formed rapidly uponheating These first-formed units may resemble ZSM-5channel intersections (4 per unit cell of 96 tetrahedralatoms), each containing essentially one TPA+ cation,and thus provide a possible mechanism for ZSM-5nucleation In this scheme, the hydrophobic effect andthe isomorphism between water and silicate structurelead to (i) formation of a clathrate-like water structurearound the template, and then (ii) conversion of theclathrate-like hydrate to a clathrate-like silicate by iso-morphous substitution of silicate for water in the embry-onic units Such units are initially randomly connectedbut in time become ordered (‘‘annealed’’) through re-peated cleavage and recombination of siloxane bonds,mediated by hydroxide ion Thus, nucleation occursthrough progressive ordering of these entities into the fi-nal crystal structure This dynamic-assembly argument

is very reminiscent of that originally put forward byFlanigen and Breck for an inorganic system[31,51](Sec-tion 3.2)

The principal concepts advanced by Chang and Bell

[63] have been extended in a series of papers in whichBurkett and Davis [64–66]examine the role of TPA asstructure-directing agent in silicalite synthesis, primar-ily by MAS NMR spectroscopy 1HA29

Si CP MASNMR results provide direct evidence for the existence

of pre-organised inorganic–organic composite tures in which the TPA molecules take up a conforma-tion similar to that adopted in the zeolite product Theinitial formation of the inorganic–organic composite isinitiated by overlap of the hydrophobic hydrationspheres of the inorganic and organic components Sub-sequent release of ordered water enables favourablevan der Waals interactions to be established Nucle-ation is then brought about through aggregation ofthese composite species Crystal growth occurs throughdiffusion of the same species to the surface of the grow-ing crystallites to give a layer-by-layer growthmechanism

struc-Broadly similar ideas have also been developed in

what has become a very extensive study by a team at

Leuven Using a wide variety of experimental

tech-niques, the work has concentrated on a detailed

charac-terisation of the MFI precursor material originally

described by Schoeman [84,85] The first papers in the

series [67–71]identified constituent ‘‘nanoslabs’’ having

dimensions 1.3· 4.0 · 4.0 nm with nine intersections

per particle, each of these containing a TPA cation

Aggregation of such nanoslabs leads to larger particles

measuring up to 15.6· 8 · 8 nm and ultimately to the

crystalline colloidal MFI-type material which forms

the final product of the synthesis More recently

[72,73], specific silicate oligomers (particularly a

penta-cyclic dodecamer) were identified as intermediates in

nanoslab evolution However, the elaborately detailed

interpretations adopted in these studies are the subject

of increasing criticism[86–88]

Some of the principal ideas from the period described

in Sections 3.1–3.7 are summarised inTable 1

4 Modelling the processes of zeolite synthesis

As computing capability has mushroomed, modelling

methods have become an increasingly important adjunct

to experimental studies It is convenient to consider

‘‘reaction models’’ and ‘‘molecular models’’ under

sepa-rate headings, although it is hoped that in due course the

two branches of the subject will grow together

4.1 Mathematical models of synthesis reactions

Reaction models as used by chemists and engineers

are of two basic types: (i) those using a kinetic approach

and (ii) those founded on a thermodynamic approach

Those in the first category[89]range from simple

empir-ical correlations to complex computer programs Of the

more complex treatments, the most important are those

based on particle numbers, such as the population

bal-ance model[90,91]developed extensively by Thompson

and co-workers and built upon the basic equation (for a

where n is a number density function (characterising the

crystal size distribution at any time), t is time, L is

crys-tal length, Q is the cryscrys-tal linear growth rate and s is

res-idence time Further relationships set boundary

conditions and the material balance Solutions for the

resulting cohort of equations can be developed to

pro-vide simulations covering a wide variety of conditions

In this way, hypothetical reactions can readily be

ex-plored to assess the effect of changing reaction variables

and introducing other components such as seed crystals

For example, predictions of crystal size and size bution can be developed to reflect changes in nucleationand growth behaviour brought about by gel ageing (Sec-tion 11.2)

distri-The only significant themodynamics-based model ofzeolite synthesis to have emerged is the equilibriummodel of Lowe[92] This was initially developed to pro-vide insight into the pH changes which occur in thecourse of high-silica zeolite syntheses [93] The modelconsiders the zeolite synthesis process as a series ofpseudo-equilibria:

amorphous solid$ solution species $ crystalline zeolite

— progress of reaction!

At the start, amorphous solid is in equilibrium withsolution species This initial equilibrium is then main-tained while product crystals grow from the supersatu-rated solution Finally, when all the amorphousprecursor has been consumed, the crystalline zeoliteequilibrates with its mother liquor Lowe
s originalsketch of this process is reproduced in Fig 2

This simple analysis enables the solution chemistry,and in particular the effects of solubility and pH, to beunderstood at a fundamental level[92] Computer mod-elling of the pH function provides a good simulation ofthe types of pH curve observed experimentally[94] Themost notable feature is the sharp rise in pH which occurswhen all of the solid gel phase has been consumed andcontrol of the solubility is transferred to the crystallineproduct The difference between initial and final pH val-ues is directly related to the difference in solubility be-tween zeolite product and gel precursor, providing ameasure of the strength of the templating effect for a ser-ies of organic additives[95] The most effective templategives the most stable (least soluble) product and hencethe largest pH rise Perhaps the key relationship is the

Fig 2 Conceptual basis for the Lowe equilibrium model [92] of the zeolite synthesis process (B.M Lowe
s original sketch) Control of solubility passes from the initial equilibrium between amorphous solid and solution species to the final equilibration between the crystalline zeolite product and its mother liquor.

ratio of precursor and product solubility constants (Ks),

since this is directly related to the solution

supersatura-tion (Secsupersatura-tion 14) and hence the driving force for the

In terms of zeolite synthesis, molecular modelling

methods have provided insight into (i) the determination

of the location and energies of the templating agents

oc-cluded within zeolite structures during synthesis and (ii)

the detailed investigation of small framework fragments

4.2.1 Modelling zeolite-template pairs

The key advantage modelling methods offer over

experimental techniques is that they permit investigation

of the energetics of template–framework interactions

However, experimental studies, most notably X-ray

dif-fraction, have yielded crucial data on template location,

enabling modelling methods to be validated Modelling

studies of zeolite-template interactions have usually

fo-cused on the use of molecular mechanics calculations

This basic methodology has been employed in several

different ways to investigate the relationship between

template molecules and the zeolite product Moini

et al.[96]optimised the geometry of the template

mole-cules using the molecular mechanics approach prior to

docking them ‘‘by eye’’ into the framework structure

In this way, they demonstrated the excellent void filling

properties exhibited by EU-1 (EUO) templates, as

shown here in an alternative representation (Fig 3)

This type of technique was also successfully employed

by Schmitt and Kennedy to derive new templates for

ZSM-18 (MEI) based on their geometric match with

the framework[97]

The molecular mechanics approach has been

en-hanced by the addition of a Monte Carlo algorithm to

dock the guest molecule inside the framework prior to

the application of an energy minimisation or simulated

annealing routine The application of these techniques

has shown convincingly that the relative non-bonded

en-ergy between the template and framework can be used

as a measure of the efficacy with which a selected

tem-plate can form a particular product Shen and

co-work-ers[98]have demonstrated the importance of including

the Gibbs Free Energy in order to fully rationalise the

synthesis of ZSM-11 by TBA ions The addition of a

simulated annealing protocol to the Monte Carlo

rou-tine is particularly useful when several templates are

optimised within the simulation box Stevens et al.[99]

successfully used this method to show that a template

which can be used to synthesise different products under

different synthesis conditions has a similar binding

en-ergy in all its products, and that levels of void fillingare also similar in each case

Zones et al have made effective use of modellingmethods to aid their search for bulky molecules thatare too large, or have the wrong geometries, to synthe-sise common default products This has resulted in thediscovery of several new structures[100] An automated

‘‘De Novo’’ approach for tailored template design hasbeen developed by Lewis and co-workers [101] In thismethod, a template molecule is ‘‘grown’’ computation-ally inside the zeolite host in order to maximise thenon-bonded interactions between template and the sur-rounding lattice This has resulted in the successful de-sign of a new template for DAF-5(a CoAPO material)

[102].4.2.2 Cluster calculationsThe properties of small silica fragments critical to thesynthesis process are difficult to study via experimentalmethods In this respect, cluster calculations, usuallybased on quantum mechanical approaches, have provedhighly valuable as a tool to probe the detailed structure,geometry and reactions for a wide range of fundamentalsilica fragments likely to be of importance in the synthe-sis process (see for example Refs [103,104]) Pereira

et al [105,106] have developed these procedures to

Fig 3 The energy minimised location for nium ions shown relative to the EU-1 channel system.

dibenzyldimethylammo-examine the interaction between solvated zeolite

frag-ments and template molecules The results from this

study demonstrate that the organic molecule plays a

key role in stabilizing clusters of framework material,

preserving their structure under the influence of water

Methods for investigating nucleation processes have

re-cently been extended by the work of Wu and Deem

order to model a silicate solution on the atomic scale

in the absence of a template molecule This study has

yielded an insight into some of the fundamental

pro-cesses associated with nucleation such as estimation of

the nucleation barrier and the critical cluster size

Part II: Synthesis mechanism

In Sections 5–8, suggestions for the most probable

mechanistic pathways in zeolite formation are described

in sequence: induction period, nucleation, crystal

growth In order to balance coherence with conciseness,

some of the background material necessary to support

the main arguments is noted only briefly Fuller

ac-counts of such subjects as the nature of growth species

and the role of aggregation processes are given later,

in the ‘‘key topics’’ part of this review (Sections 9–18)

5 The induction period

The induction period is the time (t) between the

no-tional start of the reaction and the point at which

crys-talline product is first observed It will therefore depend

on the moment chosen for setting t = 0 (often taken as

the time at which the reactants reach the working

tem-perature) and upon the method of analysis used to

de-tect the product (most usually X-ray diffraction) For

precipitation reactions, classical nucleation theory

[108]divides this period (s) into a number of subunits:

s¼ trþ tnþ tg;

tr is referred to as the relaxation time and is said to be

the time required ‘‘for the system to achieve a

quasi-steady-state distribution of molecular clusters’’ In

zeo-lite terms, this can be equated with the equilibration

reactions taking place on mixing the reagents and

allow-ing them to reach reaction temperature, durallow-ing which

period the observable distribution of silicate and

alumi-nate ions (and other species) is established (see Section

6.2) The subunits tn(the time for the formation of a

sta-ble nucleus) and tg(the time for the nucleus to grow to a

detectable size) are directly translatable into zeolite

chemistry

It will be apparent from the above that the induction

period encompasses all the significant events of zeolite

formation Before this time, there existed only the

individual reactants; after it, the reaction mixture tains a myriad of small zeolite crystals, already formedand in the course of growing larger However, thegrowth process once underway displays no discontinu-ity, so that the mechanism of further growth is essen-tially that already established in the early stages Thispoint is reinforced by the analysis of the induction per-iod made by Subotic´ and co-workers [109], where it isshown that the activation energy determined from s isessentially that of the entire crystallisation process Afurther modelling study demonstrates the importance

con-of including the lag time (tr) in analyses of zeolite tallisation and confirms the importance of gel/solutionrearrangements to this element of the induction period

crys-[110] It is also shown that size-dependent crystal bility (the Kelvin effect) is not a significant contributor

solu-to the crystal growth-rate function (see also Sections8.5and 16.4)

In Section 6, we consider further the significance of trand the nature of the equilibration processes whichtransform the amorphous phase before zeolite productappears These prove to be of prime importance in set-ting the stage for the unique event of nucleation (Section7) It will then be demonstrated (Section 8) that suchprocesses also provide the link connecting the chemistry

of nucleation with that of crystal growth

6 The evolution of order6.1 The nature of the amorphous material

It is often convenient to treat the amorphous phase as

a constant quantity which, apart from depletion, mains essentially unchanged throughout the synthesis.Such an approach is usually adopted for the purposes

re-of reaction modelling [89–92] (Section 4) However,the dynamic nature of this material was recognised inthe earliest mechanistic studies (Section 3) Flanigenand Breck envisaged [51,52] a transformation throughpolymerisation and depolymerisation catalysed by ex-cess hydroxyl ion, whilst Zhdanov [57] and Kerr [55]

saw the initial amorphous gel as coming to equilibriumwith the liquid phase and releasing active soluble speciesinto the solution, thus changing with time As will beshown, this transformation in the nature of the amor-phous phase during the early stages of the reaction isvery significant

6.2 Primary and secondary amorphous phases

At the point where the synthesis reactants are initiallymixed together, a visible gel is frequently formed Thiswill be referred to as the primary amorphous phase Insome cases (‘‘clear solution’’ syntheses), this primaryphase is colloidal, and thus invisible to the naked eye,

but its function and behaviour are essentially the same

(see Section 7.6) The primary amorphous phase

represents the initial and immediate product from the

reactants and is a non-equilibrium and probably

hetero-geneous product containing (for example) (a)

precipi-tated amorphous aluminosilicates, (b) coagulated silica

and alumina precipitated from starting materials

desta-bilised by the change in pH and increase in salt content

and (c) unchanged reactants The pH of such a mixture

is not usually a useful characteristic measurement, since

it depends on particular circumstances and will change

with age

After some time, either on standing, or—more

rap-idly—on heating at reaction temperature, the above

mixture undergoes changes due to the equilibration

reactions which occur (Section 7.8) and is converted

into a pseudo-steady-state intermediate, the secondary

amorphous phase Concurrently, the relationship

be-tween the solid and solution phases approaches an

equilibrium and a characteristic distribution of silicate

and aluminosilicate anions is established (Fig 4) A

pH measurement will now provide a useful reference

point from which, in high-silica zeolite synthesis, the

progress of the reaction can be monitored by recording

subsequent changes [92,93] In the final stage of the

synthesis reaction (usually at elevated temperature for

a prolonged period), the secondary amorphous phase

is converted into the crystalline zeolite product (Fig

5)

The concept of an equilibrated intermediate phase

is clearly expressed in Zhdanov
s representation of

the synthesis process [57] (Section 3.4) and is also

im-plied in the type-A and type-B schemes of Derouane

et al [58] (Section 3.6—see also Table 1) In order

to show that such equilibration has occurred, the

bal-ance of solution species and the partition of

compo-nents between the solid phase and solution can be

probed by a variety of analytical and spectroscopic

techniques (e.g Refs [61,111–116]) The structuring

action of cations is apparent from their role in the

organisation of the solid phase [117], as described

below In some cases, there are also strong interactions

with solution species—as demonstrated, for example,

by the correlation between the occurrence of the cubic

octamer [Si8O20]8 and the presence of the TMA

cat-ion [118]

Several groups of authors have commented upon the

existence of the secondary amorphous phase Angell and

Flank[77] report (for zeolite A synthesis) that ‘‘ the

initially formed sodium aluminosilicate gel is converted

via solution transport to an apparently amorphous

alu-minosilicate intermediate It is this latter material which

is converted to crystalline zeolite via dissolution by the

basic medium.’’ Fahlke et al in 1987 observed (for

zeo-lite Y) an immediate precipitation of a silica-rich

pri-mary gel, followed by its slow dissolution and then the

precipitation of a secondary gel having a similar Si/Alratio to that of the zeolite product [119] However, astudy by the Montpellier group in 1993[117]is particu-larly informative and merits a more detailed description

SiO 2 source Al 2 O 3 source

M+ OH– Q+ OH– partly-reacted, heterogeneous non-equilibrium mixture:

solution + solid PRIMARY AMORPHOUS PHASE

– H 2 O ionised monomers ionised polymers

OH–

EQUILIBRATED SOLUTION

increasingly ordered

Equilibrated gel

Solution

Initial gel

nucleation

growth

random

Crystalline product

Fig 5 The evolution of order, from the primary amorphous phase (a) through the secondary amorphous phase (b) to the crystalline product (c).

6.2.1 The Montpellier study

This investigation[117]is the clearest demonstration

yet published of the multi-stage nature of the zeolite

syn-thesis reaction as outlined inFig 5 In this work, Nicolle

et al prepared a series of Na, TEA-aluminosilicate gels

of varying composition (TEA = tetraethylammonium)

and heated samples of these to 150C (i) for a relatively

short period and (ii) for a longer time The solid

prod-ucts from each reaction period were isolated and

ana-lysed The actual heating times varied with the

reactivity of the particular composition but were aimed

at achieving in the first instance (i) an equilibrated gel,

and in the second case (ii), a crystalline zeolite product

The four compositions chosen (A–D) crystallised in the

second stage to (A) zeolite beta, (B) mordenite, (C)

ZSM-12 and (D) ZSM-5 However, the most significant

findings came from the analysis of the initial,

equili-brated products

X-ray diffraction revealed only broad features

indica-tive of local ordering but no crystal lines All samples

showed aluminium in tetrahedral coordination only

(27Al MAS NMR) and no zeolitic bands in their infrared

spectra Chemical analysis demonstrated that the

hydro-thermal treatment had greatly modified the composition

of the solid phase (i.e compared to that before heating),

with TEA+ replacing Na+ as the preferred cation The

TEA content was nearly constant, corresponding to

one organic cation per 16–20 T-atoms, whatever the

method of preparation or the Si/Al ratio Other, specific

tests were carried out on selected samples Equilibrated

samples A and B had (after calcination) cation exchange

capacities in close agreement with their aluminium

con-tents (
85mmol NHþ

4/100 g) For equilibrated samples(A)–(C), calcination generated a micropore system hav-

ing about 60% of the capacity shown by the zeolite beta

crystallised from composition A Amorphous sample B

showed the same micropore volume for cyclohexane as

for nitrogen but was impermeable to trimethylbenzene

The authors concluded that the amorphous,

equili-brated samples differed from the gels initially

precipi-tated at room temperature and had been formed by

dissolution and reprecipitation These secondary

prod-ucts were silicoaluminates sharing several properties

with high-silica zeolites In both cases, clusters of silica

or alumina tetrahedra had been organised (templated)

around a bulky molecule, whose decomposition left a

micropore system The essential difference between the

two classes of solids was represented by the dispersion

of the cluster geometry Zeolites featured well-defined

and repeatable site geometries whereas the equilibrated,

amorphous products presented irregular and hence

ape-riodical, local organisation From data discussed

else-where (Sections 6.3 and 12) it is clear that these are

general observations and not, for example, associated

only with the synthesis of zeolites templated by

quater-nary ammonium ions Similar phenomena occur during

the preparation of aluminous (inorganic) zeolites

[77,119].6.2.2 Related investigationsThere appears at first sight to be a discrepancy be-tween the observations of the Montpellier group [117]

and those described in a detailed investigation of zeolitebeta synthesis by other workers On the basis of thermalanalysis and charge-balance data, Perez-Pariente et al

the amorphous solid However, in later work, Camblorand Perez-Pariente reported the presence of TEA cat-ions associated with the aluminate sites in pre-crystallinesamples [121], concluding nevertheless that the overallcrystallisation pattern (by a solution transport mecha-nism) was the same in both cases The most obvious dif-ferences between the two studies were the use oftetraethyl silicate and a low crystallisation temperature(100C) in the earlier work, whereas the later investiga-tion employed amorphous silica (Aerosil) and a reactiontemperature of 135C However, a more significant dis-parity lies probably not with these variables but in thecation balance The 1991 study centred around synthesiscompositions which are fairly typical of routine zeolitebeta preparations, namely

where x was varied from 0 to 4.5and y from 0 to 1, withOH/SiO2 constant at 0.56 The range of compositionsused for the 1987 study was generally similar—withthe exception of the gel used for the analysis of theamorphous intermediate (B2at 12 h), which was7:8Na2O 0:11K2O 1:5ðTEAÞ2O Al2O3 30SiO2 360H2OThe very high M/TEA ratio will have a major effectupon the partition of cations between the liquid andgel phases, so that the low TEA content of the isolablesolid phase is perhaps not unexpected

6.3 Further evidence for pre-crystalline order fromsynthesis studies

Tsuruta et al studied the initial product in the thesis of zeolite A from concentrated solutions in thepresence of an anionic surfactant [122] Using X-rayand electron diffraction, they found that the amorphousaluminosilicate isolated under mild reaction conditions(60C, 15min) possessed a short-range order of Siand Al atoms similar to that in crystalline zeolite A.Electron diffraction, thermal analysis and FTIR spec-troscopy were used by Subotic´ and co-workers in astudy of the structural properties of X-ray amorphoussodium and potassium aluminosilicate gels [123,124].The gels were found to contain structurally ordered re-gions, or particles of a partly crystalline phase inside

syn-an amorphous matrix Subsequent hydrothermal

treat-ment in 2 M NaOH at 80C yielded zeolites A and X

Walton and O
Hare[125]examined amorphous gallium

silicates isolated in the early stages of the hydrothermal

synthesis of Ga-hydroxysodalite by means of EXAFS

and XANES spectroscopy Both Ga and Si K-edge

EX-AFS indicated some degree of medium-range order (i.e

beyond the first coordination shell), indicative of an

ex-tended network of alternating Ga(OSi)4 and Si(OGa)4

units resembling the structure of the crystalline product

A related multi-nuclear MAS NMR and neutron

diffrac-tion study of amorphous zeolite A precursors by Yang

et al.[126]detected changes in medium range order prior

to crystallisation as Al (accompanied by Na+ cations)

was incorporated into the silicate network However,

both spectroscopic and diffraction techniques indicated

that there were no well-defined structural units (e.g

SBUs) present before the formation of the zeolite

crys-tals Rather equivocal results were obtained from a

129

Xe NMR study of samples taken during the course

of various zeolite syntheses[127] Evidence of gel

struc-turing (large cavities) was found for Na,TPA-ZSM-5

and Na-Y but not for Na,TEA-ZSM-20 However, only

in the latter case did the NMR measurement show any

appreciable microcrystallite formation in the absence

of XRD crystallinity

There have been several other investigations of the

pre-crystalline state of MFI-type synthesis compositions In

the work of Burkett and Davis, 1HA29

Si CP MASNMR spectroscopy demonstrated that the TPA mole-

cules and silicate species were in close proximity in the

heated gel before the characteristic XRD and IR

finger-prints of crystalline silicalite became apparent[64] No

such correspondence was found in the unheated gel This

result was supported by an in-situ multinuclear NMR

study at 80C[128] Hydrogen bonds between the organic

and water-clathrated molecules are progressively

re-placed by hydrophobic interactions between the organic

and silica species Similar evidence of TPA occlusion

comes from SAXS [129] and SANS[130,131]

measure-ments Several groups of workers have reported FTIR

bands at 550–560 cm1 in isolated material which is

X-ray amorphous[84,130–134], these being interpreted as

indicative of the presence of material having the MFI

structure However, some caution is necessary in this

analysis since such an absorption is not unique to

ZSM-5 It is advisable to compare the bands which should

ap-pear at both 550 and 450 cm1: for a well-crystallised

ZSM-5, the 550:450 intensity ratio should be
0.7, a

fig-ure which decreases with decreasing crystallinity[135]

6.4 Summary

When first prepared, the reaction mixture for a zeolite

synthesis consists of a non-equilibrium combination of

components—heterogeneous and at best partly reacted

(primary amorphous phase) However, over a period oftime (tr), and especially on heating, silicate and alumino-silicate equilibration reactions will occur, leading to are-distribution of species and a repartition of reactioncomponents between the solid and liquid phases as theyapproach equilibrium with one another Cations play astructuring role in the organisation of the solid phase,which is generated as a new or reconstructed materialhaving a similar chemical composition to that of theeventual zeolite product but lacking in long-range, peri-odical organisation and hence amorphous to X-rays (sec-ondary amorphous phase) However, elements of localorder will be present and will impart to the amorphousintermediate some of the chemical and physical proper-ties associated with the final, crystalline, ‘‘isomeric’’ zeo-lite This is shown diagrammatically in Fig 5 and isdiscussed further when the topic of ‘‘X-ray amorphouszeolites’’ is considered (Section 12) In Section 7, thissemi-organised solid phase is shown to be pivotal inthe transforming step of zeolite nucleation

In the above discussion, a firm distinction is made tween primary and secondary amorphous phases in order

be-to emphasise the importance of the changes taking place.However, the situation in some zeolite syntheses may beless clear cut, since the different stages in the process mayoverlap, e.g nucleation may begin whilst the bulk of thereaction mixture is still far from attaining a steady state(this is more likely to occur in high temperature synthe-ses) This ‘‘relative kinetics’’ effect is an important factor

in the dependence of the outcome of a synthesis reactionupon the nature and mode of mixing of the reactants.Certainly, in cases where no intermediate samples aretaken for physico-chemical characterisation, the re-searcher may be unaware that the progression from reac-tants to product is anything other than a continuum.The extent to which the secondary amorphous phasecan be identified or isolated will also depend upon thenature of the synthesis mixture In some cases, partition

of the reactants between the liquid and solid phases may

be such that little or no material may be identifiable byconventional methods of solid–liquid separation, so thatcolloid chemical techniques may be necessary to observe

it (see Sections 7.6 and 7.7)

Finally, it should be noted that inFig 5gross neity (e.g the presence of some unchanged starting com-ponents) in the primary amorphous phase has beenignored for the sake of simplicity and the material illus-trated as a single homogeneous phase of random structure

heteroge-7 Nucleation7.1 IntroductionFrom the above (Section 6), it is clear that the equil-ibration which allows the conversion of primary to

secondary amorphous phases (Fig 5 and Section 7.8)

forms an extremely important part of the zeolite

synthe-sis process By this means, the initial gel reacts with

solution species to establish (or approach) a

pseudo-steady-state distribution of liquid phase species,

pre-dominantly ionic but with a wide range of charge/mass

ratios (Fig 4and e.g Refs.[111–116]) The changes in

the amorphous phase involve an increase in structural

ordering but without the establishment of the periodic

zeo-lite lattice itself For this, a discreet nucleation event has

to occur In this step, a statistical selection of the

recon-structed areas reach a critical nuclear size and degree of

order such that a periodic structure is able to propagate,

i.e crystal growth can begin

In the following treatment, nucleation is first

consid-ered from a classical standpoint (Section 7.2), after

which experimental methods for the determination of

nucleation patterns are described (Section 7.3) Next,

brief reviews are given on the topics of seeding (Section

7.4) and autocatalytic nucleation (Section 7.5) Finally,

the special features of zeolites are examined, leading to

a consideration of the underlying mechanism (Sections

7.6–7.8)

7.2 General considerations

Amongst the components (tr, tn, tg) comprising the

induction period (s) (Section 5), tn is the time taken to

form a stable nucleus.Fig 6illustrates the classical

con-cept of a critical nuclear size [108] Beyond this point

(where nuclear radius (r) = rc), sufficient matter has

come together in an ordered way for the cohesive energy

(DGv) to outweigh the energy expended in creating a

sur-face dividing the nucleus from the continuum (DGs)

This enables a stable unit to be formed: a viable nucleus,

capable of growth For zeolite systems, a critical size of

1–8 unit cells has been estimated by Thompson and

Dyer[90] However, the chemical and physical processes

leading up to critical nucleus formation are governed by

the structure being formed and by the experimental

con-ditions If these conditions change, then rcwill also vary,

as can be seen from the general form of the expression

for r below[108]:

r¼ 2rv

kTln S;

where r is the interfacial tension (surface energy per

unit area), v is the molecular volume and S is the

super-saturation ratio A notional nucleation rate (J) can be

defined as the rate of production of units having radius

r P rc A simple measure of J is often taken by

assum-ing J/ 1/s [31,78] However, since s is a compound

term, the danger in this procedure is apparent, unless

tr and tg are known to be small compared with tn

[109,110]

It seems reasonable to suppose that the construction

of the nucleus of a zeolite crystal may involve a morecomplex assembly process than is necessary for simplersubstances whose unit cells are smaller and contain farfewer atoms, although there should be no difference inbasic principles One possible special feature in the for-mation of microporous crystals has however been iden-tified by Pope, who has pointed out that the energetics

of zeolite nucleation may be greatly modified from that

of dense phases because of the presence of the largeinternal surface area [136]

The nomenclature for the most usually recognisedtypes of nucleation can be summarised as follows

solution) or heterogeneous (induced by foreign cles); secondary nucleation (induced by crystals) may

parti-be considered as a special case of heterogeneous ation in which the nucleating particles are crystals ofthe same phase Good general accounts of nucleation

nucle-in zeolite systems (along with further references) will

be found in reviews by Barrer [137] and Thompson

of the information available is at best that of earlygrowth behaviour rather than of nucleation itself Per-haps only the recent cryo-TEM[129]and HRTEM stud-ies[134,138,139](Section 7.7) genuinely fall outside thiscategory and can be considered as direct evidence of

+ ve

∆Gc = 3

4 πσrcr

Fig 6 The energetics of nucleation, illustrating the concept of a critical nucleus of radius r c ; beyond this size, the net energy gain from the resultant (DG) of cohesive (DG v ) and surface (DG s ) terms is favourable to growth (after Mullin [108] ).

nucleation phenomena However, much useful data can

be obtained from ‘‘circumstantial evidence’’, i.e

mea-surements of the appearance, growth and size

distribu-tion of the crystals resulting from the earlier, hidden

patterns of nucleation

7.3.1 Studies under isothermal conditions

It is still fairly common practice to estimate

nucle-ation rates in zeolite synthesis from the reciprocal of

the induction times[78], despite the acknowledged

pit-falls (Section 7.2)[31] This is largely because of the

dif-ficulty of making more appropriate and accurate

measurements, particularly if temperature variation is

to be included An improvement on this can be found

in the procedure of Zhdanov and Samulevich, which

en-ables the calculation of isothermal nucleation rate

pro-files from determinations of growth rate and crystal

size distribution [140,141] Originally implemented in

analyses of zeolite Na-A[57]and Na-X[140]

crystallisa-tion, the method has subsequently been applied to other

zeolite systems, including silicalite[142,143] If it is

sup-posed that all the crystals in a batch have the same

(known) growth rate behaviour, the total growth time

of each crystal can be calculated Assuming also that

the nucleation point for each crystal can be obtained

by linear extrapolation to zero time, the nucleation

pro-file for the whole batch can be determined, as illustrated

first to apply the above type of analytical procedure to

zeolite synthesis, Giaya and Thompson, whilst

research-ing their own numerical technique for determinresearch-ing the

crystal size distribution function [144], discovered that

the approach had been discussed in a wider context

much earlier by Bransom and Dunning[145].)

7.3.2 Ageing studies

Although the types of investigation mentioned above

yield useful data, far more detailed information on

nucleation behaviour can be obtained from ageing

stud-ies In these cases, a reaction mixture is allowed to age at

one temperature and is then crystallised at a second

(usually higher) temperature By correlating changes

(principally of size) in the product crystals with the

length of the ageing period, aspects of the nucleation

behaviour can be deduced In particular, the evolution

of order (Section 6) in the zeolite synthesis mixture

(equivalent to the concept of accumulation of

proto-nuclei) can be determined quantitatively A detailed

ac-count of ageing processes is given in Section 11

7.4 The use of seed crystals

It is a common and very useful practice to add seed

crystals to zeolite reaction mixtures and excellent

ac-counts of this technique are available [37,91,146] The

two main effects anticipated are (i) a reduction in thesynthesis time and (ii) a ‘‘direction’’ of the synthesistowards a desired phase with consequent reduction in

Fig 7 Determination of nucleation rate profiles by the method of Zhdanov and Samulevich [140] as applied to silicalite synthesis [142] From final product crystal size distribution (a), crystal linear growth rate (b) allows calculation of nucleation rate as a function of reaction time (c) Data consistency is shown by comparison of calculated mass% growth rate ((d), curve) with experimental values ((d), marked points) Note that the length axis in (a) becomes a time axis in (b)–(d).

impurities It is also possible to exert control over

prod-uct crystal size distribution [147–150] The bases for

these valuable influences are worthy of brief

consideration

The basic action of seed crystals (when added

effec-tively) is to provide surface area on which the required

product can grow This removes the necessity for such

surface to be self-generated by primary nucleation and

thus reduces the induction time (s), since the tn

compo-nent of s is eliminated (see Section 7.2) In some cases

(equivalent to the traditional scratching with a glass

rod in organic chemistry), almost any intervention

may produce the desired effect However, in others there

is a definite element of surface recognition by which one

phase can be given a kinetic advantage whilst potential

competitor products are relatively unresponsive

If seed crystals are added to a synthesis mixture, they

may behave in a number of ways, i.e the crystals may

(a) remain inert, (b) dissolve, (c) act as pure seeds, in

that mass is deposited upon them and they grow, or

(d) give rise to secondary nuclei and hence a new crop

of crystals In general, it is necessary to provide sufficient

surface area to achieve an effect, so that it is rare for a

very small quantity of large crystals to cause any

signif-icant change in the normal course of the reaction Such

crystals can usually be found at the end of the synthesis,

either intact (i.e case (a)) or showing signs of attack

(case (b)) If the added material is unstable in the

synthe-sis medium, it will normally dissolve and only influence

the course of the synthesis if added in sufficient quantity

to alter significantly the reaction stoichiometry (limit of

case (b))

The balance between seed growth (c) and secondary

nucleation (d) depends on the nature of the system,

the quantity of material added and the degree of

agita-tion If the surface area of seed crystals present is

suffi-cient to absorb most of the available flux of growth

species and thus prevent the effective solution

supersat-uration from reaching high levels, then most of the

growth in the system will take place on the seed crystals,

whose size and growth rate can be closely controlled and

predicted by kinetic modelling[79,89–91,147,148] Very

small (colloidal) zeolite crystals are, as would be

ex-pected, particularly effective [149–152] As the quantity

(or, more correctly, surface area) of added seed material

is reduced, then the natural supersaturation and quent self-nucleation will no longer be suppressed andthe product crystal size will deviate more and more fromthat predicted on the basis of linear growth on seed crys-tals only (Table 2)

conse-Secondary nucleation certainly occurs in zeolite tems but has been the subject of few detailed studies

sys-[37,91] If also assisted by agitation, then some form ofcollision breeding [37,91,108,153] is likely to be opera-tive on a macro- or micro-scale From a practical point

of view, secondary nucleation has been observed in theses which for various reasons (sometimes unknown)are very reluctant to self-nucleate even though the com-positions are believed to be near the optimum Possibly

syn-‘‘active’’ impurities [37,154] are absent, or nucleation/growth poisons are inadvertently present [155,156] Insuch cases, the use of seed crystals can be very effectivesince the response to secondary nucleation is a geometricfunction of the quantity added This is demonstrated inthe synthesis of siliceous forms of EUO-type zeolite(EU-1, ZSM-50) [16] (Fig 8) The reactions [157] wererun at 150C for three days using the systems10Na2O : xAl2O3: 60SiO2: Template : 3000H2O, whereeither x = 0, Template = 13 dibenzyldimethylammo-nium chloride (DMDBACl), or x = 0.05, Template = 10hexamethonium bromide (HEXBr2) In the absence ofseed crystals, reactions gave either very low yields orproducts heavily contaminated with impurity phases(mainly EU-2 and cristobalite) With DMDBA tem-plate, crystallisation was greatly assisted by the inclusion

of seed crystals (5mass% based on silica) Large ent’’ and small ‘‘daughter’’ crystals are clearly distin-guishable The greater available surface area provided

‘‘par-by calcined seed makes this a more effective nucleant(Fig 8b) than the uncalcined material (Fig 8a), result-ing in a larger number of smaller product crystals Thespawning of daughter crystals from a partly dissolvingcalcined seed crystal is shown inFig 8c A similar effectwas found for the HEX preparation of EU-1, where theproduct crystal size from calcined seed (Fig 8d) wasagain smaller (1 lm) than that from as-made material(2 lm, not shown) The micrographs reveal a furtherinteresting point The same batch of Na,DBDMA-EUO seed was used in all cases but it is clearly the syn-thesis reaction template which controls the productmorphology

A further aspect of seeding has been demonstrated byBronic´ and co-workers The syntheses of zeolites Y and

using a 2-compartment reactor in which seed crystalsand reactant gel were separated by a submicron mem-brane Crystal growth occurred in both compartments,

in some cases giving different zeolite phases (Y on seedside, P on gel side) However, it was found that the seedcrystals modify the crystallisation process only if they

are in physical contact with the gel Thus it can be

deduced that growth species in the liquid phase travel

freely through the membrane, whereas nucleation is

con-trolled more locally These studies are thus reminiscent

of the Kerr recirculation experiment [55,56] (Section

3.3)

7.5 Autocatalytic nucleation

One of the sources of nuclei proposed in the classic

paper by Zhdanov[57]was that some germs lie dormant

in the amorphous phase until activated by release into

the solution through gel dissolution Since the rate of

gel dissolution must increase with the rate of

consump-tion of growth species by the increasing cumulative

crys-tal surface area, the process was seen as ‘‘autocacrys-talytic’’

This approach has been elaborated and modelled by

Subotic´ and co-workers [143,160–165] However, it hasbeen pointed out[166] that the nucleation periods pre-dicted by the model are unrealistically long The ideawas therefore modified[167]by postulating that the dor-mant nuclei were located preferentially near the periph-ery of the gel particles and therefore became activatedmuch earlier in the crystal growth/gel dissolution pro-cess A subsequent study by Falamaki et al [168] onZSM-5crystallisation using this modified approach gaveexcellent agreement between the model and experiment.However, there appears to be no chemical justificationfor the assumption and any self-consistent set of modelparameters which generated nuclei as a suitably limitedfunction of synthesis time would presumably produce anequivalent result This is demonstrated by the work ofNikolakis et al.[169] who provide an alternative to thehypothesis of nuclei release from the dissolving gel by

Fig 8 Seed crystals (large) seen amongst the induced secondary crystal population in siliceous EUO-type zeolite synthesis Templates were: (a–c) DMDBA, (d) HEX Seed crystals were: (a) as-made, (b–d) calcined Note that the micrographs vary in scale See text for discussion.

treating gel dissolution and nucleation as interfacial

phe-nomena controlled by the gel–solution boundary Their

model for the gel microstructure gives rise to a

maxi-mum in the temporal surface area function which

corre-sponds to the observed early maximum in the nucleation

rate

7.6 Nucleation in zeolite systems—The nature of the

reaction sol

In Sections 7.3–7.5, nucleation has been considered

largely from an experimental standpoint The

mechanis-tic basis for these phenomena will now be discussed As

a first step, it is necessary to expand upon the

consider-ation of the synthesis reaction mixture introduced in the

original summary of experimental observations (Section

2) In a typical situation (from which there will be

vari-ants), there is likely to be present (i) amorphous

mate-rial, (ii) a solution phase and (iii) one or more

crystalline phases It is relatively easy to identify

compo-nent (iii), even if the crystals are of colloidal dimensions

The distinction between constituents (i) and (ii) is much

less clear cut and has been the source of much

misunderstanding

The simplest case is that in which most of the

amor-phous material is visibly solid in nature, as in traditional

syntheses of aluminous zeolites such as A or X At the

low Si/Al ratio and high pH of these preparations,

nearly all the Si,Al-derived nutrients are present as a

solid gel phase, with the balance existing as low

molec-ular weight species (monomers, oligomers) in a strongly

alkaline (Na+OH) solution phase [170] Separation of

solid and liquid phases by normal laboratory methods

(filtration, centrifugation) will achieve a reasonably

effective division of the reaction mixture into a liquid

phase containing (largely) true solution species

of crystalline to amorphous material can be determined

by XRD measurement

At higher Si/Al ratios and lower pH, the situation

be-comes complicated by an increasing proportion of high

molecular weight silicate and aluminosilicate

compo-nents The distinction between solution-related

poly-mers and colloidal material is a problem which lies at

the heart of the understanding of silicate chemistry

Si1, Si2, Si3, ., Sincannot be regarded as a continuum,

since at some point the larger components behave as if

they are particulate and phase-separated, resembling

an invisible precipitate [171–173] Thus, whereas one

part of the silicate and aluminosilicate loading of the

liquid phase behaves as true solution species, a second,

colloidal fraction resembles a separate amorphous

phase This latter is essentially equivalent, in terms of

energetics, to the solid, visible, amorphous gel but is

invisible to optical detection and inseparable from the

liquid phase by filtration or (standard) centrifugation.The result of this is that separation of solid and liquidphases in syntheses of this type does not remove amor-phous material from the liquid phase (Fig 9) Indeed,

in most so-called ‘‘clear solution’’ syntheses (see e.g.Refs [130,142,175–177]), all the amorphous feedstock

is stored by the system in this way, forming the nutrientwhich provides the driving force for the crystallisation

of the zeolite product Thus, any consideration of ation and crystal growth in hydrothermal zeolite synthe-sis must take into account the role of amorphouscomponents at all levels of subdivision

nucle-7.7 Nucleation in zeolite systems—Homogeneous orheterogeneous?

In zeolite synthesis, there is evidence for the pation of both primary and secondary (crystal-cataly-sed) nucleation, although the former is believed to bethe most prevalent [37,91] Even so, there is often dis-agreement on whether the predominant mechanism ishomogeneous or heterogeneous For example, a recentstudy combining both theoretical and experimentalwork on the nucleation of zeolite A concluded thatnucleation was homogeneous except in the case of seed-ing with certain high surface-area titanias[178] Relatedcomparisons of calculated and observed behaviour forzeolites A and Y by Bronic´ and Subotic´ found that thecontribution of homogeneous nucleation was negligible

partici-[179] However, it can be seen from the above discussion

of the nature of the liquid phase (Section 7.6) that there

is a more subtle ambiguity within the primary tion as to whether an apparently homogeneous nucle-ation event may be genuinely so—or in realityheterogeneous—depending on whether or not the col-loid phase is involved

classifica-Fig 9 The colloid problem: the liquid phase resulting from tional solid–liquid separation is likely to contain both true solution species and colloidal amorphous material.

conven-A study on (initially) clear solution synthesis of

alu-minous zeolites by Aiello et al.[180]showed that

heter-ogeneous nucleation appeared to occur on gel lamellae

which separated at an early stage from the solution

phase A similar phenomenon was later observed in

the clear solution crystallisation of silicalite, where the

first visible crystals were seen to nucleate on traces of

amorphous material (‘‘gel rafts’’) which could be seen

using optical microscopy[142] The authors of the latter

paper suggested that ‘‘the heterogeneous process directly

observable from the small amount of gel precipitated

early in the reaction may be indicative of a more

gen-eral heterogeneous process involving the much smaller,

invisible, colloidal gel particles which represent the

prin-cipal reservoir of silica in the system.’’ The presence of

these particles was at that time inferred from indirect

measurements More recently, this has been confirmed

by the application of improved scattering techniques

For syntheses in which a visible gel phase is present

(probably the most familiar case), examination of the

reaction mixture under an optical microscope usually

re-veals gel particles of around 1000–50,000 nm (1–50 lm)

in size These dimensions are system-dependent and also

liable to change through processes of agglomeration or

agitation-induced fission However, crystallisation in

such systems seems always to be associated with thegel phase, so that the product crystals, when observed

in mid-reaction, are found to be intimately dispersedwithin a matrix of amorphous material From observa-tions on dilute reaction mixtures [142,180], it is reason-able to assume that these crystals have indeednucleated heterogeneously within the gel phase Such asituation is credible since nutrient concentration gradi-ents will be greatest at the solid–solution interface andthe surface of the amorphous phase will provide sites

at which the free energy change necessary for nucleusformation is lower than would be the case for homoge-neous nucleation (equivalent to a reduced number of de-grees of freedom)

A more complex type of heterogeneous nucleationhas been observed by Zandbergen[182] A Na,TPA-sil-icalite synthesis mixture was heated quiescently at 65Cand the solid phase which began to form after two dayswas examined from time to time by HREM After fourdays, two types of morphologies were observed withinthe solids: (A) large (2 lm) amorphous, irregularlyshaped particles, and (B) agglomerates of small, partlycrystalline particles (0.1 lm) The ratio B/A + B andthe fraction of crystalline particles in B both increasedwith reaction time The d-spacings detected in Bsuggested the initial formation of a dense phase Large

Fig 10 Direct observation of zeolite Y nucleation on colloidal gel particles by HRTEM [139] : (a) amorphous particles in freshly prepared aluminosilicate solution; (b–d) development of crystallinity upon hydrothermal treatment at 100 C after (b) 28 h, (c) 48 h and (d) 75h (Reproduced from Ref [139] with permission.)

d-values in the ED patterns were observed only after

25days and only in the vicinity of the dense crystalline

phase, after which the presence of the MFI phase

in-creased rapidly Very small zeolite crystals (<10 nm)

were rare, indicating very rapid growth once the critical

size had been reached

Clear solution synthesis is a special case of gel

synthe-sis in which the amorphous particles remain of colloidal

or sub-colloidal dimensions (650 nm), i.e invisible by

optical examination However, the considerations of

crystal nucleation and growth are essentially the same

This has recently been beautifully demonstrated using

electron microscopy for the clear solution syntheses of

zeolites A[138], Y [139] and Si-MFI [134] by Mintova

and co-workers (Fig 10) Using high-resolution

trans-mission electron microscopy (HRTEM) in conjunction

with in situ DLS, X-ray diffraction and other

tech-niques, Mintova et al have imaged the development of

crystalline structure within amorphous gel particles of

nanometre dimensions, i.e the process depicted

sche-matically in Fig 5 Single zeolite A crystals were

ob-served [138] to nucleate in amorphous gel particles of

40–80 nm in size within three days at room temperature

The embedded zeolite A nanocrystals grew at room

tem-perature, consuming the gel particles and forming a

col-loidal suspension of 40–80 nm crystals A broadly

similar picture was found for the zeolite Y [139] and

Si-MFI[134]phases

It is worth noting that the only case in which a ‘‘clear

solution’’ synthesis does not involve some amorphous

material is that in which all components are in true

tion and crystallisation is driven by differences in

solu-bility only (see Section 14.3) This is very rare and can

yield only small amounts of product—a very dilute

sys-tem of moderate Si/Al ratio could perhaps provide an

example

A further comment can here be made concerning the

interpretation of the phenomenological and practical

differences observed between the ‘‘A’’ and ‘‘B’’ type

pentasil syntheses of Gabelica and co-workers [58–62]

(Section 3.6) Dissimilarities in reagent sources and

reaction composition lead to changes in the partition

of reaction components between solution, macroscopic

gel and colloid phases These in turn give rise to the

observed differences, particularly in nucleation

behav-iour In case ‘‘A’’, nucleation occurs mainly on

colloi-dal particles in the liquid phase In ‘‘B’’, the lower base

level, lower silica/water ratio and higher Na/TPA ratio

lead to a much higher proportion of amorphous solids

and a very high rate of nucleation within the solid gel

phase

In summary, it is concluded that (i) the most common

process of zeolite nucleation relies on a primary

nucle-ation mechanism and (ii) the most probable primary

nucleation mode is heterogeneous and centred upon

the amorphous phase of the reaction mixture (which

for most ‘‘clear solution’’ syntheses is colloidal innature)

7.8 Nucleation in zeolite systems—MechanismThis section considers the implications of the equil-ibration reactions which lead to the formation of thesemi-ordered secondary amorphous phase and howrelated processes then give rise to zeolite nuclei In

diagrammatically as a glass-like network having noelements of order Where there is a driving forcefor zeolite formation, i.e negative DG in theequation

a convenient geometric shape with which to representthe development of an ordered region It does not implyany particular role or importance for 6-rings in general.)The chemical route by which such progressive order-ing is brought about is that highlighted by Flanigen andBreck [51,52] and by Chang and Bell [63], namely thebreaking and remaking of Si,AlAOASi,Al (TAOAT)bonds catalysed by hydroxyl ion, the related condensa-tion reaction also playing a part:

TAOH þOAT  TAOAT þ OH

TAOH þ HOAT  TAOAT þ H2OHowever, the anions do not work in isolation and a cru-cial role in the ordering process is played by the cationspresent in the synthesis system These will act to attractaround themselves energetically favourable coordina-tion spheres of oxy-species and in so doing will generatecertain preferred geometries In this way, cation-depen-dent elements similar to those present in the eventualzeolite product will gradually be assembled The notion

of cation-assisted ordering goes back to the ideas ofBreck [54](Section 3.2) and was more recently general-ised in terms of ordering through minimisation of thepotential energy of molecular assemblies by Brunner

units through electrostatic and (for organo-cations)van der Waals forces as proposed by Flanigen [82,83],Chang and Bell [63] and Burkett and Davis [64–66]

has also been discussed earlier (Sections 3.6 and 3.7).The following argument builds on these ideas to suggest

a specific mechanism for the construction of the zeolite

lattice

InFig 11, a cation is envisaged as migrating to a

tem-porary site in a developing structure at which the

coor-dination geometry is suitable but not ideal Whilst in

place, the cation mediates the acquisition of new T-units

from solution (through the condensation reactions given

above), guiding them into a more favourable

coordina-tion geometry and in so doing generating a periodically

regular local structure In due course, the cation may

have sufficient of its original hydration shell replaced

by lattice oxygen donors to become a fixture at the

newly created site Otherwise (and depending upon the

charge balance), the cation may migrate to function

sim-ilarly elsewhere Statistically, some areas of the overallstructure will be more ordered than others and in duecourse, particular ‘‘islands of order’’ will be establishedrandomly throughout the now semi-ordered network(as (b) in Fig 5) These constitute the ‘‘proto-nuclei’’discussed later on (Section 11) The whole system is dy-namic and at any one moment such islands are beingcreated, up-graded or (by dissolution) destroyed How-ever, the overall trend will be towards an increasing de-gree of order Eventually, some areas of the structurebecome sufficiently ordered that a periodic lattice canpropagate, i.e nucleation has occurred This discontinu-ity takes the form of a topological rearrangement akin

to an isomerisation Such a transformation can beviewed as a first order phase transition [184]and corre-sponds to the achievement of critical radius described inSection 7.2 From this point, the kinetics of accretionsufficiently outweigh the kinetics of dissolution to result

in net growth Thus, the reversible equilibration tions characteristic of the trcomponent of the inductionperiod (s) are superceded by the initial net growth repre-sented by tg However, the underlying chemistry remainsthe same and the reactions still maintain a significant re-verse component

reac-Zeolite nucleation is therefore a discreet event whichcould be defined as ‘‘a phase transition whereby a criti-cal volume of a semi-ordered gel network is transformedinto a structure which is sufficiently well ordered to form

a viable growth centre from which the crystal lattice canpropagate.’’ Two comments may be added The criticalvolume refers to the requirement that the germ structureneeds to be of a certain minimum size, corresponding tothe critical radius (rc) of Section 7.2 Following this, itmay be surmised that the potential nucleus does nothave to be perfect in its structure – rather, it needs to

be ‘‘good enough’’ to function as a nucleus under theprevailing circumstances This reflects the condition that

rc is not a fixed quantity but varies, depending on thevalues of other key variables such as temperature andsupersaturation Just as the critical size is situation-dependent, then a variable ‘‘perfection index’’ can also

be envisaged, the necessary degree of precise periodicitymodulating with the growth environment Thus, forexample, a rather defective potential nucleus might re-main dormant (or dissolve) under quasi-equilibriumconditions but may grow into a crystal under a highsupersaturation driving force An analogy for the grad-ual evolution of such a nucleus would be the cutting of anew key for a tumbler (‘‘Yale’’-type) lock Initially, there

is no chance that the key will function However, as themachining progresses, the new key will gradually be-come more perfect and at some definite point will per-form the discreet action of opening the lock.Comparison of the new key with an original masterkey may reveal errors and imperfections but within acertain tolerance it will perform the required function

Fig 11 The basic mechanism for the cation-mediated assembly of

ordered regions: (a) nomenclature and symbolism; (b) details of in-situ

construction process by addition of solution units to a surface site The

same mechanism can be applied to zeolite crystal growth See text for

full description.

in the existing circumstances It is ‘‘good enough’’

How-ever, if circumstances change (e.g the lock becomes hot

and expands), it is possible that the ‘‘imperfect’’ key may

no longer open it

7.9 Summary

As can be seen from the above, the overall process of

zeolite nucleation is a summation of a complex chain of

events This is necessarily so since it encompasses the

en-tirety of the transformation from an initially random

structure to the beginnings of a regular, periodic crystal

lattice However, it is suggested that the basic steps are

quite simple, namely: (i) mixing of the reactants to give

a non-equilibrated, inhomogeneous starting material

(primary amorphous phase), (ii) equilibration to form

a semi-organised precursor (secondary amorphous

phase) containing ‘‘islands of order’’ or ‘‘proto-nuclei’’,

(iii) the establishment of sufficient regular structure

within a statistical distribution of ordered sites to enable

such structure to propagate (the nucleation step itself),

and (iv) the beginnings of crystal growth on the

estab-lished nuclei The reversible condensation reactions

which constantly make and break TAOAT bonds in

the dynamic reaction medium of hydrothermal synthesis

provide the chemical mechanism by which all these

changes are accomplished The accumulation of T-units

(e.g T(OH)3O) from solution is mediated by the

asso-ciated cations, which provide structural organisation for

the assembling architecture

8 Crystal growth

In Sections 5–7, the pathway of zeolite synthesis has

been followed from the assembling of reaction

compo-nents through to the inception of crystal growth Crucial

intermediate stages can be identified: the development of

local order and the enabling discontinuity of the

nucle-ation step Thus, nearly all the main features of zeolite

formation have already been described However, the

remaining stage—the growth of a nascent microcrystal

into a material entity—is of great significance, not least

because it is this final step which we most commonly

monitor in the laboratory or synthesis plant and which

gives us the tangible zeolite products, sorbents and

cat-alysts with which we are familiar The present section

discusses the experimental observations on this process

and their mechanistic basis

8.1 Experimental methods

In a typical zeolite synthesis, the first substantive

evi-dence for a successful reaction is the appearance of

crys-tals of the product As noted earlier (Section 5), this

signal for the end of the induction period is dependent

upon the method of detection—most commonly a bination of visual inspection or microscopy with X-raydiffraction Thereafter, crystal growth can be monitored

com-by the same techniques and the resulting S-shapedgrowth curve of bulk crystallinity against time is byfar the most commonly reported measurement of zeolitecrystallisation kinetics

Fig 12 A cautionary tale in particle size analysis The monodisperse but anisotropic silicalite crystals (a) have caused particle counters operating on conductivity (b) (Coulter Multisizer) and diffraction (c) (Malvern 3300 Laser Particle Sizer) to report a multimodal distribu- tion based on the angularity of the crystals and their hydrodynamic behaviour in suspension.

More closely related to the growth mechanism itself is

the crystal linear growth rate and to determine this some

form of temporal size measurement is necessary

Unfor-tunately, the crystals of synthetic zeolites are usually too

small (<1–20 lm) and the experimental conditions too

severe for the application of many of the traditional

crystal growth measurement techniques [108,153,185]

Nevertheless, data can be obtained, usually from

sam-ples taken during the course of small-scale experiments,

through the use of particle counters relying on (for

example) light scattering, conductivity across an orifice,

or diffraction effects However, unless the product

crys-tal size range is narrow, the result will be some form of

size distribution, which must be very carefully

inter-preted Problems are likely to arise the more the sample

deviates from a spherical shape (Fig 12) In the example

illustrated [186], the instruments have provided results

showing multimodal distributions, based on the

angu-larity of the crystals and their hydrodynamic behaviour

in suspension (When analysed by the measurement of

SEM micrographs, the sample in fact showed a more

uniform size distribution (±3%) than the glass

micro-sphere standard supplied for calibration purposes.)

More directly, the sizes of individual crystals can be

measured by optical or electron microscopy These

methods were used in the first determinations of linear

growth rates (for zeolites A and X) by Zhdanov and

co-workers [57,140] and were later applied to ZSM-5

synthesis by Nastro and Sand[187] In these cases,

mea-surements were carried out on syntheses sampled or

ter-minated at a series of different reaction times The first

in-situ growth rate determinations were made by Lowe

and associates [94,188], who used optical microscopy

to study the growth behaviour of ZSM-5and other

high-silica zeolites by direct observation of reaction

mix-tures contained in glass capillaries This work was

ex-tended by Sano, Iwasaki and co-workers using a

specially constructed cell in which a sample of ZSM-5

reaction mixture maintained at temperatures up to

170C could be directly observed using an optical

reflec-tion microscope[189,190] The procedure was later

fur-ther refined by the addition of an interferometric

technique which allowed the detailed observation of

growth behaviour on all three faces of silicalite crystals

[191]

The above methods permit the growth kinetics of

zeo-lite crystals in one, two or (in favourable circumstances)

three dimensions to be studied as a function of reaction

conditions In this way, inferences can be indirectly

drawn concerning the manner in which the crystal is

being assembled from the components available in the

synthesis mixture (Section 8.2) Most recently, data from

atomic force microscopy (AFM) and high-resolution

TEM (HRTEM) have provided more detailed evidence

of the surface construction process [192] AFM is one

of a new family of scanning probe microscopies

[193,194] In this technique, a mechanical probe with atip of atomic dimensions is rastered over the surface ofthe sample The derived signals can be processed intoimages which have a lateral resolution of
20 nm inthe principal (x, y) plane of the specimen but which aresensitive to features 61 nm in size in the vertical (z)dimension The results provide information on zeolitecrystal surfaces of astonishing detail (see below) In afurther new development, Cr-sputtered SEM imagesshowing unprecedented resolution of similar surfacetopography have recently been reported [195,196].8.2 Experimental observations—IntroductionThe majority of reports on zeolite crystallisationkinetics are based upon bulk crystallinity measurements,usually as determined by X-ray diffraction In thesecases, it is not possible to estimate crystal growth rates

in the absence of further information on nucleationbehaviour or crystal size distributions However, thereare now in the literature a significant number of investi-gations of zeolite synthesis in which crystal lineargrowth rates have been reported (Table 3) From thisinformation, it is possible to draw some conclusions

on the nature of zeolite crystal growth in relation to that

of other ionic, or partially ionic, substances

The linear growth rates (0.5Dl/Dt) of zeolite crystalsvary from about 0.1 lm h1 for zeolites A and X toaround 0.02 lm h1for ZSM-5(for near-optimum com-positions at
90 C), whilst the values for some zeolitessuch as EU-1 may be an order of magnitude lower[188].These growth rates are 2–4 orders of magnitude smallerthan typical values for simple ionic salts (such as alums

or alkali halides) or simple molecular compounds (e.g.sucrose)[208,209] This large difference reflects the nat-ure of the zeolite synthesis reaction, in which a largelycovalent, polymeric structure is being assembled piece

by piece as TAOAT bonds are formed one after other This is clearly a more complex and intricate pro-cess than the packing of relatively simple units in ionic

an-or molecular crystals where bonding is electrostatic an-orvan der Waals and relatively non-directional

The plots of crystal size against time typically show aconstant linear growth rate for most of the reaction with

a final tailing-off due to nutrient depletion, as illustrated

inFig 7b and also later in Section 11.3 (Fig 22a) Fromthis and also from the magnitude of the activation ener-gies (
45–90 kJ mol1), it is very probable that the ratecontrolling process is the surface integration step ofcrystal growth itself [33,37,188] If the reactions weregenerally dissolution-rate limited (as found for unreac-tive silica sources [210,211]), the linear growth plotswould be intrinsically curved as the reagent sources werecalled upon to provide ever increasing nutrient fluxes atdiminishing surface areas The activation energies aretypical of those for making and breaking TAO bonds

[212,213]and much higher than would be expected for a

process controlled by diffusion from the bulk liquid

phase to the crystal surface[214] This conclusion is

fur-ther supported by the work of Schoeman and

co-work-ers who applied Nielsen
s method of chronomal analysis

[215]to the synthesis of colloidal silicalite[216,217] The

best linear fit to their data was obtained by assuming a

first order surface reaction to be rate limiting

As indicated above, zeolite crystal linear growth rates

show a strong temperature dependence However, even

for a given structure type, the rates are also influenced

by other synthesis variables, particularly reaction

com-position In a few cases, systematic studies have been

re-ported, although most of these relate to the ZSM-5

system

8.3 Experimental observations—Studies of

macrocrystalline systems

Kacirek and Lechert[79,147]provided the first

quan-tification of the observation that the formation rate of

faujasite decreases with increasing silica content of the

product The limits of the series shown inTable 3

corre-spond to Si/Al = 1.4 (growth rate = 0.2 lm h1) and Si/

Al = 3.4 (0.00053 lm h1) A detailed in-situ

investiga-tion of the MFI system by Cundy et al.[188] examined

crystal length and width growth rates as a function of

temperature, concentration, cation type, aluminium

content and addition of salts (Na and TMA halides)

or ethanol At a given temperature (the most sensitive

parameter), the highest growth rates were found for

sil-icalite at moderate base levels, whilst addition of

alu-minium compounds, tetramethylammonium salts orethanol reduced the observed value

Several groups of workers have investigated the effect

of dilution and base level or pH upon growth rate Aspart of their in-situ study of silicalite crystallisation,Iwasaki et al [191] measured the growth rates at

150C of the (0 0 1), (1 0 0) and (0 1 0) crystal faces forvalues of H2O/SiO2 from 75to 300 at final crystallengths of
40 lm Rates progressively decreased ineach case, the c-axis value declining from 0.6 to0.2 lm h1 In approximately the same dilution range,other workers [188] concluded that growth rate wasnot a sensitive function of the concentration parameters,although the overall tendency was for the rate to in-crease with concentration and with increasing base.Tavolaro et al [205] studied silicalite crystallisation in

a fluoride system at 170C as a function of the pH ofthe initial gels For crystals of
300–500 lm in size, lin-ear growth rates were found to vary from 0.08 lm h1at

pH 2.6 to 1.0 lm h1 at pH 6.7 A plot of [log (growthrate)] against pH was found to be linear These resultsare in contrast to those determined for some colloidalsystems (see Section 8.4)

An interesting observation relates to the variation inZSM-5crystal aspect ratio with reaction conditions,showing that each unique crystal face may respond dif-ferently to changes in the system At constant composi-tion, major increases in the length/width ratio are seenwith increasing reaction temperature[188,206]and sim-ilar changes are seen as a function of reaction composi-

crystals elongate as the pH is reduced [207,218–220]

Table 3

Some published values for zeolite macrocrystal linear growth rates

Zeolite Temperature range (C) Other parameters investigated

whereas for a fluoride-mediated system, the aspect ratio

increases with pH in the range pH 3–7[205] Crystal

as-pect ratio may also vary during the course of a synthesis

environment as the reaction proceeds

Even if the reaction product and its morphology do

not change, variations in reaction conditions can still

ex-ert a powerful influence upon the course of a synthesis

by altering the nucleation behaviour Cundy et al

synthesis composition for silicalite at a fixed

tempera-ture (95C) by varying the order of mixing of reagents,

the type of reaction vessel, the stirring speed and the

de-gree of ageing In all cases, pure crystalline silicalite was

obtained in high yield at the same linear growth rate

[(1.95± 0.05)· 102lm h1] but there were major

changes in particle size distribution Analysis by the

method of Zhdanov and Samulevich[140]gave the

cor-responding nucleation profiles, which demonstrated

how both the yield and the size distribution of the

prod-uct at any given time are strongly linked to the treatment

of the reaction mixture In a separate study, the

chemi-cal changes surrounding the growth and dissolution of

Na,TPA-ZSM-5crystals were investigated in detail

was shown to be a natural consequence of the evolving

reaction environment and to reflect the composition of

intermediate samples taken during crystal growth

One of the very few zeolite crystal growth studies to

have been applied to non-batch preparations concerns

the synthesis of MFI-type materials under

semi-continu-ous conditions [148,186] It is demonstrated that the

process is controlled by an equation of the form

the ratio of the nutrient mass feed rate (u) to the

instan-taneous total surface area of the crystal population (A)

This relationship was explored for a wide variety of

reaction conditions Two effects are observed as the

nutrient feed rate is increased: (i) the product crystal

lin-ear growth rate (dx/dt)obsincreases in accordance with

the above expression, up to a fixed value (dx/dt)max

cor-responding to the maximum observed in a batch

synthe-sis, and (ii) there is an increasing tendency to nucleate

new crystals as ratio (dx/dt)obs/(dx/dt)max(the ‘‘effective

supersaturation’’) rises

8.4 Experimental observations—Studies of

nanocrystalline systems

In recent years, the increased use of scattering

tech-niques has led to a considerable number of

investiga-tions on the formation and growth of nanosized

zeolite particles (usually <100 nm) In general, the

ob-served growth behaviour is similar to that found for ger (>1 lm) crystals However, different trends areobserved in the effects of dilution and virtually no databeyond radial growth rates are available in view of thevery small size of the particles The vast majority of pub-lished work has been carried out on MFI systems, inview of the convenience of the available ‘‘clear solution’’syntheses

lar-Pioneering work on colloidal zeolite crystals anddetermination of their growth behaviour using dynamiclight scattering (DLS) has been carried out by Scho-eman and co-workers, including a detailed in-situDLS study at 70C [181] For colloidal silicalite

98C) were much lower than those reported earlierfor macrocrystals (e.g 20 nm h1 at 95C [142]) andwere essentially independent of alkalinity However,the growth rates approached the existing values upondilution (e.g growth rate ! 5.0 nm h1 for a factor

of three dilution and ! 9.3 nm h1 for a factor of 12

[85]) A similar dilution effect was observed by Tsayand Chiang following a determined growth rate value

of 1.6 nm h1 at 115C[221].Twomey et al.[222]have also investigated the nucle-ation and growth of silicalite using an ex-situ DLSmethod In addition to temperature variation, the effect

of ageing and change of cation was also investigated

At 96C, the observed growth rate of 18 nm h1 wasfound to increase to 42 nm h1 after an ageing period

of nine days The probable reason for this difference

is discussed below Substitution of K for Na in thereaction composition gave an almost identical radialgrowth rate A similar DLS measurement of the radialgrowth rate for colloidal silicalite at 100C by Cundyand co-workers gave a value of 11.7 nm h1, whilst fur-ther growth in a different reaction mixture using theinitial colloidal product as seed gave 25nm h1 [150].Other workers have employed X-ray scattering tech-niques De Moor et al [223] have used in-situ time-re-solved ultra-small angle X-ray scattering (USAXS) tomeasure the growth rates of Si-TPA-MFI crystals inthe 50–400 nm size range Linear growth rates at

125C of 36 ± 3 nm h1 were found for Si/OH values

of both 2.4 and 3.0

It is notable that the growth rates found in theexperiments of Twomey, Cundy and de Moor are veryclose to the expected macrocrystal values, although thereaction compositions are at first sight very similar tothat employed by Schoeman and co-workers The prin-cipal difference lies in cation content of the reactionmixture The preparation used by Schoeman and col-leagues is very low in alkali metals (Na/TPA = 0.02,

K much lower), whereas that of Twomey et al hasNa/TPA = 0.2 and the others are comparable orhigher It seems probable that the steric stabilisationconferred by the envelope of TPA cations surrounding

the colloidal particles may also cause mass transport

limitations Dilution, or the introduction of smaller

and more mobile cations may alleviate these, as well

as facilitating the process of crystal growth [150]

Some comparison with the MFI system is possible

from two studies on the crystallisation of zeolite A at

sizes up to
1 lm In a very significant and carefully

considered paper, Gora et al.[198]found radial growth

rates of 139 nm h1at 60C and 1100 nm h1at 80C

As with the related study on silicalite[222], the growth

rate of the crystallites was strongly influenced by ageing,

increasing at 60C to 334 nm h1after 10 h of ageing at

25C In this case, further investigation showed that the

particles formed from the aged solution were

agglomer-ates The anomalous apparent increase in growth rate is

thus explained, since polycrystalline particles have a

higher surface area per particle and can therefore

assim-ilate nutrient faster than single crystals with regular

pla-nar faces Otherwise, it is difficult to see why the ageing

of a reaction mixture should change the linear growth

rate of the resulting crystals, unless some

physico-chemical change has occurred to remove a kinetic

restraint—for example, the solubilisation of a nutrient

source having a very slow dissolution rate The

surpris-ingly large increase in growth rate between 60 and 80C

has not been accounted for, although the authors note

that the data at this higher temperature refer to the

lar-ger component of a bimodal population of product

crystals

Using a very similar reaction composition to that of

Gora and colleagues, Singh et al.[224] also used DLS

to measure the growth rate of zeolite A at temperatures

of 30–60C Once again, ageing was found to shorten

the induction period and increase the growth rate

Val-ues found for radial growth rates from fresh solutions

and those aged at room temperature for a few hours

were—at 40C: 15.5 nm h1(fresh), 28.8 nm h1(aged);

at 60C: 94 nm h1 (fresh), 165nm h1 (aged) The

growth rates at 60C therefore compare fairly well with

those determined by Gora et al and also with values

determined for the growth of larger crystals

(Table 3)

8.5 Size-dependent growth of nanocrystals

Very small particles have an unusually high surface

free energy associated with their extreme curvature

and as a consequence display anomalous properties

Nanosized liquid droplets show increased vapour

pres-sure The enhanced solubility of very small solid

parti-cles is given by a relationship variously described as

the Kelvin, Gibbs–Kelvin, Gibbs–Thomson or

where cris the solubility of small particles of size r and c*

is the equilibrium (saturation) solubility of large cles (i.e r! 1); M is the molecular mass of the solid

parti-in solution, r is the surface energy of the solid particle

in contact with the solution and q is its density.The significance of such a size-dependent solubilityfor zeolite synthesis would be to bring about a reducedcrystal growth rate at early reaction times when the crys-tals are very small The consequences of this have beendiscussed by Warzywoda et al [110], working from themoment equations contained in the Thompson popula-tion balance model [89–91] (Section 4.1) However, it

is possible to achieve a similar result algebraically asfollows:

For a spherical crystal of radius r, the radial growthrate is given by

70C was non-linear only below a mean crystal size of

20 nm However, whilst such correlations are ing, the agreement may be fortuitous There are noestablished values for the surface energy, and both sets

reassur-of calculations are based upon a simple model reassur-of radialgrowth with no allowance made for particulate addition

In reality, the growth units are ‘‘quantised’’ in mass andsize, so that, at the nanometre and sub-nanometre level,the increase in size is not continuous Nevertheless, itdoes seem probable that the Kelvin effect is only signif-icant in zeolite synthesis at the very earliest stages ofcrystal growth

8.6 Growth models

The slow linear growth rates of zeolite crystals at

fairly high supersaturations[148] and the rarity of the

observation of growth spirals[196,225,226]suggests that

the predominant growth mode is that of the adsorption

layer type[108], controlled by surface nucleation rather

than by the dislocations necessary for the

Burton–Cab-rera–Frank mechanism[227] A simple picture of this is

given inFig 13 Following Gibbs–Volmer theory[228],

a growth unit is adsorbed on the growing crystal face

and migrates to its optimum location (a) Following

completion of a given layer (b), no further growth can

occur in the absence of a dislocation This hiatus is

over-come by the creation of monolayer island nuclei

(two-dimensional nucleation) (c) from which the growth

process can continue on a layer-by-layer basis Clearly,

this is a greatly oversimplified representation As one

of many authors who have worked to make the picturemore realistic [108], Kossel[229] has provided a modelshowing the sites of high binding energy at whichgrowth units are most easily incorporated (Fig 14).Direct evidence for the layer growth model is emerg-ing from AFM studies (Fig 15), the growth terracessketched in Figs 13 and 14 being clearly identifiable

points on the surface of zeolite A are seen to merge, ing rise to scalloped growth fronts progressing acrossthe crystal The elliptical terraces on the surface of zeo-lite SSZ-42 (Fig 15b) also merge to give cusp-shapedgrowth fronts but in this case the shape persists, ratherthan growing out This is probably linked to a reducedgrowth rate at kink sites, a factor also responsible forthe elliptical nature of the terraces themselves In boththese cases, layer growth is proceeding simultaneouslyfrom many nucleation points distributed across the crys-tal surface However, for the small (1 lm) crystal of zeo-lite Y illustrated inFig 15c, the step pattern appears toarise from a much reduced rate of surface nucleation.Detail from a large (60 lm) silicalite crystal (Fig 15d)demonstrates the complex terrace structure in the vicin-ity of one of the ‘‘ramps’’ on the (0 1 0) face The circularterraces on the bulk (0 1 0) face show the radial decrease

giv-of terrace spacing typical giv-of the closing stages giv-of batchgrowth[230] However, terrace distortion in the vicinity

of the ramp suggests that the ramp corners have acted asnucleation centres It is also possible that the intrudingsurface is that of a (1 0 0) face, i.e a 90 intergrowth

[231] In all the above micrographs, the step heights respond to identifiable macro-features of the unit cell,for example the thickness of a pentasil chain (1.0 nm)

cor-in the silicalite image

Fig 13 Zeolite crystal growth via a layer-by-layer mechanism A

growth unit is adsorbed on to the growing crystal face and migrates to

a high-energy kink site at which the number of points of attachment

are maximised (a) Following completion of a layer (b), further growth

can occur only after the creation of monolayer island nuclei (c) (after

Mullin [108] ).

Fig 14 Kossel-type model of a growing crystal surface, showing a surface-adsorbed growth unit (D) on a flat surface (A) and also steps (B), kinks (C), an edge vacancy (E) and a surface vacancy (F) The kink site (C) has the highest binding energy and is thus the most likely point of attachment for a growth unit (after Kossel [229] and Mullin

[108] ).

Using a representation of the Gibbs–Volmer type,

Agger and colleagues have modelled the growth of

zeo-lite A in terms of the features seen in AFM images[232–

234] Growth probabilities are assigned to the potential

growth sites (edge, kink, etc.) and the relative values for

these probabilities are adjusted until the computed

unit-by-unit growth pattern matches that observed

experi-mentally The information gained by this procedure at

present provides no indication of the nature of the

growth units in the real system However, it is found

that, to match the observed features, growth at kink

sites has to be 500 times faster than growth on the clean

surface and 15times faster than growth at an edge site

It is also apparent that the overall advance of a growth

face must be controlled by the surface nucleation rate,

the lateral spreading rate being much faster The

chal-lenge for this type of approach is to forge a link between

the parameters used to fit the experimental data and

their chemical basis, as has been achieved in the

simula-tion of ice crystal growth by Wathen et al [235] An

early attempt to describe zeolite crystal growth by a

chemically significant model is summarised inAppendix

A

A very interesting finding from recent AFM studies

has been the observation of steps on the surface of

zeo-lite A corresponding to the sizes of a double-4-ring(D4R) [196,236] These have been interpreted [236] assuggesting that the D4R unit is the key building unitfor crystal growth in zeolite A The authors speculatethat ‘‘first Si, Al and O bond together to form theD4R, and then D4R units bond together, like buildingblocks
, to form the LTA structure’’—an idea stronglyreminiscent of the early suggestions of Barrer and ofBreck (Section 3.2) An alternative possibility is put for-ward below (Section 8.2), where this case would provideone example of the cation-mediated ‘‘in-situ’’ construc-tion of a microporous framework The observation of

a surface terminated by double-4-rings would suggestthat this configuration represents a metastable interme-diate in the layer-by-layer construction process, i.e (i)slow 2D nucleation of a new surface layer, (ii) relativelyrapid lateral spread of that layer, (iii) slow nucleation of

a further surface layer, etc On a time-averaged basis,the growing structure would be most often observed instate (ii), which would correspond to the AFM result,although this does not explain why the surface shouldapparently terminate in D4R units rather than sodalitecages A possible rationalisation for this would be thatthe cations are more tightly held in the square prisms,rendering these units relatively more stable (less soluble)

Fig 15 AFM images of zeolite crystal surfaces: (a) zeolite A, (b) SSZ-42, (c) zeolite Y and (d) silicalite Scale bar below each image = 1 lm See text for discussion (We thank Itzel Meza and Dr J.R Agger of the University of Manchester Centre for Microporous Materials for these images.)

in the reaction medium Similar work by Wakihara et al.

[237] was carried out on faujasite by observing crystals

isolated from seeded growth in dilute solution From a

seed at near equilibrium with the solution, the initially

rough surface was found to have become a well-ordered

(1 1 1) face For the solution-seed growth environment,

most of the top surfaces of the crystals were found to

be terminated by D6R units, while some presented

com-plete or incomcom-plete sodalite cages These results were

taken to show that aluminosilicate species equal to or

smaller than a 6-ring contributed to the crystal growth

8.7 Mechanism

Some of the main observations relating to the growth

of zeolite crystals have been summarised above The

question now arises as to the nature of the processes

occurring at an atomic and molecular level which are

responsible for the observed behaviour In general,

par-ticles may increase in size in two ways: (i) by addition of

growth units, and (ii) by aggregation The latter is a

spe-cial case of the former in which the particles added are

of comparable size to the existing particle, whereas

growth units are assumed to be incremental in terms

of size Whilst there is abundant evidence for the

aggre-gation and coalescence of zeolite macrocrystals during

their growth[146,148], the predominant growth

mecha-nism clearly involves the step-by-step addition of much

smaller units In recent years, far more work has been

carried out on the growth of zeolite crystals of colloidal

size than on their larger counterparts For these

nano-scale situations, the distinction between the two modes

of size increase is far less clear cut In the following

dis-cussion, it is assumed that growth is taking place on

macroscopic zeolite crystals, i.e those of maximum

dimension J 0.5 lm, and that the growth mechanism

is predominantly one of unitary addition Furthermore,

the growth species are taken to be small solution units,

predominantly the silicate and aluminate monomeric

anions Si(OH)3O and AlðOHÞ4 Other possibilities

are later treated in more detail (Section 9) At present,

the primary concern is with the chemistry of the growth

process

In Section 7.8, there is described a long-recognised,

very simple but fundamental equilibration mechanism

of TAOAT bond making and bond breaking, based

upon condensation and nucleophilic displacement

reac-tions This facilitates dynamic structural modification

through a hydroxl ion catalysed

polymerisation–depoly-merisation of the silicate or aluminosilicate gel network

Further structuring is achieved principally through the

absorption of cations at the surface and their reaction

with small growth units from the solution phase to

con-struct ordered regions During this cation-mediated

assembly process, some or all of the water in the

coordi-nation sphere of the cation becomes replaced by silicate

units In the zeolite synthesis process, the first importanttransformation brought about by this means is the con-version of the initial random gel structure (primaryamorphous phase) into a semi-ordered network (second-ary amorphous phase) In the next stage, the generation

of heterogeneous nucleation sites is brought about inthe same way The third manifestation of the equilibra-tion/cation-structuring mechanism is in crystal growth.This can follow an essentially identical pattern to thatshown inFig 11except that the process must be capable

of self-perpetuation However, this is a case of ‘‘successbreeding success’’ since the more perfect the (n 1)thstructure, the easier it is to construct without error thenth repeat

The sequence is therefore as follows A cation isenvisaged as docking at a surface site on an amorphousparticle Monomer units from solution then become at-tached to the growth site and are assembled in an or-dered array around the cation to create a cyclic unit.(Note again that the depiction of hexagonal units in

Fig 11is illustrative only and has no particular chemicalsignificance.) It is a requirement that, in the course ofthis process, a new cation site is generated so that thecoordination steps will iterate and the cycle continue.Gel equilibration, nucleation and crystal growth thusshare a common mechanism However, as the structuredevelops, different stages of the overall growth processmay have different activation energies and rates.8.8 Summary

Zeolite crystals appear to grow rather slowly, pared to ionic crystals (such as common salt) or molec-ular crystals (such as sugar) The reason for this lies inthe necessity to construct a semi-covalent three-dimen-sional lattice: a polymer of ‘‘TO2’’ The predominantgrowth mode is that of the adsorption layer type[108],where the overall rate is controlled by the surface inte-gration step and nucleation of a new layer is slower thanthe lateral spread of such a layer

com-The overt features of equilibration, nucleation andcrystal growth have been discussed in terms of an overallmodel for zeolite crystallisation The pre-requisites forthis mechanism to operate are:

1 The interactions are mediated by small solution unitssuch as silicate monomer (see Section 9)

2 The mechanism relies on an in-situ, localised struction process rather than the assembly of large,pre-fabricated units as required by some growth the-ories Also, not all the growth units forming the finalcrystal may have arrived by solution transport In asituation in which the amorphous phase is both thenutrient reservoir and host of nucleation sites, somelocal reconstruction of the gel particle may becomebuilt in to the final product (Fig 16) If there are

con-several nucleation sites on a single gel particle, this

situation must inevitably occur where 2 or more

growth areas converge and unite

3 A key feature of the postulated mechanism is that the

assembly process at the growth points is subject to

the organising influence of polar structuring agents,

usually cations

4 The system must be dynamic, with the chemical

reac-tions reversible In this way, errors which do occur in

the overall process can be corrected and the structure

can propagate Similarly, there must also be a

mech-anism for redissolution, although this may not

necessarily be the simple reverse of the synthesis

reaction

The above proposal is compatible with the known

experimental data and also with many of the published

ideas for synthesis mechanism It additionally provides

an explanation for the formation of zeolites from

poly-meric templates [31,238–241], where macromolecules

extend through, and are encapsulated by, many

frame-work unit cells With a localised construction process,

a single crystal can be assembled around

surface-adsorbed polymer molecules, convergent growth

proba-bly occurring at several sites simultaneously The

suggested mechanism also does not completely exclude

earlier proposals of key roles for particular building

units (e.g specified rings or polyhedra) although these

are regarded with caution (see Section 9) However, if

any particular units do have especial significance forthe genuine chemical construction of certain zeolites,then it may be because such units are ordered and assem-bled at the growth site by the participating cations Thus,the mere occurrence of such units in the solution phase(e.g as detected by NMR spectroscopy) is unlikely to besignificant

A complicating factor is the possible duality ofgrowth by (i) unitary addition (as above) and (ii) particleaggregation However, this is unlikely to introduce anynew chemical features, being more a variant on the basicprocess described above A full discussion is given inSection 9

Part III: KeytopicsThe explanation given in Part II (Sections 5–8) is con-sidered to represent the most likely mechanism for zeo-lite formation in the majority of hydrothermal zeolitesyntheses It is self-consistent and chemically plausible.The detail of individual steps is already well establishedand no new or special conditions have to be invoked.However, some degree of qualification may be requiredfor unusual reaction situations Also, in setting out thesesuggestions, other views and some associated topicsrelating to the nature of the synthesis mechanism haveoften been mentioned only in passing It is the purpose

of Part III of this Review (Sections 9–18) to deal morefully with these issues

Fig 16 Progress from a gel particle to crystalline zeolite (see also Figs 5, 10 and 11 ) From an initially amorphous structure (a), areas of local order are established (b), some of which develop into crystal nuclei (c) and grow by acquisition of building units from solution (d) To provide such growth units, amorphous material is dissolved (e) to supply nutrient to both distant (e-i) and nearby (e-ii) growth sites The distant site could be located on another gel particle For the nearby site, crystal growth may reduce in the limit to a local reconstruction of the host gel particle Eventually, all amorphous material is converted into an approximately equal mass of zeolite crystals (f).

9 The nature of growth species and the role of

aggregation processes

In Section 8.6, mass increase by the unitary addition

of simple solution species was taken as the most likely

growth mechanism for macrocrystals (J 0.5 lm)

How-ever, it was recognised that aggregation processes may

also play a significant role, especially for very small

crys-tallites In this case, the growth units are not soluble

spe-cies but particles of comparable size to the existing

particle The ‘‘in-situ construction’’ process proposed

for macrocrystal growth provides a realistic mechanism

for the generation of the periodic crystal lattice It is less

easy to see how such a regular, and essentially perfect,

structure can be formed by the addition of large blocks

of new matter, especially if such material is wholly or

partly amorphous Grain boundaries and residual glassy

areas would presumably be observable in the product

The nature of the species which carry mass to the

growing crystal is not known with certainty and the

res-olution of this issue presents a challenging problem The

reason for the difficulty lies partly in the dearth of in-situ

micro-techniques for probing the chemistry of the

boundary layer at the crystal surface but also in the

complexity of silicate sol chemistry, which offers such

an array of candidate species The following paragraphs

examine first the arguments for crystal growth by the

addition of both simple and also more complex solution

species Since it is tacitly assumed in these discussions

that the synthesis in question is of the traditional

OH-mediated type, there follows a short discussion

of other possibilities, for example systems based on

fluo-ride ion Finally, the features of growth by aggregation

are evaluated and the consequences of such a

mecha-nism are considered

9.1 Growth from soluble, pre-fabricated units

There exists a structural similarity between some

solution species and the building units which appear

in zeolite frameworks Several groups of workers have

accordingly adopted the idea that a zeolite containing

(for example) double-5-rings is constructed by the

extraction and polymerisation of these units from

the distribution of such species available in solution,

the equilibrium then rapidly shifting to make up the

deficit Although attractive at first sight, this theory

does not survive close examination Prominent

sup-porters of this idea, especially as applied to MFI

syn-thesis [242], have given an excellent account of its

genesis and subsequent demise [243], whilst the more

general conjecture of zeolite growth ‘‘from secondary

building units’’ (SBUs) has likewise been thoughtfully

analysed and rejected [244]

The above papers[243,244] should be consulted for

their detailed analysis but the main points are that: (i)

the real structure of silicate and (where known) nosilicate anions in solution differs considerably fromthat of most SBUs, (ii) there is no correspondence be-tween the occurrence of particular solution anions andthe nature or formation rate of a particular zeolite,(iii) zeolites of widely different structure can be formedfrom solutions containing apparently similar distribu-tions of anions, (iv) complex zeolite structures areformed in situations (high dilution, high temperature,high levels of Na and K) where the solution anionsare known to revert to the simplest species, (v) com-ponents can be added which change the course ofthe synthesis but do not materially alter the anion dis-tribution (e.g the appearance of MAZ-structuredproducts in zeolite Y synthesis upon addition ofTMA cations), (vi) structural predictions (e.g on thedensity of defects) made on the basis of an assumedD5R precursor in MFI synthesis are found to beincorrect

alumi-It is interesting that the 29Si NMR assignments ofdouble-ring structures which gave rise to some of theoriginal speculation[245]are now believed to be errone-

com-pletely rejected since there is (as in catalysis) alwaysthe possibility of an active component present at a low(and conceivably undetectable) level

Szostak [247] has introduced an interesting variant

on the ZSM-5synthesis mechanism proposed by vanSanten et al [242] The original suggestion consideredformation of ZSM-5via ring opening and polymerisa-tion of double-5-ring silicate anions (Fig 17a) In themodified version, emphasis is placed on the role of uni-versally present free silicate monomer species (Q0),which are largely responsible for further growth once

a single or dimerised D5R unit has been formed

growing chains and sheets would be made up fromthe solution phase equilibrium of silicate species, as

in Thompson
s ‘‘tugging chain’’ mechanism (Section9.2) This modified scheme has much in common withthat outlined in Section 8 and is altogether morecredible

A more recent revival of the idea that absolutely cific units are required for the synthesis of particularzeolites is seen in the nanoblock hypothesis of the Leu-ven research group[67–73](Section 3.7) Both the exper-imental observations and their interpretation are nowthe subject of active debate [86–88] Whilst it is notinconceivable that such a prescribed mechanistic coursemay be followed under a particular set of experimentalconditions, it seems overwhelmingly unlikely that such

spe-a situspe-ation represents the generspe-ality Pre-nuclespe-ationbuilding units (PNBUs) are also the key feature ofmechanisms proposed for the formation of new open-framework materials discovered recently by Fe´rey

et al (Section 18)

Variants of the above hypothesis include those

founded on topological considerations One such

suggestion (by Melchior et al [248]) is based primarily

on 29Si MAS NMR spectroscopic data and argues

from aluminium-ordering patterns that the

fauja-sites examined must have been assembled by the

pathway

4R! D6R ! FAU;

where the single four-ring (4R) is seen as the secondary

building unit leading to the double six-ring (D6R),

which itself is the immediate precursor to the FAU

lat-tice Similar schemes for the formation of zeolites A, Y

and other aluminous types by the assembly of ring unitsare based on spectroscopic studies of synthesis solutions,for example by29Si and27Al NMR spectroscopy [249],Raman spectroscopy [250–252] and UV Raman spec-troscopy [253] Although such arguments are thought-provoking, they lack chemical detail and suffer fromsimilar objections to those mentioned above since there

is not necessarily any direct link between the existence ofspecies in solution and the nature of solid phases whichcrystallise from those solutions The hypothesis of in-situ construction at the crystal surface in a reaction-crys-tallisation process (Section 8.7) does not require theexistence of pre-fabricated building blocks What isneeded is made as the structure is generated and many

Fig 17 Suggested mechanisms for the formation of zeolite ZSM-5: via (a) ring opening and polymerisation of double-5-ring silicate anions [242] , (b) nucleation by the SBU followed by growth through addition of monomeric TO 2 units [247] (Adapted from Ref [247] )

of the arguments for the presence of particular precursor

units can be adapted to this reaction-at-growth-site

model As Thompson[37]has pointed out, ‘‘the fact that

the solution has the thermodynamic tendency to make

small oligomeric structures similar to the final crystal

structure also suggests that material having the same

structure can be assembled at the crystal–solution

inter-face, i.e that the surface reaction would be expected to

create similar crystalline material.’’

9.2 Growth from simple species

A more credible idea for zeolite crystal growth is that

it occurs predominantly from simple units, especially

monomer, although all solution species may contribute

to some extent [37,94,142,188,222,244,254,255] Whilst

it is impossible (at least for the present) to provide

unequivocal proof of this theory, the circumstantial

evi-dence in its favour can be summarised as follows:

1 The in-situ construction hypothesis (Section 8.7) does

not demand the presence of any specialised building

blocks, requiring only that the simple solution silicate

species re-equilibrate very rapidly—as is known,

pre-dominantly from NMR spectroscopy, to be the case

[256,257]

2 From (1), it is apparent that solution species which

are not used directly are consumed indirectly in the

re-equilibration process This is very well expressed

in Thompson
s ‘‘tugging chain’’ model [37] which,

in common with the mechanism suggested earlier

(Sections 7.8 and 8.7), proposes that elements of the

final crystal structure are assembled at the crystal–

solution interface from solution species: ‘‘ of the

myriad silicate oligomer species present at the

crys-tal–solution interface, there may be only a few which

are incorporated, e.g monomers and dimers The

other, larger, more complex species are continually

unraveled, at a relatively fast rate, to maintain the

equilibrium distribution of oligomers, or very near

to it Thus, it is quite possible that the whole process

is governed by the ordering of silicates around the

pertinent template species adsorbed at the crystal

surface.’’

3 Under semi-continuous synthesis conditions in which

silicate monomer is by far the dominant solution

spe-cies[148], crystal linear growth rates are directly

pro-portional to the rate of supply of nutrient solution

and the crystals formed are of a very high degree of

crystallinity

4 As noted earlier (Section 3.5), Kacirek and Lechert

[79] interpreted their detailed kinetic studies on

seeded faujasite syntheses in terms of a solution

growth model, the rate-determining step being the

connection of silicate species to the surface of the

crys-tal They noted that, under their conditions, the

solution phase would contain essentially only mers and dimers during the crystallisation of zeolite

mono-X, with higher oligomers (perhaps up to Si20) present

in the synthesis of the more siliceous Y-types Sincethese two zeolites share a common structure (FAU),

it is unlikely that the construction process differsmarkedly in the two instances Hence, the likelihood

is that simple units are the growth species in bothcases On the same basis, it could be argued thatthe growth rate should be proportional to the frac-tion Sin/Sitotalof simple units in the reaction mixture(where n = 1 or 2) This is at least qualitatively true,since the calculated linear growth rate decreases asthe Si/Al ratio (and hence degree of silica polymerisa-tion) increases[79]

5 For the case of aluminate, there is no doubt that thebuilding block is AlðOHÞ4 monomer since under thenormal conditions of synthesis this is the only alumi-nate species present [258,259] (with the assumptionthat there is no direct participation of aluminosilicatepolymers)

6 In a simple, two-dimensional computer model of lite growth from poly-TO2 solution species (see

zeo-Appendix A), it has been shown that only simple cies are viable for network propagation at realisticrates Larger and more complex units incurred a timepenalty and led to potentially fatal faults as all thestatistically possible docking combinations weretried Only a few of these (depending on the symme-try and conformation of the unit) were successful inpropagating the structure With monomer as thegrowth unit, every docking that becomes chemicallylocked must be successful Every potential growthunit will have (in given circumstances) a sticking fac-tor and an error probability In the case of monomer,the latter is zero

spe-9.3 Mineralising agents other than hydroxideThis account of the background to the synthesischemistry of zeolites has concentrated almost exclusively

on OH-based systems This is partly because they arethe most common but principally because so little de-tailed information is available on any of the other con-tenders For a synthesis system to operate successfully,

it must contain some agent (the ‘‘mineraliser’’) whichperforms the following functions, acting in many ways

as a form of catalyst: (a) it must convert the startingmaterials into mobile forms (for example by taking theminto solution or the vapour phase), (b) it must conveychemical reactivity such that the mobile units (e.g sili-cate anions) can react together to form new chemicalbonds and generate the zeolite framework, and (c) itmust de-complex from the mobile units during or afterstage-(b) so that the new structure can exist as a stable

solid phase The hydroxide ion performs these tasks very

effectively

The most viable alternative to OHas a mineraliser is

the fluoride ion Introduced into zeolite synthesis by

Flanigen and Patton[260], it came to prominence only

more recently, largely due to the pioneering research

programme of the Mulhouse school [261–265] Unlike

the traditional synthesis in basic media, the

fluoride-mediated route can operate over a very wide range of

pH values The ability to function under acidic

condi-tions has proved advantageous for the introduction of

heteroelements which form unreactive precipitates in

alkaline solutions Although many zeolites and zeotypes

can be prepared by either synthetic pathway, in some

cases the fluoride method provides the only route There

is a tendency for such products to contain some

struc-tural fluoride and this may perhaps account for the

importance of its presence in the reaction mixture A

further advantage of the fluoride route is its tendency

to furnish large (in zeolite terms) crystals of the product

This suggests that the mineralising power of Fis less

than that of OHso that solubilities and effective

super-saturations are lower, a point also indicated by the

higher temperature and longer reaction times usually

necessary for fluoride-based syntheses Useful parallels

between the hydroxide and fluoride routes have been

set out by Guth et al [261], who have also provided

some insight into the synthesis chemistry Using an

elec-trode specific to F, they concluded that fluoride

syn-thesis media probably contained the species Si(OH)3F

and Al(OH)3F, with octahedral coordination (SiF26 ,

AlF36 ) becoming more prominent as Fconcentration

increased [263] More detailed investigations,

particu-larly using multinuclear NMR spectroscopy, have been

carried out on the syntheses of zeotype materials

(Sec-tion 18.2)

The only other class of mineralising agents to have

re-ceived any significant attention are chelating oxygen

li-gands such as o-dihydroxy aromatic compounds and

glycols These are well-known complexing agents for

sil-icon [172,266,267] Pyrocatechol (benzene-1,2-diol) has

been used to control the crystal size in silicalite

synthe-sis, the presence of a silicon–benzene-1,2-diol complex

being detected by 29Si-NMR spectroscopy [268] The

same pattern was found in the silica-sodalite[269]

sys-tem[270], although in this case the ethylene glycol

sol-vent alone is also known to form 5-coordinate

silicoglycollate complexes in the reaction solution[271]

9.4 Growth from particles

9.4.1 Particle aggregation

Several authors have postulated mechanisms for

zeo-lite nucleation and growth which involve the

aggrega-tion of particles Much of this speculaaggrega-tion arises from

recent work on ‘‘clear solution’’ syntheses in the

TPA-silicalite system, using advanced scattering and tion techniques The basic argument is that primarysub-colloidal particles with diameter 3–4 nm aggregateand ‘‘densify’’ in a series of steps to give clusters with

diffrac-a size of 6–7 nm diffrac-and diffrac-amorphous to X-rdiffrac-ays After this,the densified clusters aggregate to particles of around

50 nm in size which display X-ray crystallinity The mary particles correspond to the hydrated tetrapropy-lammonium silicate complexes identified by variousauthors [63–66,128,272] Aggregation models are de-scribed by Dokter[273]de Moor[274,275]and co-work-ers, Regev et al [129], Nikolakis et al [276], Corkeryand Ninham[277]and the Leuven group[67–73] A rep-resentative schematic illustration of one proposed aggre-gation model is given inFig 18

pri-Much of the evidence for growth by aggregation inthe studies outlined above is based upon the detection

of variously sized particle populations at different stages

of the synthesis process, although one group has alsomade AFM measurements of interactive forces [276].Extensive theoretical analyses have also been under-taken by three different research groups but with dis-agreement in their conclusions DLVO calculationswere carried out to ascertain whether or not particlesmigrating to a crystal surface have sufficient energy toovercome the repulsive energy barrier at the liquid–solidinterface and attach themselves Schoeman [278] deter-mined that the thermal energy of the colloidal particles

at crystallisation temperature (373 K) was not sufficient

to overcome the net repulsive energy between the tively charged particles Since similar results were foundfor particles of both comparable (3 nm) and very differ-ent (3 nm, >10 nm) sizes, it was concluded that theaggregation model was not a reasonable description ofmolecular sieve growth Kirschhock et al.[71]calculatedthat the energy barriers were net repulsive at room tem-perature but furnished a net attractive secondary energyminimum at
7 A˚ from the surface at 373 K In thework of Nikolakis and co-workers [276], the potentialenergy was net repulsive but a fraction of the particleswere estimated to have sufficient diffusional energy toovercome the energy barrier and subsequently attachrapidly to the crystal surface

nega-Most recently, Erdem-Sßenatalar and Thompson[279]

have used the Matijevic´ model for the aggregation ofcolloidal particles [280,281] to assess the likelihood ofsilicalite growth by ‘‘nanoslab’’ addition in clear solu-tion synthesis It was found that the original, diffusion-limited model predicted a very much faster process thanthat observed experimentally Modifications to accountfor surface-reaction growth limitations alone also failed

to simulate the system, showing slowed growth as thenanoslabs were depleted A further modification to al-low for a distribution of nanoslab structures gave someimprovement but did not reproduce the observed sud-den termination of crystal growth All simulations pre-

dicted a broadening of the particle size distribution over

an extended period, a feature which is not observed

experimentally The authors concluded that MFI

growth by nanoslabs could not be ruled out, although

none of the simulations considered were able to capture

all the features observed by various research groups

9.4.2 Chemical and physical consequences of an

aggregation mechanism

No detailed explanations have been provided as to

how a ‘‘growth by aggregation’’ model operates in

chemical terms As regards the particulate behaviour

of the reaction sol, a degree of agglomeration of

sub-colloidal or sub-colloidal amorphous matter seems very

probable However, the aggregation of amorphous

with crystalline particles[276] as a growth mechanism,

or the agglomeration of non-viable nuclei with other

growing nuclei [222], are more problematic If some

sort of direct amalgamation is implied, there has to

be a mechanism for the ordered formation of the

nec-essary chemical bonds, or the resulting product would

contain amorphous areas or extensive defects There

are at least two possibilities for the case in which an

amorphous particle adheres to one that is partly or

wholly crystalline Firstly, the aggregate may unify

and become more crystalline overall by the usual

equil-ibration mechanism of TAOAT bond making and

breaking (Section 7.8) Alternatively, the aggregate

may behave as a nutrient-recipient pair, with the

amor-phous particle depolymerising to primary growth units

which then feed growth sites on the adjacent (or

con-joined) crystallite The latter situation may form part

of the normal route for mass transfer in some systems

and would be consistent with Thompson
s ‘‘tugging

chain’’ mechanism [37] (see Section 9.2) It will be

appreciated that these cases share many features with

the generalised picture of nucleation and growth

illus-trated earlier inFig 16

Direct experimental evidence relating to the

signifi-cance of aggregation processes comes largely from

HRTEM studies Thus, a weakness of an aggregation

growth mechanism involving additive entities of

5nm in size has been highlighted by Davis[282] This

is its inability to account for the existence of ordered

crystalline intergrowths—a fact which is readily

ex-plained by a normal layer-by-layer mechanism of crystal

growth (Section 8) A high-resolution TEM study on

nanoclusters of zeolite L [283] has verified that each

crystallite has a single nucleation centre as its origin

Larger crystals possess grain boundaries, each one a

new domain resulting from an isolated nucleation event

and not via the agglomeration of previously discrete

col-loidal crystallites A similar study on colcol-loidal silicalite

the particles were
50 nm in size but composed of 5–

10 nm single-domain sub-particles The 50 nm silicalite

product was believed to have grown from an ation of 5–10 nm particles (partially ordered TPA+–sil-ica–water ‘‘cubosomes’’) that crystallised to becomesyntaxial only after aggregation There were no signs

agglomer-of defects Some agglomer-of the 50 nm particles were fused gether in open aggregates of 50–100 units with no pre-ferred orientation The fused regions showed a poorlydiffracting glassy structure a few tens of A˚ ngstromsthick It thus seems probable that many of the possibil-ities outlined above do occur in practice, namely (a)amalgamation and subsequent crystallisation of partlyordered particles, (b) grain boundary repair to give over-all crystallographic alignment, and (c) some random fu-sion of crystallites, with glassy boundaries

to-Other experimental observations which add further

to the understanding of the processes taking place arefound in synthetic studies In the preparation of colloi-dal MFI-type materials, a very noticeable transition inmorphology occurs as the synthesis temperature israised [150,284] At temperatures of around 90C,the typical morphology is that of poorly defined

Fig 18 An aggregation model for the formation of silicalite from a clear sol: (a) silicate/TPA clusters in solution, (b) primary fractal aggregates formed from the silicate/TPA clusters, (c) densification of the primary fractal aggregates, (d) combination of the densified aggregates into a secondary fractal structure and crystallisation, (e) densification of the secondary aggregates and crystal growth [37,273] (Redrawn from Ref [273] with permission.)

spheroids which, although seeming to behave as single

entities (e.g in the size analysis of sols), appear at high

resolution to consist of aggregates of smaller particles

synthesis temperature is raised, the morphology of

the products becomes better defined At 150–175C,

the nanoparticles are clearly discreet crystals with

well-developed facets (Fig 19b) As far as can be

ascer-tained from limited data at intermediate temperatures,

there is no abrupt discontinuity in behaviour Thus, it

seems likely that the observed pattern is the result of

two competing processes: (a) an aggregation of small

and mainly amorphous particles into larger units, and

(b) the sintering together of agglomerated particles into

a macrocrystal, which is all the time becoming more

crystalline and also growing larger Process (a) is

prob-ably governed by the aggregation processes of colloid

chemistry and is driven by interparticle forces rather

than the formation of chemical bonds Process (b)

re-lies on the ubiquitous reversible mechanism of TAOAT

bond formation discussed earlier and will include the

acquisition of further growth units from solution

The aggregation processes will occur at low

tempera-ture and may even become less effective as the

temper-ature is raised Conversely, the true crystallisation

reaction will be strongly encouraged by increasing

temperature

The temperature-dependent morphology observed in

the above MFI syntheses is clearly closely related to

the HRTEM results mentioned earlier A 2-component

growth mechanism was discussed by Schoeman in his

detailed study of aggregation and growth processes

[278] Both a quantitative treatment of colloid stability

and also extensive experimental results on the synthesis

and properties of colloidal zeolites were considered It

was concluded that a reasonable postulate for the

nucleation of silicalite is via a cluster-cluster

aggrega-tion mechanism until a certain size (in this case

possi-bly less than 1 nm) is attained, whereafter the growth

mechanism is best described as being via the addition

of low molecular weight species, most likely the

monomer

9.5 Summary

The most generally applicable crystal growth model

for zeolite systems is that based on mass gain from

sim-ple species (predominantly monomers) in solution

Fol-lowing nucleation on or near the surface of an

amorphous particle, the crystalline region is extended

by the acquisition of growth units from solution These

are replenished by the adjustment of solution equilibria

and the dissolution of amorphous (or less ordered)

material The identity of the growth species will depend

on the reaction chemistry (e.g OH- or F-based) but

the same general principles should apply to everysystem

At early synthesis times in colloidal systems, discretenanoparticles may agglomerate and then grow to-gether, in some cases leaving detectable grain bound-aries in the final crystalline particle However, acrystal growth mechanism involving growth by addi-tion of amorphous or semi-ordered colloidal particles

is not, by itself, realistic: a depolymerisation or in-situreconstruction step involving mobile growth units isalso necessary for the generation of an ordered periodicstructure The alternative growth mechanisms (directlyfrom solution or through aggregation and subsequentordering) are linked by the common chemistry of thereversible TAOAT bond-making and -breaking equili-bration reactions Investigation of the nature of thespecies which carry mass to the growing crystal is achallenging problem, which may in due course be re-solved by (for example) a combination of AFM andRaman microscopies

10 Solid state transformations

In earlier parts of this work, zeolite synthesis hasbeen described in terms of a solution-mediated reac-tion-crystallisation process, as shown in Fig 2 Amor-phous precursor material is dissolved and the zeoliteproduct crystallises from the resulting solution Thismodel has been directly demonstrated by the Kerrexperiment [55,56] (Section 3.3) in which the dissolu-tion, solution transport and crystallisation stages arephysically separated It provides the most generally sat-isfactory mechanism by which the experimental obser-vations of zeolite nucleation and crystal growth can

be explained However, there are circumstances inwhich this model appears unrepresentative For exam-ple, zeolites can be synthesised by exposing a suitablesolid precursor mixture to a vapour source (steam oramine) at elevated temperature (see Section 10.3) Inthis technique, a solid precursor is converted to zeoliteproduct in the absence of a bulk liquid phase The exis-tence of this and other, less extreme, cases of appar-ently direct conversion of amorphous componentsinto crystalline products has led to the postulation of

an alternative ‘‘solid state transformation’’ mechanismfor zeolite synthesis

The implication is of an internal, bond-switchingrearrangement from amorphous to crystalline material,although the nature of this is frequently ill-defined andhas never been the subject of a detailed, chemicallyspecific description or illustration In Sections 10.2–10.6, the justification for such a postulate is investi-gated A useful reference document for the generaltopic of heterogeneous events in zeolite crystallisation

has recently been provided by Serrano and van

Grie-ken [39]

10.1 Hydrothermal synthesis in the presence of a liquidphase

The original suggestion by Flanigen and Breck [51]

of a predominant role for the solid phase in a normal,liquid slurry zeolite synthesis has been examined else-where [31] and has been shown to reflect a participa-tive, but not exclusive, function for the solid phase(Section 3.2) The results of McNicol et al [75,76]

were taken to be indicative of a zeolite crystallisation

in the solid gel phase on the basis that changes could

be seen in the isolated and dried solid phase but not

in the separated liquid However, as is clear fromthe subsequent and more detailed study of Angelland Flank [77], there is no conflict between theobservations of McNicol et al and either the views

of Flanigen and Breck or the reaction model given

in Part II, so that either scheme could account forthem

More recently, other groups of workers haveproposed significant roles for solid-state transforma-tions or gel rearrangements in zeolite synthesis [58–

the existence of identifiably distinct ‘‘liquid phase iontransport’’ and ‘‘solid hydrogel transformation’’ pro-cesses on a wide variety of measurements (Section 3.6)but principally upon chemical and thermal analysesand SEM results[58–62] However, they are comparingtwo (and only two) syntheses from different startingmaterials and with widely divergent compositions: sub-stantial dissimilarities are not unexpected and can berationalised in terms of differences in the partition ofreaction components between solution, macroscopicgel and colloid phases (Section 7.7) The conclusions

of Gittleman and co-workers for silicalite synthesis

spectro-scopic and chemical studies, affording a picture in whichthe condensing silica network encapsulated hydrophobicTPA cations in cages having only short range order andinaccessible to ion exchange Upon heating, the cageswere believed to rearrange through the breaking and ref-ormation of siloxane bonds into the more stable silicalitestructure However, the presence of ion exchange selec-tivity in amorphous zeolite precursors has been observedpreviously[122,290]and does not necessarily lead to theconclusion of a ‘‘hydrogel-solid transformation mecha-nism’’ Quaternary ammonium templates can certainly

be strongly bound in amorphous precursor complexes

[117,291]and this is compatible with a subsequent tion-mediated growth mechanism [64–66] In the latest

solu-[292] of a series[39,293] of studies in which solid–solidtransformations are postulated to play a prominent role,Serrano and associates conclude that the synthesis ofTS-2 from clear liquid gels occurs by a dual mechanism(cf Section 9.4.2) The initial stage operates by a reorga-nisation-agglomeration mechanism and only when a

Fig 19 Morphological differences seen in particles of colloidal

silicalite synthesised using conventional heating at (a) 90 C (scale

bar 50 nm) and (b) 150 C (scale bar 100 nm) The same batch of

reaction mixture (age 5days) was used for both experiments.