Direct measurements of the crystal growth rate and nucleation behaviour of silicalite, a zeolitic silica polymorph

Journal of Crystal Growth 100 (1990) 189—202

189

North-Holland

DIRECT MEASUREMENTS OF THE CRYSTAL GROWTH RATE AND NUCLEATION BEHAVIOUR OF SILICALITE, A ZEOLITIC SILICA POLYMORPH Cohn S. CUNDY Research and Technology Department,

ICI

Chemicals and Polymers Ltd., The Heath, Runcorn. Cheshire, WA 7 4QD. UK

and Barrie M. LOWE and Douglas M. SINCLAIR Deparonent of (hem istrr. Unicersity of Edinburgh, West Mains Road, Edinburgh EH9 3JJ. Scotland, UK Received 9 June 1989: manuscript received in final form 15 September 1989

The method of Zhdanov and Samulevich has been used to analyse the crystal growth rate and nucleation behaviour of silicalite. a silica molecular sieve. Four reactions run under different conditions at 368 K gave values for the crystal linear growth rate (0.5 dl/dt) in the range (1.9—2.0)X1O~pm h ~. However, the reactions do not show the same pattern of crystal mass increase with time, the variations reflecting differences in nucleation behaviour. All the nucleation rate curves were either bimodal or trimodal, suggesting that at least two separate nucleation mechanisms were operating. Early in the reaction, nucleation is probably heterogeneous and associated with the amorphous gel or colloidal material present in the mixture. Later on. when appreciable quantities of crystalline product have been formed, a further crop of crystals nucleates either by a secondary mechanism or by release of further hctero-nuclei from within the dissolving amorphous component. Nucleation was facilitated by early addition of the mineralising organic “template” (tetrapropylammonium cation). and (probably) by rapid stirring. Although all four reaction compositions were identical, the differences seen in nucleation behaviour were linked to observable differences in chemical parameters (e.g. the pH profile of the reactions) as expected for systems governed by the competitive kinetics of many overlapping reactions. The results confirm that simple growth curves based on, for example. XRD crystallinit\ are of limited use in helping to understand the complex processes occurring in zeolite synthesis, and that a more detailed analysis is justified.

I. Introduction

synthesis reaction, which is a reaction-crystallisation at elevated temperatures, often under auto-

Although there have been many studies of the kinetics of zeolite crystallisation [1], most have followed the process of crystal growth by indirect methods, usually by monitoring the development of XRD crystallinity. In relatively few cases have crystal sizes and size distributions been determined by direct measurements. This is largely due to (a) the difficulties in measuring crystal sizes in the range typically found for zeolites (sub-micron to ca. 5 jim), (b) problems (e.g. with particlecounting equipment) in differentiating between the several solid phases which are often simultaneously present in a single reaction mixture (e.g. amorphous material plus one or more crystalline phases), and (c) the intrinsic nature of the zeolite

dave conditions (350—480 K). Crystal linear growth rates for zeolites Na-Y and N-A have been given by Kacirek and Lechert [2—4]. although the values were not derived from direct measurement but calculated from a cuberoot analysis of XRD crystalhinities. Similar methods were applied to zeolite A by Yoshida and Inoue [5] and by Gon Seo [6], although in the former case comparative data are given in (crystaltz3 form only. Some directly measured data linity) for zeolite Na-A have been given by Thompson et at, [7.8] as part of a continuing study [9,10] of zeolite crystallisation processes. and results for Na-mordenite and Na-Y appear in a recent paper by Bodart and co-workers [11]. The most detailed

0022-0248/90/803.50

(North-Holland)

Elsevier Science Publishers By.

I 91)

( .5. ( undc ci a!.

Gr,ocih mic’ and ii ac/canon of vdic a/I ic

analysis of zeohite nucleation and crystal growth has been presented for zeohite Na-X by Zhdanov and Samulevich [12] following earlier work b Zhdanov on Na-A [13]. This type of treatment was also used recently by Sand et at in a study of zeoltte Na-A crvstalhisatton carried out in preparation for an experiment to he performed in space [14]. The family of high—silica zeolttes and silica molecular sieves [15] is so far only poorly represented in crystalhisation studies by direct methods. Length and width growth rates for (Li. NH 4. TPA)-ZSM-5 (TPA = tetrapropvlamrnonium) ai-e civen h\ Nastro and Sand, together with some data for the corresponding (Na. NH4. TPA). (K. N H4. TPA) and (N H4. TPA) systems [16a]. Additional information on (NH4. TP.A)-ZSM-5 has been published more recently by Hon. Sand and Thompson [16h]. and by Hayhurst et al. [17] for sihicalite [18]. However, no detailed study of nucleation and growth has ~et appeared It vsas the objective of the present study to apply the analysis of Zhdanov and Samulevich [12] to the crystallisation of sihicahite [18]. the notionally aluminium-free end member of the ZSM-5 [19] family of zeolitic materials havtng the MFI [20] structure. A more wide-ranging investigation of the effect of reaction variables on the growth parameters of ZSM-5 and other high-silica zeohites will he published separately [21].

2. Experimental 2.1. MateriaLs The following materials were used: tetraethvl silicate (BDH, GPR grade). sodium hydroxide (Fisons. anahar grade). tetrapropylammonium hromtdc (TPABr) (Fluka, purum). Distilled water was always used.

*

Whilst the present paper was in preparation, an investigation of silicalite crystal growth sVas published by Feok tistova et at This work i’, complementary to the present study, and should he consulted in particular for values of linear growth rates at high temperatures [361.

2.2. Pt’epw’aiton

0/ rewTio,i nlLvtIlrc’s

Following some preliminary experiments, the composition chosen for this study (expressed as mole ratios of oxides) was: I .0Na~O.20 Sift. 2.0 TPARr. 1960 H ~O. 8OEtOH. Two hatches ( 1000 g each) of reaction mixture vs crc made up in one lttre pok propyletie bottles. using slightly different preparattve procedures. and each hatch was divided between two different t~pes of reactor. The sodium hydroxide was dissolved in the required amount of water. For hatch (1) the TPABr was then added to this solution and alloyed to dissolve. No TPABr ssas added to hatch (2) at this stage. Tetraethxl silicate vs added to both hatches, and the solutions vsere then stirred rapidly for at least 16h until a single phase was formed. Thereafter the reaction niixtures were kept static at room temperature for seven days. The reqLiired amount o fT PA Br was added to hatch (2) vsith vigorous stirring itiimediately before the crystallisatton reaction was started. 2.3. C’r’i’stalhsatio,i The reactions were carried out in two stainless steel 500 ml autoclaves and two I litre polvpro— pylene bottles. The aLitoclaves were agitated by niagneticallv—coupled paddle stirrers al 300 rpm. The reactions could he sampled by means of dippipes fitted on the autoclaves. Each aLitoclave vs as filled with 400 g of mixture, with hatch (I> in one autoclave and hatch (2) in the other (reactions Al and A2 respecttvelv). The stirrers were started immediately and the temperatures raised to 368 K. The reactions were timed from the moment that heat was applied. The polypropylene reactors were agitated by means of stainless steel paddle stirrers at 15)) rpm. Each reaction could he sampled through a hole in the bottle lid Each bottle was filled with 500 ° of .

.

.

.

-

solution, with hatch (1) in one reactor and hatch (2) in the other (reactions P1 and P2 respectively). The reactors were placed in a 368K water bath,

CS. Cundi eta!.

/

191

Growth rate and nucleation ofsilic’alite

Table I Summary of reaction conditions

Run

Reactor vessel

Batch size (g)

Reaction temperature (K)

Stirring speed (rpm)

Addition of TPABr

Static ageing (days) Room Reaction temperature temperature

Al

Steel autoclave Steel

400

368

300

Initial

7

0

400

368

300

Late

7

0

500

368

150

Initial

7

500

368

150

Late

7

,A2

autoclave Polypropylene Bottle Polypropylene Bottle

P1 P2 Initial

=

before static ageing. Late

=

immediately before the mixtures were brought to the reaction temperature.

allowed to remain static for 24 h and stirred thereafter. The reaction conditions and procedures are summarised in table 1.

weighed and examined by thermal gravimetric

The reaction composition chosen allowed the zeolite product to be crystallised directly from a clear sol [22]. In this way, a direct measurement of

analysis (TGA) with a Stanton Redcroft type min_t. TG770 (flowingThis air 4,5 cm3the heating thermobalance rate 10 K mm_i). gave percentage of water and organic material (normally 13.5%) in the equilibrated solid. The remainder was assumed to have the composition SiO,. For each sample taken from the reactor, the density of the liquid phase was determined and used to

the precipitated crystal mass could be obtained just as for conventional crystallisations from solution. This contrasts with the usual practice of

calculate the liquid sample volume. From all these measurements, the weight of silicalite (expressed as Si0

expressing the extent of zeolite crystallisation as

case (table 2).

percentage volume conversion from amorphous solid as calculated by X-ray powder diffraction peak intensities. The appearance of even a small amount of solid could be detected quite early in the reaction and even if it could not be weighed it could be examined by scanning electron microscopy. It was found that crystals were present very soon after the reaction temperature had been reached, and well before the crystal mass could be measured. Samples (8 cm3) were collected in glass sample bottles. They were allowed to stand for at least 24

Scanning electron micrographs of each sample were obtained with a Cambridge Stereoscan type 604 microscope. Final particle size analysis was carried out in a semi-automatic manner, by viewing the sample with an optical microscope (Nikon

2.4. Analysis

2) per unit volume was calculated in each

Table 2 Weight of calcined silicalite per unii volume of reaction mixture . (h) Time

4

z

(gem

)

X10

Al

A2

h before they were weighed and their pH mea-

70

5

sured. Any solid material sank to the bottom of the sample bottle within 24 h. After the pH had been measured the liquid layer was removed with

92 t16 161

12 28 96

187

a syringe. The solid was filtered through a 0.2 jim

208

filter membrane, washed, dried for 16 h at 368 K. then equilibrated over saturated sodium chloride solution at 293 K for 16 h. The solids were then

232 256

P1

P2 3

-

9 19 56

7 19 75

-

3 21

157

90

118

38

280

153

166

60

308

232

264

109

293

300

303

-

160 319

I ~2

( . .S. (andy

ci a!.

.

(iron iii rOtc’ and] ciiu ic’ancd/i of si/id a/i to

°•25r-—

Optiphot Pol.) ccnnected to a Vickers Magiscan. The final crystal size distribution was obtained using the computer program DIATEST. The size distribution was checked by manual meitsurements of crystal size on scanning electron micrographs. There was reasonable agreement between the two methods hut results obtained with the Magiscan were used in all calculations as many more crystals were measured by this method. When enough sample was available, it was examined by X-ray powder diffraction using a Philips powder diffractometer (Cu Ke radiation). These confirmed that the material obtained wits MFI silicalite in all cases.

O.2o~ ,,~

P

0.05 ,

0.00

6

-

~

9 -

-

-

/=4pm ~

-..

6~ /=ôpm

3. Results and discussion

~

~

3. 1, Measurement of crt’.sial grosith, and tier/eat/on o/ nucleation and mas.c growth curi’e,c

0

The linear growth rate R ( dl/dl) was ohtamed by measuring the length of between 10 and

— — — — -

I

- -.

-

00

200

—

300

i/h

1-ig. 2. Reaction Al: (a) histogram of crsstah site distribLition in final product:

hf growth of largest crystals.

Discrepancies would he expccted at very small crystal sizes, where Gibbs- Thomson (Kelvin) equation soluhility is applicable [23]. Size-dependent growth has also been observed recently (for non-zeolite systems) where crystals have been subjected to mechanical damage [24] or where stressinduced defects are present [25]. The length growth rate (0.5 d//dt ) was found to he (2.0 ±0.1) x 10 1 jim h fcr reactions Al and A2. and (1.9 ±0.1) >< 10 = p.m h for reactions P1 and P2. The measurement of the length/width (I/n-) ratio for crystals of various sizes, from thc same hatch of crystals, indicated that this ratio was constant and equal to 1.3, Thts gives a width growth rate (0.5 dw/dt) of (1.54 ± 0.1) / 10 1 jim h ‘ for Al and A2 and (1 .4~,± 0.1) / 10 p.m h for P1 and P2. The final

0 1 0

W

j

—

100

Tine

=

20 of the largest crystals present in each sample. An idealized drawing of a silicalite crystal is portrayed in fig. 1. Normally the rate of displacement of a crystal face parallel to itself is measured but, under the growing conditions used, the 101 faces of the crystal were not very well developed. However, the length (1) of the crystal is geomctrically related to the displacement of the 101 faces and is much easier to measure, so this was the dimension recorded. The major assumption made in the analysis of the results is that the linear rate of growth is the same for all sizes of crystal within the same readtion mixture. This is a standard assumption and appears to he true for many crystallising systems.

b -

0

particle size distribution for reaction Al is shown as a histogram in fig. 2. This was constructed from

101

I I

Fig. I. Idealised sihicalite crystal

the results of size measurements of over 500 crystals in the final product. The number of crystals ci, in each hand is expressed its the frac-

C. S. Cundy et at.

/

j.~

100

80

equal to the volume ratio V~/V1. This can be

/r

calculated using the equations:

7

40

~=Ky~5[I,(t)] F~=K~a

3.

(2)

1[l,]

20 a 0 2.0

In these, a~ is the fraction of the final total number of crystals with length I,, and 1(1) is the length of these crystals at time t. The shape factor K (representing the difference from cubic morphology) cancels for the ratio ~/V~ and does not require evaluation. For the purposes of this calculation, increments corresponding to a reaction time of 6 h (‘— 0.24 jim) were used. The curve for mass growth calculated in this way is

~ ‘0

(1)

3,

I/

~60

193

Growth rate and nucleation ofsilicalite

‘

shown in fig. 3. The circles in fig. 3 indicate the

-

experimentally determined values of (Z,/Z 1) x 0 0 0

100Time

t/h200

b 300

Fig. 3. Reaction Al: (a) mass growth curve calculated with R

=

0.0404 pm h~ and l~ = 8.35 pm, (0) experimental values, (b) Nucleation curve.

tion cs, of the total number of crystals N. The curve in fig. 2 smooths irregularities in the histogram (see appendix). In further calculations values from the curve are used. The linear growth rate curve for reaction Al is also shown in fig. 2. If it is assumed that the growth rates of all the crystals in this reaction are the same, even at the time of reduced growth after 190 h, then the approximate time for nucleation of any crystal can be calculated. For example, crystals of 4 p.m and 6 jim in the final product would have nucleated at approximately 108 h and 58 h, respectively, as shown by the arrows in fig. 2. It is possible to continue this process for other crystal sizes so that the product size distribution curve can be converted into a nucleation curve (fig. 3, curve b). This was carried out with a computer program using the procedure outlined in the appendix. The ratio of the mass of crystals per unit volume at time t (Z,) to the mass of crystals in the final product (Z1), Z,/Z1. is

100% calculated from the results shown in table 3 1. (theFor average of the two experimental reaction Al.last Z~was taken as 0.030values g cmof Z 1). For the other reactions, the last value of Z, was used for Z1. If all the silica in the reaction ,

The final valuesof for mixture had converted to3.silicalite, the value Z1 would be 0.0334 g cm reactions Al, A2 and P1 indicate that about 90% of the silica was converted to silicalite. Reaction P2 gave a slightly higher value (95%). As the reaction mixtures remained as clear sols throughout the crystallisations. it was relatively easy to separate the solid from the sol by filtration. X-ray powder diffraction showed that the solid was normally completely crystalline, but scanning electron microscopy revealed some amorphous material in the samples taken early in the crystallisation. Fig. 4 shows electron micrographs at various stages of the crystal growth of reaction P1. The sample obtained after 42 h shows small crystals on the surface of amorphous material (fig. 4a). For reaction Al, the experimental values of Z 1/Z1 and the calculated curve (fig. 3) are close and hence it appears that the assumptions made in the calculations are, in the main, valid. However, one factor which was not allowed for in the volume calculation was that of twinning. It was assumed that twins would be equally spread throughout the whole product range, but this may not always be

I 94

1

lOp

5 ( in,!’ ci a!

Ic,

cc/i

~

rotc’ and na /

i,, ci ci ,iii, i/ti,

lOp

I p 4 SI \-l for re,Ic.tion I’h: a) 42 h. ibi’i Ii. (ef lhl h, cO 205 Ii.

ic(’2

Ii. (I) ~4 Ii.

the case and could result in a discrepancy between

lor the other three reactions (A2, l’l and l’2) are

calculated and experimental curves,

shown in figs. 6—Il. The experimental values of

The product of reaction Al appeared to contam fewer twinned crystals than did the other products. This can he seen in fig. 5 which shows electron micrographs of the final products for all four reactions. The product distribution histograms and the crystal growth, nucleation and mass growth curves

Z,/Z

1 and the calculated mass growth curve agree well in all cases. It should he noted that the fit is sensitive to the values chosen for the linear growth rate R ( = dl/dt) and the size of the largest crystals ( 1,,,,~). These were taken from the results in figs. 2. 6, 8 and 10, and are given in the captions to the figures. Slightly different (hut acceptable) values

.5, Cunilr c’t a!.

Growth rate and nuch’atlftn ccf si/li a/in’

4 I

195

_

20p

lOp

~

2Op

Fig. 5. .SEi’sl of final produet~:ha) ,\I. (h A2: bcb Ph:

of ‘,flax give even better agreement between the observed and calculated values of Z,/Z 1. For these calculations no allowance was made for the decrease in R which occurs as the reaction mixture is depleted, hence the calculated curves are not valid for Z,/Z~> 0.95.

bdb

P2.

The second appears to be related to the growing

3.2. Nucleation

population of macroscopic crystals. In terms of the initial creation of nuclei, crystal growth measurements on the largest crystals mdicate that these crystals start to grow at virtually zero time, i.e., as soon as the solutions are heated. This suggests that nuclei are either already present or can be formed very quickly. For very insoluble

The nucleation curves for reactions Al and A2 were trimodal and those for P1 and P2 were bimodal. In both cases the initial maxima always occurred (fig. 2) before any appreciable amount of solid product formed, whereas the others appeared during the growth stage. Thus there are at least two significant nucleation processes. The first is the initial creation of nuclei from the clear sol.

[26] materials which are likely to he nucleated at high supersaturation, a homogeneous nucleation mechanism is possible. Such a mechanism has been suggested for zeolite Na-A, although the nucleation rate was found to increase with stirrer speed [8]. In the present study, initial nucleation rate maxima were higher and later for the two reactions stirred at the lower speed (reactions P1 and P2). However, since the stirring rates apply to

1 96

C. S. Cuncli’ ci a!. / Grout/i rcctc’ and nucleation of slic a/it,’ 0.25

0.25

-

E 2

~

4

6

a 10 //pm

8

a 3

1~ -

V 0

b

Time

200

300

0

_____ 100

t/h

200

Fig. 8. Reaction P1: (a) histogram of crystal size distribution in final product: (b) growth of largest crystals

/

100

/

100

/

/

/0

80

80’

~60

~60-

~40

~40-

9/

7

/0

/0/

20-

____a

0

300

Time ~/h

Fig. 6. Reaction A2: (a) histogram of crystal size distribution in final product: (b) growth of largest crystals.

20

//pm

1~

~

100

9

6

____

0

I

2.0~

,~

------—-

-~

2.0

~1.5

~::L~ ‘0

‘0

J~b

0

100 Time

200

300

i/h

Fig. 7. Reaction A2: (a) mass growth curve calculated with R = 0.0405 pm h and 1m,cx =10. 2 pm. (0) experimental values: (b) nucleation curve.

0

100

Time

200

300

i/h

Fig. 9. Reaction Pt: (a) mass growth curve calculated with R =0.0387 pm hi and ~ = 8.70 pm. (o) experimental values: (h) nucleation curve.

C. S. Cundy et a!.

/

Growth rate and nucleation ofsi/icalite

different reactors, a direct comparison is not

197

100

strictly valid without more detailed analysis of the

hydrodynamics of the two systems. In any case, it seems more likely that the effect arises from the period of static ageing at the reaction temperature

undergone by gel) P1reaction and P2.mixtures, In view of [both the gel)] complex nature probable nucleation zeolite source curves for involves the first the participation maximum perhaps aof ofmore imthe (macroscopic purities and invisible particles (colloidal visible present in of the sol,amorphous i.e.and a heterogeneous mechanism. It has already been noted that the earlier sampIes (taken at times close to the first nucleation maximum) contain a proportion of amorphous material. Optical and electron microscopic cxamination suggests an intimate association between this material and crystalline product formed early in the reaction. Support for this mechanism comes from the work of Aiello et al. [27]. These workers found that dilute, clear aluminosilicate solutions, prepared from amorphous silica powder (about 0.03 to 0.05 mol Si02 per dm3) would

80

~

0

60

-

40iJa 20 0 ______________________________

2.0

‘I’ ‘0 ‘0

1.5 1.0

o

.-~

0.5

b

0.0 ________________________________ 0 100 200 300

Time t/h Fig. 11. Reaction P2. (a) mass growth curve calculated with R = 0.0373 pm h° and ~ =10.4 pm, (0) experimental values; (b) nucleation curve,

0.25

produce amorphous static conditions at appear as crystallites lae. These lamellae

0.20 ~,0.15

d

lamellae when held under 353 K. Zeolites would then in association with the lamelthen developed holes as if

b

0.10 0.05

a

—

0.00 3

9

6

1/pm

being consumed. Nucleation of the zeolites was thus heterogeneous under the conditions used. The clear sols used in the present work are much more concentrated (about 0.5 mol Si0 3). but a 2 per dm similar process appears to operate. Amorphous gel “rafts” form first and then the zeolite nucleates on

,0’

10

or in the “raft”. Interestingly, in a recent electron microscopic study of an even more concentrated

8

~ 2

a 0

~7 100

200

system (ca. 3.0 mol Si0 3) by Zandbergen [28], a further aspect of this heterogeneous process 2 dm

is revealed. Two types of solid were observed early in the reaction: large (2jim) amorphous, irregularly shaped particles, and agglomerates of small b 300

Time t/h

Fig. ~0. Reaction P2: (a) histogram of crystal size distribution in final product; (b) growth of largest crystals.

partly crystalline particles (0.1 jim) containing metastable dense crystalline phases. ZSM-5 was observed to form only in the proximity of the latter material. In the present case, it is suggested

that the heterogeneous process directly observable

I

95

( -

(

andy ci a!. / Giant/i ratc’ and nuc/c’anon of si/icc/it,’

from the small amount of gel precipitated early in the reaction (and which later disperses. releasing its burden of crystals into the continuous phase) may he indicative of a more general heterogeneous process involving the much smaller, invisible, colloidal gel particles which represent the principal reservoir of silica in the system. The second maximum in the nucleation curves of reactions P1 and P2 and the third maximum in those for Al and A2 coincide with the rapid risc of crystal mass and consequent rapid depletion of the colloidal gel component of the reaction mixture. If associated with the increasing mass concentration of crystals in the reactor, this phenomenon is most readily explained through breeding of nuclei by macro- or micro-attrition. i.e. seeondary nucleation [23]. Some support for this is provided by the higher maximum nucleation rate for the more rapidly stirred reactions although thc reservations mentioned earlier prevent a firm conelusion. Alternatively, it is possible that the max-

100 ,

-

11.4

-~ . -

~

~‘~_

-

ii.o

~

-. -

-

1 -

-

Ni -

--

1

2 -

2

-

~40

—

~2O

~-

10 6

~

a-

/

_~

~

-

5

-

~, ~,

1

i.o~

2

/ 1 -

-

-

-

-

2 -

b

0.0

-

I

- --

0

100

200

Time

irna are linked through the role played by the amorphous material present in the reaction mixture. Following the suggestion made earlier that the initial maximum may he due to gel-mediated heterogeneous nucleation, then the later maxima may be caused by release of nuclei, or ordered nucleating structures, from (colloidal) gel particles as they dissolve, as suggested by Zhdanov [13] who observed a very similar correspondence between mass growth and bimodal nucleation curves in his study of zeolite A crystallisation. Bimodal crystal size distributions for zeolite ZSM-5 have also been reported by Mostowicz and Sand [29] and by Ball and coworkers for a fernsilicate version [30]. Mostowicz and Sand used static autoclaves at 443 K and obtained a distrihution which appears to be composed of two overlapping peaks. In the absence of any forced agitation. secondary nucleation seems an unlikely cxplanation for the later nucleation peak in this case, although effects due to convection currents within the reaction vessel cannot be ruled out. In some cases, bimodal distributions could arise through the presence of two crystal populations having different compositions. for example in the case of concurrent or consecutive crystallisation of silicalite (aluminium-deficient) and ZSM-5

•

-

2.0

~eo -

--__

Fig. 12. )a)pH values: ((Al:

I 300

i/h

(•)A2: (0) Fl: (•) P2. (iro~th

curves, full lines (I) Al. (2) A2. broken lines (1) P1. (2) P2. (hi Nucleation curves: bull lines (I) Al: (2) A2: broken lines (I) 01:

(2) 02

(aluminium-rich) phases as described by Padovan et al. [31]. 3.3. Comparison o/

reaciion.s

The linear growth rate observed for reaction Al and A2. (2.0±0,1) )< 10 = p.m Ii ‘. appeared slightly higher than that for P1 and P2. (1.9 ±Oh x 10 = p.m h although the difference is within the estimated error of the measurement. Hovs ever. despite the fact that the mean linear growth rates for all the reactions are closely similar, the mass growth curves are not the same hut reflect di)’ferences in the nucleation behaviour (fig. 12). The fastest reaction (i.e. the first to reach completion) is Al. while the slowest is P2. Although it is not possible to give a conprehensive explanation, some elements of rationalisation can he offered. If the tetrapropylammonium template is necessary for both rapid nucleation and rapid growth. then the reactions in which the TPABr was added

CS. Cundy eta!.

/

199

Growth rate and nucleation ofsilica/lie

as early as possible would be expected to be faster than those where the template was added just

I

In keeping with this, Al is faster than A2, and P1

is faster than P2. Similarly, if stirring assists before raising the mixture to reaction temperature.

1.5

2.O~

nucleation and reduces mass-transport limitations. the more rapidly-stirred reaction should in each part, be and this than is also be the case (Al > p1, case faster itsseen moretoslowly-stirred counterA2>P2).

100

-

2

1.0

-~ —

0.5

~.

80 60 40

- 20 ‘

a

___________________

N

~ 0

and P1, P2 lies in the contribution the initial

nucleation process makes to the growth curve. As stated earlier, this difference is probably related to the 24 h period of static ageing at reaction temperA major difference between reactions Al, A2

ature undergone by the P-series preparations but not by the A-series. However, other differences (type of reactor, stirring rate) preclude further speculation at present. As shown in figs. 13 and 14, this contribution is only 30—50% for Al, A2

-80 1

-~

_____,,,//J

1.5 ‘~

0~C -

60 N

- 40

2 0.5 -

-

1

0.0 0

- 20 -

.-“

100

-

b

200

300

Time t/h

2.0

1.0 1.5

—

~ o.sH4:;’ ‘0

J 1

~00

curve. Fig. 14. (a)Contribution Reaction P1: of (1)nucleation first nucleation processes event: to (2) mass total growth mass total mass growth curve.

1

80 60 40

,,2

a 20

growth curve. (b) Reaction P2: in (1) the first nucleationand event: (2) but over 95% P2. The differences between Al and A2 liefor asP1, much second third nucleation humps as in the first, whereas the differences between in the initial P1 andnucleation P2 are related process. to slight The

-,

I

‘0 0 -

N

C

initial nucleation rate for P1 is greater than that for P2: more nuclei are formed in P1, the reaction

80

is finished earlier and the crystals are smaller.

may Variations also ariseinfrom behaviour differences over the manifest four reactions in their

2.0~ 1.5—

60

1.0

-

40

0.5

-

20 1

0.0

~ 0

b

100

200

0

300

Time t/h Fig. 13. Contribution of nucleation processes to mass growth curve. (a) Reaction Al: (1) first nucleation event; (2) second nucleation event; (3) total mass growth curve. (b) Reaction A2: (1) first nucleation event: (2) second nucleation event; (3) total mass growth curve,

chemistry. For example, as shown in fig. 12, the pH values of reactions Al, A2 (11.30) are significantly higher than those of P1, P2 (11.18) during much of the initial nucleation and most of the growth stages of the reactions. It is difficult to identify the cause of this difference, although it is

often found that the pH of a reaction mixture and the course of the reaction are very sensitive to the preparation procedure [32]. In all these reaction mixtures, an increase in pH corresponds to the removal of silicate species from the solution phase

20))

C ‘.5, Cucicly ci

,i/,

/

Croci tli rate and ciuc-ii’aiion o~si/ic cc/ill’

and the release of OH ions [33.34]. TI-ic initial rise in pH (fig. 12) corresponds to at-i increase in the amount (or a decrease in the soluhility-) of the ai-i-iorphous solid phase of the gel. During most of this stage the pH values of ti-ic mixtures to which TPABr was added before static ageing are the lower of two. i.e. Al < A2. P1 < P2, Towards the end of the crystailisations the pH values rise corresponding to depletion of the silicate species in the solution phase. Perhaps the most notable feature of the pH( t) curves is ti-ic essenttal constancy of the pH over most the reaction time: during this period, the crystals are growing at a steady rate (figs. 2, 6. 8 and 10).

4.

Conclusions

The growth curves frequently obtained for zeolite formation (for example. by the determinatiort of XRD crystallinities) [1.35] are of limited use as an aid to tIte understanding of zeolite crystallisation. The shapes of these growth curves arc a result of both nucleation and crystal growth processes. As a result, it is not possible to take a growth curve in isolation and use it to deduce information about either the nucleation or the growth rate. However, by making simplifying as— sumptions which place a limit on the contribution Of nucleation processes, activation parameters can he obtained which are useful in making semiquantitative comparisons between similar readtions within a series [35]. The present study confirms [12] that a comhination of linear growth rate measurement and particle size analysis can give detailed information about both nucleation and crystal growth hehaviour. This approach gives an insight into the mechanism of nucleation, and demonstrates why apparently nlinor changes in reaction conditions can alter the reaction timescale and give rise to different crystal size distributions. li-i all of the reactions studied in the present work, nucleation is not confined to an initial induction period, hut carries on through the crystallisation region and decreases only when over 40% of the available nutrient material has been consumed. Nucleation appears effectively to stop at about the same time as the linear growth rate begins to decrease or tail

off. This is to he expected since both are a tune— tion of’ the decreasing supersaturation. Mathematicat models which attempt to fit experimental growth curves require a basis which takes these variations into accoLint if they are to he of scien— tific rather ti-ian empiric;tI value [101. Finally, conclusions about the nucleation n’tech;tnisms operating in this system may he summarised. On the basis of observations made here and by others [27.28]. it is .suggcsted that the initial period of nucleation derives fron’i a heterogeneous mechanism, whereby crystals nucleate on gel particles (macroscopic or colloidal). Such partides may- he entirely amorphous. or of a lou degree of order, or may contain elements of crvs— talline precursor phases. li-i tern-is of practical oh— servation, it may he that an’ 5- difference hetu een nucleation on colloidal particles and true homoge— nieous nucleation is academic. Crystals nucleating later in the reaction when large nunihers of macroscopic crystals are airead~- preset-it may do so either by a process of secondary nucleation ftc caused by the crystals themselves), or ma\ originate in ti-ic release of nuclei from dissolving gel (which, again. may he in ti-ic form of colloid;tl particles). The “release” theory is closely related to the heterogeneous mechanism put forward ho explain the early stages of nucleation, and such process may well contmue into the later st;tges of the reaction to contribute further to the crystal population. However, if this were the only mechanism in operation. a continuum of nucleation activitv might he expected, accelerating with the mass growth rate and the corresponding consunnption of gel. The occurrence of at least one minimum in the nucleation rate at at time when ti-ic invariance of ti-ic linear growti-i r;tte implies that the average supersaturation is constant suggests that nore than one nucleation process is operating. Since the final period of nucleation coincides with the presence of a rapidly growing rn;tss of crystals, secondary nucleation is strongly implicated. Acknowledgements

Financial support from the SERC’ and ICI PLC is gratefntilv acknowledged.

CS. Cundy eta!.

/

Growth rate and nucleation ofsilicalite

201

Table 3 Parameters for eq. (A.2) Run

h

1

l,,,~

r1

h2

‘,,,.

r

h~

‘ccci

r3

a

0.01829 0.02000 0.00000

6.190 8.300 0.000

1.0000 0.9344 0.0000

0.0021 0.0021 0.0034

0.00000

0.000

0.0000

0.0027

Al A2

0.0986 0.1950

1.842 1.670

1.0000 0.5954

0.05209 0.02300

3.543 4.040

1.000 4.535

P1 P2

0.0657 0.0979

0.976 2.195

0.7320 1.0593

0.04616 0.04000

4.344 6.333

2.749 1.779

151 A. Yoshida and K. lnoue, Zeolites 8 (1988) 94. 161 Gon Seo, Hwahak Konghak 23 (1985) 295: Chemical

Appendix

Abstracts 104 (1986) 79373w.

The histograms fraction of crystals where j.ls is 0.3 for as a function of s. the equation: a,

in figs. 2, 6, 8 and 10 give the a, in the size range S ~“3 S + L\S Al, P1 and P2, and 0.4 for A2, The values of a5 were fitted to (A.l)

F(/),

=

where 1= s + ~ It was found that F(1) could be represented by a combination of normal distributions:

171 M. Tassopoulos and R.W. Thompson, in: Proc. 7th Intern, Conf. on Zeolites, Eds, Y. Murakami, A. lijima and J.W. Ward (Kodansha/F.lsevier, Tokyo/Amsterdam, 1986) p. 153.

[81 R.W. Thompson and A. Dyer, Zeolites 5 (1985) 302. 191 R.W. Thompson and A. Dyer, Zeolites 5 (1985) 292. [101 R.W. Thompson and A. Dyer, Zeolites 5 (1985) 202:

J. Warzywoda and R.W. Thompson, Zeolites 9(1989) 341. [ill P. Bodart. J.B. Nagy. Z. Gahehica and E.G. Derouane. J.

Chim. Physique 83 (1986) 777.

[121 S.P. Zhdanov and N.N. Samulevich, in: Proc. 5th Intern.

Conf. on Zeolites, Ed. L.V.C. Rees (Heyden, London, 1980) p. 75.

3 F(

I)

=

~ h, exp( — [(I

—

~

)/r]

2)

(A.2)

where the constants h,, ‘rn’ ~ have the values given in table 3. The variance of the fit a was in all cases consistent with the precision of the mea-

1131 S.P. Zhdanov, in: Molecular Sieve Zeohites

— I, Am. Chem. Soc. Advan. Chem, Ser., No. 101 (Am. Chem. Soc., 1971) p. 20.

[14] LB. Sand, A. Sacco, Jr., R.W. Thompson and AG. Dixon, Zeolites 7 (1987) 387. [15]

PA. Jacobs and J.A. Martens, Synthesis of High-Silica Aluminosilicate Zeolites, Studies in Surface Science and

surements. Hence:

Catalysis, Vol. 33 (Elsevier, Amsterdam, 1987).

da

1 —

(A.3)

~—(F(l)),

and the nucleation curve is given by da

R

=

—~-~ [ F(In1ax —

in which

R

RI)].

1161

A. Nastro and LB. Sand, Zeolites 3 (1983) 57. (b) L.-Y. Hou, LB. Sand and R.W. Thompson, in: Proc. 7th Intern. Conf. on Zeolites, Eds, Y. Murakami, A. (a)

lijima and J.W. Ward (Kodansha/Elsevier. Tokyo/Amsterdam. 1986) p. 239.

(A.4)

is the growth rate (dl/di).

References [1] See, for example. R.M. Barrer, Hydrothermal Chemistry of Zeolites (Academic Press, London, 1982) ch, 4. p. 133. [2] H. Kacirek and H. Lechert, J. Phys. Chem. 79 (1975) 1589. [3] I-I. Kacirek and H. Lechert, J. Phys. Chem. 80 (1976) 1291. [41 H. Kacirek and H. Leehert, in: Molecular Sieves II, Am. Chem. Soc. Symp. Ser.. No, 40, Ed. JR. Katzer (Am. Chem. Soc., 1977) p. 244.

[171 Giordano, D.T. Hayhurst, A.8 Nastro, R. Aiello, F. Crea and G. Zeolites (1988) 416: D.T. Hayhurst, R. Aiello, J.B. Nagy, F. Crea. G. Giordano, A. Nastro and J.C. Lee, in: Perspectives in Molecular Sieve Science, Am. Chem. Soc. Symp. Ser., No. 368, Eds. W.H. Flank and ThE. Whyte, Jr. (Am. Chem. Soc.. 1988) p. 277. 1181 EM. Flanigen, J.M. Bennett, R.W. Grose, J.P. Cohen. R.L. Patton, R.M. Kirchner and J.V. Smith, Nature 271 (1978) 512: R.W. Grose and EM. Flanigen, US Patent 4.061,724 (1977). 1191 G.T. Kokotailo, S.L. Lawton, D.H. Olson and W.M. Meier, Nature 272 (1978) 437; R.J. Argauer and G.R. Landolt, US Patent 3,702,886 (1972). [20] W.M. Meier and D.H. Olson, Atlas of Zeolite Structure Types. 2nd ed, (Butterworths, London, 1987).

2) (2

C

S. ( ‘andy ci a/,

/ (ir,,cyih raid’ cOld! ,iuc/c’aiicni 0/si/li si/Oc

[21] (‘,S, (‘andy, hiM. Lowe and D.M, Sinclair, paper in preparation. [22] S. Ueda, I. Sera. Y. Tsuiuki. M. Koizunmi and S. Takahashi, J. (‘la~Science (Japan) 23 (1983) 6)).

Flectron Microsccipv ( Furem S5(. York, I9SS. (‘cml, Ser. 93, [dv. P,J. (ioodhew and HG. (Inst. Phym.. London—Bristol, 1985) Vol. 2. di. [291 R. Mosiowic, and LII. Sand, Zeolites 3(1983)

I 23] See,

[3))] W..l - RaIl. .1. D~ver A. Garforth and WI. Siiii tO. In: Proc. 7th Intern. (‘onf. on Zecilites, Idly. V. Murak,inii, -‘5. hijinia and .1W. Ward ( Kodansha/Elses er. T,,k’oc Amsterdam, 1986) P. 137, [311 M. Padccvan. C. Leofanti. N-I. S,,Iari and I:. \loret II. Ze,,Iitcs 4 ( 1984 295, [32] ‘5, Ara’~’:i. 1.1. Barber, Ii.M. l,,vse I),N-I, SinclaIr and -‘5, ‘5’arma. Zeohites 4 (1984) 263. [33] .1.1.. (‘asci and hIM. Lcme. Zec,lites 3 (1983) ISO

for example ,J.W. Mullin, (‘rvstall isation. 2nd ed. Iiutlerworihs. London, 1972). [24] .1. (iarside. C. Webster. RI. Davey and ,A.J. Ruddick. in: Industrial (‘rystallisation 84, F.ds. SI. Jan2d and [.1. dc long ( Elsevier. Amsterdam. 1984) p. 459, [25] 11.1.. Rhat .J.N - Shern-ood and T. Shripathi. (heni. l.ng. Sc’j. 42 (1987) 609. 26] B.M. Lowe, in: Innovation in Zeohite Materials Science. Studies in Surface Science and Catalysis. Vo). 37. Fds. P__I, (irohet, WI. Mortier. F.F. Vansant and C. SehuI7-[kloff (Elsevier. Amsterdam, 1988) p.1, and references therein. [27] R. Aiello, R.M. Barrer and IS. Kerr, in: Molecular Sieve Zeohites — I. Am. (‘hens. Soc. Advan, Chem. Ser.. No. lOt

(Am. (‘hem. Soc., 1971) p. 44. [28] H.W. Zandhergen. in: Proc. 9th European C ongr. on

[34] hIM. Lowe. Zeohites 3 [35]

‘5.

(‘ulIaz and 1.0.

1983)

Inst. Ph~s, Dickensc,n 9. p. 363. 219,

301k

Sand. in: Molecular Sieves. ‘\m,

(‘hem, Soc ..Advan, (‘hem. Ser.. Nc’. 121. [dv. W.M. Meier and lB. Uvtterhc’even (Am. (hem. Soc., 1973) P. 14)).

[36] N.N. Feoktistcw:i. 5.0. Zhdanov. W. Luti inch M. Bcih,ivs Zeolites 9 (1989)

36,