Coking during ethene conversion on ultrastable zeolite Y

Coking during ethene conversion on ultrastable zeolite Y

Applied Catalysis A: General 167 (1998) 353±362 Coking during ethene conversion on ultrastable zeolite Y Boontham Paweewan, Patrick J. Barrie*, Lynn ...

239KB Sizes 3 Downloads 22 Views

Applied Catalysis A: General 167 (1998) 353±362

Coking during ethene conversion on ultrastable zeolite Y Boontham Paweewan, Patrick J. Barrie*, Lynn F. Gladden Department of Chemical Engineering, University of Cambridge, Pembroke Street, Cambridge CB2 3RA, UK Received 8 September 1997; received in revised form 28 October 1997; accepted 28 October 1997

Abstract A multi-technique approach has been adopted to investigate both the formation of coke and its effects during the conversion of ethene on ultrastable zeolite Y at 773 K. It is shown that pulsed ®eld gradient (PFG) NMR diffusion measurements in combination with other characterization techniques (product analysis, thermogravimetric analysis, nitrogen adsorption, FTinfrared and 13 C NMR spectroscopy) gives particular insight into the deactivation process. Site poisoning is responsible for the initial loss of activity. There then follows a period in which the amount of coke increases while it is transformed into more aromatic species. At high levels of coking, pore blockage becomes the most signi®cant factor. # 1998 Elsevier Science B.V. Keywords: Coke; Deactivation; Zeolite; NMR, pulsed ®eld gradient; Diffusion

1. Introduction Zeolites are widely used as shape-selective acid catalysts in the petrochemical industry, and their deactivation due to coking has received considerable research attention over the years due to their commercial importance [1]. The word `coke' is used in this context to denote the carbonaceous residues formed from secondary reactions within the catalyst. A wide variety of analytical techniques have been used to investigate coke formation within zeolites and these have been reviewed by several groups of workers [2± 6]. In-situ measurements on coke by techniques such as infrared and NMR spectroscopy are particularly important as they provide information on the chemical nature of the coke formed without the need to extract the coke or dissolve the zeolite framework. The *Corresponding author. Fax: 441223 334 796, E-mail: [email protected] 0926-860X/98/$19.00 # 1998 Elsevier Science B.V. All rights reserved. PII S0926-860X(97)00330-X

chemical nature of the coke formed and its effect on catalytic performance depends on a number of factors: the nature of the reaction, the temperature, the zeolite catalyst used and the time on stream. Cracking reactions have been most widely studied, as it is this process in which deactivation of the zeolite catalyst has greatest commercial signi®cance. Useful data have been gained on coke formed from the cracking of n-hexane [7±11] and n-heptane [3,12], as well as from the cracking of larger molecules such as hexadecane which are similar in size to the feed of a commercial cracking process [13]. However, it is dif®cult to understand coke formation for such a reaction in detail: many products are formed, and it is the reactions of some of these products rather than the reactant itself which are most responsible for coke formation. A further complication is that there is evidence that branched-chain paraf®ns adopt different cracking mechanisms to their straight-chain isomers [14]. It is generally believed that it is the ole®ns


B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362

formed from the initial cracking step which are most responsible for coke formation within zeolite catalysts, and so workers have also investigated coke formed from the reaction of ole®ns within zeolites, rather than the cracking reaction directly. Examples include ethene conversion [15,16] and propene conversion [17,18]. In this work we have chosen to study ethene conversion, as the simplest reaction that produces a large amount of coke. Temperature has a crucial bearing on the chemical nature of the coke produced. Some workers have distinguished between `low-temperature coke' and `high-temperature coke' [4,15,16]. The former is formed at temperatures of 300±500 K, and consists mainly of branched saturated hydrocarbons with a similar H/C elemental ratio to the reactants, while the latter is formed at temperatures above 550 K and consists of aromatic and polyaromatic species with a lower H/C ratio than the reactants. A simple scheme for the formation of high-temperature coke involves crackingtoole®ns,followedbyoligomerisation,cyclisation and hydrogen transfer reactions, thus: paraf®ns!ole®ns!napthenes!aromatics!polyaromatics [2]. As a result of these reactions, it is sometimes dif®cult to compare coke formation results from different laboratories when different temperatures are used, even when so-called high-temperature coke is the only product. In this work, we have chosen to perform the reactions at 773 K, which is a typical temperature for commercial operation. Not surprisingly, the three zeolites most commonly used in industry have been the most widely studied, namely zeolite Y (normally in the `ultrastable' form, USY, but sometimes rare-earth exchanged), H-ZSM-5 and H-mordenite. The effect of pore space on coke formation has been well illustrated [3,19]. Mordenite consists of unidimensional channels, and so pore blockage has a major effect even at low levels of coke. Zeolites Y and H-ZSM-5 have three-dimensional pore networks, and so pore blocking is less critical. They do, however, differ in the number and strength of their acid sites, and have quite different pore characteristics. Zeolite Y has large cavities within which polyaromatic coke species can build up, while H-ZSM-5 consists solely of channels, the steric constraints of which inhibit the formation of polyaromatic species inside the pore network. In this work, we have chosen to look at reactions within

ultrastable zeolite Y in the hydrogen-exchanged form, HUSY. Coke formation is expected to affect activity and selectivity in two different ways. Firstly, site poisoning (consumption of active acid sites) will affect the reactions occurring, and secondly pore blocking will affect the diffusion of both reactants and products to the active sites. Consideration of both of these effects is necessary in order to understand the system. As there is still no consensus of opinion on the detailed mechanism of coke formation and, equally importantly, the effects of coke within zeolites, in this paper we apply a multi-technique approach to study coke formed within zeolite HUSY during ethene conversion at 773 K. The techniques used include gas chromatography product analysis, thermogravimetric analysis, nitrogen adsorption, FT-infrared spectroscopy, 13 C NMR spectroscopy and pulsed ®eld gradient (PFG) NMR measurements of diffusion. 2. Experimental Coked samples of HUSY were prepared after various times on stream using a stainless steel ®xed-bed reactor. The HUSY starting material had previously been fully characterized and found to have a framework Si/Al ratio of 4.15 based on unit cell parameter and a bulk Si/Al ratio of 2.65 from X-ray ¯uorescence measurements [20]. 500 mg of fresh catalyst was heated in a ¯ow of nitrogen gas at a rate of 200 K per hour until the reaction temperature (773 K) was reached. The sample was maintained at this temperature for 30 min before starting the experiment. Coke was generated by passing a mixture of pure ethene (99.9%; 20 ml/min) and nitrogen (99%; 20 ml/min) through the catalyst. The products from the reaction were analysed using a Phillips Pye 4500 gas chromatograph equipped with a ¯ame ionisation detector and a porapak Q column. After a speci®ed time on stream, the ethene ¯ow was stopped, and the reactor cooled down to room temperature under the nitrogen atmosphere. Coked samples were obtained in this way after 15, 30, 45, 60, 120 and 240 min time on stream. The integrity of the samples was checked by powder X-ray diffraction and the samples were then examined by a variety of techniques. For most of the techniques used, a heat treatment to remove water was required. This

B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362

could cause removal of any volatile coke present, but it is unlikely that light coke is present in signi®cant amounts given the high temperature at which the coke formation reaction was carried out. The BET surface area was obtained by analysing nitrogen adsorption at 77 K in a conventional volumetric apparatus (Micromeritics ASAP 2000). Each coked sample was pre-treated by heating to 673 K at 10ÿ5 Torr for 12 h in order to remove all water prior to starting the BET adsorption isotherm experiment. The weight percent of coke present in the samples was obtained by thermogravimetric analysis. Each sample was heated at a rate of 10 K per minute in an argon gas ¯ow until a temperature of 973 K was reached. At that point, air was applied to ensure complete coke combustion at this temperature. Infrared spectra were obtained on a Nicolet Magna 750 FT-IR spectrometer equipped with a SpectraTech 0030-102 diffuse re¯ectance assembly. Each coked sample was heated in a helium gas ¯ow to 673 K to eliminate all water present in the pore space before recording the IR spectrum at this elevated temperature. A background spectrum was obtained by replacing the sample with a polished mirror in an identical helium gas ¯ow. Infrared spectra of the zeolite with adsorbed pyridine were also obtained. For these measurements, the samples were activated at 673 K in helium, cooled to 573 K and then exposed to a pyridine/helium gas ¯ow, obtained using a simple room temperature saturator, at 573 K for 10 min. The sample was maintained at 573 K for a further 5 min in a pure helium ¯ow, before cooling down to room temperature and recording the spectra. Background spectra were obtained on the activated sample at 573 K in a helium gas ¯ow. For these measurements it was necessary to dilute the sample with 50 wt.% germanium powder in order to reduce the specular re¯ectance component of the SiO4/AlO4 tetrahedral stretching vibrations which would interfere with the pyridine ring vibrations in the 1700±1400 cmÿ1 region. The amount of pyridine adsorbed on the germanium powder is negligible [21]. 13 C solid-state NMR spectra were obtained at 50.32 MHz on the coked samples using cross-polarisation (CP), magic-angle spinning (MAS) and highpower proton decoupling on a Bruker MSL-200 spectrometer. The sample spinning rate was 4 kHz, and spinning sidebands were suppressed using the TOSS


method [22]. The contact time used was 1 ms, and the recycle delay between scans was 1 s. For the samples containing only low coke loadings, the spectra were acquired overnight. Chemical shifts are quoted relative to tetramethylsilane. Diffusion measurements on methane and n-butane adsorbed within the fresh zeolite and the coked samples were made using pulsed ®eld gradient (PFG) NMR spectroscopy [23,24]. Each sample was heated at 2 K/min up to 673 K and left at that temperature overnight under a vacuum of 10ÿ6 Torr in order to remove all adsorbed water. After cooling down to room temperature, methane or butane were adsorbed into each sample (at a loading of approximately 10 molecules per unit cell), and the glass tube sealed. The 1 H NMR spectra were acquired at 200.13 MHz. After measuring the T1 and T2 nuclear relaxation times to aid choice of the acquisition parameters, diffusivities were measured using the PFGLED pulse sequence. This uses a stimulated echo and minimizes any errors due to eddy current effects [25]. The signal intensity is expected to follow the form ln…I=I0 † ˆ ÿ 2 G2 2 D… ÿ =3† where I is the observed intensity, I0 is the intensity in the absence of gradient pulses, is the gyromagnetic ratio of 1 H, G is the applied ®eld gradient amplitude,  is the length of the gradient pulse,  is the interval between the gradient pulses, and D is the diffusivity [25,26]. For methane,  was varied while maintaining G at 0.22 T/m and  at 4.5 ms. For n-butane,  was again varied, with G ®xed at a value of up to 2.2 T/m and  chosen to be 10± 57 ms depending on the sample. Due to the presence of both inter- and intra-crystalline diffusion, it was necessary to analyse some of the experimental data assuming the presence of two distinct diffusion components using a least-squares ®tting routine. 3. Results and discussion 3.1. The reaction Gas chromatography analyses during the course of the reaction showed a certain amount of scatter but clear trends. A typical plot of ethene conversion against time on stream is shown in Fig. 1. It can be seen that there is a rapid drop in activity during the ®rst 20 min or so of the reaction, and the conversion


B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362

Fig. 1. Plot of ethene conversion (%) as a function of time on stream.

becomes close to zero after 60 min. In general, activities as a function of time-on-stream, t, are expected to follow the form (1‡Gt)ÿN where G and N are constants [27,28], and the line shown in Fig. 1 is a function of this form. The product selectivities are shown in Fig. 2. At the beginning of the reaction, when activity is still fairly high, the primary product is butene. However, the amount of butene in the products drops rapidly with time on stream, and propene becomes the major product after about 20 min of reaction. After a reaction time greater than 1 h, when activity is now almost zero, methane becomes the major product. Formation of these products is shown in Scheme 1 below. The product butene can form from the simple oligomerisation of ethene. However, adsorbed butene itself can be cracked to give propene and methane. The

Fig. 2. Plot showing product selectivity (in vol.%) as a function of time on stream.

reduction of butene in the product may thus be related to differences in the acid sites after partial coking, or to steric constraints once appreciable pore blockage has occurred: the larger molecule will diffuse slower, and thus have more time to react further. Higher hydrocarbons than butene were not observed as products, again due to the steric effects of the zeolite framework and the coke formed within it. Cracking reactions in zeolites may be modelled as a chain propagation reaction with both monomolecular and bimolecular reactions playing a part [29]. In this scheme, the coke is believed to form from adsorbed carbenium ions which undergo side-reactions that give products that are not capable of desorbing or reacting further in the chain propagation [29]. Bimolecular reactions are expected to be more dominant in large-pore zeolites, which suggests that more coke will be formed early in

Scheme 1.

B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362

Fig. 3. Plot of wt.% coke present in the catalyst as a function of time on stream.

the reaction at higher conversions; this is observed to be the case here. 3.2. The chemical nature of the coke formed The total amount of coke formed was determined by thermogravimetric analysis and is shown in Fig. 3. Comparison with Fig. 1 shows that it only takes a small amount of coke to have a large effect on catalytic activity. It can also be seen that coke continues to form even after 60 min reaction time when the catalyst activity is very low. In order to probe the chemical identity of the coke formed, both 13 C NMR spectroscopy and infrared spectroscopy were used. 13 C solid-state NMR spectroscopy has been used previously to characterize the coke formed [16,30± 37], as well as to investigate the nature of the products formed [38,39]. Some of these works have used 13 Cenriched feeds in order to improve signal-to-noise (e.g. [16]), while others have concentrated the coke by demineralisation using hydro¯uoric acid (e.g. [36]). In this work, we achieved acceptable quality 13 C spectra using the cross-polarisation (CP) technique even at the lowest coke loading studied at natural abundance without any chemical treatment. While a small amount of carbon may not be fully detected using the CP method due to the in¯uence of paramagnetic free radicals or conducting electrons in any graphite present [32,40], the trend between the samples investigated is very clear. Spectra are shown in Fig. 4 and resemble those of high-rank coals [41]. Two


major spectral regions can be identi®ed: the asymmetric peak at 130 ppm is associated with aromatic carbon environments (though polyole®ns also resonate in this region), while the peaks at 10±50 ppm are associated with saturated aliphatic carbon species. One frequently used parameter in coal chemistry is the aromaticity, de®ned as the fraction of carbon present in the sample that is in aromatic environments. From the spectra we obtain an aromaticity of 80% for the sample after 15 min on stream, 94% after 1 h on stream, and 98% after 4 h on stream. Thus it is clear that the coke is highly aromatic after 15 min reaction time, and becomes even more aromatic in character with further time on stream. However, even at a temperature of 773 K, saturated aliphatic environments are signi®cant in coke (except at very long reaction times); larger proportions of aliphatic components have been observed at lower temperatures and shorter reaction times [33]. It is sometimes not fully appreciated that coke is not a single compound and that its structure itself changes with time on stream. Infrared spectroscopy has been widely used to study the chemical nature of coke within zeolites [4,6]. The so-called coke band [42] indicating polyalkenes or polyaromatic species at around 1600 cmÿ1 is a good indicator of the presence of coke. FTIR spectra recorded using diffuse re¯ectance in the region 1400±1800 cmÿ1 are shown in Fig. 5, and show a large peak at 1590 cmÿ1 after 15 min reaction time indicating polyaromatic species in agreement with the 13 C NMR results. The intensity of this peak grows with reaction time, but is little changed after 1 h reaction time. There are no other peaks in this spectral region. Bands at 2960 and 2930 cmÿ1 due to the CH3 and CH2 asymmetric stretching modes of coke can also be detected for the samples after 15, 30 and 45 min reaction time. There is little change in the relative intensity of these two C±H bands between samples, suggesting that the degree of branching of the coke does not change signi®cantly during this time [4]. Only very weak broad bands were observed at wavenumbers above 3000 cmÿ1, which argues against the presence of signi®cant amounts of polyalkenes. After 60 min time on stream, the C±H bands become too weak to detect, as the coke becomes more polyaromatic in character in line with the 13 C NMR results.


B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362

Fig. 4.


C MAS NMR spectra of coked samples after a time on stream of: (a) 15 min; (b) 60 min; (c) 240 min.

Fig. 5. FTIR spectra of HUSY after a time on stream of: (a) 0 min (fresh catalyst); (b) 15 min; (c) 60 min.

3.3. The effect of coke formation Coke is expected to affect catalyst activity and selectivity through site poisoning (consumption of

active acid sites) and pore blocking, which will affect the available surface area and the molecular mobility of reactant and products. The distribution of coke within the catalyst is an important consideration.

B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362


Fig. 6. FTIR measurements of OH stretching region of HUSY after a time on stream of: (a) 0 min (fresh catalyst); (b) 60 min; (c) 120 min.

Direct measurements of this have been dif®cult to obtain, but there has been some success using the 129 Xe NMR technique [37,43±45]. Considering the small size of the reactant used, it is likely that coke will form throughout the catalyst investigated here, but this point is discussed further below. The acid sites present in the coked samples were characterized by FTIR spectroscopy, including pyridine adsorption measurements. In the fresh HUSY catalyst, there are three OH stretching bands at 3736, 3660 and 3585 cmÿ1 as shown in Fig. 6. These are attributed to terminal silanol groups, to Brùnsted acid sites in the supercages, and to Brùnsted acid sites in the smaller sodalite cages, respectively [46,47]. There is a gradual reduction in the number of all the acid groups present with increasing time on stream, indicating that site poisoning is occurring through the irreversible binding of coke to Brùnsted acid sites. The reduction in the band due to hydroxyls in the sodalite cages can be explained in terms of proton mobility [46]. While there is some site poisoning, there are still large numbers of acid sites present after one hour on stream, even though the catalyst activity is almost zero at that point. This indicates that pore blockage is having the dominant effect on deactivation at this stage of the reaction. After 2 h time on stream, there are almost no remaining hydroxyl groups, which indicates that a small molecule such as ethene can still penetrate the highly coked catalyst and react further to remove remaining Brùnsted acid sites. Catalytic activity in HUSY zeolite is not solely due to Brùnsted acid sites, but can also be affected by

Lewis acid sites due to defect aluminium sites in the framework and any extraframework aluminium species present. These may be quanti®ed at the same time as the Brùnsted sites using pyridine adsorption in conjunction with infrared spectroscopy [48]. FTIR spectra on pyridine within the fresh and a coked catalyst with the adsorption performed at 573 K are shown in Fig. 7. This shows bands due to pyridine adsorbed on both Brùnsted (1631 and 1545 cmÿ1) and Lewis acid sites (1621 and 1452 cmÿ1) and a band close to free pyridine (1490 cmÿ1) [48]. The bands at 1545 and 1452 cmÿ1 are commonly used for quantifying the number of Brùnsted and Lewis sites present as they are individual non-overlapping peaks. For the coked sample after 15 min reaction time, there is a reduction in signal intensity, with the Bùrnsted acid sites being slightly more affected than the Lewis sites. Thus both Brùnsted and Lewis acid sites appear to play a role in the coking reaction. For the samples after 60 min on stream, no pyridine signal can be observed which re¯ects the fact that pyridine molecules will not be able to enter the pores of a highly coked zeolite. One effect of pore blockage is to reduce the available surface area. This was quanti®ed using nitrogen adsorption measurements, which were analysed to give the available surface area using the standard BET model. The results are shown in Fig. 8. For this particular system there is a simple correlation between the amount of coke present (shown in Fig. 3) and the surface area. The most important aspect of pore blockage is not necessarily the reduction in surface area, but the effect


B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362

Fig. 7. FTIR measurements of adsorbed pyridine on HUSY after a time on stream of: (a) 0 min (fresh catalyst); (b) 15 min; (c) 30 min.

Fig. 8. Plot of available surface area for N2 adsorption as a function of time on stream.

Fig. 9. PFG NMR intensity results for methane adsorbed in zeolite HUSY after the specified time on stream.

of the coke on the molecular mobility of adsorbed molecules (both reactants and products). Diffusion measurements on coked zeolites have been made in the literature using volumetric uptake measurements [10,49,50], an infrared spectroscopy uptake technique [51] and by PFG NMR [24,52]. Self-diffusion measurements using PFG NMR have the advantage that the measurement is performed under equilibrium conditions and that the intracrystalline diffusivity can normally be obtained which is the parameter of interest here. For the fresh zeolite and samples at 15, 30 and 45 min reaction time, the PFG NMR results for methane show single component diffusion behaviour at room temperature: the plot of ln(I/I0) against

2 G2 2 … ÿ =3† is linear with gradient ÿD (see Fig. 9). The measured diffusivity for methane is 8.510ÿ7 m2/s which is similar to the value of 4.610ÿ7 m2/s found by PFGNMR for methane in zeolite NaX [53]. However, interparticle effects will have the major in¯uence on the measured diffusivity as the mean displacement during the measurement is larger than the crystallite size (0.5±1 mm for these HUSY samples), which means that an average of interand intra-particle effects is being observed. For reaction times of longer than 1 h, the PFG NMR results need two diffusion components in order to ®t the data. For these highly coked samples, most of the methane is in the interparticle region so a fast diffusion component is observed. However, the methane mobility

B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362

within the coked particles is now suf®ciently slow that an intracrystalline diffusivity can now be obtained. This diffusivity was found to be in the range 0.9± 2.010ÿ7 m2/s for the samples after 60, 120 and 240 min time on stream. This is still rapid diffusion through the coked sample, and so it was decided therefore to investigate a larger adsorbate in the hope that slower diffusion would allow more accurate values of the intraparticle diffusivity to be obtained. As C4 species are formed in the reaction, it was decided to use n-butane as a probe molecule rather than methane. Some results on n-butane diffusion are shown in Fig. 10. Even in the fresh zeolite, both interand intra-particle effects are apparent, and a twocomponent ®t was necessary to ®t the experimental data. The attenuation curves could in principle be affected by any coke depositions in the intercrystalline space, as well as coke at the surface and within the particles. However, the fast diffusion component, corresponding to interparticle diffusivity, was found to be about 110ÿ8 m2/s for the fresh and the coked catalysts, and depended only slightly on sample and sample packing. This suggests that the slower diffusion component does indeed correspond to the intraparticle diffusivity, and results on this as a function of time on stream are shown in Table 1. It is worth noting that the reactant used, ethene, is a small molecule that is easily able to penetrate throughout the catalyst, and interparticle coke is likely to be more important when considering cracking reactions of large molecules. It can clearly be seen in Table 1 that there is a gradual reduction in n-butane mobility with time on

Fig. 10. PFG NMR intensity results for n-butane adsorbed in zeolite HUSY after the specified time on stream.


Table 1 Diffusivity of n-butane adsorbed within coked HUSY zeolite Time on stream/min

D/10ÿ10 m2/s

0 15 30 45 60 120

23.0 17.7 5.4 1.8 2.0 <0.08

stream during the ®rst 45±60 minutes of the reaction, with the diffusivity dropping by one order of magnitude. This corresponds to the gradual increase in the amount of coke present within the sample. By 120 min reaction time, the diffusivity has become very slow (<810ÿ12 m2/s). This is in agreement with the observation that methane diffusion also slowed signi®cantly after 60 min of reaction, indicating that pore blockage is now signi®cant even for small molecules. 4. Conclusions This work is a self-contained account of the formation of coke and its effects during the ethene oligomerisation reaction on zeolite HUSY at 773 K. The initial stage of coking, the ®rst 15 min of the reaction conditions used here, is rapid and causes partial site poisoning (some loss of acid groups). Both Brùnsted and Lewis acid sites are affected. Diffusion of small molecules is rapid, and thus it is likely that coking is fairly homogeneous throughout the internal pore structure of the catalyst for the reaction studied. As the reaction proceeds, the number of acid sites is gradually reduced and the amount of coke increases. The coke gradually becomes more polyaromatic in character. The diffusivity of small molecules decreases slightly, but the pore space is still accessible, indicating that the increased levels of coke in the large supercages of HUSY are not blocking the cage windows to any great extent. After a certain time, 60 min in the reaction conditions used here, pore blockage becomes signi®cant. Diffusion of small molecules within the catalyst particles becomes slow. Acid sites are still present at this stage but they themselves are gradually consumed over time despite the slow diffusion of reactants. Thus it can be seen that the combination of PFG NMR diffusion measurements together


B. Paweewan et al. / Applied Catalysis A: General 167 (1998) 353±362

with the other characterization techniques used gives particular insight into the deactivation process. Acknowledgements We thank the Cambridge Overseas Trust and the Cambridge Thai Foundation for the funding of Boontham Paweewan's studentship. We thank Dr. Sunil Ashtekar for his help with the infrared spectroscopy experiments. References [1] [2] [3] [4] [5] [6]

[7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19]

[20] [21] [22]

E.E. Wolf, F. Alfani, Catal. Rev. Sci. Eng. 24 (1982) 329. E.G. Derouane, Stud. Surf. Sci. Catal. 20 (1985) 221. M. Guisnet, P. Magnoux, Appl. Catal. 54 (1989) 1. H.G. Karge, Stud. Surf. Sci. Catal. 58 (1991) 531. D.M. Bibby, R.F. Howe, G.D. McLellan, Appl. Catal. 93 (1992) 1. M. Guisnet, P. Magnoux, in: E.G. Derouane et al. (Eds.), Zeolite Microporous Solids: synthesis, structure and reactivity, NATO ASI Series C, Kluwer Academic Publishers, Dordrecht, 1992, p. 437. J. VoÈlter, J. Caro, M. BuÈlow, B. Fahlke, J. KaÈrger, M. Hunger, Appl. Catal. 42 (1988) 15. N. Mori, S. Nishiyama, S. Tsuruya, M. Masai, Appl. Catal. 74 (1991) 37. C. Li, Y.-W. Chen, S.-J. Yang, R.-B. Yen, Appl. Surf. Sci. 81 (1994) 465. P.D. Hopkins, J.T. Miller, B.L. Meyers, G.J. Ray, R.T. Roginski, M.A. Kuehne, H.H. Kung, Appl. Catal. A: General 136 (1996) 29. K.Y. Cheah, P. Alexander, L.F. Gladden, Appl. Catal. A: General 148 (1997) 387. P. Magnoux, P. Cartraud, S. Mignard, M. Guisnet, J. Catal. 106 (1987) 235. S.M. Holmes, A. Garforth, B. Maunders, J. Dwyer, Appl. Catal. A: General 151 (1997) 355. M. Bassir, B.W. Wojciechowski, J. Catal. 150 (1994) 1. H.G. Karge, E. Boldingh, Catal. Today 3 (1988) 379. J.-P. Lange, A. Gutsze, J. Allgeier, H.G. Karge, Appl. Catal. 45 (1988) 345. A.K. Ghosh, R.A. Kydd, J. Catal. 100 (1986) 185. K. Moljord, P. Magnoux, M. Guisnet, Appl. Catal. A: General 122 (1995) 21. M. Guisnet, P. Magnoux, K. Moljord, in: P. O'Connor, T. Takatsuka, G.L. Woolery (Eds.), Deactivation and Testing of Hydrocarbon-Processing Catalysts, ACS Symposium Series, 634, 1996, 77. A.P. Matharu, Ph.D. thesis, University of Cambridge, 1995. R. Ferwerda, J.H. van der Maas, J. Phys. Chem. 99 (1995) 14764. W.T. Dixon, J. Schaefer, M.D. Sefcik, E.O. Stejskal, R.A.

McKay, J. Magn. Reson. 49 (1982) 341. [23] J. KaÈrger, D.M. Ruthven, Diffusion in Zeolites and other Microporous Solids, Wiley, New York, 1992. [24] J. Caro, H. Jobic, M. BuÈlow, J. KaÈrger, B. Zibrowius, Adv. Catal. 39 (1993) 351. [25] S.J. Gibbs, C.S. Johnson, J. Magn. Reson. 93 (1991) 395. [26] E.O. Stejskal, J.E. Tanner, J. Chem. Phys. 42 (1965) 288. [27] B.W. Wojciechowski, Can. J. Chem. Eng. 46 (1968) 48. [28] B.W. Wojciechowski, Catal. Rev. Sci. Eng. 9 (1974) 79. [29] K.A. Cumming, B.W. Wojciechowski, Catal. Rev. Sci. Eng. 38 (1996) 101. [30] E.G. Derouane, J.P. Gilson, J.B. Nagy, Zeolites 2 (1982) 42. [31] J. Weitkamp, S. Maixner, Zeolites 7 (1987) 6. [32] R.H. Meinhold, D.M. Bibby, Zeolites 10 (1990) 121. [33] W.A. Groten, B.W. Wojciechowski, B.K. Hunter, J. Catal. 138 (1992) 343. [34] J.L. Bonardet, M.C. Barrage, J. Fraissard, in: P. O'Connor, T. Takatsuka, G.L. Woolery (Eds.), Deactivation and Testing of Hydrocarbon-Processing Catalysts, ACS Symposium Series, 634, 1996, 99. [35] C.E. Snape, B.J. McGhee, J.M. Andresen, R. Hughes, C.L. Koon, G. Hutchings, Appl. Catal. A: General 129 (1995) 125. [36] C.E. Snape, B.J. McGhee, S.C. Martin, J.M. Andresen, Catal. Today 37 (1997) 285. [37] M.C. Barrage, F. Bauer, H. Ernst, J. Fraissard, D. Freude, H. Pfeifer, Catal. Lett. 6 (1990) 201. [38] K.P. Datema, A.K. Nowak, J. van Braam Houckgeest, A.F.H. Wielers, Catal. Lett. 11 (1991) 267. [39] A.G. Stepanov, V.N. Zudin, K.I. Zamaraev, Solid State NMR 2 (1993) 89. [40] C.E. Snape, D.E. Axelson, R.E. Botto, J.J. Delpuech, P. Tekely, B.C. Gerstein, M. Pruski, G.E. Maciel, M.A. Wilson, Fuel 68 (1989) 547. [41] D.E. Axelson, Solid State Nuclear Magnetic Resonance of Fossil Fuels; Multiscience Publications Ltd., Canada, 1985. [42] P.E. Eberly, J. Phys. Chem. 71 (1967) 1717. [43] P.J. Barrie, Annu. Rep. NMR Spectrosc. 30 (1995) 37. [44] M.C. Barrage, J.L. Bonardet, J. Fraissard, Catal. Lett. 5 (1990) 143. [45] S.-B. Liu, S. Prasad, J.-F. Wu, L.-J. Ma, T.-R. Yang, J.-T. Chiou, J.-Y. Chang, T.-C. Tsai, J. Catal. 142 (1993) 664. [46] J.W. Ward, J. Catal. 13 (1969) 364. [47] D.W. Breck, Zeolite Molecular Sieves, Wiley, New York, 1974, p. 475. [48] A. Corma, Chem. Rev. 95 (1995) 559. [49] T. Masuda, N. Murakami, K. Hashimoto, Chem. Eng. Sci. 47 (1992) 2775. [50] T. Masuda, K. Hashimoto, in: P. O'Connor, T. Takatsuka, G.L. Woolery (Eds.), Deactivation and Testing of Hydrocarbon-Processing Catalysts, ACS Symposium Series, 634, 1996, 62. [51] H.G. Karge, W. Nieûen, H. Bludau, Appl. Catal. A: General 146 (1996) 339. [52] R.Q. Snurr, A. Hagen, H. Ernst, H.B. Schwarz, S. Ernst, J. Weitkamp, J. KaÈrger, J. Catal. 163 (1996) 130. [53] J. Karger, M. BuÈlow, K. Haberlandt, J. Colloid Interface Sci. 60 (1977) 386.