Journal of Catalysis 249 (2007) 195–207

www.elsevier.com/locate/jcat

Conversion of methanol to hydrocarbons over zeolite H-ZSM-5:

On the origin of the olefinic species
Morten Bjørgen a,∗ , Stian Svelle b , Finn Joensen a , Jesper Nerlov a , Stein Kolboe b ,
Francesca Bonino c , Luisa Palumbo c , Silvia Bordiga c , Unni Olsbye b
a Haldor Topsøe A/S, Nymøllevej 55, DK-2800 Lyngby, Denmark
b Center for Materials Science and Nanotechnology, Department of Chemistry, University of Oslo, P.O. Box 1033 Blindern, N-0315 Oslo, Norway
c Department of Inorganic, Physical and Materials Chemistry, NIS Centre of Excellence, and Centro di Riferimento INSTM, University of Turin,
Via P. Giuria 7, I-10125 Torino, Italy
Received 22 February 2007; revised 21 March 2007; accepted 9 April 2007
Available online 5 June 2007

Abstract
This study examined the reaction mechanism with respect to both catalyst deactivation and product formation in the conversion of methanol
to hydrocarbons over zeolite H-ZSM-5. The reactivity of the organics residing in the zeolite voids during the reaction was assessed by transient
12 C/13 C methanol-switching experiments. In contrast to previously investigated catalysts (H-SAPO-34 and H-beta), hexamethylbenzene is virtu-
ally unreactive in H-ZSM-5 and is thus not a relevant reaction intermediate for alkene formation. However, the lower methylbenzenes are reaction
intermediates in a hydrocarbon pool-type mechanistic cycle and are responsible for the formation of ethene and propene. An additional reaction
cycle not applicable for ethene also must be taken into account. The C3+ alkenes are to formed through rapid alkene methylation and cracking
steps to a considerable extent; thus, methanol is converted to hydrocarbons according to two catalytic cycles over H-ZSM-5. Moreover, in contrast
to what occurs for large-pore zeolites/zeotypes, molecules larger than hexamethylbenzenes are not built up inside the H-ZSM-5 channels during
deactivation. Thus, deactivation is explained by coke formation on the external surface of the zeolite crystallites only. This is a plausible rationale
for the superior lifetime properties of H-ZSM-5 in the methanol-to-hydrocarbon reaction.
 2007 Elsevier Inc. All rights reserved.

Keywords: ZSM-5; MFI; MTO; MTH; MTG; Methanol; Zeolite; Mechanism; Hydrocarbon pool; Coking; Deactivation

1. Introduction The MTH chemistry and its commercial potential have been
known for decades. The energy crisis in the 1970s sparked new
The current pricing of crude oil has led to increased interest interest in MTG technology. In the late 1970s, the New Zealand
in the use of natural gas or coal as a feedstock for the pro- government decided to build the world’s first facility for pro-
duction of gasoline and petrochemicals. Methanol can be made ducing gasoline from natural gas via methanol [1]. In 1986,
from such feedstocks via synthesis gas and further converted a plant with a planned annual gasoline production of 600,000
into gasoline-range hydrocarbons or alternatively light, possi- tons over a H-ZSM-5 catalyst was started up, but, due to the
bly polymer-grade alkenes. The zeolite-catalyzed conversion of subsequent drop in crude oil prices, only the methanol synthesis
methanol to hydrocarbons is commonly referred to as the MTH step was left on stream. The Topsøe integrated gasoline synthe-
reaction, but because process conditions and catalyst choice sis (TIGAS) process of Haldor Topsøe AS, also based on H-
alter product selectivity, the abbreviations MTO (methanol to ZSM-5 as the catalyst, was demonstrated on a pilot scale in the
olefins) and MTG (methanol to gasoline) are frequently used mid-1980s [2]. However, the TIGAS process was never scaled
depending on the desired products. up due to the situation in the global energy market. Later, more
attention was paid to the MTO reaction, and the manufactur-
* Corresponding author. ing of polymer-grade ethene and propene from methanol using
E-mail address: [email protected] (M. Bjørgen). the UOP/Norsk Hydro technology was demonstrated in a demo
0021-9517/$ – see front matter  2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.04.006
196 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207

plant [3]. The catalyst choice for the UOP/Norsk Hydro MTO
process is the silicoaluminophosphate H-SAPO-34. Lurgi de-
veloped a process specifically aiming at selectively converting
methanol into propene (MTP) over H-ZSM-5, and the process
was demonstrated in a joint Lurgi/Statoil demo plant.
Today, commercial interest in MTG is again increasing, and
activities involving both natural gas and coal as feedstocks are
emerging. New-generation methanol plants designed to pro-
duce larger amounts of product at significantly lower cost are
due to come online within the next 5 years, which will increase
the profit and interest in methanol downstream activities.
Over the years, the complex mechanism of the MTH reaction
has posed a challenge to catalysis investigators. In a simplified
route of the MTH reaction, methanol is dehydrated under for-
mation of dimethyl ether and water, followed by formation of
alkenes from this equilibrium mixture of oxygenates. The long- Scheme 1. Suggested hydrocarbon pool mechanism for propene formation from
the heptamethylbenzenium cation over zeolite H-beta.
standing question has been the specific mechanism behind this
latter step, in which C–C bonds are formed from oxygenates
(methanol/dimethyl ether). There is presently strong evidence To date, the most extensive fundamental mechanistic un-
disfavoring direct mechanisms in which methanol/dimethyl derstanding has been attained for the H-SAPO-34 and H-beta
ether molecules are combined to form, for example, ethene catalysts, and many aspects may vary with, for example, pore
during steady-state conversion [4–6]. Hunger and co-workers architecture and also possibly catalyst acidic strength/density.
[7,8] have published data indicating that a direct route from In the present contribution, the mechanism of the MTH re-
methanol/dimethyl ether to hydrocarbons may operate during action over H-ZSM-5, the archetype MTH catalyst, was pur-
an induction period. However, this direct route most likely op- sued. Main attention was given to the buildup and reactivity
erates at a rate so low as to be eclipsed by even trace impurities of the organics trapped within the zeolite voids. This work
of higher alcohols/hydrocarbons in the methanol feed, the cata- demonstrates that the reaction scheme for alkene formation
lyst, or the carrier gas. from methanol over H-ZSM-5 is different than those drawn
On the other hand, an all-dominating indirect route known previously for H-beta and H-SAPO-34. In addition to a mod-
as the “hydrocarbon pool mechanism” has been gradually ac- ified hydrocarbon pool mechanism, wherein ethene is formed
cepted in recent years. The hydrocarbon pool mechanism was from the lower methylbenzenes, a parallel C3+ alkene methyla-
conceptually introduced by Dahl and Kolbe in the early 1990s tion/cracking cycle is operative. We mentioned this latter prop-
[9,10], and more recent independent contributions [11–24], erty in a recent communication [25]; the present work sup-
have led to an in-depth description of the MTH reaction mech- plements this with more extensive data and a more thorough
anism. The hydrocarbon pool has been described as a catalytic discussion.
scaffold composed of larger organic molecules adsorbed in the
zeolite to which methanol/dimethyl ether is added and from 2. Experimental
which alkenes and water are formed in a closed cycle, in which
high-energy intermediates are avoided. 2.1. Catalyst and catalyst characterization
The identity and the operation of the hydrocarbon pool
have been fairly well described for some specific systems, in A commercially available H-ZSM-5 (Si/Al = 140) from Ze-
particular zeolite H-beta [16–21] and H-SAPO-34 [13–15]. olyst International was used in this study. Powder X-ray dif-
For these catalysts, the hydrocarbon pool has been identified fraction (XRD) was performed in Bragg-Brentano parafocusing
as highly methyl-substituted benzene rings (“polymethylben- geometry using a Philips X’Pert Pro vertical T-T diffractometer
zenes”) or the benzenium cations derived thereof. In the wide and CuKα radiation. The FTIR transmission spectra were col-
pore beta zeolite, the terminal benzene methylation product lected using a cryogenic cell allowing high-temperature treat-
(i.e., the heptamethylbenzenium cation) is unquestionably the ment of the sample in vacuum. The sample was pretreated at
species responsible for the major part of alkene formation. 500 ◦ C for 1 h and kept at −213 ◦ C during the subsequent
According to the hydrocarbon pool mechanism over the beta CO adsorption/desorption. A detailed description of the cryo-
zeolite, the heptamethylbenzenium cation splits off (at least) genic cell (consisting of a modified closed-circuit liquid He
propene and butene under a concomitant formation of the tetra- Oxford CCC 1204 cryostat) is available elsewhere [26]. The
and trimethylbenzenium ions, respectively. The heptamethyl- spectra were acquired at a resolution of 1 cm−1 by averaging
benzenium cation is then reformed by subsequent methylation 128 interferograms on a Bruker Equinox-55 FTIR spectrome-
of these lower polymethylbenzenium ions (after deprotona- ter. The BET surface area was obtained from a Quantachrome
tion) by methanol/dimethyl ether, thus completing the cycle. Autosorb by N2 adsorption at −196 ◦ C. The sample was out-
A simplified representation of this reaction cycle is outlined in gassed at 350 ◦ C for 18 h before the N2 adsorption measure-
Scheme 1. The hexaMB plays a similar role in H-SAPO-34. ment.
 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207 197

2.2. Catalytic studies rate of 15 ◦ C min−1 . The isotopic compositions of the retained
compounds and the high boiling fraction of the effluent were
The catalytic reactions were carried out in a fixed-bed Pyrex analyzed by a HP 6890N GC–MS (with MSD 5973, EI 70 eV),
reactor using 60 mg of catalyst. The catalyst powder was com- with a HP-5MS column [30 m, 0.25 mm i.d., (5%-phenyl)-
pressed to wafers that were crushed and sieved to obtain 0.25– methylpolysiloxane, stationary phase thickness 0.25 µm]. The
0.42 mm particles. The reaction temperature, measured using following temperature programming was applied: 10 ◦ C min−1
a thermocouple inserted in the catalyst bed, was between 290 from 10 ◦ C (1 min at the initial temperature) to 300 ◦ C (5 min
and 390 ◦ C. Methanol, either 13 C-enriched (Cambridge Iso- at the final temperature). Hexachloroethane was used as the in-
tope Laboratories, 99% 13 C purity) or ordinary 12 C methanol ternal standard for the extracts. Similar analyses for the light
(BDH Laboratory Supplies, >99.8% chemical purity) was fed effluent compounds were performed on a HP 6890 GC–MS
by passing a He carrier gas stream, maintained at 35 mL min−1 , (with MSD 5973, EI 70 eV), with a J&W GasPro column (60 m,
through a saturation evaporator kept at 20 ◦ C. The resulting feed 0.32 mm i.d.). The following temperature programming was ap-
partial pressure was 130 hPa, giving a feed rate (WHSV) of plied: 20 ◦ C min−1 from 10 ◦ C (5 min at the initial temperature)
7.0 g g−1 h−1 . to 200 ◦ C (25 min at the final temperature).
Three types of experiments were performed in the mecha-
nistic investigation: 3. Results and discussion

1. Ordinary 12 C methanol was reacted for predetermined 3.1. Physical characteristics of the catalyst
times at given reaction temperatures. Effluent analyses
were performed during the reaction, and analyses of the Powder XRD confirmed that the sample consists of a pure
organic material retained in the zeolite (see below) were and well crystalline MFI phase. The XRD peak broadening
carried out after each run. analyses suggests a crystallite size (scattering domain size)
2. Alternatively, the methanol feed was stopped after a given larger than 100 nm. SEM showed nonuniformly shaped parti-
time by bypassing the evaporator and flushing the catalyst cles in the size range 0.5–5 µm (average, around 2–3 µm). The
with carrier gas at the reaction temperature for a set time, BET surface area was 370 m2 g−1 .
followed by analysis of the retained material. Fig. 1 reports a series of IR spectra (1–9) of the zeolite in
3. Using two separate and identical feed lines, it was possi- the ν(OH) region. Spectrum 1 represents the pretreated zeo-
ble to switch from 12 C to 13 C methanol without otherwise lite before CO adsorption, whereas spectra 2–9 were recorded
disrupting the experimental conditions. The isotopic com- at increasing CO pressures at −213 ◦ C. The inset displays the
positions of both effluent compounds and retained material difference spectra 2′ –9′ in the ν(CO) region. The rather large
were determined at increasing 13 C methanol reaction times
(see Section 3.2.5).

2.2.1. Analysis of organic material retained in the zeolite

After the catalyst sample was removed from the reactor and
rapidly cooled to room temperature, a part of the catalyst mass
(40 mg) was transferred to a screw-cap Teflon vial and dis-
solved in 1.0 mL of 15% HF. Acid treatment for 1 h was
sufficient to dissolve the catalyst completely; further increases
treatment time did not change the results. Then 1.0 mL CH2 Cl2
(Fluka puriss) was added to the solution thus obtained to extract
the liberated organic molecules from the aqueous phase. After
15 min, the organic phase and the water phase were separated,
and the CH2 Cl2 extract was analyzed using a gas chromato-
graph with a mass sensitive detector (GC–MS). The technique
of dissolving the zeolite crystals in HF to recover confined
constituents by extraction has been used previously in related
studies [14,15,18,21,25,27] and was introduced by Guisnet et
al. [28].

2.2.2. Chromatography
The reactor effluent was analyzed quantitatively by online
GC (Carlo Erba, GC6000-Vega series 2, with a flame ioniza-
tion detector) with a Supelco SPB-5 column (60 m, 0.53 mm Fig. 1. FTIR spectra of the H-ZSM-5 sample. Spectrum 1 (black) was recorded
i.d., stationary phase thickness 5 µm), applying temperature for the zeolite after outgassing at 500 ◦ C. Spectra 2–9 (gray) were collected at
programming starting at 50 ◦ C (6 min at the initial tempera- increasing pressures of CO at −213 ◦ C. The inset shows the difference spectra
ture) and heating to 260 ◦ C (5 min at the final temperature) at a (spectrum 1 subtracted from spectra 2–9) in the ν(CO) region.
198 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207

crystallite size gives a pronounced scattering background start- Table 1

ing at 3000 cm−1 , and the band associated with the external Product selectivities (C%) and methanol/dimethyl ether conversion (%) at
Si–OH groups (3749 cm−1 ) has a modest intensity. The other 370 ◦ C

general features of the infrared spectrum of the sample are as TOS Conv. C1 C2 C3 C4= C4− C5 C6+
expected for a zeolite with a relatively low Al content. Two fam- (min)
ilies of OH groups can be clearly distinguished in spectrum 1: 10 85.0 0.2 6.1 32.8 16.6 5.7 13.4 25.2
15 84.9 0.2 6.2 32.9 16.8 5.5 13.3 25.1
(i) isolated Si–OH groups located on the external surface (max- 20 84.4 0.2 6.2 32.5 16.5 5.5 13.2 25.9
imum at 3749 cm−1 ) and in internal defects (low-frequency 30 85.4 0.2 6.0 32.3 16.5 5.3 13.4 26.3
shoulder/tail on the main band) and (ii) strongly acidic-bridging 60 85.1 0.2 6.2 33.0 16.9 5.3 13.0 25.4
Si–OH–Al groups in the framework of the zeolite (maximum 1200 82.5 0.2 7.1 32.9 16.6 5.2 12.8 25.2
at 3618 cm−1 ). In addition, a weaker absorption, representing (20 h)
mainly partially hydrolyzed framework Al (expected at around 20160 10.8 1.2 10.2 34.0 20.6 3.0 12.0 19.0
(14 days)
3660–3680 cm−1 ), can be discerned between these two com-
ponents. ν(OH) vibrations of extra-framework Al–OH groups,
which normally appear at 3782 cm−1 , cannot be seen for this
sample. The acidic strength of these species cannot be as-
sessed by direct measurement of the vibrational frequency of
the unperturbed OH groups, but by the shift ν caused by
the interaction with a basic molecule used as sensitive elec-
tric field probe. When adsorbing a base, the observed red shift
of the ν(OH) band will be proportional to the Brönsted acidic
strength of these species [29–32]. CO is one of the most widely
used probe molecules for acidity measurements of zeolites,
and extensive data are available [29,33]. Spectra 2–9 show the
perturbations induced on the ν(OH) modes at increasing CO
pressures. The original ν(OH) modes of the zeolite are grad-
ually consumed while the ν(OH) vibrations of the OH· · ·CO Fig. 2. GC–MS analysis of retained material after 20 min of methanol conver-
adducts grow in a parallel manner at lower frequencies. The sion at 370 ◦ C.
band at 3618 cm−1 becomes completely eroded and gives rise
to a rather broad band with maximum around 3270 cm−1 In this report, main attention is given to the buildup and sta-
(ν(OH) = −350 cm−1 ). The Si–OH groups are partially con- bility of the organic material in the zeolite pores during the
sumed and shifted to 3645 cm−1 (ν(OH) = −100 cm−1 ). reaction. Fig. 2 presents the GC–MS total ion chromatogram
Formation of the OH· · ·CO adducts is also accompanied by a of the organics present inside the zeolite after 20 min of
perturbation of the internal vibrational mode of CO. The per- methanol reaction at 370 ◦ C. The chromatogram in Fig. 2
turbation of the ν(CO) mode is displayed in the inset. Starting was obtained by thermally quenching at 20 min on stream,
from low coverage, three components are visible at 2137, 2163, dissolving the zeolite framework in HF, extracting the or-
and 2175 cm−1 . Their respective assignments are free liquid ganic material with CH2 Cl2 , and finally analyzing this ex-
like CO, CO adsorbed on weakly acidic Si–OH sites or partially tract. As Fig. 2 shows, a few compounds are dominant among
extra-lattice Al–OH, and CO adsorbed on the strongly acidic the retained organics, and virtually all peaks represent poly-
Brönsted sites. In compliance with the other characterization methylbenzenes (polyMBs). Hexamethylbenzene (hexaMB) is
techniques, these data confirm that the zeolite under study is a most abundant, followed by pentamethylbenzene (pentaMB),
representative H-ZSM-5. The low Al content has not increased tetramethylbenzenes (tetraMBs), trimethylbenzenes (triMBs),
the acidic strength of the Brönsted sites. This latter observation and p/m-xylene. Hydrocarbons with greater molecular mass
is important, because it is allows us to consider the mechanistic than hexaMB were not present in appreciable amounts. (This
conclusions of this work to be rather general for H-ZSM-5, not is not due to an experimental limitation, because the setup al-
for just one specific sample. lows analysis of compounds as least as heavy as methylpyrene.)
PolyMBs up to tetraMB were also detected in the gas phase;
3.2. Conversion of methanol over H-ZSM-5 thus, to be precise, only penta- and hexaMB can be considered
truly trapped species in this context. Previous studies on the or-
3.2.1. Gas-phase products and retained material at 370 ◦ C ganics formed inside wide-pore zeolites/zeotypes (H-beta and
and 20 min on stream H-SAPO-34) under MTH conditions have shown similar results
At 370 ◦ C, the conversion was 85% (dimethyl ether [DME] [15,21], but two essential differences should be underlined:
included in reactants) over the fresh zeolite (time on stream =
20 min), and the main products were propene, butenes, and 1. The well-defined cutoff in the distribution of confined or-
various C5 –C10 hydrocarbons. The detailed composition of the ganics at hexaMB (Fig. 2) must be a result of the restricted
effluent, which is typical for siliceous H-ZSM-5 at these condi- space in the channels of ZSM-5, because far larger polyaro-
tions, is shown in Table 1. matics are found in copious amounts in zeolites/zeotypes
 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207 199

providing more spacious voids. This point is further cor-

roborated in Section 3.2.2.
2. The accumulation of hexaMB in the zeolite cavities at this
rather high temperature (and conversion) points to a low
reactivity of this species. This is in stark contrast to the
aforementioned investigations in which hexaMB [15] (and
heptaMB+ ) [21] were found to be highly unstable and
present in only minute amounts above 325 ◦ C.

3.2.2. Retained material at increasing time on stream

at 370 ◦ C
Fig. 2 shows that traditional coke-like compounds (polyaro-
matics) cannot be found in H-ZSM-5 during the MTH reaction,
at least not after 20 min of reaction. This may be due to steric
limitations or, alternatively, an insufficiently long reaction time
(20 min) to allow the buildup of appreciable amounts of pol-
yaromatics. To clarify this, a series of experiments with an ex-
tended reaction time (from 10 min to 14 days) were performed.
Analyses of the gas-phase products were carried out in parallel
to the HF dissolution experiments. The results, given in Table 1,
show that a conversion of about 85% was maintained for sev-
eral hours, but that it dropped to 82.5% after 20 h on stream.
After 14 days of reaction, the conversion decreased to 10.8%,
and the low activity of the zeolite was accompanied by a lower
hydride transfer activity, as reflected in the lower C6+ selectiv-
ity (less aromatics) and the higher alkene/alkane ratio in, for
example, the C4 fraction.
Fig. 3 shows the GC–MS details of the CH2 Cl2 extracts ob-
tained at the same reaction times as those given in Table 1.
The concentration of all polyMBs increased consistently up to
20 h. The concentration of hexaMB increased by a factor 3.3
when the reaction time was increased from 10 min to 20 h.
The corresponding factors for the lower polyMBs were 1.1
for p/m-xylene, 1.3 for the triMBs, 1.2–1.7 for the tetraMBs Fig. 3. GC–MS analyses of retained material after methanol conversion at
(isomer-dependent), and 1.9 for pentaMB. All methylbenzenes 370 ◦ C.
attained a steady-state concentration after 60 min on stream ex-
cept hexaMB, which increased by a factor of 1.3 from 60 min to
20 h. This suggests that methylation of pentaMB into hexaMB organic constituents (isolated after HF dissolution) of a deac-
is sterically restricted, because complete methylation is thermo- tivated H-SAPO-34. The experiment was conducted at 350 ◦ C,
dynamically favored. and the distribution of retained organics spanned benzene to
The changes in absolute and relative concentrations of the methylpyrenes, with different methylated naphthalene isomers
organics in the zeolite channels observed during the first 60 min as the prevailing species. The concentration of organics in the
of reaction did not affect the effluent composition. By increas- deactivated H-SAPO-34 is orders of magnitude higher than that
ing the reaction time to 20 h, the conversion dropped slightly, found for H-ZSM-5.
with the only change seen in the confined material an increase A study in progress, specifically devoted to coke formation
in hexaMB. However, the increased amount of hexaMB in the on H-ZSM-5, has shown a considerable discrepancy between
zeolite cannot be related to the conversion drop, because the the total coke content determined by temperature programmed
14-day experiment (Fig. 3) ruled out any correlation between oxidation and the mass of the confined organics determined by
the concentration of organics in the pores and the zeolite de- HF dissolution for a deactivated H-ZSM-5 (Si/Al = 45): The
activation. It is also interesting to note that despite the fact total coke content of the deactivated sample was 28% by mass,
that the catalyst suffered from severe deactivation after 14 days whereas the material isolated after HF dissolution constituted
on stream, no larger aromatics were found. Thus, deactivation only about 0.08% of the total mass of the spent catalyst. This
must result from the formation of large graphitic species (not finding indicates that virtually all of the carbon present on a
soluble in CH2 Cl2 ) on the external surface of the crystallites. deactivated H-ZSM-5 is insoluble in CH2 Cl2 and thus is of
This is in strong contrast to the zeolites/zeotypes with more graphitic character.
spacious cavities in which coke is readily formed, resulting in Behrsing et al. investigated coke formation from methanol
rapid deactivation. This is borne out in Fig. 4, which shows the over H-ZSM-5 (with Al content ranging from 0.7 to 2.7%
200 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207

Fig. 4. GC–MS analyses of retained material in severely deactivated H-SAPO-34 (top) and H-ZSM-5 (bottom). The intensity scales are directly comparable.

w/w) using transmission electron microscopy (TEM) and found Table 2

that the catalysts became deactivated only when the coke de- Product selectivities (C%) and methanol/dimethyl ether conversion (%) at
posits contained packets of aromatic sheets of carbon parallel 20 min on steam
to the catalyst surface [34]. Similar results, also in line with Temp. Conv. C1 C2 C3 C4= C4− C5 C6+
our conclusions, were reached by Gallezot et al. in a study in (◦ C)
which n-heptane was reacted over H-ZSM-5 [35]. TEM and 290 0.003 0.0 44.0 56.0 0.0 0.0 0.0 0.0
310 4.8 0.5 17.5 33.2 17.8 8.4 9.3 13.4
electron energy loss spectroscopic measurements revealed that
330 51.6 0.2 13.7 29.7 12.4 5.1 10.2 28.7
the coke was formed as envelopes around the zeolite crys- 350 73.0 0.2 9.9 30.4 14.3 5.2 11.7 28.4
tals. 370 84.9 0.2 6.2 32.9 16.8 5.5 13.3 25.1
The current data strongly suggest that an all-dominating 390 91.0 0.2 4.8 34.5 17.4 5.2 13.9 24.0
external coking of H-ZSM-5 explains the superior lifetime
properties of this catalyst. We should address the general va-
lidity of this statement, because it is well known that the hardly observable above 300 ◦ C due to rapid decomposition
process of coke formation (i.e., cyclization and hydrogen trans- into alkenes and lower polyMBs. Consequently, we conducted
fer reactions) is closely related to the acidic site density of a temperature series for ZSM-5 to observe possible species that
the zeolite. The present sample has a rather low Al content are too reactive to be seen at more realistic temperatures. Efflu-
(Si/Al = 140), implying slow aromatization reactions. But ent analyses were carried out in parallel at the different reactor
we carried out experiments for H-ZSM-5 samples with Si/Al
temperatures (Table 2), and the reaction was stopped and the
ratios as low as 25 and found that the overall picture did
zeolite dissolved in HF after 20 min on stream. Table 2 shows
not seem to change dramatically; no additional compounds
that at the lowest temperature, the conversion was close to zero
(e.g., aromatics larger than hexaMB) appeared, even though
(0.003%), and the only detectable products were ethene and
the relative and absolute concentrations of the confined or-
propene. At the highest temperature (390 ◦ C), the conversion
ganics differed somewhat. A more comprehensive study in-
volving a broader selection of ZSM-5 samples is currently in increased to 91%.
progress. Fig. 5 presents the composition of the retained material
formed after 20 min of reaction at the different temperatures.
3.2.3. Retained material after 20 min on stream at 290–390 ◦ C Note that the chromatogram representing experiment conducted
A previous study found that it was necessary to reduce the at 290 ◦ C is shown at higher (10×) sensitivity. At 290 ◦ C, the
reaction temperature to 300 ◦ C to observe the most reactive concentration of retained organics was very low, and pentaMB
retained species in the beta zeolite during the MTH reaction seemed to dominate, whereas triMBs can hardly be discerned.
[21]. In that case, heptaMB+ (isolated as its corresponding Increasing the temperature to 310 ◦ C gave a methanol conver-
base 1,2,3,3,4,5-hexamethyl-6-methylene-1,4-cyclohexadiene sion of 4.8% and orders-of-magnitude increases in the concen-
after HF dissolution and extraction) was the dominant com- trations of the organic material in the zeolite pores. Methanol
pound in the zeolite pores at the lower temperatures, but was conversion increased to 51.6% with a further temperature in-
 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207 201

Fig. 6. GC–MS analyses of retained material after 20 min of methanol conver-

sion at 370 ◦ C (bottom) and 20 min of reaction followed by 8 min of flushing
with carrier gas at 370 ◦ C.

3.2.4. Stability of the retained organic material: Flushing with

carrier gas at 350 ◦ C
To obtain information on the stability of the confined hydro-
carbons, an experiment was carried out in which the feed was
stopped after 20 min of reaction and the catalyst was flushed
with carrier gas for 2, 4, and 8 min at the reaction temperature.
Fig. 6 shows the effect of catalyst flushing on the hydrocarbon
deposits by comparing the analysis from the 8-min flushing ex-
periment with that obtained after 20 min of reaction. In view of
what we have learned from other zeolites, the results are rather
surprising. Similar experiments on H-beta and H-SAPO-34, but
Fig. 5. GC–MS analyses of retained material after 20 min of methanol conver- at a considerably lower reaction temperature (325 ◦ C), led to a
sion. total disappearance of hexaMB at 1–2 min after methanol cutoff
[15,18].
crease to 330 ◦ C; the temperature rise had the most pronounced On flushing, the polyMBs disappeared from the zeolite chan-
effect on the 1,3,5-triMB concentration, which was increased nels at rates according to the following order: p/m-xylene
by a factor of 14. The concentrations of the remaining com- > 1,3,5-triMB > 1,2,4-triMB > 1,2,4,5-tetraMB > 1,2,3,5-
pounds increased in the order 1,2,4-triMB > 1,2,3,5-tetraMB tetraMB > pentaMB = hexaMB.
> p/m-xylene > hexaMB > 1,2,4,5-tetraMB > pentaMB. The On the basis of the findings of the flushing experiment, we
concentration of the retained material seemed to reach a max- can draw the following conclusions:
imum between 330 and 350 ◦ C. A further increase in temper-
ature led to lower concentrations of all hydrocarbons in the 1. Penta- and hexaMB have a similar decomposition rate af-
zeolite channels. Note that the concentration of 1,3,5-triMB ter methanol cutoff, provided that their diffusion rates are
suffered more than that of the 1,2,4-isomer in this temperature equal (zero).
regime. 2. Penta- and hexaMB show very slow decomposition rates
It now can be stated that no new possible intermediates in the zeolite compared with those found for other zeo-
or heptaMB+ appeared at lower temperatures, and a reaction lites/zeotypes [15,18].
route similar to the previously suggested mechanism based on 3. The tetraMBs are very persistent during flushing.
heptaMB+ is not likely to be valid for H-ZSM-5. Obviously, 4. Apparently, 1,3,5-triMB disappears faster than the 1,2,4-
the channels were too narrow to allow the terminal methylation isomer. This is opposite to what would be expected from
step in which hexaMB forms heptaMB+ . The previously men- the relative diffusion rates, because 1,3,5-triMB is the
tioned restricted methylation of pentaMB into hexaMB further most sterically demanding. The explanation for this finding
supports this conclusion. could be that 1,3,5-triMB decomposes rapidly, or simply
202 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207

that 1,3,5-triMB isomerizes into 1,2,4-triMB, which then which strongly disfavors any differences in reactivity. In con-
diffuses out of the zeolite crystals. trast, for the two tetraMB isomers, the isotopic distributions
and total 13 C contents differed significantly. At 0.5 min of
Indeed, we should mention that the observed slow decomposi- 13 C methanol reaction, 1,2,4,5-tetraMB contained 40% labeled

tion of penta- and hexaMB during the flushing experiment does carbons, whereas 1,2,3,5-tetraMB contained 52%. An explana-
not necessarily imply low reactivity in the working zeolite. Af- tion for these differences is not apparent, but 1,2,3,5-tetraMB
ter several minutes of flushing, the concentration of methanol is somewhat bulkier than 1,2,4,5-tetraMB and thus will re-
in the pores is low, and the slow decomposition of penta- and side within the zeolite channels for a longer time. Indeed, only
hexaMB may not reflect their reactivity in the presence of 1,2,4,5-tetraMB is present in appreciable amounts in the efflu-
methanol. The reactivity of the different retained species in the ent. However, as a whole, the behavior of both tetraMB isomers
presence of methanol is further corroborated in the next section. is consistent with the general trend of a decreasing rate of 13 C
incorporation with an increasing the number of methyl groups.
3.2.5. Switching to 13 C methanol after having built up
organics from 12 C methanol: rate of 13 C incorporation in the 3.2.5.2. Gas-phase products The isotopic compositions of
retained organic material and gas phase products the effluent compounds were also determined after 0.5, 1, and
To assess the reactivity patterns of the confined organics 2 min of 13 C methanol reaction; the data thus obtained are
toward methanol (i.e., toward the reactant) in the working cata- shown in Fig. 8. Some compounds are present both in the efflu-
lyst, a series of experiments were conducted in which ordinary ent and the retained material, that is, p/m-xylene, 1,2,4-triMB,
12 C methanol was reacted for 18 min followed by a rapid switch and 1,2,4,5-tetraMB. The trend for these compounds seen in the
to 13 C methanol feed, reacted for a shorter time (0.5, 1, or retained material is lost in the effluent. The data for p/m-xylene
2 min). The compositions of both retained material and effluent and 1,2,4-triMB are fairly similar in both the retained material
were practically the same in all three experiments—the same and the effluent, but 1,2,4,5-tetraMB is considerably richer in
as that shown for 20 min on stream in Fig. 5 and Table 2, re- 13 C in the gas phase. In fact, in the gas phase, the 13 C content

spectively. Isotopic analysis of both the retained material and of 1,2,4,5-tetraMB is very similar to that seen for p/m-xylene
the gas-phase products was performed after 0.5, 1, and 2 min of and 1,2,4-triMB.
13 C methanol reaction. For all effluent compounds, the 13 C methanol exchange is
fairly rapid (Fig. 8), but some important differences, particu-
3.2.5.1. Retained material Fig. 7 shows isotopic data for the larly evident at short 13 C reaction times, have valuable mecha-
material isolated after dissolving the used catalyst with HF and nistic implications. Based on the data in Fig. 8, the compounds
extracting with CH2 Cl2 . Panel (a) displays the isotopomer dis- may be divided into two groups: The rates of 13 C incorporation
tributions for the three different 13 C methanol reaction times; are significantly slower for ethene and the aromatics compared
panel (b) shows the time developments of the total 13 C contents with the C3 –C6 alkenes. This is reflected in the isotopomer dis-
in the different molecules. Clearly, there are large differences in tributions for ethene and the aromatics, which are broader than
both 13 C distributions and total content among the various com- those observed for the C3 –C6 alkenes, in the sense that the pure
pounds. Further, Fig. 7 shows an evident trend for the higher 13 C isotopomers are more dominant for the latter group. The

methylbenzenes: The incorporation of 13 C consistently slows grouping is even more striking when comparing the time de-
as the number of methyl substituents increases. After 0.5 min velopments in the total 13 C content. For the sake of clarity,
of 13 C methanol reaction, xylenes and triMBs display broad Fig. 9 collects these data in a single plot for all compounds.
distributions centered around the isotopomers that are highly Clearly, the C3 –C6 alkenes are systematically richer in 13 C than
rich in 13 C. The tetraMB isotopomers are also broadly distrib- ethene and the aromatics, and the 13 C contents virtually overlap
uted, but the molecules that do not contain any 13 C are the most within the two compound groups. These results suggest a mech-
abundant. This latter observation is even more pronounced for anistic link between the ethene formation and the aromatics in
pentaMB and in particular for hexaMB. After 1 and 2 min of the effluent. Moreover, the significantly higher 13 C content in
13 C methanol reaction, the distributions for the lower methyl- the C3 –C6 alkenes suggests a different or additional route for
benzenes are skewed toward complete 13 C incorporation, and methanol incorporation into these compounds.
the distributions become bimodal for the tetraMBs and pen-
taMB, whereas hexaMB remains dominated by isotopomers not 3.3. Reaction mechanism for the MTH reaction over H-ZSM-5
containing 13 C. The total 13 C contents displayed in Fig. 7 also
reflect the trend in reactivity among the retained methylben- In this report we have shown that methylbenzenes up to and
zenes. Similar isotope experiments conducted for the H-SAPO- including hexaMB are present in copious amounts in H-ZSM-5
34 [14] and H-beta [21] catalysts have shown the reverse trend, during the MTH reaction. Isolated, this finding is in accord
with hexaMB having the highest rate of 13 C incorporation. with similar investigations of other related catalysts. However,
Two minor points regarding the methylbenzene isomers also in stark contrast to all previous conclusions [13–15,18,21], the
should be mentioned. In the flushing experiment (discussed highest methylbenzenes do not appear to play a significant part
in Section 3.2.4), 1,3,5-triMB disappeared faster than 1,2,4- in the mechanism for alkene formation. PentaMB and hexaMB
triMB after methanol cutoff; however, identical isotopic data are present in the working zeolite at realistic reaction temper-
were obtained for the two retained triMB isomers (Fig. 7), atures, contradictory to the expected behavior of a reactive in-
 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207 203

Fig. 7. Isotopic distributions (panel a, left) and total 13 C content (panel b, right) in retained organic material after 18 min of 12 C methanol reaction followed by
a switch to 13 C methanol at 350 ◦ C. 13 C methanol was reacted for 0.5 min (black bars), 1 min (white bars), and 2 min (gray bars) after 12 C methanol reaction.
204 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207

Fig. 8. Isotopic distributions (panel a, left) and total 13 C content (panel b, right) in the effluent compounds after 18 min of 12 C methanol reaction followed by
a switch to 13 C methanol at 350 ◦ C. 13 C methanol was reacted for 0.5 min (black bars), 1 min (white bars), and 2 min (gray bars) after 12 C methanol reaction.
 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207 205

Scheme 2. Suggested mechanism for propene formation via the alkene crack-
ing/methylation route.

pointing the exact extent of propene formation via this route

is the aim of further studies already in progress. It should be
pointed out, however, that the addition of large amounts of a
reactant (benzene/toluene) in addition to methanol might lead
to results less general in nature. The present data do not suffer
from the perturbation that a co-reactant may induce, because
methanol is the only reactant in all experiments.
Fig. 9. Total 13 C content in the effluent compounds after 18 min of 12 C
methanol reaction followed by a switch to 13 C methanol and further reaction
The C3 –C6 alkenes contain more 13 C than all other prod-
for 0.5, 1.0, and 2.0 min at 350 ◦ C. Note the scale break. ucts. This enrichment must be attributed to 13 C incorporation
by another mechanism, not viable for ethene, and the only sat-
termediate, and are also stable during flushing after methanol isfactory explanation is an alkene methylation reaction. Kinetic
cutoff at 350 ◦ C. In addition, the isotope data decisively con- investigations have shown that ethene methylation is between
firm that, compared with the lower methylbenzenes, pentaMB one and two orders of magnitude slower than for propene and
and hexaMB are inert toward methanol. Finally, heptaMB+ was linear butanes [37,38]; this has been further supported by quan-
not observed, despite the broad range of reaction conditions tum chemical studies [39]. Thus, ethene methylation is a very
studied. As a direct implication, a hydrocarbon pool mechanism minor reaction under the current set of reaction conditions. It
involving the highest methylbenzenes or heptaMB+ , proven to thus seems very probable that additional 13 C is introduced into
be dominant for H-beta and H-SAPO-34 [14,15,18,21], cannot the C3 –C6 alkenes via repeated methylations, and that the car-
be responsible for any appreciable activity in the H-ZSM-5 cat- bon atoms are distributed among the product fractions through
alyst. subsequent cracking steps. Such a sequence of reactions, il-
However, the data suggest that a hydrocarbon pool mecha- lustrating how three molecules of methanol are converted into
nism involving the xylenes and triMB as the reactive intermedi- propene, is outlined in Scheme 2. In this scheme, the reac-
ate operates over H-ZSM-5. Based on the practically identical tion cycle starts with propene and ends up with hexene cracked
13 C contents and time evolutions (Fig. 9), ethene appears to be into two molecules of propene. This is, of course, a simplifica-
formed predominantly via the xylenes and/or the triMBs. The tion, because far larger alkenes may be formed before they are
C3 –C6 alkenes do not display this clear correspondence with cracked, and the cracking products may be alkenes higher than
any of the polymethylbenzenes. Haw and co-workers [23,36] propene. Dimerization also might occur, but this will not lead
have indeed shown that the lower methylbenzenes may split to additional 13 C incorporation.
off mainly ethene and that the higher analogues yield predom- As was the case for methylation, ethene dimerization is also
inantly propene. It is not possible to assess the exact mecha- slow compared with that of propene and butenes [40]. The
nism by which ethene is lost from the pool compounds from importance of alkene methylation and cracking in the MTH
the current data, but several relevant suggestions exist in the reaction was recognized early and pointed out in particular
literature. In experiments where 13 C methanol was co-reacted by Dessau and co-workers at Mobil Oil [41]. In the original
with benzene or toluene at steady state conditions over H- methylation/cracking reaction schemes, however, ethene was
ZSM-5, Mikkelsen et al. [12] observed a relation between the included, even though its slow formation through cracking was
methylbenzenes and both ethene and propene formation. In one emphasized. Indeed, in the cracking of 3,4-dimethyl-3-hexene
case where 12 C benzene and 13 C methanol were co-reacted, over the Y zeolite at 350 ◦ C, the molar yields for propene
the ethene thus formed contained significantly less 13 C than and butenes have been shown to be 14 and 33 times higher
propene, suggesting that a larger fraction of the ethene relative than the yield of ethene [42]. Dessau [41] unsatisfactorily con-
to propene was formed via methylbenzenes. This is in com- cluded that ethene was formed by a secondary re-equilibration
pliance with our results. However, when 12 C toluene and 13 C of primary alkenes. This is not in agreement with our find-
methanol were co-reacted, no significant distinction between ings, because the discrepancy between the 13 C content in ethene
ethene and propene could be made. and the C3 –C6 alkenes excludes any measurable ethene forma-
Based on these observations, it seems reasonable to assume tion from the higher alkenes, however secondary in nature. It
that a fraction of the propene is formed from the aromatics. Pin- should be noted that based on the current level of mechanis-
206 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207

1. The higher methylbenzenes (penta- and hexamethylben-

zene) are present in the zeolite pores but are virtually
unreactive, both with respect to flushing with carrier gas
and incorporation of 13 C after switching from 12 C to 13 C
methanol feed. This poor reactivity is contrary to observa-
tions made for H-beta and H-SAPO-34.
2. Deactivation is not an effect of coke formation inside the
zeolite voids. External coking appears to be the only plau-
sible cause of the loss of activity with time on stream,
which might explain the outstanding catalytic stability of
H-ZSM-5.
3. Ethene displays the same 13 C content and time evolution as
p/m-xylene and the triMBs after switching and appears to
be formed predominantly from the lower methylbenzenes
through a modified hydrocarbon pool mechanism.
4. In contrast to ethene, propene and higher alkenes are
Scheme 3. Suggested dual cycle concept for the conversion of methanol over formed to a considerable extent from alkene methylations
H-ZSM-5. and interconversions (e.g., cracking reactions).

tic insight, in particular the shift of focus from coupling of C1 The latter two statements, implying that ethene and propene
units (methanol/dimethyl ether) to the more indirect hydrocar- are formed through different routes, are of utmost importance
bon pool mechanism, the designation of products as “primary” for understanding and possibly controlling the ethene/propene
or “secondary” often leads to confusion rather than clarifica- selectivity in MTO/MTP catalysis.
tion. Clearly, the reactant molecules (methanol) undergo many
reaction steps before the formation of even the simplest C–C References
bond containing product in the zeolite voids.
Based on the present results, it can be established that [1] M. Stöcker, Microporous Mesoporous Mater. 29 (1999) 3.
[2] J. Topp-Jørgensen, Stud. Surf. Sci. Catal. 36 (1988) 293.
two mechanistic cycles run simultaneously during the MTH
[3] J.Q. Chen, A. Bozzano, B. Glover, T. Fuglerud, S. Kvisle, Catal. To-
reaction over H-ZSM-5: ethene formation from the lower day 106 (2005) 103.
methylbenzenes followed by remethylation, and a methyla- [4] W. Song, D.M. Marcus, H. Fu, J.O. Ehresmann, J.F. Haw, J. Am. Chem.
tion/cracking cycle involving only the C3+ alkenes. This is Soc. 124 (2002) 3844.
depicted in Scheme 3. Cycle I in Scheme 3 is the aromat- [5] D.M. Marcus, K.A. McLachlan, M.A. Wildman, J.O. Ehresmann, P.W.
ics/ethene (with aromatics formally represented by toluene and Kletnieks, J.F. Haw, Angew. Chem. Int. Ed. 45 (2006) 3133.
[6] D. Lesthaeghe, V. Van Speybroeck, G.B. Marin, M. Waroquier, Angew.
trimethylbenzene) cycle, whereas cycle II is the alkene-based Chem. Int. Ed. 45 (2006) 1714.
cycle. A highly interesting question regarding both mechanis- [7] W. Wang, A. Buchholz, M. Seiler, M. Hunger, J. Am. Chem. Soc. 125
tic understanding and selectivity control is whether cycles I (2003) 15260.
and II operate completely independently or if they are inter- [8] Y. Jiang, W. Wang, V.R.M. Marthala, J. Huang, B. Sulikowski, M. Hunger,
twined in some manner. According to our results, a completely J. Catal. 238 (2006) 21.
[9] I.M. Dahl, S. Kolboe, Catal. Lett. 20 (1993) 329.
independent operation can hardly be the case for H-ZSM-5. [10] I.M. Dahl, S. Kolboe, J. Catal. 149 (1994) 458.
There is a continuous production of aromatics during the re- [11] P.W. Goguen, T. Xu, D.H. Barich, T.W. Skloss, W. Song, Z. Wang, J.B.
action; hence there must be a corresponding aromatization of Nicholas, J.F. Haw, J. Am. Chem. Soc. 120 (1998) 2650.
higher alkenes formed in cycle II. This implies that cycle I [12] Ø. Mikkelsen, P.O. Rønning, S. Kolboe, Microporous Mesoporous
cannot run without cycle II over H-ZSM-5. However, the two Mater. 40 (2000) 95.
[13] W. Song, J.F. Haw, J.B. Nicholas, C.S. Heneghan, J. Am. Chem. Soc. 122
cycles might not be mutually dependent. Based on the low re- (2000) 10726.
activity of ethene toward methanol relative to that of propene [14] B. Arstad, S. Kolboe, J. Am. Chem. Soc. 123 (2001) 8137.
and butenes mentioned above, the contribution to the C3 –C6 [15] B. Arstad, S. Kolboe, Catal. Lett. 71 (2001) 209.
alkenes involving ethene is very small and possibly not re- [16] A. Sassi, M.A. Wildman, H.J. Ahn, P. Prasad, J.B. Nicholas, J.F. Haw,
quired. However, the significance of propene formation from J. Phys. Chem. B 106 (2002) 2294.
[17] A. Sassi, M.A. Wildman, J.F. Haw, J. Phys. Chem. B 106 (2002) 8768.
the methylbenzenes cannot be evaluated at this point. Even so,
[18] M. Bjørgen, U. Olsbye, S. Kolboe, J. Catal. 215 (2003) 30.
for a given catalyst topology or for a given set of conditions, we [19] M. Bjørgen, F. Bonino, S. Kolboe, K.-P. Lillerud, A. Zecchina, S. Bordiga,
may imagine that methanol can be converted solely according J. Am. Chem. Soc. 125 (2003) 15863.
to cycle II. [20] M. Bjørgen, U. Olsbye, S. Svelle, S. Kolboe, Catal. Lett. 93 (2004) 37.
[21] M. Bjørgen, U. Olsbye, D. Petersen, S. Kolboe, J. Catal. 221 (2004) 1.
[22] S. Svelle, M. Bjørgen, S. Kolboe, D. Kuck, M. Letzel, U. Olsbye, O.
4. Conclusion
Sekiguchi, E. Uggerud, Catal. Lett. 109 (2006) 25.
[23] B. Arstad, J.B. Nicholas, J.F. Haw, J. Am. Chem. Soc. 126 (2004) 2991.
This study has demonstrated that the following occur during [24] U. Olsbye, M. Bjørgen, S. Svelle, K.-P. Lillerud, S. Kolboe, Catal. To-
the conversion of methanol to hydrocarbons over H-ZSM-5: day 106 (2005) 108.
 M. Bjørgen et al. / Journal of Catalysis 249 (2007) 195–207 207

[25] S. Svelle, F. Joensen, J. Nerlov, U. Olsbye, K.-P. Lillerud, S. Kolboe, M. [32] H. Knözinger, S. Huber, J. Chem. Soc. Faraday Trans. 94 (1998) 2047.
Bjørgen, J. Am. Chem. Soc. 128 (2006) 14770. [33] M.A. Makarova, A.F. Ojo, K. Karim, M. Hunger, J. Dwyer, J. Phys.
[26] G. Spoto, E.N. Gribov, G. Ricchiardi, A. Damin, D. Scarano, S. Bordiga, Chem. 98 (1994) 3619.
C. Lamberti, A. Zecchina, Prog. Surf. Sci. 76 (2004) 71. [34] T. Behrsing, H. Jaeger, J.V. Sanders, Appl. Catal. 54 (1988) 289.
[27] S. Svelle, U. Olsbye, K.-P. Lillerud, S. Kolboe, M. Bjørgen, J. Am. Chem. [35] P. Gallezot, C. Leclercq, M. Guisnet, P. Magnoux, J. Catal. 114 (1987)
Soc. 128 (2006) 5618. 100.
[28] P. Magnoux, P. Roger, C. Canaff, V. Fouche, N.S. Gnep, M. Guisnet, Stud. [36] W. Song, H. Fu, J.F. Haw, J. Am. Chem. Soc. 123 (2001) 4749.
Surf. Sci. Catal. 34 (1987) 317. [37] S. Svelle, P.O. Rønning, U. Olsbye, S. Kolboe, J. Catal. 234 (2005) 385.
[29] A. Zecchina, S. Bordiga, G. Spoto, Phys. Chem. Chem. Phys. 7 (2005) [38] S. Svelle, P.O. Rønning, S. Kolboe, J. Catal. 224 (2004) 115.
1627. [39] S. Svelle, B. Arstad, S. Kolboe, O. Swang, J. Phys. Chem. B 107 (2003)
[30] M.A. Makarova, V.L. Zholobenko, K.M. Al-Ghefaili, N.E. Thompson, J. 9281.
Dewing, J. Dwyer, J. Chem. Soc. Faraday Trans. 90 (1994) 1047. [40] S. Svelle, S. Kolboe, O. Swang, J. Phys. Chem. B 108 (2004) 2953.
[31] J.A. Lercher, C. Gründling, G. Eder-Mirth, Catal. Today 27 (1996) [41] R.M. Dessau, J. Catal. 99 (1986) 111.
353. [42] Y.V. Kissin, Catal. Rev. 43 (2001) 85.