On reaction pathways in the conversion of methanol to hydrocarbons on HZSM-5

Article abstract of DOI:10.1016/j.jcat.2014.06.017

The underlying mechanisms of the two distinct catalytic cycles operating during conversion of methanol to olefins (MTO) on HZSM-5 have been elucidated under industrially relevant conditions. The co-existence of olefins and aromatic molecules in the zeolite pores leads to competition between the two cycles. Therefore, their importance depends on the local chemical potential of specific carbon species and the methanol conversion. Due to a faster, "autocatalytic" reaction pathway in the olefin based cycle, olefin homologation/cracking is dominant under MTO conditions, irrespective of whether aromatic molecules or olefins are co-fed with methanol. Another hydrogen transfer pathway, faster than the usual route, has been identified, which is directly linked to methanol. In agreement with that, the co-feeding of olefins resulted in a remarkable longer lifetime of the catalyst under MTO conditions, because the high rate methylation competes with the formation of more deactivating coke - presumably oxygenates- through methanol derivatives.

Full text of DOI:10.1016/j.jcat.2014.06.017

Journal of Catalysis 317 (2014) 185–197
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
On reaction pathways in the conversion of methanol to hydrocarbons
on HZSM-5
Xianyong Sun, Sebastian Mueller, Yue Liu, Hui Shi, Gary L. Haller, Maricruz Sanchez-Sanchez,
1
⇑
Andre C. van Veen , Johannes A. Lercher
Department of Chemistry and Catalysis Research Center, Technische Universität München, Lichtenbergstrasse 4, D-85747 Garching, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
The underlying mechanisms of the two distinct catalytic cycles operating during conversion of methanol
to olefins (MTO) on HZSM-5 have been elucidated under industrially relevant conditions. The co-exis-
tence of olefins and aromatic molecules in the zeolite pores leads to competition between the two cycles.
Therefore, their importance depends on the local chemical potential of specific carbon species and the
methanol conversion. Due to a faster, ‘‘autocatalytic’’ reaction pathway in the olefin based cycle, olefin
homologation/cracking is dominant under MTO conditions, irrespective of whether aromatic molecules
or olefins are co-fed with methanol. Another hydrogen transfer pathway, faster than the usual route,
has been identified, which is directly linked to methanol. In agreement with that, the co-feeding of olefins
resulted in a remarkable longer lifetime of the catalyst under MTO conditions, because the high rate
methylation competes with the formation of more deactivating coke – presumably oxygenates- through
methanol derivatives.
Received 16 April 2014
Revised 13 June 2014
Accepted 14 June 2014
Keywords:
Methanol to olefins
ZSM-5
Hydrocarbon pool
Mechanism
Ó 2014 Elsevier Inc. All rights reserved.
1
. Introduction
The catalytic conversion of methanol to olefins (MTO) has drawn
co-workers as the main reaction pathway on HZSM-5 at steady-
state conditions [12,13]. After the initial olefins are formed during
the induction period, these olefins are consecutively methylated to
form higher olefin homologues, which in turn crack into lighter
olefins such as ethene and propene. Hydrogen transfer and cycliza-
tion reactions lead to the formation of alkanes and aromatics as
end products. In parallel with the proposal of this olefin based
cycle, the important role of aromatics and unsaturated cyclic spe-
cies in the methanol reaction has also been proposed. Langner et al.
for example reported that co-feeding 36 ppm of cyclohexanol sig-
nificantly reduced the duration of the induction period [14]. In a
parallel study, experiments of co-feeding toluene or p-xylene with
methanol led Mole et al. to postulate a co-catalytic effect of meth-
ylbenzenes on methanol conversion [15,16]. As the conversion of
methanol produces aromatic molecules, one could also identify
the aromatics based reaction routes as another form of autocataly-
sis. These early investigations were very insightful and embodied
already the concept of aromatics based cycle. However, these find-
ings remained largely ignored.
particular attention in recent years, as this process is believed to
provide an alternative pathway for the production of ethene and
propene [1–4]. One of the key issues of the MTO chemistry, the con-
trol of product selectivity, necessitates a fundamental understand-
ing of the reaction mechanism. The extremely complex reaction
network makes this, however, a very challenging task [5–7].
The initial debate was mainly focused on how the first CꢀC
bond was formed. Over 20 direct coupling mechanisms were sug-
gested in spite of little experimental evidence [2,8,9]. Recent
experimental and theoretical investigations suggested, however,
that direct CꢀC coupling is not the dominating pathway due to
unstable intermediates and prohibitively high activation barriers
[
8–10]. Several mechanisms involving impurities in the feedstock
appear to offer more plausible reaction routes for the formation
of initial hydrocarbons.
In 1979, Chen and Reagan originally proposed that MTH is an
autocatalytic reaction [11]. Consistent with this proposal, an olefin
homologation/cracking route was suggested by Dessau and
It was the invention of SAPO-34 material in the 80s that stimu-
lated again the investigations in the early 1990s of the role of aro-
matics in the MTH reaction and subsequently the proposal of the
‘
‘hydrocarbon pool’’ concept by Kolboe et al. [17–19]. In the origi-
⇑
Corresponding author. Fax: +49 8928913544.
E-mail address: johannes.lercher@ch.tum.de (J.A. Lercher).
Present address: The University of Warwick, School of Engineering, Library Road,
nal proposal, the active center was a ‘‘coke-like’’ organic species
adsorbed on the surface [17–19]. The active site for MTO was later
defined as a supramolecular inorganic–organic hybrid (zeolite-
1
Coventry CV4 7AL, United Kingdom.
http://dx.doi.org/10.1016/j.jcat.2014.06.017
0
021-9517/Ó 2014 Elsevier Inc. All rights reserved.

1
86
X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
hydrocarbon species), which acts as a scaffold for light olefin for-
mation [20,7]. The unique structure of SAPO-34 material provided
the possibility of trapping carbon intermediate species and, thus,
stimulated intensive investigations in parallel by groups of Kolboe
and Haw, respectively. This established the role of aromatics, espe-
cially polymethylbenzenes and their protonated forms, as the
active hydrocarbon pool species in SAPO-34, H-BEA, and H-MOR
catalysts have large pores or cages [20,7,21–30]. Recent experi-
mental and theoretical work proposed that olefins may act as
another kind of active hydrocarbon pool species, particularly in
medium-pore zeolites, such as the ZSM-22 zeolite with 1-D 10-ring
channels, in which the internal spaces are too small to form polym-
ethylbenzenes [31–34].
In retrospect, the MTO history clearly demonstrated that the
actual course of the mechanistic understanding developed in loops,
and the key mechanistic aspects that are generally accepted today
were reported already in the very early literature [5]. Both, the ole-
fin-based cycle and the aromatic-based route are well accepted at
present by researchers favoring either the early ‘‘autocatalysis’’
proposal or the later ‘‘hydrocarbon pool’’ concept [5,6]. However,
further insights into the dynamic course of the interactions of zeo-
lite (acid sites), hydrocarbon species, and methanol are yet to be
developed, some of which we try to address in the present work.
Although a general rationale of the zeolite-specific product dis-
tribution has been achieved through understanding the kinetic
consequences of the zeolite topology and the identity of the active
hydrocarbon species, for zeolites such as H-BEA, SAPO-34 or
HZSM-5, where probably both catalytic cycles work, a quantitative
relationship between these cycles and the product distribution has
not been unequivocally established [5,6]. For instance, although
several reports have shown that aromatics, especially higher
polymethylbenzenes, are active hydrocarbon pool species in
H-BEA zeolite at 623 K [29,30], recent investigations by Ahn et al.
and Simonetti et al. [35,36] demonstrated that, over H-BEA, the
olefin based cycle can be selectively favored over the aromatics
based cycle, and the carbenium-ion chemistry dictates the forma-
2. Experimental
The employed catalyst and other reagents are identical with the
materials used in the previous paper [43]. The zeolite powder has a
Si/Al ratio of 90 and crystal size of 500 nm. The catalytic tests were
performed on either a bench-scale plug flow reaction unit (with
nitrogen dilution) or 10-fold parallel reaction unit (using water
as diluent). The pressed, crushed, and then fractionized zeolite pel-
lets are diluted, loaded and processed with the identical proce-
dures as reported in Ref. [43].
In the experiment addressing the MTO reaction cycles, a fixed
catalyst amount of 20 mg was loaded into the reactor. The metha-
nol partial pressure in the flow is held constant at 10 kPa by main-
taining the temperature of methanol saturator at 299 K. The total
flow rate was systematically changed to achieve different space
velocities. After the reaction temperature was stabilized under
50 ml/min N flow at 723 K for 1 h, the N flow was passed through
2
2
the methanol saturator to achieve 10 kPa methanol. After 5 min on
stream, the GC was started to measure the reactor effluent compo-
sition, and subsequently the valve for the feed control was
switched to pure N flow and the N flow rate was set to the target
value. This procedure was repeated for a series of space velocities.
The experiments for the aromatics co-feeding were conducted
as previously described [43]. The methanol partial pressure was
10 kPa, while the para-xylene partial pressure was 0.2 kPa, both
2
2
diluted by N . The total flow rate was maintained, as the catalyst
2
charged was changed to reach different levels of conversions. In
the experiments for the conversion of pure olefins (120 C%), i.e.,
1-pentene, 1-hexene or 1-heptene, experiments were performed
under reaction conditions as close as possible to those applied in
the MTO reaction of a feed containing 10 kPa methanol (100 C%)
and 0.4 kPa 1-pentene (20 C%). 10 kPa water vapor was introduced
with the olefin, to mimic the water concentration formed during
the MTO reaction (i.e., the outlet partial pressure of water at
100% conversion of 10 kPa methanol). Identical to the experiments
of toluene co-feeding, the total flow rate was unchanged, while the
catalyst weight was systematically changed to reach different lev-
els of conversions.
In the experiment addressing the impact of the carbon ratio of
methanol to co-feed (butene or 0-pentene), experiments were per-
formed in the 10-fold parallel reaction unit at 748 K with water
dilution. 2-butanol or 2-pentanol were used as co-feed, as they
were expected to be fully dehydrated on acidic zeolite to butenes
and pentenes, respectively. Defining the carbon based concentra-
tion in a mixture of methanol and water (weight ratio 1:2) as
100%, different compositions of the same carbon based concentra-
tion were used: (a) 100 C% from methanol, methanol partial pres-
sure 0.22 kPa, (b) 90 C% from methanol with 10 C% from co-feed (2-
butanol or 2-pentanol), (c) 83 C% from methanol with 17 C% from
co-feed (2-butanol or 2-pentanol), (d) 73 C% from methanol with
27 C% from co-feed (2-butanol or 2-pentanol), (e) 63 C% from
methanol with 37 C% from co-feed (2-butanol or 2-pentanol), (f)
37 C% from methanol with 63 C% from co-feed 2-butanol. Thus,
identical carbon concentrations were used in all experiments with
alcohol co-feeds.
4 7
tion of a product pool rich in highly branched C and C alkanes by
using low temperatures (473 K) and moderate DME partial pres-
sures (>50 kPa). Therefore, the reaction conditions, in addition to
zeolite topology, play a remarkable role in evolvement of the active
hydrocarbon species and the product selectivity.
The same situation applies to HZSM-5. The archetypical catalyst
for methanol conversion to gasoline – and in recent years light ole-
fins – has attracted tremendous efforts to elucidate the mechanism
(
e.g. [1–2,5–6]). A more recent study of the group of Olsbye sum-
marized the proposals for a dual-cycle mechanism on HZSM-5
5]. An aromatics-based cycle involved ethene and methylbenz-
enes, and the olefin-based methylation/cracking cycle produces
3+ olefins [37,38]. This has been a seminal contribution to the
[
C
interpretation of HZSM-5 specific product distributions. However,
compared to the typically chosen temperatures (6623 K) for these
mechanistic studies, higher reaction temperatures (P723 K) are
used in the case of HZSM-5 based industrial processes [5], such
as the Air Liquide’s MTP process, for which a recycling operation
of the aliphatic products other than propene is incorporated [39–
4
2]. As a result, tailoring product distributions requires insight into
the reaction mechanism under realistic reaction conditions and
specific operation modes.
3. Results and discussions
Thus, we report here the elucidation of the kinetic aspects of the
mechanism under reaction conditions closely related to the practi-
cal operations, i.e., the intrinsic selectivities toward ethene and
propene formation of the aromatics- or olefins-based cycles, and
how the dominant reaction pathways are influenced by feed com-
position, how they change during the reaction course, and, ulti-
mately, how each cycle contributes to methanol conversion and
the specific product distributions.
3.1. Autocatalysis versus hydrocarbon pool proposal
Fig. 1 depicts the effect of repeated variations of the methanol
contact time on the catalytic performance of a fixed catalyst load-
ing. When a methanol conversion of 93% was reached at a contact
ꢀ1
time of 0.11 min kgcat molMeOH, the subsequent increase of flow
ꢀ1
rate led to a decrease of contact time to 0.03 min kgcat molMeOH
,

X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
187
x
y
r ¼ ꢀk½Mꢁ ½Pꢁ
ð1Þ
where the [M] and [P] represent the concentrations of methanol and
hydrocarbon products, respectively. Although the calculated reac-
tion orders with respect to methanol deviate from zero and they
are usually slightly below one at practical MTO reaction tempera-
tures, it is well accepted that methanol adsorbs on the Brønsted acid
sites of zeolite more strongly than light hydrocarbons. Therefore,
the MTO mechanism can be seen as an analogue of the methylation
mechanism depicted in Scheme 1. The zeolite surface is predomi-
nantly covered by methanol, and the hydrocarbon species has a sig-
nificantly weaker interaction with zeolite than methanol. The
Brønsted acid sites activate methanol molecules and are responsible
for the final step of cracking of the highest hydrocarbon intermedi-
ates, acting as one type of catalytic center. On the other hand, the
entrained lower hydrocarbons accommodate the ‘‘activated’’ CH
2
entities, and the further cracking of the highest homologues pro-
duces the light olefins as products. The original lower hydrocarbon
is regenerated in the meantime. In this way, the lower hydrocarbon,
which is the starting point of a catalytic cycle, can be defined as
another type of co-catalyst that offers a faster reaction route.
Therefore, the dynamic course of the interactions of zeolite
Fig. 1. Methanol conversion in repeated cycles by varying contact time at 723 K and
methanol partial pressure of 10 kPa. Orange trace, data obtained varying the weight
of fresh catalyst for a constant flow rate. Blue trace, data for a fixed catalyst charge
at varying flow rates.
(
acid sites), hydrocarbon species, and methanol, demonstrated by
the results obtained in the repeated cycles (Fig. 1), are better
described by an autocatalysis mechanism, than by the original
hydrocarbon pool proposal as shown in Scheme 2. Although both
autocatalysis and hydrocarbon pool proposals rely on the critical
roles of acid site and lower hydrocarbons (olefins or aromatics),
the main distinction lies in the roles that the acid site and the
hydrocarbons play. In the autocatalytic route, the acid site is cov-
ered predominantly by methanol and attacked by mobile-phase
hydrocarbon; while in the strictest understanding of a hydrocar-
bon pool concept, the acid site adsorbs certain hydrocarbon spe-
cies, constituting the reaction centers and then a surface pool of
and the methanol conversion dropped to nearly zero. The following
decrease of the flow rate brought the contact time back to
ꢀ1
0
9
.11 min kgcat molMeOH and restored the methanol conversion to
3%. These repeated variations of flow rate led to fully reversible
conversion cycles as shown in blue (Fig. 1) in accordance with
the results of Olsbye et al. [5]. In addition, the results obtained with
a fixed catalyst charge and varied flow rates overlapped with those
obtained with a fixed flow rate and fresh catalyst of varied
amounts.
This complete reversibility contradicts the strictest formulation
of a hydrocarbon pool mechanism. According to this concept [17–
‘
2
‘–CH ’’ is formed via contributions by methanol, which split off
1
9], it was believed that a surface carbon pool, which was sup-
gas phase light olefins as products.
posed to be formed on the zeolite surface and to act as reservoir
of centers when a high methanol conversion was reached, should
have a longer residence time than the formed products [52]. If this
were the case, the switch of contact time from 0.08 to
Based on the above analysis of the description of the interac-
tions between zeolite, hydrocarbons and methanol, we are now
in a position to unify the autocatalysis and the hydrocarbon pool
concepts. A generalized hydrocarbon pool mechanism should
define the entrained hydrocarbons in the zeolite pores, other than
the surface carbon species, as the working hydrocarbon pool spe-
cies. Following this definition, hydrocarbon pool species, which
are free of diffusional constraints, act also as entrained co-catalysts
in the autocatalysis concept. Thus, a generalized hydrocarbon pool
mechanism is at large in accordance with the autocatalysis con-
cept, as long as one assumes a rapid exchange between the species
adsorbed in the fluid phase of the zeolite.
ꢀ
1
0
.03 min kgcat molMeOH should have generated an appreciable
methanol conversion, one higher than the conversion obtained
in the first experiment with contact time of 0.03 min
kgcat molMeOH. This was clearly not observed. The conversion upon
a
ꢀ1
ꢀ1
establishing a residence time of 0.03 min kgcat molMeOH was nearly
zero and the activity of a ‘‘pre-activated’’ catalyst was identical to
that with a fresh catalyst (Fig. 1). The fully reversible conversion
cycles indicate, therefore, that the carbon species are unstable,
decomposing or desorbing rapidly as the flow rate is increased, and
being formed rapidly after switchingback toa higher residence time.
The indirect catalytic MTO cycle starts with the consecutive
steps of methylation, the kinetics of which has been well described
3.2. Insights from aromatics and olefin co-feeding
[
44–50]. The widely accepted mechanism for methylation follows
Based on the analysis in the previous discussion of the MTO
mechanism on HZSM-5, methanol conversion proceeds via consec-
utive methylation reactions, in which the hydrocarbon species
interact with the pre-adsorbed methanol. Therefore, competitive
methylation of entrained aromatics and olefins in the zeolite pores
will influence the locally prevalent hydrocarbons, which eventu-
ally propagate each specific cycle. Thus, the reacting molecules,
except for methanol or dimethyl ether, need to remain, at large,
in the mobile phase.
An increase in the abundance of aromatics in the feed leads to
the selective propagation of the aromatics based cycle, especially
at low methanol conversions. Therefore, compared to the conver-
sion of the feed of pure methanol, propagation of the aromatics
based cycle shifts the product distribution toward ethene and
a mechanism, as depicted in Scheme 1, where a hydrocarbon mol-
ecule in the mobile phase (either olefin or aromatics) interacts
with a chemisorbed methanol, i.e., methoxonium or methyl carbe-
nium ion, and a methylated hydrocarbon is formed and desorbed
[
6]. The formed higher hydrocarbon molecules may undergo fur-
ther reactions with other chemisorbed methanol molecules until
the higher homologue has reached the size for which the rates of
cracking on the acid site are higher than the rates of methylation.
The reaction rate of methylation is reported to have a zero order
and first order dependence on the partial pressures of methanol
and hydrocarbons (olefin or aromatics), respectively [44–50]. This
is in line with the general description for the rate of an autocata-
lytic reaction where the methanol conversion follows a character-
istic S-shape as a function of contact time, as shown in
methylated aromatics. As meta-xylene has
a larger kinetic

1
88
X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
Scheme 1. Generally accepted mechanisms for methylation of aromatics or olefins, in which a gas-phase olefin or aromatic molecule reacts with a meth-oxonium ion (a) or
methyl carbenium ion (b).
methanol and 4 mol.% para-xylene. At the lowest methanol conver-
sion in this study, a carbon-based ethene selectivity of 39% and a
propene selectivity of 36% were observed. Considering that ethene
is very unreactive in further secondary reactions such as methyla-
tion or dimerization, and a small portion of propene is possibly
methylated to higher olefins even at this low conversion, it is con-
cluded that the aromatics based cycle, which has been found to be
responsible for a dominant fraction of methanol conversion at low
conversions, produces simultaneously ethene and propene with
almost equal carbon-based selectivities.
Scheme 2. Originally proposed hydrocarbon pool mechanism by Dahl and Kolboe
17–19].
This is further supported by the experiments adding aromatics
at relatively low reaction temperatures. Fig. 5(a) and (b) depicts
the aliphatic product distribution as a function of methanol conver-
sion at 623 and 673 K, respectively, for a feed containing 4 mol.%
para-xylene. It clearly shows that the extrapolated carbon based
selectivities to ethene and propene are nearly equal at zero metha-
nol conversion. This is in line with an observation by Bjørgen et al.
in which a HZSM-5 zeolite (Si/Al = 45) and a reaction temperature
of 623 K were explored [51]. However, we note that the observation
may be, in part, a result of transport effects. In an early report [52],
[
diameter than the HZSM-5 zeolite pore, it cannot effectively diffuse
into the channel to propagate the aromatics based cycle, which
leads to a significantly smaller impact of such molecule in compar-
ison with co-feeding para-xylene.
Further insight into the aromatics based cycle can be deduced
from the experiments by co-feeding of toluene and benzene. As
shown in Ref. [43], co-feeding toluene of the same molar concen-
tration as para-xylene leads to an identical impact on the propaga-
tion of the aromatics based route. In other words, as shown in
Fig. 2, toluene and para-xylene are both involved in one cycle as
active intermediates and they play an identical role. This is due
to the comparable diffusion coefficients and methylation rates for
toluene and para-xylene. On the other hand, co-feeding benzene
also leads to an identical degree of propagation of the aromatics
based cycle, but it results in stoichiometrically more methylation
of aromatics than that of toluene. As shown in Fig. 3, co-feeding
2 2 3
Haag et al. stated that ‘‘With ZSM-5 catalyst of SiO /Al O = 35–70,
temperatures of 550–750 K and conversion as low as 0.05%, the
molar ratio of propylene to ethylene was 1 or greater,’’ in agreement
with our observation. However, in the same report using HZSM-5
with a SiO /Al O = 1600 (which will have less transport effect on
2 2 3
sequential reactions), the authors measured a propene to ethene
ratio of <0.5 at a conversion <0.1 at 723 K and 1 bar. It is noted in
the present study (Figs. 4a and 5) that the extrapolated selectivity
=
to C
4
was less than 10% at zero methanol conversion. While report-
=
2
mol.% benzene led to 2% more ring methylation than the case
4
edly it is feasible for C to be formed by the aromatics based cycle as
of toluene co-feeding. This suggests that benzene itself is not an
intermediate of the aromatics based cycle; however, the compara-
ble rate coefficient for benzene methylation rapidly transformed
benzene to toluene, which is actively involved in the cycle, as
shown in Fig. 2.
a minor product compared to ethene and propene, one cannot
exclude the possibility that the olefin based cycle operates at extre-
mely low conversion as a result of transport effects, contributing to
=
the formation of a part of propene and C
4
. Considering the low car-
=
4
bon based amount of C formed, we tentatively propose that the
Fig. 4(a) depicts the carbon-based product selectivities as a
function of methanol/dimethylether conversion for a feed of
aromatics based cycle produces ethene and propene with equal car-
bon-based selectivities over HZSM-5 zeolites.

X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
189
Fig. 2. A modified dual-cycle mechanism in operation during methanol-to-hydrocarbons catalysis over HZSM-5 under industrially relevant MTO conditions.
Although the olefin based cycle plays a minor role at low meth-
anol conversions when aromatics are co-fed (as a result of the low
concentration of olefins), the significance of the olefin based cycle
increases with increasing methanol conversion. The explanation is
that the aromatics based cycle produces ethene and propene in
equal amounts. Propene is able to compete with aromatics and
propagate the olefin based cycle at higher concentrations, resulting
in the formation of most of the C3+ in the final product pool, as
shown schematically in Fig. 2. For instance, in the experiment with
a feed containing 4 mol.% p-xylene in which the aromatics based
cycle is largely dominating, a maximum of 13 C% ethene is formed
at nearly 100% methanol conversion ([43], Fig. 1(b)). If we make
the extreme assumption that all ethene was formed by the aromat-
ics-based cycle (note that the aromatics based cycle produces eth-
ene and propene with equal carbon based rates), 13% methanol is
then supposed to be converted concurrently to propene via the
aromatics based route. Thus, less than 30% methanol is converted
via the aromatics based cycle, and the remaining 70% methanol
via the olefin methylation/cracking route. Therefore, although the
aromatics based route dominates the initial conversions with a
feed abundant in aromatics, the olefin based cycle is dominant dur-
ing the overall conversion.
Fig. 3. Carbon distribution in the aromatic products by co-feeding methanol and
benzene, toluene or para-xylene, in terms of carbon in the aromatic ring and carbon
in the side chain. Methanol partial pressure was 10 kPa, aromatic co-feed partial
pressure 0.2 kPa, reaction temperature 723 K. Durene, 1,2,4,5 tetramethylbenzene
is illustrative of alkylated benzene products.
This is even more evident for the conversion of methanol in the
absence of any co-feed. At very low conversion (below 0.1%), the
Fig. 4. Carbon-based C2+ aliphatic product distributions as a function of methanol/dimethylether conversion. Reaction temperature was 723 K, methanol partial pressure
0 kPa, para-xylene co-feed partial pressure (a) 0.4 or (b) 0 kPa.
1

1
90
X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
Fig. 5. Aliphatic product distributions versus methanol conversion with a feed of mixed methanol and toluene at reaction temperatures of 623 (a) and 673 K (b). Methanol
partial pressure was 10 kPa, p-xylene co-feed partial pressure 0.4 kPa.
ratio of carbon selectivity of ethene to propene is close to one [51].
Therefore, the methanol conversion is proposed to be initiated by
the aromatics based cycle. However, olefins formed will rapidly
compete with the initial aromatic hydrocarbon species and the ole-
fin based cycle will dominate the methanol conversion thereafter,
as shown in Fig. 4(b).
Inspired by the finding that co-feeding identical concentrations
of benzene, toluene or para-xylene led to the identical impact on
the aliphatic product distribution, we designed similar experiments
6 7
MTO reaction conditions is dictated to a large extent by C –C ole-
fin cracking. Therefore, to gain mechanistic insight into olefin
transformations within the complex chemistry involved, the con-
version of pure olefins, i.e., 1-hexene and 1-heptene, was studied
at reaction conditions as close as possible to those industrially
practiced in the MTO reaction. As water is formed during the
MTO reaction, 10 kPa water vapor (i.e., the outlet partial pressure
of water at 100% conversion of 10 kPa methanol) was introduced
with the olefin.
on the olefins based cycle by co-feeding various C
2
–C
6
olefins of the
Fig. 6 illustrates the conversion of 1-hexene and 1-heptene of
the same concentration of 120 C% (equivalent to the 120% carbon
form a feed containing 100 C% of methanol and 20 C% from olefin
co-feed, but no actual methanol in the feed), respectively, as a
function of contact time. Different from the previous sections,
the contact time in this section is defined as the ratio of
carbon molar flow rate to catalyst weight, with units of
same carbon-based concentration. Indeed, very similar to the situ-
ation in the aromatics based cycle, it was demonstrated that co-
feeding a lower concentration of olefins (up to 40 C% in this study)
leads to product distributions independent of the identity of the co-
fed C3+ olefins at higher conversions [43]. As shown schematically
in Fig. 2, similar to the roles that toluene and xylene play in the aro-
ꢀ
1
matics based cycle, C
–C
3 6
olefins take part in the olefin methyla-
min kgcat moltotal carbon.
tion/cracking cycle as active intermediates via rapid scrambling
and incorporation of co-fed olefins by methylation and cracking
with comparable rate constants. Adding ethene leads to the only
exception. Contrary to benzene in the aromatics based cycle, ethene
is not effectively involved in the olefin methylation/cracking and
remains primarily unreacted, because it is at least one order of mag-
nitude less active in methylation than C3+ olefins (Fig. 2).
The observation that the final product distribution is indepen-
dent of the nature of the co-fed olefins, strongly suggests that
the olefin based cycle is the dominant reaction pathway on
HZSM-5 under the studied reaction conditions (i.e., at higher con-
versions of methanol).
On the other hand, addition of olefins in the feed favors hydro-
gen transfer and aromatization. Even though the aromatics con-
centration is low compared to the olefins, aromatics have a
relatively lower diffusion rate than the olefins. Therefore, these
aromatics tend to stay in the pores and compete with olefins. In
turn, the aromatics based hydrocarbon pool route remains active.
The overall consequence is, therefore, that co-feeding olefins does
not selectively propagate the olefin-based cycle. In other words,
the ratio of activities of aromatics and olefins based cycles does
not significantly change compared to that in the case of a pure
methanol feed.
The primary reactions of 1-hexene cracking proceed via two
pathways, forming ethene and butene (1:1) as well as two mole-
cules of propene, respectively (Scheme 3). As shown in Fig. 6, the
former primary reaction is much slower than the latter. The ratio
of the reaction rates for these two primary reactions was estimated
to be 6, based on the selectivities to propene and ethene at the low-
est contact time studied (Table 1). Fast secondary reactions of pro-
pene, which was the main product, with 1-hexene, which was
abundant at low conversions, led to butenes and pentenes, result-
ing in the deviation of butenes and ethene from unity, even at the
lowest contact time studied. In the case of 1-heptene, the predom-
inant cracking route leads to propene and butenes, essentially in
accord with the stoichiometry (Fig. 6 and Table 1). Other products
were formed with very low selectivities even at 1-heptene conver-
sion of 42%. The much faster primary cracking of 1-heptene than
that of 1-hexene might be the reason as to why propene and
butenes were observed in parity at 1-heptene conversion of 42%,
while the ethene/butenes ratio deviated from unity when 1-hex-
ene conversion was still 10–20%.
At longer contact times, the primary olefin products undergo
reactions with the reactant and among themselves, resulting in
an increased formation of olefins, which would not be expected
from monomolecular pathways mediated by carbenium ions, e.g.,
pentenes from cracking of 1-hexene. Hydrogen transfer reactions
between olefins become detectable with increasing conversion,
but the extent remains very low.
3.3. Comparison with olefin cracking
Three groups of products, i.e., propene, ethene and hydrogen
transfer products (C2–4 light alkanes plus aromatics), deserve
special attention in the context of comparing the catalytic perfor-
mance of HZSM-5 in olefin cracking (pure olefin feeds) and the
At this point we have shown that two distinct catalytic cycles
operate over HZSM-5, with the olefin methylation/cracking route
being the dominant reaction pathway in the overall methanol con-
version range, and that the final product distribution under typical

X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
191
Fig. 6. Reaction pathway for 1-hexene (a) and 1-heptene (b) cracking at 723 K. Water partial pressure 10 kPa, partial pressures for 1-hexene and 1-heptene were 2 and
1
.7 kPa, respectively. For the sake of comparison, 1-hexene and 1-heptene were expressed as 120%C. Hydrogen transfer (HT) products include aromatics and C2–4 paraffins.
process. Fig. 8 depicts the yield of these HT products, as a function
of contact time, for the methanol-olefin feed (molar ratio = 100:20)
and various pure C5–7 olefin feeds with the same total carbon con-
centration. Over the whole range of contact time, hydrogen trans-
fer was significantly higher for the methanol-containing feed
compared to pure olefin feeds. Upon reaching full conversion of
methanol, the formation rates of the HT products from the metha-
nol-containing feed and pure olefin feeds became comparable, as
indicated by the similar slopes at higher contact times. These HT
products were possibly formed via reactions between two olefin
species, which should be similar for both cases, as both methanol
and higher olefins were converted predominantly to C3–5 olefins
Scheme 3. Primary reaction routes for 1-hexene and 1-heptene cracking. Reaction
rates based on the product distribution at the lowest conversion measured.
at full reactant conversions.
Before full conversion of methanol, hydrogen transfer pro-
ceeded much faster in the methanol-olefin mixed feed than in pure
olefin feeds. Although a molecular-level understanding of the
hydride transfer routes is only beginning to emerge, the results
demonstrate clearly that a hydrogen transfer pathway exists,
which involves methanol or intermediates derived from it and
which has a higher rate than the one along the classical route
between two olefinic species.
MTO reaction (methanol-olefin mixtures). Analysis of these prod-
ucts can also provide valuable information about the olefin meth-
ylation/cracking route. At full reactant conversion, 1-hexene
cracking led to a higher concentration of propene, while 1-heptene
cracking led to a lower propene yield compared to the propene
produced from the methanol and C3–6 olefin reaction mixture
(Fig. 7).
Fig. 9 depicts the yield of ethene as a function of contact time
from four different feeds, including a methanol-olefin mixed feed
This agrees well with the olefin methylation/cracking mecha-
nism for mixed methanol-olefin feeds, in which olefin methylation
either terminates at hexenes that are cracked to propene or termi-
nates at heptenes that are cracked to propene and butenes (one
heptene molecule produces one propene molecule). The fact that
the propene yield from mixed methanol-olefin feeds was closer
to that from 1-heptene cracking suggests a higher termination
probability at heptenes for the conversion of methanol in the pres-
ence of co-fed C3–6 olefins. Note that 100% conversion of methanol
occurs at a specific residence time of about 0.12 min kgcat/moltotal
carbon, see Fig. 8, so the comparison is appropriate at or beyond this
point.
(
1
the choice of C3–6 olefin does not impact the product distribution),
-pentene, 1-hexene and 1-heptene. Before reaching >90% conver-
sion, 1-hexene cracking gave rise to the highest formation rate of
ethene among all four feeds, indicating that the kinetically primary
cracking of 1-hexene to ethene and butenes dominates under real-
istic MTP reaction conditions. Ethene formation from 1-pentene
cracking was the second highest, while 1-heptene cracking pro-
duced ethene in a much lower rate and yield over the studied range
of contact times. From the yield ratio of formed ethene, propene
and butenes, it is concluded that the bimolecular reaction pathway
(
3 n
that is, methylation/cracking of CH OH + Cn+1 to produce C +
C1–4 alkanes and aromatics that are referred to as hydrogen
ethene) plays a predominant role in the formation of ethene, which
is essentially formed by cracking of secondary products such as
transfer (HT) products in this work are undesirable in the MTO
Table 1
Conversions and product distributions at low contact times for cracking of 1-pentene or 1-hexene or 1-heptene. Reaction temperature was 723 K. Partial pressure of water vapor
was 10 kPa, and partial pressures for 1-pentene, 1-hexene or 1-heptene were 2.4, 2 and 1.7 kPa, respectively, leading to an equivalent 120 C% in each feed mixture (referred to as
the 120 C% from a feed containing 100 C% of methanol and 20 C% from the co-fed olefin). The conversion rates were given on a molecular basis.
ꢀ
1
ꢀ1
Feed
Conversion rate (mol gcat
h
)
Conversion (%)
Hydrocarbons (C%, in total 120 C%)
=
=
=
C
2
C
3
C
4
C
5
C
6
C
7+
=
1-C
1-C
1-C
7
1.05
0.17
0.02
41.9
14.5
1.6
0.15
0.76
0.22
21.15
14.20
0.55
28.0
1.93
0.43
0.71
0.51
118.15
0.25
102.59
0.65
69.74
=
6
=
5

1
92
X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
Fig. 9. Yield of ethene as a function of contact time for the feeds containing
methanol (100 C%) and 1-pentene, 1-hexene, and 1-heptene (20%C). Reaction
temperature 723 K, water partial pressure 10 kPa; partial pressures for 1-pentene
1-hexene and 1-heptene: 2.4, 2.0 and 1.7 kPa, respectively.
Fig. 7. Propene yield as a function of contact time for the feeds of 1-hexene,
-heptene, and methanol co-fed with 20 C% C3–6 olefins. Total concentration of
carbon was 120% for all feeds. Reaction temperature 723 K. For cracking of 1-hexene
and 1-heptene, water partial pressure was 10 kPa, and partial pressures for
1
1
-hexene, and 1-heptene were 2 and 1.7 kPa, respectively.
It should be emphasized that the main conclusion drawn from
Fig. 9, i.e., that hexene cracking is an important route for ethene,
is apparently against the proposal widely accepted in literature
that ethene does not form via the olefin methylation and cracking
cycle, but via the aromatics based cycle, as demonstrated by the
isotope experiments by Bjørgen et al. [37,38]. However, a detailed
comparison of the reaction conditions shows that the reaction
temperature (623 K) for most mechanistic studies [37,38] is much
lower than the typical temperature range (723–773 K) for MTO
and MTP processes, which leads to the different product dependen-
cies. It is, thus, concluded that the reaction temperature is the key
factor affecting the dominant formation pathway for ethene. The
=
olefin based route, C
6
cracking as the key step, has a significantly
higher activation energy than the aromatics based cycle. Therefore,
although ethene predominantly forms via the aromatics based
cycle at lower reaction temperatures such as 623 K, the olefin
based cycle is the most significant pathway at 723 K and above.
This conclusion about the formation route of ethene via C=
ole-
6
fin cracking is supported by the impact of reaction temperature on
the MTO product distributions, as shown in Fig. 10. With the
increase of reaction temperature, both yields of ethene and pro-
Fig. 8. Yield of hydrogen transfer products as a function of contact time for the
mixed feeds containing methanol (100 C%) and 1-pentene, 1-hexene and 1-heptene
(
20%C). Reaction temperature 723 K.
pene increased. Correspondingly the yield of C4+ aliphatics
decreases with temperature. This indicates that the higher olefin
cracking at higher reaction temperatures is the main reaction route
that leads to the favored formation of light olefins. Most impor-
tantly, the yield of hydrogen transfer products decreased with
reaction temperature, most probably due to a lower coverage of
olefinic species at higher temperature. The opposite trends of eth-
ene and aromatics demonstrated that much of ethene is mechanis-
hexene cracking [53] (Scheme 3). Compared to hexene cracking,
the ethene formation rate from the methanol-olefin feed was
lower. This implies that under realistic reaction conditions
(
methanol total conversion), hexene cracking acts as an important
route for ethene formation. Analogous to the situation for the
hydrogen transfer rates, with contact times longer than
ꢀ
1
0
.1 min kgcat molMeOH the ethene formation showed an identical
tically not linked with aromatics, but comes from the other
=
=
=
rate independently whether the feed was methanol, C
6
, or C
7
. This
pathways, i.e., C
6
olefin cracking.
again was attributed to the similar actual feed composition after
cracking, predominantly C3–4 olefins (P85 C%). Therefore, ethene
3.4. Impact of methanol to olefin co-feed ratio on product distribution
is concluded to be formed from the oligomerization of lower olefin
=
and the subsequent cracking of C
6
intermediates. The only excep-
The finding that the product distribution is independent of the
nature of the co-fed C3–6 olefin contradicts the observation of a
remarkable impact of the identity of co-fed olefins by Xiao et al.
[53]. The authors applied identical partial pressures of methanol
and co-fed C2–6 olefin, corresponding to C-based olefin-to-metha-
nol ratios of (2–7):1, which is significantly higher than those
applied in this work (at most 4:10) [53]. Similarly, it was observed
that 1-butene and 1-pentene, when co-fed with methanol in a C-
based ratio of 1:1, influenced the product distribution (Fig. 11). A
=
tion is ethene formation from
.11 min kgcat molMeOH, which has been sufficient to almost fully
C
5
.
At
a contact time of
ꢀ1
0
=
convert the other feeds (conversion P96% for C6–7 and 100% for
methanol), the overall pentene conversion was only 32%. There-
fore, the actual feed composition from 1-pentene behaves very
differently from the others and it forms increasing amounts of
ethene as the contact time is extended and further pentene
conversion is achieved.

X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
193
Fig. 10. Impact of reaction temperature on the MTO product distribution over HZSM-5. Methanol partial pressure was 22 kPa, diluted by water.
similar or even higher concentration of olefin co-feed relative to
methanol is inferred to lead to a more effective competitive
adsorption by the olefin-derived species and a greater fraction of
methanol being consumed by methylation of such olefin-derived
species. In this case, the product distribution at complete methanol
conversion is dominated by the subsequent inter-conversion of the
formed higher olefins, which reflects the impact of the nature of
co-fed olefin. In contrast, when only a small fraction of olefin is
co-fed with methanol, the catalyst surface is dominated by meth-
anol-derived species, which are, in turn, responsible for the major
part of methanol consumption. Fast olefin methylation/cracking
leads to scrambling of co-fed carbon in the products, and, thus,
the product distribution becomes independent of the co-feed
identity.
To investigate the effect of the feed composition on the
product distribution, different feeds of an identical total carbon
concentration were studied. Defining the carbon based concentra-
tion as 100% in the mixture of methanol and water with a weight
ratio 1–2, they were (a) 100 C% from methanol, methanol partial
pressure 0.22 kPa, (b) 90 C% from methanol with 10 C% from
co-feed (2-butanol or 2-pentanol), (c) 83 C% from methanol with
These results are consistent with the other findings in this work.
A decrease of the methanol concentration from 100% to 63% did not
essentially alter the surface chemistry, which was still dominated
by the presence of methanol/dimethyl ether and its derived C
1
entities, and the olefin methylation/cracking route resulted in a
similar product distribution for the nine feeds (100 C% methanol,
90 C% methanol + 10 C% C
, 73 C% methanol + 27 C% C
or C and C ). However, a further decrease of methanol concentra-
4
5
or C , 83 C% methanol + 17 C% C
4
or
C
5
4
5
or C , 63 C% methanol + 37 C% C
4
4
5
tion in the feed to 37% results in competitive adsorption of the
methanol and 1-butene species. Therefore, analogous to the case
shown in Fig. 11, a separate methylation and olefin cracking reac-
tion path will lead to a product distribution different from that
resulting from a methanol dominated hydrocarbon pool route.
With an identical carbon concentration in the feed, the feed con-
taining 100 C% methanol showed the highest yield of hydrogen
transfer products (including methane), as the methanol-involving
hydrogen transfer pathway has a higher rate than the classical
route, which solely involves olefins (Fig. 8). Although the yield of
aromatic products decreased from 1.5% to 1.0% when the concen-
tration of methanol in the feed decreased from 100% to 63%, the
yield of ethene remained very similar (6.9% compared to 7.0%). This
indicates again that at an elevated reaction temperature of 748 K,
ethene is hardly mechanistically linked with the presence of aro-
matic molecules, and the olefin cracking route becomes an impor-
tant pathway for ethene formation.
1
7 C% from co-feed (2-butanol or 2-pentanol), (d) 73 C% from
methanol with 27 C% from co-feed (2-butanol or 2-pentanol),
e) 63 C% from methanol with 37 C% from co-feed (2-butanol or
-butanol and 2-pentanol), (f) 37 C% from methanol with 63 C%
from 2-butanol, respectively, and the product distributions at
00% methanol conversion are depicted in Table 2. Less methanol
(
2
1
in the feed led to a lower concentration of methane, while identi-
cal yields of ethene were observed, irrespective of the feed
composition. A decrease of the methanol concentration in the feed
from 100 C% to 63 C% led to little change in C2+ production
distribution, and at a fixed methanol concentration, the product
distribution was independent of the nature of the co-feed. The
yield of C2–4 paraffins and aromatics decreased with decreasing
concentration of methanol. While the amount of methane was
3.5. Toward elucidating the working mechanism in MTO over HZSM-5
As outlined before, the dual cycle concept [38] depicts the key
reaction steps including olefin methylation and cracking, aromatics
methylation and dealkylation, and hydrogen transfer as a bridging
step [6]. In the context of elucidating the effect of the zeolite topol-
ogy, reaction conditions and practical operations on the final prod-
uct selectivity, more quantitative descriptions of the individual
reaction steps of the dual cycle mechanism are, while crucial, yet
to be fully developed.
2 4
about an order of magnitude lower, its correlation with C –C
paraffins and aromatics suggests that it can be viewed as a less
probable hydride transfer product. Further decrease in the
methanol concentration to 37% significantly shifted the product
spectrum from propene to butenes and C5+ aliphatics.
Two key persistent questions are for a specific cycle, (1) what
influences the intrinsic selectivities to the desired products and
(2) what accounts for the formation of and what is the nature of

1
94
X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
Fig. 11. Concentrations of products ethene (a), propene (b), butenes (c) and C5 hydrocarbons (d), as a function of contact time for feed mixtures containing 100 C% methanol
and 100 C% from 1-butene (in red) or 1-pentene (in green). Reaction temperature 723 K, methanol partial pressure 10 kPa, and partial pressures for 1-butene and 1-pentene
2
.5 and 2 kPa, respectively (equivalent to 200 C% in each feed mixture).
Table 2
Detailed product distributions as a function of feed composition. Carbon based concentration was defined as 100% in the mixture of methanol and water with a methanol partial
ꢀ
1
ꢀ1
pressure of 22 kPa. Reaction temperature 748 K, weight hourly space velocity with respect to carbon was 0.56 gcarbon
g
cat
h
.
Feed composition (%C)
Methanol 2-Butanol
00
Product distribution (%C)
=
=
=
2-Pentanol
C
1
C
2
C
3
C
4
C
5+
C
2
–C
4
paraffins
Aromatics
1
0
10
0
0
0
10
0
17
0
27
0
0.35
0.27
0.27
0.23
0.23
0.17
0.18
0.14
0.15
0.08
6.9
6.9
6.9
6.9
7.0
6.9
6.9
7.0
7.1
7.0
46.9
46.9
46.9
47.0
47.0
47.0
47.0
47.1
47.0
44.1
26.7
26.7
26.7
27.0
27.0
27.0
27.0
27.2
27.1
30.2
14.5
14.6
14.6
14.6
14.6
14.6
14.6
14.7
14.9
15.3
2.8
2.7
2.7
2.6
2.6
2.5
2.5
2.5
2.4
2.3
1.5
1.4
1.4
1.3
1.3
1.2
1.2
1.0
1.0
0.6
90
90
83
83
73
73
63
63
37
17
0
27
0
37
20
63
17
0
the prevalent active hydrocarbon species? The complex chemistry
of methanol conversion makes it challenging to answer these ques-
tions unequivocally. The key information, which has been derived
from the present study over HZSM-5 includes, however, several
points. The aromatics based cycle operates over HZSM-5 and pro-
duces ethene and propene with carbon-based intrinsic selectivities
of 1 to 1 at realistic reaction temperatures for the MTO process. The
olefin based methylation/cracking route dominates. Growth of the
catalytic cycles contributes. The turnover rate for an aromatics-
or olefin-populated site has still not been assessed. It should be
noted, however, that the olefin based route is by far the dominant
reaction pathway at high methanol conversion, and its significance
increases with methanol conversion to olefins.
Let us turn at this point to the dominant pathway of ethene and
propene formation on HZSM-5. It has been shown that the olefin
based cycle dominates in methanol conversion over HZSM-5 irre-
spective of the co-feed. Combined with the fact that this cycle is
highly selective toward C3+, it allows us to conclude that C3+ olefins
are predominantly formed from the olefin based cycle. The forma-
tion route of ethene, on the other hand, is more complex, and
depends on multiple parameters. For a feed rich in aromatics, the
aromatics based cycle is selectively promoted, and correspondingly
it acts as the dominant pathway for ethene formation. However, for
the conversion of feeds of methanol with or without olefin co-
feeds, ethene is formed via aromatics and olefin based catalytic
6 7
carbon chain terminates predominantly at C or C . The subsequent
cracking of these higher olefins via carbenium ions and b-scission
leads to a product spectrum significantly more selective to propene
and C4+ than to ethene. An estimate from C6–7 olefin cracking dem-
onstrated that the relative ratio of carbon based selectivities for
ethene to propene is 5–100.
The final product distribution, however, depends not only on
the intrinsic selectivities for a specific cycle, but also relies to a
large extent on the methanol conversion to which each of the

X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
195
cycles. In this case, the reaction conditions critically influence the
relative contributions of the two pathways. This is related to the
fact that the aromatics based cycle is comparatively more selective
in ethene formation, but it contributes significantly less to metha-
nol conversion, especially at high conversions. On the other hand,
the olefin based route is significantly less selective to ethene for-
mation, but the olefin driven autocatalysis makes it overwhelm-
ingly active in converting methanol, the selectivity to ethene
increasing with increasing reaction temperature. Therefore, the
contribution for each cycle depends on the methanol conversion
and reaction conditions. At low methanol conversions and rela-
tively low reaction temperatures, a large fraction of ethene is
formed via the aromatics based cycle. However, the olefin based
cycle contributes more significantly to the ethene formation as
the methanol conversion and reaction temperature increase.
The present results clearly identify the cracking of olefinic inter-
mediates as the dominant reaction pathway in the MTO conversion
over HZSM-5 catalysts. However, particular attention was paid to
the critical role that methanol and methanol/olefin ratio feeds play
in the reactions.
As shown in the Section 3.3, a hydrogen transfer pathway
involving methanol or surface reaction intermediates directly
derived from methanol have a much higher rate than the classical
route between two olefinic species. We tentatively propose that
this pathway is linked with intermediates of the chemisorbed
methanol (Scheme 4(b)), which can undergo a reaction analogous
to the reaction proposed for surface species derived directly from
an acid site adsorbed olefin (Scheme 4a). As methanol derived spe-
cies is believed to have a significantly higher coverage than the ole-
fin products due to a high adsorption ability, the methanol-
involved hydrogen transfer pathway would be very important at
partial methanol conversions. It is proposed that the reaction
would lead to the formation of methane and a more unsaturated
cyclic species (cyclohexadiene-like) which is in turn more active
than cyclohexene in participating in other hydrogen transfer reac-
tions. This is supported by the fact that less methanol in the feed
led to less hydrogen transfer products including methane, C2–4 par-
affins, and aromatics, as shown in Table 2.
hydrogen transfer solely occurred between olefins (see the envi-
sioned curve in Fig. 12 to guide the eyes). The reason is that the
concentration of olefins increases with increasing methanol con-
version, and, thus, hydrogen transfer between secondary products
would necessarily lead to a significant increase of C
increased methanol conversion.
4
HTI with
Inspired by the unique role that methanol plays in the hydrogen
transfer reactions, which are responsible for the formation of coke
precursors, we turn at this point to the impact of methanol on the
catalyst deactivation. Fig. 13 depicts the evolution of concentra-
tions of methanol and C2–4 olefins as a function of time on stream
for the various methanol containing feeds (100%, 90%, 83%, 73%,
63% methanol, respectively). The identical total carbon concentra-
tions (100 C%) were ensured by adding required concentrations of
n-butanol correspondingly. Independent experiments showed that
the co-fed n-butanol had an extremely high dehydration activity
under the employed test conditions; therefore n-butanol was
expected to be fully dehydrated on acidic zeolite to butenes over
the very top layer of the catalyst bed (less than 2 mg of the total
300 mg catalyst bed loading). As shown in the Section 3.4 and
Table 2, conversion of the five feeds produces identical olefin dis-
tributions at the initial time on stream. However, as shown in
Fig. 13, a substitution of 10% methanol by higher alcohol (butanol)
resulted in a doubled catalyst lifetime (lifetime was defined as the
time on stream when the methanol concentration in the reactor
effluent reached 10%). A further increase of the butanol concentra-
tion to 17 C% or even higher concentrations in the feed led to a sub-
stantially longer catalyst lifetimes. This is generally in line with the
observations that less methanol in the feed resulted in less
aromatic products. However, the slight decrease of the yield of
aromatic products (1.5% compared to 1.4%) can hardly be
associated with the doubled lifetime when 10 C% butanol was
co-fed, clearly indicating the existence of another deactivation
pathway involving methanol.
As shown in Table 2, the main distinction in the product distri-
butions is the yield of methane. 10 C% butanol in the feed led to an
appreciably lower methane formation. As illustrated in Scheme 4,
methane is proposed to form via reduction of a methyl group by
hydride transfer. Generally in line with an early proposal [55],
we propose that methanol itself acts as the hydride donor during
the initial phase when no or little olefin species exist, leading to
the formation of methane and formaldehyde, as shown in
Scheme 5a. The reactive formaldehyde molecules may undergo
Along with this proposal, it is indicated that the conversion of
methanol produces light olefins and hydrogen transfer products
in parallel reactions pathways. This is in line with a long-standing
observation in the MTH conversion [54]. The C
Index (C HTI) is often used in methanol chemistry to describe the
hydrogen transfer activity. The C
paraffin (n-butane and iso-butane) concentration to the total C
concentration. A high C
olefins to aromatics and alkanes. Fig. 12 depicts the plot of C
as a function of methanol conversion. The C HTI remains almost
4
Hydrogen Transfer
4
4
HTI is defined as the ratio of C
4
4
4
HTI indicates a high activity in converting
HTI
4
4
unchanged over a wide range of methanol conversion, indicating
a constant hydrogen transfer activity. This implies that olefin
formation and hydrogen transfers proceed in parallel rather than
in sequential reactions. This would not have been the case, if
4
Fig. 12. C Hydrogen Transfer Index (C4 HTI) as a function of methanol conversion.
Scheme 4. Proposed mechanistic pathway for the methanol enhanced hydrogen
transfer via the same interme-diate as olefin formation via methylation.
Methanol partial pressure was 22 kPa, diluted in water. Reaction temperature
748 K.

1
96
X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
from top down mainly via this type of mechanism. However, when
small concentrations of higher olefin (or alcohol) were co-fed, the
deactivation pathway described above is largely bypassed, because
the olefin based cycle would compete with a rate overwhelmingly
faster than the rate of the pathway depicted in Scheme 5a, leading
to little formation of the oxygenic coke. In the case of feed with
high content in olefins or alcohols, the main coking species is
expected to be the graphitic type (Type-2 deactivation species).
This graphitic coke is formed and deactivates the catalyst in a rel-
atively much lower rate compared to the oxygenic coke. This
accounts for the significantly longer catalyst lifetime achieved
when small amounts of a higher alcohol (olefin) are co-fed
(
Fig. 13).
To summarize, we conclude that when a pure methanol feed is
used, the aromatic based cycle starts. The first aromatic species is
proposed to be formed via oxygenic C species. This proposal
1
allows the interpretation of the observation that at extremely
low conversion, methane is the major hydrocarbon formed in line
with earlier observations of Haag et al. [52]. At still lower conver-
sion (0.05%) of a pure methanol feed, the main products were
methane (85%), and ethane and propene selectivity was close to 1.
Fig. 13. Concentrations of methanol and C2–4 olefins as a function of time on stream
for feeds (all diluted in water) of (a) 22 kPa methanol, (b) 19.8 kPa methanol and
0
.55 kPa 2-butanol, (c) 18.2 kPa methanol and 0.94 kPa 2-butanol, (d) 16 kPa
methanol and 1.48 kPa 2-butanol, (e) 13.8 kPa methanol and 2.1 kPa 2-butanol,
respectively. All feeds contained the identical carbon based concentration. Reaction
ꢀ
1
ꢀ1
temperature 748 K, weight hourly space velocity 0.56 gcarbon g h .
cat
4
. Conclusions
Formose-type reactions leading to carbon–carbon formation and
chain growth and further dehydrations (Scheme 5b). This results
in the formation of the first cyclic species, as depicted in
Scheme 5b. This hetero-atom unsaturated cyclic species aggregate
further, forming the coke (Type-1 deactivation species) which can
rapidly deactivate the acid site due to its high sticking affinity [55].
Further work is currently on-going for a detailed kinetic and phys-
iochemical analysis of the formation of coke. Such reaction path-
way might in fact be able to initiate and/or enhance the
aromatics based cycle, which in turn forms the first olefin products.
The formed olefins then compete with the hydride transfer path-
way depicted in Scheme 5a in reacting with methanol via the rapid,
The MTO reaction over HZSM-5 catalysts has been clearly iden-
tified as autocatalysis mechanism, with mobile olefins and aro-
matic products in the zeolite pore acting as competing co-
catalysts. Accordingly, two distinct reaction pathways, aromatics-
based and olefin-based, are active for the production of ethene
and propene from methanol over HZSM-5 under reaction condi-
tions relevant to practical operations. The aromatics-based cycle
starts with toluene as the lowest sufficiently active species, while
a complete olefin methylation/cracking cycle begins its turnover
with propene as the lowest sufficiently active species. The aromat-
ics-based cycle produces ethene and propene with equal carbon
based selectivities, while the olefin-based cycle favors C3+ olefins
over ethene.
‘
‘autocatalytic’’ olefin based cycle at an overwhelmingly higher
rate. On the other hand, aromatic molecules are subsequently
formed via hydrogen transfer. The adsorption of these aromatic
coke precursors and further transformation leads to the formation
of the classic ‘‘graphitic’’ coke (Type-2 deactivation species), typi-
cally observed on deactivated MTO catalysts.
Along with this proposal, it is speculated that when a 100%
methanol feed was used in the plug-flow reactor, the active cata-
lyst layer on the very top is in contact with only methanol, which
leads to the formation of coke (Type 1 deactivation species) and
subsequently a fast deactivation of this catalyst layer. With time
on stream, the whole catalyst bed deactivates and layer by layer
The co-existence of olefins and aromatics species in the zeolite
pores leads to a competition between the two cycles for chemi-
sorbed methanol. Therefore, total activity depends on the local
activities of specific hydrocarbon species and methanol conversion.
Co-processing of intermediates in each catalytic cycle of the same
concentrations results in identical involvements in the turnover
and in turn an identical impact on the product distribution, due
to the comparable rate coefficients in each step of a cycle. While
co-feeding lower substituted benzenes propagate the aromatics-
based cycle, the olefins produced by the aromatics based cycle will
subsequently propagate also the olefin based cycle. In turn, olefin
Scheme 5. Proposed mechanistic pathways for the formation of formaldehyde via hydrogen transfer and its involvement in deactivation and olefin formation.

X. Sun et al. / Journal of Catalysis 317 (2014) 185–197
197
homologation/cracking reactions are more important than the aro-
matics based cycle at higher methanol conversions, contributing to
3+ higher olefin formation irrespective of the aromatics or olefinic
[16] T. Mole, G. Bett, D. Seddon, J. Catal. 84 (1983) 435.
[
[
[
17] I.M. Dahl, S. Kolboe, Catal. Lett. 20 (1993) 329.
18] I.M. Dahl, S. Kolboe, J. Catal. 149 (1994) 458.
19] I.M. Dahl, S. Kolboe, J. Catal. 161 (1996) 304.
C
nature of co-feeds. On the other hand, the aromatics based and the
olefin based cycle operate for ethene formation, and the dominant
pathway for ethene formation depends to a large extent on the
reaction conditions. With an aromatics-enriched feed and/or at
low reaction temperatures, the aromatics based cycle contributed
predominantly, while the olefin based cycle contributes signifi-
cantly as well when a high reaction temperature is adopted (such
as in the MT(O)P process).
The results shown here also demonstrate the presence of a spe-
cific hydrogen transfer pathway involving chemisorbed methanol
intermediates, which is significantly faster than classic hydride
transfer between two olefinic species and has significant conse-
quences for catalyst lifetime under MTO process conditions.
[20] J.F. Haw, J.B. Nicholas, W. Song, F. Deng, Z. Wang, T. Xu, C.S. Heneghan, J. Am.
Chem. Soc. 122 (2000) 4763.
[
21] Ø. Mikkelsen, P.O. Rønning, S. Kolboe, Microporous Mesoporous Mater. 40
2000) 95.
[22] B. Arstad, S. Kolboe, Catal. Lett. 71 (2001) 209.
(
[
[
23] B. Arstad, S. Kolboe, J. Am. Chem. Soc. 123 (2001) 8137.
24] W. Song, J.F. Haw, J.B. Nicholas, C.S. Henghan, J. Am. Chem. Soc. 122 (2000)
10726.
[25] W. Song, H. Fu, J.F. Haw, J. Am. Chem. Soc. 123 (2001) 4749.
[
[
[
26] W. Song, H. Fu, J.F. Haw, J. Phys. Chem. B 105 (2001) 12839.
27] W. Song, H. Fu, J.F. Haw, Catal. Lett. 76 (2001) 89.
28] J.F. Haw, D.M. Marcus, Catal. Lett. 34 (2005) 41.
[29] M. Bjørgen, U. Olsbye, S. Kolboe, J. Catal. 215 (2003) 30.
[
[
30] M. Bjørgen, U. Olsbye, D. Petersen, S. Kolboe, J. Catal. 221 (2004) 1.
31] D. Lesthaeghe, J. Van der Mynsbrugge, M. Vandichel, M. Waroquier, V. Van
Speybroeck, ChemCatChem 3 (2011) 208.
[32] S. Teketel, S. Svelle, K.P. Lillerud, U. Olsbye, ChemCatChem 1 (2009) 78.
[
33] S. Teketel, U. Olsbye, K.P. Lillerud, P. Beato, S. Svelle, Microporous Mesoporous
Mater. 136 (2010) 33.
Acknowledgments
[
34] J. Li, Y. Wei, Y. Qi, P. Tian, B. Li, Y. He, F. Chang, X. Sun, Z. Liu, Catal. Today 164
(
2011) 288.
The authors acknowledge the financial support from Clariant
Produkte (Deutschland) GmbH and fruitful discussions within the
framework of MuniCat. X.S. is thankful to Anmin Zheng for helpful
discussions.
[35] J.H. Ahn, B. Temel, E. Iglesia, Angew. Chem., Int. Ed. 121 (2009) 3872.
[
[
36] D.A. Simonetti, J.H. Ahn, E. Iglesia, J. Catal. 277 (2011) 173.
37] S. Svelle, F. Joensen, J. Nerlov, U. Olsbye, K.P. Lillerud, S. Kolboe, M. Bjørgen, J.
Am. Chem. Soc. 126 (2006) 14770.
[38] M. Bjørgen, S. Svelle, F. Joensen, J. Nerlov, S. Kolboe, F. Bonino, L. Palumbo, S.
Bordiga, U. Olsbye, J. Catal. 248 (2007) 195.
[
[
[
39] H. Koempel, W. Liebner, Stud. Surf. Sci. Catal. 167 (2007) 261.
40] H. Bach, L. Brehm, S. Jensen, EP 2004/018089 A1, 2004.
41] G. Birke, H. Koempel, W. Liebner, H. Bach, EP2006048184, 2006.
References
[
[
[
[
[
1] C.D. Chang, Catal. Rev. Sci. Eng. 25 (1983) 1.
[42] M. Rothaemel, U. Finck, T. Renner, EP 2006/136433 A1, 2006.
[43] X. Sun, S. Mueller, H. Shi, G.L. Haller, M. Sanchez-Sanchez, A.C. van Veen, J.A.
Lercher, J. Catal. 314 (2014) 21.
2] M. Stöcker, Microporous Mesoporous Mater. 29 (1999) 3.
3] T. Mokrani, M. Scurrell, Catal. Rev. Sci. Eng. 51 (2009) 1.
4] G.A. Olah, Angew. Chem., Int. Ed. 44 (2005) 2636.
5] U. Olsbye, S. Svelle, M. Bjorgen, P. Beato, T.V.W. Janssens, F. Joensen, S. Bordiga,
K.P. Lillerud, Angew. Chem., Int. Ed. 51 (2012) 5810.
[44] S. Svelle, P.O. Rønning, U. Olsbye, S. Kolboe, J. Catal. 224 (2005) 385.
[45] S. Svelle, P.O. Rønning, S. Kolboe, J. Catal. 224 (2004) 115.
[46] I.M. Hill, Y.S. Ng, A. Bhan, ACS Catal. 2 (2012) 1742.
[47] I.M. Hill, S.A. Hashimi, A. Bhan, J. Catal. 291 (2012) 155.
[48] I.M. Hill, S.A. Hashimi, A. Bhan, J. Catal. 285 (2012) 115.
[49] G. Mirth, J.A. Lercher, J. Phys. Chem. 95 (1991) 3736.
[50] J. Van der Mynsbrugge, M. Visur, U. Olsbye, P. Beato, M. Bjørgen, V. Van
Speybroecka, S. Svelle, J. Catal. 29 (2012) 201.
[51] M. Bjørgen, F. Joensen, K.P. Lillerud, U. Olsbye, S. Svelle, Catal. Today 142
(2009) 90.
[52] W.O. Haag, R.M. Lago, P.G.J. Rodewald, Mol. Catal. 17 (1982) 161.
[53] W. Wu, W. Guo, W. Xiao, M. Luo, Chem. Eng. Sci. 66 (2011) 4722.
[54] U.V. Mentzel, S. Shunmugavel, S.L. Hruby, C.H. Christensen, M.S. Holm, J. Am.
Chem. Soc. 131 (2009) 17009.
[
[
[
6] S. Ilias, A. Bhan, ACS Catal. 3 (2013) 18.
7] J.F. Haw, W. Song, D.M. Marcus, J.B. Nicholas, Acc. Chem. Res. 36 (2003) 317.
8] D. Lesthaeghe, V. Van Speybroeck, G.B. Marin, M. Waroquier, Angew. Chem.,
Int. Ed. 118 (2006) 1746.
[
9] D. Lesthaeghe, V. Van Speybroeck, G.B. Marin, Waroquier, Ind. Eng. Chem. Res.
4
6 (2007) 8832.
10] W. Song, D.M. Marcus, H. Fu, J.O. Ehresmann, J.F. Haw, J. Am. Chem. Soc. 124
2002) 3844.
[
(
[
[
[
[
[
11] N.Y. Chen, W.J. Reagan, J. Catal. 59 (1979) 123.
12] R.M. Dessau, J. Catal. 99 (1986) 111.
13] R.M. Dessau, R.B. LaPierre, J. Catal. 78 (1982) 136.
14] B.E. Langner, Appl. Catal. 2 (1982) 289.
[55] G.F. Hutchings, F. Gottschalk, R. Hunter, Ind. Eng. Chem. Res. 26 (1987) 637.
15] T. Mole, J. Whiteside, D. Seddon, J. Catal. 82 (1983) 261.