Supramolecular origins of product selectivity for methanol-to-olefin catalysis on HSAPO-34

Article abstract of DOI:10.1021/ja0041167

Ethylene selectivity in methanol-to-olefin (MTO) catalysis is related to the number of methyl groups on benzene rings trapped in the nanocages of the preferred catalyst HSAPO-34. By correlating the time evolutions of the catalysts' ¹³C NMR spectra and the volatile product distribution following abrupt cessation of methanol flow, we discovered that (in the absence of other adsorbates) propene is favored by methylbenzenes with four to six methyl groups but ethylene is predominant from those with two or three methyl groups. We substantially increased ethylene selectivity by operating at lower methanol partial pressures or higher temperatures, either of which reduces the steady-state average methyl substitution. As a step toward a kinetic analysis of the MTO reaction on HSAPO-34, we treated each nanocage with a methylbenzene molecule as a supramolecule capable of unimolecular dissociation into ethylene or propene and a less highly substituted methylbenzene. Addition of a water molecule to a nanocage containing a methylbenzene produces a distinct supramolecule with unique properties. Indeed, co-feeding water with methanol significantly increased the average number of methyl groups per ring at steady state relative to identical conditions without additional water, and also increased ethylene selectivity, apparently through transition state shape selectivity.

Full text of DOI:10.1021/ja0041167

J. Am. Chem. Soc. 2001, 123, 4749-4754
4749
Supramolecular Origins of Product Selectivity for Methanol-to-Olefin
Catalysis on HSAPO-34
Weiguo Song, Hui Fu, and James F. Haw*
Contribution from the Loker Hydrocarbon Research Institute and Department of Chemistry,
UniVersity of Southern California, UniVersity Park, Los Angeles, California 90089-1661
ReceiVed NoVember 29, 2000
Abstract: Ethylene selectivity in methanol-to-olefin (MTO) catalysis is related to the number of methyl groups
on benzene rings trapped in the nanocages of the preferred catalyst HSAPO-34. By correlating the time evolutions
of the catalysts’ 13C NMR spectra and the volatile product distribution following abrupt cessation of methanol
flow, we discovered that (in the absence of other adsorbates) propene is favored by methylbenzenes with four
to six methyl groups but ethylene is predominant from those with two or three methyl groups. We substantially
increased ethylene selectivity by operating at lower methanol partial pressures or higher temperatures, either
of which reduces the steady-state average methyl substitution. As a step toward a kinetic analysis of the MTO
reaction on HSAPO-34, we treated each nanocage with a methylbenzene molecule as a supramolecule capable
of unimolecular dissociation into ethylene or propene and a less highly substituted methylbenzene. Addition
of a water molecule to a nanocage containing a methylbenzene produces a distinct supramolecule with unique
properties. Indeed, co-feeding water with methanol significantly increased the average number of methyl groups
per ring at steady state relative to identical conditions without additional water, and also increased ethylene
selectivity, apparently through transition state shape selectivity.
Introduction
Here we show that ethylene selectivity is governed by the
number of methyl substituents on the aromatic rings, and the
presence or absence of other molecules in the nanocages. At
higher methanol space velocities (partial pressures), the average
number of methyl groups per ring reaches a steady-state value
of greater than 5 at 400 °C, and the ethylene selectivity is only
ca. 25%. At far lower space velocities, the average number of
methyl groups per ring reaches a steady-state value below 2,
and the catalyst then yields over 60% ethylene. Similarly, the
average number of methyl groups per ring reaches lower steady-
state values at higher temperatures, and this contributes to higher
ethylene selectivities.
As a step toward a kinetic analysis of MTO chemistry in
HSAPO-34, we treat each nanocage containing a methylbenzene
as an independent supramolecule capable of unimolecular
decomposition to either ethylene, {n} f {n - 2, e}, or propene,
{n} f {n - 3, p}, where n implies a methylbenzene molecule
with n methyl groups, and { } designates the nanocage and its
associated Brønsted acid site. We performed a series of
experiments where we allowed a working catalyst to reach
steady state during continuous introduction of methanol before
abruptly cutting off reagent introduction. We then followed the
time evolutions of the rates of product formation and the average
number of methyl groups on benzenes trapped in the catalyst’s
nanocages. The correlation of these observables permitted us
to compare the relative magnitudes of some of the unimolecular
rate constants for the reactions for supramolecules composed
of cages containing a methylbenzene molecule, but no other
adsorbate. A rate equation is derived to account for the time
evolution of the average number of methyl groups per benzene
ring in such experiments.
Control of selectivity is a central problem in catalysis, and
this is particularly challenging with the microporous solid acid
catalysts1-3 used in most industrial catalytic processes. We
recently demonstrated that for some reactions on solid acids,
the active site is a hybrid of inorganic and cyclic organic
components.4,5 The latter function as scaffolds for bond making
and breaking steps that would otherwise be energetically
unfavorable. In the case of methanol-to-olefin (MTO) conver-
sion6 on the preferred catalyst HSAPO-34 (a silico-alumino-
phosphate of CHA topology),7 the organic components are
methylbenzenes8 that self-assemble in the nanometer-size cages
during a kinetic induction period, where they must remain until
burned out, as even benzene is too large to pass through the
0.38 nm windows connecting adjacent cages. An outstanding
problem in MTO catalysis is improving ethylene selectivity at
the expense of propene to better match the demand for the
corresponding polyolefins.
(1) VanSanten, R. A.; Kramer, G. J. Chem. ReV. 1995, 95, 637-660.
(2) Corma, A. Chem. ReV. 1995, 95, 559-614.
(3) Corma, A. Chem. ReV. 1997, 97, 2373-2419.
(4) Haw, J. F.; Nicholas, J. B.; Song, W.; Deng, F.; Wang, Z.; Heneghan,
C. S. J. Am. Chem. Soc. 2000, 122, 4763-4775.
(5) Song, W.; Haw, J. F.; Nicholas, J. B.; Heneghan, K. J. Am. Chem.
Soc. 2000, 122, 10726-10727.
(6) Sto¨cker, M. Microporous Mesoporous Mater. 1999, 29, 3-48.
(7) Wilson, S.; Barger, P Microporous Mesoporous Mater. 1999, 29,
117-126.
(8) Toluene has also been implicated as a “cocatalyst” for methanol
conversion on aluminosilicate zeolites: (a) Mole, T.; Whiteside, J. A.;
Seddon, D. J. Catal. 1983, 82, 261-266. (b) Mole, T.; Bett, G.; Seddon,
D. J. Catal. 1983, 84, 435-445. Other workers have proposed a phenom-
enological carbon-pool mechanism for MTO chemistry on HSAPO-34
without explicitly identifying the organic species responsible for the catalytic
effect: (c) Dahl, I. M.; Kolboe, S. J. Catal. 1996, 161, 304-309. (d) Dahl,
I. M.; Kolboe, S. J. Catal. 1994, 149, 458-464.
Coadsorption of a water molecule into a nanocage with a
methylbenzene produces unique supramolecules, {n, H2O}, and
we find that rate constants for their decomposition to olefinic
10.1021/ja0041167 CCC: $20.00 © 2001 American Chemical Society
Published on Web 04/28/2001

4750 J. Am. Chem. Soc., Vol. 123, No. 20, 2001
Song et al.
products, for example, {n, H2O} f {n - 2, H2O, e} are slower
than in the absence of water, leading to higher steady-state
averages for the number of methyl groups per benzene rings.
Furthermore, the presence of water in the nanocage significantly
increases ethylene selectivity at the expense of propene, and
we attribute this to enhanced transition state shape selectivity.
Experimental Section
Materials and Reagents. HSAPO-34 was prepared according to a
patent procedure.9 XRD showed a pure crystalline phase with the CHA
structure. The product was calcined at 873 K for 10 h to remove the
template agent and pressed into 10-20 mesh pellets. The Brønsted
site concentration was determined to be 1.1 mmol/g. Methanol-13C was
obtained from Isotech, Inc. In general, we used methanol-13C for
experiments in which an NMR measurement was to be made (with or
without GC) and natural abundance methanol if GC was to be the only
analytical method.
Catalysis. Experiments were performed by using the pulse quench
reactor described elsewhere10 with the exception that methanol was
delivered by a motor-driven syringe pump (Harvard Apparatus model
PHD 2000). For each experiment a bed consisting of 0.3 g of catalyst
was activated at 673 K in the reactor under 200 sccm He flow for 2 h
immediately prior to use. This carrier gas feed rate was also used during
methanol introduction in all experiments. For NMR sample preparation,
methanol flow was abruptly ceased a predetermined time (usually 0 to
60 min) prior to quench. Previous studies have shown that the
temperature of the catalyst pellets decreases 150 K in the first 170 ms
of a quench. After quenching each reacted catalyst sample, the reactor
was sealed off and transferred into a glovebox filled with nitrogen.
The catalyst pellets were ground and transferred to a 7.5-mm MAS
rotor that was sealed with a Kel-F end-cap.
Gas Chromatography. A Hewlett-Packard Model 6890 gas chro-
matograph with flame-ionization detector was used to analyze gases
sampled from the reactor product streams with a Valco valve. The
column was 150 m dh150 (Supelco) operated isothermally at 323 K to
permit sampling of the gas stream more frequently than the total analysis
time for any given sample.
Figure 1. 13C CP/MAS NMR spectra (75 MHz) showing the loss of
methyl groups as a function of time from methylbenzenes trapped in
the HSAPO-34 nanocages at 400 °C. For each case, a fresh catalyst
bed was used to convert 0.1 mL of methanol-13C at a WHSV of 8 h-1
,
and then methanol flow was abruptly cut off. The catalyst bed was
maintained at temperature with He flow (200 sccm) for the time
indicated, and then the reactor temperature was rapidly quenched to
ambient. Entire catalyst beds were loaded into MAS rotors to avoid
sampling errors, and cross polarization spectra were measured at room
temperature. The average numbers of methyl groups per ring, Meave
,
were calculated from Bloch decay spectra very similar to the cross
polarization spectra shown.
NMR Spectroscopy. 13C solid-state NMR experiments were per-
formed with magic angle spinning (MAS) on a modified Chemagnetics
CMX-300 MHz spectrometer operating at 75.4 MHz for 13C. Hexam-
ethylbenzene (17.4 ppm) was used as an external chemical shift
standard, and all 13C chemical shifts are reported relative to TMS.
Chemagnetics-style pencil probes spun 7.5 mm zirconia rotors at
typically 6.5 kHz with active spin speed control ((3 Hz).
first flowed 0.1 mL of methanol or methanol-13C onto a 300
mg bed of pelletized HSAPO-34 at 400 °C at a weight hourly
space velocity (WHSV) of 8 h-1 (for a 300 mg catalyst bed
this corresponds to 50 µL/min) before abruptly terminating
methanol flow and waiting a variable time before quenching
the catalyst temperature to ambient. Figure 1 reports solid-state
13C MAS NMR spectra of catalysts prepared with delays from
0 to 60 min. These spectra show an aromatic carbon signal
between 129 and 134 ppm, and a methyl group resonance at
20 ppm that drops with increasing delay between methanol
cutoff and thermal quench. A small, sharp resonance at 25 ppm
is due to isobutane, which like the methylbenzenes, is too large
to exit the nanocages.
Typical 13C experiments included the following: cross polarization
(CP, contact time ) 2 ms, pulse delay ) 1 s, 2000 transients); cross
polarization with interrupted decoupling (contact time ) 2 ms, pulse
delay ) 1 s, 2000 transients, dipolar dephasing time of 50 µs); and
single pulse excitation with proton decoupling (Bloch decay, pulse delay
) 10 s, 400 transients). CP and Bloch decay spectra gave very similar
values for the average number of methyl groups per aromatic ring.
Similar to other NMR work on chemically dilute carbonaceous
materials, the CP spectra here were generally “cleaner” than the Bloch
decay spectra. All spectra reported here were measured with CP, but
the average methyl group numbers reported were determined from the
Bloch decay spectra since these are more commonly regarded as
quantitative. Our conclusions would not change if we reported
integrations from the CP spectra.
The 13C NMR spectra permit direct measurement of the
average number of methyl groups per benzene ring in the
catalyst, Meave
.
6
Meave ) [methyls]/[rings] ) nf
(1)
∑
n
Results
n)0
Decomposition of Methylbenzenes Following Methanol
fn denotes the fractions of rings with n methyl groups. As shown
in Figure 1, Meave decreased from a nearly full complement of
5.6 immediately after methanol cutoff to 1.9 (e.g., xylenes on
average) after 60 min. The last methyl group could be removed
only with great difficulty. At 450 °C, Meave was 1.1 after 60
min and 0.2 after 14 h.
Cutoff. We carried out a series of experiments in which we
(9) Lok, B. M.; Messina, C. A.; Patton, R. L.; Gajek, R. T.; Cannan, T.
R.; Flanigen, E. M. U.S. Patent 4,440,871, 1984.
(10) (a) Haw, J. F.; Goguen, P. W.; Xu, T.; Skloss, T. W.; Song, W.;
Wang, Z. Angew. Chem. 1998, 37, 948-949. (b) Goguen, P. W.; Xu, T.;
Barich, D. H.; Skloss, T. W.; Song, W.; Wang, Z.; Nicholas, J. B.; Haw, J.
F. J. Am. Chem. Soc. 1998, 120, 2651-2652. (c) Xu, T.; Barich, D. H.;
Goguen, P. W.; Song, W.; Wang, Z.; Nicholas, J. B.; Haw, J. F. J. Am.
Chem. Soc. 1998, 120, 4025-4026.
Figure 2 reports gas chromatographic analyses of samples
taken from the product streams immediately prior to thermal

Methanol-to-Olefin Catalysis on HSAPO-34
J. Am. Chem. Soc., Vol. 123, No. 20, 2001 4751
Figure 2. Gas chromatography (flame-ionization detection) analyses
of the volatile products captured immediately prior to thermal quench
from the experiments used to prepare some of the samples for Figure
1. Note the significant increase in ethylene selectivity at longer times.
quench for several of the experiments used to prepare the
samples for Figure 1. One minute after methanol cutoff, the
catalyst was still producing olefinic products with an ethylene
selectivity only slightly higher than that immediately prior to
cutoff; in particular, one notes a much higher selectivity for
propene than for ethylene. As the catalyst continued to age at
400 °C with He flow but no further addition of methanol, the
total rate of olefin production necessarily fell, but the ethylene
selectivity increased dramatically, surpassing that of propene
between 6 and 12 min after cutoff. We measured the rates of
formation of ethylene, propene, and a representative C4 product
from a single experiment in which we made a number of gas
chromatographic injections as a function of time, and these data
are plotted in Figure 3a; ethylene dominated after 8 min. C4
products may form in part from secondary reactions and are
hereafter neglected. Figure 3b compares the ethylene and
propene selectivities, computed from Figure 3a, as a function
of the time evolution of the average number of methyl groups
per ring: a mapping made possible by a smooth fit of the data
in Figure 1. The relationship between molecular structure and
selectivity is clear, a changeover in selectivity to ethylene
occurred when the number of methyl groups per ring decreased
to 2.8, and further decreases in propene selectivity accompanied
further decreases in methyl substitution.
Effects of Space Velocity or Temperature. We explored
rational means of limiting the average number of methyl groups
per ring to see if these conditions did indeed increase ethylene
selectivity. Figure 4 reports 13C NMR spectra that demonstrate
that as the methanol space velocity (and hence partial pressure)
was reduced at 400 °C, the benzene rings reached steady state
with fewer methyl groups per ring. The figure shows that as
the space velocity was decreased (at 400 °C) from 8 h-1 to
0.0008 h-1, Meave decreased progressively from 5.6 to 1.2.
We varied the methanol space velocity by 4 orders of
magnitude (while keeping the He flow constant) and measured
the steady-state olefin selectivities (all at 400 °C and 100%
conversion) reported in Figure 5a. The ethylene selectivity,
which was only 25% at the highest space velocities, increased
dramatically as the space velocity was reduced, and we
Figure 3. (a) The rates of formation of ethylene, propene, and 2-butene
as a function of time from a single experiment similar to those used
for Figures 1 and 2. For these measurements, multiple gas samples
were analyzed from a single catalyst bed as it evolved over time after
cessation of methanol flow. Ethylene and propene are major primary
products of MTO chemistry. (b) Ethylene and propene selectivity as a
function of the average number of methyl groups per ring, Meave. The
data in part b are re-plotted by using an abscissa derived by empirically
fitting the time evolution of Meave in Figure 1 to a smooth curve
generated by Excel. This mapping shows that ethylene is favored by
methylbenzenes with 2 or 3 methyl groups, while propene is favored
with 4 or more methyl groups.
measured an extraordinary steady-state selectivity of 63% at
the lowest space velocity studied. With use of the Meave
measurements in Figure 4, Figure 5b maps the selectivity data
in Figure 5a onto an Meave coordinate. Under steady-sate
conditions, in a catalyst that also contains some methanol,
dimethyl ether, and water in the nanocages, the changeover in
selectivity for dominance of ethylene or propene occurred at
Meave ) 2.1 in Figure 5b compared to 2.8 without methanol,
dimethyl ether, and water (Figure 3b).
We also found that for a fixed space velocity of 8 h-1, Meave
decreased with increasing temperature, and again this correlated
with higher ethylene selectivity. Figure 5 shows that at 450 °C
Meave was 4.3 (cf. 5.6 at 400 °C) and further dropped to 1.8 at
550 °C. The ethylene selectivities were 25% at 400 °C, 34% at
450 °C, and 50% at 550 °C.
Effect of Added Water. As the first step toward studying
the decomposition of methylbenzenes in HSAPO-34 cages in
the presence of other molecules, we studied the effect of added

4752 J. Am. Chem. Soc., Vol. 123, No. 20, 2001
Song et al.
Figure 4. 13C CP/MAS NMR spectra (75 MHz) from experiments
probing the steady-state average number of methyl groups per ring at
various methanol space velocities. In each case we converted methanol
to reach steady-state selectivities, and then abruptly ceased methanol
flow and thermally quenched the catalyst bed without further delay.
Lower space velocities reduce the average number of methyl groups.
water at 400 °C. Water is a coproduct of MTO chemistry that
must be present in at least some nanocages during methanol
conversion, especially at higher space velocities. Figure 7 a,b
shows that if the methanol space velocity is kept fixed but water
is also introduced at a large space velocity, Meave will reach a
larger steady-state value, suggesting that olefin elimination is
retarded by the presence of water in the cages. GC analysis of
the volatile products revealed an ethylene selectivity of 35%
with pure methanol delivered at 5 µL/min, but this increased to
42% when 10% (v/v) methanol in water was delivered at 50
µL/min. We also investigated the time evolution of Meave
following methanol cutoff in the presence of water. Figure 7 c
shows that 60 min after cutoff Meave drops to 1.9 without water,
but with 10 µL/min of water following methanol cutoff (Figure
7d) Meave drops to only 2.5.
Figure 5. (a) Ethylene selectivity at 400 °C vs the logarithm of
methanol weight-hourly space velocity. The catalyst bed, 300 mg, was
first treated with a total of 0.1 mL of methanol at a WHSV of 8 h-1 at
400 °C to create methylbenzenes, and then was held at a given space
velocity until selectivity reached a steady-state value. Note the very
high ethylene selectivities at very low space velocities. (b) The ethylene
selectivity data in part a plotted against Meave
.
Meave versus time from experiments such as Figure 1 follow
no simple integrated rate law, because of contributions from
rings with different numbers of methyl groups and the time
evolution of the fractions of rings, fn, with n methyl groups.
Consider the possible reactions of a trimethylbenzene molecule
isolated in a nanocage. Trimethylbenzene can either lose
ethylene and leave toluene with a rate constant k({3}f{1, e})
Discussion
The overall kinetics of olefin synthesis on HSAPO-34 are
no doubt complicated, and will vary with coadsorption of
methanol, water, or other species into the nanocages containing
the methylaromatics. It is convenient to consider each nanocage
with a methylbenzene molecule to be a supramolecular entity,
each including and delimited by the cage boundaries and
possessing one acid site, but distinguished by the number of
methyl groups on the benzene ring and the presence or absence
of other adsorbed molecules. We will assume that equilibration
of isomers is rapid compared to olefin elimination, and that we
can treat, for example, a cage containing only xylene and an
acid site as a single supramolecule, and not distinguish the three
isomeric forms.
Decomposition of Methylbenzenes in HSAPO-34. The
easiest case to discuss is the cutoff experiments of Figures 1-3,
where there are no water or methanol molecules in the cages
(at least at longer times). We can treat the decomposition of
methylbenzenes to ethylene or propene and less-substituted
methylbenzenes as unimolecular dissociation steps. Plots of
e
) k3 or it can instead lose propene and leave benzene with a
p
rate constant k({3}f{0, p}) ) k3 . If all methylbenzenes
function independently, and diffusion of olefin products out of
the cages is rapid, then the rate of change of the average number
of methyl groups per ring in an experiment like Figure 1 is
described by
6
6
d(Meave
)
dfn
dt
p
e
)
n
)
n[-(k e + kn )fn + kn fn + 2
+
∑
∑
n
dt
n)0
n)0
p
kn fn + 3] (2)
where k0e ) k0p ) k1e ) k1p ) k2p ) 0, and fi ) 0 when i > 6.
Figures 1 through 3 allow us to infer the relative values of
some of the rate constants. While xylene cannot eliminate

Methanol-to-Olefin Catalysis on HSAPO-34
J. Am. Chem. Soc., Vol. 123, No. 20, 2001 4753
Figure 6. 13C CP/MAS NMR spectra (75 MHz) probing the steady-
state average number of methyl groups per ring at fixed methanol space
velocity (8 h-1) at various temperatures. In each case, the thermal
quench occurred immediately upon cessation of methanol flow. The
average number of methyl groups decreased with increasing temper-
ature.
Figure 7. 13C CP/MAS NMR spectra (75 MHz) from experiments
probing the effect of co-feeding water on the average number of methyl
groups per ring at a 400 °C reactor temperature. In every case 100 µL
of pure methanol was first delivered at WHSV ) 8 h-1 (50 µL/ min)
to form methylbenzenes, and then conditions were varied: (a) control
experiment, pure methanol at 5 µL/ min to reach steady-state and then
thermal quench immediately upon cessation of methanol flow; (b) 10%
(v/v) methanol in water at 50 µL/ min to reach steady-state and then
thermal quench immediately upon cessation of flow; (c) control
experiment, wait 60 min with no flow (other than He gas) after forming
methylbenzenes, then thermal quench; and (d) wait 60 min while
flowing 10 µL/min water with He gas after forming methylbenzenes,
then thermal quench. The effect of added water is to decrease the rate
of methylbenzene decomposition to olefins, hence an increase in Meave
relative to the control experiments. Water also increases ethylene
selectivity (see text).
e
propene, k2 is not necessarily zero. Figure 3 shows that the
changeover in selectivity at 400 °C occurs with three methyl
e
p
groups per ring, implying that k3 > k3 . We conclude that
retention of at least one methyl group on the ring stabilizes
transition states leading to olefin elimination. Also, k6p is greater
e
than k6 , and propene is favored at higher space velocities and
lower temperatures. Unfortunately, we also have the following
e
inequalities, k6p > k3p and k6e > k3 , and the conditions for the
most rapid methanol conversion on this catalyst (large Meave
)
are not those that provide the highest ethylene selectivity (small
Meave).
on the transition state, and solvate every point on the potential
energy surface. (Note that a cage with trimethylbenzene and
two water molecules would also be a distinct supramolecule
with unique rate constants for decomposition to ethylene or
propene.) Another possible reaction for a nanocage containing
a methylbenzene and water is demethylation to form methanol,
but this can be ruled out based on the endothermicity of such
reactions in the gas phase. For example, the reaction of toluene
and water to form benzene and methanol has a ∆H of +17.7
kcal/mol, implying a negligible equilibrium constant.11
The results in Figure 7 can most easily be interpreted in the
context of k({n, H2O}f{n - 2, H2O, e} and k({n, H2O}f{n
- 3, H2O, p}) relative to the corresponding rate constants in
the absence of water. Figure 7 implies that the rates of both
ethylene and propene production are lower with water in the
nanocages. Furthermore, the additional steric constraints pro-
vided by adsorbed water provide the most likely explanation
for the increased ethylene selectivity in the water co-feed
experiments.
Temperature Effect. At higher temperatures the catalyst
equilibrated to lower values of Meave and produced higher
ethylene selectivities. Of course, this correlation does not rule
out the possibility of other contributions to the temperature
dependence of olefin selectivity. In a recent theoretical analysis
of the elimination of ethylene and propene from another cyclic
molecule, we found a higher barrier for the former relative to
the latter.4 If (as we believe) ethylbenzenes and propyl or
isopropylbenzenes are precursors to olefins on HSAPO-34,
elementary consideration of the stabilities of reactive intermedi-
ates would predict a higher barrier to ethylene than to propene.
A higher barrier to ethylene would also contribute to improved
ethylene selectivity at higher temperatures, as we observed.
Thus, we suspect that both the decreased Meave and the greater
rate of barrier crossing at higher temperatures contribute the
increase in ethylene selectivity with temperature.
Effect of Water. Under realistic operating conditions of
appreciable methanol space velocity, some of the cages will
also contain methanol, dimethyl ether, and/or water molecules.
A supramolecule consisting of a nanocage, an acid site,
trimethylbenzene, and a single water molecule could still
eliminate an olefin, but we would not expect, for example, k({3,
In the more general case of a catalyst operating at appreciable
methanol apace velocity we also have reactions such as {n, CH3-
OH} f {n + 1, H2O}. The analogous reaction with dimethyl
ether would produce a methanol molecule. Furthermore, me-
e
H2O}f{1, H2O, e}) to be identical with k({3}f{1, e}) ) k3 ,
since the adsorbate would impose steric constraints, especially
(11) NIST Chemistry WebBook (http://webbook.nist.gov/chemistry/).

4754 J. Am. Chem. Soc., Vol. 123, No. 20, 2001
Song et al.
thylation could occur on the side chains to more directly give
precursors to ethylene or propene, but we do not consider the
detailed mechanism of olefin formation in this contribution. The
interplay between these various reactions, equilibration of
methanol, dimethyl ether, and water, and mass transport of
reactants and products through the catalyst are reflected in
observables such as Meave and ethylene selectivity.
Coadsorption of water produces distinct supramolecular
species, for example {n, H2O}, and these form olefins more
slowly than the corresponding species without water, {n}. Co-
feeding of water during methanol conversion results in an
increase in the average value of n (Meave) in the catalyst when
steady state is reached between methanol introduction and olefin
formation. The presence of H2O in the nanocage shifts the
selectivity in favor of ethylene: the product formed through
the smallest transition state.
Conclusions
We believe that the present study is a major step forward in
the understanding of an important problem in heterogeneous
catalysis. Also, the level of integration of volatile product
analysis and NMR data on the solid catalyst structure (Figures
3b and 5b) represents a methodological advance. Supplementary
analysis by GC has become commonplace in NMR studies of
catalysis, but these data are not typically combined to the extent
or advantage demonstrated here.
This investigation has found the origins of product selectivity
in methanol-to-olefin chemistry on HSAPO-34. At 400 °C
nanocages containing only methylbenzenes preferentially yield
ethylene when the aromatic ring has two or three methyl groups,
while four to six methyl groups are more likely to lead to
propene as a product. This insight directly led us to significant
improvements in ethylene selectivity by either reducing the
methanol space velocity or increasing the temperature: we
showed that either change reduces the average number of methyl
groups per ring at steady-state. We partially interpreted the
kinetics of this process by focusing on the unimolecular
decomposition of methylbenzenes in the nanocages viewed as
the reaction of one supramolecule forming another, for example
{n} f {n - 2, e}.
Acknowledgment. J.F.H. is supported by the National
Science Foundation (CHE-9996109) and the U.S. Department
of Energy (DOE) Office of Basic Energy Sciences (BES) (Grant
No. DE-FG03-93ER14354).
JA0041167