On the way to biofuels from furan: Discriminating Diels-Alder and ring-opening mechanisms

Article abstract of DOI:10.1021/cs4003904

We performed kinetics experiments and quantum calculations to investigate the reaction of furan to benzofuran catalyzed by the acidic zeolite HZSM-5, which is a key step in the conversion of biomass to biofuels through catalytic fast pyrolysis. The reaction was studied experimentally by placing the zeolite in contact with solution-phase furan and detecting the benzofuran product over the temperature range 270-300 C, yielding an apparent activation energy of 72 ± 3 kJ/mol. The reaction was modeled in gas and zeolite phases to determine the energetics of the following two competing pathways: a Diels-Alder mechanism often assumed in interpretations of experimental data and a ring-opening pathway predicted by the chemoinformatic software RING. Quantum calculations on the zeolite/guest system were performed using the ONIOM embedded cluster approach. We computed the energetics of reactants, products, and all intermediate steps. Locating relevant transition states fell beyond our computational resources because of system size and the ruggedness of the energy landscape. The Diels-Alder mechanism in the gas phase was found to pass through a high-energy intermediate roughly 380 kJ/mol above the reactant energy, which reduces to approximately 200 kJ/mol in HZSM-5. In contrast, the ring-opening mechanism passes through a gas-phase intermediate roughly 500 kJ/mol above the reactant energy, which falls to approximately 50 kJ/mol in HZSM-5. The energy of the ring-opening mechanism over HZSM-5 fits into the experimentally determined energy budget of 72 ± 3 kJ/mol. These experimental and computational results highlight the importance of the ring-opening mechanism for this key step in making biofuels. Our results strongly indicate that, in the cavities of HZSM-5, the condensation of two furan molecules to form benzofuran and water does not proceed by a Diels-Alder reaction between the reactants.

Full text of DOI:10.1021/cs4003904

Research Article
pubs.acs.org/acscatalysis
On the Way to Biofuels from Furan: Discriminating Diels−Alder and
Ring-Opening Mechanisms
S. Vaitheeswaran,† Sara K. Green, Paul Dauenhauer, and Scott M. Auerbach*
,‡
‡
‡
,†,‡
†
‡
Department of Chemistry and Department of Chemical Engineering, University of Massachusetts, Amherst Massachusetts 01003,
United States
*
S Supporting Information
ABSTRACT: We performed kinetics experiments and quantum calcu-
lations to investigate the reaction of furan to benzofuran catalyzed by the
acidic zeolite HZSM-5, which is a key step in the conversion of biomass to
biofuels through catalytic fast pyrolysis. The reaction was studied
experimentally by placing the zeolite in contact with solution-phase
furan and detecting the benzofuran product over the temperature range
2
70−300 °C, yielding an apparent activation energy of 72 ± 3 kJ/mol. The
reaction was modeled in gas and zeolite phases to determine the energetics
of the following two competing pathways: a Diels−Alder mechanism often
assumed in interpretations of experimental data and a ring-opening
pathway predicted by the chemoinformatic software RING. Quantum
calculations on the zeolite/guest system were performed using the
ONIOM embedded cluster approach. We computed the energetics of
reactants, products, and all intermediate steps. Locating relevant transition states fell beyond our computational resources
because of system size and the ruggedness of the energy landscape. The Diels−Alder mechanism in the gas phase was found to
pass through a high-energy intermediate roughly 380 kJ/mol above the reactant energy, which reduces to approximately 200 kJ/
mol in HZSM-5. In contrast, the ring-opening mechanism passes through a gas-phase intermediate roughly 500 kJ/mol above the
reactant energy, which falls to approximately 50 kJ/mol in HZSM-5. The energy of the ring-opening mechanism over HZSM-5
fits into the experimentally determined energy “budget” of 72 ± 3 kJ/mol. These experimental and computational results
highlight the importance of the ring-opening mechanism for this key step in making biofuels. Our results strongly indicate that, in
the cavities of HZSM-5, the condensation of two furan molecules to form benzofuran and water does not proceed by a Diels−
Alder reaction between the reactants.
KEYWORDS: biofuels, catalytic pyrolysis, furan, Diels−Alder, HZSM-5, zeolite catalysis, ONIOM, QM/MM
1
. INTRODUCTION
experiments and calculations to obtain fundamental insights
into this key step in catalytic fast pyrolysis.
Catalytic fast pyrolysis of biomass, involving rapid heating of
biomass in the presence of nanoporous zeolite catalysts such as
HZSM-5, has proven to be a promising route for converting a
There are a great many possible mechanisms by which two
furans may condense to form benzofuran and water. To assist
in the identification and enumeration of plausible pathways for
complex reactions, chemoinformatic software tools have been
created that codify a set of assumed rules of chemistry and
chemical mechanisms. One specific example is the Rule Input
1
2
variety of biomass sources into aromatic−range biofuels. While
this is an attractive approach, only a fraction of the biomass
carbon ends up in the resulting fuel (in effect increasing the
price of this cellulosic biofuel), largely because of the tendency
to form solid carbonaceous materials (i.e., “coke”) in the
process. Optimization of this technology has been hampered by
a lack of fundamental understanding of the catalytic
5
Network Generator program or RING, which is an automated
reaction network generation and analysis tool. Given certain
rules and constraints, such as the maximum number of allowed
steps, RING generates possible reaction networks connecting
given reactants and products. However, these codes do not
compute energetics, nor do they account for catalytic surfaces
or active sites. As such, we advocate taking plausible outputs
from codes like RING as inputs to accurate energy calculation
methods such as density functional theory (DFT). One
3
,4
mechanisms at play. Previous experimental studies have
shown that furan is a good proxy for vapors that arise during
cellulose pyrolysis, and that benzofuran is made in HZSM-5
nanopores as an intermediate on the way to making aromatic−
range biofuels (Figure 1). However, there is no information on
the mechanism by which benzofuran is made, nor is there
substantial evidence that the process from furan to benzofuran
occurs via Diels−Alder cycloaddition, which has been assumed
Received: May 24, 2013
Revised: July 25, 2013
3
in the literature. To address this issue, we performed
©
XXXX American Chemical Society
2012
dx.doi.org/10.1021/cs4003904 | ACS Catal. 2013, 3, 2012−2019

ACS Catalysis
Research Article
Below we investigate relatively large system sizes with many
degrees of freedom, making the search for relevant transition
states extremely challenging, especially for rugged energy
landscapes. We show below that by combining experimental
and computational data, we can still discriminate between
Diels−Alder and ring-opening mechanisms without having to
compute transition states. In particular, we find that although
the Diels−Alder pathway involves lower energies in the gas
phase, the ring-opening process is much more favored in the
zeolite, with energetics that fit well within the energy budget
determined by kinetic experiments.
The remainder of this article is organized as follows: in
section 2, we describe the experimental and computational
methods pursued below; in section 3, we provide results and
discussion on the experiments and on the calculations for both
gas-phase and zeolite-catalyzed systems; and in section 4, we
offer concluding remarks.
Figure 1. Potential chemical reactions during furan pyrolysis at 600
°C. The colored box highlights the key step that is investigated in the
present work.
challenge with this translation, which we address below, is that
RING outputs bonding patterns using the Lewis−structure
language of valence−bond theory, while accurate energies are
computed using the molecular−orbital language of DFT. In the
work described below, we compute energetics for a ring-
opening pathway suggested by RING as a mechanism
competing with Diels−Alder cycloaddition for converting
furan into benzofuran.
2
. METHODS
2.1. Experimental Section. We measured the activation energy
for the formation of benzofuran from furan, catalyzed by HZSM-5.
Experiments were conducted in a 100 mL high-pressure, high-
temperature batch reactor equipped with a gas entrainment impeller
from Parr Instrument Co. (model 4598HTHP). The reaction solution
consisted of 60 mL of 1 M furan (Acros Organics, 99%) in heptane
(Alfa Aesar, 99%) with 1 mL of n-tridecane (Acros Organics, 99%)
Several computational studies have provided mechanistic
information on pathways for processing select biomass-derived
species. For example, recent theoretical studies of the
conversion of biomass-derived compounds like fructose,
glucose, and 5-hydroxymethylfurfural (HMF) have been
investigated in the gas phase or in aqueous solution using
used as an internal standard. HZSM-5 (CBV−3024E, Zeolyst) with a
Si/Al ratio of 15 was calcined at 600 °C in a muffle furnace for at least
6 h and was used with a loading of 0.25 g catalyst to 0.06 mols reactant
(
i.e., 60 mL of 1 M furan with 0.25 g HZSM-5). The catalyst had a
2
−1
surface area of 405 m g , a Brønsted acid site density of 0.71 mmol
g , as measured using isopropylamine TPD-TGA, and a total acid site
density of 1.10 mmol g , as measured using ammonia TPD. Prior to
−1
−1
6
−10
quantum chemical methods.
Other researchers have used
each experiment, the reactor vessel was purged with N . The reactor
2
Car−Parrinello molecular dynamics with metadynamics to
vessel was then stirred at a rate of 900 rpm while heating to the desired
temperature between 270−300 °C. An initial sample was taken once
the desired temperature was achieved and subsequent samples were
taken at 10 min intervals for the first 30 min of reaction such that furan
conversion was less than 10%. Standard error was calculated for the
1
1,12
study the conversion of glucose in aqueous acids.
Cluster
models of the zeolite HZSM-5 have been used to probe the
13
relatively simple process of protonating fructose. The zeolite-
catalyzed cycloaddition of small olefins and substituted furans
14
9
7
0% confidence interval. Liquid samples were analyzed by an Agilent
890A gas chromatography system, equipped with a Restek
has also been modeled. However, these particular calculations
were performed without a representation of the cage structure
of the zeolite catalyst, leaving open the question of shape-
selective effects. Explicit inclusion of the zeolite framework in
the energy calculations is thus an important step forward in
understanding zeolite-catalyzed biomass conversion, which we
pursue in the present work by considering sufficiently large
embedded cluster models of HZSM-5.
Stabilwax−DA capillary column and an FID detector. Helium was
used as the carrier gas with a column flow rate of 30 mL/min. One
microliter liquid sample was injected for each analysis. Initial rates of
benzofuran production were fitted to the Arrhenius temperature
dependence to extract an apparent activation energy.
2.2. Computational Section. We commenced our theoretical
5
study by applying the cheminformatics program RING. Given a set of
5
Figure 2. Pathway and output produced by RING for the acid catalyzed conversion of furan to benzofuran.
2
013
dx.doi.org/10.1021/cs4003904 | ACS Catal. 2013, 3, 2012−2019

ACS Catalysis
Research Article
Figure 3. Geometry optimized configuration of benzofuran and water in the HZSM-5 cluster, viewed (a) along the straight channel and (b)
perpendicular to the straight channel. The quantum mechanical (QM) layer in the zeolite cluster is shown in tube representation, the molecular
mechanics (MM) layer in wire frame, and the adsorbates in the ball−and−stick model. Panel c shows the same perspective as panel b, but only the
adsorbates and zeolite atoms within two bonds of the acid site are shown, all in ball-and-stick representation. Atom color key: yellow, C; red, O; gray,
Si; blue, Al; white, H.
reactants, products, and reaction rules, RING generates a network of
reactions that can be used to generate a “first-guess” pathway for the
process of interest. In the absence of a comprehensive understanding
of the zeolite-catalyzed chemistry of oxygenated compounds like furan,
reaction rules have to be based on solution-phase or gas-phase organic
chemistry, or experimental observations on related systems. A priori
decisions have to be made on which chemical reactions and
intermediate species need to be considered by the program and
which ones may be disregarded. Hence, the results obtained from
RING may be subjective, reflecting the chemical intuition of the user.
Nevertheless, it is a very useful tool for identifying the chemical species
and elementary reactions that are likely to be important.
which is computationally too expensive given the large quantum
cluster size (∼400 electrons). As such, we rely below on the nexus
between experimental data and intermediate energetics to elucidate
benzofuran production.
18
1
7,18
We used a finite cluster model for HZSM-5.
Fermann et al.
have shown that such finite zeolite clusters can provide an excellent
alternative to more expensive periodic models for the calculation of
reaction energies, because true long−range forces approximately
cancel when computing energy differences between nearby config-
18
17
urations. Following Agarwal et al., the HZSM-5 cluster is centered
19
on O(13), a catalytically important site because of its location at the
intersection of the straight and zigzag channels where guest molecules
can be accommodated.
For the process of two furans reacting to yield benzofuran and
water, we used RING to generate pathways with a (i) list of possible
reactions and (ii) the following two constraints: pathways must have
no more than 8 steps, and they must be acid-catalyzed. The second
constraint was chosen based on experimental observations that the
conversion of furan over silicalite (the all-silica analog of HZSM-5)
HZSM-5 is modeled as an embedded cluster using the 2-layer
1
6
ONIOM methodology (see Figure 3). The quantum mechanical
QM) layer, which is treated at a high level of theory, is composed of
1 tetrahedral (T = Si or Al) sites, 30 oxygen atoms, and 1 acid site
(
1
3
proton. The molecular mechanics (MM) layer, which is treated with a
less expensive (and less accurate) level of theory, has 132 T sites and
does not lead to the desired aromatic products. The list of possible
reactions considered by RING included standard acid catalysis
reactions such as protonation/deprotonation, dehydration, oligomeri-
3
11 O atoms. Zeolite clusters of this size have been demonstrated to
18
1
5
converge relevant reaction energies with respect to system size. For
the high level of theory, we used density functional theory (DFT) with
the B3LYP functional
zation, ring-opening/closing, and hydrogen transfer reactions. In
1
5
addition, carbonylation/decarbonylation steps were included to
2
0,21
account for Cheng et al.’s experimental detection of carbon
and the 6-311G(d,p) basis set. This model
3
monoxide. This yielded the reaction pathway shown in Figure 2.
chemistry has been shown to capture ∼90% of barrier heights and
2
2
The pathway in Figure 2 indicates that the acid-catalyzed conversion
of furan to benzofuran is initiated by the protonation and ring-opening
of a furan molecule, followed by the alkylation of a second furan. Using
this pathway as a guide, we exhaustively explored all possible
protonated and ring-opened derivatives of furan. In subsequent
steps, we explored chemical species as thoroughly as our computa-
tional resources allowed. In the zeolite, we sampled 5−10 relevant
orientations/conformations of each guest molecule in the HZSM-5
intersection. Conformations that were judged as unlikely to lead to the
next step, for example, requiring very long-range proton transfer, were
treated as off-pathway “dead ends” and were discarded. The lowest
energy conformation of each intermediate was used to construct the
zeolite-catalyzed pathways reported below.
reaction energies for acid−base chemistry in zeolites. For the low
23
level of theory, we used the universal force field (UFF), which is
generic and hence convenient but not optimized for any particular
zeolite or composition. All atoms in the QM layer were allowed to
relax freely, while the MM layer atoms were frozen at their
17
crystallographically−determined positions, as in our previous work.
The QM layer is mechanically but not electronically embedded in the
MM layer, that is, there are no external electrostatic interactions
included in the quantum one−electron integrals. Dangling bonds on
oxygen atoms at the periphery of the QM layer were capped with
24
hydrogens. Computations were performed using Gaussian 09. In
what follows, we illustrate intermediate structures for reactions in
zeolites using the format shown in Figure 3c, where only zeolite atoms
within two bonds of the acid site are shown for visual simplicity. We
reiterate that all calculations were performed using the cluster shown
in Figure 3a and 3b.
As touched upon in the introduction, locating relevant transition
states fell beyond the scope of this study for the following two reasons:
ruggedness of the energy landscape and the relatively large system size.
In particular, the floppiness of Si−O−Si angles in the zeolite produced
substantial framework relaxation along reaction pathways, making for a
rugged energy landscape littered with many saddle points unrelated to
the process of interest. To overcome this problem, we applied the
Bonding in key structures was analyzed by computing atomic
charges using the electrostatic potential (ESP) procedure of Kollman
2
5,26
and co-workers.
This ensured that our calculations are reasonably
1
6
well converged with respect to basis set by comparing with ESP
charges from B3LYP/6-311G++(d,p), which contains more diffuse
functions. The ESP charges also pointed to the most important Lewis
structures along the reaction pathways, which assisted in interpreting
DFT results in light of RING predictions.
nudged elastic band (NEB) method within the ONIOM formalism
for finding elusive transition states (results not shown). However, even
this approach failed in the present application, because initial−guess
pathways failed to capture the nature of zeolite relaxation. The failure
of NEB thus necessitates a more exhaustive search for saddle points,
2
014
dx.doi.org/10.1021/cs4003904 | ACS Catal. 2013, 3, 2012−2019

ACS Catalysis
. RESULTS AND DISCUSSION
.1. Experiments. Figure 4 shows an Arrhenius plot of
Research Article
3
3
initial rates of benzofuran production from experiments using
Figure 5. Gas-phase geometry-optimized structures and energies of all
chemical species arising from the protonation and subsequent ring-
opening of furan. These are the first two reactions in the ring-opening
pathways from furan to benzofuran. Dotted blue arrows represent ring-
opening reactions. Energies of all species are ground-state electronic
energies plotted relative to furan, and include the energy cost of
27
stripping a proton from the zeolite framework (+1170 kJ/mol).
Figure 4. Arrhenius plot for the production of benzofuran from 1 M
furan in heptane over 0.25 g of HZSM-5.
C3 only exhibits two resonance structures. This correlates with
the fact that protonation at C2 is lower in energy than that at
C3 by about 50 kJ/mol. ESP charges on the C2 protonated
species (Figure S3 in Supporting Information) show that the
dominant Lewis structure in the mechanism shown in Figure 2
is the one in which the positive charge is localized on the C5
atom (see Supporting Information for the rationale behind this
assignment).
The dotted blue arrows in Figure 5 represent ring-opening
reactions, and identify the structures that result from the ring-
opening of each of the three protonated species. Two more
ring-opened species can be generated by single hydrogen shifts,
making a total of five protonated, ring-opened derivatives of
furan that are stable in the gas phase. We note that the
energetic order of protonated furan species does not correlate
with the corresponding ring-opened structures, producing the
line crossings in Figure 5.
The next step in the process is the alkylation of a second
furan molecule at its C2 or C3 atom by one of the positively
charged, ring-opened species. The open-ring species can react
either at the carbon atom nearest, or the one farthest from the
oxygen atom. Thus, each of the five ring-opened species can
give rise to four alkylated derivatives, making a total of twenty
alkylated species. Deprotonation, ring-closing, and dehydration
steps complete the route to benzofuran. At each step of the
process, many of the intermediates can interconvert via single
hydrogen shifts, and are relatively closely spaced in energy. An
interconnected web of pathways can thus be drawn from furan
to benzofuran. Below, we map out two representative pathways
with the lowest energy intermediates at the second
(protonation) and third (ring-opening) steps.
Figure 6 shows the structures and energies of stable
intermediates in a gas phase pathway originating with the
protonation of furan at the C2 atom (step 2). This protonated
furan ring opens (step 3) in a process that is endothermic by
roughly 180 kJ/mol, to give the highest-energy intermediate in
the pathway. In step 4, the open-ring species reacts at the
carbon atom farthest from the oxygen, and alkylates a second
furan molecule at the C2 position. Two branches of the
pathway can be identified at the next step. The positively
the methods described above. Additional plots containing the
raw data are given in Figures S1 and S2 of the Supporting
Information. The good linear fit in Figure 4 is consistent with a
single, dominant mechanism in the temperature range 270−300
°
C. The slope extracted from Figure 4 corresponds to an
apparent activation energy of 72 ± 3 kJ/mol for the HZSM-5
catalyzed process. This apparent activation energy is inclusive
of the following physical processes: diffusion, adsorption−
desorption from the zeolite catalyst, and of course, chemical
reaction. As such, this measured apparent activation energy is
likely an upper bound for the activation energy of the rate-
determining chemical reaction in benzofuran production.
3
.2. Calculations: Gas Phase Pathways. 3.2.1. Ring-
Opening Routes. To establish energetic baselines for the
zeolite-catalyzed processes of interest, we investigated the5
conversion of furan to benzofuran in the gas phase. RING
predicted that the acid-catalyzed conversion of furan to
benzofuran proceeds by a sequence of protonation, ring-
opening, ring-closing, and dehydration steps shown in Figure 2.
The overall reaction involves two furan molecules condensing
to give benzofuran and water. In our gas-phase calculations, we
exhaustively mapped out intermediates for the first two steps in
the process: the protonation and subsequent ring-opening of
one furan molecule.Figure 5 shows the geometry-optimized
structures and ground state electronic energies of the three
protonated and five ring-opened species found in our gas-phase
calculations. Energies are plotted relative to the ground state of
furan. The zeolite solid acid is accounted for, approximately, by
shifting the zero of energy by an amount equal to the energy
required to strip a proton from the zeolite, assumed to be
2
7
+
1170 kJ/mol. For example, a value of 300 kJ/mol in Figure 5
means that the gas-phase proton affinity of furan is 300 kJ/mol
less than the assumed acidity of the zeolite, rendering
protonation of furan endoergic by 300 kJ/mol. We emphasize
that this is simply a shift in the zero of energy assumed in our
gas-phase calculations, to facilitate comparison with the zeolite
catalyzed systems. Figure 2 shows that protonation at C2 is
consistent with three resonance structures, while protonation at
2
015
dx.doi.org/10.1021/cs4003904 | ACS Catal. 2013, 3, 2012−2019

ACS Catalysis
Research Article
Figure 6. Gas-phase geometry optimized structures and energies of
stable intermediates in a pathway arising from the protonation of furan
at the C2 atom. Energies are plotted relative to the ground state
electronic energy of two furans (step 1), and for positively charged
species (i.e., those with positive energies), include the energy cost of
Figure 7. Stable intermediates and their energies in a gas phase
pathway initiated by the Diels−Alder cycloaddition of two furan
molecules. Energies are plotted relative to the ground state energy of
two furans (step 1), and for positively charged species, include the
energy cost of stripping a proton from the zeolite framework (+1170
2
7
27
stripping a proton from the zeolite framework (+1170 kJ/mol).
kJ/mol).
charged alkylated species can deprotonate and restore the
aromaticity of the furan ring (Figure 6, solid red line, step 5)
and then ring close (step 6). Alternatively, the alkylated species
in step 4 can restore the aromaticity of the furan ring by a series
of proton hops, and then ring close as a positively charged
species (Figure 6, dotted blue line). Deprotonation restores the
aromaticity of the furan ring at step 6, where the two branches
of the pathway merge. Step 7 is a dehydration reaction, and a
final deprotonation in step 8 leads to benzofuran.
occurrence of dehydration instead of decarbonylation reactions
28
cannot be treated as a signature of a Diels−Alder process. In
the gas phase, the data in Figures 6, S4 (Supporting
Information), and 7 suggest that the Diels−Alder route appears
more likely because of intermediates 80−120 kJ/mol lower in
energy than in the ring-opening pathways.
We show in the next section that this conclusion does not
hold for the corresponding zeolite-catalyzed processes.
3.3. Calculations: HZSM-5 Pathways. With the gas-phase
pathways as a guide, we explore the same processes in the
HZSM-5 cluster, starting with the structural features. Figure 8
shows the geometry-optimized structures of stable intermedi-
ates in the zeolitic pathway following the ring-opening route.
The two branches, I and II, correspond to the gas phase
pathways of Figures 6 and S4 (Supporting Information)
A second representative pathway for the gas phase
conversion of furan to benzofuran is shown in Figure S4
(
Supporting Information) with energetics similar to those in
Figure 6. This pathway starts with the highest energy
intermediate at step 2, the furan derivative protonated at the
O atom. The path to benzofuran is completed by ring-opening,
alkylation, ring closing and dehydration steps, similar to the first
pathway in Figure 6. In both pathways, dehydration is the
penultimate step and occurs after the ring closing. Similar
pathways can be constructed in which the dehydration step
occurs before the ring closing (data not shown).
3
.2.2. Diels−Alder Routes. A recent study of the conversion
3
of furan to aromatic compounds over HZSM-5 postulated that
the production of benzofuran occurs via a Diels−Alder
condensation between two furans. Diels−Alder processes
have also been invoked to explain the reactions between
furanic compounds and olefins in the zeolite-catalyzed
1
4,28
production of toluene and xylene.
To begin exploring this
alternate route to benzofuran, we mapped out a representative
Diels−Alder pathway in the gas phase in Figure 7. This
pathway starts with a Diels−Alder cycloaddition reaction, in
which one furan is the diene and the other is the dienophile.
The cycloadduct can then be protonated at the bridge O atom,
a reaction that is endothermic by roughly 300 kJ/mol.
Deprotonation and dehydration steps complete this pathway
to benzofuran, without any open ring intermediates. Similar
pathways can be drawn in which a protonation step occurs
before the cycloaddition.
Figure 8. Geometry optimized structures of the stable intermediates in
the zeolite-catalyzed, minimum energy pathway. In the zeolite, only
the acid site and adjacent atoms are displayed, as in Figure 3c. Branch I
is initiated by the stable protonation of furan at the C2 atom, while
branch II proceeds by the transient protonation of furan at the O1
atom.
In both classes of pathways, ring-opening and Diels−Alder,
oxygen is removed through dehydration reactions. Hence, the
2
016
dx.doi.org/10.1021/cs4003904 | ACS Catal. 2013, 3, 2012−2019

ACS Catalysis
Research Article
respectively. The first step in the process is the adsorption of
furan in the zeolite (Figure 8A). In the minimum energy
configuration, one furan is hydrogen bonded to the zeolite acid
site, at a distance of roughly 3.4 Å from the second furan
molecule. Thus, there are two reacting furan molecules per acid
site, consistent with the experimental observation that furan
is, therefore, an off-pathway dead end. In the lowest energy, on-
pathway configuration (Figure 9B), the cycloadduct is hydro-
gen bonded to the acid site at the bridge O atom. Subsequently,
the adduct is protonated at the bridge O atom (Figure 9C),
following which, deprotonation (Figure 9D) and dehydration
(Figure 9E) reactions lead to benzofuran (Figure 9F) with the
loss of a water molecule.
We next consider the energetics of the ring-opening and
Diels−Alder zeolite-catalyzed processes in HZSM-5. The solid
red and dotted blue lines in Figure 10 show the energies of the
adsorbs in HZSM-5 at a furan-to-aluminum molar ratio of
3
1.73.
In branch I, the zeolitic proton is transferred to the furan at
the C2 position (Figure 8B), followed by the ring-opening of
the protonated furan into an alkoxy species covalently bonded
to the zeolite framework (Figure 8C). In branch II, the zeolitic
proton is transferred to the O atom of the furan, but this
protonated furan derivative is not a stable intermediate. It is
instead a transition state on the proton hopping path between
the central O atom of the zeolite cluster, and one of the other
zeolite O atoms adjacent to the Al atom (Figure S5, Supporting
Information). The transiently protonated furan ring opens to a
species that is hydrogen bonded to the zeolite framework
(
Figure 8D). At the next step branches I and II merge, and the
ring-opened species alkylate the second furan molecule (Figure
E). Deprotonation (Figure 8F), ring-closing (Figure 8G), and
8
dehydration (Figure 8H) steps lead to benzofuran (Figure 8I)
with the loss of a water molecule.
We note that this mechanism is also a potential source of the
larger polycyclic compounds, that is, coke, that can clog the
pores of the zeolite and deactivate it. Similar to the ring-
opening of the protonated furan discussed above, the positively
charged, alkylated furan derivative (Figure 8E) can ring open
and alkylate a third furan molecule. This process will result in
species that are too big to diffuse out of the pores of HZSM-5,
leading to the accumulation of carbon deposits and subsequent
deactivation of the catalyst.
Figure 10. Energies of the intermediates in Figure 8 are shown in the
solid red (branch I) and dotted blue (branch II) lines. Comparison
with Figures 6 and S4 (Supporting Information) shows that the
HZSM-5 catalyst energetically stabilizes the intermediates leading to
benzofuran. The solid black line shows the energies of intermediates in
the zeolite-catalyzed Diels−Alder route (Figure 9). The dashed black
line leads to the lowest energy configuration of the cycloadduct in the
zeolite.
We next consider the Diels−Alder route to benzofuran inside
the zeolite cavity. Figure 9 shows the geometry optimized
intermediates in branches I and II of the indirect route
respectively. The corresponding structures are shown in Figure
8. Comparison with Figures 6 and S4 (Supporting Information)
shows that the zeolite greatly stabilizes the intermediates on the
pathway leading to benzofuran with active intermediate
energies reduced from about 500 kJ/mol to roughly 50 kJ/
mol, a reduction of 90%. This intermediate energy of 50 kJ/mol
fits nicely into the “energy budget” determined by our
experimentally determined apparent activation energy of 72 ±
3
kJ/mol.
The substantial stabilization of the intermediates seen in the
zeolite, relative to the gas phase, deserves further comment. In
the gas-phase pathways, the proton donor (HZSM-5) and the
proton acceptor (furan) are assumed to be infinitely far from
each other to remove the catalytic effect while still having a zero
of energy that is common between gas-phase and zeolite-
catalyzed energy diagrams. In contrast, in the computed zeolite-
catalyzed pathways, the proton motion is over a distance of
only a few ångstroms. The approximately 500 kJ/mol gas-phase
energy scale thus reflects the difference in proton affinity (PA)
between the zeolite (with a relatively large PA) and furan (with
a relatively small PA). When the protonated furan is explicitly
inserted into the deprotonated zeolite in the zeolite-catalyzed
system, strong electrostatic attractions (responsible for the large
zeolite PA) become re-established, which serve to substantially
stabilize the system.
Figure 9. Geometry-optimized structures of the stable intermediates in
the zeolite-catalyzed, Diels−Alder pathway.
structures of the stable intermediates along the pathway
corresponding to the gas phase pathway in Figure 7. As with
the indirect process discussed above, the first step in the direct
route is the adsorption of two furans per zeolite acid site
(
Figure 9A). In the next step, the two furans undergo a Diels−
Alder cycloaddition. In its lowest energy configuration, the
cycloadduct is hydrogen bonded to the acid site at the furanic
O atom (Figure S6, Supporting Information). However, the
adduct cannot be stably protonated at the furanic O atom and
While the ring-opening and Diels−Alder pathways are
energetically similar in the gas phase, the situation for the
2
017
dx.doi.org/10.1021/cs4003904 | ACS Catal. 2013, 3, 2012−2019

ACS Catalysis
Research Article
reaction inside the zeolite is quite different. The solid black line
in Figure 10 shows the energies along the zeolite-catalyzed
Diels−Alder route. These intermediates are also stabilized
relative to the corresponding gas-phase structures, but are much
higher in energy than the ring-opening zeolite-catalyzed
pathway.
The geometries of the cycloadduct along the Diels−Alder
pathway (Figure 10, solid black line), and in the off-pathway
dead end configuration (Figure 10, dashed black line), offer a
clue to the high energies along the Diels−Alder route in the
zeolite. The hydrogen bond distance between the cycloadduct
O atom and the central O atom of the zeolite cluster is the
same (2.6 Å) in both configurations. However, the config-
uration along the Diels−Alder pathway is a “tighter fit” inside
the pore, with three other atoms of the guest molecule located
less than 2.9 Å from a zeolite framework atom; the
corresponding distance for the dead end configuration is 3.4
Å. By comparison, atomic van der Waals diameters are typically
in the range 3−4 Å.
These computational results, taken together with the
experiments reported herein, provide strong evidence that the
HZSM-5 catalyzed conversion of furan to benzofuran proceeds
by a ring-opening pathway involving open ring intermediates,
rather than the Diels−Alder route. The stable intermediates
along the Diels−Alder pathway are seen to be higher in energy
than the experimentally measured apparent activation energy.
Hence, we can rule out the Diels−Alder route to benzofuran in
HZSM-5, even without locating transition states along any of
the pathways.
While this analysis applies to the formation of benzofuran
from furan in HZSM-5, it is important to note that these results
do not necessarily apply to other potential Diels−Alder
reactions such as furan condensation with olefins. Specific
reaction mechanisms depend on the reactant species, catalyst
type, and reactant, intermediate, and product energetics. The
results of this study serve to highlight the importance of
examining individual key steps in biomass conversion to
determine which mechanisms are favorable for a given system.
reasonably well with the experimentally determined activation
energy of 72 kJ/mol.
These experimental and computational results suggest the
importance of the ring-opening mechanism for this key step in
1
4,29
making biofuels. However, recent results
suggest that
Brønsted acidity is relatively ineffective at catalyzing Diels−
Alder cycloaddition, while Lewis acidity is much more effective
at catalyzing such chemistry. In future work, we will apply the
computational methods described above to investigate whether
Lewis acidity may facilitate benzofuran production through
Diels−Alder pathways in HZSM-5 and other zeolites. We will
also pursue direct transition state calculations to provide
complete microkinetic pathways for these important processes
in biofuel production.
ASSOCIATED CONTENT
Supporting Information
Structures, energies, and atomic charges of key species in
benzofuran formation. This information is available free of
charge via the Internet at http://pubs.acs.org/.
■
*
S
AUTHOR INFORMATION
Corresponding Author
E-mail: auerbach@chem.umass.edu.
■
*
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We acknowledge generous funding from NSF (CBET-0932777
and EFRI-0937895) and from the Catalysis Center for Energy
Innovation, an Energy Frontier Research Center funded by the
U.S. Department of Energy, Office of Science, Office of Basic
Energy Sciences under award number DE-000SC0001004. We
greatly appreciate Srinivas Rangarajan’s extensive help with the
program RING and Figure 2. We also thank an unknown
reviewer for pointing out that the ring-opening mechanism can
potentially lead to coke.
REFERENCES
■
4
. SUMMARY AND CONCLUSIONS
(
1) Auerbach, S. M., Carrado, K. A., Dutta, P. K., Eds. Handbook of
We have performed kinetics experiments and quantum
calculations on embedded clusters to investigate the conversion
of furan to benzofuran in the zeolite catalyst HZSM-5, a key
step in the catalytic fast pyrolysis of biomass to biofuels. The
reaction was probed experimentally by placing the zeolite in
contact with solution-phase furan in heptane and detecting
benzofuran over the temperature range 270−300 °C, which
yielded an apparent activation energy of 72 ± 3 kJ/mol. The
reaction was modeled in gas and zeolite phases by computing
energies for the following two pathways: a Diels−Alder
mechanism assumed in interpretations of experimental data,
and a ring-opening pathway predicted by the software RING.
Quantum calculations for this reaction were performed using
the ONIOM embedded cluster approach. We computed
energies for reactants, products, and all intermediate steps;
locating transition states was not possible because of system
size and the ruggedness of the potential energy. The gas-phase
Diels−Alder pathway was found to pass through a high-energy
intermediate roughly 380 kJ/mol above the reactant energy;
this reduced to about 200 kJ/mol in HZSM-5. In contrast, the
gas-phase ring-opening mechanism passes through an inter-
mediate roughly 500 kJ/mol above the reactant energy; this
reduced by 90% in the zeolite to about 50 kJ/mol, which agrees
Zeolite Science and Technology; Marcel Dekker, Inc.: New York, 2003.
(2) Vispute, T. P.; Zhang, H.; Sanna, A.; Xiao, R.; Huber, G. W.
Science 2010, 330, 1222−1227.
(3) Cheng, Y.-T.; Huber, G. W. ACS Catal. 2011, 1, 611−628.
(
4) Cheng, Y.-T.; Jae, J.; Shi, J.; Fan, W.; Huber, G. W. Angew. Chem.,
Int. Ed. 2012, 51, 1387−1390.
5) Rangarajan, S.; Bhan, A.; Daoutidis, P. Ind. Eng. Chem. Res. 2010,
9, 10459−10470.
6) Vasiliu, M.; Guynn, K.; Dixon, D. A. J. Phys. Chem. C 2011, 115,
5686−15702.
7) Assary, R. S.; Redfern, P. C.; Hammond, J. R.; Greeley, J.; Curtiss,
(
4
(
1
(
L. A. J. Phys. Chem. B 2010, 114, 9002−9009.
(8) Assary, R. S.; Redfern, P. C.; Hammond, J. R.; Greeley, J.; Curtiss,
L. A. Chem. Phys. Lett. 2010, 497, 123−128.
(9) Assary, R. S.; Redfern, P. C.; Greeley, J.; Curtiss, L. A. J. Phys.
Chem. B 2011, 115, 4341−4349.
(
(
(
(
10) Assary, R. S.; Curtiss, L. A. Energy Fuels 2012, 26, 1344−1352.
11) Qian, X. J. Phys. Chem. A 2011, 115, 11740−11748.
12) Qian, X. Top. Catal. 2012, 55, 218−226.
13) Cheng, L.; Curtiss, L. A.; Assary, R. S.; Greeley, J.; Kerber, T.;
Sauer, J. J. Phys. Chem. C 2011, 115, 21785−21790.
14) Williams, C. L.; Chang, C.-C.; Do, P.; Nikbin, N.; Caratzoulas,
S.; Vlachos, D. G.; Lobo, R. F.; Fan, W.; Dauenhauer, P. J. ACS Catal.
2012, 2, 935−939.
(15) Rangarajan, S.; Bhan, A. Private communication.
(
2
018
dx.doi.org/10.1021/cs4003904 | ACS Catal. 2013, 3, 2012−2019

ACS Catalysis
Research Article
(
16) Vreven, T.; Morokuma, K. Annu. Rep. Comput. Chem. 2006, 2,
5−51.
17) Agarwal, V.; Huber, G. CW.; W., C. C., Jr.; Auerbach, S. M. J.
Catal. 2010, 269, 53−63.
18) Fermann, J. T.; Moniz, T.; Kiowski, O.; McIntire, T. J.;
Auerbach, S. M.; Vreven, T.; Frisch, M. J. J. Chem. Theory Comput.
005, 1, 1232−1239.
19) Olson, D. H.; Kokotailo, G. T.; Lawton, S. L.; Meier, W. M. J.
3
(
(
2
(
Phys. Chem. 1985, 85, 2238−2243.
(
(
20) Becke, A. D. J. Chem. Phys. 1993, 98, 5648−5652.
21) Stephens, P. J.; Devlin, F. J.; Chabalowski, C. F.; Frisch, M. J. J.
Phys. Chem. 1994, 98, 11623−11627.
22) Fermann, J. T.; Blanco, C.; Auerbach, S. J. Chem. Phys. 2000,
12, 6779−6786.
23) Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A.;
Skiff, W. M. J. Am. Chem. Soc. 1992, 114, 10024−10035.
24) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;
(
1
(
(
Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci,
B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H.
P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.;
Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima,
T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.; Montgomery, J. A., Jr.;
Peralta, J. E.; Ogliaro, F.; Bearpark, M.; Heyd, J. J.; Brothers, E.; Kudin,
K. N.; Staroverov, V. N.; Kobayashi, R.; Normand, J.; Raghavachari, K.;
Rendell, A.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.; Rega,
N.; Millam, J. M.; Klene, M.; Knox, J. E.; Cross, J. B.; Bakken, V.;
Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.;
Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Martin, R. L.;
Morokuma, K.; Zakrzewski, V. G.; Voth, G. A.; Salvador, P.;
Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, O.;
Foresman, J. B.; Ortiz, J. V.; Cioslowski, J.; Fox, D. J. Gaussian 09,
revision B.01; Gaussian, Inc.: Wallingford, CT, 2009.
(
1
(
25) Singh, U. C.; Kollman, P. A. J. Comput. Chem. 1984, 5, 129−
45.
26) Besler, B. H.; K., M. M., Jr.; Kollman, P. A. J. Comput. Chem.
1
(
(
990, 11, 431−439.
27) Haw, J. F. Phys. Chem. Chem. Phys. 2002, 4, 5431−5441.
28) Cheng, Y.−T.; Huber, G. W. Green Chem. 2012, 14, 3114−
3
125.
(
29) Nikbin, N.; Do, P. T.; Caratzoulas, S.; Lobo, R. F.; Dauenhauer,
P. J.; Vlachos, D. G. J. Catal. 2013, 297, 35−43.
2
019
dx.doi.org/10.1021/cs4003904 | ACS Catal. 2013, 3, 2012−2019