Kinetic Regime Change in the Tandem Dehydrative Aromatization of Furan Diels-Alder Products

Article abstract of DOI:10.1021/cs5020783

Renewable production of p-xylene from [4 + 2] Diels-Alder cycloaddition of 2,5-dimethylfuran (DMF) and ethylene with H-Y zeolite catalyst in n-heptane solvent is investigated. Experimental studies varying the solid acid catalyst concentration reveal two kinetic regimes for the p-xylene production rate: (i) a linear regime at low acid site concentrations with activation energy E_a = 10.8 kcal/mol and (ii) a catalyst-independent kinetic regime at high acid site concentrations with activation energy E_a = 20.1 kcal/mol. We carry out hybrid QM/MM calculations with a three-layer embedded cluster ONIOM model to compute the energetics along the main reaction pathway, and a microkinetic model is constructed for the interpretation of the experimental kinetic data. At high solid acid concentrations, p-xylene production is limited by the homogeneous Diels-Alder reaction, whereas at low acid concentrations, the overall rate is limited by the heterogeneously catalyzed dehydration of the Diels-Alder cycloadduct of DMF and ethylene because of an insufficient number of acid sites, despite the dehydration reaction requiring significantly less activation energy. A reduced kinetic model reveals that the production of p-xylene follows the general kinetics of tandem reactions in which the first step is uncatalyzed and the second step is heterogeneously catalyzed. Reaction orders and apparent activation energies of quantum mechanical and microkinetic simulations are in agreement with experimental values. (Chemical Presented).

Full text of DOI:10.1021/cs5020783

This is an open access article published under an ACS AuthorChoice License, which permits
copying and redistribution of the article or any adaptations for non-commercial purposes.
Research Article
pubs.acs.org/acscatalysis
Kinetic Regime Change in the Tandem Dehydrative Aromatization of
Furan Diels−Alder Products
†
,∥,⊥
†,∥,⊥
,†,∥
C. Luke Williams,§,∥,⊥ Sara K. Green,§,∥ Chun-Chih Chang,
§,∥
Ryan E. Patet,
Nima Nikbin,
§
,∥
,‡,∥
,†,∥
Wei Fan, Stavros Caratzoulas,* Paul J. Dauenhauer,* and Dionisios G. Vlachos*
†
Department of Chemical and Biomolecular Engineering, University of Delaware, Newark, Delaware 19716, United States
‡
§
∥
Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, Minnesota 55455, United States
Department of Chemical Engineering, University of Massachusetts, Amherst, Massachusetts 01003, United States
Catalysis Center for Energy Innovation (CCEI), University of Delaware, Newark, Delaware 19716, United States, www.efrc.udel.edu
*
S Supporting Information
ABSTRACT: Renewable production of p-xylene from [4 + 2] Diels−
Alder cycloaddition of 2,5-dimethylfuran (DMF) and ethylene with H−Y
zeolite catalyst in n-heptane solvent is investigated. Experimental studies
varying the solid acid catalyst concentration reveal two kinetic regimes for
the p-xylene production rate: (i) a linear regime at low acid site
concentrations with activation energy E = 10.8 kcal/mol and (ii) a catalyst-independent kinetic regime at high acid site
a
concentrations with activation energy E = 20.1 kcal/mol. We carry out hybrid QM/MM calculations with a three-layer
a
embedded cluster ONIOM model to compute the energetics along the main reaction pathway, and a microkinetic model is
constructed for the interpretation of the experimental kinetic data. At high solid acid concentrations, p-xylene production is
limited by the homogeneous Diels−Alder reaction, whereas at low acid concentrations, the overall rate is limited by the
heterogeneously catalyzed dehydration of the Diels−Alder cycloadduct of DMF and ethylene because of an insufficient number
of acid sites, despite the dehydration reaction requiring significantly less activation energy. A reduced kinetic model reveals that
the production of p-xylene follows the general kinetics of tandem reactions in which the first step is uncatalyzed and the second
step is heterogeneously catalyzed. Reaction orders and apparent activation energies of quantum mechanical and microkinetic
simulations are in agreement with experimental values.
KEYWORDS: p-xylene, 2,5-dimethylfuran (DMF), ethylene, Diels−Alder, dehydration, faujasite, microkinetic model
1
.0. INTRODUCTION
manufacture of plastic bottles, clothing, automobile compo-
1
5,16
nents, and many other products.
Efforts to maximize the
Research into the production of biorenewable fuels and
chemicals has rapidly expanded in the past decade in an effort
to reduce dependence on traditional, nonrenewable resour-
production of p-xylene from DMF have demonstrated 90% p-
17
xylene yield over H-BEA zeolite in heptane solvent. Heptane
has proven effective at greatly reducing competing side
reactions, such as alkylation of p-xylene with ethylene,
formation of polymers from aromatics and furans, and
hydrolysis of the furan by the water produced in the
1
,2
ces.
Coinciding with the increased research effort into
biomass-derived fuels and chemicals, shale gas production has
also greatly expanded in the United States. This newly tapped
energy source has the potential to provide vast quantities of C1
and C hydrocarbons but lacks the ability to provide larger
olefins and aromatic chemicals. Within this framework,
utilization of biomass-derived sugars has the potential to
provide the chemical industry with a more feedstock-diverse
source of C + and aromatic chemicals that is close to carbon-
13,18
dehydrative aromatization (Figure 1).
2
3
Despite a high yield of p-xylene formation from dimethylfur-
an, the role of the active catalytic site of H−Y zeolites and
associated kinetics remains to be understood. The rate at which
p-xylene is produced with a H−Y (Si/Al 2.6−40) zeolite
catalyst has been reported to be independent of the number of
4
4
−8
neutral.
19
available Brønsted acid sites, and thus, it was inferred that the
One major biorenewable chemical of interest is p-xylene,
which may be produced from biomass-derived 2,5-dimethylfur-
an (DMF) and ethylene. DMF can be produced from glucose
rate-limiting step should be uncatalyzed. However, in a different
study of the reaction of DMF and ethylene with WO −ZrO
x
2
9
,10
catalyst, the rate of formation of p-xylene was instead reported
via hydroxymethylfurfural,
from dehydration of biomass-derived ethanol.
4 + 2] cycloaddition of DMF and ethylene has been shown to
and ethylene can be obtained
20
1
1,12
as having linear dependence on the density of acid sites.
Diels−Alder
Quantum chemical calculations have shown that Brønsted acids
[
form an oxanorbornene cycloadduct in a single elementary step,
after which multistep dehydration forms p-xylene and
Received: December 23, 2014
Revised: February 19, 2015
1
3,14
water.
Renewable p-xylene has been proposed as a source
for the production of terephthalic acid, which is used in the
©
XXXX American Chemical Society
2367
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375

ACS Catalysis
Research Article
xylene, as shown in Figure 1) and more generally about the
kinetics of tandem reactions, which, in conjunction with
bifunctional catalysts can be utilized in the design of efficient,
one-pot synthetic processes. In this article, we present kinetic
data that show that the concentration of Brønsted acid sites
gives rise to two distinct kinetic regimes in the tandem scheme
of Diels−Alder cycloaddition and dehydrative aromatization to
p-xylene. Using hybrid quantum mechanics/molecular mechan-
ics (QM/MM) calculations and microkinetic modeling, we
show that in the first kinetic regime, in which the overall rate of
p-xylene formation grows linearly with the density of acid sites,
the kinetics are controlled by the catalyzed dehydration of the
Diels−Alder cycloadduct. In contrast, in the second regime, in
which the rate is independent of the density of acid sites, the
kinetic bottleneck is the uncatalyzed Diels−Alder reaction. At
high catalyst loadings, there are enough free active sites to
catalyze dehydration, and as a result, the homogeneous Diels−
Alder reaction becomes rate-limiting, consistent with the fact
that this is the most energetically demanding step of the
mechanism. The linear regime occurs at low catalyst loadings,
and because there are not enough active sites to catalyze it, the
dehydrative aromatization reaction becomes rate-controlling,
despite the fact that it requires a lower activation energy than
the cycloaddition reaction. Finally, we present a reduced kinetic
model to show that the two regimes are characteristic of
tandem schemes in which the first reaction proceeds
uncatalyzed while the second reaction does so catalyzed.
Figure 1. Reactions in the production of p-Xylene from DMF and
ethylene.
do not significantly catalyze the Diels−Alder reaction between
1
3,14,21
furans and ethylene.
Acidic protons have a higher affinity
for the furan than ethylene and, in particular, for the ring
carbons of the furanwhere the HOMO amplitude is the
largestwhich breaks the requisite orbital symmetry for [4 +
1
3,14,22
2.0. METHODS
2
] cycloaddition.
However, the second reaction,
dehydration of the Diels−Alder cycloadduct of DMF and
ethylene, cannot proceed uncatalyzed; Brønsted acids reduce
the kinetic energy barrier by as much as 45 kcal/mol and are
2.1. Experiment. Experiments examined the kinetics of the
reaction of DMF and ethylene with faujasite catalysts within
gas/liquid/solid reactors.
2.1.1. Reaction Setup. Reactions were performed in Parr
4560 series reactors equipped with 4848 controllers and gas
entrainment impellers to eliminate mass transport limitations.
Chemicals were used without further purification and consisted
of 2,5-dimethylfuran (Alfa Aesar 98+%), n-heptane (Alfa Aesar
1
3,14,21
essential for this step.
Variation in the dependence of the rate of p-xylene formation
on the type and concentration of Brønsted acids thus raises
intriguing questions about the kinetics of the reaction
(
cycloadduct formation and subsequent dehydration to p-
Figure 2. Kinetics of p-xylene production with H−Y zeolite catalyst: experiment. (A) Rate of p-xylene production at 200 and 250 °C relative to
strong acid site concentration. (B) Arrhenius plot at 1.3 mM H−Y acid site concentration. (C) p-Xylene reaction rate with respect to dimethylfuran
(
DMF; □) and ethylene (▲) at 1.3 mM H−Y acid site concentration. (D) Arrhenius plot at 1.3 mM H−Y acid site concentration. (E) p-Xylene
reaction rate with respect to dimethylfuran (DMF; □) and ethylene (▲) at 5.1 mM H−Y acid site concentration.
2
368
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375

ACS Catalysis
Research Article
Figure 3. Adsorption of dimethylfuran, p-xylene, oxanorbornene and hexanedione in faujasite. (A) ONIOM model (218T) used for adsorption
calculations where atoms represented by balls-and-sticks are in the high layer, tube frames are in the intermediate layer, and wireframes are in the low
layer (white, hydrogen; gray, carbon; red, oxygen; pink, aluminum; and green-gray, silicon). Adsorption complexes extracted from the ONIOM
model of (B) DMF, (C) p-xylene, (D) oxanorbornene, and (E) 2,5-hexanedione are also shown, where the distances from the active site proton to
the nearest atom of the adsorbate are provided in units of Å.
9
9+%), ethylene gas (Airgas UHP), and n-tridecane (Sigma-
with a G1513A autosampler, HP-Innowax column (to achieve
separation of o,m,p-xylene isomers), and a flame ionization
detector. Samples were collected under reaction conditions at
high temperature and pressure utilizing a double block
sampling system, which allows for samples to be taken without
opening the reaction vessel. Major species were identified by
matching retention times with pure standards. All reported data
exhibited carbon balance closure greater than or equal to 90%.
2.2. Computational. Quantum mechanics/molecular
mechanics (QM/MM) calculations have been used to calculate
the relevant adsorption energies and reaction barriers for the
chemistry of interest, using a three-layer, mechanically
embedded ONIOM cluster model. The computed energies
were then used to parametrize a microkinetic model.
Aldrich 98+%) as an internal standard. An H−Y faujasite
catalyst with a Si/Al ratio of 2.6 (Zeolyst CBV 600) was used
for all of the reactions and dried in a furnace at 200 °C prior to
use. Reaction procedure involved loading the vessel with DMF,
heptane, tridecane, and catalyst prior to heating and initiation
of the reaction by adding 200 psi of ethylene gas and
maintaining pressure throughout the course of the experiment.
̈
Additionally, the second Damkohler number was estimated to
be equal to 0.0002 ≪ 0.1, which indicates the absence of mass
transfer limitations between gaseous and solution phase
ethylene (Figure S1). The reaction rate was found to be
independent of particle size, which implies that intraparticle
diffusion is not rate limiting either.
2
.1.2. Reaction Orders and Activation Energies. Depend-
ence of reaction rate on catalyst loading was investigated using
.38 M DMF in n-heptane, with 0.082 M n-tridecane as an
2.2.1. QM/MM Calculations. The reaction pathway
calculations were performed on a HAlSi313O520 cluster model
(314 tetrahedral atoms) of H−Y. The model was cut out from
1
26
internal standard and 14 bar (200 psi) ethylene pressure. These
initial concentrations and pressures were used as the standard
reaction conditions. Experiments investigating the reaction
regimes of Figure 2 were performed at 200 and 250 °C by
varying the catalyst loading from 0.3 to 1.5 g to yield effective
Brønsted acid site concentrations between 1.1 and 5.1 mM.
Experiments investigating the parameters of reaction rate
expressions (i.e., reaction orders) were performed at low and
high catalyst loadings (1.3 and 5.1 mM) for both DMF and
ethylene at 200 °C. The concentration of DMF in DMF
reaction order experiments varied from 0.46 to 2.3 M in 0.46 M
increments. In ethylene reaction order experiments, the
ethylene pressure varied from 6.9 to 34.5 bar in 6.9 bar
increments (100−500 psi in 100 psi increments), yielding
ethylene concentrations of 0.036−0.18 M in 0.036 M
increments. Ethylene concentrations were estimated from
the periodic structure of pure faujasite zeolite, and the
dangling bonds were saturated with hydrogen atoms. Three-
27−29
layer ONIOM models
were employed to perform all
calculations within the zeolite for the purposes of capturing the
local active site and surrounding zeolite pore environments. For
reactions within the zeolite, the high layer, comprising the
active site and its environment (the active hexagonal ring and
the six surrounding 4-T rings, a total of 18 T atoms,
HAlSi O ), was treated quantum mechanically with the
17
24
25
M062X functional. The adsorbates, the Brønsted H atom,
the aluminum atom of the zeolite and the oxygen atoms in its
first coordination shell, viz. nearest neighbors, were modeled
with the 6-31G(d,p) basis set, while the rest of the high layer
zeolite atoms were modeled with the effective core potential
basis set LANL2DZ. The medium layer, the supercage in which
the reaction takes place, was modeled at the M062X/3-21G
level and kept frozen after optimization of the bare zeolite. For
the modeling of the low ONIOM layer, the rest of the zeolite
cluster, we employed the molecular mechanics force field
2
3
previous measurements. Experiments investigating the
activation energies of Figure 2 were performed at 0.3 and 1.5
g by varying the temperature from 200 to 250 °C in 10 °C
increments using the standard reaction conditions above.
30
UFF; the atoms of this layer were kept frozen in their
crystallographic positions at all times. The zeolite was
optimized in the absence of adsorbates and only the high
layer was allowed to relax in the presence of adsorbates.
2
.1.3. Reaction-Product Characterization. Characterization
of chemical components within the reaction mixture was
performed with an Agilent 6890 gas chromatograph equipped
2
369
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375

ACS Catalysis
Research Article
Figure 4. Free energy profile of p-xylene production from DMF and ethylene. Diels−Alder cycloaddition of DMF and ethylene followed by
dehydrative aromatization occurs in the absence (dashed line; homogeneous phase) and presence (solid line; heterogeneous phase) of H−Y zeolite.
Calculated intermediate and transition states account for the influence of the zeolite pore.
Binding energies are sensitive to the basis set of the quantum
mechanical region and susceptible to basis set superposition
error. Thus, for the adsorption of reactants, intermediates, and
products within the zeolite H−Y, we used a second ONIOM
model with a slightly smaller high layer which, however,
allowed us to use triple-ζ valence basis functions and additional
polarization. In this model consisting of 218 tetrahedral atoms
homogeneous Diels−Alder Gibbs free energy of reaction, and
the heterogeneous dehydration reaction barrier), which were
identified to be important through sensitivity analysis of the
microkinetic model. This methodological innovation departs
from the traditional single-parameter-change-at-a-time sensitiv-
ity analysis and is reminiscent of global sensitivity analysis
subject to constraints imposed by our experimental measure-
ments. Details of this approach can be found in the Supporting
Information, and the final set of parameters used to build the
microkinetic model can be found in Table S1. All adjustments
to the final parameters were found to be within the expected
computational error. A global sensitivity analysis was also
performed.
(
HAlSi217O352), the high layer consisted of the 14 T atoms in
the first two coordination spheres around the substituted
aluminum atom and was modeled at the M062X/6-311G-
(
2df,p) level. The intermediate layer included the 16 T atoms
in the third coordination sphere around the substituted
aluminum atom and was modeled at the M062X/3-21G level
and kept frozen. The remaining 188 T atoms in the fourth
through sixth coordination spheres were modeled with the
molecular mechanics force field UFF, with atoms being kept
frozen in their crystallographic positions. This model was
benchmarked against adsorption experimental data (not
shown). Solution phase calculations of isolated reactants and
3.0. RESULTS AND DISCUSSION
3
.1. Experimental Kinetics of Reaction of DMF and
Ethylene. The rate of p-xylene production exhibits complex
dependence on the relative concentrations of reactants (DMF
and ethylene) and strong acid sites in H−Y zeolite. As shown in
Figure 2A, the rate of p-xylene production at 200 and 250 °C
exhibits linear dependence on the concentration of strong
Brønsted acid sites at the given conditions and low acid site
concentrations (<3.0 mM). However, above ∼3.0 mM acid site
concentration, the rate of p-xylene formation becomes
independent of strong acid site concentration; for these
conditions, the addition of more catalyst has no impact on
24
products were performed with the SMD model at the
M062X/6-311G(2df,p) theory level. All calculations were
31
performed with the Gaussian 09 (Rev. A.2) program.
.2.2. Reaction Network. The reaction network consists of
2
the following reactions: homogeneous and heterogeneous
Diels−Alder cycloaddition of DMF and ethylene; homoge-
neous and heterogeneous cycloadduct dehydration comprising
of three elementary reaction steps (Figure 3); heterogeneous
+ Z
the overall formation rate of p-xylene (rp‑xylene ∝ [H ] , Z ∼ 0).
1
3,32
DMF hydrolysis;
reactants, products and stable intermediates.
.2.3. Reaction Rate Constants and Microkinetic Model-
and adsorption/desorption steps of
Differences in the two regimes of Figure 2 extend to the
measurable kinetic parameters. As depicted in Figure 2B,D,
Arrhenius plots generated at low acid site concentration (1.3
1
3,14
2
+
+
ing. The calculated free energies of activation and reaction were
used to parametrize a microkinetic model, as detailed in the
Supporting Information. Given the uncertainty in QM/MM
calculations and especially in the entropic contributions to the
free energies of reaction and activation, we introduced a design
of experiment (DOE) approach to investigate the sensitivity of
multiple, experimentally measured parameters (p-xylene
production rate, apparent activation energy, and reaction
orders) to multiple model parameters (the binding strength
of oxanorbornene, DMF, and hexanedione to the active site, the
mM [H ]) and high acid site concentration (5.1 mM [H ])
exhibit statistically different activation energies: low acid
concentration conditions lead to an activation energy of 10.8
± 2.1 kcal/mol, and high acid concentration conditions exhibit
an activation energy of 20.1 ± 1.2 kcal/mol. Moreover, the rate
of formation of p-xylene exhibits variable dependence on the
reactant concentrations between the two kinetic regimes, as
+
measured at 200 °C. At low acid concentration (1.3 mM [H ]),
the reaction rate expression for p-xylene formation has first-
X
order dependence on ethylene (rp‑xylene ∝[C H ] , X = 1.01 ±
2
4
2
370
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375

ACS Catalysis
Research Article
Figure 5. Model comparisons with experimentally available data. (A) p-Xylene initial production rate at Bronsted acid concentrations at 250 °C. p-
Xylene production rates are calculated as the average rate over the first 50 min of reaction time at the specified temperature and catalyst loading. (B)
Solution-phase composition experiments (points) and simulations (lines). Conditions: 2.0 g of H−Y (Si/Al = 2.6) for an effective acid site
concentration of 6.8 mM, 1.38 M DMF, at 250 °C. MKM results are shown as lines, and experimental data are shown as points. The values in red are
the solution phase concentration of 2,5-hexanedione from the model and the sum of all side products from the experimental data.
0
.07) and half-order dependence on dimethylfuran (rp‑xylene
∝
dimethylfuran and ethylene remains about the same, with a
barrier of 43.4 kcal/mol; however, there is a significant decrease
in the largest dehydration reaction barrier to 14.4 kcal/mol.
From the depicted reaction barriers of Figure 4, it is clear that
H−Y does not catalyze the Diels−Alder reaction between DMF
and ethylene. Dehydration of the oxanorbornene cycloadduct
is, however, orders of magnitude faster in the Brønsted acidic
zeolite H−Y. Fast proton transfer from the zeolite active site to
the bridging oxygen atom initiates the reaction. The
oxanorbornene oxygen bridge, C2−O, breaks with activation
energy of only 14.4 kcal/mol, which is 45.7 kcal/mol lower
than the uncatalyzed reaction barrier. The binding of the
proton to the bridging oxygen appears to stabilize the transition
state for the C2−O cleavage, making this reaction much more
favorable. In the acid catalyzed case, the slowest step along this
pathway is the uncatalyzed Diels−Alder cycloaddition. The
furan ring of DMF may open hydrolytically upon β-C
protonation and subsequent water addition at the adjacent α-
Y
[
DMF] , Y = 0.49 ± 0.08), determined from the data of Figure
+
2
. However, at high acid concentrations (5.1 mM [H ]), the
rate of p-xylene formation exhibits closer to first-order
dependence in both ethylene and DMF (r
∝
p‑xylene
X
Y
[
C H ] [DMF] ; X = 0.90 ± 0.13, Y = 0.79 ± 0.07), as
2 4
determined from the data of Figure 2E.
.2. QM/MM Calculations of Relevant Energies and
3
Barriers. Calculation of the adsorption of reactants,
intermediates, products, and transition states within the zeolite
pore environment provides mechanistic understanding of the
chemistry at the strong acid site within H−Y. The calculated
adsorption strengths from the HAlSi
O
17 352
ONIOM model
2
can be found in Table S1 and Figure 3. Calculations show that
among the reactants and products, DMF (17.9 kcal/mol) and
p-xylene (16.6 kcal/mol) bind more strongly to the active site
than either ethylene (9.0 kcal/mol) or water (11.1 kcal/mol).
Examining the DMF and p-xylene adsorption geometries on the
catalyst surface (Figure 3B and 3C), we see that both molecules
coordinate to the active site with their aromatic carbons atoms.
The oxanorbornene intermediate formed by the Diels−Alder
cycloaddition reaction of DMF and ethylene strongly adsorbs
to the surface of the zeolite (25.8 kcal/mol), with the bridging
oxygen atom located closest to the active site (Figure 3D). 2,5-
Hexanedione, formed by the hydrolysis of DMF with water,
also binds strongly (25.9 kcal/mol), with an oxygen from one
of the keto groups being closest to the active site (Figure 3E).
Tandem Diels−Alder reaction and subsequent dehydrative
aromatization pathways within Brønsted acidic zeolites have
32
C.
3.3. Microkinetic Model Results and Discussion. By the
calculated energies of Figure 4, one would conclude that Diels−
Alder cycloaddition should be the rate-limiting reaction for all
acid-catalyzed reactions, a result differing from the experimental
results of Figure 2, which indicates two distinct kinetic regimes.
To probe the kinetics of the acid-catalyzed system, the QM/
MM calculations for adsorption energies and reaction barriers
were used to develop a microkinetic model (MKM) by the
details provided in the Methods section and Table S1.
Figure 5A shows the predicted rate of p-xylene production as
a function of acid site concentration from the MKM. The
MKM (green line) agrees well with experimental data at 200
and 250 °C. The two kinetic regimes were captured by the
MKM: at low acid site concentrations, the p-xylene production
rate varies with the acid site concentration, whereas at high acid
site concentrations, the p-xylene production rate is independent
of acid site concentration. In addition, a global sensitivity
analysis was performed on the MKM model such that the
adsorption strength of oxanorbornene and DMF to the active
site, the homogeneous Diels−Alder Gibbs free energy of
reaction, and the heterogeneous dehydration reaction barrier
1
3,14,21
been previously discussed;
here, we provide comprehen-
sive energetics of all elementary steps, obtained from ONIOM
calculations on the HAlSi313O520 cluster model (Figure 4), and
detailed understanding of the dominant reaction pathways via a
microkinetic model (MKM). Energies for the uncatalyzed
reaction mechanism in Figure 4 show that the Diels−Alder
reaction has a barrier of 39.5 kcal/mol, and the largest barrier of
the subsequent dehydration of the oxanorbornene intermediate
is 60.1 kcal/mol. In the calculations from the ONIOM model
accounting for catalysis at the H−Y zeolite active site and
surrounding pore, the Diels−Alder reaction barrier of
2
371
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375

ACS Catalysis
Research Article
Figure 6. Abridged reaction network and simulation reaction path fluxes. Diels−Alder and dehydration reactions in homogeneous (green) and
heterogeneous (blue) phases and hydrolysis of DMF. Inactive pathways are represented in gray. Fluxes calculated by the MKM model are provided
as percentages of the total flux of DMF initially fed into the reactor averaged over the first 50 min of reaction at 250 °C at Brønsted acid site
concentrations of 0.1 mM (low loading, blue) and 5.1 mM (high loading, red).
provided a confidence interval that indicated that the
parametrized model was robust in describing the p-xylene
production rate. These energies were independently perturbed
using a normal distribution with a standard deviation of 1.5
kcal/mol. The 95% confidence interval, depicted as a green
region in Figure 5A, encompassed the data points of p-xylene
production rate at 250 °C.
both kinetic regimes have been verified by performing
sensitivity analysis of the MKM shown in Figure 7. At low
Figure 5B depicts the solution phase concentrations
predicted by the MKM with reaction time. Consumption of
DMF and formation of p-xylene predicted by the model are in
agreement with experiments. A consumption rate of DMF
slower than expected could be because the model does not
include side reactions other than DMF hydrolysis to
hexanedione. Further Diels−Alder reaction between DMF
and the cycloadduct or unproductive pathways through the
dehydrative aromatization intermediates can also be responsible
for loss of DMF, but they are not considered in the current
1
8
model. Nevertheless, the close agreement between the
experimental and model concentration profiles that 2,5-
hexanedione makes up the majority of side products present
in the reaction. The 2,5-hexanedione that is formed is able to
reversibly convert back into DMF to be further converted into
p-xylene, and therefore, as the conversion of DMF increases,
the thermodynamics of the system reduce the presence of 2,5-
hexanedione product found in solution and help drive the
system to high yields of p-xylene.
Figure 7. Sensitivity analysis of microkinetic model. The normalized
sensitivity coefficient is defined as ln(rpX,Perturbed − rpX,Unperturbed)/
ln(ki,Perturbed − k
). Reaction rate parameters were perturbed
i,Unperturbed
by +1% (blue) and −1% (red) at 250 °C for each of the catalyst
loadings shown. Dehydration step 1 represents the C−O bond
cleavage and dehydration step 2 represents the first proton transfer
from the three-step mechanism seen in Figure 4.
+
Analysis of the MKM (Figure 6) provides detailed
understanding of the active reaction pathways. For both the
acid site concentrations (0.1 mM [H ]), the normalized
sensitivity coefficients for the heterogeneous dehydration
reaction steps (∼0.45) are much greater than that of the
homogeneous Diels−Alder reaction (0.05), indicating that the
heterogeneous dehydration reaction is rate-limiting. In contrast,
+
+
high (5.1 mM [H ], red) and low (0.1 mM [H ], blue) catalyst
loading scenarios, the reaction pathway to p-xylene is the same.
The homogeneous Diels−Alder reaction accounts for 65% of
the flux of DMF in solution in a reversible reaction, and the
remaining DMF in solution adsorbs on the catalyst surface and
undergoes hydrolysis with water to 2,5-hexanedione. It can be
seen that the adsorbed DMF does not undergo heterogeneous
Diels−Alder reaction, but rather, a hydrolysis reaction. Of the
oxanorbornene formed by homogeneous Diels−Alder reaction,
it can be seen that none of it is homogeneously dehydrated into
p-xylene and water, as a result of the prohibitively high reaction
barrier. Rather, all of the oxanorbornene adsorbs on the catalyst
surface and undergoes irreversible, heterogeneous dehydration.
With the reaction path analysis revealing the major route for
the reaction, uncatalyzed Diels−Alder reaction followed by
Brønsted acid catalyzed dehydration, we can examine differ-
ences between the two kinetic regimes. Rate-limiting steps for
+
at high acid site concentrations (5.1 mM [H ]), there is a clear
shift in the normalized sensitivity coefficients (0.62 for the
homogeneous Diels−Alder reaction compared with ∼0.15 for
the heterogeneous dehydration reaction), indicating that the
homogeneous Diels−Alder reaction is rate-limiting.
Using the MKM, we have determined apparent activation
energies and rate orders in the two kinetic regimes, namely, at
+
0.1, 1.3, and 5.1 mM [H ], as shown in Table 1 and Figure 8A.
We should note that although the 1.3 mM active site
concentration clearly falls within the experimental linear regime
(Figure 5A), the same cannot be said of the model. Thus, in the
case of the MKM, we have also considered a catalyst loading of
0.1 mM to ensure that we were looking at values well within the
model’s linear regime and not in a transitional state between
2
372
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375

ACS Catalysis
Research Article
a
Table 1. Kinetic Parameters for Cycloaddition of Dimethylfuran and Ethylene at Different Brønsted Acid Site Concentrations
experiments
simulation
low catalyst loading
high catalyst loading
(5.1 mm)
low catalyst loading
(0.1 mm)
low catalyst loading
(1.3 mm)
high catalyst loading
(5.1 mm)
kinetic parameter
(1.3 mm)
DMF reaction order
0.49 ± 0.08
1.01 ± 0.07
10.8 ± 2.1
0.79 ± 0.07
0.90 ± 0.13
20.1 ± 1.2
0.30 ± 0.02
0.99 ± 0.00
18.4 ± 0.2
0.59 ± 0.03
0.98 ± 0.00
21.3 ± 0.3
0.73 ± 0.02
0.98 ± 0.00
22.7 ± 0.3
ethylene reaction order
apparent activation energy
(kcal/mol)
a
Error bars in simulations are the result of uncertainty quantification.
Figure 8. Microkinetic model output. (A) Reaction rate orders as a function of acid site concentration. MKM derived parameters are shown as lines,
whereas experimental data is shown as points. (B) Surface coverage of H−Y active sites. Surface coverage of strong acid sites vary as a function of
conversion at 250 °C at 5.1 mM of acid site concentration. Coverages do not change significantly when moving from low to high catalyst loading
−
7
regimes. Oxanorbornene coverage (not shown) is less than 10 % at all conversions.
Table 2. Simulated Reactions and Kinetic Parameters
a
b
c
ΔG‡
d
ΔG‡
d
No.
Reaction
ki
‑
k−i
250 °C,i
250 °C,−i
3.5 × 10 M−1 s−1
8.0 × 10− s−1
6.9 × 10− s−1
4
39.5
60.1
48.3
47.9
43.4
14.4
14.4
12.1
10.8
18.2
1.2 × 10 s
3
−1
23.8
58.6
54.4
89.3
46.8
43.1
12.2
28.4
43.1
21.3
1
DMF + E ⇌ CA
CA → DehyInt1
13
3.5 × 10−12 s−1
2.0 × 10−10 s−1
4.9 × 10−25 s−1
3.0 × 10−7 s−1
8
DehyInt1 → DehyInt2
DehyInt2 → pX + W
1.0 × 10− s−1
7
8.1 × 10−6M−1 s−1
DMF* + E ⇌ CA
CA* → pX + W*
7
−1
‑5
‑1
2
3
1.1 × 10 s
1.1 × 10 M− s
7
−1
−1
−1
7
−1
CA* → DehyInt1*
1.1 × 10 s
8.9 × 10 s
7
1
−1
DehyInt1* → DehyInt2*
DehyInt2* → pX + W*
9.4 × 10 s
1.5 × 10 s
8
1.1 × 10−5 M−1 s−1
3.4 × 10 s
2.7 × 10 M s−1
5
−1
1.3 × 10 s
4
−1
DMF* + W ⇌ HDI*
a
b
d
No.
4
Adsorption/Desorption
Ki
ΔGads,250 °C
CA + * ⇌ CA*
W + * ⇌ W*
5.84
0.08
−1.8
2.6
5
6
DMF + * ⇌ DMF*
HDI + * ⇌ HDI*
E + * ⇌ E*
35.1
793
−3.7
−6.9
3.0
7
0.05
0.17
0.003
pX + * ⇌ pX*
1.9
Hept + * ⇌ Hept*
6.0
a
Elementary steps with a number are used to derive the reduced model in which the numbers correspond to the kinetic parameters found in eqs
−3. Elementary steps without a number are used in the full microkinetic model. Molecules are labeled as follows: DMF = dimethylfuran, E =
b
1
ethylene, CA = cycloadduct of DMF and ethylene, pX = p-xylene, W = water, HDI = 2,5-hexanedione, Hept = Heptane, * = vacant strong acid H−Y
c
d
site. Subscript “i” implies forward reaction; “−i” implies reverse reaction. Reaction barriers and adsorption energies are tabulated in units of kcal/
mol.
the two kinetic regimes. The differences in the apparent
activation energies and reaction orders between the two kinetic
regimes are predicted by the MKM, with good agreement with
experiment. In the linear kinetic regime, the apparent activation
2
373
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375

ACS Catalysis
Research Article
energies (10.8 kcal/mol from experiment and 18.4 kcal/mol
from the MKM) are close to the dehydration reaction barrier of
first term in the denominator is dominant, and the p-xylene
production rate instead reduces to
1
4.5 kcal/mol calculated from the QM/MM calculations. The
4 kcal/mol difference between the QM/MM calculation and
⎛
rpX = ⎜
⎝
⎞
k K K
∼
2
1
4
+
⎟[E][H ]
the microkinetic model is due to contributions from the
dehydration and hydrolysis reactions included in the model.
The DMF reaction order was 0.49 ± 0.08 from experiment and
K (1 + K [W])
⎠
(3)
6
3
The rate has a linear dependence on the Brønsted acid active
site concentration, consistent with the findings in the linear
regime. In addition, the concentration of DMF appears in both
the numerator and the denominator of the rate, and the
concentration of ethylene appears only in the numerator. The
reduced model supports the variable reaction-order kinetics
with respect to DMF (from zero to one) and first-order kinetics
with respect to ethylene observed experimentally. The reduced
model also supports the experimental trends (Figure 5) and
indicates that two regimes is a signature of tandem reactions,
specifically of a noncatalyzed reaction followed by a
heterogeneously catalyzed second reaction.
0
.3 from the microkinetic model. At higher catalyst loadings,
the apparent reaction barrier increases (20.1 kcal/mol from
experiment and 22.7 kcal/mol from the MKM) and is similar to
the Diels−Alder reaction barrier obtained from the QM/MM
calculations. In addition, the reaction orders for both DMF and
ethylene are close to 1, in support of the homogeneous Diels−
Alder reaction mechanism.
The MKM also provides insight into the coverages on the
acid sites as a function of conversion (Figure 8B). At early
times, the high concentration of DMF in solution causes it to
occupy the surface of the zeolite. As DMF is converted and
forms p-xylene, water, and 2,5-hexanedione, the 2,5-hexane-
dione that is formed can be seen to almost completely cover the
surface of the zeolite. The strong adsorption strength of 2,5-
hexanedione allows it to dominate the surface of the zeolite,
despite its lower concentrations in solutions relative to DMF
and p-xylene (Figure 5B). Near complete DMF conversion,
however, we see that 2,5-hexanedione is able to reversibly
convert back into DMF, driven by the thermodynamics of the
system, causing the 2,5-hexanedione to desorb. The strong
adsorption strength of the oxanorbornene intermediate allows
it to compete with 2,5-hexanedione for active sites, enabling the
heterogeneous dehydration pathway. The rapid dehydration of
the adsorbed oxanorbornene, however, causes its concen-
trations on the surface to be very small at all conversions.
5.0. CONCLUSIONS
We have experimentally and computationally studied the
kinetics of formation of p-xylene by dehydrative aromatization
of the Diels−Alder product between 2,5-dimethylfuran and
ethylene over H−Y faujasite. This reaction is an archetype of
tandem reactions. Cycloaddition proceeds uncatalyzed in
solution with a moderate reaction barrier, and dehydration
takes place on the active Brønsted acid sites of the zeolite with a
low kinetic energy barrier. We have shown that two different
kinetic regimes exist dependent upon the reactor loading in H−
Y faujasite. At high catalyst loadings, there are enough free
active sites to catalyze the dehydration reaction, and as a result,
the homogeneous Diels−Alder reaction becomes rate-limiting,
consistent with the fact that this is the most energetically
demanding step of the mechanism. As the catalyst loading
decreases past a critical transition value, the kinetics enters the
linear regime where, as a result of an insufficient number of
active sites, the dehydrative aromatization becomes rate-
limiting, despite the fact that it requires less activation than
the cycloaddition reaction.
3
.3. Insights into the Mechanism. From insights
obtained from reaction path and model sensitivity analysis, a
simplified kinetic model for the reaction of DMF and ethylene
to produce p-xylene consists of the reactions identified in Table
2. By this simplified reaction mechanism, the rate of p-xylene
production can be solved as
+
k K K [DMF][E][H ]
2
1
4
As we noted in the Introduction, in a study of the same
reaction by Wang et al., the rate of formation of p-xylene was
reported as having a dependence on the density of acid sites.
This was explained by an alternative mechanism for the
formation of the oxanorbornene derivative that did not involve
Diels−Alder cycloaddition. This finding was at odds with the
kinetic studies of Williams et al., who reported that the rate of
rpX
=
k2
+
K (1 + K [W])[DMF] +
K [H ]
6
3
( )
4
k
(1)
−1
where k and K are the reaction rate and equilibrium constants
i
i
+
20
for reaction i, [i] is the concentration of component i, and [H ]
is the total Brønsted acid active site concentration (for details
see the Supporting Information). The functional form of the
rate law reveals the presence of two regimes, and its
dependence on the concentration of water indicates product
inhibition, as water reacts with adsorbed DMF to form 2,5-
hexanedione. The reduced model is in good agreement with the
microkinetic model; the simplified model is presented as the
blue line in Figure 5.
p-xylene production over H−Y was independent of the density
19
of active sites. We believe that the two kinetic regimes
reported and analyzed here can reconcile these two seemingly
opposing views. In Williams et al., the reaction was carried out
at 300 °C, with an effective catalyst loading range of 1.7−3.2
mM, which is in the flat, cycloaddition-limited regime and, thus,
independent of catalyst loading. On the other hand, in Wang et
al., the experiments were performed at 250 °C and significantly
lower loadings, ∼0.2 mM of effective acid site concentration,
which lies in the dehydration-limited regime and which explains
the reported dependence of the rate on the density of active
sites.
At high Brønsted acid active site concentrations, the [H+]
term in the denominator is dominant and the p-xylene
production rate reduces to,
rpX = k [DMF][E]
1
(2)
Although available acid sites and site acidity may influence
the rate of change of the rate of the reaction with acid
concentration in the dehydration regime (the slope of the curve
in Figure 5A), the plateau should remain unaffected because it
is determined by the uncatalyzed cycloaddition reaction rate.
The reaction rate depends on the homogeneous Diels−Alder
reaction and, therefore, is first-order with respect to DMF and
ethylene, in agreement with experiments (Table 1). At
sufficiently low Brønsted acid active site concentrations, the
2
374
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375

ACS Catalysis
Research Article
Thus, the cycloaddition-limited regime should be “universal” to
all (Brønsted acid) catalysts. Given that the maximum rate is
controlled by the cycloaddition reaction, a logical next step to
optimize the process is to develop methods to accelerate the
Diels−Alder reaction.
(13) Nikbin, N.; Do, P. T.; Caratzoulas, S.; Lobo, R. F.; Dauenhauer,
P. J.; Vlachos, D. G. J. Catal. 2013, 297, 35−43.
(14) Nikbin, N.; Feng, S.; Caratzoulas, S.; Vlachos, D. G. J. Phys.
Chem. C 2014, 118, 24415−24424.
15) Tomas, R. A. F.; Bordado, J. C. M.; Gomes, J. F. P. Chem. Rev.
013, 113, 7421−7469.
16) Pacheco, J. J.; Davis, M. E. Proc. Natl. Acad. Sci. U.S.A. 2014,
111, 8363−8367.
17) Chang, C.-C.; Green, S. K.; Williams, C. L.; Dauenhauer, P. J.;
Fan, W. Green Chem. 2014, 16, 585−588.
18) Do, P. T. M.; McAtee, J. R.; Watson, D. A.; Lobo, R. F. ACS
Catal. 2013, 3, 41−46.
19) Williams, C. L.; Chang, C.-C.; Do, P.; Nikbin, N.; Caratzoulas,
(
́
2
(
ASSOCIATED CONTENT
Supporting Information
The following file is available free of charge on the ACS
■
S
(
*
(
Publications website at DOI: 10.1021/cs5020783.
Investigation of gas-to-liquid mass transfer limitations,
reaction network and energetics considered in the
microkinetic model, development of rate expressions,
design of experiment (DOE) to access model un-
certainty, reduced model development, catalyst prepara-
tion and characterization, and additional references
(
S.; Vlachos, D. G.; Lobo, R. F.; Fan, W.; Dauenhauer, P. J. ACS Catal.
2012, 2, 935−939.
(20) Wang, D.; Osmundsen, C. M.; Taarning, E.; Dumesic, J. A.
ChemCatChem 2013, 5, 2044−2050.
21) Li, Y.-P.; Head-Gordon, M.; Bell, A. T. J. Phys. Chem. C 2014,
18, 22090−22095.
22) Woodward, R. B.; Hoffmann, R. Angew. Chem., Int. Ed. 1969, 8,
81−853.
(23) Zhuze, T. P.; Zhurba, A. S. Bull. Acad. Sci. USSR, Div. Chem. Sci.
Engl. Transl.) 1960, 9, 335−337.
24) Marenich, A. V.; Cramer, C. J.; Truhlar, D. G. J. Phys. Chem. B
009, 113, 6378−6396.
25) Zhao, Y.; Truhlar, D. G. Theor. Chem. Acc. 20072008, 120,
215−241.
26) Baerlocher, C.; McCusker, L. B. Database of Zeolite Structures.
(
1
(
7
(
PDF)
AUTHOR INFORMATION
Corresponding Authors
E-mail: cstavros@udel.edu.
E-mail: hauer@umn.edu.
E-mail: vlachos@udel.edu.
■
(
(
2
(
*
*
*
(
Notes
http://www.iza-structure.org/databases (accessed Feb 11, 2015).
(27) Svensson, M.; Humbel, S.; Froese, R. D. J.; Matsubara, T.;
Sieber, S.; Morokuma, K. J. Phys. Chem. 1996, 100, 19357−19363.
The authors declare no competing financial interest.
⊥
R.E.P., N.N., and C.L.W. contributed equally.
(
28) Dapprich, S.; Komar
́
omi, I.; Byun, K. S.; Morokuma, K.; Frisch,
ACKNOWLEDGMENTS
M. J. J. Mol. Struct.: THEOCHEM 1999, 461, 1−21.
■
̈
(
29) Vreven, T.; Morokuma, K.; Farkas, O.; Schlegel, H. B.; Frisch,
M. J. J. Comput. Chem. 2003, 24, 760−769.
30) Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A.;
Skiff, W. M. J. Am. Chem. Soc. 1992, 114, 10024−10035.
31) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;
We acknowledge support 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 No. DE-SC0001004. This
research used resources of the National Energy Research
Scientific Computing Center, a DOE Office of Science User
Facility supported by the Office of Science of the U.S.
Department of Energy under Contract No. DE-AC02-
(
́
(
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.;
05CH11231.
REFERENCES
■
(
1) Vlachos, D. G.; Chen, J. G.; Gorte, R. J.; Huber, G. W.; Tsapatsis,
M. Catal. Lett. 2010, 140, 77−84.
2) Binder, J. B.; Raines, R. T. J. Am. Chem. Soc. 2009, 131, 1979−
985.
3) Bruijnincx, P. C. A.; Weckhuysen, B. M. Angew. Chem., Int. Ed.
(
1
(
̈
Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, O.;
2
(
(
013, 52, 11980−11987.
Foresman, J. B.; Ortiz, J. V.; Cioslowski, J.; Fox, D. J. Gaussian 09
4) Dauenhauer, P. J.; Huber, G. W. Green Chem. 2014, 16, 382−383.
(
Rev. A.2); Gaussian Inc.: Wallingford, CT, 2009.
32) Nikbin, N.; Caratzoulas, S.; Vlachos, D. G. ChemSusChem 2013,
, 2066−2068.
5) Ragauskas, A. J.; Williams, C. K.; Davison, B. H.; Britovsek, G.;
(
6
Cairney, J.; Eckert, C. A.; Frederick, W. J., Jr.; Hallett, J. P.; Leak, D. J.;
Liotta, C. L.; Mielenz, J. R.; Murphy, R.; Templer, R.; Tschaplinski, T.
Science 2006, 311, 484−489.
(
(
6) Kwart, H.; Burchuk, I. J. Am. Chem. Soc. 1952, 74, 3094−3097.
7) Wenkert, E.; Moeller, P. D. R.; Piettre, S. R. J. Am. Chem. Soc.
1
988, 110, 7188−7194.
8) Chen, C.-H.; Rao, P. D.; Liao, C.-C. J. Am. Chem. Soc. 1998, 120,
3254−13255.
9) Roman-Leshkov, Y.; Barrett, C. J.; Liu, Z. Y.; Dumesic, J. A.
Nature 2007, 447, 982−986.
10) Chidambaram, M.; Bell, A. T. Green Chem. 2010, 12, 1253−
262.
11) van Haveren, J.; Scott, E. L.; Sanders, J. Biofuels, Bioprod.
Biorefin. 2008, 2, 41−57.
12) Phillips, C. B.; Datta, R. Ind. Eng. Chem. Res. 1997, 36, 4466−
475.
(
1
(
́
(
1
(
(
4
2
375
DOI: 10.1021/cs5020783
ACS Catal. 2015, 5, 2367−2375