Acylation of methylfuran with Br?nsted and Lewis acid zeolites

Article abstract of DOI:10.1016/j.apcata.2018.06.005

The acylation of methylfuran has been investigated using Br?nsted and Lewis acid zeolite catalysts. The highest reaction rate for acylation on a per gram basis is found on zeolite Beta with high aluminum content (Si/Al = 23) and the highest turnover frequency on a per metal site basis is found on zeolite Beta with low aluminum content (Si/Al = 138). Among Lewis acid zeolites, [Sn]-Beta shows higher turnover frequency than [Hf]-, [Zr]- or [Ti]-Beta. Similar apparent activation energies were found for [Al]-Beta with different Si/Al ratios and a lower apparent activation energy was found for [Sn]-Beta. Electronic structure calculations reveal that on both [Al]- and [Sn]-Beta the most favorable pathway follows the classic addition-elimination aromatic electrophilic substitution mechanism. The calculations also reveal that, on both [Al]- and [Sn]-Beta, the rate of methylfuran acylation is controlled by the dissociation of the C–O–C linkage of the anhydride while hydrogen elimination is the rate-determining step in the acylation of furan. The latter is in complete agreement with measured primary kinetic isotope effects. One remarkable and unexpected finding from our calculations is that the most favorable catalytic pathway in [Sn]-Beta involves Br?nsted acid catalysis by the silanol group of the hydrolyzed “open” site and not Lewis acid catalysis by the Sn metal center.

Full text of DOI:10.1016/j.apcata.2018.06.005

Accepted Manuscript
Title: Acylation Of Methylfuran With Brønsted And Lewis
Acid Zeolites
Authors: Maura Koehle, Zhiqiang Zhang, Konstantinos A.
Goulas, Stavros Caratzoulas, Dionisios G. Vlachos, Raul F.
Lobo
PII:
S0926-860X(18)30274-6
DOI:
Reference:
https://doi.org/10.1016/j.apcata.2018.06.005
APCATA 16693
To appear in:
Applied Catalysis A: General
Received date:
Revised date:
Accepted date:
23-2-2018
31-5-2018
3-6-2018
Please cite this article as: Koehle M, Zhang Z, Goulas KA, Caratzoulas S, Vlachos DG,
Lobo RF, Acylation Of Methylfuran With Brønsted And Lewis Acid Zeolites, Applied
Catalysis A, General (2018), https://doi.org/10.1016/j.apcata.2018.06.005
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

ACYLATION OF METHYLFURAN WITH BRØNSTED AND LEWIS ACID ZEOLITES
Maura Koehlea,b,‡, Zhiqiang Zhanga,c,‡, Konstantinos A. Goulas , Stavros Caratzoulas , Dionisios
a
a
G. Vlachosa,b and Raul F. Loboa,b,*
a
Catalysis Center for Energy Innovation, University of Delaware, Newark, DE 19716
Department of Chemical and Biomolecular Engineering, University of Delaware, Newark, DE
b
1
c
9716
Department of Physics and Astronomy, University of Delaware, Newark, DE 19716
‡
Joint First Authors
*
Email: lobo@udel.edu
Graphical Abstract

Highlights




Lewis and Bronsted acidic zeolites are compared for acylation of methylfuran and furan
Bronsted acid zeolite beta shows the highest specific rates at low Si/Al=12.5
[Sn]-zeolite beta shows the lowest activation energy
Acylation rate is controlled by the dissociation of the C-O-C linkage of the anhydride for
methylfuran

[Sn]-beta catalytic pathway involves Brønsted acid catalysis by the silanol group of the
hydrolyzed “open” site
ABSTRACT
The acylation of methylfuran has been investigated using Brønsted and Lewis acid zeolite
catalysts. The highest reaction rate for acylation on a per gram basis is found on zeolite Beta with
high aluminum content (Si/Al=23) and the highest turnover frequency on a per metal site basis is
found on zeolite Beta with low aluminum content (Si/Al=138). Among Lewis acid zeolites, [Sn]-
Beta shows higher turnover frequency than [Hf]-, [Zr]- or [Ti]-Beta. Similar apparent activation
energies were found for [Al]-Beta with different Si/Al ratios and a lower apparent activation
energy was found for [Sn]-Beta. Electronic structure calculations reveal that on both [Al]- and
[
Sn]-Beta the most favorable pathway follows the classic addition-elimination aromatic
electrophilic substitution mechanism. The calculations also reveal that, on both [Al]- and [Sn]-
Beta, the rate of methylfuran acylation is controlled by the dissociation of the C-O-C linkage of
the anhydride while hydrogen elimination is the rate-determining step in the acylation of furan.
The latter is in complete agreement with measured primary kinetic isotope effects. One remarkable
and unexpected finding from our calculations is that the most favorable catalytic pathway in [Sn]-
Beta involves Brønsted acid catalysis by the silanol group of the hydrolyzed “open” site and not
Lewis acid catalysis by the Sn metal center.
Keywords: acylation; furan; acid catalysis; DFT modeling; flow reactor kinetics; isotopic labeling

1
. INTRODUCTION
The first examples of Friedel-Crafts acylation were carried out using homogenous Lewis
acid catalysts such as aluminum chloride.[1] Brønsted acid zeolites were subsequently found to
catalyze this chemistry and had advantages over the classical Lewis acid catalysts in terms of
separation, isomer selectivity and true catalysis; less than stoichiometric amount of catalyst is
used.[2-4] The acylation of furans with Brønsted acid zeolites is a particularly efficient and selective
means of forming C-C bonds and adding functionality to these bio-derived compounds.[5-8] In
contrast to Brønsted acid zeolites—substituted with trivalent metals like aluminum—solid Lewis
acid zeolites are isomorphously substituted with tetravalent metals that act as isolated framework
Lewis acid sites. These materials have shown remarkable improvement in activity and selectivity
over Brønsted acid zeolites for transformations of biomass, like glucose isomerization and
reduction and etherification of hydroxymethylfurfural,[9-11] but they have yet to be investigated for
acylation. Thus, it is of interest to determine how Brønsted and Lewis acid zeolites compare in
Friedel-Crafts acylation of bio-derived furans.
Brønsted and Lewis acid zeolites with the Beta framework are the focus of this work. For
Brønsted acidic zeolites, H-[Al]-Beta results in higher conversion of methylfuran to 2-acetyl-5-
methylfuran (2A5MF) compared to H-[Al]-Y or H-[Al]-ZSM-5.[8] Beta has been shown to be
superior for the acylation of other reactants with acetic anhydride[12-14] and with other acylation
agents.[15] This zeolite framework has been the subject of many other acylation studies, where
other zeolite structures are not explicitly compared.[5, 6, 16-21] Moreover, there is an array of Lewis
acid zeolites formed from the substitution of various metals (Ti, Zr, Sn, Hf) into the Beta
framework, which can be compared to the Brønsted acid form of [Al]-Beta. In this report, the rate
and mechanism of methylfuran acylation with acetic anhydride, to form 2-acetyl-5-methylfuran

(
2A5MF) and acetic acid (Scheme 1), are compared for a number of Brønsted and Lewis acid Beta
zeolites.
The value of this reaction in the production of biomass-derived commodity chemicals has
been demonstrated as a key step in the production of para aromatic species;[8] and with fatty
anhydrides this chemistry can be applied to the production of surfactants.[22] Methylfuran may be
produced by hydrodeoxygenation of furfural, which is produced industrially from hemicellulosic
feedstocks.[23] Acetic anhydride is used as the acylating agent as it requires much lower
temperatures for reaction than acetic acid. Much of the previous literature on zeolite-catalyzed
acylation with acetic anhydride in the liquid phase focuses on toluene or anisole on H-[Al]-Beta .
These studies found that competitive adsorption between reactants and products contributes to
deactivation and lower reaction rates,[14, 18, 24] and that an acylium ion, rather than ketene, is the
acylating intermediate.[20] Similarly, the mechanism of electrophilic aromatic substitution with
homogeneous Lewis acids is known to proceed via the formation of an acylium ion from the
interaction of the acylating agent and the Lewis acid catalyst. Solid Lewis acid catalysts have not
yet been investigated for this chemistry and thus the mechanism remains to be determined. While
the mechanism is known to include the acylium ion in Friedel-Crafts acylation with acetic
anhydride,[20] the rate-determining step depends on the specific electrophile and nucleophile
involved and can be probed with kinetic isotope effects (KIE). Here the rate-determining step was
probed experimentally with a kinetic isotope effect study of furan acylation on Brønsted and Lewis
acid zeolite and supplemented with density functional theory (DFT) calculations to investigate
various catalytic pathways on H-[Al]-Beta and [Sn]-Beta.
An additional issue for Lewis acid zeolites is characterization of the true active site. While
the tetravalent metal may exist with four bonds coordinated to the framework as a “closed” site, it

has also been found that one of these bonds may hydrolyze to create an “open” site with one
hydroxyl group bound to the metal and a corresponding silanol group, as depicted in Scheme 2.
Baeyer-Villiger oxidation, Meerwein-Ponndorf-Verley reduction and glucose isomerization have
all been shown to be catalyzed by an open [Sn]-Beta site,[25, 26] which involves a Lewis-acidic tin
but also a Brønsted-acidic silanol (Si-OH). Here, experiments and modeling were used to
determine the role of the open site for methylfuran acylation on [Sn]-Beta. X-ray absorption
spectroscopy experiments were also conducted to better understand the active site of [Sn]-Beta
materials.
In this study, Brønsted and Lewis acid Beta zeolites were screened using batch reactors.
Brønsted acidic H-[Al]-Beta and Lewis acidic [Sn]-Beta were also studied in a flow system under
differential conditions. Kinetic isotope effect experiments of furan acylation were used to compare
the rate-determining step. DFT calculations of methylfuran acylation on H-[Al]-Beta and [Sn]-
Beta were performed to probe the reaction mechanism on these materials.
2
2
. METHODS
.1 Experimental
2
.1.1 Materials. The aluminum Brønsted acid forms of zeolite Beta were purchased from
Zeolyst. Specifically, for a low Al content H-Beta the hydrogen form of zeolite Beta (Zeolyst,
CP811C, Si/Al=150) was used as received and is designated as H-[Al]-Beta-150. For a high Al
content H-Beta the ammonium form of zeolite Beta (Zeolyst, CP814E, Si/Al=12.5) was calcined
-1
as follows: ramp 1 K min to 823 K, hold 12 hours. This sample is designated as H-[Al]-Beta-
1
2.5. [Si]-Beta, [Sn]-Beta, [Zr]-Beta and [Hf]-Beta were synthesized hydrothermally as reported
previously.[27] [Sn]-Beta prepared via this method is designated as [Sn]-Beta-HT. Another [Sn]-
Beta was prepared via solid-state ion exchange (SSIE) as described in literature,[28] by grinding

dealuminated H-BEA with tin (II) acetate (Aldrich) and calcining in air. This sample was
designated as [Sn]-Beta-SSIE. For control experiments, bulk SnO was obtained from Sigma
Aldrich and was used without further treatment. Al (99.97%, Alfa Aesar, γ-phase) was tested
as a control as well. Prior to use, Al was calcined at 873 K (ramp 10 K min-1, hold 2 h).
2
2
O
3
2
O
3
Si/Al ratios of the commercial zeolites were determined by X-Ray fluorescence (XRF) with
a Rigaku Supermini200 and the Si/Al for H-[Al]-Beta-12.5 was similar to the value advertised by
the manufacturer but the Si/Al for H-[Al]-Beta-150 was not. For H-[Al]-Beta-150, n-propylamine
adsorption was performed in order to ensure the Al content was as advertised. In this experiment,
Brønsted acid site density was measured via n-propylamine decomposition into propylene and
ammonia using a flow reactor with an on-line Agilent 7890A gas chromatograph (GC) equipped
with an HP-PLOT Q column. This was accomplished using a microreactor system where all gas
lines between the location of n-propylamine introduction and the GC were heat traced with
temperature maintained at or above 348 K. Approximately 50 mg of H-[Al]-Beta-150 was loaded
into a quartz tube in the reactor and was heated to 773 K at a rate of 10 K/min in flowing He (100
mL/min) and held at 773 K for 45 min. Then, the sample was cooled to 373 K and exposed to
flowing He saturated with n-propylamine for 15 min via a bubbler. The sample was subsequently
heated to 473 K and held for 90 min to desorb excess n-propylamine and ensure a 1:1 ratio of
adsorbed n-propylamine to Brønsted acid sites. Finally, the temperature was increased to 773 K at
a rate of 30 K/min. The GC sampling loop was immersed in liquid nitrogen to collect the desorbed
reaction products, which were subsequently quantified via a thermal conductivity detector (TCD).
For Lewis acid zeolites (Sn,-Zr-,Hf-Beta), Si/Metal ratio was determined by inductively
coupled plasma atomic emission spectroscopy (ICP-AES) performed by Galbraith Laboratories
(
Knoxville, TN). [Ti]-Beta was synthesized as described previously.[9] Si/Ti was determined by

XRF with a Rigaku Supermini200. Micropore volume for all samples was determined with
nitrogen physisorption in a Micromeritics 3Flex system using the t-plot method. Samples were
degassed overnight at 523 K and backfilled with nitrogen prior to analysis. X-ray diffraction
(XRD) patterns were collected using a Bruker D8 diffractometer with Cu Kα radiation. The pattern
was collected for 0.5 seconds at each increment of 0.02 degrees between 5 and 50 degrees.
Scanning electron microscopy (SEM) images were recorded on a JEOL JSM 7400F at 10 μA.
Na-exchanged [Sn]-Beta was prepared following a procedure adapted from literature.[26]
Specifically, 300 mg of [Sn]-Beta was mixed with 45 mL of 1 M NaNO
in distilled water for 24 hours at room temperature. The catalyst was recovered by filtration over
a ceramic filter (Chemglass, CG-1402-16) and washed 3 times with 1 M NaNO (50 mL solution
3
(Sigma-Aldrich, ≥99%)
3
for each wash). Then the material was calcined at 853 K for 5 hours (5 K/min ramp).
Thermogravimetric analysis (TGA) on spent catalysts was collected on a Mettler Toledo
TGA/DSC Thermal Gravimetric Analyzer with STARe software. Samples were heated at a rate of
5
K/min under 80 mL/min of air flow. The first derivative (dTGA) of the TGA curves was
calculated numerically using Origin. The differential curves were smoothed with an FFT filter with
00 points.
Extraction of retained organics on spent catalysts was performed as previously
1
described[29]: 30 mg of spent catalyst was dissolved in 1 g HF (Acros, 48%) and 2.6 g DI water
and allowed to react for 1 hour at room temperature. The organic species in the mixture were
extracted with methylene chloride and analyzed with a GC-MS (Shimadzu QP2010 Plus, HP-
Innowax column). Because the concentrations of these samples were so low, the extracted mixture
was placed in a hood overnight to allow some methylene chloride to evaporate before GC-MS
analysis. Therefore, concentrations of the retained organics were not determined.

2
.1.2 Batch Reaction Experiments. For batch reaction experiments, 5 mL of 2-methylfuran
(Sigma Aldrich, 99%, BHT-stabilized) and 15 mL of acetic anhydride (Sigma-Aldrich, ≥99%)
were loaded into a 45 mL 4714 Parr reactor with 50 mg of catalyst and a magnetic stir bar for
mixing. The reactor was pressurized with nitrogen to 14 bar, placed in an oil bath at a temperature
of 383 K, and quenched with an ice bath after the desired reaction time. The reaction product was
filtered from the catalyst with a 0.2 micron syringe filter (Corning). Product samples were analyzed
by gas chromatography (Agilent 7890A) equipped with a flame ionization detector. An HP-
Innowax column (Agilent) was used with the following temperature program: hold at 313 K for
-1
4
.5 min, 10 K min ramp to 523 K and a final hold for 3 minutes. Reaction side products were
identified with a GC-MS (Shimadzu QP2010 Plus) equipped with an HP-Innowax column
following the same temperature program.
For kinetic isotope effect experiments, the specified amount of 1 g furan (Aldrich, 99%) or
1
.06 g furan-d
loaded into a 10 mL with 50 mg catalyst and a magnetic stir bar and sealed with crimp seal septum
Chemglass, CG-4920-10). The vial was then placed in a reactor block with oil at 393 K. After 20
minutes, the catalyst was filtered from the reaction product and analyzed as described above.
.1.3 Flow Reaction Experiments. A flow reactor previously described[27] was used to study
the reaction under continuous flow conditions. Gamma-valerolactone (GVL) (Sigma-Aldrich,
99%) was used as solvent to reduce the concentration of acetic anhydride. This was necessary
because the back pressure regulator (Equilibar, EB1ULF1) diaphragm (Polyimide-5) and o-ring
Viton) is sensitive to acetic anhydride concentrations in excess of 50% v/v. A few experiments
4
(Aldrich, 98% D atom) and 3.2 g acetic anhydride (Sigma-Aldrich, ≥99%) were
(
2
≥
(
carried out without the addition of GVL were performed with a stainless steel diaphragm and
Kalrez o-ring. Typically, a solution comprised of 45.5 g methylfuran, 150 mL acetic anhydride

and 200 mL GVL was supplied at 4 mL/min with an HPLC pump (Alltech). 14 bar pressure was
applied with the back pressure regulator. The reactor consists of a ¼ inch (6.4 mm) stainless steel
tube between two VCR fittings. 50 mg of uniformly sized catalyst pellets (80-120 U.S. mesh, 125-
1
80 µm) was held by a VCR gasket (Swagelok, SS-4-VCR-2-10M) on the bottom, and glass wool
was used to hold the catalyst in place, filling the space up to the top VCR gasket. Unless otherwise
stated, the reactor was submerged in the oil bath and heated under GVL flow until the desired
temperature was reached; at that point, flow of the desired reaction mixture was started. For
reactions at a flow rate of 4 mL/min, the first sample was taken after 15 min of time-on-stream and
then periodically for the next 75 minutes. Experiments with H-[Al]-Beta-150 were run at 2
mL/min, and the initial sample was taken after 30 minutes and the periodically for the next 150
minutes. At each time point, 2 mL of sample was collected at the outlet and analyzed offline via
gas chromatography (Agilent 7890A) equipped with a flame ionization detector. An HP-1 column
(
Agilent) was used with the following temperature program: hold at 313 K for 4 min, 15 K min-1
ramp to 473 K and a final hold for 1 minute. The carbon balance in all runs was between 97.1-
03.6%.
1
2
.1.4 X-ray absorption spectroscopy. X-ray absorption spectroscopy experiments were
performed at beamline 5BM-D (DND-CAT). Spectra were recorded in transmission at the Sn K
edge (29200 eV), using sealed ionization chambers (Danfysik). A Sn foil was used for energy
calibration. Zeolite samples and a SnO standard were pressed into pellets in stainless steel six-
2
well sample holders (“shooters”), for a total absorption equal to 2 at the Sn K edge. The sample
holder was then placed in a quartz tube (25.4 mm OD), which was in turn sealed with stainless
steel fittings equipped with Kapton windows and welded ball valves. Nitrogen (Airgas 5.0) was
flowed through the sample using a rotameter at 100 ml/min and the temperature was raised at 2

K/min to 923 K while recording spectra. The spectra were processed using the Demeter suite,
using the Athena software for spectra normalization and the Artemis software for fitting. The
-1
fitting was performed using data from k = 2.5 to 13 Å for R = 1 to 2.3 Å. The amplitude reduction
factor (S ) was calibrated using SnO
0
2
.
2
.2 Computational
.2.1 Zeolite Model. In the computational studies of the catalytic pathways, the H-[Al]-
2
Beta and [Sn]-Beta zeolites were modeled with clusters that were treated quantum mechanically.
For H-[Al]-Beta, the active site was created at the T2 crystallographic site by substituting an Al
atom for a Si atom and neutralizing with a proton.[30-32] In the case of [Sn]-Beta, the exact location
of Sn in the framework is difficult to ascertain and also dependent on the synthesis method, and so
still a matter of debate and vigorous experimental work.[32-34] In this work we have opted for the
T2 site. For [Sn]-Beta, we modelled both the closed (unhydrolysed) and open (hydrolysed) active
sites (Scheme 2).
The cluster models were hewn out of the zeolite crystals using the “multi-centered distance
cutoffs” approach of Migues, et al.[35] Using geometries of the reactants in a very large cluster of
the zeolite pre-optimized at a low-theory-level, the zeolite cluster is subsequently trimmed down
by keeping only framework atoms within 5 Å of any atom of the 4T active site and of any atom of
the guest molecules (reactants). To build well-connected clusters, framework atoms that lie outside
the cutoff boundary but are bonded to atoms inside the boundary are also included in the cluster
model. The cutoff radius of 5 Å has been determined by Migues et al. to be a reliable compromise
between convergence and computational efficiency. The cluster is terminated at Si atoms whose
dangling bonds are subsequently capped with H atoms; the terminal Si-H bonds are 1.47 Å long
and in the direction of the corresponding Si-O crystallographic bond. The final clusters consisted

of 39T atoms and a total of 142 atoms for [Al]-Beta and 141 atoms for [Sn]-Beta; and incorporated
the 12-membered ring of the channel along [100] and the 4- and 5-membered rings around the
active site (see Figure 1).
2
.2.2 Electronic Structure Calculations. All calculations were performed with the Gaussian
0
9 package.[36] The Al, C, O and non-terminal H atoms were modeled at the M06-2X/6-31G(d,p)
theory level; the capping H atoms were modelled at the M06-2X/3-21G level. The Si and Sn atoms
were modelled at the M06-2X/LANL2DZ level. All the atoms were allowed to relax, except the
capping H atoms, which were kept frozen. The transition states (TS) were characterized by
vibrational frequency analyses and intrinsic reaction coordination (IRC) calculations. Population
analysis was performed using the Natural Bond Orbital (NBO) theory, implemented in the NBO
6
.0 program.[37]
Unless otherwise specified, Gibbs free energies were computed at 393.15 K using the quasi
rigid-rotor harmonic oscillator (qRRHO) approximation of Grimme[38] and of Li et al.[31] In
addition, adsorbed species were considered as mobile and entropic contributions associated with
the two-dimensional translational motion in the pore were added to the adsorbates. Details can be
found in the SI (Section S1).
Binding energies (퐸 ,) were calculated by the equation
퐵ꢀ
퐸퐵ꢀ = 퐸푎푑푠 + 퐸푧푒표 − 퐸푎푑푠ꢁ푧푒표
,
where 퐸푎푑푠 is the gas phase adsorbate energy, 퐸푧푒표 is the energy of the bare zeolite and 퐸푎푑푠ꢁ푧푒표
is the total energy of the adsorbate in the zeolite.
2
.2.3 Microkinetic Modeling. The calculated free energies profiles were used to
parametrize a microkinetic model which was used to compute apparent activation energies and to
determine rate-determining steps by sensitivity analysis. Details are provided in the SI.

3
3
. RESULTS
.1 Catalyst Characterization
Two Brønsted acid Beta zeolites were investigated: one with high aluminum content, H-
[
Al]-Beta-12.5, and one with low aluminum content, H-[Al]-Beta-150. A series of Lewis acid Beta
zeolites (Sn, Zr, Hf, Ti) were synthesized hydrothermally. A siliceous zeolite Beta (Si-Beta) was
also synthesized hydrothermally and tested as a control. Finally, a post-synthetic [Sn]-Beta-SSIE
was synthesized via solid-state ion exchange by grinding tin acetate with dealuminated H-[Al]-
Beta-12.5. This technique allows a much higher incorporation of tin compared to what is possible
via the hydrothermal technique. The [Sn]-Beta synthesized hydrothermally is referred to as [Sn]-
Beta-HT to distinguish it from its solid-state ion exchange analog.
The Si/M (M=heteroatom) and micropore volume are reported in Table 1 for each material,
including the dealuminated Beta that [Sn]-Beta-SSIE was synthesized from. Si/M ratios were
determined by ICP-AES, unless otherwise noted. In the case of H-[Al]-Beta-150, which was
purchased from Zeolyst, Si/Al=75 was determined from XRF which is twice as high as listed, and
could be due to extraframework (non-Brønsted acidic) aluminum. To address this seeming
inconsistency, we performed n-propylamine decomposition, which specifically probes Brønsted
acid site concentration. 120 µmol Brønsted acid sites per gram was measured, which corresponds
to Si/Al=138, similar to the Si/Al=150 reported by the manufacturer. Normalized rates for H-[Al]-
Beta-150 were based on this 120 µmol/g concentration. Similarly, the Brønsted acid site density
of the H-[Al]-Beta-12.5 sample was determined to be 690 µmol/g, using the n-propylamine
decomposition test. This corresponds to Si/Al=22.9, consistent with the presence of
extraframework Al.[38]

The results of nitrogen physisorption analysis are summarized in Table 1, while XRD
patterns and SEM images are reported in the SI (Section S2). Micropore volumes and XRD
patterns were all consistent with the structure of zeolite Beta for all the materials investigated.
[
Sn]-Beta-SSIE shows additional peaks in the XRD pattern consistent with the formation of SnO
2
from the decomposition of excess tin acetate in the zeolite. Consistent with this, the micropore
volume of [Sn]-Beta-SSIE was lower than that of the other zeolites, likely due to the presence of
SnO
2
occluded in the zeolite pores.
A series of X-ray absorption experiments were performed to understand the local
coordination environment of [Sn]-Beta zeolites (Table 2). At room temperature, the tin in as-
prepared [Sn]-Beta-HT zeolite is bound to six O atoms. Compared to values reported in the
literature,[39,40] the optimized Debye-Waller factor (σ ) is significantly higher, indicating greater
uncertainty in the measurement. This reflects the fact that the Sn-O distances are longer for the Sn-
2
OH bond to adsorbed water, compared to the Sn-OH and framework Sn-O bonds. Upon heating
2
under inert gas, the coordination numbers, bond distances and Debye-Waller factors decrease, as
°
initially the coordinating water is lost at temperatures lower than 120 C (Figure 2). Following that,
the condensation of adjacent silanol and stannanol groups results in the closing of the open sites
and the reduction of the coordination number to 4.
In contrast, [Sn]-Beta-SSIE shows a much higher coordination number, both at room
temperature and after treatment with inert gas at high temperature. At room temperature, the
coordination number is 7.15 ± 0.37, while after dehydration, it is equal to 5.81 ± 0.47. This suggests
that most (~90%) of the Sn in the sample is present as SnO
2
, which has a coordination number of
(2.044
6
. The average bond distance of 2.029 Å is also closer to the Sn-O bond distance in SnO
2

Å and 2.061 Å for axial and equatorial O atoms, respectively) than to the Sn-O distance in the
zeolite framework.
3
.2 Batch Reactions
The results of acylation of methylfuran with acetic anhydride to form 2A5MF (Scheme 1)
over various zeolite Beta catalysts at 383 K using a batch reactor are shown in Table 3. Turnover
frequency (TOF) is normalized by metal content of the catalyst as determined by XRF or ICP-
AES. Brønsted acidic H-[Al]-Beta-12.5 (row 1) exhibited the highest rate on a per gram basis. The
high and low concentration H-[Al]-Beta catalysts (row 1 and 2) exhibited very similar TOF
normalized by metal content, higher than any of the Lewis acid materials. The extraframework
aluminum in these materials does not contribute significantly to the acylation activity as evidenced
by row 9. The difference in TOF in the case of the [Sn]-Beta materials (row 3 and 7) is certainly
affected by the presence of SnO
rate is normalized by total metal content and SnO
inactivity of most of the tin in [Sn]-Beta-SSIE, the specific reaction rate is significantly better than
Sn]-Beta-HT. The other hydrothermally synthesized Lewis acid zeolites had comparatively low
2
(observed via XRD and EXAFS) in [Sn]-Beta-SSIE, since the
2
is inactive for this reaction (row 8). Despite the
[
reactivity, less than half the TOF measured on [Sn]-Beta (rows 4-6). [Ti]-Beta showed the least
activity, similar to previous comparisons of Lewis acid Beta zeolites for Meerwein-Ponndorf-
Verley reduction, where [Sn], [Zr] and [Hf]-Beta reaction rates are significantly faster than Ti.[41,
4
2]
Control experiments with siliceous Beta or with no catalyst (rows 10-11) produced no acylated
product. While [Sn]-Beta-HT had a TOF slightly higher than H-[Al]-Beta-12.5, its TOF was much
lower than that of H-[Al]-Beta-150 which has a similar heteroatom concentration. While it is clear
the Brønsted acid material is superior for this acylation chemistry, the [Al]- and [Sn]-Beta

materials were next investigated in a flow reactor under differential conditions to better understand
their reactivity.
3
.3 Flow Reactions
.3.1 Activation energy. Since the [Al]-and [Sn]-Beta catalysts exhibited the highest
3
turnover under batch conditions, they were investigated in a flow reactor to quantify the activation
energy. All catalysts exhibited deactivation with time on stream as shown in Figure 3. To
determine the initial rate, the product was sampled over time and the rate was extrapolated to t=0
with an exponential fit. Catalyst lifetimes were compared by determining the half-life of the
catalyst (the time at which half of the rate at t=0 is observed, see Table 4). Arrhenius plots based
on initial rates were used to determine the apparent activation energy of the reaction (Figure 4 and
Table 4). Given the low activation energy on Sn compared to Al and the large crystal size of [Sn]-
Beta-HT (Figure S2c), the possibility of mass transfer limitations was considered. However, the
crystal size of [Sn]-Beta-SSIE prepared via solid-state ion exchange was much smaller (Figure
S2g) and comparable to that of the H-[Al]-Beta-12.5 (Figure S2a) from which it was synthesized.
Since a very similar activation energy was measured on two tin catalysts with very different crystal
sizes, it is unlikely that mass transfer is affecting the apparent activation energy. Additionally, the
Weisz-Prater criterion was estimated and falls below the limit where pore diffusion is limiting (SI
Section 3).
3
.3.2 TGA of spent catalysts. All catalysts deactivated with time on stream. Among the
catalysts investigated, the half-life of the catalysts was in the range of 18-32 minutes (total length
of experiments was 180 minutes for H-[Al]-Beta-150 and 90 minutes for all others). The spent
catalyst obtained from the lowest-temperature measurement of each catalyst was analyzed using

TGA (Figure 5). H-[Al]-Beta-12.5 exhibited the largest weight loss (43%), followed by [Sn]-Beta
37%). H-[Al]-Beta-150 and [Sn]-Beta-SSIE had smallest weight loss (21% and 19%,
(
respectively). Since the catalysts were only dried under nitrogen flow at room temperature, and
the boiling points of the reactants and products are high, a portion of this weight loss is from
reactants or products left inside and outside of the catalyst pores. The boiling points of the reactants
and products in this system are: 336 K (methylfuran), 414 K (acetic anhydride), 373 K (2A5MF),
3
91 K (acetic acid), and 480 K (GVL). The differential of the TGA traces is shown in Figure S3,
but even from Figure 5 it is clear that there is significant weight loss on H-[Al]-Beta-12.5 and [Sn]-
Beta-HT between 380-400 K that is not observed on the other two catalysts. Based on boiling
points, the weight loss in this region seems consistent with loss of either acetic acid or anhydride.
While weight loss in the TGA could be from reactants and products left in the
intracrystalline voids of the samples, the TGA analysis also revealed some weight loss at higher
temperatures that might be due to the occlusion of large byproducts in the pores. Spent catalyst
from the flow reactions were dissolved with HF, and the organics were extracted and analyzed
with GC-MS (Table 5) to identify the retained molecules. GVL was found retained on all the
catalysts, which is likely due to its high boiling point since the catalyst is “dried” under nitrogen
flow at room temperature. Acetic acid was detected in the extracted organic phase for H-[Al]-Beta-
1
2.5, it is also possible that acetic acid and acetic anhydride were retained on the other catalysts,
but stayed in the aqueous HF phase due to their polarity. Products with a higher molecular weight
than the desired 2A5MF product (MW=124) were found on some of the catalysts, indicating some
deactivation from pore blockage by these large molecules. The MW=206 product detected on H-
[
Al]-Beta-12.5 could be derived from the condensation of hydrated methylfuran (1,4-

pentanedione) with 2-acetyl-5-methylfuran. The MW=290 product detected on [Sn]-Beta could be
from acylation at both ends of the double bond in the 206 molecule. Despite deactivation, previous
work has shown that H-[Al]-Beta zeolites run in batch reactions at high temperature (453 K) can
be regenerated via calcination resulting in minimal loss in conversion (<5%) and selectivity
(
<1%).[8]
3
.3.3 Effect of GVL. GVL was used as solvent to minimize acetic anhydride and acetic acid
concentration in the effluent and minimize potential damage to the back-pressure regulator.
Acetonitrile was also investigated as a solvent, but resulted in the same or lower rates on H-[Al]-
Beta-12.5 and [Sn]-Beta-HT, respectively (Figures S4 and S5). Extraction and analysis of the
organics on spent catalyst from these experiments revealed the presence of both 2A5MF and GVL
retained in the catalyst. It is possible that methylfuran is hydrated by residual water, forming an
unstable 1,4-pentanedione that then isomerizes to GVL. There are no products corresponding to
the molecular weight of 5,5-bismethylfuran-2-pentanone, formed from the reaction of two
methylfuran molecules with 1,4-pentanedione, as described by Corma et al.[42] The observation of
2
A5MF in the spent catalyst in acetonitrile but not in GVL indicates that GVL may have a
promoting effect in removing the 2A5MF product from the active site, resulting in the higher rates
observed for [Sn]-Beta with GVL as solvent. Consequently, a backpressure regulator configuration
compatible with acetic anhydride was used to investigate the reaction without solvent on H-[Al]-
Beta-150. This resulted in a dramatic decrease in the rate (no GVL in Figure 6) compared to when
GVL was used (with GVL in Figure 6), which further reveals the positive effect of GVL solvent
on this reaction. Additionally, when the bed was heated under GVL flow but then only supplied
with methylfuran and acetic anhydride once the desired temperature was reached (heat with GVL

in Figure 6), the initial rate was improved over heating under acetic anhydride flow, but still
reached a regime of no product evolution after about two hours of time on stream.
3
.3.4 Co-feeding of products. The retention of acylation products can contribute to
deactivation observed in acylation reactions.[14, 16, 18, 24] To this end, acetic acid and 2A5MF
products were co-fed to determine their effect on the reaction rate on the catalyst with the highest
turnover, H-[Al]-Beta-150. As shown in Figure 7, co-feeding just a small amount (5 mol% based
on methylfuran) of acetic acid resulted in a reduced reaction rate. Similarly, co-feeding 5 mol% of
2
A5MF decreased the initial rate, though not as severely as 5 mol% of acetic acid. A further
increase of 10 mol% 2A5MF in the feed reduced the initial rate more. The co-feeding of 2A5MF
also slowed the rate of deactivation and the catalyst half-life was increased from 32 to 44 minutes.
The reduced reaction rate upon co-feeding of products indicates product inhibition is contributing
to deactivation similar to what is reported in other acylation studies.
3
.4 Mechanistic Studies
.4.1 Kinetic isotope effect. The acylation mechanism over zeolites was probed with
labeled furan-d . A significant primary isotope effect of 2.5 and 1.8 was found for both H-[Al]-
3
4
Beta-150 and [Sn]-Beta-HT, respectively. For deuterated aromatics, a concerted, single-step
substitution of the acetyl group should result in a primary kinetic isotope effect (KIE). In contrast,
the absence of a primary KIE would suggest a two-step mechanism, involving the formation of a
Wheland intermediate followed by deprotonation (see Scheme 3).[42] This Wheland complex is the
origin of the positional selectivity observed in electrophilic aromatic substitution (EAS) reactions;
a late Wheland complex transition state leads to high positional selectivity. This complex is also

stabilized by electron donating groups, which accounts for the acceleration of EAS reactions by
electron donating substituents on the aromatic. Whether the formation of this intermediate is the
rate-determining step has been found to be dependent on the strength of the electrophile and
nucleophile. In the case of strong electrophiles or strong nucleophiles, the highest energy transition
state resembles the starting aromatic (early transition state) and the formation of the Wheland
intermediate is rate determining. In the case of weak electrophiles or weakly basic aromatics, the
highest energy transition state resembles the Wheland intermediate (late transition state) and the
deprotonation of this intermediate is rate determining.[1, 43, 44] If formation of the Wheland complex
is rate determining, only a small secondary inverse KIE would be expected due to change from sp2
3
[45]
to sp of the carbon on the aromatic.
If deprotonation of the Wheland complex is rate
determining, or if the mechanism is concerted, a significant primary KIE would be expected due
to C-H bond breaking during hydrogen elimination.[45] Based on these considerations, the
measured KIE in the case of furan-d indicate that deprotonation (either in a concerted or stepwise
4
manner) is the rate-determining step. This is also supported by the computational modeling
discussed in the next section. We should note that strong primary KIE of 3.25 and 2.25 have been
observed for toluene and benzene acylation, respectively, with acetyl fluoride catalyzed by
antimony pentafluoride.[46] However, for the acylation of p-xylene with isobutyl chloride catalyzed
by aluminum chloride, the lack of primary KIE has ruled out deprotonation as the rate-determining
step.[47] As explained by Effenberger, the magnitude of the KIE is a function of the Wheland
complex formation and deprotonation rates.[48]
Since methylfuran is a stronger nucleophile than furan, on account of the electron-donating
methyl group—and strong nucleophiles are not associated with primary KIE and an earlier rate-
determining step—it might be that the rate-determining step could be different for methylfuran

compared to furan. Indeed, calculations and microkinetic analysis (see next section) indicate a
change in the rate-limiting step in the case of 2-methylfuran. That was unexpected considering that
a primary KIE for both toluene and benzene with the same acylating agent (as discussed earlier)
indicates that the additional methyl group is not enough to change the rate-determining step.
3
.4.2 Na-exchanged [Sn]-Beta. To determine whether the active site for acylation on [Sn]-
Beta is open or closed, a [Sn]-Beta-HT sample was exchanged with sodium to replace the hydroxyl
groups at the open site, as reported in Bermejo-Deval et al.[26] When the exchanged sample was
tested under the same conditions as described for the experiments in Table 3, no 2A5MF was
formed, that is, sodium exchanged with an open site and prevented the reaction from occurring,
that is, the open site is the active site. There is still uncertainty in the quantification of open and
closed sites in these materials, particularly under reaction conditions. Currently the active site
concentration is estimated to be the same as the metal content, however, if only a fraction of Sn is
in an open configuration, this would result in an underestimate of the true turnover frequency.
Since recent work has shown that less than half of Sn atoms in [Sn]-Beta samples are open sites,[49]
the TOFs reported here are probably underestimated.
An additional consideration arising from the identification of the Sn open site as the active
site is the possibility of Brønsted acid catalysis at this site. Recent computational modeling of [Sn]-
Beta open sites revealed the silanol coordinated to the Sn of the open site to be more Brønsted
acidic than typical framework defect silanols and comparable in binding strength to the Lewis-
acidic Sn center for strong bases like ammonia and pyridine.[50] Given that this chemistry can
clearly be catalyzed by either Brønsted or Lewis acids, it is possible that the reaction occurring at
the open Sn site is actually catalyzed by the Brønsted-acidic silanol and not the Lewis-acidic Sn

atom. The reaction was investigated computationally to address this question and to compare the
mechanisms on [Al]- and [Sn]-Beta.
3
.5 Computational Modeling of Acylation on H-[Al]-Beta and [Sn]-Beta
.5.1 H-[Al]-Beta. Acetic anhydride binds more strongly than methylfuran to H-[Al]-Beta
by 11.8 kcal/mol (Figure 8). The binding energy of acetic anhydride is 12.7 kcal/mol and binding
takes place through a carbonyl oxygen, O , forming a hydrogen bond (1.39Å) with the proton of
the active site (Figure 8a). NBO analysis shows weakening of the O -H bond at the Brønsted site;
the occupancy of the anti-bonding orbital σ*(O -H) is 0.14 compared to 0.02 prior to binding. The
-H), however, is unaffected: 1.99 prior to and 1.98 after
3
1
5
5
occupancy of the bonding orbital σ(O
5
binding.
We have identified two acylation pathways on H-[Al]-Beta (Scheme 4). Both follow the
classic, two-stage, addition-elimination mechanism of electrophilic aromatic substitution which
involves formation of a stable addition intermediate (σ-complex). The two pathways differ,
however, in one respect. In the first, (H1), the active-site-bound anhydride dissociates upon
protonation to form acetic acid and an acyl cation, which is stabilized by covalent bonding to a
framework oxygen atom. The first phase of the reaction is completed by electrophilic attack on
methyfuran in the C position by the acyl cation. The electrophilic substitution is completed by
5
proton elimination from the intermediate (σ-complex) back to the conjugate base of the active site.
In contrast, in the second pathway (H2), the protonation of the anhydride, its dissociation and
formation of the acyl cation and the electrophilic attack by the latter occur in a single step.
The free energy profiles for the two pathways are shown in Figure 9. In H1, the dissociation
H
H1
‡
of the protonated anhydride (INT1 to INT1 ) has an intrinsic barrier (ΔG ) of 24.3 kcal/mol.

NBO analysis confirms a covalent bond between the acyl group and a framework oxygen atom
H1
(
INT1 ). This is consistent with work by Kresnawahjuesa et al., who have reported dissociation
of acetic anhydride over H-[Al]-ZSM-5.[51] Then the acyl group detaches from the zeolite to form
H1
‡
a surface acyl cation (INT2 , ΔG = 10.7 kcal/mol), which easily adds to the C of methylfuran
5
‡
(
ΔG =3.7 kcal/mol). The proton back-donation to the conjugate base of the Brønsted site requires
‡
ΔG =12.5 kcal/mol. In the H2 mechanism, the concerted formation of the acyl cation and its
H
H
‡
addition to methylfuran (INT1 to INT2 ) has ΔG =37.0 kcal/mol, significantly higher than in H1.
H2
The geometry of the transition state (TS ) shows that the proton of the active site has already
migrated to the acetic anhydride while the acyl cation is in close proximity to the C of
5
methylfuran. All of these factors indicate a late transition state, characteristic of an endergonic
process and in accord with Hammond’s postulate.[51] As can be seen from Figure 9, the two
H
pathways, H1 and H2, share the intermediate INT2 and all the intermediates that follow on from
it.
Clearly, the stepwise pathway H1 is energetically more favorable for the acylation of
methylfuran over H-[Al]-Beta. Although the free energy profile in Figure 9 suggests that the
hydrogen elimination is the rate-determining step, microkinetic modelling (vide infra) reveals that
the rate of acylation of methyfuran is controlled by the breaking of the C-O-C linkage in the
anhydride and the acyl formation. The rate of acylation of furan, however, is controlled by the
hydrogen elimination step (presented in Section 8 of the SI), which is consistent with the observed
KIE reported above from kinetic isotope effect experiments.
3
.5.2 [Sn]-Beta. Both the closed and open [Sn]-Beta sites were modeled; the mechanism
on the closed site was found to be slower than on the open site. Given that the sodium exchange
experiments also support the open active site, only the open site calculations are discussed here,

but a discussion of the mechanism on the closed site is available in the SI (Section S6). Owing to
the moderate acidity of the silanol proton of the [Sn]-Beta open site, which has been demonstrated
in previous DFT studies,[50, 53] the silanol could be deprotonated by acetic anhydride and catalyze
the reaction following pathways similar to those described above for H-[Al]-Beta. At the same
time, the metal Sn center may also participate in the reaction as a Lewis acid. This interplay
between Brønsted and Lewis acidity complicates the elucidation of mechanisms for the open-Sn
site. In the following, we present three competing reaction pathways for the acylation of
methylfuran (Scheme 5). Below, two of them will be classified as Brønsted acid catalysis (denoted
by D1 and D2), as they only involve the moderately acidic silanol group and are very similar to
those identified in the H-[Al]-Beta case. The third pathway (denoted by D3) engages both the
Lewis-acidic metal center and the neighboring silanol group.
Two modes of binding of acetic anhydride to the active site were considered, shown in
Figure 10. In the first mode (Figure 10a), one of the two carbonyl groups of the anhydride is
coordinated to the Sn atom, while the other carbonyl interacts with the silanol, with a total binding
energy of 4.4 kcal/mol. In this binding geometry, Sn assumes octahedral coordination.
Alternatively, as shown in Figure 10b, both carbonyl groups of the anhydride may interact with
the silanol proton, while the Sn atom remains pentacoordinated. This results in a more stable
geometry, with a binding energy of 9.4 kcal/mol and is consistent with the XAS data presented
above. NBO analysis shows that stabilization largely comes from hydrogen bonding while the lone
pairs of both carbonyl oxygen atoms donate electron density to the antibonding orbital of the
silanol OH. In all three identified reaction pathways presented in Scheme 4, the reaction proceeds
through the latter binding mode (Figure 10b).

Free energy profiles for the three mechanisms are shown in Figure 11 and Figure 12. The
mechanisms D1 and D2 (Figure 11) are essentially the same as H1 and H2 in H-[Al]-Beta (Figure
9
), respectively, and overall, the pathway D1 is energetically more favorable than D2. In the D1
‡
pathway, the dissociation of the anhydride and the formation of the acyl cation requires ΔG = 24.7
kcal/mol compared to 24.3 kcal/mol for the H1 case in H-[Al]-Beta, indicating similar Brønsted
acid strengths.
In the mechanism D3, which engages the Sn center, the reaction proceeds with donation of
the silanol proton to the bridging oxygen of the anhydride, producing acetic acid and an acylium
‡
cation, and requires ΔG = 25.7 kcal/mol. The subsequent electrophilic attack of the furan ring by
D3
the acyl cation (INT2 ) is practically non-activated, as it requires a mere 1.7 kcal/mol to climb up
the barrier. The geometries of TS1D3 and INT2D3 in Figure 12 show the Sn atom in a 6-fold
coordinated state, which is evidence of the engagement of the Sn atom. Interestingly, the proton
elimination after the acyl addition is different from the mechanisms described so far. In pathway
D3, the proton back-donation to the active site is still the rate-controlling step, but it does not take
place in a direct fashion. Rather, it is mediated by the carboxyl group of acetic acid. The intrinsic
D3
free energy barrier for the hydrogen elimination is calculated to be 17.9 kcal/mol (TS3 ), which
is 5.4 kcal/mol higher than in pathway D1. Clearly, the participation of the Lewis-acidic Sn center
does not facilitate the reaction.
What endows the silanol with stronger than usual acidity is its proximity to the Sn
framework atom—the oxygen of SiOH is not formally bonded to the Sn framework, but it is,
nevertheless, within the first coordination sphere of the latter (see Figure 13a). As a result, the OH
moiety of the silanol is polarized, becoming a good proton donor. We have compared the relative

acidities of the open [Sn]-Beta silanol and the bridging hydroxyl in H-[Al]-Beta by computing the
deprotonation energies the respective deprotonation energies (DPE):
퐷푃퐸 = 퐸푍푒표ꢂ푀
ꢃ
− 퐸푍푒표[푀] ,
where 퐸푍푒표ꢂ푀
ꢃ
and 퐸푍푒표[푀] are the electronic energies (zero-point energy corrected) of the
deprotonated and initial metasubstituted zeolites, respectively. The smaller the DPE, the easier to
remove the proton, and the stronger the acidity of the catalyst. The DPE of the open [Sn]-Beta
silanol is 293 kcal/mol, which is only 5 kcal/mol higher than that of H-[Al]-Beta (288 kcal/mol).
It is also worth pointing out that upon deprotonation of the silanol, the conjugate base is stabilized
by restoring the former Sn-O covalent bond with the framework, which effectively delocalizes the
negative charge (Figure 13b). Thus, even though we do not see effective Lewis acid catalysis by
the Sn metal center, the catalytic ability of the silanol must be attributed to the presence of the
metal atom. One would have expected that [Zr]-Beta and [Hf]-Beta would be as active as [Sn]-
Beta as the DPEs of the respective open sites are equal to 292 and 290 kcal/mol, practically the
same as that of [Sn]-Beta. It is certainly curious and unclear why [Zr]-Beta and [Hf]-Beta turned
out to be quite inactive when tested for this reaction—a matter that warrants further investigation.
3
.5.7 Microkinetic model. We have utilized the free energy profiles presented above to look
into the microkinetics of the reaction. The model and its parameterization are presented in detail
in Section S7 of the SI. By solving the model at different temperatures and constructing Arrhenius
plots, we have computed apparent activation energies of 22 kcal/mol in H-[Al]-Beta and 25.4
kcal/mol in [Sn]-Beta. In the case of H-[Al]-Beta, the agreement with the experimental values of
1
6.9 kcal/mol for H-[Al]-Beta-12.5 and 20.9 kcal/mol for H-[Al]-Beta-150 is quite satisfactory
Table 4). In the case of [Sn]-Beta, there is a discrepancy of about 16 kcal/mol between the model
and the experimental value. A plausible explanation for this discrepancy is that the ONIOM model
(

for [Sn]-Beta possibly overestimates the enthalpy of binding of the anhydride to the active site,
overpredicting the coverage in anhydride. Such overbinding does not affect the calculated intrinsic
activation energies (primarily owing to error cancellation), but it can affect the predicted apparent
activation energies, as the latter are determined by both the intrinsic barriers and heats of
adsorption; the stronger the binding, the higher the coverage and the closer the apparent activation
energy is to the intrinsic energy span.
Sensitivity analysis of the rate of the acylation of 2-methylfuran (Figure 14) has revealed
that on both H-[Al]-Beta and [Sn]-Beta the reaction is controlled not by the deprotonation of the
Wheland intermediate but rather by the dissociation of the anhydride and the formation of the
framework bound acyl group. In contrast, the acylation of furan is controlled by the last step of the
reaction, namely the hydrogen elimination that follows the addition of the acyl group to the furan
ring (Figures S9 and S10 in Section S8 of the SI). This is in complete agreement with the KIE
experiments presented earlier in Section 3.4.1. Our calculations with furan on H-[Al]-Beta also
show that its acylation is considerably slower than that of 2-methylfuran, with an apparent
activation energy of 31.7 kcal/mol, consistent with the fact that electron donating groups enhance
the nucleophilic character of the furan ring, accelerating the electrophilic aromatic substitution.
4
. CONCLUSIONS
A comparison of Brønsted and Lewis acid zeolite Beta catalysts for methylfuran acylation
reveals that H-[Al]-Beta with a low Si/Al ratio exhibits the highest specific reaction rates. When
turnover is normalized on a per site basis, H-[Al]-Beta exhibits higher turnover frequency
compared to all Lewis-acid zeolite Beta materials investigated. The difference in rates and
activation energies on [Al]- and [Sn]-Beta is interesting, since the apparent activation energy is

lower on [Sn]-Beta despite it having a lower rate. It is possible that the intrinsic activation energy
is higher for [Sn]-Beta and a more exothermic adsorption makes it appear lower, however this is
not supported by the lower binding energy computed for [Sn]-Beta compared to H-[Al]-Beta.
There could be a large compensation in the rate through entropic effects on H-[Al]-Beta that result
in a higher pre-exponential factor and higher rate despite a higher activation energy. However, the
rate on [Sn]-Beta may be underestimated because we were unable to unequivocally determine the
relative number of Sn open sites. This report demonstrates that commercially available Brønsted
acid zeolites exhibit excellent activity and selectivity for this reaction. Although all the catalysts
deactivate with time on stream, GVL has a promotional effect on the rate, which may be helpful
for scale up purposes.
Computational modeling of the acylation of methylfuran reveals that the reaction follows
classic addition-elimination (i.e., stepwise) aromatic electrophilic substitution mechanisms in both
[
Al]- and [Sn]-Beta, and that the rate-determining step is the hydrogen elimination from the
Wheland intermediate in the case of furan but the dissociation of the acetic anhydride in the case
of methylfuran. The former is supported experimentally by the observation of a primary KIE on
both catalysts. One remarkable and unexpected finding from our calculations is that the most
favorable catalytic pathway in [Sn]-Beta involves Brønsted acid catalysis by the silanol group of
the open site and not Lewis acid catalysis by the Sn metal center.
Acknowledgements
The authors gratefully acknowledge Nicholas Gould for performing the n-propylamine
decomposition measurement. Portions of this work were performed at the DuPont-Northwestern-
Dow Collaborative Access Team (DND-CAT) located at Sector 5 of the Advanced Photon Source

(APS). DND-CAT is supported by Northwestern University, E.I. DuPont de Nemours & Co., and
The Dow Chemical Company. This research used resources of the Advanced Photon Source, a
U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office
of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. This
material is based upon work supported as part of 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-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-AC02-05CH11231. Z.Z. would like to thank Dr. Tyler R. Josephson for many useful
discussions.

References
[
[
[
[
[
1] G. A. Olah, Friedel-Crafts Chemistry, John Wiley & Sons, Inc., New York, NY, 1973, p. 581.
2] B. Chiche, A. Finiels, C. Gauthier, P. Geneste, J. Graille, D. Pioch, J. Org. Chem. 1986, 51, 2128-2130.
3] A. Corma, M. J. Climent, H. Garcia, J. Primo, Appl. Catal. 1989, 49, 109-123.
4] A. Corma, H. Garcia, Chem. Rev. 2003, 103, 4307-4365.
5] V. F. D. Alvaro, A. F. Brigas, E. G. Derouane, J. P. Lourenco, B. S. Santos, J. Mol. Catal. A: Chem.
2009, 305, 100-103.
[
[
[
[
[
6] Y. Xiong, W. Chen, J. Ma, Z. Chen, A. Zeng, RSC Adv. 2015, 5, 103695-103702.
7] D. E. Resasco, B. Wang, S. Crossley, Catal. Sci. Technol. 2016, 6, 2543.
8] M. Koehle, E. Saraci, P. Dauenhauer, R. F. Lobo, ChemSusChem 2017, 10, 91-98.
9] M. Moliner, Y. Román-Leshkov, M. E. Davis, PNAS 2010, 107, 6164.
10] J. D. Lewis, S. Van de Vyver, A. J. Crisci, W. R. Gunther, V. K. Michaelis, R. G. Griffin, Y. Roman-
Leshkov, Chem. Sus. Chem. 2014, 7, 2255.
[
[
[
[
11] P. Y. Dapsens, C. Mondelli, J. Perez-Ramirez, Chem. Soc. Rev. 2015, 44, 7025.
12] U. Freese, F. Heinrich, F. Roessner, Catal. Today 1999, 49, 237-244.
13] L. Cerveny, K. Mikulcova, J. Cejka, Appl. Catal. , A 2002, 223, 65-72.
14] E. G. Derouane, G. Crehan, C. J. Dillon, D. Bethell, H. He, S. B. Derouane-Abd Hamid, J. Cat. 2000,
194, 410.
[
[
[
[
[
[
15] J. Klisakova, L. Cerveny, J. Cejka, Appl. Catal. , A 2004, 272, 79-86.
16] D. Rohan, C. Canaff, E. Fromentin, M. Guisnet, J. Cat. 1998, 177, 296-305.
17] D. Rohan, C. Canaff, P. Magnoux, M. Guisnet, J. Mol. Catal. A: Chem. 1998, 129, 69-78.
18] P. Botella, A. Corma, J. M. Lopez-Nieto, S. Valencia, R. Jacquot, J. Cat. 2000, 195, 161-168.
19] M. Guidotti, C. Canaff, J. Coustard, P. Magnoux, M. Guisnet, J. Cat. 2005, 230, 375-383.
20] M. L. M. Bonati, R. W. Joyner, M. Stockenhuber, Microporous Mesoporous Mater. 2007, 104, 217-
224.
[
21] Z. Chen, W. Chen, T. Tong, A. Zeng, J. Mol. Catal. A: Chem. 2015, 396, 231-238.