Investigation of the reaction kinetics of isolated Lewis acid sites in Beta zeolites for the Meerwein-Ponndorf-Verley reduction of methyl levulinate to γ-valerolactone

Article abstract of DOI:10.1016/j.jcat.2014.10.010

We investigate the reaction kinetics of the Meerwein-Ponndorf-Verley (MPV) reduction of methyl levulinate (ML) to 4-hydroxypentanoates and subsequent lactonization to γ-valerolactone (GVL) catalyzed by Lewis acid zeolites. Reaction kinetics studies show a first-order dependence on ML and 2-butanol, confirm that the hydride shift is the rate-limiting step, and support a dual-binding mechanism on a single Lewis acid site. All catalysts generate GVL with selectivities >97%, with Hf-Beta exhibiting the highest activity in the temperature range of 393-453 K. Sn-, Zr-, and Hf-Beta show apparent activation energies of ca. 52 kJ mol¹, which is significantly lower than that of Ti-Beta (69 kJ mol¹). Secondary alcohols consistently exhibit higher reaction rates than primary alcohols with lower apparent activation energies. Increasing polarity of the hydrogen donor leads to a decrease in reaction rates. The experimental data are used to build a kinetic model for the MPV reaction in a tubular packed-bed reactor.

Full text of DOI:10.1016/j.jcat.2014.10.010

Journal of Catalysis 320 (2014) 198–207
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Investigation of the reaction kinetics of isolated Lewis acid sites in Beta
zeolites for the Meerwein–Ponndorf–Verley reduction of methyl
levulinate to c-valerolactone
⇑
Helen Y. Luo, Daniel F. Consoli, William R. Gunther, Yuriy Román-Leshkov
Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 29 August 2014
Accepted 19 October 2014
We investigate the reaction kinetics of the Meerwein–Ponndorf–Verley (MPV) reduction of methyl
levulinate (ML) to 4-hydroxypentanoates and subsequent lactonization to -valerolactone (GVL)
c
catalyzed by Lewis acid zeolites. Reaction kinetics studies show a first-order dependence on ML and
2-butanol, confirm that the hydride shift is the rate-limiting step, and support a dual-binding mechanism
on a single Lewis acid site. All catalysts generate GVL with selectivities >97%, with Hf-Beta exhibiting the
highest activity in the temperature range of 393–453 K. Sn-, Zr-, and Hf-Beta show apparent activation
energies of ca. 52 kJ molꢀ1, which is significantly lower than that of Ti-Beta (69 kJ molꢀ1). Secondary
alcohols consistently exhibit higher reaction rates than primary alcohols with lower apparent activation
energies. Increasing polarity of the hydrogen donor leads to a decrease in reaction rates. The experimental
data are used to build a kinetic model for the MPV reaction in a tubular packed-bed reactor.
Ó 2014 Elsevier Inc. All rights reserved.
Keywords:
Liquid-phase transfer hydrogenation
c
-Valerolactone
Lewis acids
Zeolites
Reaction kinetics
1. Introduction
The reaction mechanism involves a cyclic six-membered transition
state where the carbonyl and the alcohol are both coordinated to
Biomass conversion to sustainable chemicals and fuels requires
the development of efficient catalytic processes capable of convert-
ing highly complex and oxygenated precursors to platform mole-
cules. Ideally, selective deoxygenation and hydrogenation must
occur under mild conditions while simultaneously minimizing unit
operations. Transfer hydrogenations are versatile reactions that
avoid the use of high pressures of molecular hydrogen by using
molecules capable of donating hydrogen atoms in the liquid phase
[1–5]. In particular, the Meerwein–Ponndorf–Verley (MPV)
reaction, wherein an alcohol donates a hydride to an aldehyde or
ketone, has shown promising results in several biomass-relevant
reaction schemes. Examples include the conversion of the
following: crotonaldehyde to crotyl alcohol [6]; 5-(hydroxymethyl)
furfural to 2,5-bis(ethoxymethyl)furan [7]; levulinates to c-valero-
lactone [8,9]; and intramolecular hydride transfers in hexoses,
pentoses, and other sugars [10–19].
Traditionally, homogenous Lewis acids catalyze the MPV
reaction as shown for aluminum alkoxides [20,21] in an isopropanol/
ketone system, for lanthanide complexes in the reduction of aryl
ketones [22], and more recently for actinide isopropoxides [23].
the same metal center. The reaction proceeds through a direct
hydride shift where the hydrogen on the C-1 position of the alcohol
is directly transferred to the carbonyl carbon of the ketone/
aldehyde. To overcome the problems associated with using
homogenous catalysts, such as expensive separations and moisture
sensitivity, many heterogeneous catalysts have been studied for
the MPV reaction. These include metal oxides [24,25], hydrotal-
cites [26,27], and zeolites [28–32]. In particular, zeolites containing
Lewis acid centers have shown very high activity and selectivity
under mild reaction conditions [33–36]. Lewis acidic metal centers
tetravalently incorporated into a silicate framework can coordinate
with and polarize carbonyl functional groups [37,38]. It is proposed
that these metal centers coordinate the ketone and alcohol to form
the same six-membered transition state as that formed with
homogenous MPV catalysts. Various Lewis acid zeolites are
capable of catalyzing the MPV reaction including Al-, Ti-, Sn-, and
Zr-containing zeolites.
Recently, we reported on the one-pot conversion of xylose-
derived furfural into
c-valerolactone (GVL) using the combined
action of zeolites featuring Lewis and Brønsted acid sites [8]. GVL
is a versatile platform molecule that has the functionality and
reactivity to be upgraded to fuel additives, fuels, and commodity
chemicals [39–43]. GVL is stable, non-toxic and has shown promise
as an effective solvent for sugar production by breaking down
⇑
Corresponding author.
E-mail address: yroman@mit.edu (Y. Román-Leshkov).
http://dx.doi.org/10.1016/j.jcat.2014.10.010
0021-9517/Ó 2014 Elsevier Inc. All rights reserved.

H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
199
lignocellulosic biomass [44]. In our domino-like reaction sequence,
furfural was reduced to furfuryl alcohol (FA) via an MPV reaction
with 2-butanol catalyzed by a Lewis acid; next, FA was converted
to levulinic acid (LA) through a hydrolytic ring opening catalyzed
by a Brønsted acid; finally, LA was reduced via a second MPV step
to 4-hydroxypentanoate (4HP), which underwent a spontaneous
lactonization to form GVL and water. GVL selectivity values always
exceeded 94%, but it was observed that Ti-, Sn-, and Zr-Beta
catalysts featured drastically different turnover rates for the second
MPV step, with Zr-Beta having the highest activity. Assary et al.
showed with computational techniques that stronger Lewis acids
better stabilize the six-membered transition state for the rate-
limiting hydride shift step [45]. Indeed, changing the identity of
the framework heteroatom can alter the Lewis acid character of
the zeolite; however, Lewis acidity is also highly dependent on
the nature of the solvent and reacting molecules. A difficult chal-
lenge exists in correctly predicting which catalyst will be optimal
for a specific substrate/solvent combination. For example, Boronat
et al. showed that while Sn-Beta is more active for the reduction
of cyclohexanone in 2-butanol, Zr-Beta is more active for the
reduction of benzaldehyde in 2-butanol under identical reaction
conditions [46]. The nature of the hydrogen donor also had a
dramatic impact on the reaction rates. As such, establishing a robust
kinetic framework that describes the governing parameters of the
ML to GVL transformation over a wide range of conditions is of great
relevance to understand, and ultimately predict, the performance of
Lewis acid zeolites in transfer hydrogenation reactions.
In this work, we present an experimental study aimed at
extracting relevant kinetic parameters that describe the catalytic
performance of Lewis acid catalysts during the reduction of ML
to form GVL. We propose a set of elementary steps, derive a rate
expression for the overall reaction, and perform reactivity mea-
surements to confirm reaction orders and identify the rate-limiting
step in the mechanism. We quantitatively compare the kinetic
parameters of Ti-, Sn-, Zr-, and Hf-Beta catalysts over a wide range
of temperatures and find that Hf-Beta has the highest activity of all
catalysts. We also study the effect of varying the hydrogen donor
by determining the kinetic parameters when using primary and
secondary alcohols with varying carbon chain lengths.
to a Teflon-lined stainless steel autoclave and heated to 413 K in a
static oven for 20–40 days. The solids were recovered by filtration,
washed extensively with water and ethanol, and dried at 373 K
overnight. The solids were calcined at 923 K for 10 h with a 1 K/
min ramp and 1 h stops at 423 K and 623 K at a flow rate of
300 ml minꢀ1 of dry air (Airgas, ultra zero grade) to remove the
organic content in pores of the crystalline material. After calcina-
tion, the solid yield was 80–90%. The other metal precursors used
were as follows: tin(II) chloride dihydrate (Sigma–Aldrich, 99.99%
(w/w)), zirconium(IV) oxychloride octahydrate (Sigma–Aldrich,
99.5% (w/w)), and titanium(IV) isopropoxide (Sigma Aldrich,
99.999% (w/w)). All the catalysts were synthesized to achieve a sil-
icon/metal ratio of ca. 100.
2.2. Catalyst characterization
The crystal structures of Beta zeolite catalysts were determined
from powder X-ray diffraction (PXRD) patterns collected using a
Bruker D8 diffractometer using Cu K
a radiation (40 kV, 40 mA).
Nitrogen adsorption and desorption isotherms were measured on
a Quantachrome Autosorb iQ apparatus at liquid nitrogen temper-
ature (77 K). Prior to the adsorption analysis, all samples were pel-
leted and degassed under vacuum for 12 h at 623 K. Micropore N2
uptake was recorded at P/P0 = 0.01, and total pore volume was
recorded at P/P0 = 0.95. Ultraviolet–visible (UV–vis) spectra were
recorded using a Cary 5000 (Varian) instrument equipped with a
Praying Mantis diffuse-reflectance accessory (Harrick Scientific
Products) on calcined samples using a barium sulfate blank.
Reflectance measurements were converted to absorbance using
the Kubelka–Munk function. Elemental analysis was performed
with
a CCD-based inductively coupled plasma (ICP) atomic
emission spectrometer (Activa-S, HORIBA Scientific). Samples were
dissolved in 48% HF and diluted into 3% HNO3 before analysis.
A 5-point calibration curve was built using the following ICP
standards: 1000 ppm Zr in 3% HNO3/trace HF, 1000 ppm Sn in
10% HCl, 1000 ppm Ti in 2% HNO3/trace HF, 1000 ppm Hf in 5%
HNO3/trace HF (all TraceCERTÒ) on the following spectral lines:
327.305 nm Zr line, 189.989 nm Sn line, 339.978 nm Hf line,
336.121 nm Ti line.
Fourier transform infrared (FT-IR) spectra were collected with a
Bruker Vertex 70 spectrophotometer equipped with an Hg–Cd–Te
(MCT) detector. Each spectrum was recorded by averaging 128 scans
at 2 cmꢀ1 resolution in the 4000–400 cmꢀ1 range. Zeolite-beta cat-
2. Experimental methods
2.1. Catalyst synthesis
alysts were pressed into self-supporting wafers (8–10 mg cmꢀ2
)
Lewis acid zeolites with the Beta topology were synthesized in
fluoride media following the procedure outlined by Corma et al.
[47], using different heteroatom metal precursors. For example,
Hf-Beta was prepared as follows: 27.16 g of tetraethylammonium
hydroxide (Sigma–Aldrich, 35% (w/w) in water) and 23.97 g of tet-
raethylorthosilicate (Sigma–Aldrich, P99% (w/w)) were added to
an open Teflon jar. The mixture was magnetically stirred at room
temperature for 90 min. An additional 15 ml of deionized water
was added and the mixture was cooled in an ice bath. Then,
0.37 g of hafnium(IV) chloride (Sigma–Aldrich, 98% (w/w)) was
dissolved in 2 ml of ethanol, and this solution was added dropwise
to the mixture while stirring on ice. The solution was left uncov-
ered while stirring to complete the hydrolysis of TEOS, evaporate
the ethanol, and reach approximately 2 g of water above the target
water content. Then, 2.62 g of hydrofluoric acid (Sigma–Aldrich,
48% (w/w) in water) was added dropwise and the mixture was
homogenized using a PTFE spatula, resulting in a thick paste. Next,
0.36 g of previously-made Hf-Beta was sonicated in 2 ml deionized
water and added into the mixture as seeds. The mixture was
homogenized and allowed to evaporate to a final weight of
33.96 g. The final molar composition of the gel was 1 SiO2/0.01
HfCl4/0.56 TEAOH/0.56 HF/7.5 H2O. The thick paste was transferred
that were sealed within a high temperature cell (Harrick Scientific)
with ZnSe windows. Zeolite wafers were calcined in flowing air (Air-
gas, zero grade, 25 ml minꢀ1) at 773 K for 8 h, evacuated at 573 K for
3 h (<0.01 Pa, dynamic vacuum, Edwards’ T-Station 75 Turbopump),
and cooled to 298 K in vacuum. A reference spectrum was acquired
before dosing with excess CD3CN (Sigma Aldrich, 99.8% (atom D))
vapor under static vacuum. Once dynamic vacuum was
re-established, the difference spectrum was acquired.
2.3. Kinetic studies of MPV reduction of levulinates
Flow reactions were conducted in a 0.46 cm I.D. tubular stainless
steel reactor mounted inside a 30 cm long aluminum block within
an insulated single-zone furnace (Applied Test Systems Series
3210, 850W/115 V). The catalyst was pelleted and sieved to obtain
75–150
lm particles. The catalyst particles were then diluted into 5
times their weight in inert
a-aluminum oxide (Sigma Aldrich, >99%
(w/w)) with the same particle size range, creating a bed approxi-
mately 2.5 cm long. The bed was loaded between glass wool plugs
and supported by additional
a-aluminum oxide. The reaction
temperature was monitored by a K-type thermocouple (Omega,
inconel) placed inside the bed and a PID temperature controller

200
H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
(Cole-Parmer, Digi-Sense Advanced Temperature Controller Series
89000). Prior to reaction, the packed bed was dried under flowing
N2 gas (Airgas, grade 5, 300 ml minꢀ1) at 423 K for 2 h. Liquid-phase
reactants were introduced with a Waters 515 HPLC pump, and the
entire system was pressurized under 15–20 bar of N2 using a back-
pressure regulator (Swagelok). The effluent was collected in a high-
pressure separator (Jerguson Gage and Valve Co., Series R20) at
room temperature.
3. Results and discussion
3.1. Synthesis and characterization of Ti-, Sn-, Zr-, and Hf-containing
Beta zeolites
Characterization data of all Beta catalysts are given in Table 1.
X-ray diffraction patterns (Fig. S1, Supporting Information) of Ti-, Sn-,
Zr-, and Hf-Beta were consistent with the BEA topology. The
asymmetry of the peak in the range of 2h = 7–9° in the Hf-Beta
and Zr-Beta powder patterns indicates an increase in the relative
proportion of the polymorph B in the crystal structure, an effect
previously reported for Zr-Beta by Zhu et al. [32]. No peaks due to
metal oxide or other crystalline impurities were detected. The
micropore volumes of all catalysts, calculated from the nitrogen
adsorption isotherms at 77 K (Fig. S3, Supporting Information),
were consistent with the Beta topology measuring ca. 0.20 cm3 gꢀ1
at P/P0 = 0.01. The slight variations in total pore volume and exter-
nal surface area indicate slight differences in crystal size and mor-
phology between the samples. Elemental analysis by ICP-AES shows
Si/M ratios in the range of 108–121, close to the theoretical value of
100 used in the synthesis gels. Synthesizing these zeolites in fluoride
media has been shown to produce defect-free frameworks, which
translate to pore structures with high hydrophobicity [10].
For sample quantification, a known amount of 1,3,5-tri-tert-
butylbenzene (Sigma Aldrich, 97% (w/w)) in the same solvent as
the effluent was added as an external standard. Liquid samples were
analyzed by gas chromatography (Agilent 6890N) equipped with a
methoxy-siloxane capillary column (HP-1, 30 m ꢁ 0.25 mm ꢁ
0.25 lm) connected to a flame ionization detector. All mass balances
exceeded 97% for all experiments. The conversions are determined
based on initial and final ML concentrations, and the reaction rates
are calculated based on GVL yield. In the runs where methyl levuli-
nate is used as a solvent, the rates and conversions are based on the
alcohol reactant. Conversion is defined as follows:
CML
CML;0
X ¼ 1 ꢀ
ð1Þ
ð2Þ
The presence of isolated Lewis acidic M4+ centers incorporated
in the framework of the zeolites was demonstrated qualitatively
by diffuse-reflectance UV–vis spectroscopy, which shows a single
peak at ꢂ200 nm for each catalyst (Fig. S2, Supporting Informa-
tion). The absence of peaks or shoulders in the 230–250 nm range
indicates a lack of detectable MOx species. Extraframework MOx
species can catalyze the MPV reaction, but do so at rates that are
negligible with respect to the rates observed for framework spe-
cies. IR spectra of Ti- and Sn-Beta collected after saturation with
CD3CN and subsequent evacuation at 298 K (Fig. S4, Supporting
Information) shows bands at ꢂ2308–2316 cmꢀ1 consistent with
C„N stretching vibrations for CD3CN bound to Lewis acidic cen-
ters, and at ꢂ2276 cmꢀ1 for CD3CN bound to silanol groups
[48,49]. These spectra indicate the presence of Lewis acid centers
as well as Brønsted acidic hydroxyl groups in the Beta catalysts.
CX-OH
CX-OH;0
XML;solvent ¼ 1 ꢀ
Yield is defined as follows:
CGVL
CML;0
Y ¼
ð3Þ
ð4Þ
CX¼O
YML;solvent
¼
CX-OH;0
where X is the conversion, Y is the yield, Ci is the molar concentra-
tion (mol dmꢀ3) of species i recovered in a given reactor sample,
and Ci,0 is the initial concentration of reactant fed at a constant vol-
umetric flow rate.
Kinetic studies were conducted in a differential reactor bed in
which levulinate conversion was controlled to be <10%. For some
runs, a small amount of ethyl, propyl, or butyl levulinate (XL)
production was observed when using ML as the reactant, but total
ML conversion was always kept below 10%. The other levulinates
undergo an identical reduction to ML and generate variants
of 4HP, which subsequently lactonize to GVL; consequently, they
are treated as equivalent to ML in the overall reactant pool and do
not affect GVL selectivity negatively. Selectivity is defined as
follows:
3.2. Proposed mechanism for MPV reaction over metal sites and
reaction order studies
The reduction of ML to 4HP and subsequent intramolecular lact-
onization over Hf-Beta in excess 2-butanol (BuOH) at 423 K yields
only one major product, GVL, for ML conversions ranging from 1%
to 100%. The intermediate 4HP could not be detected, which indi-
cates that the intramolecular ring closing is very fast at the reaction
conditions investigated. Abdelrahman et al. showed that in the pres-
ence of Brønsted acidity and temperatures above 323 K, lactoniza-
tion rates are high, leading to high GVL selectivity and the absence
of detectable 4HP intermediates [50]. A small amount of butyl levu-
linate (BL) is observed (<5%), indicating a concurrent acid-catalyzed
esterification of ML with BuOH. This side reaction does not alter the
reaction pathway since BL can be converted to GVL along the exact
same mechanism as that for ML.
CGVL
SGVL
¼
ð5Þ
CML;0 ꢀ CML ꢀ CXL
Volumetric flow rates for the activation energy studies and order
studies ranged from 0.5 to 2.5 ml minꢀ1 corresponding to space
velocities ranging from 0.1 to 103 (mol reactant)(mol metal s)ꢀ1
.
The concentration of reactant was kept at 0.02 mol dmꢀ3 for activa-
tion energy studies and in the range of 0.0075–0.03 mol dmꢀ3 for
reaction order studies. Activation energy experiments were per-
formed in a temperature range of 393–453 K, and all other studies
were performed at 423 K. Reactants and solvents used were as fol-
lows: 2-butanol, 2-propanol, 1-butanol, 1-propanol (Sigma Aldrich,
P99.5% (w/w) anhydrous); ethanol (Koptec, 99.5% (v/v) anhy-
drous); 2-propanol-d8 (Sigma Aldrich, 99.5% (atom D), anhydrous);
methyl levulinate (Sigma Aldrich, P98% (w/w)). Solvents were
dried with activated 4 Å molecular sieves (Sigma Aldrich, 4–8 mesh
beads) prior to use.
Table 1
Catalyst site and structural characterization.
Catalyst
Si/M
Vads(N2) (cm3 gꢀ1) microporea
Vads(N2) (cm3 gꢀ1) totalb
Ti-Beta
Hf-Beta
Zr-Beta
Sn-Beta
108
121
111
108
0.20
0.20
0.20
0.20
0.33
0.30
0.29
0.34
a
N2 volume at end of micropore filling transition (77 K) at P/P0 = 0.01.
N2 volume at P/P0 = 0.95.
b

H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
201
ꢃ
kf;HKML KBuOHCMLCBuOH
r0 ¼
D
D ¼ 1 þ KMLCML þ KBuOHCBuOH þ KGVLCGVL þ KH OCH O þ K4HPC4HP
2
2
þ KMeOHCMeOH þ KBuOCBuO
ð6Þ
In this equation, Ci terms are liquid-phase concentrations
(mol dmꢀ3) for species i, and Ki are the equilibrium constants (dm3 -
molꢀ1) relating liquid-phase species to species that are singly
ꢃ
bound to the Lewis sites (Scheme A2). KML is the equilibrium con-
stant relating free ML to ML bound specifically to sites that already
have a bound BuOH molecule. kf,H is the forward rate constant (sꢀ1
)
for the hydride shift.
In excess of BuOH, we can assume BuOH⁄ constitutes the major-
ity of bound species on the acid sites. Therefore, the third term in
the denominator dominates over the others, leading to the further
simplified expression (Eq. (7)), which shows first-order depen-
dence on ML concentration. The expression for the apparent rate
constant is given in Eq. (8), and the expression for the apparent
activation energy (Ea) consisting of the activation energy of the
hydride shift and the enthalpy of adsorption of ML onto an already
occupied site, is given in Eq. (9).
r0 ¼ ðkf;HKML ÞCML
ð7Þ
ð8Þ
ð9Þ
ꢃ
ꢃ
kapp;BuOH ¼ kf;HKML
ꢃ
Ea;app;BuOH ¼ Ea;H
þ
DHads;ML
Conversely, in excess ML, we can assume ML⁄ constitutes the
majority of bound species on the acid sites. Therefore, the second
term in the denominator dominates over the others, leading to
the further simplified expression (Eq. (10)), which shows
first-order dependence on BuOH concentration. If we assume that
Scheme 1. Proposed catalytic cycle for the MPV reduction of methyl levulinate
(ML) to 4-hydroxypentanoate (4HP) with 2-butanol (BuOH) over partially
ꢃ
KML is approximately equal to KML , then the apparent rate constant
a
hydrolyzed metal site in zeolite framework (⁄). BuOH adsorbs and deprotonates
(1), ML then adsorbs (2) and converts to 4HP through a hydride shift in a six-
membered transition state (3), 2-butanone (BuO) desorbs (4), 4HP receives a proton
from the catalyst and desorbs (5), and 4HP undergoes spontaneous lactonization
producing GVL and methanol (MeOH) (6).
in excess ML is simplified to Eq. (11). This gives an expression for
the apparent Ea (Eq. (12)), which consists of the activation energy
of the hydride shift and the enthalpy of adsorption of 2-BuOH onto
an unoccupied site.
ꢀ
ꢁ
ꢃ
kf;HKML KBuOH
r0 ¼
CBuOH
ð10Þ
KML
A proposed catalytic cycle for the MPV reaction of ML with
BuOH over a single catalytic center in Beta zeolite based on a
set of elementary steps is shown in Scheme 1. The catalytic cycle
is consistent with previously proposed mechanisms for the MPV
reaction [46]. The active Lewis acid site is hypothesized to be a
partially hydrolyzed ‘‘open site’’ where the metal is coordinated
to three OASi framework groups and one hydroxyl group [48].
The reaction proceeds in the following steps: adsorption and
deprotonation of BuOH over the Lewis acid center (step 1);
adsorption of ML (step 2) and hydride shift from deprotonated
BuOH to ML through a six-membered cyclic transition state (step
3); desorption of 2-butanone (BuO) (step 4); and hydrogen trans-
fer from the catalyst and desorption of 4HP (step 5). 4HP then
kapp;ML ¼ kf;HKBuOH
Ea;app;ML ¼ Ea;H
ð11Þ
ð12Þ
þ
DHads;BuOH
The initial turnover frequencies were measured in terms of GVL
production normalized by total metal content of zeolite, at an ML
conversion <10%. As shown in Fig. 1, first-order dependencies of
initial rates to ML concentration in solvent excess of BuOH and
of BuOH concentration in solvent excess of ML are confirmed by
slope values of the log rate vs. log initial concentration plots that
approach unity. These first-order dependencies support our pro-
posed mechanism where both reactant molecules are coordinated
to a single catalytic center. These data are in agreement with
several computational studies on the MPV reaction over Beta
zeolites using density functional theory [45,46], as well as the estab-
lished mechanism for MPV reaction on Al(III) isopropoxide [20].
undergoes
a spontaneous intramolecular ring-closure and a
Brønsted acid-catalyzed release of methanol (MeOH) to form
GVL (step 6). Based on computational studies, the hydride shift
is the slowest step in the mechanism and constitutes the
rate-determining step. We further assume that all adsorption
and desorption steps are in quasi-equilibrium, that all species in
the reaction including small amounts of residual water (H2O)
can act as singly-bound inhibitors and that the intramolecular
lactonization of 4HP is very fast and irreversible (Scheme A2,
Appendix A). The fast lactonization step results in a product sink
making the concentration of 4HP very small, and consequently,
the reverse reaction of step 3 (i.e., 4HP back to ML) is negligible.
A rate expression for the rate of GVL production normalized per
catalytic site (mol (mol metal s)ꢀ1) is given by Eq. (6).
3.3. Exclusion of mass transport limitations and confirmation of rate-
determining step
Measured initial rates would also show an apparent first-order
dependence on ML or BuOH concentration for a process limited by
the diffusion of reactants to active sites located within the zeolite
pores. To rule out intercrystalline heat and mass transport artifacts,
the linear velocity in the reactor was varied while maintaining a
constant space velocity at temperatures of 393, 423, and 453 K.

202
H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
(KIE) using 2-propanol-H8 and 2-propanol-D8 as the solvent. The
-2.5
-3.0
-3.5
-4.0
-4.5
-5.0
a
Slope = 0.954 0.008
R2 = 0.999
kapp,H/kapp,D ratio at 423 K was 2.01 0.33 (Table S1, Supporting
Information). This KIE value indicates that the elementary steps
for adsorption and deprotonation do not limit rates of ML hydroge-
nation prior to the kinetically-relevant hydride shift. Incorporation
of the deuteron in the C-4 position in the GVL product was
confirmed by 1H NMR (Fig. S7, Supporting Information). The same
effect can be seen when performing the experiment in excess ML
where the kapp,H/kapp,D ratio at 423 K was 1.90 0.10 (Table S1,
Supporting Information). Previous studies on the MPV reduction
catalyzed by homogenous aluminum(III) complexes have
yielded experimental KIE values of 2.3 and 3.0 [20,51]. The exact
nature of the vibrational mode of the bimolecular six-membered
transition state is difficult to determine for heterogeneous catalytic
sites even with density functional theory (DFT) given the need to
use tight convergences for the force constants and appropriate
dispersive interactions to locate the transition states. At the given
-5.5
-5.0
-4.5
-4.0
-3.5
-3.0
Ln(ML concentration) / mol dm-3
temperatures, assuming the transition state oscillation is
a
combination of CAH bond scissoring (ꢂ1400 cmꢀ1) and CAH bond
stretching (ꢂ2900 cmꢀ1), a theoretical KIE value between 1.9 and
3.6 is expected (see derivation in section A, Supporting
Information). Given the absence of mass transfer limitations, our
experimental KIE values imply that the bending vibrational mode
is dominant in the transition state, possibly due to confinement
in the narrow zeolite pores.
b
-5.5
-6.0
-6.5
-7.0
-7.5
-8.0
Slope = 0.992 0.070
R2 = 0.985
3.4. Temperature effects
Having established reaction orders and ruled out heat and mass
transfer limitations, the apparent rate constant kapp and Arrhenius
parameters were explicitly determined under differential reaction
conditions in a temperature range of 393–453 K (Table 2). The
apparent activation energy for the reaction over Hf-Beta catalyst
in excess of BuOH is 52.5 1.9 kJ molꢀ1 and the pre-exponential
factor is 4.1 ꢁ 106 (mol dm3)(mol metal s mol ML)ꢀ1. This apparent
activation energy agrees well with that found for levulinic
acid hydrogenation in liquid water using Ru-based catalysts
(48 kJ molꢀ1) [50]. In addition, the experimental value matches
closely with the apparent activation energy of 53.9 kJ molꢀ1 calcu-
lated by Assary et al. for ethyl levulinate in isopropanol solvent over
Zr-Beta catalyst [45]. We would expect Hf and Zr sites to behave
very similarly due to their comparable electronic properties and
atomic sizes. The activation energy for the reaction over Hf-Beta cat-
alyst in excess ML is 44.8 1.4 kJ molꢀ1, and the pre-exponential fac-
tor is 3.8 ꢁ 104 (mol dm3)(mol metal s mol BuOH)ꢀ1. The decrease in
apparent activation energy is attributed to the stronger adsorption
of ML on the metal site compared to the adsorption of BuOH on
the metal site based on the simplified rate expressions (Eqs. (7)
and (10)). This trend agrees with the theoretical difference in
adsorption enthalpies of ML and BuOH of 14.2 kJ molꢀ1 calculated
by Assary et al. [45]. In addition, the two order of magnitude
difference in the pre-exponential factors when switching solvents
leads to a large difference in the free energies of adsorption, which
is reflected by an order of magnitude decrease in the apparent rate
constant at 423 K. More specifically, kapp is 0.110 (mol dm3)
(mol metal s mol BuOH)ꢀ1 when operating in excess ML compared
to 1.33 (mol dm3)(mol metal s mol ML)ꢀ1 when operating in excess
BuOH.
-5.5
-5.0
-4.5
-4.0
-3.5
-3.0
Ln(2-butanol concentration) / mol dm-3
Fig. 1. Log–log plot of initial rates vs. methyl levulinate (ML) concentration in
solvent excess of 2-butanol (BuOH) (a), and BuOH concentration in solvent excess of
ML (b) for Hf-Beta catalyst. Rates are measured in moles of GVL produced per
second normalized per total mole metal center in the catalyst. In excess ML, rates
are measured in moles of 2-butanone produced per second normalized per total
metal center in the catalyst. The standard error of regression is calculated assuming
errors are normally distributed and the sum of square residuals is v2 distributed
with n ꢀ 2 degrees of freedom. All liquid-phase flow reactions were performed
under differential conditions in a packed-bed reactor pressurized under nitrogen at
15 bar with a flow rate of 1.0 ml minꢀ1, at temperature 423 K, with feed concen-
trations of 0.0075–0.03 mol dmꢀ3 reactant.
As shown in Fig. S5, the rate of GVL production remains constant
(within experimental error) for all linear velocities at the full range
of temperatures, indicating that no external mass transfer
limitations exist at these conditions. Next, to ensure the absence
of intracrystalline heat and mass transport artifacts, turn over
frequencies for Hf-Beta zeolites with varying Si/M ratios were mea-
sured and compared. Initial rates were determined for Hf-Beta200
(Si/Hf ratio of 190, as confirmed by ICP-AES elemental analysis)
and Hf-Beta300 (Si/Hf ratio of 336) under identical kinetic reaction
conditions to those used for the original Hf-Beta catalyst at 393, 423,
and 453 K. As shown in Fig. S6, the rates normalized per total metal
site are constant for all temperatures tested, thus confirming that
the reaction conditions used in this study satisfy the Koros–Nowak
criterion and do not exhibit internal heat or mass transport limita-
tions. All kinetic data were collected in the absence of significant
catalyst deactivation, as confirmed by replicating initial runs in a
series and obtaining identical activity values.
3.5. Comparison of Ti, Sn, Zr, Hf-Beta catalysts
The MPV reduction of ML has previously been shown to occur
with high activity and selectivity over Sn- and Zr-Beta catalysts in
batch reactions. Although previous studies in liquid-phase batch
reactors report very low conversions for Ti-Beta (ꢂ1%), we now
quantitate its turnover rate for the MPV reaction [8]. Our results
To support our hypothesis that the rate-limiting step of the
reaction is the hydride shift, we measured the kinetic isotope effect

H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
203
Table 2
Measured first-order rate constants, activation energies and pre-exponential factors for MPV reduction of methyl levulinate (ML) in hydrogen-donating solvent 2-butanol (BuOH)
with various zeolite-beta catalysts. Rates are measured in moles of GVL produced per second normalized per total mole metal center in the catalyst. All liquid-phase flow
reactions were performed in a packed-bed reactor pressurized under nitrogen at 20 bar with a flow rate range of 0.5–2.5 ml/min, with temperature range of 393–453 K.
Run Catalyst Feed
(mol dmꢀ3
Solvent Rate constant kapp at
Apparent Ea
Standard errorb
Pre-exponential
factor Ac
Standard multiplicative
errorb,c
)
423 Ka
(kJ molꢀ1
)
(kJ molꢀ1
)
1
2
3
4
5
Hf-Beta 0.02 ML
Hf-Beta 0.02 BuOH
Zr-Beta 0.02 ML
Sn-Beta 0.02 ML
Ti-Beta 0.02 ML
BuOH
ML
BuOH
BuOH
BuOH
1.33E+0
1.10Eꢀ1d
6.84Eꢀ1
1.92Eꢀ1
5.26Eꢀ4
52.5
44.8
51.8
51.7
69.0
1.9
1.4
1.0
2.1
2.2
41.3
0.382e
17.3
5.1
1.7
1.5
1.3
1.8
1.9
1.8
a
b
c
Units: (mol dm3)(mol metal s mol ML)ꢀ1
.
Standard error of regression assuming errors are normally distributed, sum of square residuals is v2 distributed with n ꢀ 2 degrees of freedom.
Units: 105 (mol dm3)(mol metal s mol ML)ꢀ1
.
d
e
Units: (mol dm3)(mol metal s mol BuOH)ꢀ1
.
Units: 105 (mol dm3)(mol metal s mol BuOH)ꢀ1
.
show that Hf-Beta converts ML to GVL in BuOH with high selectivity
and with a rate constant (1.33 (mol dm3)(mol metal s mol ML)ꢀ1
temperature range (Table 2). Although the differences in the
experimental pre-exponential factors are not statistically signifi-
cant to 90% confidence (Table S2, Supporting Information), we
expect to see differences in the free energy of binding and reac-
tion depending on the identity of the heteroatom. Changes in site
geometry or energetics likely restrict available conformations of
bound intermediates and the transition state. In addition, varying
binding strengths may contribute to differences in vibrational
frequencies of bound species. These differences in free energy
may stabilize the transition from reactants to the six-membered
ring transition state for certain heteroatoms, leading to the
observed differences in reaction rate.
The observed differences in initial rates that exist for catalysts
possessing similar activation barriers could also be a result of dif-
ferent proportions of catalytically active sites with respect to total
metal content. Previous studies of Ti- and Sn-Beta showed that the
distribution of open sites was highly dependent on catalyst
pretreatment such as calcination conditions, as well as initial
)
that is twice as large than that of Zr-Beta (0.684 (mol dm3)
(mol metal s mol ML)ꢀ1), and seven times larger than that of
Sn-Beta (0.192 (mol dm3)(mol metal s mol ML)ꢀ1) at 423 K (Table 2).
The marked differences in initial rates between the catalysts that
arise in spite of the similar apparent activation energy values of
52.5 1.9, 51.8 1.0, and 51.7 2.1 kJ molꢀ1 for Hf-, Zr-, and
Sn-Beta, respectively (Table 2), highlight the need to find the
optimal Lewis acid catalyst for each reaction system. The activation
energies and activities depend on the Lewis acidity of the sites in the
specific reaction–solvent system. Due to the similar apparent
activation energies in these zeolites, commensurate differences in
activity were observed throughout the experimental temperature
range (Fig. 2). However, for Ti-Beta, an apparent activation energy
of 69.0 2.2 kJ molꢀ1 was calculated, which translates to a rate con-
stant that is three orders of magnitude smaller than that of Hf-Beta
at 423 K.
The apparent activation energy consists of the sum of the
activation energy of the hydride shift and the enthalpy of
adsorption of ML onto a site already occupied by a bound BuOH
molecule (Eq. (9)). Our results indicate that the Ea of the hydride
shift and the adsorption enthalpy of ML are either very similar
for Sn-, Zr-, and Hf-Beta or that the differences in Ea and adsorption
enthalpies for each catalyst exactly offset to give the same appar-
ent energies. We would expect Zr- and Hf-Beta to have similar
Ea and adsorption values since Hf and Zr have a similar size and
electronic structure (s–d valence orbitals). However, it is surprising
that Sn-Beta features similar apparent activation energy values,
since Sn has a significantly smaller atomic radius and s–p valence
electron orbitals. Evidently, the barriers for the hydrogen shift
depend on the electronics of the entire reaction site, not just the
metal atom alone. The Ea,H comprises numerous energetic
contributions involving the heteroatom and its effect on the
solvent, reactants and transition state, including the following: site
geometry, charge transfer, and polarizability. In addition, it is
unknown whether the adjacent hydroxyl group of the open site
plays a role in transition state stabilization. Therefore, different
energetic contributions could offset to give the same total apparent
activation energy. Recently, Li et al. highlighted this balancing
effect with an extended QM/MM model of heteroatom-containing
zeolites for glucose–fructose isomerization [52]. It was reported
that although the metal centers and the adjacent hydroxyl groups
in Sn- and Zr-Beta have drastically different polarizabilities and
Brønsted basicities, respectively, their contribution summed to
almost the same apparent activation energy.
2
0
-2
-4
-6
-8
-10
0.0021 0.0022 0.0023 0.0024 0.0025 0.0026
(Temperature-1) / K-1
Fig. 2. Arrhenius plot for the MPV reduction of 0.02 mol dmꢀ3 ML in 2-butanol
solvent with various zeolite-beta catalysts. Hf-Beta (d), Zr-Beta ( ), Sn-Beta ( ), Ti-
Beta ( ). Rates are measured in moles of GVL produced per second normalized per
total mole metal center in the catalyst. All liquid-phase flow reactions were
performed under differential conditions in a packed-bed reactor pressurized under
nitrogen at 20 bar with a flow rate range of 0.5–2.5 ml minꢀ1, with temperature
range of 393–453 K.
Pre-exponential factors for Hf-, Zr-, Sn-, and Ti-Beta are
all within an order of magnitude, but they reflect the difference
in relative reaction rates observed across the experimental

204
H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
exposure to the reaction environment [46,53]. The metal centers
may interconvert between open and closed sites at reaction condi-
tions. Different metal centers may have different MAOASi hydroly-
sis rates, which would lead to different distributions of open and
closed sites between the metals. At this time we cannot
deconvolute these effects in our experimental data. The ability to
quantitatively compare experimental kinetic parameters will
facilitate the search for which fundamental physical properties
are most important in predicting catalyst activity.
previously reported for transfer hydrogenations using primary
and secondary alcohols over supported ruthenium catalysts [56].
Longer alcohol chains increase reaction rates relative to shorter
chains. This effect is observed primarily in the pre-exponential fac-
tors, as the activation energies are nearly the same within the pri-
mary and secondary alcohol groups. The pre-exponential factor of
2-PrOH is an order of magnitude lower than that of BuOH, and a
similar decrease is seen when switching from 1-BuOH to 1-PrOH.
In contrast, there is only a threefold decrease in pre-exponential
factor when switching from 1-PrOH to EtOH. The polarity of the
alcohol, quantitatively represented by its dielectric constant,
increases with decreasing chain length for the non-cyclic alcohols
(Table S3, Supplemental Information). Cy-PeOH has a dielectric
constant that lies in-between that of BuOH and 2-PrOH, and we
see the same trend in the pre-exponential factors. An increase in
solvent polarity most likely results in tighter binding to the Lewis
acid center, and therefore, a decrease in flexibility of the bound
intermediate and transition state. The importance of binding
flexibility is evident in the large decrease in apparent rate constant,
from 1.33 to 0.196 (mol dm3)(mol metal s mol ML)ꢀ1, when using
Cy-PeOH instead of BuOH. The rigid structure of cyclopentanol
limits the number of accessible conformations leading to a higher
penalty for binding. The differences in solvent polarity may also
lead to different partitioning in the hydrophobic zeolite pores
and affect the strength of solvation of ML [57–60]. We note that
the sizes of all studied alcohols are smaller than the size of a Beta
zeolite pore [30]. Therefore, this work captures the trends of
solvents that should be largely unaffected by constriction due to
the pore size. It is important to determine how catalyst activity
is affected by small changes in substrate geometry in order to
increase the flexibility of catalytic processes to adapt to varying
feedstocks.
3.6. Comparison of hydrogen donor properties
We studied the effect of varying the position of the hydroxyl
group and the carbon chain length of the hydrogen donor. Lewis
acidity is highly dependent on the solvent environment, so it is
important to study how solvent-catalyst systems may be
optimized for higher activity [54,55]. The activation energies, pre-
exponential factors, and rate constants at 423 K for the MPV reduc-
tion of ML over Hf-Beta are shown in Table 3. We found the most
dramatic effect on the reaction rates was achieved by switching
from a secondary to a primary alcohol donor (Fig. 3). BuOH, 2-pro-
panol (2-PrOH), and cyclopentanol (Cy-PeOH) have activation ener-
gies of 52.5 1.9, 46.3 2.7, and 56.9 2.2 kJ molꢀ1, respectively,
whereas 1-butanol (1-BuOH), 1-propanol (1-PrOH), and ethanol
(EtOH) have activation energies of 73.5 4.4, 69.2 2.9, and
70.2 4.2 kJ molꢀ1, respectively. We hypothesize that the lower
apparent activation barrier is primarily due to the stabilization of
the hydride shift transition state by the electron-donating alpha
terminal methyl group in secondary alcohols, which leads to a
lower Ea,H. This can be quantitatively represented by the oxidation
enthalpies of the respective alcohols, which clearly segregate
between the primary and secondary alcohols (Table S3, Supporting
Information). The pre-exponential factors for the reactions with pri-
mary alcohols are nearly two orders of magnitude greater than
those for the secondary alcohols; a statistically significant result
at 90% confidence (Table S2, Supporting Information). The absence
of the alpha terminal methyl group likely reduces steric hindrance
on both the bound intermediates and the transition state. Also, pri-
mary alcohols have access to a second C-1 hydrogen that can be
transferred to ML during the six-membered ring transition state.
This contributes both conformational and rotational degrees of
freedom to the free energy along the reaction pathway. However,
the larger pre-
3.7. Kinetic model of packed-bed reactor
An ideal packed-bed reactor kinetic model was developed to
quantitatively describe the reduction of ML in BuOH over an
Hf-Beta catalyst. The reactor was assumed to operate under
isothermal and constant pressure condition, with no radial
gradients in concentration, temperature, or reaction rate. For the
derivation, we used the simplified rate expression (Eq. (13)), which
has first-order dependence on ML concentration (mol dmꢀ3) and
the temperature dependent apparent rate constant kapp
((mol GVL dm3)(mol metal s mol ML)ꢀ1). This apparent rate con-
stant can be calculated from the Arrhenius parameters (Eq. (14))
determined from initial rate experiments listed in Tables 2 and 3.
exponential factors of primary alcohols do not compensate for their
unfavorable activation energies at the temperatures investigated;
the reaction rate constants for primary alcohols are an order of
magnitude lower than their secondary counterparts at 423 K
(Table 3). The higher rates for secondary alcohols have been
r0 ¼ kappCML
ð13Þ
Table 3
Measured first-order rate constants, activation energies and pre-exponential factors for MPV reduction of methyl levulinate (ML) in hydrogen-donating solvents (BuOH, 2-
butanol; 1-BuOH, 1-butanol; 2-PrOH, 2-propanol; 1-PrOH, 1-propanol; EtOH, ethanol; Cy-PeOH, cyclopentanol) with Hf-Beta catalyst. Rates are measured in moles of GVL
produced per second normalized per total mole metal center in the catalyst. All liquid-phase flow reactions were performed in a packed-bed reactor pressurized under nitrogen at
20 bar with a flow rate range of 0.5–2.5 ml/min, with temperature range of 393–453 K.
Run
Catalyst
Feed
Solvent
Rate constant
Apparent Ea
Standard errorb
Pre-exponential
factor Ac
Standard multiplicative
errorb,c
(mol dmꢀ3
)
kapp at 423 Ka
(kJ molꢀ1
)
(kJ molꢀ1
)
1
6
7
8
9
Hf-Beta
Hf-Beta
Hf-Beta
Hf-Beta
Hf-Beta
Hf-Beta
0.02 ML
0.02 ML
0.02 ML
0.02 ML
0.02 ML
0.02 ML
BuOH
1.33E+0
1.64Eꢀ1
7.35Eꢀ1
5.41Eꢀ2
1.44Eꢀ2
1.96Eꢀ1
52.5
73.5
46.3
69.2
70.2
56.9
1.9
4.4
2.7
2.9
4.2
2.2
41.3
1508.7
3.4
189.6
68.8
1.7
3.5
2.1
2.3
3.3
1.9
1-BuOH
2-PrOH
1-PrOH
EtOH
10
Cy-PeOH
20.6
a
b
c
Units: (mol dm3)(mol metal s mol ML)ꢀ1
.
Standard error of regression assuming errors are normally distributed, sum of square residuals is v2 distributed with n ꢀ 2 degrees of freedom.
Units: 105 (mol dm3)(mol metal s mol ML)ꢀ1
.

H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
205
Fig. 4 shows the experimental values superimposed over the
concentration profiles predicted by the model at space velocities
in a range of 10–300 (mol ML0)(mol metal s)ꢀ1 and conversions
ranging from 10% to 80%. There exists good agreement between
the experimental and theoretical values obtained with the
activation energies and pre-exponential factors calculated in previ-
ous sections. As such, we can model the behavior of the ML to GVL
reaction using each of the different hydrogen-donating solvents as
well as each of the different Beta catalysts with high fidelity.
2
1
0
-1
-2
-3
-4
-5
4. Conclusions
Hf-Beta was determined to be a highly active and selective cata-
lyst for the MPV reduction of ML to 4HP and subsequent ring closing
to GVL in 2-butanol. The reaction was confirmed to be first order in
ML in 2-butanol solvent, and first order in 2-butanol in ML solvent.
This supports a dual-binding mechanism where the ketone and
alcohol interact with a single metal site forming a six-membered
transition state. The absence of intercrystalline and intracrystalline
mass and heat transport limitations was confirmed by varying lin-
ear velocities and varying Beta zeolite metal content. A KIE value
of ca. 2 corroborated that the rate-determining step is in fact the
hydride shift, and that the transition state is dominated by bending
rather than stretching vibrational modes. The apparent activation
energy of the reaction over Hf-Beta was 52.5 kJ molꢀ1, which is sim-
ilar to activation energies for LA hydrogenation and computational
studies. We found that Hf-Beta was the most active out of Ti-, Sn-,
and Zr-Beta catalysts. Sn-, Zr-, and Hf-Beta all had very similar acti-
vation energies, while that of Ti-Beta was higher at 69.0 kJ molꢀ1. A
study of varying the hydrogen donor properties showed that sec-
ondary alcohols are much more active than primary alcohols with
activation energies that are ca. 20 kJ molꢀ1 lower, most likely due
to transition state stabilization of the C-1 position during the
hydride shift. In addition, decreasing chain length of the alcohol
did not have an effect on the activation energies but did decrease
the pre-exponential factors leading to three- to ten-fold decreases
in rates in the experimental temperature range. We developed a
kinetic model using the experimentally determined activation ener-
gies and pre-exponential factors to accurately predict conversions
in a packed-bed reactor.
-6
0.0021 0.0022 0.0023 0.0024 0.0025 0.0026
(Temperature-1) / K-1
Fig. 3. Arrhenius plot for the MPV reduction of ML with Hf-Beta using different
hydrogen donors (BuOH, 2-butanol; 1-BuOH, 1-butanol; 2-PrOH, 2-propanol;
1-PrOH, 1-propanol; EtOH, ethanol; Cy-PeOH, cyclopentanol). 0.02 mol dmꢀ3 ML
in 2-BuOH ( ), 0.02 mol dmꢀ3 ML in 1-BuOH ( ), 0.02 mol dmꢀ3 ML in 2-PrOH
(
), 0.02 mol dmꢀ3 ML in 1-PrOH
(
), 0.02 mol dmꢀ3 ML in EtOH
(
),
0.02 mol dmꢀ3 ML in Cy-PeOH ( ). Rates are measured in moles of GVL produced
per second normalized per total mole metal center in the catalyst. All liquid-phase
flow reactions were performed under differential conditions in
a
packed-bed
a flow rate range of
reactor pressurized under nitrogen at 20 bar with
0.5–2.5 ml minꢀ1, with temperature range of 393–453 K.
This study provides the first quantification of kinetic parame-
ters for the MPV reduction of liquid-phase substrates on zeolite
Lewis acid catalysts. In addition, we provide a fully quantitative
comparison of Ti-, Sn-, Zr-, and Hf-Beta initial rate constants,
apparent activation energies, and pre-exponential factors for the
reduction of ML in 2-butanol. Experimentally quantifying the
effects of varying Lewis acid metal centers and substrates is crucial
for the design of effective heterogeneous catalysts for applications
in biomass conversion.
Fig. 4. Comparison between experimental and model-predicted conversion of ML
to GVL using 2-butanol with Hf-Beta catalyst at various space velocities. Liquid-
phase flow reactions were performed in a packed-bed reactor pressurized under
nitrogen at 15 bar with a flow rate range of 0.3–5.0 ml minꢀ1, 0.02 mol dmꢀ3 ML at
423 K.
Acknowledgments
The authors acknowledge financial support for this work from
the U.S. Department of Energy, Office of Science, Office of Basic
Energy Sciences, Chemical Sciences, Geosciences and Biosciences
Division under Grant No. DE-FG02-12ER16352. We also thank
Marco B. Wonink and Anthony Crisci (Massachusetts Institute of
Technology, Chemical Engineering) for their support in building
the infrared spectroscopy in situ dosing system.
ꢀ
ꢁ
Ea;app
RT
kapp ¼ Aapp exp
ð14Þ
The concentration profile for a packed-bed-reactor is given in
Eq. (15). The detailed derivation for this equation is shown in sec-
tion C of the Supporting Information.
ꢀ
ꢁ
Appendix A. Abridged derivation of ML MPV turnover rate
expression
CML;0kapp
SV
X ¼ 1 ꢀ exp
ð15Þ
X is the conversion in terms of GVL production, SV is the space
velocity (mol ML0)(mol metal s)ꢀ1, and CML,0 is the initial concentra-
tion of ML.
Here, we show an abridged derivation of the rate expression
for GVL production (Eq. (6)) from the MPV reduction of ML with
2-butanol (BuOH) as the hydrogen donor. The sequence of

206
H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
of 4HP to be close to 0, which is supported by GC analysis of
KB
u
OH
reaction solutions. This means the concentration of bound
intermediate 4HP species is also close to 0. This allows us to
simplify Eq. (A1) to only the forward reaction. Replacing bound
intermediates with solution phase species using the equilibrium
Eqs. (A2) and (A3), we obtain a rate expression in terms of free
active sites.
1
2
u
+
B OH*
u
u
+
+
ML
B
H
O
*
KML*
u
-
B
B
H ML*
O
B
H*
O
ML
kfH
kr
u
O
-
u
O
-
H ML*
4HP*
B
3
4
H
ꢃ
rrds ¼ kf;HKML KBuOHCMLCBuOH
C
ðA4Þ
ꢃ
KB
u
O*
Combining the Lewis acid total site balance with the
equilibrium constant expressions and rearranging allows us to
u
-
+
+
u
B
O
B
4HP*
4HP*
4HP
O
K4HP
isolate C .
⁄
4HP*
kfL
5
6
7
*
C
¼ C ð1 þ KMLCML þ KBuOHCBuOH þ KGVLCGVL þ K OC
ꢃtot
ꢃ
H2
H2O
þ K4HPC4HP þ KMeOHCMeOH þ KBuOCBuO
Þ
ðA5Þ
+
e
4HP
VL
G
M
H
O
Substituting the expression for C into (A4) gives us:
⁄
KH2O
H
+
*
H
2O
*
2O
ꢃ
ꢃ
kf;HKML KBuOHCMLCBuOHCtot
rrds
¼
KGVL
KM
D
8
+
*
VL
VL*
G
G
e
D ¼ 1 þ KMLCML þ KBuOHCBuOH þ KGVLCGVL þ KH OCH O þ K4HPC4HP
2
2
e
OH
þ KMeOHCMeOH þ KBuOCBuO
ðA6Þ
9
+
*
e
M H*
O
M
H
O
If we normalize the rate by the concentration of total acid sites,
which is assumed to be the metal content in the zeolite, we get the
expression:
KML
KB
10
+
*
ML
ML*
u
O
ꢃ
+
*
11
kf;HKML KBuOHCMLCBuOH
u
B
u
B
O
*
O
r0 ¼
D
Scheme A2. Proposed elementary steps for the catalytic cycle of the MPV reduction
of methyl levulinate (ML) to 4-hydroxypentanoate (4HP) with 2-butanol (BuOH)
over a partially hydrolyzed metal site in zeolite framework (⁄). Singly-bound species
can reversibly deactivate the catalytic site as shown in equations steps 7–11.
Doubly-bound inhibitors are assumed to be negligible.
D ¼ 1 þ KMLCML þ KBuOHCBuOH þ KGVLCGVL þ KH OCH O þ K4HPC4HP
2
2
þ KMeOHCMeOH þ KBuOCBuO
ðA7Þ
This rate has units of (mol (mol metal)ꢀ1 sꢀ1). Further simplifi-
cations of this rate expression when using a large excess of either
BuOH or ML are given in Section 3.2 Eqs. (7)–(12).
elementary steps is shown in Scheme A2. A complete derivation is
given in section B in the Supporting Information. The kinetically-
relevant step is the hydride shift (step 3) based on the observed
H/D KIE of ꢂ2 at 423 K on Hf-Beta when using 2-propanol-D8
reactant (Section 3.3). All the adsorption and desorption steps are
assumed to be quasi-equilibrated, including steps 7–11, which
allow species to bind and inhibit active sites. These steps are
depicted as single adsorption and desorption steps even though
they may represent a combination of elementary steps, such as
deprotonation to form alkoxides. We also assume that the intramo-
lecular ring closing (step 6) is very fast and essentially irreversible.
This assumption is reasonable given the reaction conditions at
423 K, the presence of Brønsted acidity in the zeolite catalyst,
and no detectable 4HP when analyzing reaction solutions. The rate
expression for the rate-determining step is (mol dmꢀ3 sꢀ1):
Appendix B. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2014.10.010.
References
[1] M.J. Climent, A. Corma, S. Iborra, Green Chem. 16 (2014) 516.
[2] M. Dusselier, M. Mascal, B.F. Sels, Top chemical opportunities from
carbohydrate biomass: a chemist’s view of the biorefinery, in: Top. Curr.
Chem., Springer, Berlin, Heidelberg, 2014, pp. 1–40.
[3] C. Aellig, F. Jenny, D. Scholz, P. Wolf, I. Giovinazzo, F. Kollhoff, I. Hermans, Catal.
Sci. Technol. 4 (2014) 2326–2331.
[4] S. Van de Vyver, J. Geboers, P.A. Jacobs, B.F. Sels, ChemCatChem 3 (2011) 82–
94.
[5] M.J. Climent, A. Corma, S. Iborra, M.J. Sabater, ACS Catal. 4 (2014) 870–
891.
ꢃ
ꢃ
rrds ¼ kf;HCBuOHꢀML ꢀ kr;HCBuOꢀ4HP
ðA1Þ
[6] V.L. Sushkevich, I.I. Ivanova, S. Tolborg, E. Taarning, J. Catal. 316 (2014) 121–
129.
[7] J.D. Lewis, S. Van de Vyver, A.J. Crisci, W.R. Gunther, V.K. Michaelis, R.G. Griffin,
Y. Román-Leshkov, ChemSusChem 7 (2014) 2255–2265.
[8] L. Bui, H. Luo, W.R. Gunther, Y. Roman-Leshkov, Angew. Chem. Int. Ed. Engl. 52
(2013) 8022–8025.
[9] M. Chia, J.A. Dumesic, Chem. Commun. (Camb.) 47 (2011) 12233–12235.
[10] R. Gounder, M.E. Davis, J. Catal. 308 (2013) 176–188.
[11] H.J. Cho, P. Dornath, W. Fan, ACS Catal. 4 (2014) 2029–2037.
[12] M. Moliner, Y. Roman-Leshkov, M.E. Davis, Proc. Natl. Acad. Sci. USA 107
(2010) 6164–6168.
[13] R. Gounder, M.E. Davis, ACS Catal. 3 (2013) 1469–1476.
[14] Y. Román-Leshkov, M. Moliner, J.A. Labinger, M.E. Davis, Angew. Chem. Int. Ed.
49 (2010) 8954–8957.
[15] W.R. Gunther, Y. Wang, Y. Ji, V.K. Michaelis, S.T. Hunt, R.G. Griffin, Y. Román-
Leshkov, Nat. Commun. 3 (2012) 1109.
[16] W.R. Gunther, Q. Duong, Y. Román-Leshkov, J. Mol. Catal. A: Chem. 379 (2013)
294–302.
[17] P.P. Pescarmona, K.P. Janssen, C. Delaet, C. Stroobants, K. Houthoofd, A.
Philippaerts, C. De Jonghe, J.S. Paul, P.A. Jacobs, B.F. Sels, Green Chem. 12
(2010) 1083–1089.
where Ci are the concentration of species i (mol dmꢀ3), and ki (sꢀ1
are the forward and reverse rate constants for the hydride shift. The
quasi-equilibrium assumption gives us the following expressions in
)
kads
kdes
terms of the adsorption/desorption equilibrium constant K ¼
(dm3 molꢀ1) for each species. We show only the expression for
steps 1 and 2.
ꢃ
CBuOH
CBuOH
KBuOH
¼
ðA2Þ
ðA3Þ
C
ꢃ
ꢃ
CBuOHꢀML
CBuOH CML
ꢃ
KML
¼
ꢃ
We assume pseudo-steady-state on the concentration of 4HP
intermediate, and that the rate constant for lactonization (kf,L),
is extremely large. After rearranging, we find the concentration

H.Y. Luo et al. / Journal of Catalysis 320 (2014) 198–207
207
[18] E. Taarning, S. Saravanamurugan, M. Spangsberg Holm, J. Xiong, R.M. West,
C.H. Christensen, ChemSusChem 2 (2009) 625–627.
[40] J.Q. Bond, D.M. Alonso, D. Wang, R.M. West, J.A. Dumesic, Science 327 (2010)
1110–1114.
[19] C.M. Lew, N. Rajabbeigi, M. Tsapatsis, Microporous Mesoporous Mater. 153
(2012) 55–58.
[20] R. Cohen, C.R. Graves, S.T. Nguyen, J.M. Martin, M.A. Ratner, J. Am. Chem. Soc.
126 (2004) 14796–14803.
[41] L. Qi, I.T. Horváth, ACS Catal. 2 (2012) 2247–2249.
[42] D.M. Alonso, S.G. Wettstein, J.A. Dumesic, Green Chem. 15 (2013) 584.
[43] E.I. Gürbüz, D.M. Alonso, J.Q. Bond, J.A. Dumesic, ChemSusChem 4 (2011) 357–
361.
[21] E.J. Campbell, H. Zhou, S.T. Nguyen, Org. Lett. 3 (2001) 2391–2393.
[22] D.A. Evans, S.G. Nelson, M.R. Gagne, A.R. Muci, J. Am. Chem. Soc. 115 (1993)
9800–9801.
[23] B.P. Warner, J.A. D’Alessio, A.N. Morgan III, C.J. Burns, A.R. Schake, J.G. Watkin,
Inorg. Chim. Acta 309 (2000) 45–48.
[44] J.S. Luterbacher, J.M. Rand, D.M. Alonso, J. Han, J.T. Youngquist, C.T. Maravelias,
B.F. Pfleger, J.A. Dumesic, Science 343 (2014) 277–280.
[45] R.S. Assary, L.A. Curtiss, J.A. Dumesic, ACS Catal. 3 (2013) 2694–2704.
[46] M. Boronat, A. Corma, M. Renz, J. Phys. Chem. B 110 (2006) 21168–21174.
[47] A. Corma, L.T. Nemeth, M. Renz, S. Valencia, Nature 412 (2001) 423–425.
[48] M. Boronat, P. Concepción, A. Corma, M. Renz, S. Valencia, J. Catal. 234 (2005)
111–118.
[49] S. Roy, K. Bakhmutsky, E. Mahmoud, R.F. Lobo, R.J. Gorte, ACS Catal. 3 (2013)
573–580.
[50] O.A. Abdelrahman, A. Heyden, J.Q. Bond, ACS Catal. 4 (2014) 1171–1181.
[51] P. Nandi, Y.I. Matvieiev, V.I. Boyko, K.A. Durkin, V.I. Kalchenko, A. Katz, J. Catal.
284 (2011) 42–49.
[52] Y.-P. Li, M. Head-Gordon, A.T. Bell, ACS Catal. 4 (2014) 1537–1545.
[53] R. Bermejo-Deval, R.S. Assary, E. Nikolla, M. Moliner, Y. Román-Leshkov, S.-J.
Hwang, A. Palsdottir, D. Silverman, R.F. Lobo, L.A. Curtiss, Proc. Natl. Acad. Sci.
USA 109 (2012) 9727–9732.
[54] C. Laurence, J.-F. Gal, Lewis Basicity and Affinity Scales: Data and
Measurement, John Wiley & Sons, 2009.
[55] C. Laurence, J. Graton, J.-F. Gal, J. Chem. Educ. 88 (2011) 1651–1657.
[56] P. Panagiotopoulou, N. Martin, D.G. Vlachos, J. Mol. Catal. A: Chem. 392 (2014)
223–228.
[57] J. Stelzer, M. Paulus, M. Hunger, J. Weitkamp, Microporous Mesoporous Mater.
22 (1998) 1–8.
[24] V. Ivanov, J. Bachelier, F. Audry, J. Lavalley, J. Mol. Catal. 91 (1994) 45–59.
[25] M.a.A. Aramendıa, V. Borau, C. Jiménez, J.M. Marinas, J.R. Ruiz, F.J. Urbano,
´
Appl. Catal. A 244 (2003) 207–215.
[26] P.S. Kumbhar, J. Sanchez-Valente, J. Lopez, F. Figueras, Chem. Commun. (1998)
535–536.
[27] C. Jiménez-Sanchidrián, J.R. Ruiz, Appl. Catal. A 469 (2014) 367–372.
[28] E. Creyghton, S. Ganeshie, R. Downing, H. Van Bekkum, J. Mol. Catal. A: Chem.
115 (1997) 457–472.
[29] J.C. van der Waal, K. Tan, H. van Bekkum, Catal. Lett. 41 (1996) 63–67.
[30] A. Corma, M.E. Domine, S. Valencia, J. Catal. 215 (2003) 294–304.
[31] A. Corma, M.E. Domine, L. Nemeth, S. Valencia, J. Am. Chem. Soc. 124 (2002)
3194–3195.
[32] Y. Zhu, G. Chuah, S. Jaenicke, J. Catal. 227 (2004) 1–10.
[33] M. Moliner, Dalton Trans. 43 (2014) 4197–4208.
[34] M. Dusselier, P. Van Wouwe, S. De Smet, R. De Clercq, L. Verbelen, P. Van
Puyvelde, F.E. Du Prez, B.F. Sels, ACS Catal. 3 (2013) 1786–1800.
[35] H.Y. Luo, L. Bui, W.R. Gunther, E. Min, Y. Román-Leshkov, ACS Catal. 2 (2012)
2695–2699.
[36] J.J. Pacheco, M.E. Davis, Proc. Natl. Acad. Sci. USA 111 (2014) 8363–8367.
[37] W.R. Gunther, V.K. Michaelis, M.A. Caporini, R.G. Griffin, Y. Román-Leshkov, J.
Am. Chem. Soc. 136 (2014) 6219–6222.
[38] Y. Román-Leshkov, M.E. Davis, ACS Catal. 1 (2011) 1566–1580.
[39] I.T. Horváth, H. Mehdi, V. Fábos, L. Boda, L.T. Mika, Green Chem. 10 (2008) 238.
[58] E.E. Mallon, I.J. Babineau, J.I. Kranz, Y. Guefrachi, J.I. Siepmann, A. Bhan, M.
Tsapatsis, J. Phys. Chem. B 115 (2011) 11431–11438.
[59] R. Gounder, M.E. Davis, AlChE J. 59 (2013) 3349–3358.
[60] R. Gounder, Catal. Sci. Technol. 4 (2014) 2877–2886.