An examination of the intrinsic activity and stability of various solid acids during the catalytic decarboxylation of γ-valerolactone

Article abstract of DOI:10.1039/c4cy00307a

Rates of γ-valerolactone (GVL) decarboxylation were measured in the gas phase under anhydrous conditions from 523-723 K over a series of solid acids including amorphous silica alumina, MFI zeolites, supported phosphotungstic acid, and γ-Al₂O₃. Through consideration of decarboxylation rates obtained under differential conditions, we examine the roles of Bronsted and Lewis acidity, deprotonation energy, and catalyst morphology in defining the intrinsic activity and stability of each material. In aluminosilicates, Bronsted sites associated with framework aluminum appear to contribute the majority of decarboxylation activity. Of the aluminosilicates tested, Bronsted sites in MFI are more intrinsically active than analogous sites in ASA; however, zeolite micropores hinder GVL diffusion and lead to mass transfer limitations at high temperatures. Relative to bridging hydroxyls, coordinatively unsaturated aluminum sites are substantially less active and do not contribute significantly to decarboxylation rates in materials having both framework and extraframework aluminum. Decarboxylation barriers scale with the deprotonation energy of Bronsted acid sites; however, lower deprotonation energies do not necessarily imply higher intrinsic activity in GVL decarboxylation. At 623 K, catalyst stability is highest in materials having large pore dimensions, Lewis sites as primary catalytic centers, and Bronsted sites with relatively high deprotonation energies.

Full text of DOI:10.1039/c4cy00307a

Catalysis
Science &
Technology
View Article Online
View Journal | View Issue
PAPER
An examination of the intrinsic activity and
stability of various solid acids during the catalytic
decarboxylation of γ-valerolactone
Cite this: Catal. Sci. Technol., 2014,
4, 2267
*
Aimee B. Kellicutt, Roozbeh Salary, Omar Ali Abdelrahman and Jesse Q. Bond
Rates of γ-valerolactone (GVL) decarboxylation were measured in the gas phase under anhydrous condi-
tions from 523–723 K over a series of solid acids including amorphous silica alumina, MFI zeolites,
supported phosphotungstic acid, and γ-Al2O3. Through consideration of decarboxylation rates obtained
under differential conditions, we examine the roles of Brønsted and Lewis acidity, deprotonation energy,
and catalyst morphology in defining the intrinsic activity and stability of each material. In aluminosilicates,
Brønsted sites associated with framework aluminum appear to contribute the majority of decarboxylation
activity. Of the aluminosilicates tested, Brønsted sites in MFI are more intrinsically active than analogous
sites in ASA; however, zeolite micropores hinder GVL diffusion and lead to mass transfer limitations at high
temperatures. Relative to bridging hydroxyls, coordinatively unsaturated aluminum sites are substantially
less active and do not contribute significantly to decarboxylation rates in materials having both framework
and extraframework aluminum. Decarboxylation barriers scale with the deprotonation energy of Brønsted
acid sites; however, lower deprotonation energies do not necessarily imply higher intrinsic activity in GVL
decarboxylation. At 623 K, catalyst stability is highest in materials having large pore dimensions, Lewis sites
as primary catalytic centers, and Brønsted sites with relatively high deprotonation energies.
Received 10th March 2014,
Accepted 4th April 2014
DOI: 10.1039/c4cy00307a
www.rsc.org/catalysis
Broadly, GVL decarboxylation over solid acids can occur
directly from GVL or proceed through intermediate formation
Introduction
In many respects, light alkenes comprise the foundation of
the petrochemical industry. They are a remarkably versatile
industrial feedstock, and proven upgrading strategies can
readily transform them into transportation fuels, plastics,
and commodity chemicals.1–7 Considering the diverse value
chains they enable, light alkenes are attractive intermediate
targets in biomass refining, where paradigms are frequently
based upon coupling small, reactive molecules.8 Potentially,
light alkenes can be derived from biomass using a number of
different technologies, such as ethanol or butanol dehydra-
tion,9,10 pyrolysis oil upgrading,11 or syngas-based techno-
logies leveraging either Fisher–Tropsch chemistry12–15 or
methanol synthesis.16–18 Alternatively, alkenes can be pro-
duced via acid-catalyzed decarboxylation of cyclic lactones or
alkene-acids.19–21 Of this latter category, γ-valerolactone is
perhaps the most likely candidate for production in a bio-
refinery since it is readily derived through hydrogenation of
levulinic acid,22–36 which can be produced from both C5 and
C6 sugars present in lignocellulosic biomass.37–41
of pentenoic acids, and it yields butene isomers and carbon
dioxide (Fig. 1). Prior reports have demonstrated that a number
of materials, including amorphous silica-alumina (ASA),
γ-Al2O3, and WO42−/Al2O3 all catalyze the reaction.19–21
Although the transformation was originally proposed to be
Brønsted-mediated in ASA,20 Wang and Dumesic have recently
shown that GVL decarboxylation also occurs over Lewis-acid
sites in γ-Al2O3.21 By varying Brønsted and Lewis site densities,
the authors effected control of alkene distributions obtained
through decarboxylation. Specifically, Lewis acid sites allow
GVL decarboxylation but are less active for secondary butene
isomerization (in the presence of water). In contrast, Brønsted
sites promote both decarboxylation and (to a larger extent)
alkene isomerization subsequent to decarboxylation. For this
reason, relatively Brønsted-rich materials, such as ASA, are
slightly more active but deliver an equilibrium-controlled mix-
ture of butene isomers, whereas Lewis-rich materials allow
selective production of linear alpha olefins (LAOs).21 The obser-
vation that Brønsted and Lewis sites appear to catalyze unique
decarboxylation pathways motivates consideration of the extent
to which each site contributes in materials like ASA, which
have significant quantities of both framework and
extraframework aluminum. A detailed understanding of decar-
boxylation at each site could facilitate insightful design of
Department of Biomedical and Chemical Engineering, Syracuse University,
329 Link Hall, Syracuse, NY 13244, USA. E-mail: jqbond@syr.edu;
Tel: +1 315 443 2550
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2267–2279 | 2267

View Article Online
Paper
Catalysis Science & Technology
Fig. 1 Illustration of the proposed GVL decarboxylation pathway occurring over solid acids. Figure adapted from ref. 19.
materials with, for example, complementary Brønsted and
Lewis functionalities to target specific aspects of decarboxyl-
ation. In this context, it is additionally important to consider
the roles of other parameters, such as Brønsted site deproton-
ation energy and the local environment of the acid site, which
can influence intrinsic activity and provide additional parame-
ters for tuning decarboxylation activity.42–46 Finally, prior stud-
ies have indicated that solid acids employed during GVL
decarboxylation deactivate due to coke formation with time on
stream.20 Deactivation can be mitigated to some extent
through a steam co-feed in decarboxylation reactors;6 however,
stability may also correlate with specific physicochemical prop-
erties of solid acids such that it can be introduced by design
once governing relationships are established.
To address these questions, we have considered the activity
and stability of a range of solid acids including γ-Al2O3, which
is generally considered to be Lewis acidic;47,48 ASA, which is
amorphous, mesoporous, and has both Brønsted sites associ-
ated with framework aluminum and Lewis sites associated
with extraframework aluminum;49,50 several crystalline, high
silica, MFI zeolites that are microporous and primarily exhibit
Brønsted sites associated with framework aluminum;42,50–52
and a supported heteropolyacid (phosphotungstic acid, PWA),
which displays Brønsted sites that should deprotonate readily
compared to the above described bridging hydroxyls in ASA
and MFI.53 Samples selected also have a range of physico-
chemical properties, allowing us to probe their effects on both
activity and stability.
ammonium sulfate (>99%, Acros), phosphotungstic acid
hydrate (Reagent grade, Sigma-Aldrich), ethanol (99.99%,
Pharmco-AAPER), and amorphous silica (>99%, Davisil
Grade 633, 200–425 mesh, Sigma-Aldrich) were employed for
catalyst syntheses. Prior to use in kinetic studies,
γ-valerolactone (GVL, >98%, Sigma-Aldrich) was dried over
silica gel (5% w/v, 12–24 mesh, Strem Chemicals).54
4-Pentenoic acid (97%, Sigma-Aldrich), 3-trans-pentenoic acid
(95%, TCI America), 2-pentenoic acid (98%, Sigma-Aldrich),
1-butene (0.1% in He, Scott), propylene (1% in He, Praxair),
and CO2 (5% in N2, Airgas) were used as supplied by the
manufacturers for standard preparation and instrument
calibration. Isopropylamine (IPA, 99%, Acros) and pyridine
(>99.0%, Sigma-Aldrich) were used as probe adsorbates for
acid site titration. IPA was used without purification, and
pyridine was dried over potassium hydroxide (85% Acros).55
Purified water used in sample preparation was prepared in
house by sequential reverse osmosis, UV oxidation, and dou-
ble deionization. He (99.999%, Airgas), N2 (99.999%, Airgas),
and Air (Medical Grade, Airgas) were used in multiple appli-
cations as described in subsequent sections. To eliminate
water adsorption during FTIR sample preparation, we
employed purge gas generated in-house from dried com-
pressed air (−70 °C Dew Point, Peak Scientific) for catalyst
pretreatment (e.g., calcination).
Catalyst preparation
Five amorphous silica-alumina (ASA) samples were consid-
ered in this study, and each is identified using the convention
ASA-XXX.x, where the variable portion of the label specifies
the atomic ratio of silicon to aluminum in that sample.
ASA-005.3a and ASA-05.3b were supplied by Grace-Davison
(SIAL 3113 and 3901, respectively), while ASA-067.6, ASA-004.4,
Materials and methods
Reagents
Sodium metasilicate (Sigma-Aldrich), sodium aluminate
(Sigma-Aldrich), sulfuric acid (95.0–98.0%, Sigma-Aldrich),
2268 | Catal. Sci. Technol., 2014, 4, 2267–2279
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
and ASA-002.0 were synthesized using a co-gelation proce-
dure.56 For all synthesized ASA samples, aqueous solutions of
sodium metasilicate and sodium aluminate were combined in
desired molar ratios, resulting in gel formation. The gel was
then aged for 30 minutes at 339 K before decreasing the solu-
tion pH to 9 using a 10 wt% aqueous solution of sulfuric acid.
Recovered solids were then washed with deionized water,
dried at 398 K, and ion exchanged using aqueous ammonium
sulfate (10% w/v). Ammonium-exchanged ASA samples were
recovered by filtration, washed with deionized water to remove
residual sulfates, dried in air at 398 K, ground to a uniform
particle size, and calcined in air (100 ml min−1, 773 K, 4 h, 3 K
min−1) to obtain H+ form ASA. Prior to use, commercial ASA
samples were calcined using an identical protocol but were
not otherwise modified.
MFI samples of varied aluminum content were obtained
from commercial suppliers. This family of analogs is desig-
nated using the convention MFI-XXX.x, where the variable
portion of the label specifies their atomic ratio of silicon to
aluminum. MFI-500.0 (Acros), MFI-200.0, (Alfa Aesar), MFI-
080.0 (Alfa Aesar), and MFI-030.0 (Alfa Aesar) were supplied
in either H+ or ammonium form. Prior to use, all MFI sam-
ples were calcined in flowing air (100 ml min−1, 4 h, 773 K,
3 K min−1).
Fig. 2 Schematic of experimental apparatus used in measuring GVL
decarboxylation rates over various solid acids.
γ-Al2O3 (Strem) was calcined (100 ml min−1,4 h, 773 K, 3 K
min−1) prior to use. Supported phosphotungstic acid (PWA)
was prepared by impregnation of phosphotungstic acid
hydrate dissolved in ethanol onto amorphous silica as
required to achieve desired PWA loadings. PWA/SiO2 samples
were subsequently sealed in a vial, aged for 12 hours at ambi-
ent temperature, dried in an oven at 393 K, and treated in
flowing air (100 ml min−1, 4 h, 623 K, 3 K min−1).44,53
model 110). The liquid feed was vaporized in a temperature-
regulated gas–liquid contactor by combining with a He
stream (50 ml min−1) controlled by mass flow controller
(Brooks, 5850S). The combined GVL/He stream (19.0 mol%
GVL) was then preheated to reaction temperature. In typical
operation, the catalyst bed was bypassed until steady-state
GVL concentrations were observed in the feed stream, at
which point the combined feed was diverted into the reactor
inlet. The reactor effluent was sent to a vapor–liquid separa-
tor (298 K), where low vapor pressure species (GVL and PEA
isomers) were condensed, diluted in water, and quantified
using a GC-FID (GC 7890A, Agilent) equipped with an
Innowax column. Decarboxylation products (butene isomers,
CO2) were retained in the vapor phase and sent to an in-line
gas chromatograph (GC 7890A, Agilent) equipped with sepa-
rate injectors, columns, and detectors for hydrocarbon and
CO2 analysis. Butenes were resolved using an HP-PONA col-
umn, and their concentrations were determined via cali-
brated FID response. CO2 was resolved using a HP-PLOT/Q
column and quantified via TCD response relative to a helium
reference.
Catalytic activity testing
GVL decarboxylation was carried out in a gas phase, down-
flow reactor. The apparatus is illustrated in Fig. 2. Catalyst
samples were loaded into a 316 stainless steel tube having
either a 1/4" or 1/2" outer diameter as required to minimize
pressure differentials with varied catalyst loadings. Catalyst
beds were held in the reactor with quartz wool end plugs,
and bed lengths were generally 1–3 cm. Dead volume
upstream and downstream of the catalyst bed was packed
with fused silica granules (850–2000 μm, 99.9%, Sigma-
Aldrich), and each end of the reactor was capped with a plug
of quartz wool. The reactor was mounted inside an alumi-
num block within an insulated furnace (Applied Test Sys-
tems). Reactor temperature was monitored using a Type K
thermocouple (Omega) positioned against the exterior reactor
wall at the center of the catalyst bed, and temperature was
regulated using a PID controller (Omega CN7500). Prior to
introducing reactant feeds to packed beds, catalyst samples
were calcined in situ under flowing air (50 ml min−1) using
protocols described in the catalyst preparation section.
For all experiments summarized here, reactors operated
differentially (<4% GVL conversion), and the only reaction
products observed were butenes, CO2, and pentenoic acid iso-
mers. As such, we assume that GVL partial pressures do not
change significantly across the bed. Further, we assume that
product partial pressures (pentenoic acids, butene, and CO2)
remain sufficiently close to zero such that their bulk
Neat GVL was fed into the apparatus at ambient tempera-
ture using a syringe pump (0.05 ml min−1, Cole-Parmer,
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2267–2279 | 2269

View Article Online
Paper
Catalysis Science & Technology
concentrations do not strongly influence the kinetics of GVL
decarboxylation. We note that this does not exclude the pos-
sibility that decarboxylation can occur through intermediate
formation of pentenoic acid; it only implies that the bulk
concentration of pentenoic acid isomers remains low
throughout the catalyst bed. Carbon balances based on com-
bined analysis of gas and liquid phases typically closed to
within 5%. Since conversion ranges for differential operation
are within the precision range expected of mass balance clo-
sure and GC analysis, GVL conversion was determined based
on product formation as defined in eqn (1), where nC4 and
nPEA are the total molar quantities of butenes and pentenoic
acids recovered in a given time period, and nGVL0 is the total
molar quantity of GVL fed into the system in the same time
period.
decarboxylation rates were less than 3% of the extensive
butene production rates at all temperatures considered.
Catalyst characterization
All materials were characterized by physisorption of N2 at
77 K (Micromeritics ASAP 2020). Prior to N2 dosing, samples
were outgassed under vacuum (623 K, 4 h). Total and micro-
pore surface areas were determined by BET and t-plot analy-
ses, respectively. Pore diameters for mesoporous samples
were determined from BJH analysis of the desorption branch
of N2 uptake isotherms.57 Pore dimensions for microporous
samples were determined using the Saito–Foley modification
to the Horvath–Kawazoe method.58 Pore volumes were esti-
mated from the cumulative amount of nitrogen uptake at a
relative pressure of 0.995.
nC4 + nPEA
XGVL
=
(1)
Brønsted site densities were determined from molar quan-
tities of propylene evolved between 575 K and 700 K during
temperature programmed desorption (TPD) of iso-
propylamine (IPA).59 Approximately 100 mg of powdered cata-
lyst were loaded into a quartz tube (1/4") with a fritted
distension (1/2") in the center, and the tube was positioned
in a high temperature furnace (Omega) and connected to a
gas flow manifold. Prior to analysis, catalysts were calcined
in situ according to standard protocols (see above Catalyst
preparation section), cooled to 296 K, and purged (60 ml min−1)
with He that was dried over molecular sieves. After
pretreatment, the sample was held at 296 K and contacted
for 60 minutes with an IPA/He blend prepared by flowing He
(60 ml min−1) through the headspace of an IPA saturation
chamber held in an ethanol/liquid nitrogen bath. The cell
was then purged with He (296 K, 60 min) and subsequently
ramped to 873 K (10 K min−1). During the temperature ramp,
the cell effluent was monitored using a mass selective detec-
tor (Stanford Instruments RGA 100). Signals corresponding to
IPA (m/z = 44) and propylene (m/z = 41) were monitored con-
tinuously, and Brønsted site densities were calculated from
evolved propylene based on the assumption that one mole-
cule of propylene forms at one accessible Brønsted site.
Relative quantities of Brønsted and Lewis sites were deter-
mined using transmission FTIR spectroscopy (Nicolet 6700,
DTGS detector) of adsorbed pyridine on catalyst samples.
Catalysts were pressed into self-supporting wafers (10–15 mg
catalyst, 13 mm diameter, 0.5–5 tons force), loaded into an
aluminum sample holder, and positioned in a glass flow cell
having separate zones for spectral acquisition and sample
pretreatment. Samples were pretreated in an externally
heated, temperature-controlled zone positioned directly
above the spectral acquisition zone. Spectra were acquired
from samples positioned at the center of an unheated glass
tee sealed with CaF2 end caps, and the sample holder was
vertically mobile to permit transfer of samples between the
two sections.
n
GVL0
Intensive rates of reaction are presented here on both
mass-normalized and acid site-normalized bases (i.e., turn-
over frequencies, TOF). Mass normalized rates of reaction are
calculated from measured butene production rates according
to eqn (2).
RC4
rDC
=
(2)
mcat
Where RC4 represents the extensive butene production rate
determined by GC analysis, and mcat is the total mass of cata-
lyst in the bed. TOFs are reported on either a Brønsted or
Lewis site basis. In either case, they are calculated by dividing
the mass-normalized rate of decarboxylation by the molar
density of Brønsted or Lewis sites in a given catalyst (see Cat-
alyst characterization section). TOFs reported here have units
of moles butene per mole of acid site per minute, which is
expressed as min−1 for convenience. All of the solid acids
tested deactivate when first placed on stream, yet all of the
characterization methods we have employed describe the cat-
alyst in a pristine state. To facilitate meaningful comparison
between unique sites in different materials, all kinetic data
summarized here represent theoretical decarboxylation rates
at zero time on stream. These were estimated by regression
of decaying decarboxylation rates as a function of time.
Decarboxylation of GVL does occur over quartz packing at
temperatures above 573 K. Accordingly, background decar-
boxylation rates were measured in both 1/4" and 1/2"
stainless steel tubes packed with quartz granules at all tem-
peratures where kinetic data are reported here. In all cases,
we observed that background activity does not contribute sig-
nificantly to extensive butene production rates measured in
the presence of catalyst beds. Specifically, for ASA, MFI, and
PWA samples, background GVL decarboxylation rates were
less than 1% of the extensive butene production rates at all
temperatures considered. For γ-Al2O3, background GVL
In a typical pyridine adsorption experiment, samples were
calcined in purge gas (100 ml min−1, 4 h, 723 K, 3 K min−1).
Subsequently, the sample was cooled to 423 K, and the cell
2270 | Catal. Sci. Technol., 2014, 4, 2267–2279
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
was purged under He flow. Prior to introducing into the gas
manifold, He was purified in sequential cryogenic and molec-
ular sieve traps to remove trace moisture and hydrocarbon
impurities. Under a continuous purge, the sample was
lowered into the spectral acquisition section, and a back-
ground spectrum was recorded. The sample was then trans-
ferred to the heated zone (423 K) and the helium purge flow
was switched to a pyridine/He blend prepared by passing He
(100 ml min−1) through the head space of a saturator
containing dry pyridine and held at 263 K using an NaCl/ice
bath. Pyridine was dosed onto the sample (423 K) for
60 minutes, and the sample was subsequently purged with
helium (423 K) for another 60 minutes to remove physically
adsorbed pyridine. The sample was then lowered into posi-
tion between the CaF2 windows under a continuous He
purge, and a spectrum of adsorbed pyridine was acquired.
Brønsted and Lewis site ratios were determined from differ-
ence spectra by integration of IR bands at 1545 cm−1 and
1455 cm−1.60–62 In our analysis, we employed molar extinction
coefficients published by Satsuma, which are 1.23 and 1.73 cm
μmol−1 for Brønsted and Lewis sites, respectively.63 For all
materials except γ-Al2O3, Lewis site densities were estimated
based on Brønsted site densities determined from IPA desorp-
tion and Brønsted : Lewis ratios measured by FTIR spectra of
adsorbed pyridine. In the case of γ-Al2O3, we observed no
pyridinium ion formation (1545 cm−1) in FTIR, and its Lewis
site density was determined directly from FTIR difference
spectra of adsorbed pyridine based on the band at 1455 cm−1.
rates of GVL decarboxylation (DC rate, 623 K, 0.19 bar GVL)
at zero time on stream. For a visual comparison, each catalyst
is additionally ranked in order of decreasing DC rates in
Fig. 3. In general, the catalyst properties summarized in
Table 1 vary according to expectations with e.g., aluminum
content and morphology. Since it is outside the focus of this
study, we have not pursued a rigorous explanation for the
observed increase in Brønsted–Lewis ratio for decreasing
Si : Al ratios in low-silica ASA samples, but we speculate that
it arises from structural heterogeneity, noting that it is
difficult to control the distribution of framework and
extraframework aluminum using the co-gelation method
employed here. As there is uncertainty in titrating accessible
Results and discussion
Fig. 3 Comparison of mass-normalized decarboxylation rates mea-
sured over the solid acids documented in Table 1. Decarboxylation
rates were measured at 623 K and 0.19 bar GVL at a total system pres-
sure of 1 bar with a helium balance. Weight-hourly space velocities
were varied for each catalyst to maintain differential conversions.
Catalyst characterization and productivity in GVL
decarboxylation
Table 1 summarizes the physical and chemical properties of
all catalysts considered here alongside their mass-normalized
Table 1 Summary of physicochemical properties and measured decarboxylation rates of solid acids considered in this study. Decarboxylation
rates reported here were measured at 623 K and 0.19 bar GVL at 1 bar system pressure with a helium balance. Weight-hourly space velocities were
varied as necessary to maintain differential conversion
BET surface
area (m2 g−1
Micropore
Pore
diameter (Å)
Pore
Brønsted sites
DC rate
Catalyst
)
area (m2 g−1
)
volume (cm3 g−1
)
(μmol g−1
)
B : L
(μmol min−1 g−1
)
ASA
ASA-067.6
ASA-005.3a
ASA-005.3b
ASA-004.4
ASA-002.0
550
610
420
320
120
130
0.0
0.0
7.0
7.0
32
42
45
67
0.40
0.81
0.61
0.58
0.41
52
3.0
66
340
105
180
63
0.34
0.47
0.64
0.96
440
150
290
95
140
MFI
MFI-500.0
MFI-200.0
MFI-080.0
MFI-030.0
360
360
390
380
100
140
230
240
5.3
5.8
5.4
5.3
0.22
0.22
0.27
0.28
39
79
320
560
12
13
9.6
8.7
230
440
690
750
Other
γ-Al2O3
5% PWA
SiO2
230
470
480
5.0
1.0
2.0
60
56
57
0.47
0.84
0.87
70.0
30
0.0
0.0
0.52
—
1.5
7.8
0.05
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2267–2279 | 2271

View Article Online
Paper
Catalysis Science & Technology
Brønsted sites in supported heteropolyacids,53 it is worth
commenting on the Brønsted site density measured here for
5% PWA/SiO2 (30 μmol g−1) by TPD of IPA after calcination in
air at 623 K. A site density of 30 μmol g−1 corresponds to
roughly 1.7 accessible protons per polyoxymetallate cluster in
5% PWA/SiO2, which is lower than the theoretically antici-
pated value of 3.0. This result generally agrees with Brønsted
site densities for 5% PWA/SiO2 reported by Iglesia (1.4–2.9 H+
per Keggin unit) using in situ titration with pyridine and
di-tertbutyl pyridine under various reaction conditions.43,53,64
We thus consider that the Brønsted site density measured
here by IPA desorption provides a reasonable basis for calcu-
lating turnover frequencies in PWA/SiO2. Further discussion
of specific properties and their impact on activity and stabil-
ity will be deferred to subsequent sections.
Of the materials tested here, MFI zeolites are, on average,
the most active, achieving initial DC rates ranging from
220–750 μmol min−1 g−1, and DC rates generally increase with
aluminum content in MFI. ASA samples are slightly less
active than MFI zeolites, demonstrating DC rates ranging
from 66–440 μmol min−1 g−1. Unlike the MFI series, there is
no clear correlation between aluminum content and DC rates
in the ASA samples. Both 5% PWA/SiO2 and γ-Al2O3 are sub-
stantially less active than either MFI or ASA, achieving decar-
boxylation rates of 7.8 and 1.5 μmol min−1 g−1, respectively,
at 623 K and 0.19 bar GVL. From a practical viewpoint, alumi-
nosilicates seem broadly well-suited to GVL decarboxylation
based on their high DC rates, and we generally conclude that
a high aluminum content benefits overall catalyst productiv-
ity. This aligns with the expectation that low Si : Al ratios
imply high acid site densities in aluminosilicates; however,
this presentation does not delineate contributions from
multiple underlying factors. For example, most of the mate-
rials summarized in Table 1 have heterogeneous site distribu-
tions and display both Brønsted and Lewis acidity. Either
type of site is proposed to catalyze GVL decarboxylation, but
contributions from each are not apparent in mass-
normalized rates.21 Further, the local structure and deproton-
ation energy of Brønsted sites vary among the materials
tested, and either may influence reactivity and stability.42 In
subsequent sections, we decouple these effects to the extent
possible to define the intrinsic activity and stability of varying
acid sites.
Amorphous silica alumina and γ-Al2O3
Examining the ASA data summarized in Table 1, we observe
that, despite spanning a considerable range of surface areas
(120–610 m2 g−1), pore diameters (32–140 Å), and Brønsted :
Lewis ratios (0.3–3.0), mass-normalized DC rates in ASA
depend only on the Brønsted site density of a particular sam-
ple. This conclusion is emphasized in Fig. 4a, which illus-
trates decarboxylation rates plotted against Brønsted site
counts on a logarithmic scale. Linear regression of this data
set reveals a slope of 0.99 0.21, indicating that DC rates in
ASA have a first order dependence on Brønsted site density.
This observation allows the conclusion that the Koros–Nowak
criteria are rigorously satisfied for this ASA series at
623 K.65,66 Thus, rates observed over each ASA sample reflect
kinetic control; the intrinsic DC activities of Brønsted sites in
ASA do not vary with aluminum content; and Brønsted sites
in each ASA sample are described by a single turnover
frequency.
As illustrated in Fig. 4b, DC rates over ASA at 623 K are
well-described by a line having a slope of 1.36 0.13 min−1
and passing through the origin. The calculated slope repre-
sents the average TOF of a Brønsted site in ASA, and a zero-
intercept implies that, in the absence of Brønsted sites, no
decarboxylation activity is expected in ASA. This result is con-
sistent with our observation that amorphous SiO2, which
Fig. 4 Illustration of the relationship between mass-normalized decarboxylation rates and Brønsted site density measured over ASA samples of
varying aluminum content at 623 K and 0.19 bar GVL at 1 bar total pressure in a helium balance. (○) ASA-067.6, (△) ASA-002.0, (□) ASA-005.3b,
(◇) ASA-004.4, (▽) ASA-005.3a. a) Display of linearity in fulfillment of Koros–Nowak criteria for illustration of kinetic control during GVL decarbox-
ylation in ASA. b) For ASA samples of varying Brønsted : Lewis ratios, decarboxylation rates depend only on the Brønsted site density at 623 K. In
both Fig. a and b, decarboxylation rates are reported in μmol min−1 g−1 and H+ site densities are reported in μmol g−1
.
2272 | Catal. Sci. Technol., 2014, 4, 2267–2279
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
displays no Brønsted signatures in pyridine FTIR or IPA TPD,
shows little activity during GVL decarboxylation at 623 K
(0.05 μmol min−1 g−1). No significant deviation from the illus-
trated trend is evident in the ASA samples despite consider-
able variation in Brønsted : Lewis ratio (0.3–3.0), indicating
that GVL decarboxylation is insensitive to Lewis sites in ASA.
Consistent with prior observations by Wang and
Dumesic,21 decarboxylation activity was observed here above
background levels at 623 K and 0.19 bar GVL over γ-Al2O3
(1.5 μmol min−1 g−1); however, it is substantially lower than
analogous rates in ASA. Tentatively, the low activity of γ-Al2O3
could be attributed to weakly (Brønsted) acidic surface
hydroxyls, coordinatively unsaturated (Lewis) aluminum sites,
or unanticipated acidity arising from residual impurities. As
in prior spectroscopic studies of pyridine adsorption, we
observe appreciable Lewis acidity but no clear pyridinium ion
formation on γ-Al2O3, suggesting a catalyst without Brønsted
acidity.47,48 However, we did observe roughly 70 μmol g−1 of
propylene evolution from γ-Al2O3 during IPA TPD, suggesting
at least a modest surface density of Brønsted sites. As to the
nature of the Brønsted sites detected by IPA desorption, we
speculate that it is likely attributed to residual impurities
rather than surface hydroxyls associated with aluminum since
the latter have been demonstrated by Gorte to not catalyze IPA
deamination.47 As such, the DC activity of γ-Al2O3 reported
here is most likely attributed either to Brønsted acidity con-
ferred by residual impurities or to a Lewis-mediated decarbox-
ylation pathway, with prior studies by Dumesic supporting the
role of Lewis acidity.21 Based on FTIR spectra of adsorbed pyri-
dine on γ-Al2O3, we estimate a Lewis-site density of roughly
148 μmol g−1. If DC activity is attributed to coordinatively
unsaturated aluminum sites in γ-Al2O3, they appear to have a
turnover frequency of 0.01 min−1 and are thus two orders of
magnitude less active than bridging hydroxyls in ASA. If this
turnover frequency is taken as representative of the intrinsic
activity of Lewis sites in ASA, we calculate (from Table 1) that
Lewis sites contribute anywhere from 0.2–2.2% of the initial
GVL decarboxylation rate in ASA samples, and that Brønsted
sites account for the remainder. This estimate is consistent
with the trend in Fig. 4b, which indicates that DC rates
depend only on Brønsted site counts in ASA.
Previously, Wang and Dumesic reported a steady state
decarboxylation rate of 12 μmol min−1 g−1 at 673 K for the
decarboxylation of pentenoic acid over γ-Al2O3 under anhy-
drous conditions (0.07 bar PEA in He), which agrees well with
a steady state rate of 10 μmol min−1 g−1 measured here for
GVL decarboxylation over γ-Al2O3 at 673 K (0.19 bar GVL in
He). In prior studies, when the helium balance was replaced
with water, a five-fold enhancement in decarboxylation rates
is observed over γ-Al2O3, while less pronounced21 or inhibi-
tory20 effects are observed over ASA. This result suggests
that water uniquely promotes decarboxylation pathways at
coordinatively unsaturated aluminum centers and that
Lewis-catalyzed decarboxylation may comprise a larger portion
of total butene production rates in the presence of water.
MFI
In contrast to ASA samples, mass normalized decarboxylation
rates over MFI samples of varied aluminum content do not
correlate uniformly with Brønsted site density at 623 K
(Fig. 5a). Rather, we observe that production rates appear to
scale directly with Brønsted site density only in the highest
silica samples (Si : Al > 80). Thereafter, increases in DC rates
diminish with further increases in aluminum content, and
the MFI series approaches an overall maximum production
rate of roughly 750 μmol min−1 g−1 at 623 K in the MFI-030.0
sample. This leads to the conclusion that turnover frequen-
cies in MFI decrease with aluminum content, and we
consider that both kinetic and transport phenomena may
underlie this observation.
Based on available data for γ-butyrolactone and
ε-caprolactone, GVL has an approximate kinetic diameter
between 5.0 and 6.0 Å (calculated based on critical molar
Fig. 5 Illustration of the relationship between mass-normalized decarboxylation rates and Brønsted site density measured over MFI zeolites of
varying aluminum content at 623 K and 523 K, 0.19 bar GVL, and 1 bar total pressure in a helium balance. MFI-500.0 (◇), MFI-200.0 (□), MFI-080.0 (▽),
MFI-030.0 (○) a) 623 K b) 523 K.
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2267–2279 | 2273

View Article Online
Paper
Catalysis Science & Technology
volumes as per Jae67), which is on the same order as the aver-
age pore size of the MFI samples employed here (5.5 Å). It is
thus reasonable to expect that pore access and diffusion of
GVL are hindered in MFI compared to mesoporous ASA. Fur-
ther, MFI-080.0 and MFI-030.0 have relatively high areal den-
sities of Brønsted sites (0.80 and 1.46 μmol m−2, calculated
from Table 1) compared to the most Brønsted-dense ASA
samples, ASA-005.3a and ASA-004.4 (0.55 and 0.57 μmol m−2).
The combination of a high surface density of acid sites
coupled with small pore diameters suggests that diffusion
limitations could be significant in MFI despite not being
observed in ASA. Increasing mass transfer limitations at
higher site densities would explain the trend illustrated in
Fig. 5a; however, this conclusion discounts the possibility
that Brønsted sites in MFI zeolites are nonequivalent in the
range of Si : Al ratios tested here. Although Brønsted sites in
high-silica MFI (i.e., Si : Al > 10) should display uniform
deprotonation energies, framework aluminum atoms can
occupy a range of tetrahedral sites and local environments in
the MFI framework.52 As previously described by Gounder,
the nanoscale effects of confinement and solvation frequently
control reactivity in zeolites; thus, it is possible that heteroge-
neous distributions of framework aluminum in MFI could
lead to actual variations in intrinsic activity as opposed to
apparent ones arising from transport limitations.68 Rigorous
consideration of diffusion limitations in this system is there-
fore warranted, and relevant data supporting this discussion
is summarized in Fig. 6, which illustrates observed trends in
DC rates over MFI zeolites as a function of Brønsted site den-
sity at 523 K and 623 K.
with increasing aluminum content, DC rates were measured
at 523 K under otherwise identical conditions, with the rea-
soning that a 100 K decrease in temperature should affect DC
rates and GVL diffusion to different extents and thus permit
discrimination between kinetic and transport effects. As illus-
trated in Fig. 6, regression of the low temperature data set
indicates a slope of 0.84 0.34, suggesting that DC rates over
MFI are shifted toward kinetic control at 523 K.57 This is con-
sistent with the interpretation that mass transfer, rather than
varying intrinsic activity, is the source of apparent decreasing
turnover frequencies with increasing aluminum content in
MFI zeolites at 623 K. Emphasizing this point, Fig. 5b shows
that decarboxylation rates over MFI at 523 K vary linearly
with Brønsted site density and are captured by a line passing
through the origin. Thus, Brønsted sites in MFI are the pri-
mary catalytic centers for GVL decarboxylation, and they
appear to be equivalent in this context over a broad range of
Si : Al ratios (30–500).
Since kinetic control is not rigorously demonstrated for all
MFI samples at 623 K, we are cautious in defining a turnover
frequency for MFI at this temperature; however, based on the
trend illustrated in Fig. 5a,
a meaningful comparison
between MFI and ASA can be established. Specifically, Fig. 5a
illustrates that the linear trend in decarboxylation predicted
by the MFI-500.0 and MFI-200.0 samples does extrapolate to
the origin. This suggests that at low aluminum contents,
there is a first order relationship between decarboxylation
activity and Brønsted site density, i.e., the data represent
kinetic control. Based on the regression illustrated in Fig. 5a
(which includes the background level activity of amorphous
silica), we estimate that Brønsted sites in MFI have a DC
turnover frequency of 5.54 0.36 min−1 at 623 K and are thus
roughly four times more intrinsically active than analogous
sites in ASA (1.36 min−1).
Regression of the high temperature data set (623 K) in
Fig. 6 indicates a slope of 0.43 0.16. Assuming structure
insensitivity in GVL decarboxylation over MFI, this result sug-
gests that internal diffusion controls the rate of decarboxyl-
ation in MFI at 623 K. To probe the possibility that site
heterogeneity could lead to decreasing turnover frequencies
The kinetic effects of acid site variation
We have thus far demonstrated that GVL decarboxylation
over aluminosilicates appears relatively insensitive to Lewis
acidity such that overall activity in anhydrous environments
is determined by the intrinsic activity and total number of
characteristic Brønsted sites. Despite having Brønsted sites
with comparable deprotonation energies (≈1200 kJ mol−1),42
MFI and ASA present turnover frequencies that differ by a
factor of four, and we explore the source of this discrepancy
in this section. We further compare the activity of ASA and
MFI to that of 5% PWA/SiO2, which has Brønsted sites char-
acterized by
a significantly lower deprotonation energy
(DPE = 1050–1100 kJ mol−1) than either of the alumino-
silicates.42,64 We additionally expand our consideration of
γ-Al2O3 to improve our understanding of its relatively low
activity in GVL decarboxylation.
For PWA/SiO2, reference experiments were carried out
under differential conditions at 623 K with samples having
both 5 wt% and 20 wt% PWA. Over these two catalysts, we
observed equivalent, Brønsted-normalized turnover frequencies,
Fig. 6 Correlation between decarboxylation rates and Brønsted site
densities in MFI at 623 K (open symbols) and 523 K (filled symbols).
MFI-500.0 (◇), MFI-200.0 (□), MFI-080.0 (▽), MFI-030.0 (○).
Decarboxylation rates are reported in μmol min−1 g−1 and H+ site
densities are reported in μmol g−1
.
2274 | Catal. Sci. Technol., 2014, 4, 2267–2279
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
suggesting that mass transfer limitations are absent during
GVL decarboxylation over PWA/SiO2 at 623 K. We did not test
explicitly for internal diffusion limitations in γ-Al2O3; however,
its pore dimensions are comparable with the ASA series (~50 Å)
and its mass-normalized decarboxylation rate (1.5 μmol min−1 g−1,
623 K) is 2 orders of magnitude smaller than ASA, on average.
Since kinetic control was rigorously demonstrated in the ASA
series at 623 K, we assume that this conclusion extends to the
less reactive γ-Al2O3.
Calculated turnover frequencies (623 K, 0.19 bar GVL) for
each material are summarized in Fig. 7. The most active sites
under the conditions tested appear to be protons in MFI
(5.54 min−1) and ASA (1.36 min−1). Brønsted sites in 5%
PWA/SiO2 display a substantially lower turnover frequency
(0.26 min−1) than either of the aluminosilicates. This suggests
that a more easily deprotonated acid site does not necessarily
imply a more active catalyst during GVL decarboxylation and
may point to the significance of structural effects in this reac-
tion. It is worth mentioning that Keggin structures swell to
varying extents depending on the titrant and reaction environ-
ment, and it is possible that protons in PWA/SiO2 are more or
less accessible to GVL than they are to IPA. There is thus
uncertainty in our apparent TOF. In the event that protons in
the Keggin structure are substantially less accessible to GVL
than they are to IPA, Brønsted sites in PWA/SiO2 could be sig-
nificantly more active than indicated by the values reported
here. As described earlier, decarboxylation over γ-Al2O3 is
most likely attributed to Lewis sites, and we have accordingly
calculated its turnover frequency on this basis to be roughly
0.01 min−1, indicating it is the least intrinsically active of the
materials considered here.
Fig. 8 Arrhenius plots illustrating the temperature dependence of GVL
decarboxylation over γ-Al2O3 (○), PWA/SiO2 (□), ASA-005.3a (x), MFI-
500.0 (◇). Decarboxylation turnover frequencies reported here have
units of min−1
.
sample, kinetic rate control has been adequately demon-
strated at 623 K and below. Rate data were additionally col-
lected at higher temperatures, but we note no deviation from
low-temperature trends in high-temperature data, permitting
the assumption that where higher temperature data are
reported, they also reflect kinetic control. Arrhenius parame-
ters for each material are summarized alongside calculated
turnover frequencies in Table 2. In this presentation, both
pre-exponential factors and activation barriers are apparent
and reflect lumped contributions from multiple elementary
phenomena.
γ-Al2O3 has the highest decarboxylation barrier of any
material tested (172 kJ mol−1), which aligns with its relatively
low intrinsic activity. Whether its catalytic activity is attrib-
uted to coordinatively unsaturated aluminum or trace impuri-
ties, either type of site appears to facilitate a higher energy
decarboxylation pathway than catalysts that have a more pro-
nounced Brønsted character. This is consistent with the
results of preceding sections that suggest extraframework alu-
minum contributes little to the catalytic activity of alumino-
silicates in the absence of water. MFI and ASA have
essentially indistinguishable apparent activation barriers of
138 and 130 kJ mol−1. This could indicate that DC barriers
over Brønsted sites are largely controlled by their deproton-
ation energy, which should be comparable (≈1200 kJ mol−1)
in these two materials.42 Supporting this conclusion, we
For a more comprehensive view, Fig. 8 presents Arrhenius
plots for the four materials compared in this section:
MFI-500.0, ASA-005.3a, 5% PWA/SiO2, and γ-Al2O3. For each
Table
2 Apparent Arrhenius parameters measured for MFI-500.0,
ASA-005.3a, 5% PWA/SiO2, and γ-Al2O3. Turnover frequencies were
measured under differential conditions at 623 K and 0.19 bar GVL and
1 bar total pressure in a helium balance. Errors in apparent activation
barriers were determined at a 95% confidence level
Fig.
7 Initial turnover frequencies measured under differential
conditions over MFI, ASA, 5% PWA/SiO2, and γ-Al2O3 at 623 K and
0.19 bar GVL at 1 bar total pressure in a helium balance. For MFI and
ASA, estimated turnover frequencies reflect the average of samples for
which kinetic control has been demonstrated. The turnover frequency
indicated for γ-Al2O3 is calculated on a Lewis site basis. All other turn-
over frequencies are calculated on a Brønsted site basis.
Catalyst
A (min−1
)
EA (kJ mol−1
)
(kJ mol−1
)
TOF (min−1
)
MFI-500.0
2.6 × 1012 138
13
5.54
1.36
0.26
0.01
ASA-005.3a 6.4 × 1010 130
24
10
36
PWA/SiO2
γ-Al2O3
1.3 × 107
1.7 × 1012 172
92
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2267–2279 | 2275

View Article Online
Paper
Catalysis Science & Technology
observe a substantially lower DC barrier over 5% PWA/SiO2
(92 kJ mol−1), which has Brønsted sites characterized by a sig-
deprotonation energies do appear to have reduced barriers for
GVL decarboxylation, but the extent to which deprotonation
energy or other factors (e.g., confinement) control the intrin-
sic activity of an acid site remains unclear.
nificantly lower deprotonation energy (1050–1100 kJ mol−1
)
than what is commonly reported for aluminosilicates.42
Despite the fact that measured barriers appear to scale
with deprotonation energy, decarboxylation activity is not
necessarily anticipated by observed activation barriers. Specif-
ically, ASA and MFI have comparable barriers, but their turn-
over frequencies vary by a factor of four. In contrast, DC
barriers are relatively low over 5% PWA/SiO2, but Brønsted
Catalyst stability
Fig. 9 illustrates the activity of selected materials as a func-
tion of time on stream. Here, dimensionless activity is calcu-
lated by normalizing decarboxylation rates measured over a
given catalyst at the indicated time on stream by the decar-
boxylation rate for the same catalyst at zero time on stream.
Fig. 9a compares MFI-500.0 to two ASA samples at 623 K, pro-
viding a comparison of materials having Brønsted sites with
comparable deprotonation energies but different physical
properties (Table 1). Fig. 9b compares ASA-005.3a, 5 wt%
PWA/SiO2, and γ-Al2O3 at 623 K to highlight stability differ-
ences between materials having comparable physical proper-
ties (Table 1) but varying acid sites and decarboxylation
barriers. In each case, deactivation is attributed to coke for-
mation based on color change in spent samples, and all cata-
lysts lose activity with time on stream. A rigorous comparison
of catalyst stability is difficult because each material displays
unique, non-integer deactivation orders, and we were not
able to capture the on-stream activity of each sample with a
single deactivation model. However, at short times on stream
(<30 min), all deactivation profiles exhibit pseudo-first order
kinetics. We therefore estimated first order deactivation con-
stants from initial periods on stream to allow a quantitative
comparison of catalyst stability. Deactivation constants are
presented alongside potentially relevant physical and chemi-
cal properties in Table 3.
sites therein display
a diminished turnover frequency
(0.26 min−1) compared to either MFI (5.54 min−1) or ASA
(1.36 min−1). Examining the pre-exponential factors given in
Table 2, we observe that the differences in activity between
MFI, ASA, and PWA/SiO2 are captured by the magnitude of
their prefactors; however, it is difficult to interpret this result
with further resolution. First, the variability in Arrhenius
parameters observed here is large and arises from the relative
imprecision of collecting kinetic data in deactivating systems.
At 623 K and assuming a common pre-exponential factor for
all materials, variation of barriers within the indicated preci-
sion could explain the majority of observed differences in
turnover frequencies. Further, since no detailed microkinetic
description of GVL decarboxylation has been established,
concentration dependencies here are lumped into the pre-
exponential factor. We have previously observed GVL decar-
boxylation to be zero-order in GVL over ASA,20 but this has
not been rigorously extended to MFI, PWA/SiO2, and γ-Al2O3.
Finally, we have already noted the uncertainty in determining
accurate, Brønsted-normalized TOFs for PWA/SiO2 using TPD
of IPA, and this uncertainty is similarly lumped into the pre-
exponential factors reported here. Given the high temperature
nature of this reaction (523–723 K), entropic effects may
indeed be significant in defining the intrinsic activity of a
catalyst; however, multiple other factors could additionally
underlie the observed variations. Brønsted sites with lower
From Fig. 9a, it is apparent that the stability of the alumi-
nosilicates decreases in the order ASA-005.3a > ASA-067.0 >
MFI-500.0. For clarity, the remaining ASA and MFI samples
are omitted from this figure, but their behaviors are
Fig. 9 Deactivation profiles observed over various solid acids with time on stream. a) Comparison of catalyst stability in (□) MFI-500.0,
(△) ASA-067.6, and (○) ASA-005.3a. b) Comparison of catalyst stability in (x) PWA/SiO2, (○) ASA-005.3a, and (▽) γ-Al2O3.
2276 | Catal. Sci. Technol., 2014, 4, 2267–2279
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
Table 3 Summary of first order deactivation constants and relevant physicochemical properties
Catalyst
BET surface area (m2 g−1
)
Micropore area (m2 g−1
)
Acid site density (μmol m−2
)
Average pore diameter (Å)
kd (h−1
)
ASA-005.3a
ASA-067.6
MFI-500.0
γ-Al2O3
610
550
360
230
470
0
130
101
5
0.55
0.09
0.11
0.13
0.06
42
32
5.3
60
56
0.4
1.5
5.2
0.03
0.5
PWA/SiO2
1
qualitatively captured by the illustrated data. Specifically,
with the exception of ASA-067.0 (illustrated in Fig. 9a), all
ASA samples displayed an activity profile comparable to
ASA-005.3a (illustrated in Fig. 9a). The activity of all MFI sam-
ples decayed rapidly, and none differed significantly from the
behavior of MFI-500.0 (illustrated in Fig. 9a). Analyzing the
physical and chemical properties of this aluminosilicate
series, we conclude that the stability of materials having com-
parable deprotonation energies correlates most strongly with
microporosity and average pore diameter. Most ASA samples
tested here had pore diameters on the order of 40–60 Å, and
they displayed similar stability when placed on stream,
whereas ASA-067.0 – the least stable of the ASA series – had a
relatively large micropore area (130 m2 g−1) and smaller pores
on average (33 Å). This conclusion is further evidenced by the
fact that MFI-500.0, with a higher micropore fraction and a
smaller average pore diameter (5.3 Å), deactivates the most
rapidly of any material considered. Since ASA-005.3a has both
a higher areal acid site density and better stability than ASA-
067.0 (Table 3), it appears that increasing acid site proximity
does not promote coke deposition during GVL decarboxyl-
ation. Finally, since multiple ASA samples of varying
Brønsted : Lewis site ratios display comparable deactivation
profiles, we suggest that Lewis acidity does not promote coke
formation during GVL decarboxylation. The results summa-
rized to this point suggest that mesoporous aluminosilicates
with low Si : Al ratios (high acid site densities) should offer
the best compromise of activity and stability during GVL
decarboxylation. Since increasing acid site proximity does not
appear to diminish stability during GVL decarboxylation,
high Brønsted site densities in mesoporous aluminosilicates
can offset their reduced intrinsic activity compared to micro-
porous zeolites (e.g., MFI).
Although not as dramatic an effect as changing pore
dimensions, variation in the chemical nature of the acid site
can also lead to changes in stability (Fig. 9b). Deactivation
constants for this series range from 0.03 h−1 for the most sta-
ble material, γ-Al2O3, to roughly 0.5 h−1 for the least stable
material, 5% PWA/SiO2. Assuming that pore dimensions in
each of these materials are sufficiently close to not alter deac-
tivation kinetics (40–60 Å), we conclude that materials with
relatively low deprotonation energies (e.g., PWA/SiO2, 1050–
1100 kJ mol−1) are less stable than those with relatively high
deprotonation energies (e.g., ASA, ≈1200 kJ mol−1), implying
that facile deprotonation of Brønsted sites exacerbates coke
formation during GVL decarboxylation. Lewis sites in γ-Al2O3
appear to be more stable than analogous Brønsted sites in
ASA. Observed decarboxylation barriers are relatively high
over γ-Al2O3 (172 kJ mol−1), which implies lower reaction
rates and higher operating temperatures; however, stable
on-stream performance and unique selectivity to LAOs make
γ-Al2O3 an intriguing catalyst for GVL decarboxylation.21
Conclusions
Rates of GVL decarboxylation in anhydrous environments
correlate most strongly with the Brønsted acidity of the cata-
lyst, and we have observed invariant turnover frequencies
over Brønsted sites in both ASA and MFI over a range of
Si : Al and Brønsted : Lewis ratios. Lewis sites associated with
extraframework aluminum appear less active than Brønsted
sites, and they do not contribute significantly to total decar-
boxylation rates in aluminosilicates. GVL decarboxylation
barriers scale with the deprotonation energy of a Brønsted
site; however, sites with low apparent barriers are not neces-
sarily more intrinsically active than those with high apparent
barriers. For example, supported PWA is less active than crys-
talline and amorphous aluminosilicates despite having a sub-
stantially lower apparent activation barrier. Additionally, even
though Brønsted sites in MFI and ASA have similar deproton-
ation energies and experimentally observed decarboxylation
barriers, the two aluminosilicates display substantially differ-
ent turnover frequencies. The above observations may sug-
gest that the local structure of a Brønsted site is important in
defining its intrinsic decarboxylation activity; however, a lim-
ited microkinetic description of GVL decarboxylation and
uncertainty in turnover frequency determination prevent
rigorous support of this conclusion.
Based on the materials considered, aluminosilicates look
to be the most practical for GVL decarboxylation under work-
ing conditions. Within this class of catalysts, pore diameter
appears to be one of the most critical design parameters,
influencing both activity and stability. Specifically, Brønsted
sites in microporous MFI are more active than those in meso-
porous ASA; however, GVL diffusion is hindered in MFI,
leading to transport limitations at high temperatures.
Furthermore, the loss of activity in microporous samples due
to coke formation is more severe than in mesoporous ana-
logs. Coke deposition seems to be associated with Brønsted
acidity, but it does not scale with Brønsted site proximity.
This suggests that catalysts with high acid site densities can
be realistically employed for GVL decarboxylation. As a final
recommendation, we propose that well-defined, mesoporous
aluminosilicates (e.g., Al-MCM41) with low Si : Al ratios may
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2267–2279 | 2277

View Article Online
Paper
Catalysis Science & Technology
offer the best compromise of activity and stability in GVL
decarboxylation.
15 A. de Klerk, Green Chem., 2008, 10, 1249–1279.
16 S. A. Tabak and S. Yurchak, Catal. Today, 1990, 6, 307–327.
17 F. J. Keil, Microporous Mesoporous Mater., 1999, 29, 49–66.
18 G. A. Olah, Angew. Chem., Int. Ed., 2005, 44, 2636–2639.
19 J. Q. Bond, D. Martin Alonso, R. M. West and J. A. Dumesic,
Langmuir, 2010, 26, 16291–16298.
20 J. Q. Bond, D. Wang, D. Martin Alonso and J. A. Dumesic,
J. Catal., 2011, 281, 290–299.
21 D. Wang, S. H. Hakim, D. M. Alonso and J. A. Dumesic,
Chem. Commun., 2013, 49, 7040–7042.
22 J. C. Serrano-Ruiz, D. J. Braden, R. M. West and
J. A. Dumesic, Appl. Catal., B, 2010, 100, 184–189.
23 D. J. Braden, C. A. Henao, J. Heltzel, C. C. Maravelias and
J. A. Dumesic, Green Chem., 2011, 13, 1755–1765.
24 S. G. Wettstein, J. Q. Bond, D. M. Alonso, H. N. Pham,
A. K. Datye and J. A. Dumesic, Appl. Catal., B, 2012, 117,
321–329.
25 O. A. Abdelrahman, A. Heyden and J. Q. Bond, ACS Catal.,
2014, 4, 1171–1181.
26 L. E. Manzer, Appl. Catal., A, 2004, 272, 249–256.
27 H. Mehdi, V. Fabos, R. Tuba, A. Bodor, L. T. Mika and
I. T. Horvath, Top. Catal., 2008, 48, 49–54.
Abbreviations
GVL γ-valerolactone
PWA phosphotungstic acid
ASA amorphous silica alumina
DC
decarboxylation
PEA pentenoic acid
DPE deprotonation energy
LAO linear alpha olefin
TOF turnover frequency
TPD temperature programmed desorption
IPA isopropylamine
Acknowledgements
The authors acknowledge financial support for this work
from the New York State Energy Research and Development
Authority (NYSERDA, agreement number 25920) and wish to
thank Anargyros Chatzidimitriou for his assistance with cata-
lyst characterization. JQB wishes to thank Will Knaeble,
Raj Gounder, Ray Gorte, Dong Wang, and Jim Dumesic for
sharing insights and helpful suggestions.
28 H. Heeres, R. Handana, D. Chunai, C. B. Rasrendra,
B. Girisuta and H. J. Heeres, Green Chem., 2009, 11, 1247–1255.
29 L. Deng, J. Li, D. M. Lai, Y. Fu and Q. X. Guo, Angew. Chem.,
Int. Ed., 2009, 48, 6529–6532.
30 L. Deng, Y. Zhao, J. A. Li, Y. Fu, B. Liao and Q. X. Guo,
ChemSusChem, 2010, 3, 1172–1175.
References
1 R. H. Grubbs and S. Chang, Tetrahedron, 1998, 54,
4413–4450.
2 B. J. Nikolau, M. A. D. N. Perera, L. Brachova and B. Shanks,
Plant J., 2008, 54, 536–545.
31 X. L. Du, L. He, S. Zhao, Y. M. Liu, Y. Cao, H. Y. He and
K. N. Fan, Angew. Chem., Int. Ed., 2011, 50, 7815–7819.
32 A. M. Hengne and C. V. Rode, Green Chem., 2012, 14,
1064–1072.
3 H. H. Brintzinger, D. Fischer, R. Mulhaupt, B. Rieger and
R. M. Waymouth, Angew. Chem., Int. Ed. Engl., 1995, 34,
1143–1170.
4 S. D. Ittel, L. K. Johnson and M. Brookhart, Chem. Rev.,
2000, 100, 1169–1203.
33 M. G. Al-Shaal, W. R. H. Wright and R. Palkovits, Green
Chem., 2012, 14, 1260–1263.
34 C. Delhomme, L. A. Schaper, M. Zhang-Presse, G. Raudaschl
Sieber, D. Weuster-Botz and F. E. Kuhn, J. Organomet.
Chem., 2013, 724, 297–299.
5 D. Martin Alonso, J. Q. Bond, J. C. Serrano-Ruiz and
J. A. Dumesic, Green Chem., 2010, 12, 992–999.
6 J. Q. Bond, D. Martin Alonso, D. Wang, R. M. West and
J. A. Dumesic, Science, 2010, 327, 1110–1114.
7 D. Martin Alonso, J. Q. Bond, D. Wang and J. A. Dumesic,
Top. Catal., 2011, 54, 447–457.
8 J. Q. Bond, D. Martin Alonso and J. A. Dumesic, in Aqueous
Pretreatment of Plant Biomass for Biological and Chemical
Conversion to Fuels and Chemicals, ed. C. E. Wyman, Wiley
Blackwell, Oxford, 2013.
35 W. H. Luo, U. Deka, A. M. Beale, E. R. H. van Eck,
P. C. A. Bruijnincx and B. M. Weckhuysen, J. Catal.,
2013, 301, 175–186.
36 X. L. Du, Y. M. Liu, J. Q. Wang, Y. Cao and K. N. Fan, Chin.
J. Catal., 2013, 34, 993–1001.
37 E. I. Gurbuz, S. G. Wettstein and J. A. Dumesic,
ChemSusChem, 2012, 5, 383–387.
38 D. Martin Alonso, S. G. Wettstein, M. A. Mellmer,
E. I. Gurbuz and J. A. Dumesic, Energy Environ. Sci., 2013, 6,
76–80.
9 A. Morschbacker, Polym. Rev., 2009, 49, 79–84.
10 B. G. Harvey and H. A. Meylemans, J. Chem. Technol.
Biotechnol., 2011, 86, 2–9.
39 A. Gandini, Polym. Chem., 2010, 1, 245–251.
40 J. P. Lange, W. D. van de Graaf and R. J. Haan,
ChemSusChem, 2009, 2, 437–441.
11 T. P. Vispute, H. Zhang, A. Sanna, R. Xiao and G. W. Huber,
Science, 2010, 330, 1222–1227.
12 A. de Klerk, D. J. Engelbrecht and H. Boikanyo, Ind. Eng.
Chem. Res., 2004, 43, 7449–7455.
41 G. M. G. Maldonado, R. S. Assary, J. A. Dumesic and
L. A. Curtiss, Energy Environ. Sci., 2012, 5, 8990–8997.
42 R. Gounder and E. Iglesia, Chem. Commun., 2013, 49,
3491–3509.
13 A. de Klerk, Energy Fuels, 2006, 20, 1799–1805.
14 A. de Klerk, Energy Fuels, 2006, 20, 439–445.
43 R. T. Carr, M. Neurock and E. Iglesia, J. Catal., 2011, 278,
78–93.
2278 | Catal. Sci. Technol., 2014, 4, 2267–2279
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
44 D. A. Simonetti, R. T. Carr and E. Iglesia, J. Catal., 2012, 285,
19–30.
57 M. A. Vannice, Kinetics of Catalytic Reactions, Springer, New
York, 2005.
45 G. Sastre and A. Corma, J. Mol. Catal. A: Chem., 2009, 305,
3–7.
58 S. Storck, H. Bretinger and W. F. Maier, Appl. Catal., A,
1998, 174, 137–146.
46 E. G. Derouane, J. Mol. Catal. A: Chem., 1998, 134, 29–45.
47 S. Roy, G. Mpourmpakis, D. Y. Hong, D. G. Vlachos, A. Bhan
and R. J. Gorte, ACS Catal., 2012, 2, 1846–1853.
48 J. F. DeWilde, H. Chiang, D. A. Hickman, C. R. Ho and
A. Bhan, ACS Catal., 2013, 3, 798–807.
49 E. J. Hensen, D. G. Poduval, V. Degirmenci, D. A. J. Michel
Ligthart, W. Chen, F. Maugé, M. S. Rigutto and J. R. V. Veen,
J. Phys. Chem. C, 2012, 116, 21416–21429.
50 A. Corma, Chem. Rev., 1995, 95, 559–614.
51 W. E. Farneth and R. J. Gorte, Chem. Rev., 1995, 95, 615–635.
52 U. Eichler, M. Brändle and J. Sauer, J. Phys. Chem. B,
1997, 101, 10035–10050.
53 J. Macht, R. T. Carr and E. Iglesia, J. Am. Chem. Soc.,
2009, 131, 6554–6565.
59 J. G. Tittensor, R. J. Gorte and D. M. Chapman, J. Catal.,
1992, 138, 714–720.
60 T. R. Hughes, H. M. White and R. J. White, J. Catal.,
1969, 13, 58–64.
61 C. A. Emeis, J. Catal., 1993, 141, 347–354.
62 T. Barzetti, E. Selli, D. Moscotti and L. Forni, J. Chem. Soc.,
Faraday Trans., 1996, 92, 1401–1407.
63 M. Tamura, K. Shimizu and A. Satsuma, Appl. Catal., A,
2012, 433, 135–145.
64 J. Macht, M. J. Janik, M. Neurock and E. Iglesia, Angew.
Chem., 2007, 119, 8010–8014.
65 R. M. Koros and E. J. Nowak, Chem. Eng. Sci., 1967, 22, 470.
66 R. J. Madon and M. Boudart, Ind. Eng. Chem. Fundam.,
1982, 21, 438–447.
54 D. R. Burfield, K. H. Lee and R. H. Smithers, J. Org. Chem.,
1977, 42, 3060–3065.
55 G. A. Olah and M. Watkins, Org. Synth., 1978, 58, 75.
56 E. W. Albers, R. J. Lussier, J. S. Magee and G. J. Surland, US
Pat. 3974099 A, 1976.
67 J. Jae, G. A. Tompsett, A. J. Foster, K. D. Hammond,
S. M. Auerbach, R. F. Lobo and G. W. Huber, J. Catal.,
2011, 279, 257–268.
68 R. Gounder and E. Iglesia, J. Am. Chem. Soc., 2009, 131,
1958–1971.
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2267–2279 | 2279