Ruthenium-catalyzed hydrogenation of levulinic acid: Influence of the support and solvent on catalyst selectivity and stability

Article abstract of DOI:10.1016/j.jcat.2013.02.003

The catalytic performance of 1 wt% Ru-based catalysts in the hydrogenation of levulinic acid (LA) has been studied at 40 bar H₂ and 473 K. This was done by assessing the influence of the support acidity (i.e., Nb ₂O₅, TiO₂, H-β, and H-ZSM5) and solvent (i.e., dioxane, 2-ethylhexanoic acid (EHA), and neat LA). The Ru/TiO₂ gave excellent selectivity to γ-valerolactone (GVL) (97.5%) at 100% conversion and was remarkably stable even under severe reaction conditions. Ru/H-ZSM5 showed a 45.8% yield of pentanoic acid (PA) and its esters in dioxane, which is the first example of this one-pot conversion directly from LA at 473 K. The gradual deactivation of zeolite-supported catalysts in EHA and neat LA was mainly caused by dealumination, as confirmed by ²⁷Al MAS NMR. Coke buildup originated from angelicalactone and, remarkably, occurred preferentially in the zigzag channels of H-ZSM5 as shown by systematic shifts in the XRD patterns. The GVL ring-opening step is considered to be the rate-determining step on the pathway to PA, necessitating an acidic support.

Full text of DOI:10.1016/j.jcat.2013.02.003

Journal of Catalysis 301 (2013) 175–186
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Ruthenium-catalyzed hydrogenation of levulinic acid: Influence of the support
and solvent on catalyst selectivity and stability
Wenhao Luo a, Upakul Deka a, Andrew M. Beale a, Ernst R.H. van Eck b, Pieter C.A. Bruijnincx a,
,
⇑
Bert M. Weckhuysen a,
⇑
a Inorganic Chemistry and Catalysis Group, Debye Institute for Nanomaterials Science, Utrecht University, Universiteitsweg 99, 3584 CG Utrecht, The Netherlands
b Institute for Molecules and Materials, Radboud University Nijmegen, Heyendaalsweg 135, 6525 AJ Nijmegen, The Netherlands
a r t i c l e i n f o
a b s t r a c t
Article history:
The catalytic performance of 1 wt% Ru-based catalysts in the hydrogenation of levulinic acid (LA) has
been studied at 40 bar H2 and 473 K. This was done by assessing the influence of the support acidity
(i.e., Nb2O5, TiO2, H-b, and H-ZSM5) and solvent (i.e., dioxane, 2-ethylhexanoic acid (EHA), and neat
Received 8 December 2012
Revised 2 February 2013
Accepted 5 February 2013
LA). The Ru/TiO2 gave excellent selectivity to
c-valerolactone (GVL) (97.5%) at 100% conversion and
was remarkably stable even under severe reaction conditions. Ru/H-ZSM5 showed a 45.8% yield of pen-
tanoic acid (PA) and its esters in dioxane, which is the first example of this one-pot conversion directly
from LA at 473 K. The gradual deactivation of zeolite-supported catalysts in EHA and neat LA was mainly
caused by dealumination, as confirmed by 27Al MAS NMR. Coke buildup originated from angelicalactone
and, remarkably, occurred preferentially in the zigzag channels of H-ZSM5 as shown by systematic shifts
in the XRD patterns. The GVL ring-opening step is considered to be the rate-determining step on the path-
way to PA, necessitating an acidic support.
Keywords:
Levulinic acid
Zeolites
Ruthenium
Deactivation
Pentanoic acid
Dealumination
Ó 2013 Elsevier Inc. All rights reserved.
1. Introduction
solvents, such as water [7], alcohols [8,9], dioxane [10], and super-
critical CO2 [11]. Less attention has been devoted to the second
Levulinic acid (LA) has been identified as a promising, sustain-
able platform molecule as it can be produced easily and economi-
part of Scheme 1, in which GVL is further converted to the depicted
deeper hydrogenation products. GVL can, for instance, be further
reduced to 1,4-pentanediol (PD) and subsequently dehydrated to
MTHF [4], a process for which high yields were reported over a
PtRe/C catalyst [12].
cally from lignocellulosic biomass via
hydrolysis process [1–3]. LA can be further converted into many
valuable derivatives of which the conversion to -valerolactone
a simple and robust
c
(GVL) has been studied most. GVL can find use as a renewable sol-
vent, fuel additive or can be subsequently converted into a whole
slate of valuable chemicals, such as 1,4-pentanediol (PD), meth-
yltetrahydrofuran (MTHF), pentanoic acid (PEA), pentanoic acid
(PA), or its esters (PE) [4–6]. The various routes to value-added
chemicals and fuel components from LA by sequential hydrogena-
tion, (de)hydration, and hydrogenolysis reactions are depicted in
Scheme 1. LA can be converted to GVL via either dehydration to
angelicalactone (AL) followed by reduction or via reduction to
4-hydroxypentanoic acid (4-HPA) and subsequent dehydration.
These reactions are generally performed at a moderate reaction
temperature in the range of 298–523 K with hydrogen pressures
from 1 to 150 bar. Many examples of commercial heterogeneous
ruthenium catalysts on neutral supports, mostly Ru/C, were re-
ported to catalyze the conversion of LA into GVL in different
Recently, two new directions have been proposed for the valo-
rization of LA via the alternative PEA pathway. Lange et al. [6] re-
ported on the production of so-called valeric (pentanoic) biofuels,
via Pt/TiO2-catalyzed LA hydrogenation to GVL, followed by GVL
hydrogenation to PA over Pt/H-ZSM-5 and finally esterification to
give the desired pentanoic acid esters (PE). Related to this are
two patents by the Shell laboratories in which a two-step process
of LA to GVL followed by GVL to PA conversion is described. A 70%
yield of PA was in this case achieved with a Pt/H-ZSM-5/SiO2 cat-
alyst [13]. PA and PE could also be obtained in 62% yield in a direct
one-pot reaction from ethyl levulinate over a Ru/H-b/silica catalyst
[14].
The Dumesic group, on the other hand, has shown that GVL can
be converted to PA over a bifunctional metal/acid catalyst, such as
Pd/Nb2O5, after which PA can be upgraded to 5-nonanone by keto-
nization [7]. In both approaches, a combination of acid functional-
ity and hydrogenation function is sought to reduce the oxygen
content and produce better suited fuel components by ring open-
ing, dehydration, or hydrogenation. Alternatively, Dumesic and
⇑
Corresponding authors. Fax: +31 30 251 10 27 (B.M. Weckhuysen).
E-mail addresses: p.c.a.bruijnincx@uu.nl (P.C.A. Bruijnincx), b.m.weckhuysen@
uu.nl (B.M. Weckhuysen).
0021-9517/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2013.02.003

176
W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
H/C=1.2
Angelicalactone (AL)
H/C=2
O
H/C=2.4
O
O
H/C=1.6
O
-H2O
H/C=1.6
O
OH
-H2O
H2
H2
H2
O
O
OH
Methyltetrahydrofuran
(MTHF)
OH
(α)
(β)
1,4-Pentanediol (PD)
O
O
-H2O
O
Levulinic acid (LA)
O
O
H2
OH
γ-Valerolactone
OH
OH
OH
(GVL)
Pentenoic acid isomers
(PEA)
4-Hydroxypentanoic acid
Pentanoic acid
(PA)
(HPA)
H/C=2
H/C=2
H/C=1.8
Scheme 1. Hydrogenation platform of levulinic acid (H/C: hydrogen-to-carbon ratio of each compound).
coworkers also reported an integrated strategy for the production
of C8+ alkenes via GVL ring opening to PEA, followed by decarbox-
ylation and oligomerization of the obtained butenes [15]. This ring-
opening/decarboxylation sequence does not require hydrogen and
is catalyzed by an acidic catalyst, SiO2/Al2O3, in the absence of a
metal hydrogenation function [16].
723 K for 4 h with a heating ramp 2 K/min under N2 (100 mL/
min), followed by reduction at 673 K, for 6 h, under H2 flow
(80 mL/min).
2.2. Catalyst characterization
Indeed, many examples of heterogeneous catalysts are now
available for the selective conversion of LA to GVL, most of which
use a hydrogenating metal (preferably ruthenium) on a neutral
support, usually under dilute and mild solvent conditions. Further-
more, the Shell and Dumesic examples of using a metal on an
acidic support for the ring-opening/hydrogenation of GVL to PA
show the potential of the PEA pathway for the valorization of lev-
ulinic acid. Building on these promising results, the direct, one-pot
reaction of LA to PA, without isolation of the GVL intermediate,
would be desired to further simplify the process. To achieve this,
more insight is required into the influence of the support, in partic-
ular the role of support acidity, on selectivity in LA hydrogenation.
In addition, the stability of the catalyst, both of the metal compo-
nent as well as the support, should be assessed under the harsh
conditions that can be expected for a highly polar and corrosive
LA conversion process. Here, we present the activity, selectivity,
and stability of four 1 wt% ruthenium catalysts on non-acidic
(Nb2O5 and TiO2) and strongly acidic (H-ZSM5 and H-b) supports
in the hydrogenation of LA under increasingly harsh conditions,
using dioxane, 2-ethylhexanoic acid, and neat LA as solvents. It
was found that zeolite-supported catalysts are capable of the direct
synthesis of PA from LA in a one-pot reaction without isolation of
the intermediate GVL. An extensive characterization study of the
fresh and spent catalysts provided insight into the stability and re-
lated deactivation of the catalyst materials under investigation.
2.2.1. Temperature-programmed desorption
Catalyst acidity was investigated by temperature-programmed
desorption (TPD) of NH3 under He flow (25 mL/min) using a
Micromeritics AutoChem II equipped with a TCD detector. 0.15–
0.2 g of catalyst was loaded and dried at 873 K for 1 h, after which
the sample was cooled down to 373 K. Subsequently, pulses of
ammonia were introduced up to saturation of the sample. The tem-
perature-programmed desorption was performed up to 873 K,
with a heating ramp of 5 K/min. The total number of acid sites
(mmol NH3/g zeolite) was determined from the total amount of
desorbed ammonia.
2.2.2. Physisorption measurements
N2 physisorption isotherms were recorded to determine surface
areas and pore volumes using a Micromeritics Tristar 3000 setup
operating at 77 K. All samples were outgassed for 12 h at 473 K
in a nitrogen flow prior to the physisorption measurements. BET
surface areas were determined using 10 points between 0.06 and
0.25. Micropore volumes (cm3/g) were determined by t-plot anal-
ysis for t between 3.5 and 5.0 Å to ensure inclusion of the mini-
mum required pressure points.
2.2.3. Atomic absorption spectrophotometry
The liquid phase was tested after reaction by flame atomic
absorption spectrophotometry (AAS) for the presence of Ru and
Al. All elements (Ru, Al, Fe, and Cr) were analyzed on a ContrAA
700 apparatus using the following conditions: Ru: 349.9 nm
(air/acetylene flame); Al: 396.1 nm (N2O/acetylene flame); Fe:
248.3 nm (air/acetylene flame); Cr: 357.9 nm (N2O/acetylene
flame).
2. Experimental section
2.1. Catalyst preparation
Four 1 wt% Ru catalysts on supports of different acidity were
prepared. The supports were treated as follows before wet impreg-
nation: TiO2 (P25, Evonik) was dried at 393 K for 4 h; Nb2O5 was
obtained by calcination of niobic acid (HY-340, CBMM) at 673 K
for 4 h under air flow with a temperature ramp of 10 K/min; H-
ZSM5 and H-b (H-ZSM5: CBV2314, Si/Al = 11.5; H-b: CP814E,
Si/Al = 12.5, Zeolyst) were converted into the H-form by heating
the samples at 1 K/min to 393 K for 1 h and at 2 K/min to 823 K
for 4 h. The 1 wt% supported ruthenium catalysts were prepared
by wet impregnation using aqueous solutions of ruthenium nitro-
syl nitrate (RuNO(NO3)3, Alfa Aesar). After evaporation of the
water, the catalysts were dried at 333 K overnight, calcined at
2.2.4. Thermal gravimetric analysis
Thermal gravimetric analysis (TGA) coupled to mass spectrom-
etry (MS) was performed with a Perkin–Elmer Pyris 1 apparatus.
The sample was initially heated to 423 K for 1 h with a tempera-
ture ramp of 10 K/min in a 20 mL/min flow of argon to exclude
physisorbed water and acetone, followed by a ramp of 5 K/min to
873 K in a 10 mL/min flow of oxygen to burn off the coke. Analysis
was performed with a quadrupole Pfeiffer Omnistar mass spec-
trometer, which was connected to the outlet of the TGA apparatus.
Ion currents were recorded for m/z values of 18 and 44.

W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
177
2.2.5. Transmission electron microscopy
3QMAS spectra (see Fig. S11 for an example) and set to a range
from À20 to 8 ppm for octahedral aluminum (Al(VI)), 8–41 ppm
for penta-coordinated aluminum (Al(V)), and 41–70 ppm for tetra-
hedral aluminum (Al(IV)) for Ru/H-b. For both Ru/H-ZSM5 samples,
only a trace amount of Al(V) was detected in the 3QMAS, while the
Al(IV) signal extended down to 20 ppm. Hence, integration limits
were set from À20 to 20 ppm for Al(VI) and 20 to 70 ppm for Al(IV)
for these samples.
Transmission electron microscopy (TEM) measurements in
bright field mode were conducted with a Tecnai 20FEG microscope
operating at 200 kV. The ruthenium particle diameters (more than
200 particles for each sample) were measured using iTEM software
(soft Imaging System GmbH). For non-symmetrical particle shapes,
both the largest and the shortest diameter was measured to obtain
an average value.
2.2.6. X-ray diffraction
2.3. Catalyst testing
Powder X-ray diffraction (XRD) patterns were measured with a
Bruker D2 and D8 Advance powder X-ray diffractometer equipped
with automatic divergence slits, a Vantec detector and a cobalt
In a typical reaction, the batch autoclave reactor was loaded with
catalyst, substrate, and solvent, purged three times with argon after
which the reaction mixture was heated to reaction temperature and
charged with H2 to 40 bar. This was taken as the starting point of the
reaction. After the reaction was cooled to room temperature, the H2
was released, and 2 wt% anisole was added as internal standard. The
catalyst was separated by centrifugation, filtration, and finally
washed with acetone. The reaction products were analyzed using
a Shimadzu GC-2010A gas chromatograph equipped with a CP-
K
a1,2 (k = 1.78897/1.79026 Å) source. Diffraction patterns were
collected between 5–40 or 5–55 2h (°) with an increment of
0.017 (in 2h) and an acquisition time of 1 s per step. Unit cell
parameters on Ru/H-ZSM5 samples were obtained by performing
Le Bail extractions between 5 and 29 2h (°) using the Rietica LHPM
package. Background corrections on all patterns were performed
using a 5th order polynomial. The peak profile parameters were
calculated using a pseudo-Voigt function. The initial lattice param-
eters were taken from the IZA database (www.iza-structure/data-
base) for the MFI framework. Silicon (NBS 640) was used as an
internal standard, physically mixed with the powdered samples
in each measurement. To account for 2h shifts caused by instru-
mental errors, that is, sample height or displacement, the Si
(111) reflection was centered at 33.149 2h for all patterns before
performing any analysis.
WAX 57-CB column (25 m  0.2 mm  0.2
Products were identified with a GCÀMS from Shimadzu with a CP-
m). The gas-phase reac-
lm) and FID detector.
WAX 57CB column (30 m  0.2 mm  0.2
l
tion products were analyzed by an online dual channel Varian
CP4900 micro-GC equipped with a COX column and TCD detector,
for analysis of H2, CO2, CO, and CH4.
Dioxane runs were performed with 10 wt% levulinic acid (2.5 g,
21.5 mmol) in dioxane (22.5 g) over a series of 1 wt% Ru catalysts
(0.3 g) containing different supports. The reactions were run in a
50 mL Parr batch autoclave at a temperature of 473 K for 4 h using
a hydrogen pressure 40 bar and a stirring speed of 1600 rpm.
2-Ethylhexanoic acid runs were performed with 10 wt% levu-
linic acid (6.0 g, 51.7 mmol) in 2-ethylhexanoic acid (54 g) with
1 wt% Ru catalysts (0.6 g). The reactions were run in 100 mL Parr
batch autoclave at a temperature of 473 K for 10 h using a hydro-
gen pressure 40 bar and a stirring speed of 1600 rpm. 1 mL of solu-
tion was sampled at various intervals during the reaction.
Reactions with intermediates GVL (2.2 g, 21.5 mmol) or PEA
(2.2 g, 21.5 mmol) in EHA (22.8 g) were conducted with 0.3 g of
catalyst in the 50 mL Parr batch autoclave.
2.2.7. Infrared spectroscopy
FT-IR spectra were recorded on a Perkin–Elmer 2000 instru-
ment after pyridine desorption at various temperatures. Self-sup-
ported catalyst wafers (12–19 mg/16 mm) were pressed at 3 kbar
pressure for 10 s. The wafer was placed inside a synchrotron cell
with a CaF2 window. The cell was evacuated to 10À6 bar followed
by drying of the sample at 573 K (3 K/min) for 1 h. The cell was
cooled down to 323 K, and the sample was brought into contact
with pyridine vapors (3.1 mbar) for 10 s, kept for 30 min to reach
equilibrium, and followed by vacuum for 30 min in order to re-
move physisorbed and loosely bound pyridine. FT-IR spectra were
recorded under vacuum at different temperature (3 K/min) from
323 to 723 K. For each spectrum, 25 scans were recorded with a
Finally, neat LA runs were performed with LA (20 g, 172 mmol)
over 1 wt% Ru catalysts (0.3 g or 0.5 g). The reactions were run in a
50 mL Parr batch autoclave at a temperature of 473 K for 4 h and
10 h using a hydrogen pressure 40 bar and a stirring speed of
1600 rpm.
resolution of 4 cmÀ1
.
2.2.8. Nuclear magnetic resonance
27Al nuclear magnetic resonance (NMR) measurements were
performed on a 600 MHz Varian solid-state NMR spectrometer
using
a
1.6 mm T3 design probe. The probe was tuned to
3. Results and discussion
599.99 MHz for protons and 156.341 MHz for 27Al. Magic angle
spinning at 25 kHz was employed. All samples were weighed to
enable quantitative measurements. T1 saturation recovery experi-
ments showed that a repetition time of 0.4 s was more than suffi-
cient for all magnetization to relax back to Boltzmann equilibrium.
To ensure that intensities are quantitative regardless of the quad-
rupole coupling constant [17], the single-pulse measurements
were done using rf-field strength of 10 kHz and a pulse length of
3.1. Catalyst characterization
The 1 wt% ruthenium catalysts were synthesized by wet
impregnation of the TiO2, Nb2O5, H-ZSM5 (Si/Al 11.5), and H-b
(Si/Al 12.5) supports. The fresh catalysts were characterized by
XRD, TEM, N2 physisorption and NH3-TPD, IR after pyridine adsorp-
tion and solid-state 27Al (1D MAS, 2D 3QMAS) NMR. The bright
field TEM images of the fresh catalysts (see Fig. S3) show that
the average particle sizes of the Ru/TiO2, Ru/H-ZSM5, and Ru/H-b
were very similar and around 3.5 nm mean particle size. Unfortu-
nately, the ruthenium particle size of the Ru/Nb2O5 catalyst could
not be determined by TEM due to lack of contrast. No Ru phase
could be observed in the XRD diffractograms of the fresh catalysts
(Fig. S6).
2 ls (equaling a tip angle of 7°). Proton decoupling did not improve
spectral quality in the 1D experiments. Care was taken to subtract
the small background signal from the MAS rotor, prior to integra-
tion of the spectra. Intensities were also corrected for the presence
of a small amount of satellite sideband intensity that overlaps with
the central transition signal by reducing the intensity by an
amount equal to the intensity of the 1 sideband; however, this
did not change relative intensities significantly.
The acid properties, namely the quantity and strength of the
acid sites, of the fresh catalysts were determined by NH3-TPD.
Table 1 lists the total amount of acidic sites of the four catalysts
Relative amounts of aluminum were obtained by integrating
the 1D spectra. The limits of integration were determined with

178
W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
under investigation, as well as a distinction between weak and
strong acid sites. Woolery et al. attributed the TPD peak at high
temperature to Brønsted acid sites (strong acid) connected with
framework Al (FAl) sites and the low temperature peak to extra-
framework Al species (EFAl), which act as Lewis acid sites (weak
acid) [18]. Here, the low- and high-temperature regions are defined
as 373–523 K and 523–673 K, respectively (Fig. S2). The differences
in acidity between the different catalysts are evident both from the
amount and from the strength of the acid sites, with acidity
decreasing in the order: Ru/H-ZSM5 > Ru/H-b > Ru/TiO2 > Ru/
Nb2O5. Ru/TiO2 and Ru/Nb2O5 can therefore be considered as
non-acidic catalysts with only a very small amount of Lewis acid
sites and their activity and selectivity in the LA hydrogenation be
compared with the strongly acidic zeolite-supported Ru/H-ZSM5
and Ru/H-b catalysts, which have both Brønsted and Lewis acid
sites in the respective MFI and BEA framework types.
acidic Ru/H-ZSM5 yielded more PA and PE, but at the expense of
a slight increase in loss of mass balance (4% for Ru/H-ZSM5 vs.
3.3% for Ru/H-b). The PE derivatives are the result of esterification
of PA with dioxane decomposition products, as the solvent was
found not to be fully stable in the presence of the acidic catalysts
under the applied conditions. Indeed, decomposition of dioxane
(2–4% in total) to ethanol, butanol isomers, and other dioxane-de-
rived small molecules (2-ethoxy-ethanol, diethyl ether) was de-
tected after the runs. The n-butanol included in the n-butyl ester
of pentanoic acid is probably the product of a Guerbet-type reac-
tion of the ethanol formed. The total combined yields of PA and
PE from LA are 37.5% with Ru/H-b and 45.8% with Ru/H-ZSM5.
These results show that the direct conversion to PA is feasible in
a one-pot reaction already at the fairly low temperature of 473 K.
Interestingly, a metal/zeolite-catalyzed process from GVL to PA
was patented by Shell [14,20], but not a process from LA directly
to PA; furthermore, a 62% yield of PA and PE was claimed over a
1 wt% Ru/H-b/silica catalyst at 523 K and 80 bar H2 with ethyl lev-
ulinate rather than LA as the substrate in a batch autoclave reactor
[14]. The gas-phase composition of the dioxane runs was also
checked by online GC analysis, with only minute amounts of CO
and CO2 (carbon balance <0.03%) being detected, which suggests
that decarboxylation, for example, by PA conversion to butene does
not occur under the applied reaction conditions.
The surface areas and pore volumes of the catalysts were deter-
mined by N2 physisorption and are listed in Table 2. Of the non-
acidic catalysts, Ru/Nb2O5 has a larger surface area than Ru/TiO2,
while Ru/H-b has a higher micropore surface and external surface
area than Ru/H-ZSM5.
3.2. Catalytic reactions in dioxane
3.2.1. Catalyst performance
The influence of the support on catalytic activity and selectivity
was initially screened at 473 K and 40 bar H2 in dioxane as the sol-
vent. To exclude any influence of external mass transfer limita-
tions, reactions were run at different stirrer speeds (Fig. S1) and
to ensure that the hydrogenation reactions take place in the kinetic
regime, all experiments were subsequently run at a stirrer speed of
1600 rpm.
Fig. 1 shows the large differences in selectivity between the four
supported catalysts in the hydrogenation of a 10 wt% solution of
levulinic acid in dioxane. Full conversion of LA was achieved with
all catalysts except for Ru/Nb2O5. The non-acidic-supported cata-
lysts mainly show formation of GVL (yields of 61.8% for Ru/
Nb2O5 and 92.3% for Ru/TiO2) with minor amounts of MTHF
(2.6% for Ru/Nb2O5 and 2.3% for Ru/TiO2) and PD (1.7% for Ru/
TiO2), with mass balances of 92.0 and 96.3% for Ru/Nb2O5 and
3.2.2. Catalyst stability
Spent catalysts were extensively characterized to assess stabil-
ity and detect any changes in either the metal phase or the support.
The ruthenium particle sizes and support morphologies of different
catalysts were examined by TEM (Fig. S3). The average diameters
of the ruthenium particles with standard deviations of the fresh
and spent catalysts are given in Fig. 2. The lack of contrast did
not allow the particle size to be determined for the niobia samples.
The average Ru particle diameter of the fresh catalysts was
around 3.5 nm for all three catalysts. All catalysts showed a small
increase in particle size in the spent catalysts to around 4 nm,
now with a slightly broader distribution around the mean. In gen-
eral, the observed sintering after a 4-h reaction in dioxane was lim-
ited, though. The increased standard deviation of the spent
catalysts nonetheless illustrates that more large Ru particles are
formed during reaction with Ru/TiO2, H-b, and H-ZSM5. The extent
of Ru loss to the liquid phase by leaching was determined by AAS
after 4 h reaction and found to be marginal. Ru/H-ZSM5 showed
the highest loss, but at only 2.4% of the original weight of ruthe-
nium (Table 3).
A comparison of the powder XRD patterns of the fresh and spent
catalysts (Fig. S6) showed patterns similar in peak intensity and
peak center, indicating that there was no phase change of the sup-
ports under the applied conditions and that the zeolites main-
tained their structural integrity. N2 physisorption data provided
insight into the extent of coke buildup. The formation of carbon
residues or coke on the catalyst surface resulted in a decrease in
surface area and pore volume (Table 2) for the niobia and zeo-
lite-supported catalysts, but not for the titania-supported one.
The results obtained in dioxane thus show large selectivity differ-
ences for the ruthenium catalysts on supports of different acidity,
Ru/TiO2, respectively. Trace amounts (<0.1%) of AL (
a and b), but
no 4-HPA, were detected at the beginning of the reaction (i.e., at
low LA conversion levels) with all four catalysts. Ru/TiO2 showed
the highest GVL yield of 92.3% (selectivity = 95.8%) under these
conditions. These results are comparable to previously reported
GVL yields over Ru/TiO2 catalysts in various solvents. Corma
et al., for instance, obtained a 93.0% yield of GVL after 5 h reaction
over a 0.6 wt% Ru/TiO2 at 423 K, 35 bar hydrogen pressure in water
[19]. Palkovits et al. reported a yield of 71.2% after 160 min of reac-
tion time over 5 wt% Ru/TiO2 at 403 K, 12 bar hydrogen pressure in
ethanol and water [9]. Ruthenium supported on the strongly acidic
zeolites showed a quite different product distribution. Full conver-
sion is achieved in both cases, but in addition to GVL, the deep
hydrogenation product PA and its ester derivatives (PE) were ob-
tained in considerable amounts with Ru on H-ZSM5 or H-b. PD
was not observed with the zeolite catalysts, however. The more
Table 1
Acid concentrations calculated from NH3-TPD.
Catalyst
Total acidity (mmol/gcat
)
Low T region (mmol/gcat
)
High T region (mmol/gcat)
Ru/H-ZSM5
Ru/H-b
Ru/TiO2
1.35
1.07
0.26
0.14
0.93
0.71
0.26
0.14
0.42
0.36
0
Ru/Nb2O5
0

W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
179
Table 2
N2 physisorption data of the fresh and spent catalyst materials under investigation.
Catalyst
BET(m2/g)
Micropore surface
area (m2/g)
External surface
area (m2/g)
Micropore volume
(cm3/g)a
Coke contents
(wt%)b
Fresh
Ru/Nb2O5
Ru/TiO2
98
56
98
56
Ru/H-b
Ru/H-ZSM5
515
343
329
258
184
85
0.161
0.126
Spent, 4 h in dioxane
Spent, 10 h in EHA
Spent, 10 h in LA
Ru/Nb2O5
Ru/TiO2
Ru/H-b
70
55
436
229
70
55
172
47
264
182
0.129
0.090
Ru/H-ZSM5
Ru/Nb2O5
Ru/TiO2
Ru/H-b
55
54
478
152
55
54
180
36
1.3
0.8
3.4
5.5
292
116
0.143
0.057
Ru/H-ZSM5
Ru/TiO2
52
52
1.8
Ru/H-b
373
200
172
0.098
10.7
a
Data obtained by the t-plot method.
Data determined by TGA.
b
3.3. Catalytic reactions in 2-ethylhexanoic acid
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
unknown
PE
The reactions run in dioxane were well suited for initial studies
into the selectivity, activity, and stability of the catalysts, but the
use of this particular solvent also showed to have some disadvan-
tages. In particular, in the presence of the strongly acidic catalysts,
dioxane was found to be unstable, hampering product analysis due
to dioxane-derived byproduct formation, that is, the pentanoic acid
esters. The use of an alternative solvent could alleviate this and
would also provide an opportunity to subject the catalysts to
harsher conditions to further assess their stability under the typi-
cally highly polar and acidic conditions encountered in a more
commercial levulinic acid conversion process. With this in mind,
2-ethylhexanoic acid (EHA) was selected as solvent. EHA is stable
under the typical hydrogenation conditions and, having a similar
pKa, is a good mimic of LA. This allowed the study of catalyst activ-
ity and selectivity in the conversion of LA, but, importantly, also of
intermediates such as GVL. In addition, catalyst stability under
conditions that are similar to running the reaction in neat LA could
be determined. The latter is important as supported metal catalysts
can suffer from leaching of both the metal phase as well as of the
support under the corrosive and highly polar liquid-phase condi-
tions typically employed in the conversion of renewable platform
PA
PD
MTHF
GVL
LA
Ru/Nb2O5
Ru/TiO2
Ru/H-ß
Ru/ZSM-5
Fig. 1. Catalytic hydrogenation of 10 wt% levulinic acid in dioxane at 473 K, 40 bar,
4 h of reaction time, 1 wt% Ru on different supports, stirrer speed 1600 rpm. LA:
levulinic acid; GVL:
pentanediol; PA: pentanoic acid; PE: pentanoic acid esters (butyl pentanoate
isomers, 2-ethoxyethyl pentanoate and valeric pentanoate); and unknown: missing
mass balance.
c-valerolactone; MTHF: methyltetrahydrofuran; PD: 1,4-
while the limited leaching and sintering points at good catalyst
stability under the applied reaction conditions.
Fig. 2. Ru particle size (nm) determined by TEM for the fresh and spent catalysts after reaction in dioxane, 2-ethylhexanoic acid (EHA) and neat levulinic acid (NLA) (blank
pattern: fresh catalyst; and diagonal pattern: spent catalyst).

180
W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
Table 3
Ruthenium leaching after reaction in dioxane, 2-ethylhexanoic acid (EHA) and neat levulinic acid (LA) as determined by AAS of the liquid phase after
reaction.
Spent catalyst
Ru leaching, 4 h in dioxane (%)
Ru leaching, 10 h in EHA (%)
Ru leaching, 10 h in neat LA (%)
Ru/TiO2
Ru/Nb2O5
Ru/H-b
2.1
1.8
1.6
2.4
2.9
0.9
0.7
0.9
0.5
–
0.5
–
Ru/H-ZSM5
molecules such as LA, and more insight is needed into possible
deactivation pathways.
of ꢀ20%), which follow the sequence: Ru/H-ZSM5 (0.204 sÀ1) > Ru/
TiO2 (0.164 sÀ1) > Ru/H-b (0.131 sÀ1) > Ru/Nb2O5 (0.088 sÀ1).
Only GVL is observed as the product for both the Ru/Nb2O5 and
the Ru/TiO2 catalysts as the support is too weakly acidic to open
the ring of GVL under the applied conditions. The acid-supported
catalysts Ru/H-ZSM5 and Ru/H-b again showed the formation of
the deep hydrogenation product PA with PA yields of 15.5% (Ru/
H-ZSM5) and 6.3% (Ru/H-b) after 10 h of reaction. Again, only min-
ute amounts of MTHF (<1%) were detected, and no PD was ob-
served. The PA yields in EHA were lower than those observed in
dioxane, illustrating that the substrate LA and the intermediate
GVL, which is the precursor to PA, compete with the solvent for
adsorption on or interaction with the catalyst. Related to this, the
extent and origin of catalyst deactivation is different in this solvent
(see below). The data thus illustrate that while the influence of the
3.3.1. Catalyst performance
Concentration profiles of substrate and products are depicted in
Fig. 3. As with the runs in dioxane, clear differences in selectivity
and activity are observed. Both the Ru/Nb2O5 and the Ru/TiO2 cat-
alysts are highly selective for GVL in EHA, with only very small
amounts of MTHF formed as the only byproduct (<1 mol%). The
reaction over Ru/TiO2 is faster, however, as full conversion is ob-
tained after 4 h, while 2.4 mol% of LA was still left after 5 h reaction
with the Ru/Nb2O5 catalyst. The Ru/H-b catalyst also showed full
LA conversion in 4 h, while Ru/H-ZSM5 is the fastest catalyst
achieving full LA conversion in only 3 h. The activity of the catalyst
is illustrated by the turnover frequencies (TOF) (at LA conversions
Fig. 3. Time profiles of the catalytic hydrogenation of 10 wt% levulinic acid in 2-ethylhexanoic acid. Conditions: 473 K, 40 bar, 10 h of reaction time, 1 wt% Ru loading with
different supports, stirrer speed of 1600 rpm LA: levulinic acid; GVL: -valerolactone; MTHF: methyltetrahydrofuran; PA: pentanoic acid.
c

W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
181
acid functionality is rather subtle in the first LA to GVL step, the
strong acid sites on the zeolites are clearly required for the consec-
utive GVL to PA conversion.
converts GVL to PA in EHA, with a yield of 7.2% after 15 min, which
increases to 34.0% after 10 h. This PA yield is higher than the one
obtained in the LA run, pointing at the key role of the acid sites
in the GVL to PA conversion and partial deactivation by acid-site
loss (be it due to blockage or dealumination) in the first LA to
GVL step (Table 4, entries 4–9).
The selectivity patterns of the different catalysts are compared
at different LA conversion levels in Fig. 4. At low LA conversion
(ꢀ30%), all catalysts show high GVL selectivity (SGVL > 97%), with
only Ru/H-b producing already some PA. No PA is observed at this
point with Ru/H-ZSM5, which might be attributed to a more lim-
ited accessibility of GVL to the acid sites in H-ZSM5. Increasing
LA conversion (ꢀ60% and 90%) leads to increased selectivity for
PA over the acidic zeolite supports, with GVL selectivity still over
98% for the less acidic supports TiO2 and Nb2O5. Full LA conversion
was achieved after 10 h of reaction with all catalysts. Although Ru/
H-b initially produced more PA, PA selectivity was lower for Ru/H-
b than for Ru/H-ZSM5 after 10 h, suggesting that deactivation is
more rapid for Ru/H-b than for Ru/H-ZSM5 (see below).
Unexpectedly, a standard check of the activity of the support it-
self in the reaction (Table 4, entries 10, 11, 13), showed that PA
was also obtained with only H-b present. Trace amounts of residual
Ru were held responsible for this hydrogenation activity, but the
(limited) activity of just H-b remained even after extensive disman-
tling and cleaning of the batch autoclave. AAS analysis of the liquid
phase after reaction did not detect any Ru or Cr [21], but trace
amounts of iron were found(Table S1). The activity in the blank reac-
tions with just the zeolite support is thus attributed to acid-cata-
lyzed ring opening by the zeolite to form PEA, which is
subsequently hydrogenated to PA by any iron leached from the auto-
clave walls. Indeed, runs with PEA showed hydrogenation activity to
yield some PA in the absence of the zeolite (Table 4, entry 16), while
in the presence of the zeolite, equilibration with GVL as well as PA
formation were observed. Importantly, blank reactions (no support,
no catalyst) did not show any conversion (Table 4, entries14 and 15).
The background hydrogenation reaction therefore does not influ-
ence the catalytic results as it is much slower than the Ru/zeolite-
catalyzed hydrogenation reactions. These results show that ring
openingis the rate-limiting step in theGVL to PA conversion and that
PEA hydrogenation is very straightforward. Furthermore, ring open-
ing from GVL to PEA is reversible (entry 17) with the equilibrium
3.3.2. Hydrogenation of intermediates
As is clear from the LA hydrogenation results presented above,
the gradual catalyst deactivation limits the yield of PA under these
more severe conditions. To get more insight into the inherent abil-
ity of the zeolite-supported catalysts to catalyze the steps after GVL
formation, that is, ring opening and hydrogenation to PA, reactions
with the implicated intermediates GVL and PEA were also per-
formed in EHA. Ru/TiO2 and Ru/H-b were chosen to compare the
non-acid and acid-supported catalysts (Table 4).
As expected, PA was not observed with Ru/TiO2 after 4 h or even
10 h of reaction (Table 4, entries 1–3). Ru/H-b slowly, but cleanly
Fig. 4. Selectivity comparison of different catalysts at different levulinic acid (LA) conversion levels; LA conversion at: (a) 30%; (b) 60%; (c) 90%; (d) after 10 h of reaction time
with full conversion of LA.

182
W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
Table 4
Results for the catalytic hydrogenation of
c-valerolactone (GVL) or pentanoic acid (PEA) in 2-ethylhexanoic acid (PA, pentanoic acid; and MTHF, methyltetrahydrofuran).
Entry
Catalyst
Time (h)
C-balance (%)
GVL conv. (%)
PA yield (%)
MTHF yield (%)
PEA yield (%)
GVL
1
2
3
4
5
6
7
8
Ru/TiO2
Ru/TiO2
TiO2
Ru/H-b
Ru/H-b
Ru/H-b
Ru/H-b
Ru/H-b
Ru/H-b
H-b
H-b
H-ZSM5
H-b
No Cat
No Cat
4
10
4
0.25
1
2
3
4
10
4
4
97.9
95.8
99.9
94.5
92.5
93.8
93.7
95.1
93.7
96.3
97.1
99.3
95.2
102.9
99.6
2.3
4.9
0.2
0
0
0
0.2
0.7
0.1
0
0
0
0
0
0
0
0
0
0
0
12.7
17.3
23.1
26.7
28.7
40.5
17.6
19.2
6.0
7.2
9.8
16.9
20.4
23.7
34.0
11.0
14.8
5.3
23.2
0
0
0.1
0.2
0
0
0
0
0.1
0.2
9
0
10
11
12
13
14
15
2.9
1.3
0
2.0
0
4
10
4
30.0
-2.8
0.6
10
0
0
Time (h)
C-balance (%)
PEA conv. (%)
PA Yield (%)
GVL yield (%)
PEA
16
17
H-b
No Cat
4
4
99.3
99.9
99.6
15.5
18.0
12.1
80.9
3.3
Conditions: 473 K, 40 bar, 1 wt% Ru on different supports, stirrer speed of 1600 rpm, 21.6 mmol of GVL or PEA in EHA (equimolar to 10 wt% LA).
favoring GVL under the applied conditions. H-b showed higher yield
of PA than H-ZSM5 (Table 4, entries 10–12), further pointing at the
better accessibility of the acid sites in the larger pore H-b.
3.3.3. Catalyst stability
The stability of the catalysts under these more severe condi-
tions was again studied by TEM and AAS to quantify any sintering
or leaching of ruthenium (Fig. S4, Fig. 2 and Table 3). Similar to the
results obtained in dioxane, limited sintering of the Ru particles
was observed for the spent Ru/TiO2 (particle size increase from
2.9 to 3.8 nm, Table 3) and Ru/H-b (3.8–4.5 nm) after 10 h reaction
in EHA. Some bigger clusters were found on the spent catalysts,
however, causing broadening of the distribution. A major differ-
ence with the reactions run in dioxane is that severe sintering
was observed for the spent Ru/H-ZSM5 catalyst, as TEM showed
that some of the H-ZSM5 particles contained no Ru, while others
showed big clusters of Ru, as confirmed by EDX (Fig. S4). Loss of
ruthenium to the liquid phase was marginal (<3%), nonetheless,
for all four catalysts (Table 3).
The gradual deactivation of the catalyst, in particular with re-
spect to its ability to catalyze the second GVL to PA conversion,
is illustrated by a longer 24 h run with Ru/H-b in which additional
fresh catalyst was added after 4 and 14 h of reaction. Changes in
composition of the liquid phase are depicted in Fig. 5. The reaction
was run with half the amount of catalyst (0.3 g) compared to the
results in Fig. 5 for Ru/H-b with the second half of catalyst (0.3 g)
only being added after 4 h. As a result, conversion dropped from
full (0.6 g catalyst) to 68.9% (0.3 g). The yield of PA after 10 h of
reaction was higher, however, for the sequential addition experi-
ment (10.1% vs. 6.3%), which shows that the deactivation of the
catalyst (by loss of acid sites, see below) happens already in the
first stage of the reaction, which is concerned mostly with the con-
version of LA to GVL. PA production seems to level off after 10 h,
indicating that the acid sites are no longer accessible. Addition of
a fresh batch of catalysts restored activity with PA yields increasing
almost linearly over the next 14 h.
3.3.3.1. Coke formation. The lower PA yield that is observed in EHA
with the Ru/zeolite catalysts points at deactivation of the catalyst
material. In addition to any changes to the metal phase, changes to
the support, for example, deactivation by coking, loss of active sites
by dealumination, or phase changes can also be expected. TGA mea-
surements did show different extents of coke formation on the dif-
ferent supports (Table 2). More carbon residue or coke is deposited
on the acid-supported catalysts. The weight loss measured by TGA
decreases in the order of H-ZSM5 > H-b > Nb2O5 > TiO2. N2 physi-
sorption measurements of the spent catalysts also show a decrease
in surface area and pore volume (Table 2). TGA–MS analysis of the
spent Ru/H-b catalyst (from a run in neat LA, vide infra) provided
some further insight into the probable coke-precursors. Indeed,
the H/C ratio, determined from the CO2 and H2O MS signals, reaches
a plateau at H/C 1.4 (Fig. S10). Given the H/C ratios of the various
intermediates in the LA hydrogenation platform (see Scheme 1),
AL is the only intermediate with a H/C ratio lower than 1.4 and must
therefore be involved in the coke formation (although it is most
probably not the sole source). This is further corroborated by the
GC and GC–MS results obtained at low conversion, in which AL,
but not 4-HPA, is detected in minor amounts.
+ 0.3 g Cat
+ 0.3 g Cat
LA
100
80
60
40
20
0
GVL
MTHF
PA
0
200
400
600
800
Time (min)
1000
1200
1400
1600
The XRD patterns of the spent catalysts showed no evidence for
formation of new phases after 10 h of reaction in EHA (Fig. S7). Clo-
ser inspection of the X-ray diffraction pattern of the spent Ru/H-
ZSM5, however, showed a shift of peaks to lower 2h values (higher
corresponding d-values), suggesting an expansion of the unit cell
(Fig. 6). Considering the respective bond distances of Al–O and
Fig. 5. Catalytic hydrogenation of levulinic acid (LA) and formation of c-valerolac-
tone (GVL), pentanoic acid (PA) and methyltetrahydrofuran (MTHF) as a function of
time at 473 K, 40 bar with 1 wt% Ru/H-b in 2-ethylhexanoic acid with twice the
addition of fresh catalyst.

W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
183
Si–O (ꢀ1.75 Å and ꢀ1.61 Å, respectively), dealumination of frame-
work aluminum would result in the opposite effect, that is, a shift
of peaks to higher 2h values (lower d-values) as a result of a
‘‘shrinking’’ unit cell. As also shown by solid-state 27Al NMR, there
was indeed loss of aluminum (Fig. 8) in the Ru/H-ZSM5 sample.
Further considering the initial amount of Al in the fresh sample
(3.1 wt%), this would not be reflected in lattice parameter changes,
as was already previously noted [22,23].
3.3.3.2. Support dealumination. Although the Ru/H-ZSM5 catalyst
showed more coke deposition (Table 2), a lower PA selectivity at
low LA conversion level (Fig. 4) and more sintering of Ru (Fig. S4)
during the reaction in EHA, a higher yield of PA than with Ru/H-b
was still obtained with this catalyst after 10 h of reaction. As the
GVL to PA conversion is critically dependent on the acid sites of
the zeolite, factors other than active site blockage, for example, loss
of active sites by dealumination should therefore also be taken into
account. Al loss was studied by IR after pyridine adsorption and so-
lid-state 27Al NMR.
However, the porosity studies indicated a significant loss in
the internal pore volume of the H-ZSM5 zeolite, which could
be associated with possible coke buildup within the zeolite
channels. This could lead to an expansion of the unit cell, consis-
tent with the observation from the diffraction patterns. To test
this hypothesis further, the spent Ru/H-ZSM5 sample was ex-
posed to a stream of pure oxygen at 723 K for 4 h to remove
as much coke from the material as possible. The XRD pattern
of the recalcined Ru/H-ZSM5 sample is compared to the fresh
and spent material in Fig. 6. Clearly, the pattern corresponding
to the spent catalyst is shifted to the right as compared to the
fresh catalyst suggesting an ‘‘expanded’’ unit cell for the former.
After recalcination of the spent catalysts, the pattern is compara-
ble, however, to the fresh pattern suggesting the unit cells are
similar in size. Table 5 lists the lattice parameters obtained for
each pattern. While lattice parameter ‘‘a’’ remains constant for
all three patterns, unit cell parameters ‘‘b’’ and ‘‘c’’ are seen to
change (as expected) over the three samples. The spent sample
shows the largest unit cell, while treating the sample in an oxy-
gen-rich feed appears to result in a unit cell comparable to the
fresh sample.
It is important to recall that H-ZSM5 is made up of a 3-
dimensional porous network with large straight channels that
are oriented perpendicular to smaller zigzag channels. From a unit
cell perspective, the former are oriented perpendicular to axes a
and c, while the latter are (almost) perpendicular to axes b and c.
Since the analysis shows lattice parameter a to be constant and
parameters b and c to increase in the spent sample, it can be de-
duced that coke forms within the zigzag channels of the zeolite
(oriented parallel to axis a), which attributes to the deactivation
of the catalyst. After treatment in an oxygen-rich flow, most of
the coke gets burnt off, and values for b and c closer to the values
seen for the fresh catalyst are again obtained. The preferential
deposition of coke in the zigzag channels is further substantiated
by the larger shifts of the peaks corresponding to 0kl planes com-
pared to peaks corresponding to h0 0 planes.
The absorbance FT-IR spectra in Fig. 7 show pyridine adsorbed
on the two types of zeolite catalysts. With fresh Ru/H-ZSM5, pyri-
dine adsorption resulted in the formation of protonated (1636,
1489, and 1542 cmÀ1
) and coordinated (1620, 1489, and
1454 cmÀ1) species [24–26]. The former interaction (specifically
the 1542 cmÀ1 vibration) can be assigned to pyridine interacting
with a Brønsted acid site (BAS), while the 1454 cmÀ1 vibration is
indicative of an interaction with Lewis acid sites (LAS). The de-
crease in BAS after the 10 h reaction is very limited, as only a slight
decrease in intensity of the band at 1542 cmÀ1 is observed, while a
small increase in the intensity and a shift to 1450 cmÀ1 of the
vibration originally at 1454 cmÀ1 (Fig. 7a) suggests an increase in
LAS, probably by framework dealumination [27].
On fresh Ru/H-b, vibrations assigned to adsorbed protonated
pyridine (1637, 1490, and 1545 cmÀ1) and coordinated pyridine
(1622, 1490, and 1455 cmÀ1) is observed as well. After reaction, a
decrease in all the bands is seen with spent Ru/H-b, indicating loss
of acid sites. The two new bands at 1603 and 1446 cmÀ1 (Fig. 7b)
most likely correspond to pyridine adsorbed on a new type of
LAS in the spent catalyst, as a result again of dealumination [28].
The severe loss of BAS and LAS can also be seen in the FT-IR spectra
of the Ru/H-b catalyst without pyridine, as the bands at 3779 (LAS),
3662 (LAS), and 3607 cmÀ1 (BAS) corresponding to the various OH
groups that are observed in the fresh catalyst, disappear after reac-
tion in EHA (Fig. S9).
The 1D 27Al MAS NMR spectra of the fresh and spent Ru/zeolite
catalysts are depicted in Fig. 8. The 27Al 3QMAS NMR spectra of
fresh and spent Ru/H-ZSM5 show two resonances in both the tet-
rahedral and the octahedral aluminum region, a broad and a nar-
row one (see Fig. S11). The sharp resonance at 54 ppm is
assigned to tetrahedral framework aluminum species (Al(IV), FAl,
BAS), while the broad tetrahedral resonance with an isotropic shift
of 57 ppm represents distorted extraframework aluminum species
(Al(IV), EFAl, LAS) [29,30]. There are also two types of octahedral
Si
Re-
calcined
Recalcined
Spent
Spent
Fresh
Fresh
26.4 26.6 26.8 27.0 27.2 27.4 27.6 27.8 28.0 28.2 28.4 28.6 28.8 29.0
2 Theta
5
10
15
20
25
30
35
2 Theta
Fig. 6. Top: XRD powder diffraction patterns of a fresh Ru/H-ZSM5 catalyst, after reaction in 2-ethylhexanoic acid (EHA), and after recalcination. Silicon (NBS 640) was used as
internal standard by physical mixing.

184
W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
Fig. 7. FT-IR absorbance spectra of adsorbed pyridine of (a) Ru/H-ZSM5 and (b) Ru/H-b (top: fresh catalyst; bottom: spent catalyst after reaction in 2-ethylhexanoic acid).
Fig. 8. Top: 27Al MAS NMR spectra of fresh and spent Ru catalysts. (a) Fresh Ru/H-ZSM5; (b) spent Ru/H-ZSM5 (reaction in 2-ethylhexanoic acid for 10 h); (c) fresh Ru/H-b; (d)
spent Ru/H-b (reaction in 2-ethylhexanoic acid for 10 h); (e) spent Ru/H-b (reaction in neat levulinic acid for 10 h). Bottom: relative amounts of tetrahedral, penta- and
octahedral coordinated aluminum normalized to the aluminum content in the fresh Ru/H-ZSM5 catalyst as determined by 27Al MAS NMR (accuracy 2%).
aluminum to be seen: a narrow line originating from octahedral
aluminum (Al(VI), EFAl, LAS) in a highly symmetric environment,
such as for instance Al(OH)3Á3H2O in addition to a broad octahedral
resonance also at 0 ppm that has been assigned to extra framework

W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
185
Table 5
acid sites of Ru/H-b hampers the GVL to PEA conversion (Scheme 1).
Although Ru/H-ZSM5 suffers more from sintering and coke forma-
tion (Table 2), less acid sites are lost, resulting in a higher PA yield
in EHA.
Unit cell lattice parameters of the 1 wt% Ru/H-ZSM5 catalyst before and after reaction
and after regeneration.
Ru/H-ZSM5 samples
Lattice constants (Å)
a
b
c
3.4. Catalytic reactions in neat levulinic acid
Fresh
Spent
Recalcined
20.100
20.100
20.101
19.909
19.957
19.924
13.376
13.419
13.399
3.4.1. Catalyst performance
The acidity of the support was also shown to influence activity
and selectivity in the catalytic hydrogenation of neat LA by Ru/TiO2
or Ru/H-b (Table 6). Indeed, Ru/H-b showed higher activity for the
hydrogenation of LA. Reaching full conversion with 0.3 g Ru/H-b in
4 h (TOF > 0.403 sÀ1) (Table 6, entry 4), compared to 57.8% LA con-
version with 0.3 g Ru/TiO2 (TOF = 0.233 sÀ1) (Table 6, entry 1). Con-
version could be increased to 98.1% after 4 h when 0.5 g Ru/TiO2
was used (TOF = 0.239 sÀ1) (Table 6, entry 2). These results are sim-
ilar to those reported by Manzer for the conversion of neat LA using
aluminum by some [31], but others have referred to this as frame-
work Al(VI) (Fig. 8a and b) [32,33].
The distorted tetrahedral aluminum is much more prevalent
than the tetrahedral framework Al in the spent Ru/H-ZSM5 which
shows that conversion of FAl to EFAl sites takes place during the
reaction in the hot acid EHA. Not only is tetrahedral FAl converted
to EFAL, but there is also an overall decrease in the amount of alu-
minum as seen by the decrease in intensity of the spectra. The dea-
lumination is more severe for the octahedral species than for the
tetrahedral aluminum (Al(IV): 12% decrease, Al(VI): 35% decrease).
The loss of acid sites for Ru/H-ZSM5 after reaction is also indicated
by the pyridine adsorption data (Fig. 7). Moreover, the reduction in
tetrahedral FAL coincides with the reduction in intensity of the
1542 cmÀ1 FTIR line, both pointing toward a reduction in the num-
ber of BAS. At the same time, the increase in tetrahedral EFAL to-
gether with growth of the band at 1450 cmÀ1 both support an
increase in LAS.
a
5 wt% Ru/Al2O3 catalyst at 473 K and 33.4 bar hydrogen
(TOF = 0.405 sÀ1, only high conversion data available) [37]. Palko-
vits et al. recently showed that full conversion of neat LA can also
be achieved with Ru/C (5 wt%) at very mild conditions (298 K,
12 bar H2), but at the longer reaction time of 50 h [9]. A comparison
of TOF numbers shows that the LA to GVL conversion is accelerated
by the acid sites on Ru/H-b; the TOF numbers obtained in neat LA
runs are larger than those of the EHA runs (Ru/TiO2: 0.239 vs.
0.164; Ru/H-b: >0.403 vs. 0.131), illustrating that the substrate
LA competes with the solvent EHA for adsorption on the catalyst.
Minute amounts of the deeper hydrogenation products PD, PA,
and MTHF were observed with Ru/TiO2 (Table 6, entry 3), while
higher but still low yields of PA were achieved with Ru/H-b with
the yields increasing with catalyst amount and reaction time. PD
was not observed in the runs with Ru/H-b (Table 6, entries 4–6).
The 27Al MAS NMR spectra of the Ru/H-b samples are shown in
Fig. 8c–e. In the fresh Ru/H-B sample (Fig. 8c), three tetrahedral
species can be discerned in the MQMAS. However, these are se-
verely overlapping in the 1D MAS spectra. The tetrahedral region
consists of both FAL and EFAL aluminum. As in ZSM-5, two octahe-
dral species can be detected, both with an isotropic shift around
0 ppm, one type in a highly symmetric environment and one more
distorted. Trace amounts of penta-coordinated aluminum species
Al(V), with a peak center at 33 ppm, are also detected in the 27Al
3QMAS NMR (Fig. S11) [34,35]. After 10 h of reaction in EHA, dea-
lumination is illustrated by the decrease in intensity of all peaks
(Fig. 8d) with a very sharp decrease in six-coordinated suggesting
the loss of both FAl and EFAl species (after 10 h reaction in EHA,
Al(IV): 41%, Al(V): 64%, Al(VI): 72% decrease. The 27Al NMR spec-
trum of the reaction of Ru/H-b in neat levulinic acid (NLA, see be-
low) shows that the loss of acid sites in NLA is less severe than in
EHA (after 10 h in NLA, Al(IV): 25%, Al(V): 30%, Al(VI): 36% de-
crease). This can be attributed to the larger total amount of acid
used (54 g EHA + 6 g LA vs. 20 g LA in the NLA run) and the fact that
in the NLA run, the acid reacts away to form GVL. The 27Al NMR
spectra agree with the pyridine data as they both reflect the loss
of BAS as well as LAS for Ru/H-b.
3.4.2. Catalyst stability in neat levulinic acid
Interestingly, both very limited sintering (Fig. S5) and limited
Ru loss (Table 3) were observed with Ru/TiO2 in pure LA. Ru/TiO2
thus showed good stability not only in dioxane, but also under
more severe conditions such as in EHA or in neat LA. Some sinter-
ing was observed with Ru/H-b as in EHA, but less Al was lost to the
liquid phase compared to the EHA runs, which can again be attrib-
uted to the disappearance of the acid function upon conversion of
LA to GVL. No phase change occurred for both supports (Fig. S8). A
very small drop in surface area was found for the titania catalyst,
consistent with the limited formation of coke that was observed
by TGA. More coke formed on the acidic catalyst, resulting in a lar-
ger drop in surface area and pore volume. More coke was deposited
on both catalysts in neat LA than in EHA (Ru/H-b: 10.7 vs. 3.4 wt%;
Ru/TiO2: 1.8 vs. 0.8 wt%), illustrating that a LA derivative and not
EHA is the precursor for coke formation (Table 2).
H-b is thus more easily dealuminated than H-ZSM5 in the pres-
ence of the hot acid EHA. Müller et al. already showed that H-b was
more easily dealuminated than H-ZSM5 by a hydrothermal treat-
ment in the presence of oxalic acid [36]. The severe dealumination
of H-b corresponds with the lower yield of PA in the EHA runs with
Ru/H-b as compared to Ru/H-ZSM5 (Fig. 4). The continuous loss of
4. Conclusions
Our study demonstrates that the catalytic performance of the
four supported Ru catalysts for the hydrogenation of levulinic acid
(LA) is greatly affected by the nature of the support and the solvent
Table 6
Catalytic performances for the hydrogenation of neat levulinic acid (LA) with GVL,
c
-valerolactone; PA, pentanoic acid; MTHF, methyltetrahydrofuran and PD, 1,4-pentanediol.
Entry
Catalyst
Time (h)
Catalyst amount (g)
C-balance
LA conv. (%)
GVL yield (%)
PA yield (%)
MTHF yield (%)
PD yield (%)
1
2
3
4
5
6
Ru/TiO2
Ru/TiO2
Ru/TiO2
Ru/H-b
Ru/H-b
Ru/H-b
4
4
10
4
4
0.3
0.5
0.5
0.3
0.5
0.5
100.4
99.6
98.9
92.8
91.7
91.3
57.8
98.8
100.0
100.0
100.0
100.0
57.8
97.7
97.5
91.0
89.2
86.6
0.3
0.4
0.4
1.7
2.3
4.0
0.1
0.2
0.4
0.1
0.2
0.7
0.0
0.1
0.6
0.0
0.0
0.0
10
Conditions: 473 K, 40 bar, 0.3 g 1 wt% Ru catalyst, stirrer speed of 1600 rpm, 20 g LA.

186
W. Luo et al. / Journal of Catalysis 301 (2013) 175–186
choice. The non-acidic catalyst materials selectively gave
c-
References
valerolactone (GVL) as the main and final product, while the
zeolite-supported acidic catalyst proved capable of the direct con-
version of LA to pentanoic acid (PA) under relatively mild condi-
tions in dioxane as solvent. The strongly acidic sites on the
support material accelerate the LA to GVL conversion and are
essential for the subsequent, most difficult step in the sequence,
that is, the ring-opening step in the conversion of GVL to PA. PA
can also be obtained under more severe conditions, using 2-ethyl-
hexanoic acid or neat LA as solvent, but yields dropped as a result
of gradual deactivation of the zeolite-supported catalysts.
Although many factors may influence the deactivation of the cata-
lyst material, pyridine adsorption and solid-state 27Al NMR data
clearly show that the deactivation of the catalyst materials can
mainly be attributed to loss of acid sites by dealumination. It
was found that H-ZSM5 was more resistant to dealumination than
H-b. Most coke is formed on the zeolite-supported catalyst materi-
als in neat LA, with XRD indicating preferential coke buildup in the
zigzag channels of H-ZSM5. The results also show that angelicalac-
tone (AL) is involved as an intermediate in the hydrogenation reac-
tion and that it is a direct precursor for the coke that is deposited
mostly on the acidic catalysts. Ru/TiO2 proved to be remarkably
stable and selective for GVL formation, also in the reactions in neat
LA with very limited coke formation, sintering and leaching of
ruthenium being observed. Further studies on the stability of the
mainly the acid sites in the zeolite-supported catalysts, under the
highly polar and corrosive conditions of LA conversion, should im-
prove the performance of these first promising examples for the di-
rect conversion of LA to PA.
[1] B. Girisuta, L.P.B.M. Janssen, H.J. Heeres, Chem. Eng. Res. Des. 84 (2006) 339–
349.
[2] D.W. Rackemann, W.O.S. Doherty, Biofuels, Bioprod. Biorefin. 5 (2011) 198–
214.
[3] A. Corma, S. Iborra, A. Velty, Chem. Rev. 107 (2007) 2411–2502.
[4] J.J. Bozell, L. Moens, D.C. Elliott, Y. Wang, G.G. Neuenscwander, S.W. Fitzpatrick,
R.J. Bilski, J.L. Jarnefeld, Resour. Conserv. Recyc. 28 (2000) 227–239.
[5] I.T. Horvath, H. Mehdi, V. Fabos, L. Boda, L.T. Mika, Green Chem. 10 (2008) 238–
242.
[6] J.-P. Lange, R. Price, P.M. Ayoub, J. Louis, L. Petrus, L. Clarke, H. Gosselink,
Angew. Chem. Int. Ed. 49 (2010) 4479–4483.
[7] J.C. Serrano-Ruiz, D. Wang, J.A. Dumesic, Green Chem. 12 (2010) 574–577.
[8] Z.-p. Yan, L. Lin, S. Liu, Energy Fuels 23 (2009) 3853–3858.
[9] M.G. Al-Shaal, W.R.H. Wright, R. Palkovits, Green Chem. 14 (2012) 1260–1263.
[10] L.E. Manzer, Appl. Catal. A: Gen. 272 (2004) 249–256.
[11] R.A. Bourne, J.G. Stevens, J. Ke, M. Poliakoff, Chem. Commun. (2007) 4632–
4634.
[12] D.C. Elliott, J.G. Frye, US Patent 5883266, Battelle Memorial Institute, 1999.
[13] J.-P. Lange, US Patent 2011/0112326, Shell International B.V., 2011.
[14] P.J. Van den Brink, K.L. Von Hebel, J.-P. Lange, L. Petrus, WO/2006/[067171],
Shell International B.V., 2006.
[15] J.Q. Bond, D.M. Alonso, D. Wang, R.M. West, J.A. Dumesic, Science 327 (2010)
1110–1114.
[16] J.Q. Bond, D.M. Alonso, R.M. West, J.A. Dumesic, Langmuir 26 (2010) 16291–
16298.
[17] E. Lippmaa, A. Samoson, M. Magi, J. Am. Chem. Soc. 108 (1986) 1730–1735.
[18] G.L. Woolery, G.H. Kuehl, H.C. Timken, A.W. Chester, J.C. Vartuli, Zeolites 19
(1997) 288–296.
[19] A. Primo, P. Concepcion, A. Corma, Chem. Commun. 47 (2011) 3613–3615.
[20] P.M. Ayoub, J.-P. Lange, WO2008/142127, Shell International B.V., 2008.
[21] D. Di Mondo, D. Ashok, F. Waldie, N. Schrier, M. Morrison, M. Schlaf, ACS Catal.
1 (2011) 355–364.
[22] J. Kornatowski, W.H. Baur, G. Pieper, M. Rozwadowski, W. Schmitz, A.
Cichowlas, J. Chem. Soc., Faraday Trans. 88 (1992) 1339–1343.
[23] C.S. Triantafillidis, A.G. Vlessidis, L. Nalbandian, N.P. Evmiridis, Micropor.
Mesopor. Mater. 47 (2001) 369–388.
[24] T. Barzetti, E. Selli, D. Moscotti, L. Forni, J. Chem. Soc., Faraday Trans. 92 (1996)
1401–1407.
[25] E.P. Parry, J. Catal. 2 (1963) 371–379.
Acknowledgments
[26] T. Yashima, N. Hara, J. Catal. 27 (1972) 329–333.
[27] A. Martin, U. Wolf, S. Nowak, B. Lücke, Zeolites 11 (1991) 85–87.
[28] J.P. Marques, I. Gener, P. Ayrault, J.C. Bordado, J.M. Lopes, F. Ramôa Ribeiro, M.
Guisnet, Micropor. Mesopor. Mater. 60 (2003) 251–262.
[29] D. Freude, H. Ernst, I. Wolf, Solid State Nucl. Magn. Reson. 3 (1994) 271–
286.
[30] E. Brunner, H. Ernst, D. Freude, T. Fröhlich, M. Hunger, H. Pfeifer, J. Catal. 127
(1991) 34–41.
[31] J. Kanellopoulos, A. Unger, W. Schwieger, D. Freude, J. Catal. 237 (2006) 416–
425.
The authors gratefully thank the Smart Mix Program of the
Netherlands Ministry of Economic Affairs and the Netherlands
Ministry of Education, Culture and Science within the framework
of the CatchBio Program. Cor van de Spek is acknowledged for
the TEM measurements. The authors thank the CatchBio User Com-
mittee for their valuable suggestions and discussions. Support of
NWO for the ‘‘Solid-State NMR facility for Advanced Materials Sci-
ence’’ is gratefully acknowledged. We would like to thank Hans
Janssen, Gerrit Janssen and Jan van Os for technical support.
[32] Z. Yu, A. Zheng, Q. Wang, L. Chen, J. Xu, J.-P. Amoureux, F. Deng, Angew. Chem.
Int. Ed. 49 (2010) 8657–8661.
[33] E.R.H. van Eck, J.A.Z. Pieterse, A.P.M. Kentgens, Solid State Nucl. Magn. Reson.
39 (2011) 99–105.
[34] J.-P. Gilson, G.C. Edwards, A.W. Peters, K. Rajagopalan, R.F. Wormsbecher, T.G.
Roberie, M.P. Shatlock, J. Chem. Soc., Chem. Commun. (1987) 91–92.
[35] J. Chen, T. Chen, N. Guan, J. Wang, Catal. Today 93–95 (2004) 627–630.
[36] M. Muller, G. Harvey, R. Prins, Micropor. Mesopor. Mater. 34 (2000) 135–
147.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2013.02.003.
[37] L.E. Manzer, US Patent 2006100449, E. I. Du Pont de Nemours and Company,
2006.