Selective conversion of furfural to methylfuran over silica-supported NiFe bimetallic catalysts

Article abstract of DOI:10.1016/j.jcat.2011.09.005

The conversion of furfural in H₂ over SiO₂-supported Ni and NiFe bimetallic catalysts has been investigated at 1 bar in the 210-250 °C temperature range. Over the monometallic Ni catalyst, furfuryl alcohol and furan are primary products resulting from hydrogenation and decarbonylation, respectively. These primary products are further converted in secondary reactions. Furan yields C₄ products (butanal, butanol, and butane) via ring opening, while furfuryl alcohol produces 2-methylfuran via CO hydrogenolysis. By contrast, 2-methylfuran is not produced to a great extent on pure Ni at any level of overall conversion. But, on FeNi bimetallic catalysts, the yield of 2-methylfuran greatly increases while the yields of furan and C₄ products decrease. That is, the addition of Fe suppresses the decarbonylation activity of Ni while promoting the CO hydrogenation (at low temperatures) and the CO hydrogenolysis (at high temperatures). DFT analysis of the possible surface species on the mono- and bimetallic surfaces suggests that the differences in selectivity displayed by these catalysts can be attributed to the stability of the η²-(C,O) surface species, which is higher on the NiFe than on pure Ni. As a result, this η²-(C,O) species can be readily hydrogenated to furfuryl alcohol and subsequently hydrogenolyzed to 2-methylfuran on the bimetallic alloy due to a strong interaction between the carbonyl O and the oxyphilic Fe atoms. Without Fe, on the pure Ni surface, the η²-(C,O) species can be converted into a surface acyl species, which can be decomposed to produce furan and CO. Detailed XRD and TPR characterization indicate the formation of FeNi alloys in all the bimetallic catalysts.

Full text of DOI:10.1016/j.jcat.2011.09.005

Journal of Catalysis 284 (2011) 90–101
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Selective conversion of furfural to methylfuran over silica-supported NiAFe
bimetallic catalysts
⇑
Surapas Sitthisa, Wei An, Daniel E. Resasco
School of Chemical, Biological and Materials Engineering and Center for Biomass Refining, University of Oklahoma, Norman, OK 73019, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 14 July 2011
Revised 2 September 2011
Accepted 6 September 2011
Available online 7 October 2011
The conversion of furfural in H2 over SiO -supported Ni and NiAFe bimetallic catalysts has been investi-
gated at 1 bar in the 210–250 °C temperature range. Over the monometallic Ni catalyst, furfuryl alcohol
and furan are primary products resulting from hydrogenation and decarbonylation, respectively. These
4
primary products are further converted in secondary reactions. Furan yields C products (butanal, buta-
nol, and butane) via ring opening, while furfuryl alcohol produces 2-methylfuran via CAO hydrogenoly-
sis. By contrast, 2-methylfuran is not produced to a great extent on pure Ni at any level of overall
conversion. But, on FeANi bimetallic catalysts, the yield of 2-methylfuran greatly increases while the
2
Keywords:
Furfural
Furfuryl alcohol
Methylfuran
NiAFe alloys
Hydrogenolysis
Hydrogenation
Decarbonylation
Bio-oil upgrading
4
yields of furan and C products decrease. That is, the addition of Fe suppresses the decarbonylation activ-
ity of Ni while promoting the C@O hydrogenation (at low temperatures) and the CAO hydrogenolysis (at
high temperatures). DFT analysis of the possible surface species on the mono- and bimetallic surfaces
suggests that the differences in selectivity displayed by these catalysts can be attributed to the stability
2
2
of the
g -(C,O) surface species, which is higher on the NiAFe than on pure Ni. As a result, this g -(C,O)
species can be readily hydrogenated to furfuryl alcohol and subsequently hydrogenolyzed to 2-methyl-
furan on the bimetallic alloy due to a strong interaction between the carbonyl O and the oxyphilic Fe
2
atoms. Without Fe, on the pure Ni surface, the
g
-(C,O) species can be converted into a surface acyl spe-
cies, which can be decomposed to produce furan and CO. Detailed XRD and TPR characterization indicate
the formation of FeANi alloys in all the bimetallic catalysts.
Ó 2011 Elsevier Inc. All rights reserved.
1
. Introduction
Metal catalysts appear as promising candidates for the selective
removal of oxygen from the carbonyl group of furfural under rela-
Furfurals are a family of aromatic compounds produced from
the dehydration of sugars and are major components of bio-oil
1,2]. Due to their high reactivity, these compounds require further
upgrading to more stable products if the goal is to incorporate
them to the gasoline/diesel pool. Several approaches for furfural
conversion have been attempted. Aldol condensation of furfural
with small ketones is a promising approach to produce larger com-
tively mild operating conditions. We have recently shown that dif-
ferent metals exhibit very different product distributions
depending on the type and relative stability of the surface species
that they are able to form. For example, we recently showed that
Cu-based catalysts are highly selective for hydrogenation of furfu-
ral to furfuryl alcohol (>95% selectivity) [7]. Microkinetic studies,
DFT (Density Functional Theory), and IR spectroscopy have been
combined to investigate the nature of the adsorbed species. It
[
8
pounds (C AC15) that may fall in the fuel range [3–6]. High yields
of condensation products by direct condensation of furfural with
acetone have been obtained in the presence of basic catalysts.
However, condensation is only the first step. Selective deoxygen-
ation is necessary to obtain compounds acceptable as components
of transportation fuels. A highly desirable outcome of this deoxy-
genation step is preserving CAC bonds while effectively breaking
the CAO bonds. Therefore, it is important to investigate what cat-
alysts are effective for this task, how do they operate, and what are
the possible reaction pathways.
was found that the adsorption of furfural on Cu results in an
1
g
(O)-surface species, in which the carbonyl group is bound to
the metal through the O lone pair while the rest of the molecule
is pushed away from the surface due to a net repulsion between
1
C and Cu. The preferred adsorption in the
g
(O)-mode is responsi-
ble for the high hydrogenation selectivity to furfuryl alcohol, typi-
cally observed on Cu-based catalysts [8–10]. By contrast, Group
VIII metals (Ni, Pd, and Pt) display a very different behavior. While
at low temperatures, they only exhibit hydrogenation activity,
with furfuryl alcohol as the major product [11–13], at high temper-
atures, the decarbonylation reaction dominates, with furan as the
main product [14].
⇑
Corresponding author. Tel.: +1 405 325 4370.
E-mail address: resasco@ou.edu (D.E. Resasco).
0
021-9517/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2011.09.005

S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
91
From the point of view of fuel production, neither hydrogena-
tion nor decarbonylation is desirable. While the former does not
remove O, the latter loses C in the process. Hydrogenolysis to 2-
methylfuran (via C1AO1 hydrogenolysis) would be the most desir-
able product, since not only has intrinsically good fuel properties
LaB6). Before TEM analysis, the samples were reduced ex situ in
pure H (100 ml/min) at 450 °C for 1 h. The reduced samples were
then mixed with 2-propanol, sonicated, deposited onto the TEM
(Cu) grids, and dried. The BET surface area (S ) was measured by
conventional N physisorption on a Micromeritics ASAP 2010 unit,
2
g
2
(
high octane number, RON = 131, low water solubility, 7 g/L [15])
after evacuation at 350 °C for 3 h.
but also can be considered an archetypical product of the desired
reaction paths in bio-oil upgrading, i.e., removing O while keeping
a high C yield.
2.2. Catalytic activity measurements
Therefore, one of the goals of this study was to find a catalyst
that is able to produce 2-methylfuran while suppressing the for-
mation of furan. Our recent study showed that bimetallic alloys
can greatly alter the furfural reaction paths. For example, we
showed that adding Cu to Pd and forming bimetallic PdACu alloys
greatly suppresses the production of furan from furfural. However,
only furfuryl alcohol, rather that 2-methylfuran, was observed
The vapor-phase conversion of furfural over the NiAFe catalyst
series was evaluated in a tubular quartz reactor. In each run, a pel-
letized catalyst sample (size range: 250–425 lm) was placed at the
center of the reactor tube between two layers of glass beads and
quartz wool and pre-reduced in H2 flow (60 ml/min, Airgas,
99.99%) for 1 h at 450 °C. After reduction, the catalyst was cooled
down to the selected reaction temperature (210–250 °C) under
the same H2 flow. Prior to the reaction, furfural (Sigma–Aldrich,
99.5%) was purified by vacuum distillation to remove residues
and any oligomers formed during storage. The purified liquid
was kept under He atmosphere until its use in the reaction test.
A 0.5 ml/h (0.006 mol/h) flow of liquid furfural was fed continu-
ously from a syringe pump (Cole Palmer) and vaporized into a H2
stream of 60 ml/min. The reaction products were analyzed by on-
line GC (Agilent model 6890), using an HP-5 capillary column
and a FID detector. The carbon balance was checked in every run
and found to be higher than 95% in every case. The product yield
and selectivity for each product were calculated as follows:
[
16]. In the present work, we have investigated the conversion of
furfural over NiAFe bimetallic catalysts. The effect of adding Fe
on the catalytic properties of silica-supported Ni is investigated.
Samples with different Ni/Fe ratios were prepared and characterized
by XRD, TEM, TPR, and BET techniques to evaluate the structural
changes and the extent of alloy formation. Structure–activity rela-
tionships were investigated by combining the results of catalytic
activity measurements with characterization and theoretical DFT
(
Density Functional Theory) calculations.
2
. Experimental
mol of the product produced
Yield ð%Þ ¼
ꢁ 100
mol of furfural fed
2.1. Catalyst synthesis and characterization
mol of the product produced
ꢁ 100
mol of furfural consumed
The NiAFe/SiO
co-impregnation, using an aqueous solution containing both metal
precursors, Ni(NO O (98%, Alfa Aesar) and Fe(NO
98% Sigma–Aldrich). Prior to impregnation, the silica support
SiO , Hisil 233) was dried overnight at 120 °C. Then, the aqueous
2
catalysts were prepared by incipient wetness
Selectivity ð%Þ ¼
3
)
2
ꢀ6H
2
3
)
3 2
ꢀ9H O
(
(
2.3. Density functional theory calculations
2
solution was added to the support, keeping an incipient wetness li-
quid/solid ratio of 1 cc/g. The Ni loading was kept constant at
All DFT calculations were performed with the Vienna ab initio
simulation package (VASP) [17,18]. A spin-polarized GGA PBE func-
tional [19], all-electron plane-wave basis sets with an energy cutoff
of 400 eV, and a projector augmented wave (PAW) method were
adopted [20,21]. The Brillouin-zone of the p(4 ꢁ 4) lateral supercell
was sampled at 3 ꢁ 3 ꢁ 1 k-points using the Monkhorst–Pack
scheme [21]. First-order Methfessel–Paxton smearing [22] of
0.2 eV was employed in the integration to speed up the conver-
gence. The conjugate gradient algorithm was used in the optimiza-
5
.0 wt.% on all samples, while the Fe loading was varied from 0.0
to 5.0 wt.%. After impregnation, the catalysts were first dried over-
night at room temperature and then placed in an oven at 120 °C for
1
2 h. The oven-dried catalysts were finally calcined for 4 h at
5
00 °C, with a linear heating ramp of 10 °C/min, under 100 ml/
min flow of pure air. The catalyst powders thus obtained were
pressed at 1500 psi, crushed, and sieved to 40–60 mesh.
ꢂ4
The reducibility of the calcined samples was determined by
temperature programmed reduction (TPR). In these measure-
ments, 20 mg of a sample was placed in a quartz reactor and
heated at 30 °C/min up to 500 °C under a He flow of 20 ml/min,
and held at this temperature for 1 h. The reactor was then cooled
tion. The convergence threshold was set to 10 eV in total energy
ꢂ2
and 10 eV/Å in force on each atom. All reported energies are
B
extrapolated to k T = 0 eV.
The Ni(111) surface was modeled by a three-layer slab with the
bottom two layers fixed at their equilibrium bulk phase positions
(calculated lattice constant = 3.522 Å), while the top layer was al-
lowed to relax. The two successive slabs were separated by a
10 Å vacuum region to ensure that the adsorbate (i.e., furfural)
and the subsequent slab would not interact. The p(4 ꢁ 4) supercell
has a dimension of 9.962 Å ꢁ 9.962 Å ꢁ 14.068 Å to ensure that the
effect of adsorbate–adsorbate interactions between adsorbed fur-
fural molecules is negligible.
2
down to 30 °C and the sample exposed to a stream of 5% H /Ar
at a flow rate of 20 ml/min. Subsequently, the sample was heated
up to 800 °C at a heating rate of 5 °C/min. The variation in hydro-
gen uptake was monitored on a TCD detector as a function of tem-
perature. The molar H uptake per gram of sample was quantified
2
from the peak area in the TPR profiles and calibrated with a CuO
standard.
Several physical techniques were employed to characterize the
The bimetallic NiFe(111) alloy surface [i.e., Ni0.5Fe0.5(111), sub-
script is omitted thereafter for simplicity] was cleaved from an
optimized NiFe fcc unit cell (calculated lattice constant
a = c = 3.553 Å, b = 3.582 Å) [22] and was modeled in the same
fashion as the pure Ni(111) surface. The p(4 ꢁ 4) supercell has a
dimension of 10.050 Å ꢁ 10.091 Å ꢁ 14.114 Å. Our calculations
show that NiFe fcc unit cell with lattice constant a = b = c = 3.553
or 3.582 Å is slightly higher in energy (0.10 and 0.12 kcal/mol,
respectively) than the one we employed in the calculations with
structure of the NiAFe/SiO
terns (XRD) for all the samples were collected on a D8 Series II X-
ray Diffractometer (BRUKER AXS), using Cu K radiation generated
at 40 kV and 35 mA. The samples were reduced ex situ under pure
(100 ml/min) at 450 °C for 1 h prior to the measurements. The
2
catalysts. X-ray powder diffraction pat-
a
H
2
scans covered the 2h range from 30° to 60°.
Morphology and size of the NiAFe clusters were characterized
by transmission electron microscopy (TEM, JEOL model JEM-2100

9
2
S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
a = c – b. While it is not uncommon that surface reconstruction
may occur on bimetallic clusters, the NiAFe system is one in which
surface segregation is less favored. For example, Ruban et al. [23]
have shown that the surface segregation energies of NiFe alloys
are unfavorable; positive energies of 0.05–0.3 eV suggest antiseg-
regation. In addition, our own calculations show that formation
of islands of a single component on the same plane (i.e., Ni or Fe
patches in the parallel direction) is also thermodynamically unfa-
vorable. The energy cost is 0.39 kcal/mol per surface atom for the
formation of islands. Therefore, a uniform bulk alloy model is a
good representation of the bimetallic clusters, experimentally ver-
ified by XRD. In fact, the same model has been employed in recent
DFT studies [24].
Ni
Ni0.5Fe0.5
Fe
5
wt.% Fe/SiO2
5
5
wt.%Ni-5 wt.% Fe/SiO2
wt.%Ni-2 wt.% Fe/SiO2
The adsorption energy (Eads) in this work is defined as Eads
=
5 wt.%Ni-1 wt.% Fe/SiO
2
E
(al/slab) ꢂ Eslab ꢂ E(al), where E(al/slab), Eslab, and E(al) are the total
energy of furfural/slab, clean surface, and gas-phase furfural in a
supercell.
5
wt.%Ni-0.5 wt.% Fe/SiO2
5
wt.% Ni/SiO2
3
. Results
3.1. Catalyst characterization
4
0
42
44
2
46
48
50
The XRD patterns of monometallic (Ni/SiO
bimetallic (various NiAFe/SiO ) samples, pretreated in H
2 2
2
and Fe/SiO
2
) and
at
50 °C for 1 h, are shown in Fig. 1. No diffraction peaks due to
θ/degree
4
Fig. 1. XRD patterns of monometallic Ni, Fe, and bimetallic NiAFe catalysts reduced
at 450 °Cfor1 h.
2
the SiO support were observed. In agreement with previous obser-
vations [25], the Ni(111) and Fe(110) peaks for the monometallic
samples were observed at 2h = 44.41° and 44.64°, respectively. The
bimetallic NiAFe/SiO
gion, and this gradually shifted from the position of pure Ni
44.41°) to smaller angles as the Fe content increased. This shift
2
samples exhibited a single peak in this re-
larger lattice constants (3.542–3.584 Å) were obtained for the
NiAFe bimetallic samples, in agreement with the DFT calculations
and previous reports [26].
(
gives clear evidence for the formation of NiAFe solid solution al-
loys with fcc structure [25]. The absence of any intensity near
The average particle sizes for Ni and NiAFe alloy particles were
estimated from the XRD peak broadening, using the Scherrer equa-
tion. The values estimated by this method are in the range 10–
12 nm for the monometallic Ni and NiAFe bimetallic samples.
These results agree with those estimated from TEM (see Fig. 2) in
the sense that there are no significant size differences between
monometallic Ni and NiAFe bimetallic particles. By contrast, the
monometallic Fe sample has much larger particles than those of
4
4.64° suggests that no unalloyed Fe is present in significant
amounts (i.e., detectable by XRD). The lattice constants of Ni and
NiAFe alloys were estimated from the position of the XRD peaks
assuming a cubic symmetry for the crystal structure as shown in
Table 1. The lattice constant of Ni is 3.530 Å, which is close to
the value obtained from DFT calculation (3.522 Å). Increasingly
Table 1
Physical properties of catalysts.
Silica-supported catalysts
Metals loading (mmol gcꢂa1t
)
H2 consumption from TPR (mmol gcꢂa1t
)
Diameter (nm)
Lattice constant (Å)
BET (m /g)
2
Ni
Fe
TEM
XRD
XRD
DFT
Std.
5
5
5
5
5
5
wt.%Ni
0.85
0.85
0.85
0.85
0.85
0.00
0.00
0.09
0.18
0.36
0.90
0.90
1.01
1.16
1.27
1.48
2.20
–
16.8
16.9
16.7
17.1
15.7
23.4
11.2
11.9
11.9
10.0
10.0
19.1
3.530
3.542
3.534
3.567
3.584
2.868
3.52
–
–
3.52
–
–
126
130
118
130
115
128
wt.%Ni–0.5 wt.%Fe
wt.%Ni–1 wt.%Fe
wt.%Ni–2 wt.%Fe
wt.%Ni–5 wt.%Fe
wt.%Fe
–
3.55
–
–
a
3.58a
2.87
a
Ni0.5Fe0.5
.
5Ni
5Ni5Fe
5Fe
4
0 nm
40 nm
40 nm
Fig. 2. TEM image of 5 wt.%Ni/SiO
2
, 5 wt.%Ni–5 wt.%Fe/SiO
, and 5 wt.%Fe/SiO catalysts after H pretreatment at 450 °C for 1 h.
2 2 2

S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
93
Ni or NiAFe samples, as shown by both XRD (d = 19.1 nm) and TEM
23.4 nm). This difference also supports the conclusion that the
(
bimetallic samples do not contain significant amounts of unalloyed
Fe, which would tend to form larger particles.
The large extent of alloy formation is further demonstrated by
TPR. As shown in Fig. 3, the reduction profile for the monometallic
Ni catalyst displays a main reduction peak at 365 °C with shoulder
peaks at lower and higher temperatures (336 and 425 °C, respec-
tively). According to previous studies [27,28], the first peak is the
result of an endothermic phase transition occurring simulta-
neously with a partial reduction of the NiO. The main peak is
attributed to the reduction of NiO to Ni metal [29]. The third peak
at higher temperatures could be due to the reduction of small nick-
el oxide crystallites strongly attached to the support, even forming
Ni silicate species (nickel hydroxysilicates, antigorite, or montmo-
rillonite) [30]. The reducibility pattern of unalloyed Fe is very dif-
ferent from that of Ni. As seen in Fig. 3, the broad hydrogen
5
wt.% Fe/SiO2
5
5
wt.%Ni-5 wt.% Fe/SiO2
wt.%Ni-2 wt.% Fe/SiO2
wt.%Ni-1 wt.% Fe/SiO2
5
2
consumption peaks for the reduction of Fe/SiO appeared shifted
5
wt.%Ni-0.5 wt.% Fe/SiO2
to much higher temperatures from those of Ni. Similar behavior
has been reported for monometallic Fe catalysts in previous stud-
ies [31]. The first broad peak ranging from 410 to 570 °C has been
5
wt.% Ni/SiO2
800
ascribed to the reduction of
broad peak ranging from 590 to 730 °C with the maximum at
650 °C is typically attributed to the reduction of iron oxides to
2 3 3 4
a-Fe O into a-Fe O . The second
2
00
400
600
o
Temperature, C
a-
Fe. In sharp contrast to the monometallic Fe sample, the TPR pro-
files of the NiAFe samples do not show any consumption on the
high temperature range, which is typical of unalloyed Fe. Instead,
as shown in Fig. 3, the addition of Fe only causes a gradual shift
to higher reduction temperatures relative to that of pure Ni, in
agreement with previous studies [26,32].
In addition to the consistent trend in peak position, the H up-
2
take linearly increased with increasing Fe loading in the bimetallic
catalysts (see Table 1). Since the Ni loading was kept constant at
Fig. 3. Temperature programmed reduction (TPR) of monometallic Ni, Fe, and
bimetallic NiAFe catalysts.
40
30
20
10
0
Furan
C4 products
FOL
0
2
.85 mmol/g, this proportional increase in H uptake with Fe load-
MF
ing clearly demonstrates the reduction of Fe. The unalloyed Fe is
not able to get reduced by itself below 500 °C, as previously re-
ported [33]. Therefore, the almost complete reduction of Fe ob-
HFOL
served in the bimetallic catalysts demonstrates
a
close
interaction with Ni, consistent with the formation of an alloy.
3.2. Conversion of furfural over monometallic Ni catalyst
The product distribution obtained over the Ni/SiO
2
catalyst at
2
10 °C is shown in Fig. 4. The most abundant products at this tem-
ꢂ1
perature are furfuryl alcohol (FOL, yield = 31% at 0.1 h ) and furan
ꢂ1
(
yield = 32% at 0.1 h ), which as shown in Scheme 1, derive from
the hydrogenation and decarbonylation reactions, respectively.
0
0.03
0.06
0.09
0.12
The next abundant products are those lumped together as C
ucts (yield = 14% at 0.1 h ), which include butanal, butanol, and
4
prod-
W/F (h)
ꢂ1
Fig. 4. Product distribution from the reaction of furfural over a monometallic
4
butane. The C products derive from the ring-opening reaction
5
wt.%Ni/SiO
2
catalyst, at 210 °C, H/Feed ratio = 25, pressure = 1 atm.
via CAO hydrogenolysis of the furan ring (i.e. C2AO2, see
hydrogenation
H2
+
H O
2
OH
Hydrogenolysis
H2
O
O
O
O
-
CO
H2
C4 products
decarbonylation
Ring opening
O
Scheme 1. Reaction pathways for furfural conversion.

9
4
S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
60
50
40
30
20
10
0
Table 2
2
Conversion and yield of products from the reaction of furfural over 5 wt.%Ni/SiO and
Furan
C4 products
FOL
5
wt.%Ni–2 wt.%Fe/SiO
2
catalysts.
Catalysts
5 wt.%Ni/SiO
2
5 wt.%Ni–2 wt.%Fe/SiO
0.025 0.05
28.2 51.1
2
W/F (h)
Conversion (%)
Furan
FOL
MF
0.025
50.9
23.1
7.2
1.4
18.1
–
0.1
96.3
12.1
9.5
39.1
27.6
7.9
MF
3.8
5.0
5.9
6.4
5.0
12.4
1.0
16.8
19.9
2.1
C
4
products
POL
2
Reaction condition: temp. = 250 °C, H /Feed ratio = 25, pressure = 1 atm.
conversion but strongly depend on reaction temperature and type
of metals [16,35].
Finally, it is important to note that on the monometallic Ni cat-
alyst, the increase in yield of 2-methylfuran with temperature is
much less pronounced than that of the other products, reaching
only ꢃ12% at 250 °C.
200
220
240
260
o
Reaction temp. ( C)
3.3. Conversion of furfural (FAL) over NiAFe bimetallic catalysts
Fig. 5. Yield of products from the reaction of furfural over a monometallic 5 wt.%Ni/
SiO catalyst as a function of temperature. Reaction conditions: W/F = 0.1 h, H /Feed
ratio = 25, pressure = 1 atm.
2
2
A comparison of the conversion levels and product distributions
2
obtained at 250 °C over monometallic 5 wt.%Ni/SiO and bimetallic
5
wt.%Ni–2 wt.%Fe/SiO catalysts at various space times (W/F) is
2
shown in Table 2. The first difference observed is in the level of
conversion obtained over the two catalysts under the same reac-
tion conditions. For example, at the same temperature and at W/
F = 0.025 h, the overall furfural conversion on the bimetallic cata-
lyst (i.e., 28.2%) was about half of that obtained on the monometal-
lic Ni catalyst (i.e., 50.9%).
In addition, interesting differences in product distribution are
observed when comparing the two catalysts. Specifically, compar-
ing them at the same level of conversion (ꢃ50%), a clear enhance-
ment in the yield of 2-methylfuran is observed over the bimetallic
catalyst. At the same time, the yield of furan greatly decreases on
this catalyst. Furthermore, as opposed to the case of monometallic
Ni catalyst that did not make any C5 products, the NiAFe catalyst
produced pentanol (POL) in significant amounts, particularly at
high W/F.
Fig. 11a for the assignment of each C, O, and H atom), a typical
reaction characteristic of Ni catalysts [14,34]. Interestingly, no C
products were observed, indicating that the C products mostly de-
rive from furan, rather than from furfural or furfuryl alcohol.
Finally, 2-methylfuran (2MF) produced as a secondary product
by the CAO hydrogenolysis of furfuryl alcohol (i.e. C1AO1) is only
5
4
observed in measurable amounts at high W/F (yield <2% at
ꢂ1
0
.1 h ). Similarly, tetrahydrofurfuryl alcohol (HFOL), resulting
from hydrogenation of the furanyl ring of furfuryl alcohol, is only
ꢂ1
observed in significant amounts at high W/F (yield = 5% at 0.1 h ).
As shown in Fig. 5, a remarkable change in product distribution
is observed as the reaction temperature increases. While the yields
of furan and its secondary ring-opening C products increase with
4
temperature, furfuryl alcohol rapidly decreases, becoming insignif-
icant near 250 °C. It is clear that decarbonylation and hydrogena-
tion are two parallel routes not strongly dependent on
It appears that pentanol is formed in a secondary ring-opening
reaction of the furanyl ring of 2-methylfuran, which is not
6
5
4
3
0
0
0
0
70
60
50
40
30
20
10
Furan
C4 products
FOL
(
a)
(b)
MF
Furan
C4 products
FOL
HFOL
POL
MF
HFOL
POL
20
10
0
0
0
1
2
3
4
5
6
0
1
2
3
4
5
6
Fe loading (wt%)
Fe loading (wt%)
Fig. 6. (a) Yield and (b) selectivity of products from the reaction of furfural over bimetallic NiAFe catalysts as a function of Fe loading. Reaction conditions: temp. = 250 °C, W/
F = 0.1 h, H /Feed ratio = 25, pressure = 1 atm.
2

S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
95
70
60
50
40
30
20
10
0
Furan
C4 products
FOL
MF
5
wt%Ni/SiO
1
wt.%Ni/SiO2
2
+
2
3
4
5
5 wt.%Ni-5 wt.%Fe/SiO 5 wt.%Fe/SiO
2
2
5
wt.%Fe/SiO
2
(physical mixed)
2
Fig. 7. Product distribution from furfural conversion over various types of catalysts. Reaction conditions: temp. = 250 °C, W/F = 0.1 h, H /Feed ratio = 25, pressure = 1 atm.
5
4
3
2
1
0
0
0
0
0
0
present conditions, the monometallic Fe/SiO
for furfural conversion.
To show that the enhanced rate in production of 2-methylfuran
at low Fe loadings catalysts is due to a direct interaction between
2
exhibited no activity
MF
FOL
Furan
C4 products
POL
Fe and Ni, a physical mixture of monometallic 5 wt.%Ni/SiO
wt.%Fe/SiO was tested and the results are compared in Fig. 7.
They confirm that the product distribution and conversion level
are almost the same as those of pure 5 wt.%Ni/SiO . That is, furan
and its related C products are dominant and the production of
-methylfuran is very small. By contrast, on the bimetallic
wt.%Ni–5 wt.%Fe/SiO catalyst, 2-methylfuran is the main prod-
2
and
5
2
2
4
2
5
2
uct. Therefore, it is clear that the change in products selectivity
of Ni catalyst observed in our experiments is due to the formation
of NiAFe bimetallic alloys, which have been undoubtedly identified
by physicochemical techniques.
The effect of reaction temperature on product distribution over
the bimetallic catalyst can be analyzed in Fig. 8. The results show
that while furfuryl alcohol and 2-methylfuran are the main prod-
ucts at low temperatures (210°), the yield of 2-methylfuran signif-
icantly increases as a function of temperature and it becomes the
main product at high temperatures (250 °C). In addition to 2-meth-
ylfuran, the ring-opening products from furan and 2-methylfuran
200
210
220
230
240
o
250
260
Reaction temp. ( C)
Fig. 8. Product distribution as
of furfural over 5 wt.%Ni–2 wt.%Fe/SiO
W/F = 0.1 h, H /Feed ratio = 25, pressure = 1 atm.
a
function of temperature from the reaction
2
also increased with temperature. In contrast to 5 wt.%Ni/SiO cata-
lyst in which furan was the main product at 250 °C, less than 10%
2
.
Reaction conditions: temp. = 250 °C,
2
yield of furan was observed on the bimetallic 5 wt.%Ni–2 wt.%Fe/
SiO catalyst at the same temperature. Again, this result indicates
2
that the presence of NiAFe alloys suppresses the decarbonylation
activity while it promotes C1AO1 hydrogenolysis reactions, for
both the cleavage of the carbonyl group and the opening of the
ring. The fact that hydrogenation to furfuryl alcohol was dominant
at low temperatures while C1AO1 hydrogenolysis dominated at
high temperatures is due to the higher activation of the latter.
Based on the TPR of the NiAFe alloys, we conducted the pre-
reduction at 450 °C before conducting the reaction in the 200–
250 °C temperature range. That is, the alloy is fully reduced before
the reaction starts. One could question whether the NiFe alloy re-
mains reduced under reaction conditions. While the reaction con-
significantly formed over the monometallic Ni catalyst. It is also
interesting to note that the yield of C products also increased by
the addition of Fe, even though the yield of furan was drastically
reduced. Since methane was not observed in the products, the C
products do not result from the C products. Therefore, it can
4
4
5
be concluded that the addition of Fe not only promotes the ring-
opening reaction of 2-methylfuran but also of furan.
The effect of Fe loading in the bimetallic catalysts on product
distribution is summarized in Fig. 6. It is clear that the addition
of Fe on bimetallic catalysts has both a promoting and a suppress-
ing effect on product yield and selectivity. That is, the yield of 2-
methylfuran first increases with increasing Fe content to a maxi-
mum value at 2%Fe but then decreases, while the selectivity to-
ward 2-methylfuran keeps increasing at all Fe contents. In
2
ditions are very reducing (H /Furfural ratio = 25), water is
produced via hydrogenolysis. However, the interaction of water
with Ni and NiAFe is relatively weak. Our DFT calculations for
adsorption of an H
show binding energies of ꢂ6.6 to ꢂ8.4 kcal/mol, respectively,
which indicate that, under reaction conditions, H O should readily
2
O molecule on Ni(111) and NiFe(111) surfaces
contrast, the yield and selectivity of furan and C
4
products contin-
uously decrease with Fe loading. It should be noted that, under the
2

9
6
S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
Table 3
Furfural conversion over various types of reaction conditions.
Treduction (°C)
Pre-reaction steps
Conversion (%)
Yield (%)
Furan
C4 products
MF
FOL
POL
450
450
450
250
–
96
83
38
13
12
10
4
28
28
12
2
39
33
13
2
10
5
7
8
6
2
0
a
2
Pass H through sat. water at 5 °C
a
Flow 5% O
–
2
/He
1
7
Condition: Catalyst: 5 wt.%Ni–2 wt.%Fe/SiO
2 2
, Temp. = 250 °C, W/F = 0.1 h, H /Feed ratio = 25, pressure = 1 atm.
a
For 10 min.
desorb from the metals upon formation. Moreover, previous stud-
ies [36,37] have shown that H O dissociation on Ni(111) has a high
occur as a secondary reaction with the alcohol acting as an inter-
mediate (see Scheme 1).
2
activation barrier (22.4 kcal/mol) compared to that of H
2
dissocia-
Having furfuryl alcohol as the feed, the main primary products
on pure Ni are furan and 2-methylfuran. As the Fe loading in-
creases, the yield of furan decreased while 2-methylfuran in-
creased, as with the furfural feed. The yield of pentanol (POL),
which derives from C2AO2 hydrogenolysis of 2-methylfuran, also
increased significantly when Fe was incorporated.
tion (ꢃ0 on Ni(111)).
To determine how an oxidized surface would behave, we com-
pared the conversion and product distribution of furfural over the
5
2
wt.%Ni–2 wt.%Fe/SiO catalyst with the following pre-reaction
steps:
While similar distribution of products are obtained with furfu-
ryl alcohol and furfural feeds, the most important difference is the
high activity for 2-methylfuran formation from furfuryl alcohol
compared to that from furfural, as shown in Fig. 9b as a function
of Fe loading in the NiAFe bimetallic catalysts. A significantly high-
er yield of 2-methylfuran is obtained from furfuryl alcohol than
from furfural in everyone of the catalysts, but the difference be-
comes more pronounced as the Fe/Ni ratio increases.
(
a) pre-oxidation with water vapor at 250 °C; after the initial
reduction at 450 °C,
(
b) pre-oxidation in 5% O
50 °C,
c) reduction at 250 °C.
2
at 250 °C; after the initial reduction at
4
(
As shown in Table 3, a continuous decrease in activity is ob-
served with increasing extent of oxidation severity. In fact, it is ex-
pected that the more oxidized is the surface, the lower the extent
of hydrogenolysis (i.e. lower MF yield) will be. That is, the fully re-
duced alloy, kept under the reducing conditions, is the one with the
highest hydrogenolysis activity.
3.5. Conversion of tetrahydrofurfuryl alcohol (HFOL) over NiAFe
bimetallic catalysts
It is important to determine whether the much faster deoxy-
genation observed with the alcohol compared to that with the
aldehyde is due to enhanced C1AO1 hydrogenolysis reaction or a
combination of partial hydrogenation/dehydration. One could
speculate that instead of the direct activation of the CAO bond
via a di-sigma intermediate mentioned above, a partial hydrogena-
tion of the furanyl ring, followed by dehydration of the alcohol,
may operate on these catalysts leading to 2-methylfuran. There-
fore, to eliminate this possibility, we measured the conversion of
tetrahydrofurfuryl alcohol (HFOL) over the same 5 wt.%Ni–
3
.4. Conversion of furfuryl alcohol (FOL) over NiAFe bimetallic
catalysts
To further investigate the reaction pathways of oxygenates over
the NiAFe bimetallic catalysts and shed more light on the results
obtained with furfural, we conducted reaction measurements
using additional oxygenated molecules on the same catalyst series.
We first chose to study the conversion of furfuryl alcohol under the
same reaction conditions as those used with furfural to determine
whether the hydrogenation of the carbonyl group makes hydrog-
enolysis easier. If that is the case, C1AO1 hydrogenolysis could
2
2 wt.%Fe/SiO catalyst, under the same reaction conditions as those
used with furfuryl alcohol and furfural. Contrary to the reaction
with furfural and particularly with furfuryl alcohol for which a high
1
00
80
(
a)
Furan
C4 products
MF
HFOL
POL
(b)
Feed:
FAL
FOL
8
6
4
2
0
0
0
0
0
60
40
20
0
0
1
2
3
4
5
6
0
1
2
3
4
5
6
Fe loading (wt%)
Fe loading (wt%)
Fig. 9. (a) Yield of products from the reaction of furfuryl alcohol over NiAFe bimetallic catalysts as a function of Fe loading. (b) Comparison of methylfuran yield from furfural
and furfuryl alcohol feeds. Reaction conditions: temp. = 250 °C, W/F = 0.1 h, H /Feed ratio = 25, pressure = 1 atm.
2

S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
97
1
00
conversion of benzyl alcohol over the same catalysts series as that
used for the furanyl compounds. To determine to what extent the
relative decarbonylation/hydrogenolysis reaction rates can be
modified by the addition of Fe, we evaluated the formation of ben-
zene and toluene as direct products of decarbonylation and
hydrogenolysis, respectively. The major difference with the prod-
ucts from furfural and furfuryl alcohol was the lack of ring-opening
products, which indicate that the aromatic ring remains unper-
turbed during the reaction.
The most important similarity was the enhancement of hydrog-
enolysis and reduction in decarbonylation activity as the Fe/Ni ra-
tio increased. Fig. 10 clearly shows that as the Fe content increases
in the bimetallic catalysts, the product distribution drastically
changes from almost 60% benzene on pure Ni to almost 100% tol-
Benzene
Toluene
8
6
4
2
0
0
0
0
0
2
uene on the 5 wt.%Ni–2 wt.%Fe/SiO catalyst.
3.7. Adsorption of furfural and furfuryl alcohol on Ni(111) and
NiFe(111) surfaces
0
0.5
1
1.5
2
2.5
Fe loading (wt%)
Density Functional Theory (DFT) calculations have been made
to investigate geometries and relative stabilities of the possible
furfural species adsorbed on the surface of pure Ni(111) and a
bimetallic NiFe(111) alloy. We have also evaluated the differences
observed on heats of adsorption and bond lengths when furfuryl
alcohol is considered as an adsorbate. These calculations have
helped us to interpret the changes observed in product selectivities
from monometallic Ni to NiAFe bimetallic catalysts and to eluci-
date possible reaction pathways.
First, the optimized structures of furfural in the gas phase and
adsorbed on the Ni(111) and NiFe(111) surfaces in an upright
and planar configuration are illustrated in Figs. 11 and 12 respec-
tively. The corresponding adsorption energies (Eads) and bond
lengths of a planar configuration are summarized in Table 4. It
can be seen that for both surfaces under the conditions investi-
gated (i.e., low furfural coverage), the upright configuration with
only the carbonyl O sitting on the metal surface is less favored than
the planar adsorption configuration. For example, the adsorption
energies on for upright and planar configurations Ni(111) are
ꢂ8.9 and ꢂ15.6 kcal/mol, respectively. The optimized geometries
of planar configuration show that the carbonyl group (C@O) ad-
sorbed on a bridge site and the furanyl ring plane sitting parallel
to the surface across two 3-fold hollow sites on.
Fig. 10. Yield of products from the reaction of benzyl alcohol over NiAFe bimetallic
catalysts as
W/F = 0.1 h, H
a
function of Fe loading. Reaction conditions: temp. = 250 °C,
/Feed ratio = 25, pressure = 1 atm.
2
yield of 2-methylfuran was observed (ꢃ40%), with the fully satu-
rated tetrahydrofurfuryl alcohol, the yield of 2-methyltetrahydrofu-
ran was very low (ꢃ1.5%) and 2-methylfuran was not observed. The
much lower reactivity of tetrahydrofurfuryl alcohol demonstrates
that dehydration is not an important reaction path on these
catalysts. By contrast, the strong interaction of the furanyl ring
with the metal surface seems to play an important role in the
CAO hydrogenolysis reaction. In fact, it is generally known that
furan has a strong interaction with the surface of Group VIII metals
due to the interaction of the p bonds of the ring with the d orbitals
of the metal [38,39].
3
.6. Conversion of benzyl alcohol (BZOL) over NiAFe bimetallic
catalysts
To investigate whether the presence of an aromatic ring in the
molecule plays a similar role in reactivity as that of the furanyl ring
in the conversion of the oxygenated group, we investigated the
(
a)
Ni
Fe
Eads = -8.9 kcal/mol
Eads = -9.8 kcal/mol
(
b)
(c)
Fig. 11. Optimized furfural structures of (a) gas phase, (b) adsorbed on Ni(111), and (c) adsorbed on NiFe(111) surfaces with an upright configuration.

Ni
NiFe (111
)-I
NiFe (111)-II
9
8
S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
NiFe (111)-III
[Ni-C1-O1-Ni]
[
Ni-C1-O1-Ni]
[Fe-C1-O1-Fe]
[Ni-C1-O1-Fe]
Eads = -15.6 kcal/mol Eads = -20.9 kcal/mol Eads = -21.8 kcal/mol Eads = -20.4 kcal/mol
(
a)
(b)
(c)
(d)
Ni
Fe
(
e)
Fig. 12. Optimized structures of furfural adsorbed on (a) Ni(111) surface and NiFe(111) surface (b–d), and side view (e) with a planar configuration.
As shown in Table 4 for the more stable planar configurations,
This is not the case with the addition of Fe to Ni, which causes a
significant increase in the resulting furfural adsorption energy (i.e.
15.6 kcal/mol on Ni to 20.4–21.8 kcal/mol on NiFe). Some varia-
tions in energy are observed depending on the position of the mol-
ecule on the NiFe(111) surface. Adsorption with the C atom on top
of Ni and the O atom on Fe showed the highest energy (NiFe(111)-
II) indicating a strong interaction between the carbonyl O atom and
Fe on the surface. This result is consistent with those of acrolein
adsorption on Pt and PtFe surfaces [40]. It was concluded that acro-
lein adsorbs in a parallel configuration on Pt(111) surface with
both double bonds interacting with the surface. When Fe atoms
are present in the surface, adsorption geometries including OAFe
interactions become much more stable.
the bond lengths for the CAC bonds in the furanyl ring increase
on both surfaces, suggesting a slight enlargement of the ring,
resulting from the interaction with the surface. In addition, a more
significant increase in bond length upon adsorption is observed for
the CAO bond of the carbonyl group. It is well known that alde-
2
hydes on Group VIII metals adsorb in an
g
-(C,O) configuration
with both C and O atoms linked to the metal surface [16,42]. There-
fore, an increased CAO bond length is expected on the adsorbed
furfural species on Ni. On the other hand, of greater interest is
the small but consistent increase in CAO bond length observed
on the NiAFe bimetallic surfaces compared to the pure Ni surface.
This difference not only shows that the presence of Fe does not in-
2
hibit the formation of the
g
-(C,O) configuration, but rather it en-
Fig. 13 shows calculated projected-DOS for C1AO1 in the car-
bonyl group of the adsorbate and two metal atoms of the metal
bridge site on which they are bound. For comparison with these
electronic structures, the discrete (C1AO1)-2p states for gas-phase
hances the interaction with the carbonyl group. We recently
described an opposite behavior working with the PdACu bimetallic
system [16]. In that case, experimental and theoretical data gave
2
evidence to the inhibition of the formation of
g
-(C,O) when Cu
furfural and the PDOS for the clean metal surfaces are included in
⁄
was added to Pd. Addition of Cu generates a strong repulsion to
Figs. 13f and e, respectively. In particular, it is observed that the
p
-
1
2
the C atom that favors the formation of
g
-(O) over the
g
-(C,O)
states of carbonyl C1AO1, which are unoccupied in the gas-phase
furfural and centered near 3 eV above the Fermi level, are down-
shifted below the Fermi level and significantly broadened in the
species. Previously we had also shown that a similar repulsion oc-
curs between the monometallic Cu(111) surface and the furanyl
ring, due to the overlap of the 3d band of the surface Cu atoms
adsorbed state. This shift favors the electron-back donation from
⁄
and the aromatic furan ring [7]. As a result, the adsorption of fur-
the metal 3d-states to the anti-bonding
p
-states of the carbonyl,
1
fural on Cu(111) is weak and can only occur in the
configuration, via the carbonyl O atom.
g
(O)-aldehyde
which is in line with the observed increases in C1AO1 bond length,
as reported in Table 4. It is interesting to note the slight differences
Table 4
Comparison of bond lengths (Å) of furfural in gas phase and adsorbed on Ni(111) and NiFe(111) surfaces. See Fig. 12 (a–d) for adsorption of furfural in planar configurations and
(
(
f) for the assignment of each C, O, and H atom. A comparison is also made with furfuryl alcohol dissociatively adsorbed on the same surfaces (see Fig. 14). Adsorption energies
ads) in kcal/mol are included.
E
d
Furfural
Furfural
Ni(111)
Furfuryl alcohol
Ni(111)
Gas phase
NiFe(111) I
NiFe(111) II
NiFe(111) III
NiFe(111) II
ClAHl
ClAOl
1.120
1.229
1.448
1.382
1.419
1.373
1.359
1.378
1.103
1.319
1.439
1.440
1.429
1.450
1.465
1.408
1.101
1.340
1.425
1.438
1.448
1.455
1.456
1.389
1.100
1.347
1.440
1.444
1.443
1.449
1.459
1.403
1.102
1.349
1.446
1.448
1.420
1.452
1.476
1.428
1.092
1.377
1.420
1.441
1.437
1.451
1.466
1.404
1.094
1.433
1.429
1.441
1.451
1.454
1.469
1.399
C1AC2
C2AC3
C3AC4
C4AC5
C5AO2
C2AO2
E
ads
kcal/mol
ꢂ15.6
ꢂ20.9
ꢂ21.8
ꢂ20.4
ꢂ27.6
ꢂ27.1

S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
99
Fig. 13. Calculated PDOS for furfural adsorbed on Ni(111) and NiFe(111) surface, corresponding to the adsorption configuration in Fig. 12 (a–d), respectively. PDOS for clean
surface (e) and gas-phase furfural (f) also included for comparison.
in the position of this peak below the Fermi level for the various
adsorbate/metal species. In particular, for Ni and NiFe in the con-
figuration I, it appears around 3.0 eV below the Fermi level, but it
shifts to 3.3 and 3.7 eV for configurations II and III, respectively.
Indicating that certain configurations on the bimetallic surface
may result in a greater extent of back-donation, which may further
weaken the CAO bond.
The experimental results indicate that the CAO hydrogenolysis
is much faster with furfuryl alcohol than with furfural; so, it is
highly possible that the path for formation of 2-methylfuran goes
through an alcohol intermediate. Therefore, we have conducted
additional DFT calculations on Ni(111) and bimetallic NiFe(111)
As summarized in Table 4, the calculations indicate that on both
surfaces, the adsorption strength is significantly enhanced com-
pared to that of furfural. However, the most interesting result is
the significant lengthening of the C1AO1 bond on the bimetallic
NiFe(111) surface compared to that on pure Ni(111) surface (i.e.,
1.433 vs. 1.377 Å). This bond lengthening on the bimetallic surface
reflects a further weakening of the CAO bond, which is consistent
with the higher CAO hydrogenolysis rate observed experimentally.
4. Discussion
The proposed reaction pathways for furfural conversion are
summarized in Scheme 1. Furfuryl alcohol and furan are primary
products typically observed over Ni-based catalysts [14,41], and
they can further convert to 2-methylfuran and ring-opening C4
5 4
by considering a hydroxyalkyl intermediate (C H O(OH)) that is
expected to result from the dissociative adsorption of furfuryl
alcohol (C
5
5
H O(OH)) (see Fig. 14).
(
a)
(b)
O1
O2
C1
Ni
C5
C2
Fe
C4
C3
(
c)
(d)
Fig. 14. Optimized structures of furfuryl alcohol dissociatively adsorbed on Ni(111) surface (a) and NiFe(111) surface (b). Side view of surface and gas-phase hydroxyalkyl
intermediate structures are shown in (c) and (d), respectively.

1
00
S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
H
C
H
H δ-
OH
δ+
O
O
OH
C
H
H2
-H2
O
H2
H
δ-
OH
+
H O
2
δ+
O
CH2
O
C
O
η2(C-O)-furfural
O
O
Δ
-CO
C
H
O
η1(C)-acyl
Scheme 2. Possible species on the surface during conversion of furfural.
products, respectively. We have shown here that this product dis-
tribution can vary significantly with reaction temperature and the
presence of Fe in the bimetallic catalysts.
Our results show that on monometallic Ni catalyst, decarbony-
lation is favored over hydrogenation, particularly at high reaction
temperatures. A behavior that is similar to that of Pd catalysts
of the stability of the surface species. In this case, the DFT calcula-
tions suggest that the oxophilic nature of Fe makes the di-bonded
2
g
(C,O)-furfural more stable than on the pure Ni surface. The long-
2
er C1AO1 bond length of the adsorbed g (C,O)-furfuryl alcohol
indicates that the overall stronger interaction of the C@O group
with the NiAFe alloy surface results in the weakening of the
C1AO1 bond. As a result, it is reasonable to speculate that the for-
mation of acyl intermediate would require a higher energy barrier,
which explains the drastic reduction in the rate of decarbonylation
observed on the bimetallic catalysts (see Scheme 2).
The enhanced yield of C1AO1 hydrogenolysis products over
NiAFe bimetallic catalysts from furfuryl alcohol and benzyl alcohol
further support this concept. On the monometallic Ni catalysts, fur-
an and benzene are the main products resulting from decarbonyla-
tion of furfuryl alcohol and benzyl alcohol, respectively. These
products greatly decrease on the bimetallic catalysts, which mainly
[
14,16]. This behavior can be explained in terms of the stability
of the surface species at different temperatures. Spectroscopic re-
sults and DFT calculations show that, on group VIII metals, alde-
hydes tend to adsorb with the carbonyl group parallel to the
2
surface (
g
(C,O)-aldehyde) [42–44]. As summarized in Scheme 2,
this configuration favors the hydrogenation of the carbonyl group,
yielding furfuryl alcohol, as experimentally observed at low tem-
2
perature. However, at high temperature,
g
(C,O)-furfural tends to
decompose into a more stable acyl species, in which the C atom
of the carbonyl remains strongly attached to the surface. This acyl
species may in fact be a precursor for the decarbonylation reaction,
yielding furan and CO [35]. As shown in Fig. 5, the selectivity to
yield C1AO1 hydrogenolysis products due to the stabilization of
2
the
g
(C,O) intermediate compared to the acyl intermediate.
decarbonylation significantly increases as a function of tempera-
2
ture, which is consistent with an activated conversion of
g
(C,O)-
5
. Conclusion
furfural into the acyl species that increases decarbonylation while
reducing hydrogenation. This observation is in excellent agree-
ment with the HREELS experiments of acetaldehyde adsorbed on
2
The conversion of furfural on pure Ni/SiO catalysts yields pri-
marily decarbonylation products due to the favorable formation
of acyl species, which can readily decompose into furan and CO
at high temperature. In addition, furan can further react to ring-
opening products due to the strong interaction of the furan ring
with Ni surfaces. In contrast, furfuryl alcohol is produced in signif-
icant low amounts, which in turn results in low formation of 2-
methylfuran.
By adding Fe to the catalysts, NiAFe alloys are formed, as evi-
denced by TPR and XRD. The product selectivity is drastically
shifted to 2-methylfuran while decarbonylation products and their
clean Pt(111) [45] which shows that, at low temperatures, an
2
g
(C,O) surface species was formed, but it was readily decomposed
above 330 K, producing an acetyl intermediate that was stable up
to 440 K.
In addition to the differences in interaction with the carbonyl
group itself, different metals have a different extent of interaction
with the furanyl ring. For example, while Cu exerts a repulsion on
the ring [7], Group VIII metals tend to interact strongly with aro-
matic [46] and furanyl rings [16], as well as C@C double bonds
[
47]. In particular, on Ni surfaces the furanyl ring/metal interaction
derivatives are suppressed. This reduction in decarbonylation can
be attributed to an increase of the stability of
is so strong that CAO bond in the ring weakens and leads to a sig-
nificant extent of ring opening (i.e. C2AO2 hydrogenolysis), a reac-
tion that is also favored as temperature increases.
The selectivity to 2-methylfuran is an interesting case. Since, as
shown above, the C1AO1 hydrogenolysis occurs mainly after furfu-
ral has been hydrogenated to furfuryl alcohol and this product is
not favored on pure Ni, particularly at the high temperatures re-
quired for CAO hydrogenolysis, very small amounts of 2-methylfu-
ran are produced on monometallic Ni.
2
g
(C,O)-surface spe-
cies. The increase in the interaction of the carbonyl O with surfaces
through FeAO interaction hinders the formation of acly species,
which is responsible for decarbonylation. In addition, the CAO
hydrogenolysis of the adsorbed furfuryl alcohol is improved by this
interaction.
Acknowledgments
This situation is drastically altered when Fe is added to the Ni-
based catalysts. First of all, decarbonylation is greatly suppressed
on the bimetallic catalysts, which can be again explained in terms
This work was partially supported by the National Science
Foundation EPSCOR (Grant 0814361) and the Department of

S. Sitthisa et al. / Journal of Catalysis 284 (2011) 90–101
[21] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758.
101
Energy (Grant DE-FG36GO88064). The authors are grateful to
[
[
[
22] M. Hayase, M. Shiga, Y. Nakamura, J. Phys. Soc. Jpn. 34 (1973) 925.
23] A.V. Ruban, H.L. Skriver, J.K. Norskov, Phys. Rev. B 59 (1999) 15990.
Dr. Tawan Sooknoi for very useful discussions. The XRD analysis
was conducted with the assistance of Dr. Rolf E. Jentoft and his
encouragement is gratefully acknowledged.
24] C. Fan, X.G. Zhou, D. Chen, H.Y. Cheng, Y.A. Zhu, J. Chem. Phys. 134 (2011)
134704.
[
25] L. Wang, D. Li, M. Koike, S. Koso, Y. Nakagawa, Y. Xu, K. Tomishige, Appl. Catal.
A: Gen. 392 (2011) 248.
References
[26] T. Ishihara, K. Eguchi, H. Arai, Appl. Catal. 30 (1987) 225.
27] Y. Colombo, F. Gazzarini, G. Lanzavecchia, Mater. Sci. Eng. 2 (1967) 125.
[
[
[
[
[
1] R. He, X.P. Ye, B.C. English, J.A. Satrio, Bioresour. Technol. 100 (2009) 535.
2] D.C. Elliott, T.R. Hart, Energy Fuels 23 (2009) 631.
3] J.N. Chheda, J.A. Dumesic, Catal. Today 123 (2007) 59.
[28] J. Szekely, C.I. Lin, H.Y. Sohn, Chem. Eng. Sci. 28 (1973) 1975.
[29] M.A. Ermakova, D.Y. Ermakov, Catal. Today 77 (2002) 225.
[30] B. Mile, D. Stirling, M.A. Zammitta, A. Lovell, M. Webb, J. Mol. Catal. 62 (1990)
179.
4] R.M. West, E.L. Kunkes, D.A. Simonetti, J.A. Dumesic, Catal. Today 147 (2009)
1
15.
[31] H. Jung, W.J. Thomson, J. Catal. 128 (1991) 218.
[
[
[
[
[
5] G.W. Huber, J.A. Dumesic, Catal. Today 111 (2006) 119.
6] L. Faba, E. Diaz, S. Ordonez, Catal. Today 164 (2011) 451.
7] S. Sitthisa, T. Sooknoi, Y. Ma, P.B. Balbuena, D.E. Resasco, J. Catal. 277 (2011) 1.
8] G. Seo, H. Chon, J. Catal. 67 (1981) 424.
[32] A.L. Kustov, A.M. Frey, K.E. Larsen, T. Johannessen, J.K. Norskov, C.H.
Christensen, Appl. Catal. A: Gen. 320 (2007) 98.
[33] E.E. Unmuth, L.H. Schwartz, J.B. Butt, J. Catal. 61 (1980) 242.
[34] Z. Xinghua, W. Tiejun, M. Longlong, W. Chuangzhi, Fuel 89 (2010) 2697.
[35] R.D. Srivastava, A.K. Guha, J. Catal. 91 (1985) 254.
9] R. Rao, A. Dandekar, R.T.K. Baker, M.A. Vannice, J. Catal. 171 (1997) 406.
[
10] B.M. Reddy, G.K. Reddy, K.N. Rao, A. Khan, I. Ganesh, J. Mol. Catal. A 265 (2007)
76.
11] H. Li, H. Luo, L. Zhuang, W. Dai, M. Qiao, J. Mol. Catal. A 203 (2003) 267.
12] B. Liaw, S. Chiang, S. Chen, Y. Chen, Appl. Catal. A 346 (2008) 179.
13] J. Kije n´ ski, P. Winiarek, T. Paryjczak, A. Lewicki, A. Mikołajska, Appl. Catal. A
[36] M. Pozzo, G. Carlini, R. Rosei, D. Alfe, J. Chem. Phys. 126 (2007) 164706.
[37] B.I. Liu, M.A. Lusk, J.F. Ely, J. Phys. Chem. C 113 (2009) 13715.
[38] M.K. Bradley, J. Robinson, D.P. Woodruff, Surf. Sci. 604 (2010) 920.
[39] C.J. Kliewer, C. Aliaga, M. Bieri, W. Huang, C.-K. Tsung, J.B. Wood, K.
Komvopoulos, G.A. Somorjai, J. Am. Chem. Soc. 132 (2010) 13088.
[40] R. Hirschl, F. Delbecq, P. Sautet, J. Hafner, J. Catal. 217 (2003) 354.
[41] R.M. Lukes, C.L. Wilson, J. Am. Chem. Soc. 73 (1951) 4790.
[42] M. Mavrikakis, M.A. Barteau, J. Mol. Catal. A: Chem. 131 (1998) 135.
[43] R. Shekhar, R.V. Plank, J.M. Vohs, M.A. Barteau, J. Phys. Chem. B 101 (1997)
7939.
2
[
[
[
2
33 (2002) 171.
[
[
[
14] S. Sitthisa, D.E. Resasco, Catal. Lett. 141 (2011) 784.
15] Y. Roman-Leshkov, C.J. Barrett, Z.Y. Liu, J.A. Dumesic, Nat. Lett. 447 (2007) 982.
16] S. Sitthisa, T. Pham, T. Prasomsri, T. Sooknoi, R.G. Mallinson, D.E. Resasco, J.
Catal. 280 (2011) 17.
[
[
[
[
17] G. Kresse, J. Hafner, Phys. Rev. B 48 (1993) 13115.
18] G. Kresse, J. Furthmuller, Phys. Rev. B 54 (1996) 11169.
19] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.
20] P.E. Blochl, Phys. Rev. B 50 (1994) 17953.
[44] J.L. Davis, M.A. Barteau, J. Am. Chem. Soc. 111 (1989) 1782.
[45] M.A. Henderson, Y. Zhou, J.M. White, J. Am. Chem. Soc. 111 (1989) 1185.
[46] C. Morin, D. Simon, P. Sautet, J. Phys. Chem. B 108 (2004) 12084.
[47] F. Delbecq, P. Sautet, J. Catal. 211 (2002) 398.