Efficient Decarbonylation of Furfural to Furan Catalyzed by Zirconia-Supported Palladium Clusters with Low Atomicity

Article abstract of DOI:10.1002/cssc.201601232

Decarbonylation of furfural to furan was efficiently catalyzed by ZrO₂-supported Pd clusters in the liquid phase under a N₂atmosphere without additives. Although Pd/C and Pd/Al₂O₃have frequently been used for decarbonylation, Pd/ZrO₂exhibited superior catalytic performance compared with these conventional catalysts. Transmission electron microscopy and X-ray absorption fine structure measurements revealed that the size of the Pd particles decreased with an increase in the specific surface area of ZrO₂. ZrO₂with a high surface area immobilized Pd as clusters consisting of several (three to five) Pd atoms, whereas Pd aggregated to form nanoparticles on other supports such as carbon and Al₂O₃despite their high surface areas. The catalytic activity of Pd/ZrO₂was enhanced with a decrease in particle size, and the smallest Pd/ZrO₂was the most active catalyst for decarbonylation. When CeO₂was used as the support, a decrease in Pd particle size with an increase in surface area was also observed. Single Pd atoms were deposited on CeO₂with a high surface area, with a strong interaction through the formation of a Pd?O?Ce bond, which led to a lower catalytic activity than that of Pd/ZrO₂. This result suggests that zero-valent small Pd clusters consisting of more than one Pd atom are the active species for the decarbonylation reaction. Recycling tests proved that Pd/ZrO₂maintained its catalytic activity until its sixth use.

Full text of DOI:10.1002/cssc.201601232

DOI: 10.1002/cssc.201601232
Full Papers
Efficient Decarbonylation of Furfural to Furan Catalyzed
by Zirconia-Supported Palladium Clusters with Low
Atomicity
Tamao Ishida,*[a, b] Kurumi Kume,[a] Kota Kinjo,[a] Tetsuo Honma,[c] Kengo Nakada,[c]
Hironori Ohashi,[d] Takushi Yokoyama,[a] Akiyuki Hamasaki,[a] Haruno Murayama,[a]
Yusuke Izawa,[e] Masaru Utsunomiya,[f] and Makoto Tokunaga*[a, g]
Decarbonylation of furfural to furan was efficiently catalyzed
by ZrO2-supported Pd clusters in the liquid phase under a N2
atmosphere without additives. Although Pd/C and Pd/Al2O3
have frequently been used for decarbonylation, Pd/ZrO2 exhib-
ited superior catalytic performance compared with these con-
ventional catalysts. Transmission electron microscopy and X-
ray absorption fine structure measurements revealed that the
size of the Pd particles decreased with an increase in the spe-
cific surface area of ZrO2. ZrO2 with a high surface area immo-
bilized Pd as clusters consisting of several (three to five) Pd
atoms, whereas Pd aggregated to form nanoparticles on other
supports such as carbon and Al2O3 despite their high surface
areas. The catalytic activity of Pd/ZrO2 was enhanced with a de-
crease in particle size, and the smallest Pd/ZrO2 was the most
active catalyst for decarbonylation. When CeO2 was used as
the support, a decrease in Pd particle size with an increase in
surface area was also observed. Single Pd atoms were deposit-
ed on CeO2 with a high surface area, with a strong interaction
through the formation of a PdÀOÀCe bond, which led to
a lower catalytic activity than that of Pd/ZrO2. This result sug-
gests that zero-valent small Pd clusters consisting of more
than one Pd atom are the active species for the decarbonyla-
tion reaction. Recycling tests proved that Pd/ZrO2 maintained
its catalytic activity until its sixth use.
Introduction
Transformation of biomass into valuable chemicals has been
important for the development of sustainable living. Furfural is
an attractive feedstock because it can be obtained by the hy-
dration of pentose sugars derived from the nonfood biomass,
hemicellulose.[1] Decarbonylation of furfural to furan followed
by the hydrogenation of furan to tetrahydrofuran (THF) is
a promising alternative pathway to the current industrial pro-
duction of THF from fossil fuel. Thus, the decarbonylation of
furfural has been intensively studied both in the gas[2] and
liquid phase.[3,4] Gas-phase decarbonylation is simple to per-
form and involves easy separation of the catalyst. Zeolites
have been used in this reaction, but a very high temperature
(>3508C) is required.[2] Supported Pd and other metal catalysts
allow the use of lower reaction temperatures,[3] but the reac-
tion is still performed above 2508C, which causes deactivation
of the catalysts and polymerization of furfural and furan deriva-
[a] Dr. T. Ishida, K. Kume, K. Kinjo, Prof. Dr. T. Yokoyama, Dr. A. Hamasaki,
Dr. H. Murayama, Prof. Dr. M. Tokunaga
Department of Chemistry, Graduate School of Science
Kyushu University
Fukushima University
1 Kanayagawa
Fukushima 960–1296 (Japan)
[e] Y. Izawa
744 Motooka
Nishi-ku, Fukuoka 819-0395 (Japan)
E-mail: mtok@chem.kyushu-univ.jp
Mitsubishi Chemical Corporation Yokkaichi Plant
1 Toho-cho
Yokkaichi, Mie 510-8530 (Japan)
[b] Dr. T. Ishida
[f] Dr. M. Utsunomiya
Current address: Research Center for Gold Chemistry, Department of Ap-
plied Chemistry, Graduate School of Urban Environmental Sciences
Tokyo Metropolitan University
1-1 Minami-osawa
Hachioji, Tokyo 192-0397 (Japan)
Mitsubishi Chemical Corporation
1-1 Marunouchi 1-chome
Chiyoda-ku, Tokyo 100-8251 (Japan)
[g] Prof. Dr. M. Tokunaga
International Research Center for Molecular Systems (IRCMS)
Kyushu University
744 Motooka
[c] Dr. T. Honma, Dr. K. Nakada
Japan Synchrotron Radiation Research Institute (JASRI)
SPring-8, 1-1-1 Kouto
Nishi-ku, Fukuoka 819-0395 (Japan)
Sayo, Hyogo 679–5198 (Japan)
Supporting Information and the ORCID identification number(s) for the
author(s) of this article can be found under http://dx.doi.org/10.1002/
cssc.201601232.
[d] Dr. H. Ohashi
Division of Environment System Management, Faculty of Symbiotic Systems
Science
ChemSusChem 2016, 9, 1 – 8
1
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
tives to produce high molecular weight products. In addition,
the presence of H2 in the reaction gas is necessary for main-
taining Pd in its metallic state.
decarbonylation of furfural to furan; however, furfural was
almost completely consumed (entry 1). This result is probably
owing to the polymerization of furan. When the reaction was
performed in the presence of PdO/ZrO2_91 under pressurized
H2, furan was obtained in 32% yield together with furfuryl al-
cohol in 11% yield as a result of the hydrogenation of the
formyl group (entry 2). These results suggested that PdII was
inactive for the decarbonylation, but was reduced by H2 during
the reaction, becoming catalytically active. For entry 3, PdO/
ZrO2_91 was pretreated in cyclohexane under H2 (2 MPa) at
1208C for 3 h prior to use to give Pd/ZrO2_91, and the reaction
was performed under N2 (0.1 MPa). This suppressed the forma-
tion of furfuryl alcohol and afforded furan in 16% yield. Several
solvents other than cyclohexane were examined, and the use
of non-polar (entry 4) and polar solvents (entries 5–7) im-
proved the furan yields. However, no furan was formed when
DMSO was used as the solvent (entry 8). Among the solvents
tested, THF gave the highest furan yield (entry 6). Although
strong coordination of DMSO to Pd suppressed the adsorption
of furfural, the moderate coordination ability of THF might
assist the desorption of CO molecules from Pd when compared
to other solvents. An increase in reaction temperature gave
better results (entry 9). The calcination temperature of Pd/
ZrO2_91 also affected the catalytic activity; the furan yield was
improved by calcination at 5508C (entry 10).
Conversely, liquid-phase decarbonylation takes place under
an inert atmosphere and milder reaction conditions near the
boiling point of furfural (1628C). Heterogeneous Pd catalysts[4]
have been developed for the liquid-phase decarbonylation of
furfural or hydroxymethylfurfural. However, an inorganic base
such as K2CO3 is needed for the neutralization of the trace
amounts of acid in furfural, which would otherwise poison the
Pd catalyst.[4a–c] Recently, decarbonylation of furfural over Pd
catalysts in the absence of a base has been reported. However,
high pressure in supercritical CO2,[4d] high metal loading,[4e–f] or
molecular sieves (MS-4A)[4g] were required. Naturally, the reac-
tion under milder conditions and without additives is more de-
sirable. Both gas- and liquid-phase decarbonylation of furfural
over Pd/C and Pd/Al2O3 have been well-studied, whereas other
metal-oxide-supported Pd catalysts have been scarcely repor-
ted.[4g] Moreover, the effect of the size of the Pd particles has
not been studied in detail. In this work, we examine various
kinds of supported Pd catalysts and discuss the support and
size specificity of Pd particles for the decarbonylation of furfu-
ral. We report that ZrO2 can stabilize Pd0 clusters consisting of
several Pd atoms and that Pd/ZrO2 exhibits higher catalytic ac-
tivity than conventional Pd/C and Pd/Al2O3 catalysts.
Next, the catalytic activities of different metal-oxide-support-
ed Pd catalysts were compared using 0.2 mol% Pd under the
optimized conditions (Table 2, entries 1–7). Pd/ZrO2_91 exhibit-
Results and Discussion
Decarbonylation of furfural
Table 2. Decarbonylation of furfural over supported Pd catalysts.[a,b]
As Pd/ZrO2 exhibited good catalytic performance in an initial
catalyst screening, it was used for the optimization of the reac-
tion conditions (Table 1). Hereafter, catalysts are abbreviated as
Pd/support_specific surface area. PdO/ZrO2 was prepared by
impregnation of Pd(NO3)2 followed by calcination in air at
4008C. PdO/ZrO2 without H2 pretreatment did not catalyze the
Entry
Catalyst
Temp. [8C]
Conv. [%][c]
Yield [%][c]
1
2
3
4
5
6
7
8
9
Pd/C_3000
Pd/MgO_4
150
150
150
150
140
150
150
140
140
140
140
140
140
140
140
150
150
140
140
40
10
20
58
59
58
100
65
24
2
15
57
43
57
82
53
72
79
89
98
37
70
81
trace
14
trace
0
Pd/Al2O3_233
Pd/SiO2_200
Pd/TiO2_312
Pd/Co3O4_1
Pd/ZrO2 _91
Pd/ZrO2_30
Pd/ZrO2 _91
Pd/ZrO2_94
Pd/ZrO2_194
Pd/ZrO2_280
Pd/CeO2_15
Pd/CeO2_81
Pd/CeO2_173
Pd(PPh3)4
Table 1. Optimization of reaction conditions for decarbonylation of furfu-
ral over Pd/ZrO2_91.[a]
90
Entry
Solvent
Temp. [8C]
Conv. [%][b]
Yield [%][b]
10
11
12
13
14
15
16[d]
17[d,e]
18[f]
19[g]
100
100
100
48
1[c]
2[c,d]
3
4
5
cyclohexane
cyclohexane
cyclohexane
toluene
1,4-dioxane
THF
t-BuOH
DMSO
THF
THF
140
140
140
140
140
140
140
140
150
150
91
>99
>99
61
trace
32
16
40
49
74
58
0
90
83
100
trace
15
6
7
8
9
>99
>99
>99
>99
>99
Pd2(dba)3
Pd/ZrO2_280
Pd/ZrO2_280
4
0
78
83
10[e]
[a] Reaction conditions: furfural (1.0 mmol), 1 wt% Pd catalyst (Pd
0.2 mol%), THF (1.5 mL), N2 (0.1 MPa), 12 h. [b] Pd catalysts were prepared
by impregnation followed by air calcination at 5508C for 4 h and pre-
treatment with H2 (2 MPa, 1208C, 3 h) in THF before use. [c] GC analysis
using tridecane as an internal standard. [d] Pd 0.5 mol%. [e] PCy3 was
added as a ligand. [f] PPh3 (0.22 mol%) was added as a ligand. [g] CO
(0.5 MPa).
[a] Reaction conditions: furfural (1 mmol), 1 wt% Pd/ZrO2_91 (Pd
1 mol%), solvent (1.5 mL), N2 (0.1 MPa), 12 h. Pd catalyst was prepared by
impregnation followed by calcination in air at 4008C and pretreatment
with H2 (2 MPa, 1208C, 3 h) before use. [b] GC analysis using tridecane as
an internal standard. [c] PdO/ZrO2_91, 24 h [d] H2 (1 MPa). [e] Calcined in
air at 5508C and pretreated with H2 (2 MPa, 1208C, 3 h) before use.
&
ChemSusChem 2016, 9, 1 – 8
www.chemsuschem.org
2
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Table 3. Scope of substrates for decarbonylation of aldehydes over Pd/
ed the highest catalytic activity, affording furan in 82% yield,
despite a lower Pd loading (entry 7) than in the experiments
reported in Table 1. Although Rauchfuss et al. reported that
furan was obtained in high yield using 5 mol% Pd/C,[4f] Pd/C_
3000 showed a relatively low catalytic activity using a much
lower amount of Pd (entry 1). Pd/MgO showed poor catalytic
activity (entry 2). Pd/Al2O3 has also been frequently used for
furfural decarbonylation, but Pd/Al2O3_233 (entry 3) showed
a low catalytic activity similar to that of Pd/C_3000. Pd/SiO2_
200 (entry 4), Pd/TiO2_312 (entry 5), and Pd/Co3O4_1 (entry 6)
showed moderate catalytic activity. The highest furan yield was
obtained with Pd/ZrO2_91 (entry 7). We also examined the
effect of the specific surface area of Pd/ZrO2 on the catalytic
activity. Support materials with a higher specific surface area
gave better furan yields when compared to the same metal
oxides (entries 8–12). Accordingly, Pd/ZrO2_280 exhibited the
best catalytic performance, giving furan in 98% at full conver-
sion (entry 12). A similar tendency was observed for Pd/CeO2
(entries 13–15).
ZrO2_280.[a]
Entry
Substrate
Product
Conv. [%][b]
100
Yield [%][b]
1
2
3
4
86
69
96
81
100
100
100
5
6
100
100
77
90
[a] Reaction conditions: furfural (1 mmol), 1 wt% Pd/ZrO2_280 (Pd
0.2 mol%), solvent (1.5 mL), 1508C, N2 (0.1 MPa), 12 h. Pd catalysts were
pretreated with H2 (2 MPa, 1208C, 3 h) before use. [b] GC analysis using
tridecane as an internal standard.
The catalytic activity was also compared with that of homo-
geneous Pd catalysts. Pd(PPh3)4 (entry 16) and Pd2(dba)3 (dba:
dibenzylideneacetone) with a PCy3 (tricyclohexylphosphine)
ligand (entry 17) showed negligible or low catalytic activities,
and Pd black was formed in the reaction. The addition of PPh3
as a ligand to the reaction solution in the presence of Pd/
ZrO2_280 suppressed the reaction, probably owing to the
blocking of the Pd active sites by PPh3 (entry 18). When the re-
action was performed under CO (0.1 MPa), the reaction did not
proceed (entry 19). The carbonylation of furan to furfural also
did not proceed over Pd/ZrO2. These results suggested that CO
inhibited the reaction by covering the surface active sites. We
also performed recycling tests using Pd/ZrO2_280. Pd/ZrO2
could be used at least six times without the loss of catalytic ac-
tivity (Figure 1).
nating groups were transformed into the corresponding decar-
bonylated products in good to excellent yields. The decarbony-
lation of 4-nitrobenzaldehyde gave lower yields than the other
substrates, probably owing to its partial decomposition
(entry 2). The yield of 1,3-dimethoxybenzene was slightly de-
creased because of steric hindrance (entry 5).
Characterization of Pd/ZrO2 and Pd/CeO2
High-angle annular dark field scanning transmission electron
microscopy (HAADF-STEM), X-ray absorption fine structure
(XAFS) measurements, and X-ray diffraction (XRD) were used to
characterize the supported Pd catalysts. According to HAADF-
STEM (Figure S1 and S2 in the Supporting Information), large
Pd nanoparticles were deposited onto Pd/MgO (14.5 nm) and
Pd/Co3O4 (7.5 nm), probably owing to the very low specific sur-
face area of the metal oxides. The mean diameters of the Pd
particles in Pd/C, Pd/Al2O3, Pd/SiO2, and Pd/SiO2 were 4.2, 2.6,
3.6, and 4.3 nm, respectively. Conversely, the mean diameters
of the Pd particles in all the Pd/ZrO2 catalysts were in the
range of 1–1.7 nm (Figure S1 and Table 4). A decrease in parti-
cle size below 2 nm markedly enhanced the catalytic activity.
To further elucidate the chemical state and the size of the
Pd particles on ZrO2 and CeO2, XAFS measurements were per-
formed. Pd K-edge X-ray absorption near edge structure
(XANES) spectra of Pd/ZrO2 and Pd/CeO2 are shown in
Figure 2. The XANES spectra of Pd/ZrO2_30 resembled that of
Pd foil and the absorption edges appeared in the same region
as for Pd foil, suggesting the chemical state of Pd in Pd/ZrO2_
30 was Pd0 (Figure 2a). With an increase in the specific surface
area of ZrO2, the peak at 24.40 keV decreased, and the spectra
became featureless, indicating the formation of Pd clusters.
Spectral changes resulting from changes in the specific sur-
face area were also observed for Pd/CeO2. The XANES spectra
of Pd/CeO2_15 also resembled Pd foil (Figure 2b). In contrast,
The scope of the substrates for the decarbonylation of aro-
matic aldehydes catalyzed by Pd/ZrO2_280 (Table 3) was exam-
ined. Substrates bearing electron withdrawing and electron do-
Figure 1. Recycling of Pd/ZrO2_280 for decarbonylation of furfural. Reaction
conditions: furfural (1.0 mmol), 1 wt% Pd catalyst (Pd 0.6 mol%), THF
(1.5 mL), N2 (0.1 MPa), 12 h. After the reaction, the Pd catalyst was collected
by filtration, washed with THF, and dried under vacuum. The catalyst was
then pretreated with H2 (2 MPa, 1208C, 3 h) in THF and used for the next
run.
ChemSusChem 2016, 9, 1 – 8
www.chemsuschem.org
3
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
Table 4. EXAFS fit parameters[a] estimated mean diameters of Pd particles, and specific surface area of Pd/ZrO2 and Pd/CeO2.
[e]
[f]
Sample
Shells
CN[b]
R[c]
[ꢁ]
s2[d]
[ꢁ2]
Rfactor
dXAFS
[nm]
dTEM
[nm]
dCO
SSA[g]
[m2 gÀ1
]
[nm]
Pd foil
PdO
Pd/ZrO2_30
Pd/ZrO2_91
Pd/ZrO2_94
Pd/ZrO2_194
Pd-Pd
Pd-O
12.0[h]
2.739Æ0.003
2.022Æ0.021
2.744Æ0.003
2.749Æ0.006
2.742Æ0.006
2.726Æ0.011
1.993Æ0.028
2.776Æ0.036
2.007Æ0.020
2.740Æ0.004
2.007Æ0.017
3.194Æ0.066
0.0055Æ0.0005
0.0024Æ0.0026
0.0067Æ0.0006
0.0082Æ0.001
0.0083Æ0.0011
0.0120[i]
0.001
0.016
0.002
0.006
0.006
0.010
0.027
0.047
0.013
0.002
0.030
0.030
–
–
–
–
–
–
–
–
19
65
77
52
4.0[i]
Pd-Pd
Pd-Pd
Pd-Pd
Pd-Pd
Pd-O
Pd-Pd
Pd-O
Pd-Pd
Pd-O
8.8Æ0.8
6.6Æ1.0
5.7Æ0.9
2.6Æ0.2
2.6Æ0.4
1.7Æ0.5
2.9Æ0.4
9.2Æ0.9
4.0[i]
1.2–2.0
0.7–1.1
0.6–0.8
–
1.2Æ0.4
1.0Æ0.3
1.2Æ0.4
1.7Æ0.5
1.9 (58.7)
–
–
–
0.0080[i]
Pd/ZrO2_280
0.0120[i]
0.0060[i]
0.0070Æ0.0006
0.0041Æ0.0026
0.0208Æ0.0109
–
1.4Æ0.3
1.1 (98.1)
–
35
[j]
Pd/CeO2_15
Pd/CeO2_173
1.3–2.6
–
13
131
Pd-Ce
4.0Æ3
[a] k3: k-range=2–10.5 ꢁÀ1, r-range=1.7–2.95 ꢁ for Pd-Pd and 1.2–1.85 ꢁ for Pd-O. Intrinsic loss factor, S02 =0.82 (Pd-Pd) and 0.88 (Pd-O) from Pd foil and
PdO data, respectively. [b] First shell coordination number. [c] Bond length. [d] Debye-Waller factor. [e] Goodness-of-fit index. [f] Diameter was estimated by
CO chemisorption. The values in parentheses were Pd dispersion (%). [g] Specific surface area (SSA) of catalysts after air calcination at 5508C for 4 h.
[h] The coordination number was fixed as that of an FCC lattice. [i] Fixed. [j] Not estimated owing to the low contrast between Pd and Ce.
Figure 2. Pd K-edge XANES spectra of (a) Pd/ZrO2 (inset: enlarged spectra
near the absorption edge) and (b) of Pd/CeO2.
Figure 3. Pd K-edge radial structure functions of (a) Pd/ZrO2 and (b) Pd/
CeO2.
the absorption edge of Pd/CeO2_173 was shifted to higher
energy and the spectral features became similar to those of
PdO. However, the EXAFS oscillation was inconsistent with that
of PdO (Figure S3d), excluding the formation of PdO particles.
Figure 3 shows radial structure functions (RSF) of Pd/ZrO2
and Pd/CeO2. For Pd/ZrO2, peaks were observed at 1.53 and
2.64 ꢁ (phase-uncorrected), corresponding to the PdÀO and
PdÀPd distances, respectively. The Fourier-transformed magni-
tude of the peak at 2.64 ꢁ decreased with increasing specific
surface area of ZrO2, suggesting that the Pd particles de-
creased in size accordingly. The contribution of the PdÀO bond
&
ChemSusChem 2016, 9, 1 – 8
www.chemsuschem.org
4
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
was more remarkable for Pd/ZrO2_194 and Pd/ZrO2_280. How-
ever, no peak was observed at 3.13 ꢁ, corresponding to the
PdÀPd distance of the second coordination shell observed in
PdO (Figure 3a). This spectrum suggested that PdO nanoparti-
cles were not present in Pd/ZrO2. Okumura et al have reported
XAFS measurements of monomeric Pd0 on zeolites, in which
the XAFS peak corresponding to the PdÀO bond was ascribed
to the interaction between Pd and the surface O atom in the
zeolites. They also concluded that the valence state was Pd0,
which was suggested by the similar XANES absorption edge
position to that of Pd foil.[5] Therefore, the Pd0 particles on
ZrO2_194 and ZrO2_280 are thought to be small clusters, and
the interaction between Pd and the O atom of the ZrO2 surface
was more obvious for small Pd particles.
In the RSF of Pd/CeO2_15, Pd was present as Pd0, as suggest-
ed by the peak at 2.64 ꢁ (phase-uncorrected) (Figure 3b). In
contrast, the peak at 2.64 ꢁ was not present for Pd/CeO2_173.
Instead, peaks were observed at 1.56 and 3.01 ꢁ, which corre-
spond to the PdÀO and PdÀCe distances, respectively. Highly
dispersed Pd species by the formation of a PdÀOÀCe bond has
been reported.[6] The absence of a PdÀPd bond and a discrep-
ancy in the EXAFS oscillation between Pd/CeO2_173 and PdO
suggested the formation of a single Pd atom on CeO2_173.
The higher energy shift of the absorption edge (Figure 2) was
probably owing to the strong interaction between Pd and O in
CeO2.
Figure 4. Ab initio simulations of the Pd K-edge XANES spectra: Pd1/ZrO2
(111), P42/mmc-PdO, FCC-Pd, and Pd55 cluster.
ters are shown in Table 4. The sizes of the Pd particles were es-
timated from the coordination number (CN) of the PdÀPd
bond. The CN values of the PdÀPd bond decreased with in-
creasing specific surface area of ZrO2. The diameters estimated
by XAFS and TEM were almost consistent for Pd/ZrO2_30, Pd/
ZrO2_91, and Pd/ZrO2_94, in which the CN is higher than 6. In
contrast, the CN values of Pd/ZrO2_194 and Pd/ZrO2_280 were
markedly decreased to 2.6 and 1.7, respectively. These CNs cor-
responded to Pd3–5 clusters, which was different from the di-
ameters observed by TEM. As Pd clusters smaller than 1 nm on
ZrO2 could barely be distinguished by HAADF-STEM, the mean
diameter by HAADF-STEM might be overestimated. Although
the possibility of an underestimation of Pd3–5 clusters cannot
be fully excluded, XAFS spectra and EXAFS fit parameters of
Pd/ZrO2_194 and Pd/ZrO2_280 are certainly different from Pd/
ZrO2_30, Pd/ZrO2_91, and Pd/ZrO2_94. Given that a CN of 6
corresponds to approximately 10 atoms for Pd/ZrO2_30, Pd/
ZrO2_91, and Pd/ZrO2_94, we believe that clusters, which are
at least smaller than 10 atoms, are deposited on ZrO2_194 and
ZrO2_280. Curve fitting analysis proved that the PdÀCe bond
distance was 3.19 ꢁ, which was similar to that reported previ-
ously.[6]
Ab initio simulation of the XANES spectrum was performed
to verify the model structure predicted by the RSF (Figure 3).
Several model structures have been hypothesized for the pur-
pose of verification. A Pd atomic absorption model on the tet-
ragonal ZrO2 (111) surface was prepared as a model for Pd1/
ZrO2 (111). Pd55 is a cluster comprising 55 Pd atoms. Structures
made of bulk P42/mmc PdO from the Inorganic Crystal Struc-
ture Database, ICSD-185482 and face-centered cubic (FCC) Pd
were also prepared (see Supporting Information). Structural
optimization was performed on these model structures using
the Vienna ab initio simulation package (VASP).[7] A charge den-
sity analysis[8] was performed based on the derived electronic
structure. The results revealed that the charge of the Pd atom
bonded to the oxygen atom was practically zero in Pd1/ZrO2
(111). Similarly, the valence of the Pd atom in the Pd55 cluster
was also practically zero.
We also performed CO chemisorption to estimate the size of
the Pd particles (Table 4). The Pd dispersions (D, %) of Pd/
ZrO2_30 and Pd/ZrO2_280 were estimated to be 58.7 and
98.1%, respectively. This result also suggested that the size of
the Pd particles of Pd/ZrO2_280 was smaller than that of Pd/
ZrO2_30. In addition, a dispersion of 98.1% indicates that
almost all the Pd atoms in Pd/ZrO2_280 were present as sur-
face Pd0 atoms, which is in agreement with the XAFS results.
Given that the Pd particles on Al2O3 and SiO2 are larger than
those on ZrO2 and CeO2 even though they have a high specific
surface area, the Pd particle size is affected predominantly by
the kind of metal oxides employed. However, the high specific
surface area of the metal oxides is a key factor for the deposi-
tion of small Pd clusters when the same metal oxide is used. It
is likely that Pd particles are strongly attached to ZrO2 and
CeO2, stabilizing the Pd clusters. As CeO2 stabilizes preferential-
Next, the optimized structure was used to calculate the
XANES spectrum with the finite difference method near edge
structure (FDMNES) code[9] using the real-space finite differ-
ence method. The calculations for the XANES spectra of the Pd
K-edge are shown in Figure 4. P42/mmc PdO and FCC Pd are
the XANES respective bulk spectra. This result qualitatively ex-
plains the experimental results shown in Figure 2. The shape
of the rising edge of the spectrum shown in Figure 2a is also
practically identical. The spectrum of a single atom adsorption
of Pd1/ZrO2 (111) is relatively flat. These calculations qualitative-
ly explain the results depicted in Figure 2a. The spectrum of
a large-sized Pd cluster resembles that of bulk Pd, as shown in
the Supporting Information.
Curve fitting analysis was performed using a calculation soft-
ware of XAFS and XANES, FEFF8.4[10] and the EXAFS fit parame-
ChemSusChem 2016, 9, 1 – 8
www.chemsuschem.org
5
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
ly single Pd atoms rather than Pd clusters, ZrO2 seems to be
the most suitable support for Pd0 clusters.
Conclusions
In the XRD patterns of Pd/ZrO2_30, Pd/ZrO2_91, and Pd/
ZrO2_94, the main peaks were observed at 24, 28, 31, 34, 35,
and 408, which corresponded to a monoclinic phase (ICDD
PDF No. 01-086-1449). Although ZrO2_194 and ZrO2_280
showed only halo patterns, peaks were observed after calcina-
tion in air and were assigned as tetragonal ZrO2 (ICDD PDF No.
01-088-1007), suggested by the main peaks at 30, 34 35, 50,
and 608. No peak corresponding to Pd was observed for all the
catalysts (Figure S7). The specific surface area decreased after
calcination (Table 4), suggesting that a high specific surface
area is required during the impregnation of Pd precursors but
does not affect the catalytic activity of the Pd after calcination.
The catalytic activity of Pd was strongly related to the size of
the Pd particles. Pd/CeO2_173 showed higher catalytic activity
than homogeneous Pd complexes, implying that the stabiliza-
tion of the Pd atoms by CeO2 was more effective than that by
phosphine ligands. However, single Pd atoms on CeO2_173
lowered the catalytic activity when compared to Pd3–5 clusters
on ZrO2_194 and ZrO2_280.
The decarbonylation of furfural to furan was efficiently promot-
ed by ZrO2-supported Pd clusters with low atomicity. The size
of the Pd particles decreased with increasing specific surface
area of ZrO2. Pd clusters consisting of several Pd atoms were
formed on ZrO2 with a high specific surface area (194 and
280 m2 gÀ1). The catalytic activity of Pd/ZrO2 was improved
with a decrease in the size of Pd particles, and Pd3–5 clusters
on ZrO2_280 exhibited the highest catalytic activity for the de-
carbonylation under N2 without additives. Conversely, single
Pd atoms were stabilized by CeO2 with a high surface area
owing to strong interaction with the CeO2 surface, and single
Pd atoms on CeO2 showed lower catalytic activity than Pd/
ZrO2. From these results, ZrO2 was determined to be the best
support for stabilizing Pd clusters of low atomicity. In addition,
Pd clusters consisting of several Pd atoms could be more
active Pd species than single Pd atoms in view of the preferred
m(C,O) coordination of aldehydes on Pd surfaces.
Experimental Section
A plausible reaction mechanism is shown in Scheme 1. Ini-
tially, oxidative addition occurs on the Pd0 clusters, followed by
decarbonylation. Reductive elimination gives furan and Pd0 is
regenerated. After the desorption of CO from the Pd atoms, Pd
clusters catalyze the next cycle. When CO cannot be desorbed
from the Pd atoms under a large excess of CO (Table 1,
entry 19), Pd is deactivated. When Pd catalysts are labile to be
oxidized, the catalysts are poorly active or easily deactivated.
Thus, the selection of a suitable support for the stabilization of
Pd as clusters and maintaining zero valent Pd is critically im-
portant. The Pd(111) surface has been reported to preferential-
ly adsorb furfural by m(C,O) coordination,[11] and catalysts con-
sisting of at least two Pd atoms are likely active and selective
for decarbonylation. Therefore, clusters consisting of several Pd
atoms on ZrO2 are the best catalyst for these reactions.
Materials
Activated carbon (MAXSORB, MSC-30) with a specific surface area
of 3000 m2 gÀ1 was supplied from Kansai Coke and Chemicals Co.,
Ltd. MgO and Co3O4 were purchased from Wako Pure Chemical.
Al2O3 (JRC-ARO-5) was obtained from the Catalysis Society of
Japan. SiO2 (CARiACT Q-15) was supplied from Fuji Silysia. TiO2
(anatase) with a specific surface area of 312 m2 gÀ1 (ST-111) was ob-
tained from Titan Kogyo, Ltd. ZrO2_30, ZrO2_94, ZrO2_194, and
ZrO2_280 corresponding to JRC-ZRO-3, JRC-ZRO-4, JRC-ZRO-5, and
JRC-ZRO-6 were obtained from the Catalysis Society of Japan and
ZrO2_91 (RC-100) was supplied by Daiichi Kigenso Kagaku Kogyo.
CeO2 with surface areas of 15 and 173 m2 gÀ1 (DKE-R9 and cerium
oxide HS, respectively) were both supplied by Daiichi Kigenso
Kagaku Kogyo. CeO2 (JRC-CEO-3) with a specific surface area of
81 m2 gÀ1 was obtained from the Catalysis Society of Japan.
Pd(NO3)2 was purchased from Tanaka Kikinzoku Kogyo. Furfural
was passed through a short activated alumina column (Wako,
300 mesh) before use. Other reagents were purchased and used
without further purifications.
Instruments
HAADF-STEM observations were performed using a JEOL ARM-
200F operating at 200 kV at Kyushu University or a JEOL JEM-
3200FS at Tokyo Metropolitan University. Pd K-edge XAFS measure-
ments were carried out at BL14B2 of SPring-8 (Hyogo, Japan).[12]
For XAFS, PdO/ZrO2 and PdO/CeO2 were treated in an autoclave
under pressurized H2 (2 MPa) at 1208C for 3 h to obtain Pd/ZrO2
and Pd/CeO2, respectively. The autoclave was transferred to
a glove box and the obtained catalysts were ground with boron ni-
tride in an agate mortar and were made into pellets using a tablet
molding machine for XAFS measurements. We also performed
in situ XAFS measurements using an autoclave with poly(ether
ether ketone) (PEEK) windows made by UNISOKU Co., Ltd. See
Supporting Information for experimental detail. The storage ring
energy was 8 GeV with a typical current of 99.5 mA. Pd K-edge
(24.32 keV) XAFS spectra were measured using a Si(311) double
crystal monochromator in transmission mode. Ionization chambers
Scheme 1. Plausible reaction mechanism for the decarbonylation of furfural
to furan over Pd/ZrO2_280.
&
ChemSusChem 2016, 9, 1 – 8
www.chemsuschem.org
6
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
were used to measure the intensity of the incident and transmitted
X-rays and the quick scan technique (QXAFS) was used for this
measurement. Spectral analyses were performed using the XAFS
analysis software packages Athena and Artemis.[13] The extraction
of the EXAFS oscillations from the spectra, normalization by edge-
jump, and Fourier transformation were performed by Athena. The
curve-fitting analysis was performed in R-space by Artemis. In the
curve-fitting analysis, the backscattering amplitude, the phase shift,
and the mean-free path of the photoelectron were calculated
using FEFF8.4,[10] and the number of neighboring atoms, the intera-
tomic distance between the absorbed atom to the neighboring
atom, the Debye–Waller factor, and the absorption edge energy
were treated as fitting parameters. The intrinsic loss factor was ob-
tained by curve-fitting analysis of the EXAFS data of Pd foil. The di-
ameters of the Pd particles were also estimated by adsorption of
CO pulse using a BELCAT, MicrotracBEL. The catalyst (50 mg) was
pretreated in He (50 mLminÀ1) from RT to 4008C at a rate of
188CminÀ1 and 5% O2 in He (50 mLminÀ1) at 4008C for 15 min.
Research Center, Kyushu University. We also thank Mr. E. Wata-
nabe of Tokyo Metropolitan University for helping with TEM ob-
servations.
Keywords: aldehyde · decarbonylation · furfural · palladium ·
XAFS
[1] a) G. J. S. Dawes, E. L. Scott, J. Le Nꢂtre, J. P. M. Sanders, J. H. Bitter,
Green Chem. 2015, 17, 3231–3250; b) J. W. Medlin, ACS Catal. 2011, 1,
1284–1297; c) A. Corma, S. Iborra, A. Velty, Chem. Rev. 2007, 107, 2411–
2502.
[2] a) W.-L. Fanchiang, Y.-C. Lin, Appl. Catal. A 2012, 419–420, 102–110;
b) J. P. Diebold, R. J. Evans, (US Energy, Lakewood, CO, USA) US4764627,
1988.
[3] a) W. Zhang, Y. Zhu, S. Niu, Y. Li, J. Mol. Catal. A 2011, 335, 71–81; b) H.
Zhang, X.-K. Gu, C. Canlas, A. J. Kropf, P. Aich, J. P. Greeley, J. W. Elam,
R. J. Meyers, J. A. Dumesic, P. C. Stair, C. L. Marshall, Angew. Chem. Int.
Ed. 2014, 53, 12132–12136; Angew. Chem. 2014, 126, 12328–12332;
c) S. Sitthisa, D. E. Resasco, Catal. Lett. 2011, 141, 784–791; d) S. Bhoges-
wararao, D. Srinivas, J. Catal. 2015, 327, 65–77; e) S. Sitthisa, T. Pham, T.
Prasomsri, T. Sooknoi, R. G. Mallinson, D. E. Resasco, J. Catal. 2011, 280,
17–27; f) V. V. Pushkarev, N. Musselwhite, K. An, S. Alayoglu, G. A. So-
morjai, Nano Lett. 2012, 12, 5196–5201; g) H. Singh, M. Prasad, R. D. Sri-
vastava, J. Chem. Technol. Biotechnol. 1980, 30, 293–296; h) M. Sugita,
H. Tsuji, (Mitsubishi Chemical, Yokohama, Kanagawa, Japan) JP Patent
P2009-149634A, 2009.
[4] a) P. Lejemble, A. Gaset, P. Kalck, Biomass 1984, 4, 263–274; b) K. J.
Jung, A. Gaset, Biomass 1988, 16, 63–76; c) P. Lejemble, Y. Maire, A.
Gaset, P. Kalck, N. de Dave, Chem. Lett. 1983, 1403–1406; d) J. G. Ste-
vens, R. A. Bourne, M. V. Twigg, M. Poliakoff, Angew. Chem. Int. Ed. 2010,
49, 8856–8859; Angew. Chem. 2010, 122, 9040–9043; e) A. Modak, A.
Deb, T. Patra, S. Rana, S. Maity, D. Maiti, Chem. Commun. 2012, 48,
4253–4255; f) J. Mitra, X. Zhou, T. Rauchfuss, Green Chem. 2015, 17,
307–313; g) Y.-B. Huang, Z. Yang, M.-Y. Chen, J.-J. Dai, Q.-X. Guo, Y. Fu,
ChemSusChem 2013, 6, 1348–1351; h) D. Scholz, C. Aellig, I. Hermans,
ChemSusChem 2014, 7, 268–275.
After cooling, the sample was treated in 5% H2 in Ar (50 mLminÀ1
)
at 508C for 15 min. After pretreatment, CO pulse was introduced
into the cell at 508C, and the amount of CO adsorbed was mea-
sured. The Pd/CO ratio was set as 1 for the calculation of Pd dis-
persion (D). The diameters of the Pd particles (d, nm) were estimat-
ed by the following equation: d=1.1/D.[14] XRD patterns were ob-
tained using a Rigaku Miniflex 600. Conversions and product yields
of the decarbonylation were obtained by gas chromatography
using an Agilent GC 6850 Series II equipped with a CO2 gas cylin-
der, flame ionization detector, and a capillary column (J&W HP-1).
Catalyst preparation
Supported PdO catalysts were prepared by impregnation. The Pd
loading was 1 wt% for all Pd catalysts. The support material was
added to an aqueous solution of Pd(NO3)2, and the mixture was
stirred at room temperature for 30 min. After impregnation, water
was removed by vacuum freeze-drying. The solid was calcined in
air for 4 h.
[5] K. Okumura, H. Matsui, T. Tomiyama, T. Sanada, T. Honma, S. Hirayama,
M. Niwa, ChemPhysChem 2009, 10, 3265–3272.
[6] S. Hinokuma, H. Fujii, M. Okamoto, K. Ikeue, M. Machida, Chem. Mater.
2010, 22, 6183–6190.
[7] a) G. Kresse, J. Hafner, J. Appl. Crystallogr. 2011, 44, 1272–1276; b) G.
Kresse, Thesis, Technische Universitꢃt Wien, 1993; c) G. Kresse, J. Furth-
mꢄller, Comput. Mater. Sci. 1996, 6, 15–50; d) G. Kresse, J. Furthmꢄller,
Phys. Rev. B 1996, 54, 11169–11186.
[8] a) G. Henkelman, A. Arnaldsson, H. Jꢅnsson, Comput. Mater. Sci. 2006,
36, 354–360; b) E. Sanville, S. D. Kenny, R. Smith, G. Henkelman, J.
Comput. Chem. 2007, 28, 899–908.
Decarbonylation of furfural
PdO/ZrO2, reaction solvent, and a magnetic stirring bar were
added to an autoclave. The catalyst was pretreated in H2 (2.0 MPa)
at 1208C for 2 h to obtain Pd/ZrO2. The autoclave was cooled to
room temperature and pressurized H2 was carefully released. Furfu-
ral was added to the autoclave, and the autoclave was purged and
filled with N2 (atmospheric pressure). The mixture was stirred at
1408C for 12 h. After the reaction, the catalyst was removed by fil-
tration and the filtrate was analyzed by GC using tridecane as an
internal standard.
[9] Y. Joly, Phys. Rev. B 2001, 63, 125120.
[10] A. L. Ankudinov, A. I. Nesvizhskii, J. J. Rehr, Phys. Rev. B 2003, 67, 115120.
[11] a) M. Mavrikakis, M. A. Barteau, J. Mol. Catal. A 1998, 131, 135–147;
b) P. M. Williams, S. H. Pang, J. W. Medlin, J. Phys. Chem. C 2014, 118,
27925–27932; c) B. Liu, L. Cheng, L. Curtiss, J. Greeley, Surf. Sci. 2014,
622, 51–59.
[12] a) T. Honma, H. Oji, S. Hirayama, Y. Taniguchi, H. Ofuchi, M. Takagaki, AIP
Conf. Proc. 2010, 1234, 13–16; b) H. Oji, Y. Taniguchi, S. Hirayama, H.
Ofuchi, M. Takagaki, T. Honma, J. Synchrotron Radiat. 2012, 19, 54–59.
[13] B. Ravel, M. Newville, J. Synchrotron Radiat. 2005, 12, 537–541.
[14] C. N. Costa, S. Y. Christou, G. Georgiou, A. M. Efstathiou, J. Catal. 2003,
219, 259–272.
Acknowledgements
This work was financially supported by Mitsubishi Chemical and
NEDO (P13006). The synchrotron radiation experiments were per-
formed at the BL14B2 in SPring-8 with the approval of JASRI (No.
2013B1598, 2014A1534, 2014B1832 2014B1897, and 2015A1702).
HAADF-STEM observation was performed at the Ultramicrosopy
Received: September 5, 2016
Published online on && &&, 0000
ChemSusChem 2016, 9, 1 – 8
www.chemsuschem.org
7
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

FULL PAPERS
T. Ishida,* K. Kume, K. Kinjo, T. Honma,
K. Nakada, H. Ohashi, T. Yokoyama,
A. Hamasaki, H. Murayama, Y. Izawa,
M. Utsunomiya, M. Tokunaga*
&& – &&
Efficient Decarbonylation of Furfural
to Furan Catalyzed by Zirconia-
Supported Palladium Clusters with
Low Atomicity
Furfural to furan! The high surface area
of zirconia stabilizes Pd clusters consist-
ing of several Pd atoms, as character-
ized by X-ray absorption fine structure
measurements. Small Pd clusters on
ZrO2 exhibited high catalytic activity for
liquid-phase decarbonylation of furfural
to furan under mild reaction conditions
without additives.
&
ChemSusChem 2016, 9, 1 – 8
www.chemsuschem.org
8
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!