Synergic Catalysis of PdCu Alloy Nanoparticles within a Macroreticular Basic Resin for Hydrogen Production from Formic Acid

Article abstract of DOI:10.1002/chem.201501760

Highly dispersed PdCu alloy nanoparticles have been successfully prepared within a macroreticular basic resin bearing -N(CH₃)₂ functional groups. This previously unappreciated combination of alloy is first proven to be responsible fo

Full text of DOI:10.1002/chem.201501760

DOI: 10.1002/chem.201501760
Full Paper
&
Hydrogen Production
Synergic Catalysis of PdCu Alloy Nanoparticles within
a Macroreticular Basic Resin for Hydrogen Production from
Formic Acid
[a, b]
[b]
[a]
[b, c]
Kohsuke Mori,*
Hiromi Yamashita*
Hiromasa Tanaka, Masahiro Dojo, Kazunari Yoshizawa,
and
[a, b]
Abstract: Highly dispersed PdCu alloy nanoparticles have
been successfully prepared within a macroreticular basic
resin bearing ÀN(CH ) functional groups. This previously
alloying effect but also cooperative action by the ÀN(CH )
3
2
groups within the resin play crucial roles in achieving excep-
tional catalytic performances. In addition to the advantages
such as, facile preparation method, free of additives, recycla-
ble without leaching of active component, and suppression
of unfavorable CO formation less than 3 ppm, the present
catalytic system is cost-effective because of the superior cat-
alytic activity compared with that of well-established pre-
cious PdAg or PdAu catalysts. The present catalytic system is
particularly desirable for an ideal hydrogen vector in terms
of potential industrial application for fuel cells.
3
2
unappreciated combination of alloy is first proven to be
responsible for the efficient production of high-purity H2
from formic acid (HCOOH) dehydrogenation for chemical
hydrogen storage. By the addition of Cu, the electronically
promoted Pd sites show significantly higher catalytic activity
as well as a better tolerance towards CO poisoning as com-
pared to their monometallic Pd counterparts. Experimental
and DFT calculation studies revealed not only the synergic
Introduction
shape, and composition plays a crucial role in designing highly
functionalized catalysts.
The precise control of geometric and electronic effect of
bimetallic nanoparticles (NPs), in which the architectural con-
figuration of two metals as alloys or with a core–shell struc-
ture, has received considerable attention in attaining superior
catalytic activity and selectivity to the monometallic counter-
Among various alternative energy strategies, hydrogen (H2)
has been considered as a promising energy carrier to solve the
long-term environmental and energy problems originating
[
3]
from fossil fuels. In addition, quite efficient energy conversion
can be realized by combining hydrogen with proton exchange
[
1]
[4]
parts. The coordination of one metal to the other can provide
specific new active sites and often alters the electronic proper-
ties owing to the unexpected interplay of the neighboring
different metals. In addition to such synergic effects, the
replacement of the portion of precious noble metal NPs with
membrane fuel cell (PEMFC) technology. However, technical
problems in controllable storage and safe delivery of hydrogen
severely limit its widespread application as a transportation
fuel. A feasible option is to store hydrogen in an organic liquid
medium, which can release hydrogen in situ under mild condi-
[
2]
[5]
inexpensive metals contributes to the atomic economy. Thus,
the successful synthesis of bimetallic NP with controllable size,
tions for direct use in fuel cells. Recently, formic acid
(HCOOH), which is one of the major products formed during
biomass processing, has emerged as a non-toxic and conven-
ient hydrogen carrier because it is an avirulent nonflammable
À1
[6]
liquid and contains 4.4 wt% (53.4 gL ) hydrogen. Additional-
ly, a carbon-neutral hydrogen-storage cycle can be attained
[
a] Prof. Dr. K. Mori, M. Dojo, Prof. Dr. H. Yamashita
Graduate School of Engineering, Osaka University
1
-2 Yamadaoka, Suita, Osaka 565-0871(Japan)
when formic acid can be regenerated through CO hydrogena-
2
E-mail: mori@mat.eng.osaka-u.ac.jp
yamashita@mat.eng.osaka-u.ac.jp
tion. Formic acid decomposition proceeds by two possible
routes: Selective dehydrogenation (HCOOH (l)!H (g)+CO (g)
2
2
[
b] Prof. Dr. K. Mori, Dr. H. Tanaka, Prof. Dr. K. Yoshizawa,
Prof. Dr. H. Yamashita
À1
DG =À48.4 kJmol ) is indispensable for the production of
298
Elements Strategy Initiative for Catalysts & Batteries ESICB
Kyoto University, Katsura, Kyoto 615-8520 (Japan)
ultra-pure H without undesirable dehydration (HCOOH (l)!
2
À1
H O(l)+CO(g) DG =À28.5 kJmol ), because CO impurity is
2
298
[
c] Prof. Dr. K. Yoshizawa
toxic to Pt catalysts in fuel cells. Recently, an in situ surface-
enhanced IR study disclosed that COad is produced by the
Institute for Materials Chemistry and Engineering and
International Research Center for Molecular Systems
Kyushu University, Fukuoka 819-0395 (Japan)
reduction of CO (dehydrogenation product from HCOOH) on
2
Pd surface containing adsorbed H species rather than by
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/chem.201501760.
[
7]
a direct dehydration pathway.
Chem. Eur. J. 2015, 21, 12085 – 12092
12085
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Although there have been extensive studies on formic acid
decomposition using homogeneous metal complex catalysts at
[8]
ambient temperatures, the combined use of excess amine
leads to severe difficulties in device fabrication due to separa-
tion issues. Current research has been focused on the exploita-
[9]
tion of practical heterogeneous catalysts, wherein Au, Pt, and
especially Pd exhibit excellent catalytic activity. Unfortunately,
monometallic catalysts still suffer from poor selectivity toward
dehydrogenation pathway, and are inclined to deactivation
due to adsorption of poisonous CO byproduct. Alloying with
another metal that possesses a superior CO anti-poisoning
ability is a promising strategy. Recently, a B-doped Pd catalyst
was developed to boost the formic acid decomposition at
Figure 1. TEM image and size distribution diagrams of Pd50Cu50/resin 1. The
inset shows the HR-TEM image.
[
10]
room temperature. Unfortunately, the catalysts reported so
far are mainly focused on Pd alloyed with a coinage metal like
Au or Ag. Because such catalysts are based heavily on precious
elements, it is highly desirable to develop the alternative
nanocatalysts with improved activity and CO tolerance by
integration of economically viable components.
sis, indicating the formation of PdCu NPs in an alloy structure.
Point energy-dispersive X-ray spectroscopy (EDS) measurement
also confirmed that the Pd/Cu molar ratio of PdCu alloy was
about 50:50, which is in good agreement with that determined
by ICP analysis. Because of their small size, the peaks attributa-
ble to the fcc lattice could not be confirmed by XRD analysis.
X-Ray absorption measurements were further carried out to
elucidate the structure of the Pd Cu /resin 1. In the Cu K-
In this paper, we have demonstrated for the first time that
the monodispersed PdCu NPs combined with the basic resin
exhibited a dramatic enhancement in the selective hydrogen
production from formic acid decomposition. Such improve-
ment can be attributed to not only the positive alloying effect
of Pd with Cu but also the cooperative action between active
NPs and basic ÀN(CH ) groups within the macroreticular resin,
50
50
edge normalized X-ray absorption near-edge structure (XANES)
spectra (Figure 2A), the first peak on the arising edge at
8982 eV of the Cu foil is assigned to the 1 s!4p transition. A
single absorption peak in the range 8983–8985 eV for Cu O is
3
2
2
as evidenced by both experimental and DFT calculation
studies. In addition to advantages such as simple work-up
procedures and the durability of the catalyst lifetime, the cata-
lytic system proposed here possesses other stimulating charac-
teristics: 1) No requirement for additives; 2) Suppression of
concomitant CO formation (to a level of less than 3 ppm), and
assigned to the electronic dipole-allowed transition characteris-
I
tic of Cu. The spectrum of Pd Cu /resin 1 is similar to that of
50
50
the Cu foil but differs from that of Cu O, verifying that the
2
Pd Cu /resin 1 exists in the metallic form. Small differences
50
50
with Cu foil may be attributed to alloying with Pd. In the Fouri-
3
) Superior catalytic activity as compared with the PdAg or
PdAu alloy.
Results and Discussion
Characterization and catalytic performances
The commercially available resin, IRA96SB (resin 1), with
a weakly basic ÀN(CH ) group was employed for the synthesis
3
2
of PdCu catalysts. The deposition of both Pd and Cu within
the basic resin 1 was performed by using a simple ion
exchange method from an aqueous solution of PdCl2 and
CuNO ·3H O. After the deposition, resins were crushed by
3
2
a ball mill (600 rpm for 10 min). Subsequently, the samples
were pre-reduced with NaBH . For example, the resulting
4
catalyst with the molar ratio of Pd/Cu=50:50 is denoted as
Pd Cu /resin 1.
50
50
2
À1
The BET-specific surface area of the sample was 45 m g .
The transmission electron microscopy (TEM) image of Pd Cu /
50
50
resin 1 exhibited the formation of monodispersed NPs with
a narrow size distribution, whose average particle size was de-
termined to be 1.9 nm (Figure 1). A representative high-resolu-
tion TEM image shows a d-spacing of 2.18 Š, which corre-
sponds to the (111) lattice spacing of face-centred cubic (fcc)
PdCu (2.175 Š) calculated from the Rietveld refinement analy-
Figure 2. Cu K-edge XANES (A) and FT-EXAFS (B) spectra of a) Pd50Cu50
resin 1, b) Cu foil, c) Cu O; Pd K-edge XANES (C) and FT-EXAFS (D) spectra of
a) Pd50Cu50/resin 1, b) Pd foil, c) PdCl
/
2
2
.
Chem. Eur. J. 2015, 21, 12085 – 12092
www.chemeurj.org
12086
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
er transform (FT) spectra of Cu K-edge extended X-ray absorp-
tallic Pd and Cu as well as Pd Au , Pd Ag nanocatalysts
50
50
50
50
tion fine structure (EXAFS) spectra (Figure 2B), Pd Cu /resin 1
supported on resin 1 were also prepared by same ion ex-
5
0
50
and the reference Cu foil showed an intense singlet peak at
around 2.4 Š ascribed to the contiguous metal–metal bonding,
which was slightly shifted to a longer interatomic distance as
compared with the pure Cu foil by the introduction with Pd
atoms. The inverse FT of the peak around 0.7–2.7 Š was well
fitted using a CuÀCu bond and CuÀPd bond (the Supporting
Information, Figure S1). The interatomic distances (R) and coor-
dination numbers (CN) were found to be 2.53 and 4.4 for the
CuÀCu bond and 2.59 Š and 5.9 for the CuÀPd bond, respec-
tively (the Supporting Information, Table S1). These results sug-
gest the formation of heteroatomic PdÀCu bonding. No peaks
change method followed by the reduction with NaBH . TEM
4
analysis revealed that the average diameters of NPs were
almost same with Pd Cu /resin 1 (1.9 nm) except for
50
50
Pd Au /resin 1; Pd resin 1 (1.7 nm), Cu resin 1 (2.0 nm),
50
50
Pd Au /resin 1 (3.1 nm), and Pd Ag /resin 1 (1.7 nm) (Sup-
50
50
50
50
porting Information, Figures S2–S5). For comparison, we also
synthesized the Pd Cu /resin 2 using a strongly acidic resin 2
50
50
possessing ÀSO H functioned group, in which average diame-
3
ter of PdCu NPs was determined to be 2.3 nm (Supporting In-
formation, Figure S6).
The time course for the hydrogen production from formic
acid decomposition using various catalysts were monitored
periodically, as shown in Figure 3A. No induction period was
observed, and the ratios of H to CO generated during the
due to CuÀO and CuÀOÀCu bonds, detectable in that of Cu O
2
at around 1.5 and 2.9 Š, were observed.
Figure 2C shows the XANES spectra at the Pd K-edge of
Pd Cu /resin 1 as well as standard samples including Pd foil
2
2
course of the reaction were almost 1 for all samples. The
50
50
and PdCl . The absorption edge varies with respect to the oxi-
Pd Cu /resin 1 with basic ÀN(CH ) groups performed signifi-
2
50
50
3 2
dation state of the palladium, and the XANES spectrum of the
cantly higher than the pure Pd/resin 1 catalyst, in which
almost 100% conversion was attained after 4 h. The reaction
using pure Cu/resin 1 was seriously sluggish. The enhance-
ment effect was almost negligible when the physical mixture
of pure Pd and Cu catalysts was employed. These results
indicate that a synergic effect exists between Pd and Cu
atoms. Surprisingly, Pd Cu /resin 1 exhibited considerably
PdCl has one broad peak at around 24365 eV, whereas Pd foil
2
showed two distinct peaks at around 24362 and 24385 eV,
[11]
corresponding to the allowed 1s!5p transition. The shapes
of the XANES spectra and the edge position of Pd Cu /resin
5
0
50
1
resembled those of Pd foil but differed from those of PdCl ,
2
0
suggesting that the oxidation state of Pd was almost Pd . A
more detailed inspection revealed that two characteristic
peaks in the Pd Cu /resin 1 seem a little weak compared with
50
50
high activity as compared with the corresponding Pd Au
50
50
and Pd Ag catalysts. Recently, there have been reports
50
50
50
50
that of the pure Pd foil. This indicates that the symmetry of
the Pd metal fcc structure is slightly disordered due to the in-
troduction of Cu atoms. The edge position (measured at the
half-height of the edge jump) depends on the electronic
charge of the Pd atoms, and the energy of Pd Cu /resin 1 is
slightly lower than that of the Pd foil (ꢀ1 eV), which might be
attributed to the change of electronic state of Pd deriving
from the electron-donating ability of Cu. The electronic charg-
es could transfer from Cu atoms to Pd atoms because the ioni-
zation potentials of Pd and Cu are 8.34 and 7.72 eV, respective-
ly. As will be discussed later, the negatively charged Pd atoms
play important roles in the hydrogen generation from the
formic acid. The Pd K-edge FT-
dealing with PdAu or PdAg catalysts for the decomposition
of formic acid, whereas the trial of the use of PdCu catalysts
remains unsuccessful so far. To the best of our knowledge,
this is the first demonstration for the exceptional improvement
was attained by using the PdCu alloy catalyst. Additionally,
the catalytic activity was also greatly influenced by the proper-
ties of resins: Pd Cu /resin 2 with strongly acidic ÀSO H
5
0
50
5
0
50
3
group showed low activity despite its relatively small average
PdCu NP diameter of 2.3 nm. Accordingly, we suggest that
not only the type of the second metal but also the properties
of the resin are important factors in attaining high catalytic ac-
tivity.
EXAFS spectra of Pd Cu /resin
5
0
50
1
exhibited a single peak due to
the contiguous PdÀPd bond in
the metallic form (Figure 2D).
The PdÀPd distance, however,
accompanied the shouldered
peak at around 2.2 Š. The best
fit was achieved with a PdÀPd
bond having R and CN of 2.72 Š
and 4.0 and a PdÀCu bond of
2
.59 Š and 5.9, respectively (Sup-
porting Information, Figure S1
and Table S1). This is clear evi-
dence for the formation of PdCu
alloy.
Figure 3. A) Time course of reaction in the H
parentheses are average diameters of NPs determined by TEM analysis; B) TOF versus mole fraction of Cu for
2
production from formic acid using various resin catalysts. Values in
To evaluate the catalytic ability
of the Pd Cu /resin 1, monome- PdCu/resin catalysts at different Pd and Cu compositions.
50
50
Chem. Eur. J. 2015, 21, 12085 – 12092
www.chemeurj.org
12087
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
The composition of the PdCu nanocatalysts (Pd Cu ,
76
24
Table 1. KIE in the decomposition of formic acid.
Pd Cu , Pd Cu , Pd Cu ) were easily adjusted by varying
67
33
50
50
30
70
the initial molar ratio of PdCl and CuNO ·3H O, and further an-
alyzed by ICP analysis. Figure 3B shows the plot of TOF versus
the mole fraction of Cu at different compositions, in which
Entry
Catalyst
Formic
acid
Reaction rate
H D
k /k
2
3
2
À1
[mmolmin
]
HCOOH
HCOOD
DCOOH
HCOOH
HCOOD
DCOOH
HCOOH
HCOOD
DCOOH
HCOOH
HCOOD
DCOOH
3.32
3.05
2.10
0.63
0.37
0.39
2.83
2.41
1.43
2.36
2.28
1.38
Pd50Cu50/resin 1
(basic resin)
1
2
3
4
1.06
1.58
–
1.70
1.62
–
1.17
1.98
–
1.04
1.71
maximum activity was obtained in Pd Cu /resin 1. Such volca-
5
0
50
no-type activity order based on the Pd/Cu composition clearly
suggests that the formation of uniform PdCu alloy structure
within the macroreticular domains of resin and the unusual
synergic effect originated from the integration of Pd with Cu.
Remarkably, the formic acid decomposition using Pd Cu /
Pd50Cu50/resin 2
(acidic resin)
Pd/resin 1
(basic resin)
50
50
À1
resin 1 gives a high turnover frequency (TOF)=810 h based
on Pd at 348 K using HCOOH/HCOONa=9:1 aqueous solution.
Notably, this achieved TOF value is higher than those reported
Pd50Ag50/resin 1
(basic resin)
À1
for other active catalyst systems, such as PdÀAu/C (27 h ,
[
9a]
À1
[9c]
À1
3
4
65 K), Pd/C (255 h , 373 K), Mo C/C PdÀAu/C (437 h ,
2
[
9d]
À1
[9f]
23 K),
and Ag@Pd/C (626 h , 363 K).
Moreover, the
(1.06) was smaller than that obtained using acidic Pd Cu /
50 50
present catalytic system efficiently suppressed unfavorable CO
impurity with less than 3 ppm of CO during the course of the
reaction. This concentration was significantly lower than that
obtained by conventional gas reforming from methanol, etha-
nol, and methane, and meets the criteria of the PEMFC stan-
resin 2 (1.70) (Table 1, entry 1 vs. 2). This is clear evidence that
the basic C->N(CH ) groups have a positive effect on OÀH
3
2
bond cleavage of formic acid. It is reported that high catalytic
activity can be obtained for formic acid decomposition in the
presence of amine in both homogeneous and heterogeneous
[
12]
[8d,9i]
dard, which requires a CO concentration lower than 10 ppm.
systems.
In the present catalytic system, it is reasonable to
Furthermore, the lifetime and leaching of active metal species
into solution are important points to consider when heteroge-
neous catalysts are used. The Pd Cu /resin 1 almost keeps its
original catalytic activity even after the recycling experiments.
ICP analysis of the filtrate confirmed that the metal content
was below the detection limit, suggesting no leaching of Pd or
Cu species. It should be noted that ion-exchange resins, which
possess reasonable thermal and mechanical stability, stabilize
highly dispersed metal NPs inside macroreticular domains.
conclude that the existence of suitable amine functionality in
the vicinity of active PdCu centers exerts a cooperative action
without the additional dosage of amine reagents.
Consequently, catalytically active species generated within
a functional resin matrix essentially differ from conventional
5
0
50
[
14]
solid supported metal catalysts.
Recently, PdAg and PdAu bimetallic catalysts including alloy,
core–shell structure, and monodispersed NPs were extensively
studied for the decomposition of formic acid. In general, such
bimetallic catalysts were reported to possess improved
HCOOH dehydrogenation activity and stability relative to mon-
ometallic Pd catalysts. The discussion has been made to eluci-
date the role of second metals. It has been proposed that the
activity of formic acid decomposition is related to the intrinsic
electronic properties of the metallic particles; the closer the
d-band center is to the Fermi level of a metal, the higher the
Mechanistic investigations
As mentioned before, the H2 production from formic acid
decomposition using basic resin with the ÀN(CH ) group was
3
2
significantly higher than that using acid resin with the ÀSO H
3
group. The addition of HCOONa to the reaction mixture slight-
[
15]
ly increased the H productivity, indicating the participation of
adsorption energy of the formic acid. Thus, Pd atoms exhibit
a higher activity than Ag or Au atoms. Meanwhile, the integra-
tion of Pd with Ag induces the charge transfer from Ag atoms
to Pd atoms owing to the net difference in ionization potential
(Ag: 7.57 eV and Pd: 8.34 eV), which optimizes the electronic
structure of Pd atoms for formic acid dehydrogenation. In the
case of the PdAu alloy catalysts, the improvement in catalytic
2
[6d]
Pd-formate species as a reaction intermediate. These obser-
vations make us deduce a possible catalytic cycle for formic
acid decomposition, in which the ÀN(CH ) group cooperative-
3
2
ly participate in reaction pathway (Supporting Information,
[
13]
Scheme S1).
First, OÀH bond dissociation of formic acid
(
HCOOH) is facilitated by the assist of the basic ÀN(CH ) group
3
2
as
a
proton scavenger, providing
a
metal-formate (MÀ activity have been explained by the higher resistance to CO
À
+
[
HCOO] ) species along with a À HN(CH ) group. The formed
poisoning, since Au dose not form stable complex with CO
molecules. Despite extensive studies about PdAg and PdAu bi-
metallic catalysts, there is currently no report about the investi-
gation of PdCu alloy catalysts in the dehydrogenation of
formic acid, which is because the synthesis of a highly effective
and monodispersed PdCu catalyst has not been attained so
far.
3
2
metal-formate species undergoes b-hydride elimination to
À
afford CO and a metal-hydride, MÀ[H] , species. Finally, the re-
2
+
action of the hydride species with À HN(CH ) produces mo-
3
2
lecular hydrogen, accompanied with regeneration of the metal
species. Pd atoms in the alloy act as the main active species in
the present catalytic system, since the reaction using only Cu
catalysts was extremely sluggish. In the kinetic isotope effect
The synthesis of smaller size of NPs is the key to obtaining
high catalytic activity in the heterogeneous catalysts. However,
in the present study, the size of NPs does not account for the
(
KIE) using HCOOH and HCOOD, the k /k values obtained
H D
from competitive reactions using the basic Pd Cu /resin 1
5
0
50
Chem. Eur. J. 2015, 21, 12085 – 12092
www.chemeurj.org
12088
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
[
18]
high catalytic activity of the Pd Cu /resin 1, because the aver-
sorption as compared with the pure Pd metals. The time
5
0
50
age diameters estimated by TEM revealed no significant differ-
profile for the H2 production using the Pd Cu /resin 1
50
50
ences in their NPs size for Pd resin 1 (1.7 nm), Pd Cu /resin
showed the linear increase without significant deactivation
(Figure 3A). Initially, a relatively high reaction rate is observed
in the case of even pure Pd/resin 1. There was no significant
differences in the reaction rate and activation energy at the ini-
5
0
50
1
1
(1.9 nm), Pd Au /resin 1 (3.1 nm), and Pd Ag /resin
50 50 50 50
(1.7 nm). A possible reason for the high catalytic activity of
the Pd Cu /resin 1 can be explained by the efficient energy
50
50
À1
À1
transfer from Cu atoms to Pd atoms owing to the net differ-
ence in ionization potential (Cu: 7.72 eV and Pd: 8.34 eV). In
the XPS spectra of alloy samples, the peaks due to the Pd 3d
are slightly shifted to lower binding energies compared with
those of Pd/resin 1, and a more significant shift was observed
in the case of Pd Cu /resin 1 than that observed for Pd Ag /
tial stage; Pd Cu /resin 1: 2.2 mLmin , E =26.4 kJmol and
50
50
a
À1
À1
Pd/resin 1: 1.7 mLmin , E =28.0 kJmol , respectively (the
a
Supporting Information, Figures S7 and S8). However, a drastic
decrease in catalytic activity was observed after 30 min. This
drastic decrease in the pure Pd catalyst is presumably ascribed
to the poisoning by CO, which may prevent the adsorption of
formic acid. On the other hand, the addition of Cu enables effi-
cient inhibition of CO adsorption as compared with the pure
Pd, which prolongs the lifetime of the catalyst.
50
50
50
50
resin 1 and Pd Au /resin 1 (Figure 4). This phenomenon is
50
50
quite different from the results observed in the PdCu/carbon
[16]
catalyst with a relatively larger particle size. It can be expect-
ed that extremely small PdCu NPs with unique electronic struc-
DFT Calculations
To understand the role of a basic resin and Cu in the present
catalytic system, we have performed density functional theory
(
DFT) calculations. Metal NPs are modeled by four atomic (111)
surface layers, where each layer consists of a 4”4 surface unit
cell of Pd, Cu, and a Pd Cu alloy (see details in the Support-
50
50
ing Information, Figure S9). A macroreticular basic resin with
the N(CH ) group (resin 1) is represented by a molecule of
3
2
N(CH ) . Table 2 summarizes the calculated adsorption energies
3
3
(
E ) of HCOOH and an acid–base pair HCOOHÀN(CH ) on
ad
3 2
metal surfaces in water as
a
solvent.
A
number of
computational studies have been reported for the adsorption
[
19]
behavior of HCOOH on Pd(111). A free HCOOH molecule has
two isomeric structures, the trans and cis forms, and the
former is energetically more favorable in the gas phase. In the
lowest-energy adsorption structure, HCOOH adopts the trans
form and interacts with two surface metal atoms through the
carbonyl oxygen atom and the acidic hydrogen atom as
shown in Figure 5. The trans OÀH(O)-bridged configuration Ia
À1
is 15.5 (Pd) and 12.9 (Cu) kJmol more stable than the cis OÀ
Figure 4. XPS spectra of Pd and Pd alloys on 1.
O(H)-bridged one Ib, in which the carbonyl and hydroxyl
À1
ture is specifically formed within
the macroreticular resin matrix,
which may influence on the
rate-determining CÀH bond dis-
sociation step from the metal-
Table 2. Relative energies (DE) and adsorption energies (E ) [kJmol ] of the adsorption structures optimized
ad
for (a) HCOOH and (b) the acid–base pair HCOOHÀN(CH
3
)
3
on the Pd(111), Cu(111), and Pd50Cu50 (111) surfaces.
HCOOH
[a]
Pd(111)
Cu(111)
Pd50Cu50 (111)
HCOOH
trans
cis
trans
cis
trans
cis
[17]
formate
intermediate.
As
[b]
bridging
DE
OÀH(O)
0.0
24.3
OÀO
15.5
12.8
OÀH(O)
0.0
13.4
OÀO
12.9
4.6
OÀH(O)
0.0–5.3
21.2–26.5
OÀO
shown in Table 1, k /k values
13.5–20.2
10.4–17.1
H
D
E
ad
obtained from competitive reac-
tions of HCOOH and DCOOH
using the Pd Cu /resin 1 (1.58)
HCOOHÀN(CH
3
)
3
[a]
50
50
Pd(111)
Cu(111)
50 50
Pd Cu (111)
was smaller than that obtained
HCOOH
trans
cis
cis
trans
cis
cis
trans
cis
cis
[
b]
using Pd Ag /resin 1 (1.71) or
Pd/resin 1 (1.98).
Another explanation for the
enhancement of catalytic activity
by the addition of Cu may be
the efficient inhibition of CO ad-
bridging
DE
OÀH(C) OÀO OÀH(C)
OÀH(C) OÀO OÀH(C)
OÀH(C)
OÀO
OÀH(C)
50
50
3.1
0.0
6.0
0.0
6.3
8.5
3.3–8.7
0.0–3.7
11.4–13.9
E
ad
56.8
65.2
59.2
43.5
42.4 40.2
51.3–56.7 61.4–65.2 51.2–53.8
[a] For the PdCu alloy surface, four bridged configurations (PdÀPd, PdÀCu, CuÀPd, and CuÀCu) are possible for
each isomer of HCOOH. The upper and lower limits are presented in the Table. [b] Notation of bridged configu-
rations is described in Figures 5 and 6.
Chem. Eur. J. 2015, 21, 12085 – 12092
www.chemeurj.org
12089
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
to two Pd atoms. In both configurations, the base molecule
abstracts the acidic hydrogen atom of HCOOH. The OÀH dis-
tance is increased from 0.985 Š in free trans-HCOOH to 1.698 Š
in IIa and 1.740 Š in IIb. The NÀH distance also indicates the
formation of an NÀH bond (1.068 Š in IIa and 1.056 Š in IIb).
Based on the adsorption structures, we expect that a formate
À
species HCOO is spontaneously generated on the surface of
a catalyst nanoparticle, once an incoming HCOOH molecule is
coupled with ÀN(CH ) group in resin 1. The formation of the
3
2
formate moiety affects the energy differences in the three con-
figurations derived from the two isomers of HCOOH. As shown
in Table 2, the total energy of the trans configuration IIa is very
close to those of the cis configurations IIb and IIc, À3.1 and
Figure 5. Two bridged configurations Ia and Ib of HCOOH adsorbed on the
Pd(111) surface. Optimized interatomic distances in [Š].
À1
+2.9 kJmol relative to IIa, respectively. In the absence of the
base, the adsorption structure of formate with the OÀH(C)-
À1
bridged configuration is calculated to be 61.8 kJmol less
oxygen atoms interact with two surface metal atoms (Table 2).
We considered four bridged configurations for each isomer of
HCOOH adsorbed on the Pd Cu (111) surface because of the
stable than that with the OÀO(H) bridged one. The presence
of an NÀH···O hydrogen bonding in the acid–base pair may
play a role in the stabilization of the adsorbed formate with
the OÀH(C)-bridged configuration, leading to the facile forma-
5
0
50
presence of Pd and Cu atoms on the surface. The lowest-
energy adsorption structure of HCOOH adopts the trans PdOÀ tion of CO through a CÀH bond cleavage. Similar trends are
2
PdH(O)-bridged configuration, in which the carbonyl oxygen
atom and acidic hydrogen atom are coordinated to two Pd
atoms. The energy difference in the trans configurations is very
observed in the optimized structures and energy differences
calculated for the acid–base pair adsorbed on the Cu(111) and
Pd Cu surfaces. Their adsorption structures are described in
50
50
À1
small (less than 5.3 kJmol ), whereas the cis configurations
Figure S10 (the Supporting Information).
À1
are 20.1 kJmol at a maximum less stable than the trans
Finally, the adsorption of CO to the Pd(111), Cu(111), and
[
20]
PdOÀPdH(O)-bridged one. The adsorption energy of HCOOH in
Pd Cu surfaces was investigated. It is experimentally and
50 50
À1
[19c–e,21]
vacuo is 45.6 kJmol for Pd(111), which is comparable with
theoretically
demonstrated that CO on Pd(111) possesses
À1 [19]
the literature values (34.7–59.3 kJmol ). On the other hand,
a hollow site with an orientation perpendicular to the surface.
In the lowest-energy adsorption structures optimized here,
a CO molecule occupies an fcc hollow site of the three surface
involvement of solvation effects of water is found to substan-
tially weaken the binding of HCOOH to the metal surfaces: The
À1
largest value of Ead are 24.3, 13.4, and 26.5 kJmol for the Pd,
metals. As shown in Table 3, the adsorption energies of CO at
À1
Cu, and Pd Cu surfaces, respectively. These results suggest
an fcc hollow site are calculated to be 183.8 kJmol
for
50
50
À1
the importance of Pd atoms at the surface of the catalyst
nanoparticle.
Pd(111) and 82.5 kJmol for Cu(111), both of which are close
À1
to the theoretical values reported by Chen et al. (179 kJmol
À1
[21]
Next, we examined the adsorption of the acid–base pair
for Pd(111) and 83 kJmol
for Cu(111)).
For the CO
(
HCOOHÀN(CH ) ) to the above three metal surfaces. The for-
adsorption on the Pd Cu surface modeled here, there are
3
3
50
50
mation of the acid–base pair should be reasonable
because an incoming HCOOH molecule must en-
counter a basic group of the resin surrounding a cata-
lyst NP before the adsorption on the surface of the
catalyst. Table 2 summarizes the adsorption energies
calculated for the Pd(111), Cu(111), and the Pd Cu
5
0
50
alloy (111) surfaces. In the presence of water as a sol-
vent, the binding of HCOOH to the metal surfaces
are significantly enhanced: The largest values of Ead
À1
are 65.2, 43.5, and 65.2 kJmol
for Pd, Cu, and
Pd Cu , respectively. Optimized structures of the
50
50
acid–base pair on the metal surfaces would provide
a clue for the increased adsorption energies. Figure 6
describes three configurations of the adsorption
structures obtained for the acid–base pair on the Pd
surface. In the trans OÀH(C)-bridged configuration
IIa, the carbonyl oxygen atom and hydrogen atom at
the carbon atom interact with two Pd atoms. In the
cis OÀO(H)-bridged configuration IIb, on the other
Figure 6. Three bridged configurations IIa–IIc of the acid–base pair HCOOHÀN(CH
3 3
)
hand, the two oxygen atoms of HCOOH are bound adsorbed on the Pd(111) surface. Optimized interatomic distances in [Š].
Chem. Eur. J. 2015, 21, 12085 – 12092
www.chemeurj.org ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
12090

Full Paper
ratios was stirred at room temperature for 24 h. The mixture was
filtered and the solid obtained was washed repeatedly with dis-
tilled water and air-dried overnight. After the deposition, resins
were crushed by a ball mill (600 rpm for 10 min). Subsequently, the
À1
Table 3. Adsorption energies (Ead) [kJmol ] of CO adsorbed at an fcc
hollow site of the Pd(111), Cu(111), and Pd50Cu50 (111) surfaces.
Pd(111)
Pd
Cu(111)
Cu
Pd50Cu50 (111)
3
3
Pd
2
Cu
PdCu
2
samples were pre-reduced with NaBH , giving PdCu nanocatalysts
4
with different molar ratios.
E
ad
183.7
82.5
148.8
126.2
Characterization
two possible hollow sites consisting of Pd Cu and PdCu . The
Brunauer–Emmett–Teller (BET) surface area measurements were
performed using a BEL-SORP max (Bel Japan, Inc.) at 77 K. The
sample was degassed in vacuum at 373 K for 2 h prior to data col-
lection. Inductively coupled plasma optical emission spectrometry
2
2
calculated values of Ead exhibit that CO prefers the Pd Cu
2
À1
À1
hollow site (148.8 kJmol ) to the PdCu one (126.2 kJmol ).
2
Similar to the adsorption of HCOOH, surface Pd atoms enhance
(
ICP-OES) measurements were performed using a Nippon Jarrell-
the binding of CO to the metal surfaces; the hollow site with
Ash ICAP-575 Mark II. TEM micrographs were obtained with a Hita-
chi Hf-2000 FE-TEM equipped with a Kevex energy-dispersive X-ray
detector operated at 200 kV. XPS spectra were measured by Shi-
madzu AXIS-ULTRA DLD. Pd K-edge and Cu K-edge XAFS spectra
were recorded using a fluorescence-yield collection technique at
the beam line 01B1 station with an attached Si (111) monochroma-
tor at SPring-8, JASRI, Harima, Japan (prop. No. 2013B1041,
À1
the largest value of Ead is Pd (183.8 kJmol ), followed in order
3
by Pd Cu (148.8), PdCu (126.2), and Cu (82.5). These results
2
2
3
support the experimental proposal that the PdCu alloy catalyst
efficiently inhibits the adsorption of CO as compared with the
pure Pd catalyst.
2
014A1045). The data reduction was performed by the REX2000
Conclusion
program (Rigaku). Backscattering amplitude and phase shift param-
eters for a curve fitting analysis were theoretically calculated with
FEFF8.40 code.
Both the deposition of Pd and Cu into the basic resin bearing
ÀN(CH ) functional groups generates PdCu nano-alloy catalyst
3
2
within its macroreticular structure, which contributed to a dra-
Decomposition of formic acid
matic improvements of catalytic activity in the H production
2
In the case of closed system, the resin catalyst (0.05 g) was placed
into a reaction vessel (30 mL) containing H O (9.6 mL), which was
then sealed with a rubber septum. The resulting mixture was bub-
bled with Ar for 15 min followed by the addition of HCOOH
from formic acid as compared with the pure Pd catalyst. It is
reasonable to determine that cooperative action by the amine
functionality as well as synergic catalysis of bimetallic PdCu ac-
counts for its high catalytic activity. The basicity of the resin
plays a positive role on the OÀH bond cleavage of formic acid
and the addition of Cu influence on the rate-determining CÀH
bond dissociation step from metal-formate intermediate owing
to the formation of electron-rich Pd species. Time profile meas-
2
(
3
0.39 mL), and subsequently reacted with magnetic stirring at
37 K. The progress of the reaction was monitored by a Shimadzu
GC-14B equipped with MS5Š columns for H2 and porapakQ for
CO . In the case of open system reaction, the catalyst was placed
2
into a reaction vessel (30 mL) with a reflux condenser and
equipped with gas buret. After the purging with Ar, HCOOH/
HCOONa=9:1 aqueous solution (1m, 5 mL) was added to the
reaction vessel at 348 K with magnetic stirring.
urements for the H production and theoretical calculations on
2
the adsorption of CO suggest that the efficient inhibition of
toxic CO adsorption as compared with the pure Pd metals also
accounts for the high efficiency and long lifetime of PdCu alloy
catalyst. The advantages of the present catalytic system, such
as facile preparation method, free of additives, high recyclabili-
ty, and superior catalytic activity as compared with PdAu, PdAg
and pure Pd catalysts, as well as suppression of unfavorable
CO formation to a level of less than 3 ppm, are particularly
desirable for an ideal hydrogen vector in terms of potential
industrial application for PEMFCs.
Computational methods
3
All DFT calculations were performed with the DMol program in
[22]
Materials Studio 6.1.
The generalized gradient approximation
(
GGA) exchange-correlation functional proposed by Perdew, Burke,
[23]
and Ernzerhof (PBE) was combined with the double numerical
basis set plus polarization functions (DNP). The inner electrons of
Pd and Cu atoms were replaced by DFT semi-core pseudopoten-
tials. Brillouin zone integrations were performed on a Monkhorst–
[24]
Pack k-point grid with a k-point spacing of 0.04 Š. The Pd and
PdCu alloy nanoparticles experimentally prepared were modeled
by a supercell slab consisting of a 4”4 surface unit cell with four
atomic (111) surface layers, where the bottom two layers were
fixed at the corresponding bulk position and the top two layers
were allowed to relax during geometry optimizations. The slab was
separated by a vacuum space with a height of 20 Š. A slab model
of the Cu(111) surface was also prepared to understand the role of
Cu. Because no information was available for the exact structure of
Pd Cu nanoparticles, we assumed a complete mixing of Pd and
Experimental Section
Materials
PdCl2 and CuCl ·2H O was purchased from Nacarai tesque. All
2
2
resins were Amberlite, supplied by Rohm and Hass. Solvents and
all commercially available organic compounds for catalytic
reactions were used as received.
50
50
Cu atoms in the alloy nanoparticle. As a result, in the modeled
Pd Cu (111) surface layers, Pd and Cu atoms are alternately ar-
Synthesis of PdCu/resin catalyst
50
50
A suspension containing the resin (5.0 g) and 200 mL aqueous
ranged along the a and b axes (see the Supporting Information,
Figure S9). Despite the application of such a simple model, the cal-
À4
PdCl solution (11.3”10 m) and CuCl ·2H O with different molar
2
2
2
Chem. Eur. J. 2015, 21, 12085 – 12092
www.chemeurj.org
12091
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
culated lattice spacing (2.191 Š in average) is surprisingly close to
the experimental value (2.18 Š) taken from the TEM image in
Figure 1. For comparison, the values of the lattice spacing of the
Pd(111) and Cu(111) surface layers are calculated to be 2.263 and
[8] a) B. Loges, A. Boddien, H. Junge, M. Beller, Angew. Chem. Int. Ed. 2008,
4
7, 3962–3965; Angew. Chem. 2008, 120, 4026–4029; b) Y. Himeda,
Green Chem. 2009, 11, 2018–2022; c) A. Boddien, B. Loges, H. Junge, F.
Gärtner, J. R. Noyes, M. Beller, Adv. Synth. Catal. 2009, 351, 2517–2520;
d) B. Loges, A. Boddien, F. Gärtner, H. Junge, M. Beller, Top. Catal. 2010,
2
.056 Š in average, respectively. Geometry optimizations were car-
5
3, 902–914.
9] a) X. Zhou, Y. Huang, W. Xing, C. Liu, J. Liao, T. Lu, Chem. Commun.
008, 44, 3540–3542; b) M. Ojeda, E. Iglesia, Angew. Chem. Int. Ed.
ried out by considering solvent effects of water (e=78.54) with
[
[25]
[26]
COSMO. The dipole slab correction was then applied for the
2
optimized structures. The adsorption energy Ead is defined as E =
2009, 48, 4800–4803; Angew. Chem. 2009, 121, 4894–4897; c) D. A. Bu-
lushev, S. Beloshapkin, J. R. H. Ross, Catal. Today 2010, 154, 7–12; d) .
Koós, F. Solymosi, Catal. Lett. 2010, 138, 23–27; e) Y. Huang, X. Zhou, M.
Yin, C. Liu, W. Xing, Chem. Mater. 2010, 22, 5122–5128; f) K. Tedsree, T.
Li, S. Jones, C. W. A. Chan, K. M. K. Yu, P. A. J. Bagot, E. A. Marquis,
G. D. W. Smith, S. C. E. Tsang, Nat. Nanotechnol. 2011, 6, 302–307; g) F.
Solymosi, . Koós, N. Liliom, I. Ugrai, J. Catal. 2011, 279, 213–219; h) Y.
Zhao, L. Deng, S.-Y. Tang, D.-M. Lai, B. Liao, Y. Fu, Q.-X. Guo, Energy Fuels
ad
Eadsorbate +EslabÀEadsorbate/slab, in which Eadsorbate is the total energy of
free adsorbate, Eslab is the total energy of the bare slab, and
Eadsorbate/slab is the total energy of the adsorbate–slab system.
Acknowledgements
2011, 25, 3693–3697; i) Q.-Y. Bi, X.-L. Du, Y.-M. Liu, Y. Cao, H.-Y. He, K.-N.
The present work was partially supported by MEXT program
Fan, J. Am. Chem. Soc. 2012, 134, 8926–8933; j) N. Yi, H. Saltsburg, M.
Flytzani-Stephanopoulos, ChemSusChem 2013, 6, 816–819.
10] K. Jiang, K. Xu, S. Zou, W.-B. Cai, J. Am. Chem. Soc. 2014, 136, 4861–
“
Elements Strategy Initiative to Form Core Research Center”,
[
MEXT; Ministry of Education Culture, Sports, Science, and Tech-
nology, Japan. We acknowledge Dr. Eiji Taguchi and Prof. H.
Yasuda at the Research Center for Ultra-High Voltage Electron
Microscopy, Osaka University, for their assistance with the TEM
measurements. XAFS spectra were recorded at the beam line
4
864.
[11] a) K. Mori, T. Hara, T. Mizugaki, K. Ebitani, K. Kaneda, J. Am. Chem. Soc.
2004, 126, 10657–10666; b) K. Mori, T. Hara, T. Mizugaki, K. Ebitani, K.
Kaneda, J. Am. Chem. Soc. 2003, 125, 11460–11461.
[
12] a) G. A. Deluga, J. R. Salge, L. D. Schmidt, X. E. Verykios, Science 2004,
03, 993–997; b) W. C. Shin, R. S. Besser, J. Power Sources 2007, 164,
328–335.
[13] M. Yadav, T. Akita, N. Tsumori, Q. Xu, J. Mater. Chem. 2012, 22, 12582–
2586.
14] K. Mori, A. Hanafusa, M. Che, H. Yamashita, J. Phys. Chem. Lett. 2010, 1,
675–1678.
3
0
2
1B1 station in SPring-8, JASRI, Harima, Japan (prop. No.
013B1041, 2014A1045). K.M. and H.Y. thank Grants-in-Aid for
1
Scientific Research (Nos. 26220911, 25289289, and 26630409,
6620194) from the Japan Society for the Promotion of Science
JSPS) and MEXT. K.Y. and H.T. thank Grants-in-Aid for Scientific
[
2
1
(
[15] A. Ruban, B. Hammer, P. Stoltze, H. L. Skriver, J. K. Nørskov, J. Mol. Catal.
A 1997, 115, 421–429.
16] a) S. Hu, S. Ha, L. Scudiero, Electrochim. Acta 2013, 105, 362–370; b) L.
Lu, L. Shen, Y. Shi, T. Chen, G. Jiang, C. Ge, Y. Tang, Y. Chen, T. Lu, Electro-
chim. Acta 2012, 85, 187–194; c) L. Wang, J.-J. Zhai, K. Jiang, J.-Q.
Wang, W.-B. Cai, Int. J. Hydrogen Energy 2015, 40, 1726–1734.
Research (Nos. 22245028, 24109014, 26888008, 22225004)
from JSPS and MEXT, as well as the MEXT Projects of “Integrat-
ed Research on Chemical Synthesis” and “Elements Strategy
Initiative to Form Core Research Center”.
[
[
[
17] Y.-l. Qin, J. Wang, F.-z. Meng, L.-m. Wang, X.-b. Zhang, Chem. Commun.
2013, 49, 10028–10030.
Keywords: copper · formic acid · hydrogen · nanoparticles ·
palladium
18] a) K. Judai, S. Abbet, A. S. Wçrz, U. Heiz, L. Giordano, G. Pacchioni, J.
Phys. Chem. B 2003, 107, 9377–9387; b) F. Abild-Pedersen, M. P. Ander-
sson, Surf. Sci. 2007, 601, 1747–1753.
[
19] a) S. Zhou, C. Qian, X. Chen, Catal. Lett. 2011, 141, 726–734; b) Q. Luo,
G. Feng, M. Beller, H. Jiao, J. Phys. Chem. C 2012, 116, 4149–4156; c) R.
Zhang, H. Liu, B. Wang, L. Ling, J. Phys. Chem. C 2012, 116, 22266–
[
[
1] K. Mori, H. Yamashita, Phys. Chem. Chem. Phys. 2010, 12, 14420–14432.
2] a) M. B. Boucher, B. Zugic, G. Cladaras, J. Kammert, M. D. Marcinkowski,
T. J. Lawton, E. C. H. Sykes, M. Flytzani-Stephanopoulos, Phys. Chem.
Chem. Phys. 2013, 15, 12187–12196; b) X. Jiang, N. Koizumi, X. Guo, C.
Song, Appl. Catal. B 2015, 170–171, 173–185.
3] S. Park, J. M. Vohs, R. J. Gorte, Nature 2000, 404, 265–267.
4] a) P. Costamagna, S. Srinivasan, J. Power Sources 2001, 102, 242–252;
b) B. C. H. Steele, A. Heinzel, Nature 2001, 414, 345–352.
22280; d) C. Hu, S.-W. Ting, K.-Y. Chan, W. Huang, Int. J. Hydrogen Energy
2012, 37, 15956–15965; e) Y. Wang, Y. Qi, D. Zhang, C. Liu, J. Phys.
Chem. C 2014, 118, 2067–2076.
[
[
[
[
[
[
20] P. Sautet, M. K. Rose, J. C. Dunphy, S. Behler, M. Salmeron, Surf. Sci.
2
000, 453, 25–31.
21] Z.-X. Chen, K. M. Neyman, K. H. Lim, N. Rçsch, Langmuir 2004, 20,
068–8077.
22] a) B. Delley, J. Chem. Phys. 1990, 92, 508–517; b) B. Delley, J. Chem.
Phys. 2000, 113, 7756–7764.
[
5] a) K. Fuku, R. Hayashi, S. Takakura, T. Kamegawa, K. Mori, H. Yamashita,
Angew. Chem. Int. Ed. 2013, 52, 7446–7450; Angew. Chem. 2013, 125,
8
7
594–7598; b) H. Cheng, T. Kamegawa, K. Mori, H. Yamashita, Angew.
Chem. Int. Ed. 2014, 53, 2910–2914; Angew. Chem. 2014, 126, 2954–
958.
23] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865–
2
3868.
[
6] a) F. Joó, ChemSusChem 2008, 1, 805–808; b) M. Grasemann, G. Lau-
renczy, Energy Environ. Sci. 2012, 5, 8171–8181; c) S. Enthaler, J. von
Langermann, T. Schmidt, Energy Environ. Sci. 2010, 3, 1207–1217; d) K.
Mori, M. Dojo, H. Yamashita, ACS Catal. 2013, 3, 1114–1119; e) M.
Martis, K. Mori, K. Fujiwara, W.-S. Ahn, H. Yamashita, J. Phys. Chem. C
[
[
24] H. J. Monkhorst, J. D. Pack, Phys. Rev. B 1976, 13, 5188–5192.
25] a) A. Klamt, G. Schuurmann, J. Chem. Soc. Perkin Trans. 1 1993, 799–
805; b) B. Delley, Mol. Simul. 2006, 32, 117–123.
[26] a) J. Neugebauer, M. Scheffler, Phys. Rev. B 1992, 46, 16067–16080; b) L.
Bengtsson, Phys. Rev. B 1999, 59, 12301–12304.
2
013, 117, 22805–22810.
7] a) K. Jiang, H.-X. Zhang, S. Zou, W.-B. Cai, Phys. Chem. Chem. Phys. 2014,
6, 20360–20376; b) J.-Y. Wang, H.-X. Zhang, K. Jiang, W.-B. Cai, J. Am.
Chem. Soc. 2011, 133, 14876–14879.
[
1
Received: May 5, 2015
Published online on July 14, 2015
Chem. Eur. J. 2015, 21, 12085 – 12092
www.chemeurj.org
12092
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim