A Comprehensive Study of Formic Acid Oxidation on Palladium Nanocrystals with Different Types of Facets and Twin Defects

Article abstract of DOI:10.1002/cctc.201500094

Palladium has been recognized as the best anodic, monometallic electrocatalyst for the formic acid oxidation (FAO) reaction in a direct formic acid fuel cell. Here we report a systematic study of FAO on a variety of Pd nanocrystals, including cubes, right

Full text of DOI:10.1002/cctc.201500094

DOI: 10.1002/cctc.201500094
Full Papers
A Comprehensive Study of Formic Acid Oxidation on
Palladium Nanocrystals with Different Types of Facets and
Twin Defects
[
a, b]
[c]
[c]
[a]
[a]
Sang-Il Choi,
Jeffrey A. Herron, Jessica Scaranto, Hongwen Huang, Yi Wang,
[
a]
[a]
[a]
[a]
[c]
Xiaohu Xia, Tian Lv, Jinho Park, Hsin-Chieh Peng, Manos Mavrikakis, and
[a]
Younan Xia*
Palladium has been recognized as the best anodic, monome-
tallic electrocatalyst for the formic acid oxidation (FAO) reac-
tion in a direct formic acid fuel cell. Here we report a systematic
study of FAO on a variety of Pd nanocrystals, including cubes,
right bipyramids, octahedra, tetrahedra, decahedra, and icosa-
hedra. These nanocrystals were synthesized with approximate-
ly the same size, but different types of facets and twin defects
on their surfaces. Our measurements indicate that the Pd
nanocrystals enclosed by {100} facets have higher specific ac-
tivities than those enclosed by {111} facets, in agreement with
prior observations for Pd single-crystal substrates. If comparing
nanocrystals predominantly enclosed by a specific type of
facet, {100} or {111}, those with twin defects displayed greatly
enhanced FAO activities compared to their single-crystal coun-
terparts. To rationalize these experimental results, we per-
formed periodic, self-consistent DFT calculations on model
single-crystal substrates of Pd, representing the active sites
present in the nanocrystals used in the experiments. The calcu-
lation results suggest that the enhancement of FAO activity on
defect regions, represented by Pd(211) sites, compared to the
activity of both Pd(100) and Pd(111) surfaces, could be attrib-
uted to an increased flux through the HCOO-mediated path-
way rather than the COOH-mediated pathway on Pd(211).
Since COOH has been identified as a precursor to CO, a site-
poisoning species, a lower coverage of CO at the defect re-
gions will lead to a higher activity for the corresponding nano-
crystal catalysts, containing those defect regions.
Introduction
Direct liquid fuel cells involving electrochemical oxidation of
liquid fuels on the anodes have received considerable interest
as the next-generation power sources for portable devices
as crossover flux through the membrane. Ethanol, another
promising liquid fuel, also generates CO during its electrooxi-
dation. Most of these issues can be addressed by switching to
a liquid fuel based on formic acid. For example, the level of CO
produced from formic acid oxidation (FAO) could be lower
than that from the oxidation of methanol or ethanol by em-
[1–3]
such as smartphones and laptops.
For several decades,
methanol has been widely explored as a liquid fuel for such
devices owing to its high energy density. However, methanol is
toxic and its electrooxidation is plagued by problems such as
production of CO, and thus poisoning of the catalyst, as well
[
4–6]
ploying a Pd catalyst (see below).
It is also feasible to use
formic acid at a higher concentration than methanol because
of its lower toxicity and crossover flux. As a result, FAO on vari-
ous types of electrocatalysts has received considerable atten-
[
a] Dr. S.-I. Choi, H. Huang, Y. Wang, Dr. X. Xia, T. Lv, J. Park, H.-C. Peng,
Prof. Y. Xia
[4–24]
tion in recent years for fuel cell applications.
The Wallace H. Coulter Department of Biomedical Engineering
School of Chemistry and Biochemistry
and School of Chemical and Biomolecular Engineering
Georgia Institute of Technology, Atlanta, Georgia 30332 (USA)
E-mail: younan.xia@bme.gatech.edu
It has been established that FAO can proceed through two
[
4,7]
different pathways.
In the direct oxidation pathway, CO is
2
formed after two steps of proton/electron transfer, and CO2
can readily desorb from the surface of a catalyst as a result of
its weak binding, liberating active sites for further catalytic re-
actions. In the indirect oxidation pathway, CO is formed after
COÀO bond scission in a carboxyl (COOH) intermediate. CO
binds strongly to the surface of a catalyst and acts as a poison,
[
b] Dr. S.-I. Choi
Current address: Department of Chemistry
Kyungpook National University, Daegu, 702-701 (Korea)
[
c] Dr. J. A. Herron, Dr. J. Scaranto, Prof. M. Mavrikakis
Department of Chemical and Biological Engineering
University of Wisconsin-Madison
which must be oxidized to CO in the following step to be re-
2
Madison, Wisconsin 53706 (USA)
moved from the catalyst. The direct pathway is dominant at
low potentials of approximately 0.4 V (vs. reversible hydrogen
electrode, RHE) and the indirect pathway occurs at potentials
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201500094.
This publication is part of a Special Issue on Palladium Catalysis. To view
the complete issue, visit:
http://onlinelibrary.wiley.com/doi/10.1002/cctc.v7.14/issuetoc
[
8]
over 0.8 V. Typically, it is more efficient to oxidize formic acid
through the direct pathway, but the gradual accumulation of
ChemCatChem 2015, 7, 2077 – 2084
2077
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
CO at potentials of approximately 0.4 V may still diminish the
catalytic activity over time. An ideal catalyst for FAO should
work at potentials of approximately 0.4 V, together with an
ability to prevent the formation of CO and/or resist the poison-
ing by CO.
Palladium is considered the most effective catalyst for FAO
owing to its high activity in directly oxidizing formic acid to
[7]
CO at relatively low potentials. The high activity can be at-
2
tributed to the low CO formation flux, thereby mitigating the
poisoning effect of CO as compared with Pt, another catalyst
[9,10]
that has been explored for FAO.
Although the Pd catalyst
Figure 1. Pd nanocrystals with different shapes, and the facets exposed on
their faces and twin boundries.
can still be deactivated because of the gradual accumulation
of CO during operation, the high activity can be replenished
by increasing the potential
[
11]
beyond 1.0 V.
The activity of
a Pd catalyst is typically deter-
mined by the arrangement of
atoms on the surface. For the
low-index facets on Pd single-
crystal substrates, their FAO ac-
tivities increase in the order of
Table 1. Summary of the FAO activities of Pd nanocrystals with different shapes.
[
a]
[a]
Shape of Pd
nanocrystal
Size of
nanocrystal
[nm]
Number of
twin boundaries
SA at 0.4 V
Anodic peak
potential
[V]
SA at anodic
À2
[mAcm
]
peak potential
À2
[mAcm
]
cube
15
15
13
15
15
17
–
1
–
–
5
5.9
9.1
5.0
3.0
7.4
8.2
0.54
0.56
0.47
0.40
0.47
0.46
10.1
20.8
6.0
3.0
10.0
10.4
right bipyramid
octahedron
tetrahedron
decahedron
icosahedron
[10]
Pd(110)<Pd(111)<Pd(100).
Over the past decade, there has
been a strong interest in control-
ling the shape of Pd nanocrystals
to preferentially expose the
most active {100} facets. In a pre-
vious study, we tested Pd cubes
30
[
a] Specific activity.
[
7, 10,12–18]
and octahedra as two model systems to understand the de-
mize any possible variations or errors.
Both the cubes
[14]
pendence of FAO activity on crystal facet. In good agree-
ment with the results obtained from single-crystal substrates,
Pd cubes enclosed by {100} facets were found to be more
active than octahedra enclosed by {111} facets. In addition to
the type of facet, the twin defect or stacking fault on the sur-
face of a Pd nanocrystal can also affect its performance in cata-
and right bipyramids are enclosed by {100} facets, whereas
the octahedra, tetrahedra, decahedra, and icosahedra are all
enclosed by {111} facets. The Pd nanocrystals enclosed by
{100} facets showed higher specific activity than those en-
closed by {111} facets, in agreement with previous observa-
[
10]
tions on Pd single-crystal substrates. For Pd nanocrystals en-
closed by the same type of facet but possessing a single-crys-
tal or twin structure, those with twin defects on the surfaces
showed higher specific activities. At 0.4 V, a potential responsi-
ble for the direct oxidation pathway, both the Pd decahedra
and icosahedra exhibited greater specific activities than the
octahedra and tetrahedra. In addition, the FAO activity was
found to be affected more significantly by the twin defects
relative to the difference in low-index facets. In this respect,
the decahedra and icosahedra exhibited higher specific activi-
ties than the cubes, even though they were enclosed by {111}
facets with a lower activity than the {100} facets on cubes.
Among all the nanocrystals we have tested, the right bipyra-
mids exhibited the highest specific activity at 0.4 V. To under-
stand the experimental results, we performed periodic self-
consistent DFT calculations (GGA-PW91) on model single-crys-
tal substrates of Pd. Because of the enhancement in flux
through the HCOO-mediated pathway on the Pd(211) surface,
representing the structure of defect sites on the nanocrys-
[
16–24]
lyzing FAO.
Recently, we demonstrated that the specific
À2
activity (j in mAcm ) toward FAO on Pd icosahedra was two-
fold higher than that of Pd octahedra, even though both geo-
[16]
metries are predominantly exposing (111) facets. This en-
hancement could be attributed to the presence of twin defects
on the surface of Pd icosahedra. We also studied the effect of
twin defects on the FAO activity by comparing Pd right bipyra-
mids with single-crystal Pd cubes, and a two-fold enhancement
[
17]
was observed. The enhancement in activity could be attrib-
uted to the {211} facets exposed on the twin defect of a Pd
right bipyramid. The reaction on this high-index facet prefers
the HCOO-mediated pathway, helping reduce the formation of
CO.
Herein, we report, for the first time, a comprehensive and
systematic study of FAO on Pd nanocrystals with different
types of facets and twin defects, including cubes, right bipyra-
mids, octahedra, tetrahedra, decahedra, and icosahedra
[25]
(
Figure 1). To minimize the effect of size, all the Pd nano-
crystals were prepared with a similar size in the range of 13–
7 nm for edge lengths or diameters (Table 1). Their corre-
[
24]
tals, the level of CO is reduced relative to that on Pd(100)
and Pd(111), leaving more catalytic sites on the nanocrystals
for the electrooxidation of formic acid to CO2.
1
sponding FAO activities were then measured by using the
same setup and under essentially identical conditions to mini-
ChemCatChem 2015, 7, 2077 – 2084
www.chemcatchem.org
2078
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
rahedra with average edge lengths of 13 and 15 nm, respec-
tively, in a mixture of ethylene glycol (EG) and tetraethylene
glycol (TTEG). The Pd decahedra (Figure 2E) were prepared
with an average diameter of 15 nm by adding Na PdCl into
2
4
a diethylene glycol (DEG) solution containing poly(vinyl pyrroli-
[
19]
done) (PVP) and Na SO . The Pd icosahedra were synthesized
2
4
by adding HCl into the polyol synthesis, which could control
the pH value of the solution and thus manipulate the reaction
[
16]
kinetics. The use of HCl at an optimal concentration could
lead to the formation of Pd icosahedra with high purity (>
9
4%) and uniformity. The sample shown in Figure 2F had an
average diameter of 17 nm. All the detailed descriptions for
the synthesis of Pd nanocrystals with different shapes are pre-
sented in the Experimental Section.
Comparison of Pd nanocrystals with different types of twin
structures
In Scheme 1, simple geometric models of two different types
of twin defects that are involved in the Pd nanocrystals evalu-
ated by this work are shown. The first type of twin defect is
a mirror symmetry plane that can be introduced into the lat-
tice of a crystal without causing any strain to the lattice. A typ-
ical example is the right bipyramid, which is bisected by a sym-
metry plane along the <111> direction in the middle of the
nanocrystal and enclosed by six right-isosceles triangular {100}
Figure 2. TEM images of various types of Pd nanocrystals used in the pres-
ent study: A) cubes, B) right bipyramids, C) octahedra, D) tetrahedra, E) deca-
hedra, and F) icosahedra. The nanocrystals had edge lengths of A) 15, B) 15,
C) 13, and D) 15 nm, respectively. The diameters of Pd decahedra and icosa-
hedra were 15 and 17 nm, respectively.
Results and Discussion
Synthesis of Pd nanocrystals with different types of facets
and twin defects
The synthesis of Pd cubes was performed in the presence of
À
KBr because of the important role of Br ions in promoting the
[14,26]
formation of {100} facets.
the as-obtained Pd cubes with an average edge length of
5 nm is shown to demonstrate good uniformity in shape and
In Figure 2A, a TEM image of
1
narrow distribution in size. In Figure 2B, a TEM image of Pd
right bipyramids synthesized in the presence of NaI is shown,
À
where I ions can act as an oxidative etchant for the selective
Scheme 1. Schematic illustrations of different types of twin defects on A) a
right bipyramid, B) a decahedron, and C) an icosahedron. A) The right bipyr-
amid has a single twin defect, by which mirror symmetry is introduced into
the crystal lattice. B) The decahedron can be considered as an assembly of
five single-crystal tetrahedra. There is a gap of 7.358 in the lattice as a result
of the unique structure. C) The icosahedron can be viewed as a densely
packed array of twenty single-crystal tetrahedra. The icosahedron with a gap
of 1.54 steradians (sr) will cause an internal lattice strain and a disordered
region at the twin boundaries (similar to the twin zone of a decahedron).
removal of multiple-twinned Pd nanoparticles and as a selective
[17]
capping agent for the {100} facets. The Pd nanocrystals with
other shapes such as octahedra and tetrahedra (Figure 2, C
and D) were prepared by seed-mediated growth with the use
of different combinations of seeds and precursors, including
[15]
Na PdCl and Pd(acac) . Specifically, 5 nm Pd cuboctahedra
2
4
2
were used as seeds for the syntheses of Pd octahedra and tet-
ChemCatChem 2015, 7, 2077 – 2084
www.chemcatchem.org
2079
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
facets, as shown in Scheme 1A. This type of twin defect only
leads to the formation of three {211} high-index facets on the
nanocrystal, and it does not cause strains to the crystal lat-
peaks for the Pd nanocrystals enclosed by {111} facets all ap-
[
28]
peared below 0.53 V (Figure S3B). The ECSAs of the catalysts
were calculated by integrating the stripping charges of CuUPD.
[17]
À2
tice.
Charges of 420 and 490 mCcm were used for the Pd{100}
[
7,16,17]
The other type of twin defect corresponds to the boundary
region between two tetrahedral units in a decahedron or ico-
sahedron. As shown in Scheme 1B, a decahedral nanocrystal
can be considered to form from five tetrahedral units (all cov-
ered by {111} facets) by sharing one common edge along the
five-fold axis. A total of five twin boundaries are required to
generate the decahedral particle because it is impossible to
completely fill the space of a particle in five-fold symmetry
with a single-crystal lattice only. Specifically, the projection
angle of 70.538 for each tetrahedral unit brings a gap of 7.358
when the five tetrahedral units are assembled together by
sharing one common edge, leading to the formation of strain-
ed regions at the boundaries. Johnson and co-workers recently
demonstrated that the twin boundary of a decahedron is char-
acterized by disclination and graded strains in the crystal lat-
and Pd{111} facets, respectively.
We then conducted electrochemical measurements of FAO
for the Pd nanocrystals with different shapes using the same
setup and under essentially identical conditions (see the Exper-
imental Section) to minimize any possible variations or
[
7,10,12–18]
errors.
We first compared the electrocatalytic activities
between the nanocrystals enclosed by {100} and {111} facets,
and then compared those with a single-crystal or twinned
structure. In Figure 3, the CVs of FAO for the catalysts recorded
between 0.08 and 1.1 V in N -saturated solutions containing
2
[21]
tice. From the rigid-body rotation measurement, they found
that a gap of 4.38 could be filled through atomic disclination
but the remaining gap had to be filled through graded shear
[21]
strains on the lattice. An icosahedron consists of twenty tet-
rahedral units with thirty twin boundaries and twenty {111}
[
16,22–24]
facets.
As shown in Scheme 1C, the twin defect of a Pd
[
27]
icosahedron corresponds to a gap of 1.54 steradians.
As
every edge on an icosahedron is a result of twin boundaries,
the density of twin boundaries on the surface of an icosahe-
dron is higher than that on the decahedron. The facets corre-
sponding to the twin boundaries of a decahedron can be in-
dexed as {211} based on its projection along the [0 1¯ 1] zone
axis, as shown in Figure S1 in the Supporting Information. As
an icosahedron is comprised of twelve interpenetrating deca-
hedra, the twin boundaries of an icosahedron can also be as-
signed to the {211} facets.
Electrochemical measurements of the Pd nanocrystals
The surfaces of the as-synthesized Pd nanocrystals were cov-
À
À
ered by capping agents such as Br , I , and PVP. Therefore, we
performed plasma etching for 30 min, followed by holding the
electrical potential at À0.05 V (vs. RHE) for 60 s to remove the
[
7,17]
capping agents prior to electrochemical measurements.
To
Figure 3. Anodic cyclic voltammograms for the FAO on Pd nanocrystals en-
[25]
minimize the effect of particle size on FAO, all the Pd nano-
crystals were prepared with edge lengths or diameters in the
range of 13–17 nm (Figure 2). In Figure S2, typical cyclic vol-
tammograms (CVs) between 0.08 and 0.8 V for the Pd nano-
closed by A) {100} and B) {111} facets obtained in a N -saturated solution
2
À1
containing 0.5m HClO
4
and 0.5m HCOOH. Scanning speed=50 mVs . The
currents were normalized to the electroactive surface areas of the corre-
sponding Pd nanocrystals derived from the charges for the underpotential
deposition of Cu. The different shapes of Pd nanocrystals are indicated.
crystals obtained in N -saturated solutions containing 0.1m
2
HClO are shown. From the curves, we could easily identify the
4
À1
hydrogen adsorption/desorption peaks. To evaluate the elec-
trocatalytic surface areas (ECSAs), we performed stripping of
0.5m HClO and 0.5m HCOOH at a scan rate of 50 mVs are
4
shown. The specific activity was obtained by normalizing to
underpotentially deposited Cu (CuUPD) prepared in a N -saturat-
the ECSA of the catalyst derived from the charges of CuUPD.
2
ed solution containing 0.05m H SO and 0.05m CuSO . As
shown by the CuUPD curves in Figure S3, the Pd nanocrystals
exhibited different behaviors. For the nanocrystals enclosed by
The anodic CVs of FAO for Pd nanocrystals enclosed by {100}
facets, such as cubes and right bipyramids, are shown in Fig-
ure 3A. At the anodic peak potential, the specific activity of
the right bipyramids was much higher than that of the cubes
(Table 1). In Figure 3B, the anodic CVs of FAO for Pd nanocrys-
2
4
4
{
100} facets, a single CuUPD peak appeared at 0.56 V for cubes
[28]
and at 0.57 V for RBPs, respectively (Figure S3A). The CuUPD
ChemCatChem 2015, 7, 2077 – 2084
www.chemcatchem.org
2080
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 4. Comparison of the specific FAO activities on Pd nanocrystals with
different shapes and twin structures at 0.4 V. The data were derived from
Figure 3A,B.
Figure 5. DFT-calculated thermochemical potential energy surfaces for
formic acid (HCOOH) oxidation at 0.4 V through HCOO-mediated (solid lines)
and COOH-mediated (dashed lines) pathways on Pd(211) (red), Pd(111)
(
H
yellow), and Pd(100) (green). The reaction stoichiometry is balanced with
+
À
and e . For example, the energy level indicated by HCOO* corresponds
tals enclosed by {111} facets are given, including octahedra,
tetrahedra, decahedra, and icosahedra. For both the decahedra
and icosahedra with twinned structures, they displayed an en-
hancement in specific activity at the anodic peak potential rel-
ative to the single-crystal octahedra and tetrahedra (Table 1).
To compare the catalytic activities of these Pd nanocrystals
quantitatively, their specific activities at 0.4 V, a potential for
+
À
to HCOO*+H +e .
Table 2. Relative stability of HCOO versus COOH on model sites based
on Pd(211), Pd(111), and Pd(100) surfaces.
[a]
Relative stability of isomers [eV]
[7]
the direct oxidation pathway, are summarized in Figure 4 and
Table 1. The cubes enclosed by {100} facets showed enhance-
ment in specific activity relative to both the octahedra and tet-
rahedra enclosed by {111} facets, in agreement with the results
adsorbate
HCOO
COOH
Pd(211)
0.00
0.01
Pd(111)
0.11
0.00
Pd(100)
0.22
0.00
[a] The energy of the more stable isomer on each surface is defined as
[10]
of previous studies for single-crystal substrates. The nano-
crystals with twin defects (i.e., right bipyramids, decahedra,
and icosahedra) demonstrated enhanced specific activity with
respect to their single-crystal counterparts (i.e., right bipyra-
mids were more active than cubes, and the decahedra or ico-
sahedra were more active than octahedra or tetrahedra). Be-
tween decahedra and icosahedra, the latter sample exhibited
slightly higher activity than the former one. This result could
be explained by the larger number of twin boundary regions
on the surface of an icosahedron than that on a decahedron
the zero-energy reference for that surface. More positive entries signify
less stable species. HCOO is practically isoenergetic with COOH on
Pd(211) whereas COOH is more stable than HCOO on both Pd(111) and
Pd(100).
two reaction pathways (see Figure 5, the binding energies
used to construct the potential energy surface are provided in
Table S1 in the Supporting Information): carboxyl-mediated
and formate-mediated. As shown in Table 2, the carboxyl
(COOH) species, which has been identified as a precursor to
CO formation, is more stable than formate (HCOO) by 0.22 eV
on Pd(100) and more stable by 0.11 eV on Pd(111). In contrast,
HCOO and COOH are isoenergetic on Pd(211). The relative sta-
bility of HCOO over COOH drives the reaction flux and selectiv-
ity toward the HCOO-mediated pathway rather than the
COOH-mediated pathway. This result is important because, on
all these surfaces, it is thermochemically favorable for COOH to
be dissociated into CO and OH. If formed, the CO will compete
with HCOO for the surface sites, poisoning the surface and
lowering its catalytic activity. We note that in our calculations,
the potential energy surface did not include the stabilization
effect of water toward adsorbed OH, which has been calculat-
(
Table 1). In addition, both decahedra and icosahedra showed
enhanced specific activity at 0.4 V relative to the cubes even
though they were enclosed by {111} facets that are less active
than {100}. Taken together, it can be concluded that the FAO
activity could be enhanced more significantly by including
twin defects rather than by varying the low-index facets from
{
111} to {100}.
DFT calculations for FAO on Pd nanocrystals
To rationalize the experimental observations, we performed pe-
riodic self-consistent DFT calculations (GGA-PW91) on model
single-crystal surfaces, which would capture the various types
of active sites that may be present in the nanocrystals our ex-
periments were performed on. The twin defect on the surface
of a right bipyramid, decahedron, or icosahedron was modeled
as a Pd(211) surface (Figure S1). The terraces in the defect
zone were modeled as Pd(111) and Pd(100) surfaces. We eval-
uated the potential energy surfaces at 0.4 V for FAO through
[
29]
ed to be ꢀ0.5 eV on Pt(111), which, in turn, would make CO
oxidation by OH to CO2 more difficult than it appears in
Figure 5. Therefore, by enhancing flux through the HCOO-
mediated pathway on Pd(211), the formation of CO will be re-
duced relative to those on Pd(100) and Pd(111), leaving rela-
ChemCatChem 2015, 7, 2077 – 2084
www.chemcatchem.org
2081
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
tively more catalytic sites on Pd(211) free of CO (compared to
AA (60 mg), and KBr (500 mg) was placed in a 20 mL vial. The mix-
ture was preheated at 808C for 10 min under magnetic stirring.
Then, a 3.0 mL volume of an aqueous solution containing Na PdCl
4
(
100) and (111)) for FAO to directly produce CO2.
We also investigated the role of surface strain (up to Æ5%)
2
(
57 mg) was rapidly injected into the vial using a pipette. The reac-
on the binding properties of Pd(211) surfaces to examine the
impact of dislocation and shear gradients resulting from the
twin defects on decahedra and icosahedra. There were only
very small changes to the binding characteristics (see the Sup-
porting Information, as well as Table S2 and Figure S4). Addi-
tionally, the effect of strain on the relative stability of HCOO
versus COOH intermediates was probed on Pd(111) with 0–3%
tion solution was allowed to proceed at 808C for 3 h.
For the synthesis of Pd right bipyramids with an edge length of
[17]
1
5 nm along the <100> direction, a 5.0 mL volume of EG con-
taining PVP (400 mg) and NaI (150 mg) was added into a 25 mL
three-neck, round-bottom flask. The mixture was preheated at
1
608C for 10 min under magnetic stirring. Then, a 1.0 mL volume
of EG containing Na PdCl4 (15 mg) was rapidly injected into the
2
[
30]
tensile strain (see the Supporting Information and Table S3),
flask with a pipette. The reaction was allowed to proceed at 1608C
and the effect was again found to be minimal, almost negligi-
ble if compared to the magnitude of change between different
types of facets. Overall, this simple analysis provides one possi-
ble explanation for the observed enhancement in activity on
the nanocrystals with twin defects compared to their single-
crystal counterparts. A more detailed mechanistic analysis may
be required to evaluate other possible mechanisms responsible
for the observed activity enhancement.
for 2 h.
Syntheses of Pd octahedra, tetrahedra, decahedra, and
icosahedra
For the syntheses of Pd octahedra and tetrahedra (with edge
lengths of 13 and 15 nm, respectively), a 2.5 mL volume of TTEG
containing PVP (10 mg) and a 0.1 mL volume of a suspension of
[15]
À1
the 5 nm cuboctahedral Pd seeds in EG (1.8 mgmL in concentra-
tion) were placed in a 20 mL vial. The mixture was preheated at
Conclusions
1
408C for 10 min under magnetic stirring. Meanwhile, a 0.5 mL
We have systematically investigated the formic acid oxidation
volume of TTEG solution containing Na PdCl (2 mg) or an equal
2
4
(
FAO) activities of Pd nanocrystals with different shapes and
molar amount of Pd(acac) was prepared for the synthesis of Pd
2
octahedra or tetrahedra, respectively. After the precursor had been
completely dissolved, the solution was quickly injected into the
vial by using a pipette. The reaction solution was allowed to pro-
ceed at 1408C for 1 h.
twin structures, including cubes, right bipyramids, octahedra,
tetrahedra, decahedra, and icosahedra. The nanocrystals en-
closed by {100} facets were found to show higher specific ac-
tivities than those enclosed by {111} facets. For nanocrystals
enclosed by the same type of facet but with a single-crystal or
twin structure, those with twin defects on the surfaces showed
higher specific activities. It is interesting to note that both dec-
ahedra and icosahedra exhibited higher specific activities than
cubes even though the {111} facets are less active than the
[19]
For the synthesis of Pd decahedra 15 nm in diameter, a 2.0 mL
volume of DEG containing PVP (80.0 mg) and Na SO4 (40.0 mg)
2
was placed in a 20 mL vial. The mixture was preheated at 1058C
for 20 min under magnetic stirring. Then, a 1.0 mL volume of DEG
containing Na PdCl4 (15.5 mg) was rapidly injected into the vial
2
with a pipette. The reaction was allowed to proceed at 1058C for
{
100} facets. In these cases, the presence of twin defects im-
2
4 h.
poses a stronger impact on the catalytic activity toward FAO
than the type of facet. To understand the correlation between
the specific activity and the twin defect on the nanocrystal,
DFT calculations were performed on model single-crystal surfa-
ces of Pd. The formation of CO is reduced on Pd(211) if com-
pared to both Pd(100) and Pd(111), retaining a higher fraction
of the defect sites free of CO, for FAO.
[
16]
For the synthesis of Pd icosahedra 17 nm in size,
a 2.0 mL
volume of EG containing PVP (30 mg) was placed in a 20 mL vial.
The mixture was pre-heated at 1608C for 20 min under magnetic
stirring. Meanwhile, H PdCl was prepared by dissolving PdCl in
a mixture of EG and 37 vol% HCl, in which the molar ratio of HCl
to PdCl was set to 4:1 and the concentration of Pd to 50 mm.
2
4
2
II
2
Then, the H PdCl solution (1 mL, 50 mm) was added into the vial
2 4
in one shot. A specific amount of HCl was also added to achieve
a final concentration of 134 mm in the reaction mixture. The reac-
tion was allowed to proceed at 1608C for 3 h.
Experimental Section
All the syntheses were quenched by immersing the vials in an ice-
water bath and the products were washed with acetone once and
DI water five times by centrifugation prior to the electrochemical
measurements.
Chemicals and materials
Palladium(II) chloride (PdCl , 99.9 %), Na PdCl (99.99%), Pd(acac)
2
2
2
4
(
(
99.0%), PVP (MWꢀ55000), l-ascorbic acid (AA, 99.0%), KBr
99.0%), NaI (99.5%), Na SO (99.0%), DEG (99.0%), TTEG (90%),
2
4
and HCl (37%) were all purchased from Sigma–Aldrich and used as
received without further purification. EG (99.0%) was obtained
from J. T. Baker. Deionized (DI) water with a resistivity of
Characterization
1
8.2 MWcm was used for all experiments.
The TEM images were taken by using a microscope (HT7700, Hita-
chi) operated at 120 kV by drop-casting the nanoparticle disper-
sions on carbon-coated copper grids and drying under ambient
conditions. The particle concentration of each suspension of Pd
nanocrystals was determined by using inductively coupled plasma
mass spectrometry (ICP–MS, NexION 300Q, PerkinElmer).
Syntheses of Pd cubes and right bipyramids
For the synthesis of Pd cubes with an edge length of 15 nm, an
[14]
8
.0 mL volume of an aqueous solution containing PVP (105 mg),
ChemCatChem 2015, 7, 2077 – 2084
www.chemcatchem.org
2082
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Electrochemical measurements
The binding energies (E ) of the FAO intermediates were calculated
b
with respect to the clean (relaxed, if applicable) surfaces (Esubstrate
and the respective free adsorbate in the gas phase (Egas-phase adsorbate),
, where Etotal is the energy of
)
Samples of approximately 0.2 mg of the Pd nanocrystals and com-
mercial Pd black (0.2 mg, Strem Chemicals) were dispersed in
a 1.0 mL volume of DI water and treated by ultrasonication for
i.e., E =E ÀE
ÀE
b
total
substrate
gasÀphase adsorbate
the surface with the adsorbate adsorbed on it. The surface inter-
1
0 min. To prepare the working electrode, a 10 mL volume of an
mediates considered were COOH, HCOO, CO, and OH. Their E are
b
aqueous suspension of the catalyst was dropped onto the pre-
cleaned glassy carbon electrode (Bioanalytical Systems Inc.). Upon
drying in an oven at 508C for 10 min, the electrode was covered
with Nafion aqueous solution (5 mL, 0.05%) and allowed to dry in
an oven set to 508C for another 10 min. Then, plasma etching (PE-
presented in Table S1.
The thermochemistry of the elementary steps in the following re-
action network was calculated using the binding energy:
þ
À
HCOOHðgÞ þ * ! HCOO* þ H þ e
ð1Þ
ð2Þ
ð3Þ
ð4Þ
ð5Þ
5
0, Plasma Etch Inc.) was performed for 30 min to remove the PVP
on the surfaces of the Pd catalysts. An Ag/AgCl electrode and a Pt
þ
À
HCOO* ! CO ðgÞ þ H þ e
2
2
mesh (1”1 cm ) were used as the reference and counter electro-
þ
À
des, respectively. The potentials are presented with reference to
HCOOHðgÞ þ * ! COOH* þ H þ e
À
RHE. To further remove the capping agents such as PVP, Br , and
þ
À
À
[7]
COOH* ! CO ðgÞ þ H þ e
2
I , the electrical potential was held at À0.05 V for 60 s. CVs were
obtained by cycling the potential between 0.08 and 0.8 V for
COOH* þ * ! CO* þ OH*
1
5
0 cycles in N -saturated 0.1m HClO solutions at a scan speed of
2 4
0 mVs . The ECSAs were obtained from the charges associated
À1
The calculated reaction thermochemistry for proton/electron trans-
fer reactions (elementary steps 1–5) were adjusted to an artificial
electrode potential of 0.4 V by using an approach similar to that of
with the stripping of CuUPD on the Pd nanocrystals by assuming
À2
4
20 and 490 mCcm for a full monolayer coverage of Cu on Pd
À2
[17,43–46]
enclosed by {100} and {111} facets, respectively, and 460 mCcm
for commercial Pd black. The CuUPD was conducted in a N -satu-
Nørskov and co-workers.
First, we chose the RHE as a refer-
[7]
2
ence. Under standard conditions, hydrogen gas is in equilibrium
with protons and electrons, at a defined potential of 0 V. A change
in the electrode potential by U will shift the free energy of each
electron exothermically by eU, where e is the absolute charge of
an electron. For this study, we have neglected the entropic and
zero-point energy contributions to the free energy of the reactions
because these corrections will not vary significantly between the
surfaces.
rated solution containing 0.05m H SO and 0.05m CuSO . To obtain
2
4
4
FAO activity, the catalyst was tested between 0.08 and 1.1 V for
two cycles in a N -saturated solution containing 0.5m HClO and
2
4
À1
0
.5m HCOOH at a scan speed of 50 mVs . All the electrochemical
measurements were conducted with a CHI600E potentiostat (CH
Instrument).
Acknowledgements
Theoretical calculations
[
31,32]
This work was supported by startup funds from the Georgia Insti-
tute of Technology and by DOE–BES, Office of Chemical Sciences,
grant DE-FG02-05ER15731. The computational work was per-
formed in part by using supercomputing resources from the fol-
lowing institutions: EMSL, a national scientific user facility at Pa-
cific Northwest National Laboratory (PNNL); the Center for Nano-
scale Materials at Argonne National Laboratory (ANL); and the
National Energy Research Scientific Computing Center (NERSC).
EMSL is sponsored by the Department of Energy’s Office of Bio-
logical and Environmental Research located at PNNL. CNM and
NERSC are supported by the U.S. Department of Energy, Office of
Science, under contracts DE-AC02-06CH11357 and DE-AC02-
All calculations were performed using DACAPO.
A 3”3 unit
cell was used to construct a four layer Pd(100) slab (36 slab atoms
in total). A 1”3 unit cell was used to construct a ten layer Pd(211)
slab (30 slab atoms in total). A 3”3 unit cell was used to construct
a three layer Pd(111) slab (27 slab atoms in total). The surface cov-
erage of adsorbates on all three surfaces was set to 1/9 ML, with
only one adsorbate per unit cell. Each unit cell was repeated in
super cell geometries with successive slabs separated by a vacuum
region of at least 10 Š. The optimized bulk lattice constant for Pd
was calculated to be 3.99 Š, in good agreement with the experi-
[33]
mental value of 3.89 Š. Adsorption was only allowed on one of
the two exposed surfaces, with the electrostatic potential adjusted
[34,35]
accordingly.
The top two layers of the Pd(100) slab were re-
laxed and the top five layers of the Pd(211) slab were relaxed,
whereas all layers of the Pd(111) slab were fixed because the
effect of relaxation for this surface was negligible. The surface
05CH11231, respectively.
[36]
Brillouin zone was sampled using a 4”4”1 Monkhorst-Pack k-
Keywords: crystal growth · density functional calculations ·
fuel cells · reaction mechanisms · palladium
[37]
point mesh for Pd(211) and Pd(100), and was sampled using 18
[38]
special k-points for the Pd(111) surface. The Kohn–Sham one-
electron valence states were expanded in a basis of plane waves
with kinetic energies up to 25 Ry, and ionic cores were described
[1] W. Vielstich, A. Lamm, H. A. Gasteiger, in Handbook of Fuel Cells - Funda-
mentals, Technology and Applications, Wiley, Chichester, 2003.
[2] B. C. H. Steele, A. Heinzel, Nature 2001, 414, 345–352.
[39]
by ultrasoft Vanderbilt pseudopotentials. The exchange-correla-
tion potential and energy were described self-consistently using
[3] C. Rice, S. Ha, R. I. Masel, P. Waszczuk, A. Wieckowski, T. Barnard, J.
Power Sources 2002, 111, 83–89.
[40,41]
the GGA-PW91 functional.
The electron density was deter-
[4] A. Capon, R. Parsons, J. Electroanal. Chem. Interfacial Electrochem. 1973,
mined by iterative diagonalization of the Kohn–Sham Hamiltonian,
4
4, 1–7.
Fermi population of the Kohn–Sham states (k T=0.1 eV), and Pulay
B
42]
[
[
5] X. Yu, P. G. Pickup, J. Power Sources 2008, 182, 124–132.
6] Y.-W. Rhee, S. Y. Ha, R. I. Masel, J. Power Sources 2003, 117, 35–38.
[
mixing of the resulting electronic density.
The total energies
were then extrapolated to k T=0 eV. Convergence with respect to
[7] M. Shao, J. Odell, M. Humbert, T. Yu, Y. Xia, J. Phys. Chem. C 2013, 117,
B
various calculation parameters was tested.
4172–4180.
ChemCatChem 2015, 7, 2077 – 2084
www.chemcatchem.org
2083
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[
[
8] H. Okamoto, W. Kon, Y. Mukouyama, J. Phys. Chem. B 2005, 109, 15659–
5666.
9] A. Miki, S. Ye, M. Osawa, Chem. Commun. 2002, 1500–1501.
[28] A. Cuesta, L. A. Kibler, D. M. Kolb, J. Electroanal. Chem. 1999, 466, 165–
168.
[29] V. Viswanathan, H. A. Hansen, J. Rossmeisl, J. K. Nørskov, ACS Catal.
2012, 2, 1654–1660.
[30] J. Wu, L. Qi, H. You, A. Gross, J. Li, H. Yang, J. Am. Chem. Soc. 2012, 134,
11880–11883.
1
[
10] N. Hoshi, K. Kida, M. Nakamura, M. Nakada, K. Osada, J. Phys. Chem. B
006, 110, 12480–12484.
11] S. Ha, R. Larsen, R. I. Masel, J. Power Sources 2005, 144, 28–34.
2
[
[
12] Z. Niu, Q. Peng, M. Gong, H. Rong, Y. Li, Angew. Chem. Int. Ed. 2011, 50,
[31] B. Hammer, L. B. Hansen, J. K. Nørskov, Phys. Rev. B 1999, 59, 7413–
7421.
6
315–6319; Angew. Chem. 2011, 123, 6439–6443.
[
13] X. Huang, S. Tang, X. Mu, Y. Dai, G. Chen, Z. Zhou, F. Ruan, Z. Yang, N.
Zheng, Nat. Nanotechnol. 2011, 6, 28–32.
[32] J. Greeley, J. K. Nørskov, M. Mavrikakis, Annu. Rev. Phys. Chem. 2002, 53,
319–348.
[
[
14] M. Jin, H. Zhang, Z. Xie, Y. Xia, Energy Environ. Sci. 2012, 5, 6352–6357.
15] Y. Wang, S. Xie, J. Liu, J. Park, C. Z. Huang, Y. Xia, Nano Lett. 2013, 13,
[33] CRC Handbook of Chemistry and Physics, 92nd ed, CRC Press, New York,
2011.
2
276–2281.
[34] L. Bengtsson, Phys. Rev. B 1999, 59, 12301–12304.
[35] J. Neugebauer, M. Scheffler, Phys. Rev. B 1992, 46, 16067–16080.
[36] A. Kokalj, M. Causa, J. Phys. Condens. Matter 1999, 11, 7463–7480.
[37] H. J. Monkhorst, J. D. Pack, Phys. Rev. B 1976, 13, 5188–5192.
[38] D. J. Chadi, M. L. Cohen, Phys. Rev. B 1973, 8, 5747–5753.
[39] D. Vanderbilt, Phys. Rev. B 1990, 41, 7892–7895.
[40] J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson,
D. J. Singh, C. Fiolhais, Phys. Rev. B 1992, 46, 6671–6687.
[41] J. A. White, D. M. Bird, Phys. Rev. B 1994, 50, 4954–4957.
[42] G. Kresse, J. Furthmuller, Comput. Mater. Sci. 1996, 6, 15–50.
[43] J. K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J. R. Kitchin, T. Bli-
gaard, H. Jonsson, J. Phys. Chem. B 2004, 108, 17886–17892.
[44] J. Rossmeisl, P. Ferrin, G. A. Tritsaris, A. U. Nilekar, S. Koh, S. E. Bae, S. R.
Brankovic, P. Strasser, M. Mavrikakis, Energy Environ. Sci. 2012, 5, 8335–
8342.
[
[
[
[
[
[
[
16] T. Lv, Y. Wang, S.-I. Choi, M. Chi, J. Tao, L. Pan, C. Z. Huang, Y. Zhu, Y. Xia,
ChemSusChem 2013, 6, 1923–1930.
17] X. Xia, S.-I. Choi, J. A. Herron, N. Lu, J. Scaranto, H.-C. Peng, J. Wang, M.
Mavrikakis, M. J. Kim, Y. Xia, J. Am. Chem. Soc. 2013, 135, 15706–15709.
18] Y. Xia, Y. Xiong, B. Lim, S. E. Skrabalak, Angew. Chem. Int. Ed. 2009, 48,
6
0–103; Angew. Chem. 2009, 121, 62–108.
19] H. Huang, Y. Wang, A. Ruditskiy, H.-C. Peng, X. Zhao, L. Zhang, J. Liu, Z.
Ye, Y. Xia, ACS Nano 2014, 8, 7041–7050.
20] J. H. Conway, S. Torquato, Proc. Natl. Acad. Sci. USA 2006, 103, 10612–
1
0617.
21] C. L. Johnson, E. Snoeck, M. Ezcurdia, B. Rodriguez-Gonzalez, I. Pastori-
za-Santos, L. M. Liz-Marzan, M. J. Hytch, Nat. Mater. 2008, 7, 120–124.
22] M. R. Langille, J. Zhang, M. L. Personick, S. Li, C. A. Mirkin, Science 2012,
3
37, 954–957.
[
[
[
[
23] W. Zhou, J. Wu, H. Yang, Nano Lett. 2013, 13, 2870–2874.
24] R. Wang, H. Zhang, M. Farle, C. Kisielowski, Nanoscale 2009, 1, 276–279.
25] Y. Suo, I.-M. Hsing, Electrochim. Acta 2009, 55, 210–217.
[45] P. Ferrin, A. U. Nilekar, J. Greeley, M. Mavrikakis, J. Rossmeisl, Surf. Sci.
2008, 602, 3424–3431.
[46] A. U. Nilekar, M. Mavrikakis, Surf. Sci. 2008, 602, L89–L94.
26] H. C. Peng, S. Xie, J. Park, X. Xia, Y. Xia, J. Am. Chem. Soc. 2013, 135,
3
780–3783.
[
27] A. Haji-Akbari, M. Engel, A. S. Keys, X. Zheng, R. G. Petschek, P. Palffy-
Received: January 30, 2015
Muhoray, S. C. Glotzer, Nature 2009, 462, 773–777.
Published online on April 13, 2015
ChemCatChem 2015, 7, 2077 – 2084
www.chemcatchem.org
2084
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim