The effect of platinum nanoparticle distribution on oxygen electroreduction activity and selectivity

Article abstract of DOI:10.1002/cctc.201300987

The catalytic activity and selectivity of Pt nanoparticles towards the oxygen reduction reaction (ORR) were investigated as a function of the Pt catalyst distribution. By means of the sputtering deposition technique, it was possible to fabricate Pt catalysts with different loadings that consisted of dispersed 2-3 nm particles, nanoparticle agglomerates and extended particulate layers. The transition from dispersed nanoparticles to extended layers led to a decrease in the electrochemical surface area (ECSA, m²_Pt g_Pt^-1) and to a shift of the platinum oxide reduction peak to more positive potentials, which indicates a decrease in the adsorption energy for oxygenated species. The latter finding was correlated to the observed decrease in specific activity with the increasing ECSA, that is, in the case of isolated nanoparticles, the higher adsorption energy for oxygenated species causes a reduction in the specific activity towards the ORR as larger amounts of active sites are blocked compared to extended surfaces. The presented data of specific and mass activity versus ECSA were found to follow a "master curve" obtained by comparing normalised Pt activities from different studies. The transition from dispersed Pt nanoparticles to extended layers also influences the Pt selectivity. At a decreased interparticle distance, a significant increase in the H₂O₂ production was observed below 0.6 V versus the reversible hydrogen electrode, which indicates the important role of a H₂O₂ desorption-readsorption reaction mechanism during the ORR on Pt nanoparticles. The effect of a close neighbor: Pt nanoparticles assembled in extended particulate layers show a higher specific activity and selectivity towards the oxygen reduction reaction (ORR) than dispersed Pt nanoparticles. The higher specific ORR activity for extended Pt layers is correlated to a lower adsorption energy for oxygenated species, which leads to a lower amount of blocked active sites for the ORR compared to isolated Pt nanoparticles.

Full text of DOI:10.1002/cctc.201300987

CHEMCATCHEM
FULL PAPERS
DOI: 10.1002/cctc.201300987
The Effect of Platinum Nanoparticle Distribution on
Oxygen Electroreduction Activity and Selectivity
Emiliana Fabbri,*[a] Susan Taylor,[b] Annett Rabis,[a] Pieter Levecque,[b] Olaf Conrad,[b]
Rꢀdiger Kçtz,[a] and Thomas J. Schmidt[a]
The catalytic activity and selectivity of Pt nanoparticles towards
the oxygen reduction reaction (ORR) were investigated as
a function of the Pt catalyst distribution. By means of the sput-
tering deposition technique, it was possible to fabricate Pt cat-
alysts with different loadings that consisted of dispersed
2–3 nm particles, nanoparticle agglomerates and extended par-
ticulate layers. The transition from dispersed nanoparticles to
extended layers led to a decrease in the electrochemical sur-
face area (ECSA, m2Pt gPtꢀ1) and to a shift of the platinum oxide
reduction peak to more positive potentials, which indicates
a decrease in the adsorption energy for oxygenated species.
The latter finding was correlated to the observed decrease in
specific activity with the increasing ECSA, that is, in the case of
isolated nanoparticles, the higher adsorption energy for oxy-
genated species causes a reduction in the specific activity to-
wards the ORR as larger amounts of active sites are blocked
compared to extended surfaces. The presented data of specific
and mass activity versus ECSA were found to follow a “master
curve” obtained by comparing normalised Pt activities from
different studies. The transition from dispersed Pt nanoparticles
to extended layers also influences the Pt selectivity. At a de-
creased interparticle distance, a significant increase in the H2O2
production was observed below 0.6 V versus the reversible hy-
drogen electrode, which indicates the important role of a H2O2
desorption–readsorption reaction mechanism during the ORR
on Pt nanoparticles.
Introduction
In the aim towards a sustainable energy economy, many ef-
forts have been devoted to improve the performance of poly-
mer electrolyte fuel cells (PEFCs) and minimise their overall
cost. Pt is the catalyst of choice both for the anodic and the
cathodic reaction, even though the oxygen reduction reaction
(ORR) that occurs at the positive electrode is hampered by
poor kinetics, which results in over-potentials of approximately
0.3–0.4 V.[1] Structure–activity relationships for the ORR on Pt
catalysts have been the topic of numerous investigations, both
from experimental and theoretical points of view. Many studies
have dealt with Pt single crystal electrochemical properties
and the contribution of different crystal planes to the ORR ac-
tivity of Pt.[2] Another large research focus has been the
change in specific and mass activity in the comparison of Pt
nanoparticles with different particle sizes. Some studies have
suggested since the early 1970s a variation of Pt activity with
particle size .[3] Pt nanoparticles in the range of 1–12 nm
showed a maximum in mass activity at 3 nm, with a significant
decrease in the specific activity below 3 nm.[4] To give a funda-
mental explanation for these findings, Kinoshita[5] suggested
that for highly dispersed Pt nanoparticles the maximum in
mass activity observed at approximately 3 nm particle size is
attributed to the variation of the major crystal faces with the
particle size. Later, it was pointed out that the differences in
activity among surfaces with different crystal faces arise from
the structure sensitivity towards the adsorption of anions,
which impedes the reaction.[2a] Although newer studies con-
firmed a maximum in mass activity for 2–3 nm particle sizes
and a constant decrease of the specific activity for a particle
size lower than 10 nm,[6] there have also been contradictory re-
sults that show size-independent specific activity[7] and mass
activity[8] or a continuous increase in mass activity with de-
creasing particle size.[7] Particularly, Nesselberger et al.[7b] found
that the ORR specific activity was almost constant for particle
sizes in the range of 5–1 nm, whereas a rapid increase of spe-
cific activity was observed going from Pt nanoparticles sup-
ported on high-surface-area carbon to unsupported Pt black
and to polycrystalline Pt. The contradictory results reported in
the literature on the particle size effect must be analysed by
also taking into account that different experimental procedures
are often used to determine the catalyst activity (e.g., Ohmic
drop correction, values from anodic or cathodic sweeps, capac-
itive current corrections, etc.). However, normalising the activi-
ty values from different studies to an arbitrary specific surface
area, it is possible to achieve an empirical master curve that
shows an increase in specific activity and a decrease in mass
activity as a function of the catalyst specific surface area.[1]
[a] Dr. E. Fabbri, A. Rabis, Dr. R. Kçtz, Prof. T. J. Schmidt
Electrochemistry Laboratory
Paul Scherrer Institut
Villigen-PSI, 5232 (Switzerland)
Fax: (+41)56-310-21-99
E-mail: emiliana.fabbri@psi.ch
[b] S. Taylor, Dr. P. Levecque, Dr. O. Conrad
HySA/Catalysis Centre of Competence
University of Cape Town
Rondebosch, 7701, South Africa
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1410

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Although substantial efforts have been devoted in the last
decades to study the effect of particle size on the ORR activity,
few studies have investigated the influence of the Pt interparti-
cle distance on the ORR activity and selectivity. It was initially
shown by Watanabe and co-workers[9] that for carbon-support-
ed Pt nanoparticles, if the inter-crystallite distance was lower
than 18 nm, the specific activity decreased significantly for a va-
riety of Pt crystallite sizes and carbon supports. The authors as-
cribed the reduction in specific activity to a mutual inhibition
for the ORR caused by the vicinity of the Pt crystallites. In
a more recent publication,[10] it has been reported that the spe-
cific activity increased significantly if the edge-to-edge distance
diminishes below 1 nm for Pt particles with different sizes. The
authors attributed this effect to a change in the electric
double-layer structure and potential distribution for a small in-
terparticle distance (<than the Debye length), which leads to
a variation of the OH coverage and, thus, of the specific activi-
ty. Other recent publications have shown further different rela-
tionships between the Pt interparticle distance and the specific
activity. Yang and co-workers[11] studied the influence of the in-
terparticle distance on the ORR activity for arrays of Pt nano-
particles deposited on glassy carbon. The ORR activity was not
influenced by the interparticle distance above 0.8 V vs. the re-
versible hydrogen electrode (RHE), whereas differences in the
kinetic currents were observed below 0.8 V vs. RHE. The au-
thors attributed the decrease in specific activity below 0.8 V vs.
RHE with the increase of the interparticle distance to a change
from a four- to a two-electron process (i.e., a larger amount of
H2O2 formation) as the Pt density decreased.[11] In a different
study, Seidel and co-workers[12] have also shown that for arrays
of Pt particles (diameter ~100 nm) deposited on glassy carbon,
H2O2 formation also increased with the decreasing Pt particle
density, whereas the specific activity was invariant with respect
to the Pt coverage. An increase in H2O2 formation on decreas-
ing the Pt loading in a Pt/Vulcan system was reported by
Inaba and co-workers.[13] All these studies point to the impor-
tant role of a H2O2 desorption–readsorption reaction mecha-
nism during the ORR on Pt particles.
only on the uppermost layer of the high-surface-area carbon
support, which allows the elimination of any possible concerns
about Pt nanoparticle distribution in a 3D volume. At the same
time, the use of high-surface-area carbon allows the interparti-
cle distance in a real fuel cell catalyst to be to better mimicked
and also a better anchoring of the Pt nanoparticles compared
to, for example, mirror-polished glassy carbon substrates. In
addition to the Pt activity, the ORR selectivity has been evalu-
ated by rotating-ring disk electrode measurements (RRDE). A
variation of the ORR activity and selectivity as a function of the
electrochemical surface area (ECSA) has been shown and possi-
ble explanations have been provided for both relationships.
Results and Discussion
The Pt nanoparticles were prepared by magnetron sputtering
on high-surface-area carbon Vulcan XC72. Rutherford back-
scattering[14] and ellipsometry measurements were used to de-
termine the Pt loading as a function of the deposition time.
Several Pt loadings of 2, 10, 20 and 50 mgPt cmꢀ2 were investi-
gated. The primary particle size was determined by sputtering
Pt nanoparticles on amorphous carbon film TEM grids. As
shown by the TEM images in Figure 1, primary particles of ap-
proximately 2–3 nm in diameter can be observed together
Figure 1. TEM images with different magnifications for primary particles
sputtered on amorphous carbon film TEM grids (which correspond to a load-
ing of 2 mgPt cmꢀ2).
From an analysis of earlier and more recent studies on the
influence of the Pt interparticle distance on the ORR, it is clear
that contradictory results are reported. As discussed above for
the particle-size effect, differences in the experimental proce-
dures as well as in the Pt morphology, electrochemical active
surface area, and support can also lead to divergent results for
the influence of the particle distribution on the activity and se-
lectivity towards ORR. The present study aims to gain a better
understanding of the influence of the Pt particle distribution
on the ORR mass and specific activity. Particularly, model elec-
trodes that present a morphology transition from isolated
nanoparticles to agglomerates and to extended surfaces were
prepared by magnetron sputtering of Pt nanoparticles under
defined conditions[14] on high-surface-area carbon. By varying
the deposition time, it was possible to investigate different Pt
distributions systematically on going from isolated Pt nanopar-
ticles of approximately 2–3 nm in diameter to extended Pt sur-
faces. It is worth highlighting that by the use of the sputtering
preparation technique, the Pt nanoparticles were synthesised
with agglomerates of a few particles. It is worth highlighting
that although on a flat surface a Pt loading of 2 mgPt cmꢀ2 re-
sults in some particle agglomeration (Figure 1), on high-sur-
face-area carbon such a low loading is expected to result in
isolated particles.[15]
As reported in a previous investigation,[15] if the sputtering
deposition time is increased, a constant increase of the Pt
amount on top of the carbon layer is obtained, but owing to
the process conditions, the primary particle size remained con-
stant (2ꢁ1 nm in diameter). For loadings lower than
3 mgPt cmꢀ2 dispersed Pt nanoparticles were observed, whereas
Pt agglomerates were formed for loadings >20 mgPt cmꢀ2; if
the Pt loading (i.e., deposition time) was further increased, ex-
tended particulate Pt layers were obtained. To further confirm
the same Pt loading/distribution trend, the present samples
were also analysed by energy-dispersive X-ray spectroscopy
(EDX) and X-ray photo-electron spectroscopy (XPS) to follow
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1411

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
the progressive Pt coverage of the Vulcan support. EDX map-
ping showed an increasing Pt signal and a corresponding de-
crease in the carbon signal with the increase of the Pt loading,
whereas it was possible to follow the decrease in the carbon
C1s and fluorine F1s peak intensities (which originate from
the Nafion used as a binder for the Vulcan support) with the
increase of the Pt loading by XPS analysis. Both techniques
confirmed that only for 50 mgPt cmꢀ2, the Vulcan surface was
mostly covered by a Pt layer. Particularly, the disappearance of
the fluorine energy line in the XPS spectra indicates the forma-
tion of a Pt layer thicker than approximately 5 nm (mean free
path of photo-electrons).
face (normalisation factor of 420 mCcmꢀ2). The Pt ECSA as
a function of the Pt loading determined by the integration of
the Hupd region and by the CO-stripping method are shown in
Figure 2b. In both cases, a decrease of the ECSA with the in-
crease of the Pt loading was observed. In an ideal case in
which the particle size is constant and no agglomerates are
formed, the ECSA (m2Pt gPtꢀ1) should be constant for different
loadings (mgPt cmꢀ2). Therefore, the present ECSA decrease for
higher Pt loadings indicates that with the increase of the Pt
coverage the Pt particles tend to form progressively larger ag-
glomerates. This finding is consistent with our initial model,
that is, the creation of Pt model electrodes that bridge from
isolated Pt nanoparticles to Pt agglomerates and finally to Pt
extended surfaces. Although the ECSA measured by the Hupd
and CO-stripping methods showed the same trend as a func-
tion of the Pt loading, it can be observed that particularly for
the low Pt loadings, the CO-stripping method led to higher
ECSA values. Different reasons could explain this finding. If the
Pt loading is decreased, more carbon surface is exposed and,
thus, the double-layer contribution from the carbon support
can become dominant in the Hupd region, which leads to
a large error in the calculation of the ECSA.[14] Similar results
were also reported by Mayrhofer et al.[16] who determined the
ECSA for various Pt catalysts by the Hupd and CO-stripping
methods; the authors attributed the lower ECSA measured by
Hupd to a large error in the
The cyclic voltammograms (CVs) recorded for each Pt load-
ing in Ar-saturated 0.1m HClO4 at scan rate of 50 mVsꢀ1 are
shown in Figure 2a. The CVs are characteristic of polycrystal-
line Pt, with distinct signals for the hydrogen adsorption/de-
sorption region between 0 and 0.35 V vs. RHE, a double-layer
region up to approximately 0.6 V vs. RHE and the signals for
platinum oxide formation/reduction at approximately 0.7–0.8 V
vs. RHE. From the region in which the hydrogen is deposited
at an under-potential (Hupd), the Pt ECSA was calculated by as-
suming a monolayer hydrogen charge of 210 mCcmꢀ2. The
ECSA of Pt catalysts was also evaluated by the CO-stripping
method; in this case the ECSA was determined through the ox-
idation of a monolayer of CO species adsorbed on the Pt sur-
double-layer correction. An addi-
tional effect can be expected
from the negative shift of the
potential of total zero charge on
going from extended surfaces to
small nanoparticles as previously
described by Mayrhofer et al.[17]
Clearly, one can expect the for-
mation of the Hupd layer to shift,
according to the shift of the po-
tential of total zero charge, to
negative potentials and only
a partial monolayer to adsorb in
the examined potential region.
We consider that the ECSA mea-
sured by the CO-stripping
method is the most accurate,
therefore, in the following the
specific catalytic activity was cal-
culated based on the ECSA de-
termined by CO stripping; fur-
thermore, any relationship be-
tween Pt activity and ECSA
refers to the ECSA measured by
CO stripping.
Apart from the decrease in
Figure 2. a) CVs measured for different Pt loadings at scan rate of 50 mVsꢀ1, RT, Ar-saturated 0.1m HClO4; b) Cor-
relation between Pt loading and the ECSA measured by the integration of the Hupd region and by the CO-strip-
ping method; c) CO-stripping voltammograms for different Pt loadings (scan rate 20 mVsꢀ1, Ar-saturated 0.1m
HClO4 after CO adsorption at 0.1 V vs. RHE for 15 min in CO-saturated 0.1m HClO4); d) potential shift for the plati-
num oxide reduction peak for different Pt loadings extrapolated from the CVs (scan rate 50 mVsꢀ1, Ar-saturated
0.1m HClO4).
ECSA with the increase of the Pt
loading, a further insight into
the change in morphology for
the Pt model electrodes is given
by the analysis of the CO-strip-
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1412

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
ping peak shown in Figure 2c.
Although for low-loading sam-
ples a single CO-stripping peak
is observed, a double CO oxida-
tion peak appeared for the
50 mgPt cmꢀ2 catalyst. A double
CO oxidation voltammetric peak
has already been reported for
other Pt catalysts and it was as-
sociated with the CO oxidation
reaction that occurs on Pt ag-
glomerates and on isolated par-
ticles.[18] The more negative and
the more positive peaks corre-
spond to CO oxidation on Pt ag-
Figure 3. a) ORR curves and b) Tafel plots obtained by the Ohmic and mass-transport-corrected ORR curves mea-
sured on a negative-going scan for different Pt loadings (scan rate 5 mVsꢀ1, RT, 1600 rpm, O2-saturated 0.1m
HClO4). The colour code is valid for a) and b).
glomerates and on isolated particles, respectively, which indi-
cates that Pt agglomerates/surface show remarkably enhanced
catalytic activity for CO stripping in comparison to isolated Pt
nanoparticles. In the present case, the more negative CO oxi-
dation peak was significantly more pronounced than the one
at higher over-potential, which suggests a Pt morphology that
consists of an extended particulate layer with minor Pt nano-
particles on top of it. Furthermore, the most intense CO oxida-
tion peak for the 50 mgPt cmꢀ2 sample occurred at similar po-
tential to that reported in the literature for Pt extended surface
catalysts,[18b,c] which confirms that an extended Pt surface was
formed mostly. The CO oxidation peak is not only a fingerprint
of the Pt morphology (isolated particles vs. agglomerates), but
the reaction potential is also particle-size dependent.[18a,b] The
CO oxidation peak potential for the 2 mgPt cmꢀ2 density was at
approximately 0.805 V vs. RHE, consistent with isolated Pt par-
ticles of 2–3 nm in diameter.[18b,c] Therefore, in the present
study the CO-stripping voltammetry could be used not only to
determine the catalyst ECSA, but also as a fingerprint of the
particle size and the extent of particle agglomeration, that is,
to discriminate between isolated particles and extended Pt sur-
faces.
that isolated nanoparticles indeed show a higher adsorption
energy of oxygenated species compared to extended surfaces.
The ORR polarisation curves at 1600 rpm on a negative-
going potential scan for the different Pt catalysts are plotted in
Figure 3a. A typical mixed mass transport/kinetically mixed
region at low over-potential and a mass-transport-limited
region at higher over-potential were observed for all the sam-
ples. If we compare the ORR curves of the different samples
we can observe that: a) for Pt loadings <20 mgPt cmꢀ2 the
mass-transport-limited current starts to gradually decrease
with decreasing Pt loading and b) the potential for the transi-
tion from the mixed kinetic/mass-transport-limited region and
the purely mass-transport-limited region shifted to a lower po-
tential on decreasing the Pt loading. As discussed above, on
decreasing the Pt loading, a transition from a Pt particulate
layer to isolated particles and, therefore, a decrease in the Pt
coverage occurred. The decrease in the mass-transport-limited
current with decreasing Pt coverage is consistent with the ORR
curves reported for Pt nanoparticles with different interparticle
distances[11,12] or different loadings.[13,16,20] Most authors attrib-
uted the lower value of the mass-transport-limited current
compared to the theoretical one to the impracticality for low-
loading catalysts to fully cover the support geometric area and
thus to the occurrence of oxygen diffusion along the glassy
carbon substrate. However, it must be clear that the diffusion-
limited current is not a function of the Pt loading (by definition
this is not the case), but of the Pt particle coverage and distri-
bution, that is, one effectively over-estimates the geometric
surface area in the case of an isolated particle catalyst. The re-
duction of the mass-transport-limited current was also ob-
served for Pt/carbon catalysts with increasing Nafion content
because of an increase of the mass transport resistance in the
Nafion film.[21] However, in the present case Nafion was only
used as binder for the Vulcan layer support, and the Pt catalyst
surface should in principle be mostly free of Nafion. At poten-
tials more cathodic than 0.1 V vs. RHE, all the samples showed
a decrease in the ORR current, which is more pronounced for
the low Pt loadings. For single-crystal electrodes, the reduction
of the mass-transport-limited current at potentials more catho-
dic than 0.3 V was ascribed to a an increase in the Hupd cover-
age, which leads to the production of a larger amount of H2O2
during the ORR.[22] The kinetic current density for the different
Changes in the Pt morphology are expected to influence its
electrochemical properties; this is in first instance indicated in
Figure 2d, which shows a shift of the platinum oxide reduction
peak (extracted from the CV in Ar-saturated electrolyte) as
a function of the Pt loading. Particularly, the platinum oxide re-
duction peak showed a constant increase in potential as the Pt
loading was increased, that is, as the ECSA decreased. Previous
studies have also reported a change of the platinum oxide re-
duction peak and of the OHads coverage with the variation of
the Pt particle size.[4,8,17,18] Furthermore, DFT studies have con-
firmed that the adsorption of oxygenated species on Pt nano-
particles differs significantly compared to that on bulk surfaces,
with an increase in the adsorption energy as the particle size is
decreased.[19] However, in the present case the primary particle
size remained constant, but the Pt distribution changed from
isolated nanoparticles to an extended particulate layer. There-
fore, the shift to more negative potential must be interpreted
as a change in the OHads adsorption properties on moving
from extended surfaces to isolated nanoparticles and indicates
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1413

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
x
Pt loadings was calculated according to Equation (1):
i ¼ nFKcO ð1 ꢀ VadsÞ expðꢀaFE=RTÞ expðꢀgrVads=RTÞ
ð2Þ
2
iꢀ1 ¼ id þ ik
ꢀ1
ꢀ1
ð1Þ
in which F, x, a, g and R are constants, n is the number of
electrons, cO is the oxygen concentration in the solution, K is
2
the chemical rate constant, E is the applied potential, rVads is
a parameter that characterises the rate of change of DGad with
surface coverage and Vads is the fraction of the electrode sur-
face sites covered by all the adsorbed species.[2a] From Equa-
tion (2) it is evident that an increase in the OHads coverage
(Vads) would lead to a decrease in the ORR current, as observed
in the present study on moving from an extended surface to
isolated particles.
in which id is the mass-transport-limited current density and
ik represents the kinetic current density (Figure 3b). If we con-
sider the kinetic current density at 0.9 V vs. RHE, Figure 3b
shows that ik increased with the increase of the Pt loading and
the decrease in ECSA, that is, from isolated nanoparticles to ex-
tended Pt layers.
Mass and specific activity for the different samples are plot-
ted as a function of the ECSA in Figure 4. The mass activity in-
The increase in mass activity
and the decrease of the specific
activity with increasing ECSA ob-
served in this work can be corre-
lated to a master curve obtained
by normalising the values from
different studies[7,14,17,25,26] to an
arbitrary Pt specific surface area
(35 m2Pt gPtꢀ1). This arbitrary nor-
malisation helps to compare re-
sults acquired using different ex-
perimental
protocols
(e.g.,
Ohmic drop correction, values
from anodic or cathodic sweeps,
capacitive current corrections,
etc.) and with different morphol-
ogies. Mass and specific activity
Figure 4. Correlation between Pt ECSA and a) mass and b) specific ORR activity calculated from the kinetic data at
0.9 V vs. RHE (shown in Figure 3b).
creased only slightly for ECSA below 40 m2Pt gPt
(10 mgPt cmꢀ2), whereas an almost double value was observed
for the largest ECSA sample. This finding can be explained in
terms of Pt dispersion; for catalysts made of relatively thick
layers or agglomerates, a large amount of bulk Pt does not
participate in the ORR and, therefore, the mass activity is lower
compared to highly dispersed nanoparticles.
from different Pt catalysts (which includes the data from the
present work) all follow a master curve that shows an increase
in mass activity and a decrease in specific activity with increas-
ing ECSA (Figure 5). An increase in ECSA can result either from
a decrease in particle size or in a change in Pt morphology,
that is, a transition from a Pt extended surface to Pt dispersed
nanoparticles as in the present study. In both cases, an in-
crease in the energy of adsorption of oxygenated species
might explain the decrease in specific activity shown in Fig-
ure 5b for several Pt catalysts. The experimental master curves
shown in Figure 5 closely resemble the plot reported by Tritsa-
ris et al.[28] for normalised Pt specific activities versus the ECSA,
determined by a mathematical model.
ꢀ1
The decrease in specific activity with the increase of the
ECSA is consistent with previous studies that report a change
of the specific activity with the interparticle distance/load-
ing.[10,14] This finding can be correlated to the variation in the
platinum oxide reduction peak potential as a function of the
Pt loading/ECSA. For the lowest loading, the highly dispersed
Pt nanoparticles remained oxidised at more cathodic potentials
compared to Pt catalysts made of large agglomerates or ex-
tended layers. Therefore, for isolated nanoparticles, the larger
surface coverage of OHads species results in the blockage of
more active sites and, therefore, in a lower specific catalytic ac-
tivity. The role of adsorbates on the specific activity of Pt-
based catalysts has been investigated widely. Adsorbates not
only block active Pt sites for the ORR but they can also alter
the adsorption energy of ORR intermediates.[23] Given that the
first electron transfer to the adsorbed oxygen is the rate-deter-
mining step, for a first-order dependence of the kinetics of the
ORR, the current density (i) has been proposed[2a] to be deter-
mined either by the free Pt sites available for the adsorption of
O2 (1ꢀVads) and/or by the change of Gibbs energy of adsorp-
tion of reactive intermediates (DGad):
Apart from the variation of the specific and mass activity as
a function of the ECSA, the influence of the Pt distribution on
the ORR selectivity has been investigated by using the RRDE
technique. RRDE measurements (ring held at 1.2 V vs. RHE) for
the catalysts with different ECSAs are shown in Figure 6. Larger
ring currents, which account for the oxidation of the H2O2 spe-
cies produced on the working electrode, were observed with
decreasing Pt loadings, that is, increasing ECSAs. Interestingly,
a sharp increase in the ring current was observed for all the
samples at potentials more cathodic than ~0.1 V vs. RHE in
correspondence with the decrease in the ORR current as dis-
cussed above.
To better compare the H2O2 formation as a function of the
ECSA, the mole fraction [%] of H2O2 species produced (xH O
)
2
2
was determined by the evaluation of the disk working elec-
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1414

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Figure 5. a) Mass and b) specific activity towards ORR at 0.9 V vs. RHE for different Pt catalysts normalised to the respective current values at 35 m2 gꢀ1Pt. Be-
sides data from the present work, ORR activities are extracted from RDE measurements by Schwanitz et al.,[14] Schmidt et al.,[24] Gasteiger et al.,[25] Nesselberg-
er et al.,[7b] Mayrhofer et al.,[17] Sheng et al.,[7a] Antoine et al.[26] and Stevenson and Pattrick.[27]
Figure 6. Bottom panels: ORR currents measured by scanning the disk electrode cathodically at 5 mVsꢀ1 and 1600 rpm. Top panels: H2O2 oxidation currents
recorded for the Pt ring held at 1.2 V vs. RHE in O2-saturated 0.1m HClO4 for loadings of a) 2, b) 10, c) 20 and d) 50 mgPt cmꢀ2
.
trode current (Id), the ring current (Ir) and its collection efficien-
cy (N) according to Equation (3):[2e]
with ECSA lower than 26 m2Pt gPtꢀ1, whereas greater H2O2 for-
mation is detected with increased ECSA, which reaches 30%
for the catalyst with the highest ECSA at 0.1 V vs. RHE. Even at
higher potentials (0.1–0.6 V vs. RHE), the H2O2 formation
ranges between 30 and 10% for the catalyst made of dis-
persed nanoparticles. The present results show clearly that
modification of the Pt distribution affects the ORR selectivity;
on moving from particulate layers and agglomerates to highly
dispersed nanoparticles, the ORR process does not proceed
fully by a four-electrons process but results in the parallel for-
mation of water and H2O2. In view of the present results, the
2Ir=N
ð3Þ
xH O ½%ꢂ ¼ 100
2
2
Id þ Ir=N
The analysis of the RRDE data is summarised in Figure 7,
which shows the H2O2 formation as a function of the applied
potential for different ECSA catalysts (Figure 7a) and of the
ECSA for selected applied potentials (Figure 7b). No significant
differences in the H2O2 formation can be observed for catalysts
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1415

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
For the highest ECSA catalyst,
the increase in H2O2 formation
occurs at higher potentials than
those typical for the Hupd region
(Figure 7); therefore, the shift
from a four- to a two-electron
process cannot be ascribed to
only a change in the Hupd cover-
age. It has been reported recent-
ly that decreasing the Pt loading
leads to a higher coverage of
anionic impurities, which inhibits
the H2O2 reduction and enhan-
ces the desorption of H2O2 inter-
mediates at potentials higher
than those typical for the Hupd
region. Indeed, as described in
Figure 7. Mole fraction of produced H2O2 determined by RRDE measurements as a function a) of the applied po-
tential for different ECSA catalysts and b) of the ECSA for selected applied potentials.
Equation (2), adsorbate spectators can alter the adsorption
energy of the ORR intermediates.[23] However, spectators other
than anionic impurities can contribute to H2O2 desorption,
such as, for example, oxygenated species. As discussed above,
moving from particulate layers to isolated nanoparticles results
in an increased adsorption energy of oxygenated species. As
reported in Ref. [14], for low Pt loadings the OH coverage on
the Pt surface is extended down to a potential of 0.5 V vs. RHE.
Therefore, a change in the OH coverage for isolated nanoparti-
cles can also contribute to a more pronounced H2O2 desorp-
tion. Furthermore, in the case of a highly dispersed Pt nano-
particle catalyst, the probability that desorbed H2O2 is re-ad-
sorbed and then further reduced to water is definitively lower
than in the case of a particulate layer. As shown in the sketch
in Figure 9a, in the case of highly dispersed nanoparticles, the
desorbed H2O2 might have an equal probability to be re-ad-
sorbed and further reduced on a nearby Pt particle (case 1) or
move out of the diffusion layer before it reaches an active Pt
site (case 2). Differently, in the case of a continuous Pt particu-
late layer (Figure 9b), H2O2 intermediates can be i) further re-
duced on the same Pt site, ii) diffuse on the surface and then
be reduced or iii) if desorbed, easily find a Pt active site in the
proximity on which it can re-adsorb and be further reduced to
water.
Figure 8. Possible reaction scheme for the ORR on Pt-based catalysts.[2a]
large H2O2 formation at potentials below 0.6 V vs. RHE for the
2 mgPt cmꢀ2 catalyst loading could mostly account for the sig-
nificantly lower mass-transport-limited current compared to
the theoretical value (Figure 3a).
A simplified reaction scheme for the ORR is presented in Fig-
ure 8:[2a] according to reaction k1 a “direct” pathway that in-
volves a four-electron transfer occurs to describe oxygen re-
duction to water without the formation of H2O2 intermediates.
Reaction k2 describes an electrochemical reduction of oxygen
in which only two electrons are involved to form H2O2 prod-
ucts. The H2O2 can be then further reduced to H2O by two ad-
ditional electrons (k3). Alternatively, the H2O2 can dispropor-
tionate chemically according to k4 or can be desorbed (k5). If
reactions k2 and k3 occur consecutively, the ORR is described to
proceed through a “serial” pathway that involves overall four
electrons (note that RRDE cannot distinguish between the k1
and k2+k3 reaction pathways unambiguously).
The results shown in Figure 7 provide clear evidence of an
ORR selectivity dependence on the Pt particle distribution/
ECSA. Although a significant amount of H2O2 was detected for
dispersed Pt nanoparticles, continuous Pt layers or agglomer-
ates have shown small amounts of dissolved H2O2 during the
ORR. This result does not imply that for an extended Pt surface
the ORR mechanism only proceeds through a “direct” pathway
(k1) but rather that the H2O2 produced can be easily further re-
duced on Pt layers or agglomerates. As shown in Ref. [29], Pt
activity towards H2O2 reduction is extremely high below 0.8 vs.
RHE. However, Ref. [29] reports the catalytic activity towards
H2O2 reduction for Pt polycrystalline electrodes, which might
be different to the catalytic activity of particulate layers or iso-
lated nanoparticles as studied in this work.
Conclusions
We have shown a change in the activity and selectivity of
2–3 nm Pt particles if the catalyst distribution changes from
highly dispersed nanoparticles to extended particulate layers.
We correlated the decrease in specific activity for the samples
with a higher electrochemical surface area (ECSA) to the in-
creased adsorption energy of the oxygenated species on
moving from extended surfaces to isolated particles. The re-
sults indicate that the OHads coverage is also a function of the
Pt nanoparticles distribution, that is, highly dispersed nanopar-
ticles or particle agglomerates. Consequently, the specific
oxygen reduction reaction (ORR) activity also depends on the
Pt distribution/ECSA.
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1416

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
ergies were scanned from 2.0–4.0 eV in 0.1 eV intervals.
For the data analysis, a stratified layer model was as-
sumed that consisted of Si substrate/SiO2 interlayer/Pt
film. The model electrodes were also characterised by
SEM and EDX spectroscopy. TEM (FEI Morgagni 268,
120 kV, image acquisition in bright field) images were ac-
quired for Pt nanoparticles deposited as described above
on amorphous carbon film TEM grids. The electrochemi-
cal characterisation was performed by using a three-elec-
trode compartment electrochemical cell in 0.1m HClO4
electrolyte using a hydrogen reference electrode and
a Pt mesh as the counter electrode at ambient tempera-
ture and pressure. The hydrogen electrode was calibrat-
ed by the hydrogen oxidation/evolution reaction in the
same electrolyte. The CV curves were acquired in Ar-sa-
turated electrolyte at 50 mVsꢀ1. For the CO-stripping ex-
periments, CO adsorption was achieved by holding the
potential at 0.1 V vs. RHE in a CO-saturated solution for
15 min; then the electrolyte was purged with Ar (40 min)
to remove the dissolved CO, and CVs at 20 mVsꢀ1 were
measured. RRDE measurements were performed by scan-
ning the disk electrode cathodically between 1–0.05 V at
5 mVsꢀ1 whilst the Pt ring electrode was held at 1.2 V
(vs. RHE); the resulting disk polarisation curves were cor-
rected by the Ohmic drop in the electrolyte measured
by electrochemical impedance spectroscopy. The collec-
Figure 9. Simplified reaction scheme for the ORR that occurs through a “serial” pathway
a) on highly dispersed Pt nanoparticles and b) on an extended Pt particulate layer.
At over-potentials higher than those typical for the calcula-
tion efficiency of the Pt ring was 0.2ꢁ0.02 and was determined ac-
cording to Ref. [21b].
tion of specific activity, the Pt distribution still plays an impor-
tant role to determine the selectivity of the catalyst towards
the ORR. Dispersed nanoparticles lead to considerable H2O2
formation, probably because the large amount of desorbed
H2O2 species is unlikely to find a Pt nanoparticle on which it
can be re-adsorbed and reduced to water.
Acknowledgements
The authors wish to acknowledge M. Stevenson and G. Pattrick
for providing the data for Figure 5. T.J.S. wishes to thank N. M.
Markovic (Argonne National Laboratory) and P. N. Ross (Law-
rence Berkeley National Laboratory) for making some previously
unpublished results accessible for Figure 5. S.T., P.L. and O.C.
thank the South African Department of Science and Technology
for financial support in the form of HySA/Catalysis Centre of
Competence programme funding (O.C., P.L.) and a HySA/Catalysis
student bursary (S.T.). S.T. further thanks the PSI for research visit
funding.
If we assume that the mechanism for the ORR does not
change with the Pt distribution, this result indicates that the
ORR occurs in a large extent through a serial pathway in which
O2 is first reduced to H2O2 that is then further reduced to
water. The understanding of H2O2 formation is of fundamental
importance not only from a mechanistic point of view, but also
from an applied point of view because H2O2 ions are consid-
ered to be very aggressive towards the polymer electrolyte
fuel cell membrane and the catalysts.
Keywords: electrochemistry · fuel cells · oxygen · platinum ·
reaction mechanisms
Experimental Section
The preparation of the model electrodes was performed in two
steps. At first, a Vulcan suspension made from Vulcan XC72
(75 mg), Nafion 117 solution (100 mL, 5%, Fluka) and 2-propanol
(25 mL Normapur, VWR Prolabo) was prepared. The suspension
was treated ultrasonically at RT for 30 min. The suspension (20 mL)
was dispersed onto a mirror-polished glassy carbon disk electrode
and dried overnight at RT in an Ar atmosphere. In the second step
of the electrode preparation, different Pt loadings (2, 10, 20 and
50 mgPt cmꢀ2) were sputtered onto the Vulcan-based electrodes in
a magnetron sputtering set-up according to the procedure in
Ref. [14]. XPS measurements were performed by using a VG ESCA-
LAB 220iXL spectrometer (Thermo Fischer Scientific) equipped with
an AlKa monochromatic source (spot size: 500 mm, power: 150 W)
and a magnetic lens system. Ellipsometry was performed by using
a spectroscopic ellipsometer (MOSS model ES 4G, Sopra, France) at
an angle of incidence of 708. In our measurements the photon en-
[1] A. Rabis, P. Rodriguez, T. J. Schmidt, ACS Catal. 2012, 2, 864–890.
´
´
[2] a) N. M. Markovic, T. J. Schmidt, V. Stamenkovic, P. N. Ross, Fuel Cells
2001, 1, 105–116; b) T. J. Schmidt, B. N. Grgur, R. J. Behm, N. M. Markov-
´
ic, P. N. Ross, Phys. Chem. Chem. Phys. 2000, 2, 4379–4386; c) T. J.
´
Schmidt, V. Stamenkovic, M. Arenz, N. M. Markovic, P. N. Ross, Electro-
´
chim. Acta 2002, 47, 3765–3776; d) V. Stamenkovic, T. J. Schmidt, P. N.
´
Ross, N. M. Markovic, J. Phys. Chem. B 2002, 106, 11970–11979; e) T. J.
´
´
Schmidt, V. Stamenkovic, P. N. Ross, N. M. Markovic, Phys. Chem. Chem.
´
Phys. 2003, 5, 400–406; f) V. R. Stamenkovic, B. Fowler, B. S. Mun, G. F.
´
Wang, P. N. Ross, C. A. Lucas, N. M. Markovic, Science 2007, 315, 493–
497.
[3] a) K. F. Blurton, D. R. Rutt, H. G. Oswin, P. Greenber, J. Electrochem. Soc.
1972, 119, 559; b) L. J. Bregoli, Electrochim. Acta 1978, 23, 489–492.
[4] M. Peuckert, T. Yoneda, R. A. D. Betta, M. Boudart, J. Electrochem. Soc.
1986, 133, 944–947.
[5] K. Kinoshita, J. Electrochem. Soc. 1990, 137, 845–848.
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1417

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
[6] a) M. H. Shao, A. Peles, K. Shoemaker, Nano Lett. 2011, 11, 3714–3719;
b) Z. Xu, H. M. Zhang, H. X. Zhong, Q. H. Lu, Y. F. Wang, D. S. Su, Appl.
Catal. B 2012, 111, 264–270; c) F. J. Perez-Alonso, D. N. McCarthy, A.
Nierhoff, P. Hernandez-Fernandez, C. Strebel, I. E. L. Stephens, J. H. Niel-
sen, I. Chorkendorff, Angew. Chem. Int. Ed. 2012, 51, 4641–4643; Angew.
Chem. 2012, 124, 4719–4721.
[7] a) W. C. Sheng, S. Chen, E. Vescovo, Y. Shao-Horn, J. Electrochem. Soc.
2012, 159, B96–B103; b) M. Nesselberger, S. Ashton, J. C. Meier, I. Kat-
sounaros, K. J. J. Mayrhofer, M. Arenz, J. Am. Chem. Soc. 2011, 133,
17428–17433.
Weinkauf, U. Stimming, Phys. Chem. Chem. Phys. 2005, 7, 385–393; c) S.
Kunz, K. Hartl, M. Nesselberger, F. F. Schweinberger, G. Kwon, M. Hanzlik,
K. J. J. Mayrhofer, U. Heiz, M. Arenz, Phys. Chem. Chem. Phys. 2010, 12,
10288–10291.
[19] B. C. Han, C. R. Miranda, G. Ceder, Phys. Rev. B 2008, 77, 075410.
[20] I. Katsounaros, J. C. Meier, K. J. J. Mayrhofer, Electrochim. Acta 2013, 110,
790–795.
[21] a) E. Higuchi, H. Uchida, M. Watanabe, J. Electroanal. Chem. 2005, 583,
69–76; b) U. A. Paulus, T. J. Schmidt, H. A. Gasteiger, R. J. Behm, J. Elec-
troanal. Chem. 2001, 495, 134–145; c) T. J. Schmidt, H. A. Gasteiger,
G. D. Stab, P. M. Urban, D. M. Kolb, R. J. Behm, J. Electrochem. Soc. 1998,
145, 2354–2358.
[8] A. Gamez, D. Richard, P. Gallezot, F. Gloaguen, R. Faure, R. Durand, Elec-
trochim. Acta 1996, 41, 307–314.
´
[9] M. Watanabe, H. Sei, P. Stonehart, J. Electroanal. Chem. 1989, 261, 375–
387.
[22] N. M. Markovic, H. A. Gasteiger, P. N. Ross, J. Phys. Chem-Us 1995, 99,
3411–3415.
´
[10] M. Nesselberger, M. Roefzaad, R. FayÅal Hamou, P. U. Biedermann, F. F.
Schweinberger, S. Kunz, K. Schloegl, G. K. H. Wiberg, S. Ashton, U. Heiz,
K. J. J. Mayrhofer, M. Arenz, Nat. Mater. 2013, 12, 919–924.
[11] H. Z. Yang, S. Kumar, S. Z. Zou, J. Electroanal. Chem. 2013, 688, 180–188.
[12] Y. E. Seidel, A. Schneider, Z. Jusys, B. Wickman, B. Kasemo, R. J. Behm,
Faraday Discuss. 2009, 140, 167–184.
[13] M. Inaba, H. Yamada, J. Tokunaga, A. Tasaka, Electrochem. Solid-State
Lett. 2004, 7, A474–A476.
[14] B. Schwanitz, A. Rabis, M. Horisberger, G. G. Scherer, T. J. Schmidt,
Chimia 2012, 66, 110–119.
[23] N. M. Markovic, H. A. Gasteiger, B. N. Grgur, P. N. Ross, J. Electroanal.
Chem. 1999, 467, 157–163.
[24] T. J. Schmidt, N. M. Markovic, P. N. Ross, unpublished results.
´
[25] H. A. Gasteiger, S. S. Kocha, B. Sompalli, F. T. Wagner, Appl. Catal. B 2005,
56, 9–35.
[26] O. Antoine, Y. Bultel, R. Durand, J. Electroanal. Chem. 2001, 499, 85–94.
[27] M. Stevenson, G. Pattrick, unpublished results.
[28] G. A. Tritsaris, J. Greeley, J. Rossmeisl, J. K. Norskov, Catal. Lett. 2011,
141, 909–913.
[29] I. Katsounaros, W. B. Schneider, J. C. Meier, U. Benedikt, P. U. Bieder-
mann, A. A. Auer, K. J. J. Mayrhofer, Phys. Chem. Chem. Phys. 2012, 14,
7384–7391.
[15] B. Schwanitz, H. Schulenburg, M. Horisberger, A. Wokaun, G. G. Scherer,
Electrocatalysis 2011, 2, 35–41.
´
[16] K. J. J. Mayrhofer, D. Strmcnik, B. B. Blizanac, V. Stamenkovic, M. Arenz,
´
N. M. Markovic, Electrochim. Acta 2008, 53, 3181–3188.
´
[17] K. J. J. Mayrhofer, B. B. Blizanac, M. Arenz, V. R. Stamenkovic, P. N. Ross,
Received: November 17, 2013
Revised: December 20, 2013
´
N. M. Markovic, J. Phys. Chem. B 2005, 109, 14433–14440.
[18] a) O. V. Cherstiouk, P. A. Simonov, E. R. Savinova, Electrochim. Acta 2003,
48, 3851–3860; b) F. Maillard, S. Schreier, M. Hanzlik, E. R. Savinova, S.
Published online on February 23, 2014
ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 1410 – 1418 1418