J. Phys. Chem.

B 2002, 106, 4181-4191 4181

Oxygen Reduction on Carbon-Supported Pt-Ni and Pt-Co Alloy Catalysts

U. A. Paulus,* A. Wokaun, and G. G. Scherer

Paul Scherrer Institut, Laboratory for Electrochemistry, CH-5232 Villigen PSI, Switzerland

T. J. Schmidt, V. Stamenkovic, V. Radmilovic, N. M. Markovic, and P. N. Ross

Materials Sciences DiVision, Lawrence Berkeley National Laboratory, UniVersity of California,
Berkeley, California 94720
ReceiVed: September 10, 2001; In Final Form: January 23, 2002

We describe a comparative study of the oxygen reduction reaction on two carbon-supported Pt-based alloy
catalysts in aqueous acidic electrolyte at low temperature. Both alloys have the bulk compositions of 50 and
75 at. % Pt, with the alloying elements being Ni and Co. Comparison is made to a pure Pt catalyst on the
same carbon support, Vulcan XC-72, having the same metal loading (20 wt %) and nominally the same
particle size (4 ( 2 nm). High-resolution electron microscopy was used to determine the size and shape of
the particles as well as the particle size distribution on all catalysts. Electrochemical measurements were
performed using the thin-film rotating ring-disk electrode method in 0.1 M HClO4 at 20-60 °C. Hydrogen
adsorption pseudocapacitance was used to determine the number of Pt surface atoms and to estimate the
surface composition of the alloy catalysts. Kinetic analysis in comparison to pure Pt revealed a small activity
enhancement (per Pt surface atom) of ca. 1.5 for the 25 at. % Ni and Co catalysts, and a more significant
enhancement of a factor of 2-3 for the 50 at. % Co. The 50 at. % Ni catalyst was less active than the Pt
standard and unstable at oxygen electrode potentials. Ring-current collection measurements for peroxide
indicated no significant differences between the Pt-Co catalysts or the 25 at. % Ni catalyst and pure Pt,
while the 50 at. % Ni catalyst had a higher peroxide yield. Together with the observed Tafel slopes and
activation energies, it was concluded that the kinetic enhancement is contained in the preexponential factor
of the conventional transition state theory rate expression. It is, however, not clear why the alloying with Ni
or Co produces this change in the preexponential factor.

1. Introduction or Pt-Cr (75% Pt) in comparison to pure Pt. Both the latter
The understanding of the kinetic limitations of the oxygen studies also reported that the transition metal was lost from the
reduction reaction (ORR) on different catalysts is a problem of alloy during testing as an oxygen electrode in a PAFC. More
outstanding interest in fuel cell research. Carbon-supported recent half-cell experiments in 190 °C phosphoric acid reported
platinum catalysts still serve as the most widely used electro- enhancement factors of 2 to 5 when using Pt-Cr, Pt-Co or
catalyst in the air electrode in low-temperature polymer Pt-Ni versus pure Pt.10 In the case of testing in PEFCs,
electrolyte fuel cells (PEFCs).1 Kinetic limitations for the ORR Mukerjee et al.11-13 and Tamizhmani et al.14 found a similar
lead to cell voltage losses (versus the theoretical potential) of activity enhancement (factors of 2-3) when using Pt-Co, Pt-
at least 0.3 to 0.4 V under typical PEFC operating conditions.1 Ni or Pt-Cr versus Pt. Toda et al.15 reported enhancement
Ineffective electrode structures increase the voltage losses even factors of as much as 10 for sputtered films of Pt-Ni, Pt-Co
further. However, intensive engineering of the membrane and Pt-Fe in comparison to a similarly prepared pure Pt film
electrode assembly has removed most of those losses, resulting in 0.1 M HClO4 at room temperature. On the other hand, ORR
in the use of extremely thin catalyst layers with Pt loading as measurements in this laboratory with well-characterized
low as 0.12 mg/cm2.2,3 Further improvements in PEFC cell Pt-Ni and Pt-Co bulk alloy electrodes in aqueous acidic
voltage will require finding a catalyst more active than Pt for electrolyte at PEFC conditions again found much smaller
the ORR. A variety of binary and ternary Pt-alloy systems have enhancements (factors of 1.5-2) versus similarly well-
been discussed and investigated during the last twenty years characterized polycrystalline Pt.16
for use in either phosphoric acid fuel cells (PAFC) or polymer With the exception of the results by Toda et al.15 all of the
electrolyte fuel cells (PEFC). An overview with emphasis on reports of enhancement with transition metal alloying to Pt have
the patent literature was presented by Stonehart.4 The results been reported using carbon-supported catalysts. Studies using
are contradictory. For example, Luczak et al.5-7 claimed an well-characterized bulk alloys with nominally the same com-
increase of a factor of 1.5 to 2.5 in mass activity (mA/mg Pt at position failed to observe this enhancement. As pointed out
0.9 V) using Pt-Co-Cr or Pt-Cr “alloy” catalysts instead of previously by Beard and Ross,8 the comparison between alloy
pure Pt in PAFCs. On the other hand, Beard and Ross8 and catalysts and pure Pt must be done carefully, recognizing that
Glass et al.9 did not find an increased mass activity for Pt-Co the activity of the Pt and/or the alloy may depend on particle
* Corresponding author: Phone: (41) 56 310 20 59. Fax: (41) 56 310 size, and regardless of the normalization procedure one should
44 15. E-mail: [email protected]. compare results at or near the same particle size. That was
10.1021/jp013442l CCC: $22.00 © 2002 American Chemical Society
Published on Web 03/29/2002
4182 J. Phys. Chem. B, Vol. 106, No. 16, 2002 Paulus et al.

generally not the case in the previous studies cited above. We with a bi-potentiostat (Pine Instruments). For the collection
report here the results from a study of two different carbon- experiments, the Pt ring electrode was potentiostated at 1.2 V
supported alloy catalysts in comparison to pure Pt on the same versus the reversible hydrogen electrode (RHE) where the
support with the same metal loading (20 wt %) and close to detection of peroxide is diffusion limited. The collection
the same particle size, 4 ( 2 nm. High-resolution transmission efficiency, N, for these thin-films electrodes was determined to
electron microscopy (HRTEM) analysis provided information be 0.2.21 The reference electrode is a saturated calomel electrode
about the particle size and shape. Quantitative kinetic measure- (SCE), separated from the working electrode compartment by
ments for the ORR were made using the thin-film rotating ring- a closed electrolyte bridge in order to avoid chloride contamina-
disk electrode (RRDE) method.17 This technique allows the tion. The potentials in this study, however, all refer to that of
accurate determination of kinetic data such as Tafel slopes, the RHE.
reaction orders and (apparent) activation enthalpies in the The thin catalyst layers were prepared as described in our
absence of mass transport effects. Ring current measurements preceding publications.17,18 In short, glassy carbon disk elec-
provide the parallel determination of the product distribution trodes (Sigradur G, Hochtemperatur Werkstoffe GmbH, 6 mm
(H2O2 versus H2O formation) under PEFC relevant condi- diameter) polished to a mirror-finish prior to each experiment
tions18,19 and give insight to the reaction pathway. Finally, we (0.1 µm alumina, Buehler) served as substrates for the supported
compare our results of these supported alloys to results obtained catalysts. In each case 20 µL of ultrasonically re-dispersed
with carefully prepared smooth polycrystalline bulk alloys of catalyst suspension (1.0 mgcatalyst/mL) in triply pyrodistilled
the same stoichiometry.16 water was pipetted onto the glassy carbon substrate, resulting
in a constant metal loading of 14 µgmetal/cm2. After drying in
2. Experimental Section flowing high-purity argon at room temperature, the deposited
catalyst was then covered with 20 µL of a diluted aqueous
We used commercially available Pt-based fuel cell catalysts Nafion solution prepared similarly to the method described by
provided by E-TEK (Natick, MA), 20 wt % Pt supported on Denton et al.22 After drying, the final thickness of the layer,
Vulcan XC72 carbon black, and two alloy catalysts, also 20 wt measured with a profilometer, was typically ∼0.2 µm. After
% metal on Vulcan XC72 (Cabot) carbon black with overall preparation, electrodes were immersed under potential control
stoichiometries of 50 at. % Ni (Co) and 25 at. % Ni (Co). The (at 150 mV) in argon-purged electrolyte (Bay Gas Research
catalysts were all used as-received. X-ray diffraction analysis Purity). For the oxygen reduction experiments, the electrolyte
of the four alloy catalysts using a conventional powder was saturated with oxygen (Bay Gas Research Purity) bubbled
diffractometer (Siemens D-500) revealed only a single fcc phase through the electrolyte. The electrolyte itself, 0.1 M HClO4,
in each, with lattice constants of 0.365 nm (25 at. % Ni) and was prepared from 70% HClO4 (Baker Ultrex) with triply
0.375 nm (50 at. % Ni) for the Pt-Ni, and 0.384 nm (25 at. % pyrodistilled water.
Co) and 0.374 nm (50 at. % Co) for the Pt-Co catalyst. These
values are in reasonable agreement with the lattice constants 3. Results and Discussion
for the corresponding bulk solid solutions.20 Hereafter, we shall
refer to these two alloy catalysts as Pt3Ni (Co) and PtNi (Co), 3.1. Catalyst Characterization. The carbon-supported cata-
respectively, although it should be understood that these are lysts were characterized by HREM. Figures 1-5 present the
solid solutions and not ordered phases (or intermetallic com- HREM analysis of Pt, Pt3Ni, PtNi, Pt3Co, and PtCo, respectively.
pounds). Pt-Co does form ordered phases at 25 and 50 at. % Low-magnification images (part A) illustrate that the distribution
compositions, and the ordered phase Pt3Co was reported by of the metal particles on the carbon support is uniform and in
Beard and Ross8 in their carbon-supported Pt-Co catalyst heat- a narrow particle size range. The histograms of the particle size
treated at relatively high temperatures (e.g., >900 °C). distribution (part B), which include analyses of several different
2.1. High-Resolution Electron Microscopy (HREM). Speci- regions, reflect quantitatively the uniform distribution with an
mens were prepared for transmission electron microscopy by average mean particle size of 3 to 4 nm for all catalysts. It is
making suspensions of the catalyst powders in ethanol, using evident that the particle size distribution is slightly broader in
an ultrasonic bath. These suspensions were dropped onto clean case of the alloy catalysts with an enhanced fraction of particles
holey carbon grids, which were then dried in air. Samples were larger than 5 nm. The mean particle diameter, d, was calculated
examined using the OO2B and ARM high-resolution electron according to eq 1, and the surface averaged dispersion, D
microscopes at the National Center for Electron Microscopy (ratio of atoms in the particles that are on the surface to the
(NCEM), Berkeley. The particle shapes were determined by real- total number of atoms in the particles), was calculated from
space crystallography using high-resolution images taken from eq 2,
particles near or on the edge of the carbon black substrate.
Structural information on single particles was obtained by ∑i nidi
numerical Fourier filtering of the digitized image intensity d) (1)
spectrum. Dark-field imaging was also utilized in order to ni
investigate the overall distribution of metallic particles on the
carbon black.
2.2. Electrode Preparation and Electrochemical Measure-
∑
V i
nidi2
ments. Electrochemical measurements were conducted in a D)6 (2)
∑i nidi
thermostated standard three-compartment electrochemical cell s 3
with reference and counter electrodes in separate compartments
from the working electrode. The working electrode is a
commercial (Pine Instruments) ring-disk electrode with an where ni is the frequency of occurrence of particles of the size
insertable (interchangeable) disk, in this case glassy carbon di, V is the volume per metal atom in the bulk, and s is the
covered by a thin layer of catalyst imbedded in a recast Nafion average area occupied by a metal atom in the surface. The
polymer electrolyte film. Ring and disk potential were controlled prefactor 6 is valid for spherical particles. For determining the
O2 Reduction on C-Supported Pt-Ni and Pt-Co Catalysts J. Phys. Chem. B, Vol. 106, No. 16, 2002 4183

Figure 1. (a) Low-magnification HREM micrograph. (b) Particle size distribution determined by HREM micrographs. (c) HREM micrograph with
atomic resolution of the standard supported Pt catalyst.

TABLE 1: Comparison of Physical Properties of the geometric area of the electrode and the dispersion and is the
Different Investigated Catalysts number of surface metal atoms per unit area of electrode. If the
catalyst mean particle diameter d (nm) dispersion D (%) surface composition of the metallic particle is known, then the
Pt/Vulcan 3.4 ( 1.2 30 absolute number of Pt surface atoms per unit area electrode can
PtNi/Vulcan 4.1 ( 1.2 23 be obtained, as shown in column 4 of Table 2. This is discussed
Pt3Ni/Vulcan 4.0 ( 0.9 25 further in the next section.
PtCo/Vulcan 4.4 ( 2.6 19
Pt3Co/Vulcan 3.4 ( 0.7 29 Elemental distributions were also obtained using X-ray
emission spectroscopy. A Ni(Co) map image at low magnifica-
particle size distribution from low-magnification images, all tion (as in part A) obtained via the X-ray emission energy
particles were considered to be spherical, but in fact, the most
filtering imaging technique (data not shown) indicated that every
frequent particle shape is cubooctahedral. These resulting values
particle contains both Pt and Ni(Co) atoms. Thus, we can
of d and D are summarized in Table 1. According to these data,
exclude the presence of unalloyed domains of Pt or Ni(Co)
the alloy particles are roughly 20% larger than the pure platinum
particles. Table 2 summarizes several other measures of the nanoclusters on the supported catalysts. Quantitative analysis
distribution of metal atoms in the catalysts. The first (column performed on several particles revealed compositions of 75 (
1) is the metal atom density per geometric electrode area. Due 5% platinum for the Pt3Ni(Co) catalyst and 50 ( 5% platinum
to the substitution of Pt by the smaller and lighter Ni (Co) atoms, for the PtNi(Co) catalyst, respectively. This was a reasonable,
equal metal loading means higher densities of metal atoms per but not definitive, indication that all metallic particles have the
geometric area using the alloy catalysts. The surface atomic same composition as the bulk stoichiometry, i.e., Pt:Ni(Co) )
density (column two) is calculated using the atomic density per 3:1 and 1:1.

TABLE 2: Comparison of Theoretical Atomic Densities and the Experimentally Determined Active Surface Area
metal atom density metal atom surface charge Hupd-region active (Pt) surface estimated surface
catalyst (nmol/cm2) density (nmol/cm2) (mC/cm2) atoms (nmol/cm2) composition (%)
Pt/Vulcan 72.4 21.7 1.97 20.4 (100)
PtNi/Vulcan 111.4 25.6 0.5 5.0 20
Pt3Ni/Vulcan 87.8 22.0 1.5 15.8 70
PtCo/Vulcan 111.4 21.2 0.72 7.5 35
Pt3Co/Vulcan 87.8 25.2 1.4 14.5 58
4184 J. Phys. Chem. B, Vol. 106, No. 16, 2002 Paulus et al.

Figure 2. (a) Low-magnification HREM micrograph. (b) Particle size distribution determined by HREM micrographs. (c) HREM micrograph with
atomic resolution of the Pt-25 at. % Ni (Pt3Ni) catalyst.

Finally part C of Figures 1-5 show a high-magnification featureless, and none of the Hupd peaks typical for polycrystalline
HREM micrograph along with a microdiffraction pattern from platinum are resolved.
a single particle in each of the three catalysts. Some twinned To determine the electrochemically active surface area, the
particles were observed in all five catalysts, even in the pure Hupd region of the different base voltammograms was integrated
Pt. The Pt-Ni catalysts had the highest number of twinned applying the procedure described in ref 17. The obtained Hupd
particles, but in all cases twinned particles were relatively rare. charges are listed in Table 2 (column 3). It is known that in
The most frequent structure of a typical Pt particle shows a case of pure Pt, the number of surface atoms can be obtained
cubooctahedral form, while the two alloy catalysts, however, from the charge in the Hupd region.23 The calculated number of
both revealed a more spherical form. electrochemically active surface atoms is in excellent agreement
3.2. Electrochemical Measurement. 3.2.1. Cyclic Voltam- with the value calculated from electron microscopy, i.e., the
metry. The catalysts were first characterized electrochemically particle size distribution. If we assume that alloying with
by cyclic voltammetry. Therefore electrodes were cycled several Ni(Co) does not effect the binding of Hupd to the Pt surface
atoms, we can again estimate the number of platinum surface
times between 0 and 0.9 V at room temperature in an argon
atoms present on the alloy particles from the determined Hupd
purged electrolyte. The resulting voltammograms are presented
charge (note that this method was proven to be well suitable
in Figure 6 for the supported Pt, PtNi and Pt3Ni catalysts, and
on UHV prepared well-defined Pt3M bulk electrodes24). The
in Figure 7 for the Pt, PtCo, and Pt3Co catalysts. Except for the number of active surface atoms obtained thereby (column 4)
PtNi catalyst, no changes in the voltammograms were observed allows the estimation of the particle surface composition
during this initial cycling procedure. Aside from this difference, (column 5). For the Pt3Ni(Co) catalyst, we find 70% Pt on the
the base voltammetry of the Ni and Co catalysts are essentially surface of the Pt3Ni particles and 58% Pt on the surface of the
identical, and in the following the features are discussed as Pt3Co particles. The value for Pt3Ni is very close to the bulk
related to both. The area in the hydrogen underpotential composition of 75 at. % Pt and indicates that no segregation
deposition (Hupd) region, i.e., the adsorption pseudocapacitance, has taken place, whereas in case of the Pt3Co a slight surface
decreases with decreasing Pt-content. The shape of the Hupd segregation of Co is observed. For the PtNi catalyst, however,
region of the 75% Pt catalyst is very similar to that of the Pt we find only about 20 at. % surface Pt and for the PtCo catalyst
catalyst and shows the same typical H-adsorption and desorption 35% surface Pt, indicating Ni(Co) segregation to the surface.
features in the potential region between 50 and 300 mV. However, unlike the PtCo catalyst, the base voltammetry of PtNi
However, the Hupd region of the 50% Pt catalyst is nearly changes substantially over the whole period of measurement
O2 Reduction on C-Supported Pt-Ni and Pt-Co Catalysts J. Phys. Chem. B, Vol. 106, No. 16, 2002 4185

Figure 3. (a) Low-magnification HREM micrograph. (b) Particle size distribution determined by HREM micrographs. (c) HREM micrograph with
atomic resolution of the Pt-50 at. % Ni (PtNi) catalyst.

(including the measurement of ORR kinetics), particularly after be observed in the Hupd potential region. This is a clear indication
measurements above room temperature. The nature of these of enhanced peroxide production on both the Pt and the alloy
changes is that the base voltammetry becomes more “Pt-like”, catalysts in the Hupd region. Qualitatively similar results were
probably indicative of leaching of Ni from the surface. The reported previously for Pt single-crystal electrodes, and the full
instability of the PtNi catalyst in the ORR environment makes discussion of the change in the ORR pathway in this potential
interpretation of the results with this catalyst problematic. region is presented there.25 Presumably the same reaction
Finally, from the base voltammetry we note that the onset of mechanism applies to these supported Pt-rich nanoparticles; i.e.,
Hupd blocks the Pt ensembles needed to break the O-O bond
oxide formation shifts to more positive potentials with increasing
leading to peroxide production.
Ni or Co content. For pure Pt, oxide formation in the anodic
going sweep starts at around 700 mV, whereas on Pt3Ni(Co) The PtNi catalyst produced significantly different results (not
the onset can be observed at around 750 mV, and on PtNi(Co) shown). The ring and disk current curves look qualitatively
between 750 and 800 mV. This may indicate that alloying with similar but are shifted in potential. The disk currents are shifted
to a lower potential, e.g., about 60 mV at 3 mA/cm2, in
Ni(Co) effects a fundamental change in the way Pt atoms
comparison to the other supported catalysts. The peroxide
interact with H2O and possibly O2, and this change could be
production increases continuously starting from potentials below
manifested in the kinetic measurements.
0.7 V, reaches a kind of plateau between 150 and 300 mV, and
3.2.2. Oxygen Reduction Reaction (ORR) Kinetics. Figures increases rapidly for potentials below 0.2 V. At 0.4 V at room
8 and 9 show a sample set of ring-disk measurements with temperature, the peroxide yield from PtNi is more than twice
the alloy catalysts in comparison with the standard Pt catalyst that from Pt or Pt3Ni. However, as we noted in the previous
in O2-saturated 0.1 M HClO4 at 60 °C. Following a potential section, the behavior of this catalyst is quite unstable in the
region where the ORR is under mixed kinetic-diffusion control, ORR potential region, especially at 60 °C. For that reason we
well-defined diffusion limiting currents were recorded in the have not shown the curves in a figure, nor have we attempted
potential range between 0.1 to about 0.6 V. The ring currents to analyze the results in any detail.
amount to negligible fractions of the disk currents in the potential The inset D in Figures 8 and 9 presents a plot of i-1 vs ω-0.5
region above 0.6 V, which indicates that the ORR proceeds for various potentials obtained on the 75% Pt alloy catalysts at
without production of peroxide in solution throughout the room temperature. Essentially identical plots were obtained with
potential region for oxygen reduction. A slight decrease in the the standard Pt catalyst. The slope of the lines, the so-called B
disk currents with a concurrent increase in the ring currents can factor, allows us to check the for consistency with the theoretical
4186 J. Phys. Chem. B, Vol. 106, No. 16, 2002 Paulus et al.

Figure 4. (a) Low-magnification HREM micrograph. (b) Particle size distribution determined by HREM micrographs. (c) HREM micrograph with
atomic resolution of the Pt-25 at. % Co (Pt3Co) catalyst.

TABLE 3: Comparison of Tafel Slopes and Enthalpies of of the film diffusion resistance, 1/if, in eq 3. Using the values
Activation Determined on Pt/Vulcan and the Alloy Catalyst of CfDf for Nafion-covered Pt solid electrodes,28 from the
Tafel slope @ 20 °C Tafel slope 60 °C ∆H/kJmol-1 @ η ) intercepts we estimate the Nafion film thickness to be about
catalyst (mV/dec) lcd/hcda (mV/dec) lcd/hcda 0.30/0.35/0.40 V 0.1 µ (in good agreement with the physical measurement made
Pt/Vulcan 59/107 52/92 21/20/23 with a profilometer). However, the film resistance is sufficiently
Pt3Ni/Vulcan 51/86 38/75 24/22/26 small, e.g., typically 0.01 mA-1 cm2, that for most of the
PtCo/Vulcan 59/73 46/67 24/28/24
Pt3Co/Vulcan 51/80 45/70 25/28/26
polarization curve the kinetic currents can be obtained from the
simpler relation
a lcd, low current density region; hcd, high current density region.
1 1 1
values according to the relation19 ) + (4)
i ik id
1 1 1 1 1 L 1
) + + ) + + (3) The inset C in Figures 8 and 9 compares the Tafel plots for the
i ik if id ik nFCfDf BC ω1/2 mass transport corrected mass specific current densities (ob-
0
tained via eq 4) at a rotation rate of 1600 rpm for Pt and the
where ik is the kinetic current density, if is the diffusion limiting Pt-Ni(Co) alloy catalysts. In addition to having unstable
current density in the Nafion film covering the catalyst layer, behavior as a cathode material, the PtNi catalyst also exhibited
CfDf is the O2 solubility-diffusivity product in the film, L is the initial performance that was substantially lower than either the
film thickness, and id is the diffusion limiting current density Pt standard or Pt3Ni. Therefore, more detailed analysis of the
through the solution boundary layer. The other terms have their ORR kinetics was done only on the Pt3Ni catalyst and the two
usual meaning. We find BC0 ) (14.44 ( 3%) × 10-2 mA cm-2 Pt-Co catalysts.
rpm-0.5, a value that agrees well with the calculated value of 3.2.3. Kinetic Analysis of the ORR. Figures 8C and 9C show
14.28 × 10-2 mA cm-2 rpm-0.5 for a four electron (n ) 4) the Tafel plots of the kinetic current densities (per unit geometric
process using literature data for O2-solubility (1.18 mol/m3) 26 area) obtained on Pt and the different alloy catalysts at room
and O2 diffusivity (1.9 × 10-9 m2 s-1) in low concentrated temperature and 1600 rpm in the anodic sweep direction (also
electrolytic solutions26 and the viscosity of water (0.893 × 10-2 the data are not shown, the qualitative trend for the current
cm2 s-1).27 As expected for Nafion film covered electrodes, the densities obtained at 60 °C is identical). First of all we note
intercept, even at potentials clearly in the so-called limiting that the activity of pure Pt and the alloy catalyst are nearly
current plateau, are not truly zero. This is due to finite values identical with respect at an identical (total) metal loading. It is
O2 Reduction on C-Supported Pt-Ni and Pt-Co Catalysts J. Phys. Chem. B, Vol. 106, No. 16, 2002 4187

Figure 5. (a) Low-magnification HREM micrograph. (b) Particle size distribution determined by HREM micrographs. (c) HREM micrograph with
atomic resolution of the Pt-50 at. % Co (PtCo) catalyst.

Figure 6. Base voltammograms of the standard supported Pt catalyst Figure 7. Base voltammograms of the standard supported Pt catalyst
compared to the supported Pt3Ni and PtNi catalyst in argon saturated compared to the supported Pt3Co and PtCo catalyst in argon saturated
0.1 M HClO4 at room temperature, 900 rpm, 50 mV/s; current densities 0.1 M HClO4 at room temperature, 900 rpm, 50 mV/s; current densities
normalized to the geometric electrode area. normalized to the geometric electrode area.

evident that the Tafel slope changes continuously in the reported for single-crystal Pt electrodes,30 or previously reported
examined potential range, as noted in previous studies of values for carbon-supported Pt18 in 0.05 M H2SO4, where ca.
supported Pt catalyst.25,29 We nevertheless fitted the experimental -2.3RT/F were obtained at low overpotentials (E > 0.85 V)
data to two Tafel slope regions at low and high overpotential and ca. -2‚2.3RT/F at high overpotentials (E < 0.8 V). Some
to enable a comparison with the literature data for solid of this difference can be attributed to the different supporting
electrodes. The resulting Tafel slopes at two different temper- electrolyte. Comparing single-crystal data obtained in 0.05 M
atures are listed in Table 3. The Tafel slopes for the standard H2SO4 to that in 0.1 M HClO4,25 we find the same qualitative
Pt catalyst in 0.1 M HClO4 are lower compared to values differences in Tafel slopes (lower slopes in 0.1 M HClO4),
4188 J. Phys. Chem. B, Vol. 106, No. 16, 2002 Paulus et al.

Figure 8. Disk (A) and ring (B) currents obtained during the orr in the anodic sweep (solid lines) and cathodic sweep (dashed lines) on the Pt3Ni
catalyst at room temperature, 5 mV/s. (C) Tafel plots for the ORR at room temperature, anodic sweep 5 mV/s, 1600 rpm on standard Pt, PtNi, and
Pt3Ni catalyst. (D) Levich-Koutecky Plots for the ORR on Pt3Ni at various potentials extracted from data in (A); current densities normalized to
the geometric electrode area.

especially for the Pt(111) and Pt(100) orientations. Since the in Figure 11. The activation enthalpies determined from least-
HREM results clearly indicate that the Pt nanoparticles used in squares regressions are essentially identical for the different
this study consist mostly of Pt(111) and Pt(100) facets, the Tafel catalysts, in the considered potential region, and are in the range
slopes are consistent with the above cited literature data. The between 20 and 25 kJ/mol (see Table 3). These values are in
lower Tafel slopes observed in 0.1 M HClO4 in comparison to close agreement with values reported for Pt single-crystal
data obtained in H2SO4 can be attributed to the influence of electrodes in sulfuric acid of ∼25 kJ/mol,30 for polycrystalline
anion adsorption.29 platinum in perchloric acid at 0.8 V (pH ) 1.9)31 of ∼20
Figure 10 shows plots of the turnover frequency (TOF) as a kJ/mol, and for values reported previously for carbon-supported
function of potential. These plots are derived from the kinetic Pt in sulfuric and perchloric acid,18 26 and 28 kJ/mol,
current densities and using the estimate of Pt atom surface respectively. An activation energy of 28 kJ/mol was also
density in Table 2, according to the relation, reported using a Nafion covered Pt microelectrode.32 However,
a significantly larger activation energy of ∼60 kJ/mol was
ik reported by Beattie et al.33 also using a Nafion-coated Pt
TOF ) (5) microelectrode technique, and by Mukerjee et al.12 using
neNs
supported Pt catalysts in conventional fuel cell type electrodes
TOF is an absolute reaction rate, the number of oxygen with a Nafion membrane. In the former the authors attributed
molecules A being reduced per unit time and per active surface their higher activation enthalpies compared to Parthasarathy et
atom, where ik is the kinetic current density, n is the number of al.32 to organic impurities from the Nafion solution used to make
electrons involved in the reaction (n ) 4 for potentials higher the recast film, a suggestion with which we concur. In a previous
than 0.7 V), and Ns is the number of active surface atoms (in study from this laboratory16 using solid electrodes of Pt and
this case we assume only Pt atoms are active). A real activity Pt3Ni bulk alloy, we reported activation enthalpies between 20
enhancement per Pt surface atom of about a factor of 1.5-2 and 25 kJ/mol for both electrodes in a comparable potential
can be observed for both of the 75% Pt catalysts, and nearly a range in perchloric acid.
factor of 5 for the PtCo catalyst. Assuming that the first electron-transfer step is the rate
The activation enthalpies (∆H*) are evaluated at a fixed determining step,
overpotential using the Arrhenius equation below and following
the procedure described previously,18 Pt(O2)ad + e- f Pt(O2-)ad (7)

∂(log ik)
∂(1/T) |
η
)
∆H*
2.3R
(6)
Markovic et al.34 derived a rate expression using conventional
transition state theory given by

using a temperature range between room temperature and 60 i ) nFkcO2(1 - θad)x exp(-βFE/RT) exp(-γrθad/RT) (8)
°C. The Arrhenius plots for three different overpotentials (η )
300, 350, and 400 mV) for three different catalysts are shown where i is the observed current density, n is the number of
O2 Reduction on C-Supported Pt-Ni and Pt-Co Catalysts J. Phys. Chem. B, Vol. 106, No. 16, 2002 4189

Figure 9. Disk (A) and ring (B) currents obtained during the orr in the anodic sweep (solid lines) and cathodic sweep (dashed lines) on the Pt3Co
catalyst at room temperature, 5 mV/s. (C) Tafel plots for the ORR at room temperature, anodic sweep 5 mV/s, 1600 rpm on standard Pt, PtCo, and
Pt3Co catalyst. (D) Levich-Koutecky Plots for the ORR on Pt3Co at various potentials extracted from data in (A); current densities normalized to
the geometric electrode area.

electrons, k is the (chemical) rate constant, cO2 is the concentra- unlikely one can obtain true kinetic currents. So curved Tafel
tion of O2 in the solution, θad is the total surface coverage of plots at potentials below ≈0.8 V cannot be considered unex-
adsorbed species (both intermediates and adsorbed anions), E pected. Further examination of the voltammetry in Figures 7
is the applied potential, β and γ are symmetry factors (assumed and 8 and the reaction rate theory we have just discussed provide
to be 0.5), and rθad is a parameter characterizing the change in a rationale for the slightly higher activity of the alloy catalysts.
the apparent Gibbs energy of adsorption of oxygeneated species As we have seen, both the activation enthalpy and the Tafel
with surface coverage. Note that in general a Tafel plot will be slope are essentially identical for the pure Pt and the alloy
curved, i.e., (∂ log i/∂E)T is a function of potential, only if the catalysts, which means that the (chemical) rate constants K and
rθad term is a nonlinear function of potential. The species, which the exponential terms are essentially identical, and the difference
constitute rθad are the intermediates of O2 reduction and the must lie in the coverage dependent part of the preexponential
anions of the supporting electrolyte, which may or may not be term of equation 8, (1 - θad). In perchloric acid electrolyte,
adsorbed. Of the possible intermediates only the coverage of this term is primarily the OHad coverage, and as we see from
OHad is known independently from cyclic voltammetry, and the the voltammetry in Figures 6 and 7, the potential region for
adsorption isotherm is available only for some anions mostly OHad is shifted positively by about 50 mV on the alloys versus
on ideal surfaces, like Pt(111). It is reasonable to assume that pure Pt. Qualitatively, this would cause the (1 - θad) term to
in perchloric acid we can exclude anion adsorption as an be somewhat larger for the alloy catalysts than for pure Pt, and
important factor. The potential dependence of OHad on these the factor of 1.5-2 is consistent with the shift of ca. 50 mV in
supported metallic particles we can see easily in the cyclic the oxidation potential. It should also be noted that unless θad
voltammetry of Figures 6 and 7. Since the pseudocapacitive for pure Pt is very close to 1, which occurs only at potentials
current is proportional to the derivative of the coverage of an above ca. 0.95 V, the maximum effect one could see from this
adsorbed specie,35,36 we can easily see from the current plateau term is about a factor of 2-5, not an order of magnitude.
at E > 0.8 V that the coverage by OHad has a linear potential An alternative and/or additional factor to the preexponential
dependence, and thus the Tafel plot should be linear in this term of eq 8 contributing to the difference in kinetics could be
potential region, but with a slope that is not 2RT/F (β ) 1/2). the activation entropy, i.e., the configurational entropy of
This is the classical result obtained for bulk Pt electrodes (both formation of (O2-)ad. There are no other contributions to the
polycrystal and single crystal) in different electrolytes by many kinetics that we believe would satisfy the observed parameters.
research groups over many years (see review by Adzic37). It is not clear why Ni or Co in the Pt surface would cause either
Fundamentally, there is no reason why the Tafel plot for the shift of ca. 50 mV in the oxidation potential of the Pt surface
supported Pt should not also be linear in this potential region, atoms, and/or change the configurational entropy of formation
unless there are factors unaccounted for in the rate expression of (O2-)ad. We offer here a speculation, not more than that, on
given above. At present, we do not know what these factors why Ni or Co in the Pt surface might cause an anodic shift in
are. In the potential region below ≈0.8 V, the Tafel plot should the oxidation potential of the Pt surface atoms. If the Ni(Co)
show a transition to another linear region where Langmuirian atoms are stable in the surface under ORR conditions, they are
conditions of adsorption apply; i.e., the rθad term is zero and probably “oxidized” and have one or more OH ligands attached
the slope is 2RT/F (β ) 1/2). The observation of this transition throughout the ORR potential region. These OH ligands could
is more controversial even with solid electrodes, and with give rise to an effect on OH formation on neighboring Pt atoms
supported Pt the reaction rates (TOFs) are so high that it is analogous to the “common-ion” effect in solution chemistry;38
4190 J. Phys. Chem. B, Vol. 106, No. 16, 2002 Paulus et al.

Figure 11. Arrhenius plots for the kinetic current densities for the
ORR on (A) standard Pt and Pt3Ni and (B) PtCo obtained from the
anodic sweeps (5 mV/s, 1600 rpm) at different overpotentials in the
temperature range between room temperature and 60 °C.
Figure 10. Turn over frequencies (TOF) for (A) standard Pt and Pt3-
Ni and (B) standard Pt, PtCo and Pt3Co at 60 °C calculated from the
kinetic current densities using eq 3.

i.e., the presence of OH species on the surface repels like species

from forming nearby. That explanation would be consistent with
the activity per unit Pt surface atom being higher for the PtCo
catalyst versus the 75% Pt surfaces. Why only Ni and Co? Many
transition metals might be expected to provide the same function,
and this may explain why so many Pt-transition metal alloys
are reported to have some beneficial effect. They key to this
enhancement is (a) the stability of the transition metal in the
surface at ORR conditions and (b) achieving the optimum
surface composition.
Finally, we comment briefly on the kinetic enhancement we
observe with carbon-supported Pt-Ni and Pt-Co in comparison
to those reported by others. We observed a small but statistically Figure 12. Histogram showing the kinetic current densities at different
potentials (geometric current densities) for the catalysts in this study
significant activity enhancement per geometric area on elec- obtained at 60 °C.
trodes with equal overall metal loading, very similar particle
size, and a very similar Pt surface atom density. An overview qualitative agreement with the factors of 2-5 (mass activity)
of the obtained kinetic current densities per geometric area is obtained using very similar catalysts in phosphoric acid,10 and
given in Figure 12. The so-called mass activity, activity per earlier studies by Mukerjee et al.11-13 or Tamizhmani et al.14
unit weight of Pt in the catalyst, is enhanced by factors of about in PEM electrolyte. In sharp contrast to these results are studies
1.5-3 for the alloy catalysts. Based on the Pt surface atom from Toda et al.15 on several different Pt-Co, Pt-Ni, and Pt-
density, the rate per Pt atom (the TOF) is a factor of 1.5-2 Fe catalysts where an activity (equivalent to our TOF) enhance-
higher for Pt3Ni(Co) and as much as a factor of 5 for PtCo. We ment of more than a factor of 10 was observed in comparison
found very similar TOF factors using well-characterized bulk to pure Pt (in 0.1 M HClO4 at ambient temperature). We will
metal electrodes of polycrystalline Pt and polycrystalline Pt3- not attempt to reconcile these differences with our results. We
Ni.16 The activity enhancements observed in this study are do, however, note that the general level of activity for the “pure
generally smaller than those reported by others but are in Pt” case reported by Toda et al. (Figure 10 in ref 16) is
O2 Reduction on C-Supported Pt-Ni and Pt-Co Catalysts J. Phys. Chem. B, Vol. 106, No. 16, 2002 4191

significantly lower than data for polycrystalline Pt at exactly (6) Luczak, F. J.; Douglas, A. Ordered Ternary Fuel Cell Catalysts
the same conditions reported by Damjanovic and co-workers.39,40 Containing Platinum and Cobalt. U.S. Patent No. 4711829, 1987.
(7) Luczak, F. J.; Landsman, D. A. Ordered Ternary Fuel Cell Catalysts
We also note that the catalysts used by Toda et al. are sputtered Containing Platinum and Cobalt and Method for Making the Catalysts. U.S.
alloy films that have subjected to an elaborate pretreatment that Patent No. 4677092, 1987.
intentionally corrodes the transition metal from the surface (8) Beard, B.; Ross, P. J. Electrochem. Soc. 1990, 137, 3368.
region leaving a so-called Pt “skin”. The physical structure of (9) Glass, J. T.; Cahen, G. L.; Stoner, G. E.; Taylor, E. J. J.
Electrochem. Soc. 1987, 134, 58.
these catalysts makes comparison using either well-characterized (10) Min, M.; Cho, J.; Cho, K.; Kim, H. Electrochim. Acta 2000, 45,
bulk alloy or conventional supported catalysts problematic and 4211.
possibly misleading. (11) Mukerjee, S.; Srinivasan, S. J. Electroanal. Chem. 1993, 357, 201.
(12) Mukerjee, S.; Srinivasan, S.; Soriaga, M. P. J. Phys. Chem. 1995,
99, 4577.
4. Conclusion (13) Mukerjee, S.; Srinivasan, S.; Soriaga, M. P.; McBreen, J. J.
We investigated oxygen reduction kinetics on carbon-sup- Electrochem. Soc. 1995, 142, 1409.
(14) Tamizhmani, G.; Capuano, G. A. J. Electrochem. Soc. 1994, 141,
ported Pt-Ni and Pt-Co alloy catalysts using the thin film 968.
RRDE method in 0.1 M HClO4 applying a temperature range (15) Toda, T.; Igarashi, H.; Uchida, H.; Watanabe, M. J. Electrochem.
between room temperature and 60 °C. Comparison was made Soc. 1999, 146, 3750.
to a pure Pt catalyst on the same carbon support, Vulcan XC- (16) Stamenkovic, V.; Schmidt, T. J.; Paulus, U. A.; Radmilovic, V.;
Markovic, N. M.; Ross, P. N. Meeting Abstracts, ECS Fall Meeting, The
72, having the same metal loading (20 wt %) and nominally Electrochemical Society, Inc., San Francisco, CA., 2001; Abstract No. 1053.
the same particle size (4 ( 2 nm). High-resolution electron (17) Schmidt, T. J.; Gasteiger, H. A.; Stäb, G. D.; Urban, P. M.; Kolb,
microscopy was used to determine the size and shape of the D. M.; Behm, R. J. J. Electrochem. Soc. 1998, 145, 2354.
(18) Paulus, U. A.; Schmidt, T. J.; Gasteiger, H. A.; Behm, R. J. J.
particles as well as the particle size distribution on all catalysts. Electroanal. Chem. 2001, 495, 134.
Hydrogen adsorption pseudocapacitance was used to determine (19) Schmidt, T. J.; Paulus, U. A.; Gasteiger, H. A.; Alonso-Vante, N.;
the number of Pt surface atoms and to estimate the surface Behm, R. J. J. Electrochem. Soc. 2000, 147, 2620.
composition of the alloy catalysts. Kinetic analysis in compari- (20) Pearson, W. B. A Handbook of Lattice Spacings and Structures of
Metals and Alloys; Pergamon Press: London, 1958.
son to pure Pt revealed a small activity enhancement (per Pt (21) Markovic, N. M.; Gasteiger, H. A.; Ross, P. N., Jr. Langmuir 1995,
surface atom) of ca. 1.5 for the 25 at. % Ni and Co catalysts, 11, 4098.
and a more significant factor of 2-3 for the 50 at. % Co. The (22) Denton, J.; Gascoyne, J. M.; Thompsett, D. Materials for use in
50 at. % Ni catalyst was less active than the Pt standard and catalytic electrode manufacture. European Patent No. EP 731520A1, 1996.
(23) Conway, B. E.; Angerstein-Kozlowska, H.; Sharp, W. B. A. J.
unstable at oxygen electrode potentials at 60 °C. Ring-current Chem. Soc., Faraday Trans. 1978, 74, 1373.
collection measurements for peroxide indicated no significant (24) Paulus, U. A.; Schmidt, T. J.; Stamenkovic, V. R.; Markovic, N.
differences between the Pt-Co catalysts or the 25 at. % Ni M.; Ross, P. N. Oxygen Reduction Activity of Pt and Pt-Alloys in Acid
catalyst and pure Pt, while the 50 at. % Ni catalyst had a higher Electrolyte; In 1st European PEFC Forum, Proceedings, Büchi, F. N.,
Scherer, G. G., Wokaun, A., Eds.; European Fuel Cell Forum: Oberrohrdorf,
peroxide yield. Together with the observed Tafel slopes and 2001; p 51.
activation enthalpies, it was concluded that the kinetic enhance- (25) Markovic, N.; Gasteiger, H.; Ross, P. N. J. Electrochem. Soc. 1997,
ment is contained in the coverage dependent part of the 144, 1591.
(26) Gubbins, K. E.; Walker, R. D., Jr. J. Electrochem. Soc. 1965, 112,
preexponential factor of eq 8 of the conventional transition state 469.
theory rate expression. It is, however, not clear why the alloying (27) Weast, R. E. In CRC Handbook of Chemistry and Physics, 66 ed.;
with Ni or Co produces this change in the preexponential CRC Press: Boca Raton, Fl., 1986.
factor. (28) Zecevic, S. K.; Wainright, J. S.; Litt, M. H.; Gojkovic, S. L.;
Savinell, R. F. J. Electrochem. Soc. 1997, 144, 2973.
(29) Uribe, F. A.; Springer, T. E.; Gottesfeld, S. J. Electrochem. Soc.
Acknowledgment. The participants in this research from the 1992, 139, 765.
Lawrence Berkeley National Laboratory acknowledge support (30) Grgur, B. N.; Markovic, N. M.; Ross, P. N. Can. J. Chem. 1997,
from the Office of Advanced Automotive Technologies of the 75, 1.
(31) Sepa, D. B.; Vojnovic, M. V.; Vracar, L. M.; Damjanovic, A.
U.S. Department of Energy under contract DE-AC03-76SF00098. Electrochim. Acta 1986, 31, 91.
The operation of the National Center for Electron Microscopy (32) Parthasarathy, A.; Srinivasan, S.; Appleby, A. J.; Martin, C. R. J.
(NCEM) is funded by the Office of Science, Division of Electrochem. Soc. 1992, 139, 2530.
Materials Sciences, of the U.S. Department of Energy. U.A.P.’s (33) Beattie, P. D.; Basura, V. I.; Holdcroft, S. J. Electroanal. Chem.
1999, 468, 180.
research stay at the Lawrence Berkeley National Laboratory was (34) Markovic, N. M.; Gasteiger, H. A.; Grgur, B. N.; Ross, P. N. J.
partly financially supported by the Paul Scherrer Institut and Electroanal. Chem. 1999, 467, 157.
the Swiss Federal Office of Energy. (35) Conway, B.; Angerstein-Kozlowska, H. J. Electroanal. Chem. 1980,
113, 63.
References and Notes (36) Conway, B.; Angerstein-Kozlowska, H.; Ho, F. J. Vac. Sci. Technol.
A 1977, 14, 351.
(1) Gottesfeld, S.; Zawodzinski, T. A. Polymer Electrolyte Fuel Cells. (37) Adzic, R. Recent Advances in the Kinetics of Oxygen Reduction.
In AdVances in Electrochemical Science and Engineering, 1st ed.; Alkire, In Electrocatalysis; Lipkowski, J., Ross, P. N., Eds.; Wiley-VCH: New
R. C., Gerischer, H., Kolb, D. M., Tobias, C. W., Eds.; Wiley-VCH: York, 1998; p 197.
Weinheim, 1997; Vol. 5; p 195. (38) Conway, B. Ionic Hydration in Chemistry and Biophysics. In Ionic
(2) Wilson, M. S.; Gottesfeld, S. J. Appl. Electrochem. 1992, 22, 1. Hydration in Chemistry and Biophysics; Elsevier Scientific Publishing: New
(3) Wilson, M. S.; Valerio, J. A.; Gottesfeld, S. Electrochim. Acta 1995, York and Amsterdam, 1981; p 417.
40, 355. (39) Paucirova, M.; Drazic, D. M.; Damjanovic, A. Electrochim. Acta
(4) Stonehart, P. Ber. Bunsen-Ges. Phys. Chem. 1990, 94, 913. 1973, 18, 945.
(5) Luczak, F. J.; Landsman, D. A. Ternary Fuel Cell Catalysts (40) Sepa, D. B.; Vojnovic, M. V.; Damjanovic, A. Electrochim. Acta
Containing Platinum, Cobalt and Chromium. U.S. Patent No. 4447506, 1984. 1981, 26, 781.