Average turn-over frequency of O2 electro-reduction for Fe/N/C and Co/N/C catalysts in PEFCs

Article abstract of DOI:10.1016/j.electacta.2007.03.045

This work aims to evaluate the average turn-over frequency of O₂ electro-reduction for the catalytic sites resulting from the heat-treatment of iron- or cobalt-acetate and carbon black in ammonia at high temperature. This task is complex because at least three factors may control the activity of such catalysts: their metal content, nitrogen content and micropore specific area. In this work, the activity was measured for metal contents from 0.005 to 5 wt.%. The time of heat-treatment was tuned to keep the micropore specific area constant. At Fe content ≤0.2 wt.% the activity increases linearly with Fe content, thus enabling the average turn-over frequency of the Fe/N/C site to be determined. At Co content ≤1.0 wt.% the activity increases approximately as the square root of the Co content. Because no linear relation was found, the turn-over frequency of the Co/N/C site could not be determined. For both Fe and Co catalysts, the activity drops dramatically at contents >1 wt.%. This is concurrent with a drop in the micropore specific area of the catalysts.

Full text of DOI:10.1016/j.electacta.2007.03.045

Electrochimica Acta 52 (2007) 5975–5984
Average turn-over frequency of O electro-reduction
2
for Fe/N/C and Co/N/C catalysts in PEFCs
∗
∗,1
∗,1
Fr e´ d e´ ric Jaouen
, Jean-Pol Dodelet
´
INRS Energie, Mat e´ riaux et T e´ l e´ communications, 1650 Bd Lionel Boulet,
Varennes, Qu e´ bec, Canada J3X 1S2
Received 5 January 2007; received in revised form 6 March 2007; accepted 15 March 2007
Available online 21 March 2007
Abstract
This work aims to evaluate the average turn-over frequency of O
2
electro-reduction for the catalytic sites resulting from the heat-treatment of
iron- or cobalt-acetate and carbon black in ammonia at high temperature. This task is complex because at least three factors may control the activity
of such catalysts: their metal content, nitrogen content and micropore specific area. In this work, the activity was measured for metal contents from
0
.005 to 5 wt.%. The time of heat-treatment was tuned to keep the micropore specific area constant. At Fe content ≤0.2 wt.% the activity increases
linearly with Fe content, thus enabling the average turn-over frequency of the Fe/N/C site to be determined. At Co content ≤1.0 wt.% the activity
increases approximately as the square root of the Co content. Because no linear relation was found, the turn-over frequency of the Co/N/C site
could not be determined. For both Fe and Co catalysts, the activity drops dramatically at contents >1 wt.%. This is concurrent with a drop in the
micropore specific area of the catalysts.
©
2007 Elsevier Ltd. All rights reserved.
Keywords: Polymer electrolyte fuel cell; Non-noble catalyst; Metal loading; Oxygen reduction
1
. Introduction
or by improving mass-transport properties of electrodes at high
current density.
Polymer electrolyte fuel cells (PEFCs) are an alternative to
i) internal combustion engines in the transportation market and
The present paper is concerned with the difficult task of
replacing Pt by non-noble catalysts at the PEFC cathode where
the oxygen reduction reaction (ORR) takes place. Several fam-
ilies of such catalysts for the ORR in acidic medium are under
investigation today. They may be divided into two groups, either
using a heat-treatment in the catalyst preparation procedure, or
not. In the first group, Fe- or Co-based catalysts are obtained
after heat-treating: (i) a metal-centered macrocyclic compound
(porphyrin, phthalocyanine) adsorbed on a carbon support [6–9],
(ii) an unsupported chelate in the presence of a foaming agent
[10,11], (iii) a metal salt in the presence of carbon and nitrogen
[12–24], or (iv) a sputter-deposited FexC1−x−yNy film [25]. In
the second group, Co-based catalysts are obtained without heat-
treatingtheCoprecursor. Forsuchprocedure, Co-basedcatalytic
sites are either obtained after plasma treatment of a carbon sup-
ported porphyrin [26] or based on the interaction of Co with
polypyrrole [27]. Recent reviews covered the subject of non-
noble catalysts for PEFCs [28,29]. Assuming that the above fam-
ilies of non-noble catalysts for the ORR are either stable or can
be stabilized in PEFC environment, another important question
(
to (ii) primary or secondary batteries in the portable electron-
ics market. One impediment to their commercialization is the
high materials cost of which the catalyst, platinum, occupies
a major place. While mass production of PEFCs would result
in lowered costs for bipolar plates and membranes, the Pt cost
is incompressible due its scarcity [1]. Thus, it is important to
either reduce the loading of Pt or Pt-group metals or replace
them with non-noble catalysts. Reducing the Pt loading (in g/kW
at given cell voltage) can be realized by increasing its intrinsic
catalytic activity through alloying with non-noble metals such
as Cr, Mn, Fe, Co or Ni [2,3] or by enhancing its utilization
through the deposition of Pt monolayers on nanoparticles [4,5]
∗
∗
Corresponding author. Tel.: +1 450 929 8142; fax: +1 450 929 8198.
Corresponding author. Tel.: +1 450 929 8143; fax: +1 450 929 8102.
E-mail addresses: jaouen@emt.inrs.ca (F. Jaouen),
∗
dodelet@emt.inrs.ca (J.-P. Dodelet).
1
ISE member.
0
013-4686/$ – see front matter © 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.electacta.2007.03.045

5
976
F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
is: can these catalysts reach an activity that meets the require-
ment for a commercial use? Even if the cost of such non-noble
catalysts is negligible compared to Pt, their acceptable level of
activity cannot be orders of magnitude lower than that of Pt:
the thickness of efficient cathodes in PEFCs is limited because
of mass- and charge-transport limitations. Thus, the decrease in
activity when switching from Pt to a non-noble catalyst must be
compensated for by making the cathode thicker. If the decrease
in activity is so large that it cannot be wholly compensated,
this means that the materials cost (cost per kW of electricity) in
membrane or bipolar plate will rise when using the non-noble
catalyst. This is not acceptable from a commercial standpoint.
Gasteiger et al. set the target of ORR activity for non-noble
metal atom) and per second [3]. Thus, the turn-over frequency
of such catalysts, and hence their practical potential, remains
unclear.
The object of this paper is thus to measure the turn-over
frequency for O2 electro-reduction of Fe/N/C and Co/N/C cat-
alytic sites obtained from a heat-treatment, and to compare it
to the target set by Gasteiger et al. in well-defined experimen-
◦
tal conditions (fuel cell, 80 C, pO 1 bar, 0.8 V versus SHE). It
2
is reminded that previous studies on Fe/N/C catalysts demon-
strated that two catalytic sites are simultaneously present in such
materials [33]. No such conclusion could be reached for Co/N/C
catalysts [34]. Therefore, the turn-over frequency measured in
this study is, for the Fe catalysts, a value averaged over the
different active sites. When estimating the average turn-over
frequency of such catalytic sites, it is important to investigate
catalysts with low metal contents in which, presumably, all metal
atomsformactivesites. Alastdifficultyinmeasuringtheaverage
turn-over frequency resides in the importance of the microp-
ore specific area of the carbon support [23]. An experimental
strategy was chosen in order to maintain the micropore spe-
cific area constant, and this independently of the metal loading.
The present study will first investigate, at room temperature and
in sulphuric acid solution, the activity against the metal con-
tent (0.005–5 wt.%). It will be shown that a range of Fe content
exists in which the activity increases linearly. In contrast, Co
catalysts in the same range display an activity increasing only as
the square root of the Co content. Thus, the turn-over frequency
could be determined only for the Fe/N/C site. Then, the activity
(Fe and Co) and average turn-over frequency (Fe only) are mea-
th
catalysts in automotive applications as 1/10 of the Pt state-of-
the-art activity [3]. This compromise on the activity is viable
under the assumption that efficient, 100-␮m thick cathodes for
non-noble catalysts (against typically 10-␮m thick for current Pt
cathodes) will become feasible. Following, the target for non-
noble catalysts in automotive applications according to these
3
authors is a volumetric activity of 130 A/cm of electrode at
◦
0
1
.8 V versus SHE, 80 C and under an O2 absolute pressure of
3
bar (Table 6 in Ref. [3]). The previous figure (A/cm electrode
at 0.8 V) for any catalyst is equal to the mathematical prod-
3
uct of the site density (number of active sites/cm ), turn-over
frequency (number of electrons reduced per active site and per
second at 0.8 V) and electric charge of a single electron (p. 30
in Ref. [3]). The turn-over frequency depends on the electro-
chemical overpotential while the site density is a constant. It
is therefore of interest to appraise as precisely as possible the
turn-over frequency of such catalysts in order to predict if the
above target can be reached and what practical site density it
implies.
Previously, the turn-over frequency (ToF) of three non-noble
catalysts has been estimated on the basis of published works [3].
However, all three estimated values are subject to large errors
due to (i) the assumption of an activation energy of 55 kJ/mol O2
that was used to correct two out of three literature data measured
◦
sured on carefully chosen catalysts at 80 C in fuel cell. From
◦
the knowledge of the average turn-over frequency at 80 C in
PEFC, it is then possible to calculate the theoretical site den-
3
sity necessary to reach the target of activity (A/cm electrode
at 0.8 V versus SHE) for non-noble catalysts for an automotive
application.
2. Experimental
◦
◦
at 20 or 50 C to the reference temperature of 80 C chosen by
Gasteiger et al., and (ii) the assumption that all metal atoms were
efficiently used in active sites for the catalysts described in the
literature. For assumption (i), an error of up to a factor 10 on
the turn-over frequency may incur, as noted by the authors [3].
2.1. Catalyst synthesis (fully detailed in [23])
The furnace black (Sid Richardson Carbon Company) has a
2
−1
2
−1
BET area of 71 m g (micropore area 5 m g ). It is the same
furnace black than the one used in Ref. [23] and is characterized
by N, O, and S contents of 0.66, 0.44, and 0.65 at.%, respec-
tively. For targeted metal loadings >0.05 wt.%, metal acetate
is weighed (mMeAc) and dissolved in distilled water. For tar-
geted metal loadings <0.05 wt.%, the mass of metal acetate is
too small to be accurately weighed and, instead, a dilute solu-
tion is prepared, from which the required aliquot is pipetted
(to which corresponds a known mass of metal acetate, mMeAc).
Then, if x is the mass fraction of metal (e.g. 0.2 wt.% metal
◦
Extrapolating an activity measured at 20 C to that expected at
0 C may lead to a decade overestimation if the true energy of
◦
8
activation is, e.g. 22 instead of 55 kJ/mol O2 (even for the exten-
sively studied Pt, activation energies spanning 22–76 kJ/mol O2
have been reported [30,31]). For assumption (ii), several papers
have shown that the activity levels off at Fe loadings >0.15 wt.%
(
iron acetate precursor) [16,32] or >1 wt.% (iron porphyrin pre-
cursor) [18] for such catalysts, meaning that some of the metal
atoms may not form active sites. Thus assumption (ii) could
also lead to an error of a factor 10 on the turn-over frequency,
e.g. in the case of a catalyst containing 2 wt.% Fe, of which only
−
3
yields x = 2 × 10 ) desired in the (carbon black + metal) pow-
der, then the mass of carbon to weight is mMeAct(1 − x)/x with
t = 0.3212 being the mass of iron in iron acetate; or t = 0.2366 that
of cobalt in cobalt acetate hydrated with four water molecules.
Commonly, 550–600 mg of carbon is added and dispersed in
100–120 mL of the water solution containing the metal acetate.
0
.2% formed active sites. Due to the combined or separate effect
of these two sources of error, it is not surprising that the turn-
over frequency of three Fe-based non-noble catalysts spanned
a wide range from 0.02 to 1.7 electrons per active site (i.e. per

F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
5977
The suspension is stirred and then dried overnight in an oven
at 80 C. About 500 mg of the resulting powder is ground and
placed in a quartz boat. The exact mass before heat-treatment,
of the catalyst is prepared by mixing 10 mg of catalytic pow-
der, 95 ␮L of a 5 wt.% Nafion solution (Aldrich) and 350 ␮L of
ethanol for 30–60 min. Seven microliters of the ink are pipetted
◦
2
m , is measured. The quartz boat with the powder, as well as a
onto a glassy carbon disk (0.2 cm ), resulting in a catalyst load-
i
−
1
glass rod/interior magnet assembly, is placed in a quartz tube,
but outside of the oven. The tube is first purged with argon.
The oven is then switched on and allowed to stabilize at its set
ing (iron plus carbon) on the glassy carbon of 0.8 mg cm
.
−1
Cyclic voltammograms with a sweep rate of 10 mV s are
recorded in the potential range of −0.25 to 0.75 V versus a satu-
rated calomel electrode (SCE). When the electrode is not rotated,
in the downward potential scan, a peak reduction current occurs,
owing to O2 reduction kinetics and O2 depletion. The potential
at which this peak occurs, Vpr, is a measure of the catalytic activ-
ity. A second measure is obtained by rotating the electrode at
1500 rpm, correcting the total current for the capacitive current
and for limitation by O2 diffusion in the electrolyte according
to the Koutecky–Levich equation. These two measures of the
activity are reliable (supplementary information of Ref. [23]).
In the present work, the kinetic current density (in A g of cat-
alytic powder) at 0.5 V versus SCE is used in the figures. The
Vpr values are reported in Table 1 in order to allow the reader
to compare the present activities with older ones reported by
our laboratory. Next, the H O yield is measured with the rotat-
◦
temperature (940 C; corresponding to an inner-tube tempera-
◦
ture of 950 C) for 2 h (supplementary information of Ref. [23]).
After 100 min of stabilization, the argon flow is switched to pure
ammonia with a flow rate of 2000 sccm. Twenty minutes later,
the boat is pushed into the middle of the oven and the chronome-
ter is started. The heat-treatment is ended by removing the quartz
tube from the oven. The powder after the heat-treatment is again
weighed and yields m . The weight loss percentage of carbon
f
during the heat-treatment in NH3 is
−1
m − m
i
f
W = 100 ×
(1)
m
i
This weight loss is the result of a gasification reaction of carbon
◦
black by NH3 occurring at a fast rate above 800 C and yielding
2
2
mainly HCN and H2 [35].
ing ring disk electrode. Before each measurement, the Pt ring is
cleaned by sweeping its potential between −0.3 and 1.1 V ver-
−
1
2
.2. Activity and selectivity at room temperature: rotating
sus SCE at 50 mV s for about 5 min. For the measurement, the
(
ring) disk electrode technique
disk potential is swept between −0.3 and 0.8 V versus SCE at
−
1
2
mV s and the disk-ring is rotated at 200 rpm. Meanwhile, the
The catalytic activity and H2O2 yield are measured in a solu-
ring potential is held at 1.1 V versus SCE. The disk current mea-
tion of sulphuric acid of pH one saturated by pure O2. An ink
sured under O is corrected by retrieving the current recorded
2
Table 1
Synthesis parameters for the catalysts
Metal content before
heat-treatment (targeted)
Time of heat-treatment at
950 C in NH3 (min)
Weight loss percentage
of carbon (wt.%)
Metal content after
heat-treatment (measured by
neutron activation) (wt.%)
Catalytic activity, Vpr
(mV vs. SCE)
◦
(
wt.%)
Initial carbon black
0
15
36.7
0.004 Fe, 0.0002 Co
162
Iron-based catalysts
0
0
0
0
0
0
0
0
2
.005
.01
.02
.05
.1
.2
.4
.8
.0
15
15
15
15
16
20
25
50
80
32.8
31.7
31.1
30.0
32.2
34.6
34.2
37.5
29.1
0.009
0.013
0.026
0.064
0.12
0.24
0.55
1.10
2.70
271
288
370
440
445
445
459
466
221
Cobalt-based catalysts
0
0
0
0
0
0
0
0
1
1
2
5
.005
.01
.02
.05
.1
.2
.4
.8
.0
15
15
15
15
15
15
19
22
20
18
20
18
30.6
32.6
33.2
30.6
30.6
28.3
31.7
34.2
36.4
31.3
33.3
35.5
0.0051
0.0093
0.015
0.057
0.094
0.20
0.37
0.87
1.12
1.73
293
347
332
379
385
394
406
410
263
180
334
191
.5
.0
.0
2.40
6.90

5
978
F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
under N2. The H2O2 yield reported is the one read at 0 V versus
SCE during the first potential scan, which tends to represent the
maximum yield over the range −0.3 to 0.8 V versus SCE. It is
calculated according to Eqs. (2) and (3) [36,37]:
3. Results and discussion
3.1. Measurements at room temperature in H SO solution
of pH 1
2
4
4
− n
3
.1.1. Conditions for a correct measurement of the
%
H2O2 = 100 ×
(2)
2
turn-over frequency
where n, the apparent number of electrons transferred during
ORR, is calculated according to
The metal loading on carbon was spanned from 0.005 to
5 wt.%. As already mentioned in Section 1, in order to study the
sole effect of metal loading on the ORR activity, it is extremely
important that the compared catalysts contain a same amount of
micropores. It has been recently demonstrated that the microp-
ore surface area controls the activity of catalysts prepared from
iron acetate and with a metal loading of 0.2 wt.% [23]. It is also
known, from Fig. 2B in the same study, that the rate of car-
bon gasification by NH3 is slowed down when the Fe loading
is increased. This effect is due to the decomposition reaction on
metallic iron of some of the NH3 into N2 and H2. Thus, heat-
treating all catalysts in NH3 for a unique duration would result
in various weight loss percentages of carbon, and therefore in
various micropore surface areas. This would be a poor experi-
mental strategy: the true relation between the metal content and
the activity for the ORR would be skewed by the change in
the catalyst microstructure. Here instead, the duration of heat-
treatment was tuned with respect to the metal loading in order
to keep the weight loss percentage of carbon constant (Table 1).
The targeted weight loss percentage was 30–35 wt.%, which
corresponds to the point at which the micropore area is maxi-
mized for the present furnace black [23]. It is then verified if
holding the weight loss percentage constant actually results in
a constant micropore surface area (Fig. 1). This proves correct
in the range 0–1 wt.% of Fe or Co. At higher metal loading, the
micropore area decreases (Fig. 1), even for a constant weight
loss of carbon (Table 1). This effect can also be measured, in
a simpler manner, by looking at the capacitive current (mea-
sured in H2SO4 of pH 1 at 0.7 V versus SCE, where no Faradaic
reaction occurs) of the catalysts (Fig. 2). Since the micropore
4
I
d
n =
(3)
Id + (Ir/N)
where I , Ir, and N are the disk current, ring current and ring
d
collection efficiency, respectively.
◦
2
.3. Activity at 80 C: fuel cell experiments
Membrane electrode assemblies are prepared using a
Nafion 117 membrane, commercial anodes (E-TEK ELAT,
−
2
0
.35 mg Pt cm ) and cathodes made of catalyst inks deposited
onto commercial electrodes (E-TEK ELAT). The formula
for the cathode ink is: 10 mg of catalyst/217 ␮L of Nafion
solution 5 wt.%/272 ␮L ethanol/136 ␮L water. This formula
yields a Nafion to catalyst mass ratio of 1 in the dry cath-
ode. The ink is mixed under ultrasonic agitation for 1 h. A
7
1
1-␮L aliquot is deposited on an ELAT electrode of area
.14 cm . The cathode is then dried and weighed, and the
2
anode and cathode are hot-pressed against a Nafion 117 mem-
brane. This procedure yields a catalyst loading at the cathode
−
2
of 1 mg cm ± 10% (mass of carbon + metal). The fuel cell
◦
◦
temperature is 80 C and the humidifier’s is 105–110 C. Humid-
ified H2 and O2 are fed in the cell at a pressure of 2 bar
absolute each, resulting in an absolute pressure of 1.5 bar each
◦
(
pH O sat = 0.5 bar at 80 C). The polarization curve is recorded
2
−
1
at a scan rate of 0.5 mV s after a break-in at 0 V for 1 h.
The cell resistance is measured by impedance spectroscopy
at open circuit after the polarization curve. It is typically
2
0
.2 ꢀ cm .
2
.4. Determination of elemental surface and bulk
concentrations
X-ray photoelectron spectroscopy is performed with a VG
Escalab 200i instrument using the Al K␣ line (1486.6 eV).
Narrow scans are recorded for the C1s and N1s core levels.
The bulk concentration of metal in the catalysts (after heat-
treatment) is determined by neutron activation analysis at the
´
Ecole Polytechnique de Montr e´ al. The error on the measure-
ment is usually ± 5%, but slightly higher for the lowest Fe
loadings.
2
.5. Note
In the text and in all figures, data given per gram of catalyst
2
−1
−1
(
(
e.g. m g or A g ) means per gram of the catalytic powder
including metal, carbon and nitrogen).
Fig. 1. Micropore specific area against metal content for (ꢀ) Fe-based and (ꢁ)
2
Co-based catalysts. Area is given in m /g of catalyst (carbon + metal).

F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
5979
Fig. 2. Capacitive current against metal content for (ꢀ) Fe-based and (ꢁ)
Fig. 3. Activity for O2 electro-reduction at room temperature against metal
Co-based catalysts. Measured in H2SO4 pH 1, room temperature, pO = 1 atm
content for (ꢀ) Fe-based and (ꢁ) Co-based catalysts. The lines were obtained
2
b
absolute.
by fitting a law of the type y = ax (Eq. (5)). Measured in H2SO4 pH 1, room
−
1
temperature, pO = 1 atm absolute. Current density is given in A g of catalytic
2
powder (carbon + metal).
area decreases at high metal loading, then the total BET area
decreases too and this yields a decreasing capacitive current.
The conclusion is that only the catalysts with a metal content
below the threshold value (1 wt.% for Fe or Co) can be used to
investigate the average turn-over frequency of the active sites.
It must also be noted that the metal content before and after the
heat-treatment has changed (Table 1). A concentration effect
arises from the carbon gasification because metal atoms do not
form volatile compounds during heat-treatment in NH3. It is
therefore important to measure accurately the metal content
after the heat-treatment. This was done by neutron activation
analysis. It is the post-heat-treatment content that is used in all
figures.
Once the above experimental traps avoided, the activity of
such catalysts can be meaningfully plotted as a function of their
metal concentration. Fig. 3 shows, in a log–log plane, how the
activity for the ORR (in A g of catalyst at 0.5 V versus SCE)
varies with the Fe or Co concentration. At low metal content
−
1
(
≤0.2 wt.% for Fe and ≤1 wt.% for Co) the activity increases
with increasing metal content. This means that, in that region,
neither the N content nor the micropore specific area of the
catalysts limits their activity; only the metal content limits the
activity. The Tafel slope, estimated from the E–log i curves
shown in Fig. 4, is 70–80 and 60–70 mV per decade for the Fe
and Co catalysts, respectively. In Fig. 3, the maximum activity
Another practical aspect that must be considered before relat-
ing the activity for the ORR to the metal content of the catalysts
is the presence of native metal atoms, i.e. atoms that come not
from the adsorption of metal acetate onto the carbon black but
that were present in the pristine carbon black itself. The latter
Fe concentration is 0.004 wt.% (first line of Table 1) while the
concentration of native Co is 20 times smaller. Thus, the prob-
lem resides in these 0.004 wt.% Fe biasing slightly the ORR
activity measured for the Co catalysts; at least for ultra-low Co
contents. This bias is corrected by retrieving from the kinetic
current at a given potential the kinetic current measured for the
carbon black heat-treated alone (no metal acetate added prior to
heat-treatment).
I
kinetic,corr(fixed E) = Ikinetic,measured(fixed E)
kinetic carbon HT alone(fixed E)
At 0.5 V versus SCE, the kinetic current measured on the carbon
−
I
(4)
−
1
black heat-treated alone is 0.039 A g of catalyst (mass of car-
bon + Fe + N). For the Co catalysts, only the corrected activity
will be displayed and discussed in the present paper. For the Fe
Fig. 4. RDE–polarisation curves in H2SO4 solution of catalysts with various
metal loadings in wt.%. Upper panel: Fe catalysts; lower panel: Co catalysts. The
current has been corrected for (i) double layer current and (ii) for O2 diffusion
limitation in the solution. Measured at pH 1, room temperature, rotation rate of
−
4
catalysts, no correction for the presence of 2 × 10 wt.% Co
−
1
was made.
1500 rpm and potential sweep rate of 10 mV s
.

5
980
F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
influence of the carbon black support on the peroxide yield [41].
The latter could reach at 0 V versus SCE (i) values as high as 80%
H2O2 for catalysts made with Black Pearls and with a develop-
mental high specific area graphite labeled HS300, and (ii) values
as low as 15% H2O2 for catalysts made with a developmental
carbon black from Sid Richardson Carbon Corporation, doped
with nitrogen in its bulk. This carbon was labelled RC1 in Ref.
[41]. In the latter work, the % H2O2 decreased with increased
nitrogen concentration at the surface of the catalysts. The low %
H2O2 values measured for the catalysts prepared for this study
may be understood in terms of nitrogen content as well. Indeed,
high values of 2–3 at.% of nitrogen were measured for the Co-
based catalysts depicted in Fig. 5, in agreement with the low
yield of peroxide (<4%) reported for the Co-based catalysts in
that figure. % H2O2 < 10% were also found by other groups for
Co catalysts [42,43].
Fig. 5. Selectivity for O2 electro-reduction at room temperature against metal
3
.1.2. Analysis of the increase in ORR activity at low metal
content
In this section the catalytic activity in the region where the
content for (ꢀ) Fe-based and (ꢁ) Co-based catalysts. Measured in H2SO4 pH
1
, room temperature, pO = 1 atm absolute.
2
catalysts are not saturated in metal atoms (left handside of the
vertical dashed lines in Fig. 3) is analyzed in more detail. If
one assumes that, below the threshold metal content, each metal
atom forms one active site and that the activity is proportional
to the number of active sites; then the mass activity (A g of
catalytic powder at fixed potential) against the metal content in a
log–log plane should display a slope of unity. The experimental
data in the metal-unsaturated region were fitted by straight lines
reached by the Fe catalysts is larger than that obtained by the Co
catalysts, which is in accordance with a previous study [38]. At
the threshold metal content (dashed vertical lines) the activity
levels off and, upon larger metal contents (>1 wt.% for Fe and
Co), even decreases dramatically by one to two decades. For
example, the activity of the 6.9 wt.% Co catalyst is as poor as that
of the carbon heat-treated with no addition of metal (Table 1: Vpr
of 191 and 162 mV versus SCE, respectively). The main expla-
nation for this is the drop in micropore area (Fig. 1). The surface
concentration of N atoms was 3–4 at.% for all Fe catalysts and
−1
(Fig. 3) yielding the following empirical relations between the
mass activity and the metal content:
I
k,Fe(0.5 V versus SCE) = 16.4cFe,
0.66
2
–3 at.% for all Co catalysts and was about constant with metal
loading. Thus, the N content cannot explain the activity drop.
In summary, a first difference is unravelled between the Fe
catalysts and the Co ones: the threshold content at which the
metal atoms saturate the carbon support is about 0.2 wt.% for
Fe against about 1 wt.% for Co; for this carbon black.
Ik,Co(0.5 V versus SCE) = 0.46cCo
(5)
−
1
where I is the kinetic current in A g of catalyst (mass of
k
C + metal) and c is the metal loading, in weight percentage. The
error on the pre-exponential or exponential factor is 10%. The
same analysis may be performed at other potentials with simi-
lar results for the exponential factors. So, a second difference is
unravelled between the Fe catalysts and the Co ones: the former
see their catalytic activity increase proportionally with the Fe
content in the region 0–0.2 wt.% while the latter see their cat-
alytic activity increase about as the square root of the Co content
in the region 0–1 wt.%.
Fig. 5 depicts the change of peroxide production with the
metal content of the catalysts. Only a slight decrease of the H2O2
yield from about 4 to 1.5% is observed when the Fe or Co con-
tent is increased from 0.005 to 1 wt.%. This oxygen reduction
mechanism of almost four-electrons for all catalysts indicates
that the nature of the catalytic sites is unchanged with the metal
loading. Only the site density seems to change with the metal
content. However, it is believed that this apparent four-electrons
reduction of O2 is mainly a 2 + 2 electron reduction; the perox-
ide generated by the first 2e reduction on one catalytic site being
able to be further reduced by another two electrons on another
catalytic site in the catalyst film. This could explain why increas-
ing the site density decreases slightly the peroxide yield, simply
by increasing the probability for a peroxide molecule to meet
another catalytic site on its way out of the catalytic film.
For the Fe catalysts, the linear increase in activity in the
region 0–0.2 wt.% metal content is easily explainable if all Fe
atoms form active sites during the heat-treatment. This result is
in accordance with previous experimental results obtained by
time-of-flight secondary-ion mass spectroscopy (time-of-flight
SIMS) [18,33]. A family of catalysts synthesized in conditions
not too different from those used in the present study (0.2 wt.%
Fe adsorbed on an N-rich carbon support; then pyrolyzed in
Ar/H2 at various temperatures) was characterized by time-of-
flight SIMS and by electrochemical methods [33]. It was shown
that the activity for the ORR correlated well with the time-of-
It is to be noted that Fe and Co catalysts of this study are
characterized by a similar and quite low peroxide yield (<4% in
Fig. 5). Previous works reported higher peroxide yields for Co
catalysts [39–41]. As far as the work performed by our group
is concerned, previous studies on Co catalysts reported a strong
+
flight SIMS intensity of the family of ions labelled FeN2Cy ,
+
and especially well with the ion FeN2C4 . The latter ion could

F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
5981
+
account, for the most active catalysts, for up to 80% of the
time-of-flight SIMS signal summed on all FeNxCy ions. These
signal of a single ion, FeN2C4 , accounted for up to 80% of
+
+
all FeNxCy ions, measurements on the Co catalysts did not
previous time-of-flight SIMS results say that, at 0.2 wt.% Fe con-
tent, and under heat-treatment conditions in which the activity
is maximum (properly chosen combination of carbon support,
time and duration of heat-treatment), a majority of Fe atoms
are involved in a first active site, labelled FeN2/C in Ref. [33].
The remaining Fe ions, about 20%, are involved in a second
catalytic site labelled FeN4/C in the same work. These catalytic
sites are also expected to be present in about the same propor-
tions in the present work since similar synthesis conditions were
applied. Thus, the turn-over frequency measured in the present
work represents an average of the turn-over frequency of the
individual catalytic sites FeN2/C and FeN4/C.
reveal any ion showing a particularly strong signal [34]. All
four families of ions (CoNCy , CoN2Cy , CoN3Cy , CoN4Cy )
+
+
+
+
represented each about 10–30% of the total signal summed on
+
all CoNxCy ions. Therefore, it is conceivable that only one of
+
these families of CoNxCy ions originates from the ORR active
sites, and that, consequently, only a fraction of the Co atoms is
found in active sites for the ORR.
Last, it is not surprising to see that, unlike for the Fe cat-
alysts, the time of heat-treatment to reach 30–35 wt.% loss is
almost constant with the Co loading (Table 1). Indeed, since the
famous screening experiments performed almost 100 years ago
by Haber and Bosh, it is known that the best metals to catalyze
the synthesis of NH3, and also therefore its decomposition, are
Ru and Fe [44–46].
A practical consequence of this section is that it is not possi-
ble to determine the turn-over frequency of the Co site because
the activity of Co catalysts does not increase linearly with metal
content (Eq. (5)). In contrast, for the Fe catalysts, the linear
relation between activity and metal content observed between
Another possibility exists that would also explain the linear
increase in activity with Fe content observed for Fe loadings
<
0.2 wt.%. Possibly, only a fraction of Fe atoms could form
active sites for the ORR, provided that this fraction does not
change with the metal loading. This hypothesis is however con-
tradicted by the apparent lack of Fe crystalline phases for Fe
loadings <0.2 wt.%, as explained now. Table 1 reports the time
◦
of heat-treatment at 950 C in NH3 necessary to reach a constant
0
and 0.2 wt.% loading enables us to determine now the aver-
weight loss of 30–35% for various Fe loadings on the catalysts.
This time must be increased a lot for loadings >0.2 wt.%. This
may be understood in terms of dissociation of some NH3 in the
catalyst bed into N2 and H2; a reaction known to be catalyzed
at the Fe surface [44–46]. This reasoning implies that, for Fe
loadings >0.2 wt.%, some iron precursor forms metallic iron
particles. On the other hand, for loadings <0.2 wt.%, no metallic
Fe exists since the duration of heat-treatment remains constant
at 15–16 min. As a consequence, all Fe atoms in the latter range
of Fe contents are believed to form iron ions in the catalytic sites
of the ORR.
The behaviour of the catalytic activity against metal content
of the Co catalysts is now discussed (Fig. 3). In contrast to the
Fe catalysts, for the Co catalysts the ORR activity increases
about as the square root of the metal content (region 0–1 wt.%).
It has been proposed by some authors that the Co active site
consists of two Co atoms instead of only one [42,47]. However,
such a hypothesis should result in an activity increasing faster
than linearly with Co content (meaning a slope >1 in a log–log
plane, Fig. 3). The opposite is observed here. Possibly, Co is
able to form several species, among which only one catalyzes
the ORR. If the fraction of the Co species active for the ORR
to the total Co species decreases with increasing Co content,
then this could give rise to an increase of the ORR activity with
Co content being less than proportional. Another possibility is
that some inactive Co species agglomerate onto, or somehow
render inactive, the Co species catalyzing the ORR. The latter
hypothesis could explain why at low Co content (<0.01 wt.%)
activities for the Fe and Co catalysts get closer and also why
the threshold metal content is higher for the Co catalysts than
for the Fe ones (1 wt.% against 0.2). The assumption that only
a minor part of all Co atoms are active for the ORR, and this
even in the region of 0–1 wt.% Co, is also reinforced by recent
time-of-flight SIMS measurements on similar Co catalysts [34].
In contrast to measurements on Fe catalysts [33] where the
age turn-over frequency of the Fe/N/C catalytic site at room
temperature.
3.1.3. Average turn-over frequency (ATF) at room
temperature of the Fe/N/C site
Since the kinetic current measured at 0.5 V versus SCE, Ik,Fe
,
increases linearly with the iron concentration, cFe, in Eq. (5), it
is possible to calculate, at that voltage, the average turn-over
frequency, ATF, of a single catalytic site. In order to do so, it
is assumed that all Fe atoms are coordinated in active sites and
that each active site hosts only one Fe atom. Then the theoretical
relation between the kinetic current, the metal content and the
ATF is
cFe Na
Ik(fixed E)=1
ATF(fixed E) × e=SD ATF(fixed E)×e
00 MFe
(
6)
−1
where I is the kinetic current (A g ), cFe is the Fe loading
k
23
(
wt.%), Na is Avogadro’s number (6.02 × 10 ), MFe is the molar
−1
mass of Fe (55.8 g mol ) and e is the electric charge of a single
−19
electron (1.6 × 10
C). In Eq. (6), the parameter group before
−1
ATF represents the site density, SD (active sites g ). The ATF
can be calculated equating Eq. (5) for Fe and Eq. (6). This yields
−1
−1 −1
ATF (0.5 V versus SCE) = 0.95 e site
s .
In order to extrapolate this value to the reference potential of
.8 V versus SHE chosen in [3], the offset of 59 mV between
.5 V versus SCE and 0.8 V versus SHE has to be taken into
0
0
account. The ATF can be corrected for this offset according to
a Tafel equation:
ATF(0.8 V versus SHE)
=
ATF(0.5 V versus SCE)
ꢀ−
ꢁ
59
ln(10) ≈ 0.14 e−1 site
−1 −1
s
×
exp
(7)
70

5
982
F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
Table 2
Average turn-over frequency, site density and volumetric activity at 0.8 V vs. SHE and under absolute O2 pressure of 1 bar
◦
−1
−1 −1
−3
−3
Ik,v (A cm )
T ( C)
ATF (e site
s
)
SDV (sites cm
)
2
2
1
1
1
0
0
8
8
8
4
7 wt.% Pt/C reference
80
80
80
80
80
80
20
25
2.5
0.38
0.35, 0.31
0.40
–
3.2 × 10
3.2 × 10
2.6 × 10
5.6 × 10
12.9 × 10
–
1300
130
Non-Pt-group metal catalyst target
0
0
0
0
.06 wt.% Fe
.13 wt.% Fe
.30 wt.% Fe
.87 wt.% Co
0.16
0.31, 0.28
0.82
0.04
–
Fe catalysts range 0–0.25 wt.%
0.14
–
sponding, for0.7 mgofPt + C cm 2, to1 A cm )provesthatthe
−
−2
A Tafel slope of 70 mV per decade was taken, as experimentally
−
1
−1 −1
s represents the
measured (Fig. 4). The value 0.14 e site
average turn-over frequency at 0.8 V versus SCE in the range of
–0.25 wt.% metal. If Tafel slopes of 60 and 80 mV per decade
experimental set-up performs well. The departure from the Tafel
law for the non-noble catalysts at currents larger than 30 A g
must thus be due to non-optimized cathode structure. For the
−1
0
were assumed instead, ATF values of 0.10 and 0.17 would be
found, respectively, giving thus an idea of the error in the value
given in Eq. (7). This value is reported in Table 2.
0
0
0
.87 wt.% Co catalyst, the polarization curve extrapolated from
.77 to 0.8 V in conditions of Fig. 6 gives a current density of
−
1
.15 A g at 0.8 V. For the Fe-based catalysts, the measured
−1
current densities at 0.8 V are 0.61, 1.17 and 3.08 A g with
0.06, 0.13 and 0.30 wt.% Fe, respectively. Moreover, the fuel
cell experiment was repeated for the 0.13 wt.% Fe catalyst to
verify reproducibility. The latter is good as it may be seen by
comparing the two values in Table 2 for 0.13 wt.% Fe.
◦
3
.2. Measurements in fuel cell at 80 C
3
.2.1. Determination of current densities at 0.8 V
In order to measure activity parameters at 80 C in PEFC,
◦
three Fe catalysts (0.06, 0.13 and 0.30 wt.%) and one Co cat-
alyst (0.87 wt.%) were chosen. For Fe, these catalysts stand
in the linear region (Fig. 3) with 0.30 wt.% being close to, or
slightly above, the metal saturation. These three Fe catalysts
were re-synthesized especially for the fuel cell tests, thus their
exact Fe content is not found in Table 1 but is given here. The
3
0
.2.2. Calculation of activity, ATF and site density of the
.06 wt.% Fe catalyst in reference conditions
From the reading of the current density at 0.8 V (assumed to
represent 0.8 V versus SHE, losses at the anode being negligi-
ble with pure H2) the derivation of values for the site density
and average turn-over frequency (ATF) is now detailed for the
catalyst with 0.06 wt.% Fe. From this example, it will be eas-
ily understood how the calculation is applied to the two other
catalysts.
0
.87 wt.% Co catalyst represents the most active Co catalyst. The
iR-corrected polarisation curves for these catalysts as well as for
a commercial Pt Gore MEA are shown in Fig. 6. The straight
−
1
Tafel slope observed for the Pt Gore MEA up to 1.5 A g (corre-
•
Assuming that all Fe atoms form active sites at such a low
content, the site density SD (sites per gram of catalyst) is
equal to cFe/100 × Na/MFe (Eq. (6)) with cFe = 0.06.
3
In Ref. [3] the site density and activity were given per cm
3
of electrode; assuming a density of 0.4 g of carbon per cm in
the porous electrode. The site density defined as such, SDv,
18
−3
is equal to 0.4 SD. This gives SDv = 2.6 × 10 sites cm for
.06 wt.% Fe.
The kinetic current, I , read at 0.8 V versus SHE is 0.61 A g
0
−1
•
•
k
of catalyst (carbon plus metal) (Fig. 6). As above, this yields
a current per electrode volume Ik,v = 0.4I . Next, a correction
k
must be made for the oxygen pressure which is 1.5 bar abso-
lute in Fig. 6 against 1 bar absolute in reference conditions.
−3
Then Ik,v in reference conditions is 0.4I /1.5 = 0.16 A cm
.
k
The ATF in reference conditions is equal to I /(SD e) =
k
18
−19
Ik,v/(SDv e). Thus ATF = 0.16/(2.6 × 10 × 1.6 × 10 ) =
−
1
−1 −1
s .
◦
0.38 e site
Fig. 6. iR-free fuel cell polarisation curves at 80 C. (A) 0.87 wt.% Co; (B)
0
.06 wt.% Fe (nominal 0.05 wt.%); (C) 0.13 wt.% Fe (nominal 0.1 wt.% Fe);
(
0
D) 0.30 wt.% Fe (nominal 0.2 wt.% Fe); (E) Commercial Gore 5510 MEA,
These values, as well as the ones obtained for the two other
−
2
.3 mg Pt cm and 45 wt.% Pt/C. H2 and O2 absolute pressures of 1.5 bar. The
−1
Fe catalysts and for the Co catalyst, are gathered in Table 2
where Pt reference data and non-noble target data from [3] are
also found. The ATF value obtained for the three Fe catalysts
current density is in A g of catalyst (carbon + metal). The catalyst cathode
−2
loading is 1 mg cm ± 10% for the non-noble catalysts. Thus the values in
−
1
−2
A g are in this case almost equivalent to mA cm .

F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
5983
is fairly constant, meaning that all Fe atoms form active sites
at these metal loadings or, at least, that the utilization is equal
for these three catalysts. The ATF value for the Fe/N/C site
tion of available micropores with catalytic sites below a metal
content of 1 wt.%, followed by the decrease of the microporous
surface area for metal contents above 1 wt.%. The other factor
that might influence the activity of such catalysts, the nitrogen
content, seems to be non-limiting here since it remains high and
constant for all catalysts of this study.
−
1
−1 −1
s . It falls in-
in reference conditions is 0.36 ± 0.03 e site
between the wide interval of 0.02–1.7 found in [3] but is very
−
1
−1 −1
s made previously by
close to the estimation of 0.4 e site
us from a single PEFC measurement [35].
The maximum activity reached by Fe catalysts is about a
decade larger than that reached by Co catalysts. This is valid
For the Co catalyst, neither the ATF nor the SD can be cal-
culated because the Co catalysts never show a linear increase of
activity with metal content (Fig. 3 and Eq. (5)).
◦
◦
both at 20 C in H2SO4 solution and at 80 C in a PEFC.
The average turn-over frequency at 0.8 V versus SHE of the
−
1
−1 −1
Fe/N/C site is 0.14 ± 0.03 e site
s at room temperature in
3
.2.3. Is the target for automotive application reachable by
a solution of sulphuric acid pH 1 saturated by pure O2 under
1 atm pressure. The average turn-over frequency at 0.8 V versus
such non-noble catalysts?
◦
−1
−1 −1 ◦
at 80 C in
The volumetric activity in PEFC conditions at 80 C of the
bestFecatalystofthisstudyis159timesbelowthetargetfornon-
noble catalysts (Table 2, last column). The volumetric activity is
the product of the volumetric site density, SDV, with the average
turn-over frequency, ATF. The SDV is a factor ∼25 below the
target and the ATF a factor ∼7 below the target. Thus, either one
or both kinetic parameters must be increased to come closer to
the target.
SHE of the Fe/N/C site is 0.36 ± 0.03 e site
s
a PEFC under an oxygen absolute pressure of 1 bar.
Acknowledgements
This work is supported by NSERC and General Motors of
Canada. The authors are indebted to the Sid Richardson Carbon
Company for providing the carbon black.
The ATF should be a parameter characteristic of the site, so
if the sites are unchanged, this parameter cannot be improved.
Thus, for the non-noble catalysts synthesized in the conditions of
the present study, all improvements have to come from SDV. In
order to estimate the possibility of improvement of this param-
eter, one must think of how much Fe loading is theoretically
possible. The active sites are believed to be made of Fe (Co)
bound to at least four N atoms and to an undetermined num-
ber of C and N atoms from the C support. Molecules like
phthalocyanine or porphyrin contain at least 24 or 40 atoms
of C + N, respectively. Assuming that the active site cannot be
made smaller than a phthalocyanine or a porphyrin molecule, an
equivalent upper Fe concentration of 11–19 wt.% could theoret-
ically be reached. Compared to the present utilizable Fe loading
References
[
1] M. Mathias, H. Gasteiger, R. Makharia, S. Kocha, T. Fuller, T. Xie, J. Pisco,
ACS Div. Fuel Chem. 49 (2004) 471 (Preprints).
[2] S. Mukerjee, S. Srinivasan, M.P. Soriaga, J. Mc Breen, J. Electrochem. Soc.
42 (1995) 1409.
1
[
3] H.A. Gasteiger, S.S. Kocha, B. Sompalli, F.T. Wagner, Appl. Catal. B:
Environ. 56 (2005) 9.
[
4] K. Sasaki, Y. Mo, J.X. Wang, M. Balasubramanian, F. Uribe, J. McBreen,
R.R. Adzic, Electrochim. Acta 48 (2003) 3841.
[5] J. Zhang, F.H.B. Lima, M.H. Shao, K. Sasaki, J.X. Wang, J. Hanson, R.R.
Adzic, J. Phys. Chem. B 109 (2005) 22701.
[
6] V.S. Bagotzky, M.R. Tarasevich, K.A. Radyushkina, O.E. Levina, S.I.
Andrusyova, J. Power Sources 2 (1977) 233.
[
7] R.W. Joyner, J.A.R. van Veen, W.M.H. Sachtler, J. Chem. Soc., Faraday
Trans. I 78 (1982) 1021.
(
about 0.3 wt.%), this represents a maximum possible improve-
ment in SDV of a factor 30–60. This is more than what is needed
to reach the SDV target, but falls short for the target of volumetric
activity if the ATF of these catalytic sites cannot be improved.
The situation is still worse for the Co-based catalysts which
show a mass activity a decade smaller than the Fe ones (Fig. 6).
It seems that, with the present synthesis procedure, only a small
fraction of the Co atoms form active sites.
[8] B. van Wingerden, J.A.R. Van Veen, C.T.J. Mensch, J. Chem. Soc., Faraday
Trans. 84 (1988) 65.
[
9] L. Zhang, C. Song, J. Zhang, H. Wang, D.P. Wilkinson, J. Electrochem.
Soc. 152 (2005) A2421.
[
10] P. Bogdanoff, I. Herrmann, M. Hilgendorff, I. Dorbrandt, S. Fiechter, H.
Tributsch, J. New Mater. Electrochem. Syst. 7 (2004) 85.
[11] I. Herrmann, P. Bogdanoff, G. Schmithals, S. Fiechter, ECS Trans. 3 (2006)
11.
12] S. Gupta, D. Tryk, I. Bae, W. Aldred, E. Yeager, J. Appl. Electrochem. 19
1989) 19.
2
[
[
(
4
. Conclusions
13] D. Ohms, S. Herzog, R. Franke, V. Neumann, K. Wiesener, S. Gamburcev,
A. Kaisheva, I. Iliev, J. Power Sources 38 (1992) 327.
The activity of the Fe catalysts increases proportionally with
[14] G. Lalande, R. C oˆ t e´ , D. Guay, J.P. Dodelet, L.T. Weng, P. Bertrand, Elec-
trochim. Acta 42 (1997) 1379.
Fe content up to about 0.2 wt.% metal while the activity of the Co
catalysts increases about as the square root of the Co content up
to 1 wt.%. With the present synthesis conditions, for metal con-
tents below 0.2 wt.% Fe and 1.0 wt.% Co, neither the nitrogen
content nor the micropore specific surface area of the catalysts
limits the ORR activity of the obtained catalysts. The metal con-
tent solely limits the activity. For metal contents above these
values, the catalytic activity first levels off for Fe, then drops
for both Fe and Co catalysts once the metal content is larger
than 1 wt.%. This behaviour essentially reflects the full occupa-
[
15] R. C oˆ t e´ , G. Lalande, D. Guay, J.P. Dodelet, G. D e´ n e´ s, J. Electrochem. Soc.
45 (1998) 2411.
1
[
16] G. Faubert, R. C oˆ t e´ , J.P. Dodelet, M. Lef e` vre, P. Bertrand, Electrochim.
Acta 44 (1999) 2589.
[17] A.L. Bouwkamp-Wijnoltz, W. Visscher, J.A.R. Van Veen, S.C. Tang, Elec-
trochim. Acta 45 (1999) 379.
[
18] M. Lef e` vre, J.P. Dodelet, P. Bertrand, J. Phys. Chem. B 104 (2000)
1238.
1
[
19] M. Bron, J. Radnik, M. Fieber-Erdmann, P. Bogdanoff, S. Fiechter, J.
Electroanal. Chem. 535 (2002) 113.
[20] M. Bron, S. Fiechter, P. Bogdanoff, H. Tributsch, Fuel Cells 2 (2003) 137.

5
984
F. Jaouen, J.-P. Dodelet / Electrochimica Acta 52 (2007) 5975–5984
[
[
[
[
[
[
21] F. Jaouen, S. Marcotte, J.P. Dodelet, G. Lindbergh, J. Phys. Chem. B 107
[33] M. Lef e` vre, J.-P. Dodelet, P. Bertrand, J. Phys. Chem. B 106 (2002) 8705.
[34] M. Lef e` vre, J.-P. Dodelet, P. Bertrand, J. Phys. Chem. B 109 (2005)
16718.
[35] F. Jaouen, F. Charreteur, J.P. Dodelet, J. Electrochem. Soc. 153 (2006)
A689.
[36] U.A. Paulus, T.J. Schmidt, H.A. Gasteiger, R.J. Behm, J. Electroanal.
Chem. 495 (2001) 134.
[37] M. Lef e` vre, J.P. Dodelet, Electrochim. Acta 48 (2003) 2749.
[38] P. He, M. Lef e` vre, G. Faubert, J.P. Dodelet, J. New Mater. Electrochem.
Syst. 2 (1999) 243.
[39] P. Gou e´ rec, A. Biloul, O. Contamin, G. Scarbeck, M. Savy, J. Riga, L.T.
Weng, P. Bertrand, J. Electroanal. Chem. 422 (1997) 61.
[40] T. Okada, M. Gokita, M. Yusada, I. Sekine, J. Electrochem. Soc. 145 (1999)
815.
[41] S. Marcotte, D. Villers, N. Guillet, L. Rou e´ , J.P. Dodelet, Electrochim. Acta
50 (2004) 179.
(
2003) 1376.
22] M. Yuasa, A. Yamaguchi, H. Itsuki, K. Tanaka, M. Yamamoto, K. Oyaizu,
Chem. Mater. 17 (2005) 4278.
23] F. Jaouen, M. Lef e` vre, J.-P. Dodelet, M. Cai, J. Phys. Chem. B 110 (2006)
5
553.
24] N.P. Subramanian, S.P. Kumaraguru, H. Colon-mercado, H. Km, B.N.
Popov, T. Black, D.A. Chen, J. Power Sources 157 (2006) 56.
25] E.B. Easton, A. Bonakdarpour, J.R. Dahn, Electrochem. Solid-State Lett.
9
(2006) A463.
26] I. Herrmann, V. Br u¨ ser, S. Fiechter, H. Kersten, P. Bogdanoff, J. Elec-
trochem. Soc. 152 (2005) A2179.
[27] R. Bashyam, P. Zelenay, Nature 443 (2006) 63.
[
28] J.-P. Dodelet, Oxygen reduction in PEM fuel cell conditions: heat-treated
non-precious metal–N4 macrocycles and beyond, in: J. Zagal, F. Bedioui,
J.-P. Dodelet (Eds.), N4-Macrocyclic Metal Complexes: Electrocatalysis,
Electrophotochemistry & Biomimetic Electroanalysis, Springer Science +
Business Media Inc., New York, 2006 (Chapter 3).
[42] A. Biloul, O. Contamin, G. Scarbek, M. Savy, B. Palys, J. Riga, J. Verbist,
J. Electroanal. Chem. 365 (1994) 239.
[
29] L. Zhang, J. Zhang, D.P. Wilkinson, H. Wang, J. Power Sources 156 (2006)
[43] P. Gou e´ rec, A. Biloul, O. Contamin, G. Scarbeck, M. Savy, J.M. Barbe, R.
Guilard, J. Electroanal. Chem. 398 (1995) 67.
1
71.
[
30] U.A. Paulus, A. Wokaun, G.G. Scherer, T.J. Schmidt, V. Stamenkovic, N.M.
Markovic, P.N. Ross, Electrochim. Acta 47 (2002) 3787.
31] S. Mukerjee, S. Srinivasan, J. Electroanal. Chem. 357 (1993) 201.
32] H. Wang, R. C oˆ t e´ , G. Faubert, D. Guay, J.P. Dodelet, J. Phys. Chem. B 103
[44] J. Lepp a¨ lathi, P. Simell, E. Kurkela, Fuel Process. Technol. 29 (1991) 43.
[45] H. Mori, K. Asami, Y. Ohtsuka, Energy Fuels 10 (1996) 1022.
[46] P. Schaaf, Progr. Mater. Sci. 47 (2002) 1.
[47] A. Biloul, F. Coowar, O. Contamin, G. Scarbek, M. Savy, D. Van den Harn,
J. Riga, J.J. Verbist, J. Electroanal. Chem. 350 (1993) 189.
[
[
(
1999) 2042.