Ptx Coy catalysts degradation in PEFC environments: Mechanistic insights: II. Preparation and characterization of particles with homogeneous composition

Article abstract of DOI:10.1149/1.3385000

In this paper, we present experimental results on the preparation and the electrochemical characterization of Pt_x Co_y electrocatalytic particles with homogeneous composition, modeled by Franco [J. Electrochem. Soc., 156, B410 (2009)]. Preparation is made through the direct liquid injection metallorganic chemical vapor deposition technique previously developed at CEA, and electrochemical analysis is carried out by using a rotating disk electrode and membrane-assembled gas diffusion electrode. Degradation structural changes are characterized by using transmission electron microscopy, X-ray diffraction, and X-ray photoelectron spectroscopy. Optimal Co compositions are identified for better oxygen reduction reaction activity and durability, validating the modeling studies.

Full text of DOI:10.1149/1.3385000

Pt x Co y Catalysts Degradation in PEFC Environments: Mechanistic
Insights : II. Preparation and Characterization of Particles with
Homogeneous Composition
Pascal Fugier, Sylvain Passot, Christelle Anglade, Laure Guétaz, Nicolas Guillet, Eric De Vito,
Sophie Mailley and Alejandro A. Franco
J. Electrochem. Soc. 2010, Volume 157, Issue 6, Pages B943-B951.
doi: 10.1149/1.3385000
Email alerting
service
Receive free email alerts when new articles cite this article - sign up in the box at the
top right corner of the article or click here
To subscribe to Journal of The Electrochemical Society go to:
http://jes.ecsdl.org/subscriptions
©
2010 ECS - The Electrochemical Society
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
B943
0
013-4651/2010/157͑6͒/B943/9/$28.00 © The Electrochemical Society
Pt Co Catalysts Degradation in PEFC Environments:
x
y
Mechanistic Insights
II. Preparation and Characterization of Particles with Homogeneous
Composition
a,z
a
a
b
Pascal Fugier, Sylvain Passot, Christelle Anglade, Laure Guétaz,
b
b
c
a, ,z
Nicolas Guillet, Eric De Vito, Sophie Mailley, and Alejandro A. Franco *
a
b
Laboratoire des Technologies des Surfaces, Département des Technologies des Nanomatèriaux, Laboratoire
des Composants pour Piles à Combustibles et de Modélisation, Département de l’Electricité et de
c
l’Hydrogène pour les Transports, and Laboratoire des Batteries Avancées, Département de l’Electricité et
de l’Hydrogène pour les Transports, Commissariat à l’Energie Atomique et aux Energies Alternatives/
Centre de Grenoble LITEN, 38054 Grenoble Cedex 9, France
In this paper, we present experimental results on the preparation and the electrochemical characterization of Pt Co electrocatalytic
x
y
particles with homogeneous composition, modeled by Franco et al. ͓J. Electrochem. Soc., 156, B410 ͑2009͔͒. Preparation is made
through the direct liquid injection metallorganic chemical vapor deposition technique previously developed at CEA, and electro-
chemical analysis is carried out by using a rotating disk electrode and membrane-assembled gas diffusion electrode. Degradation
structural changes are characterized by using transmission electron microscopy, X-ray diffraction, and X-ray photoelectron spec-
troscopy. Optimal Co compositions are identified for better oxygen reduction reaction activity and durability, validating the
modeling studies.
©
2010 The Electrochemical Society. ͓DOI: 10.1149/1.3385000͔ All rights reserved.
Manuscript submitted September 4, 2009; revised manuscript received March 15, 2010. Published May 3, 2010.
The specific activity and stability of polymer electrolyte fuel cell
PEFC͒ electrodes can be improved by alloying Pt with some tran-
of Pt–Co based on the preparation of a Co/C composite support
5
͑
followed by removing the excess Co in a H SO solution.
2
4
sition metals ͑e.g., Co, Ni, and V͒. These electrodes therefore
Other methods based on surface technologies have also been
studied such as the chemical vapor deposition ͑CVD͒ process.
Chemical deposition processes used for the formation of thin films
or supported nanoparticles are based on chemical reactions between
inorganic or organic metal-containing species, leading to the forma-
tion of a thin layer on an appropriate substrate. All these processes
are one-step processes, in general, simpler and less expensive than
physical deposition processes. They are commonly used for the pro-
duction of various coatings ͑metals, oxides, and composites͒, the
main advantage being that they allow an easy scaling up to large
size systems and a conformal coverage of complex shapes. The
chemical reactions used in CVD are numerous and include thermal
decomposition ͑pyrolisis͒, reduction, hydrolysis, and oxidation,
which can be used either singly or in combination. These reactions
can be activated by several methods, the most important one being
thermal activation and plasma activation. Depending on the applica-
tions and specifications, the reactants ͑precursors͒ have to be cor-
rectly chosen from the following general groups: halides, carbonyls,
metallorganics ͑MOs͒, and hybrids.
achieve lower overall Pt loadings compared to pure Pt-based elec-
1
trodes. In our previous paper, we have proposed a multiscale ki-
netic model describing the oxygen reduction reaction ͑ORR͒ activity
and degradation of Pt Co nanoparticles in PEFC environments.
x
y
This model provides mechanistic insights on the impact of the nano-
structure and operating conditions on the Pt Co nanoparticles dura-
x
y
bility. On the basis of ab initio data, we identify favorable pathways
of the ORR on Pt Co nanoparticles and of the competitive Pt–Co
x
y
dissolution in acidic media. The derived ab initio-based kinetics is
coupled to a description of the atomic reorganization at the nanopar-
ticle level as a function of the cumulated Pt and Co mass losses.
This nanoscale model is coupled with a transport microscale model
of charges and O through a PEFC cathode, and sensitivity simula-
2
tions to several operation conditions and initial compositions/
morphologies are performed. Two kinds of particles are simulated:
1
. Annealed Pt Co nanoparticles ͑representative of complete al-
x y
loys͒ by using the Monte Carlo method.
2
. Homogeneous Pt Co particles, i.e., with the surface compo-
In our group at CEA, a modified CVD method, the metallorganic
chemical vapor deposition ͑MOCVD͒ process, has been developed
to prepare PEFC electrodes directly on a gas diffusion layer ͑GDL͒
x
y
sition equal to the volume composition ͑representative of bulk-
truncated structures or incomplete alloys͒.
7
-9
support. MOs are compounds in which the atom of a metallic
element is bound to one or more carbon atoms of an organic hydro-
carbon group. Most of the elements used in MOCVD are from the
IIa, IIb, IIIb, IVb, Vb, and VIb groups.
In this paper, we focus on the preparation and the electrochemi-
cal characterization of the second kind of Pt Co particles. The goal
x
y
here is not to prepare highly active and stable electrocatalytic sys-
tems but to prepare model electrodes with structures/morphologies
In an MOCVD process, thin films or nanoparticles are formed on
a surface from the thermal decomposition of the gaseous precursor
molecules. The total pressure used in the gas phase is generally a
low pressure, in the range of a few Torrs.
Precursors for MOCVD should, in general, have an evaporation
rate that is high enough, stable, and constant in time. The precursor
molecule should be chemically and thermally stable during the
transport through the gas phase to the surface. The precursor should
be
that can be compared with the Pt Co homogeneous particles that
x
y
we modeled in Ref. 1.
Preparation methods for pure Pt or Pt-based electrocatalysts are
generally aqueous methods using various reducing agents. The con-
ventional method consists in depositing Pt on carbon and then Co on
2
-4
Pt/C. For example, Xiong and Manthiram synthesized Pt–M ͑M
=
Fe, Co, Ni, and Cu͒ catalysts by using a commercial Pt catalyst
2
supported on carbon black. Two other synthesis methods are pre-
sented in Ref. 5 and 6. Li et al. developed a method for the synthesis
1
2
3
. Relatively easy to synthesize.
. Not be dangerous and should not produce dangerous products.
. Must be soluble and stable in a suitable solvent without for-
*
Electrochemical Society Active Member.
E-mail: pascal.fugier@cea.fr; alejandro.franco@cea.fr
z
mation of precipitates ͑in a droplet-derived deposition process͒.
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

B944
Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
N2
Classical MOCVD processes are based on the evaporation of
solid MO precursors and meet the problem of irreproducibility of
vapor generation and transport, thus making the growth process and
consequently the film properties less reproducible. Due to these rea-
sons, new liquid delivery systems have been introduced a few years
ago to overcome the problem of thermal stability of some complexes
such as the pulsed liquid injection metallorganic chemical vapor
deposition ͑PLI-MOCVD͒ process or the direct liquid injection met-
O + N2
2
Injector
Tube containing
liquid precursors
Evaporator
1
0-12
allorganic chemical vapor deposition ͑DLI-MOCVD͒ process.
The main advantages of these liquid delivery systems are:
System of gas
flow repartition
1
. A single precursor mixture can be used for multi component
Reactor
systems.
2
3
. Very high growth rates can be obtained.
. Precursors ͑solid or liquid͒ can be kept at low temperature
Substrate
and atmospheric pressure.
. Only small amounts of materials ͑a few microliters͒ are
volatilized.
4
The evaporation of precursors with low vapor pressure even un-
Vacuum
der atmospheric pressure is possible. In PLI-MOCVD, the solution
is injected directly by a microvalve. The injector injects microdoses
Figure 1. Schematic diagram of the DLI-MOCVD process reactor.
͑
a few microliters͒ of an organic solution containing a dissolved
mixture or a pure MO liquid precursor. After flash evaporation, the
resulting vapor mixture is transported by gas toward the heated sub-
strate. The frequency, the opening time, and the solution concentra-
tion are the main parameters controlling the growth rate. To date, on
the basis of this liquid delivery system, high quality noble metal
d
support. No ionomer impregnation was applied after the catalyst
elaboration to avoid ionomer poisoning effects on the catalysts’ elec-
trochemical behavior during the experimental tests.
1
3
Structural and chemical characterization methods.— The metal
loadings in the fresh samples were measured by inductively coupled
plasma atomic emission spectroscopy ͑ICPAES͒. The morphology
and the particle size distribution were observed by electron micros-
copy ͓transmission electron microscopy ͑TEM͒ and field-emission
gun scanning electron microscopy ͑FEGSEM͔͒. Observations of the
GDL support surface after deposition were performed using a LEO
1530 FEGSEM coupled with an energy-dispersive X-ray spectrom-
etry ͑EDXS͒ analysis. TEM images and EDXS analyses were also
realized using a JEOL 2000 FX microscope. The particle structures
were observed by high resolution transmission electron microscopy
nanoparticles or oxide films have been obtained in our institute.
This method is very versatile and is thus suitable for in situ deposi-
tion of thin films or complex systems such as metal/oxide compos-
ites, carbon nanotube, nanoparticles, and multilayers.
In this paper, we discuss the structural and electrochemical prop-
erties of model Pt Co PEFC cathodes made using the DLI-
x
y
MOCVD process depositing on a carbon support Pt and Co at the
same time. Results are reported for cathodes operated under differ-
ent operating conditions and operating times. This compliments our
modeling work on Pt Co -based nanoparticles already published in
x
y
͑
͑
HRTEM͒ using a JEOL 4000 EX microscope. The X-ray diffraction
XRD͒ patterns were obtained from a Bruker D5000 diffractometer.
Ref. 1 and 14.
The paper is organized as follows. First, we present the experi-
mental preparation of the cathodes and characterization setups. Sec-
ond, we discuss the characterization results regarding the activity
and stability properties according to the Pt Co nanostructure, in
The surface of the nanoparticles was characterized by X-ray photo-
electron spectroscopy ͑XPS͒. The XPS device used in this study
consisted of an SSI S-Probe spectrometer equipped with a mono-
chromatized Al K␣ source. The pass energy was chosen to be 50
eV while the energy resolution was kept at 850 meV.
x
y
connection with our previously published modeling results. Finally,
we conclude and indicate further directions in our research.
Chemical leaching was carried out ͑24 h, H SO 0.5 M, 50°C͒
2
4
on the pure Pt catalyst and on the three Pt Co catalysts to distin-
x
y
guish the dissolution induced by the acidic media from the one
induced by the potential cycling. Then, platinum and cobalt were
also dosed by ICPAES.
Experimental
Preparation of the Pt Co electrocatalysts.— Pt Co
y
electro-
x
y
x
catalysts were prepared using the DLI-MOCVD method developed
Electrochemical characterization methods.— Our materials were
characterized by two electrochemical methods. Rotating disk elec-
trode ͑RDE͒ experiments were used first because of their easy
implementation and because they allowed us to characterize the
catalysts in the most severe conditions of stability. Half-cell experi-
ments were used for fine testing the most interesting materials under
conditions closer to the fuel cell environment. Both RDE experi-
ments and half-cell experiments were carried out directly on gas
at CEA. For the deposition of the bimetallic electrocatalysts in the
II
reactor, dimethyl͑1,5-cyclooctadiene͒ platinum͑II͒ ͓Pt ͑Me ,cod͒,
2
III
Strem͔ and cobalt͑III͒ acetylacetonate ͓Co ͑acac͒ , Strem͔ were
3
used as precursors. All precursors are soluble in toluene ͑Aldrich͒
for the used concentrations ͑Ͻ0.03 M͒. Simultaneous liquid injec-
tion was done to obtain bulk-truncated or incomplete Pt Co alloys.
x
y
The Pt Co ratio is controlled by the precursor concentration. The
x
y
precursor solutions were pushed by nitrogen. The solvent was then
evaporated and cracked by thermal effect and oxidative atmosphere
d
The DLI-MOCVD reactor is a tool dedicated for preparing Pt–Co catalysts. In our
catalysts preparation, we have implemented precursors with high purity level
͑
O 80%, N 20%͒. Both precursors were also decomposed by the
2 2
temperature and gas pressure inside the reactor ͑Fig. 1͒. Pt Co elec-
x
y
͑99.9%͒ and high quality solvents typically used in microelectronics. Based on ICP
trocatalysts were formed at 350°C on a GDL ͑ELAT LT1400W,
E-Tek BASF͒ playing the role of microporous carbon catalyst
characterizations, we have not observed any presence of impurities on the fresh
catalysts and on the GDL support.
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
B945
diffusion electrodes ͑GDEs͒ ͑GDE = GDL + catalysts deposit͒. All
the electrochemical experiments were carried out at room tempera-
ture in a 0.5 M H SO electrolyte.
W
(GDE)
Gas inlet
2
4
N
2
bubbling
CE
platinum)
Ex situ characterizations: RDE performance experiments.— The
cell used for the RDE experiments is constituted by a classical three-
electrode system with a working electrode consisting of a glassy
carbon disk ͑Sigradur Grade G͒ supporting our substrate. The
counter electrode is a reversible hydrogen electrode ͑RHE͒. Before
the tests, 1 ␮L of ethanol was put on the GDE and rinsed with
deionized water. The electrochemical three-electrode cell was con-
nected to a digital potentiostat ͑Autolab PGSTAT30͒, with an RHE
as a reference electrode.
(
REF
(
MSE)
Gas outlet
Sample
The linear sweep voltamperometry was used to allow the deter-
mination of the “electroactive catalyst surface area” ͑ECSA͒ ͑or at
least, the H adsorption properties on the catalysts͒ ͑under nitrogen
+
bubbling͒ and to quantify the properties of our catalysts regarding
the ORR ͑under oxygen bubbling͒. The effective electroactivity of
the Pt Co regarding ORR was determined from voltammetry ex-
periments ͑2 mV s ͒ between 1 and 0.4 VRHE. Before each vol-
tammetry, the electrode was kept for 1 min at a potential of 1 VRHE
to ensure a reproducible electrocatalytic surface state. During all the
experiments, the N2 or O2 concentration was kept constant at its
saturation value by permanent gas bubbling. The RDE angular rota-
tion rate ␻ was varied in the 200–900 rpm range. Levich approach is
used to calculate the current limited by the convective transport
x
y
Figure 2. Representation of the three-electrode device for GDE testing ͑half-
cell͒.
−
1
carried out at I = 0 A ͑open circuit͒ for 7 days ͑160 h͒ at 60°C in
H SO /0.5 M saturated O electrolyte and using a saturated calomel
2
4
2
electrode as a reference electrode connected to an EGG Par 273
digital potentiostat. The same electrochemical characterizations as
described above were done before and after aging.
͑
voltammetry sigmoidal wave height or Levich current͒
I ͑␻͒ = 0.620 ϫ n ϫ F ϫ A ϫ D ϫ v−1/6 ϫ C ϫ ␻
2
/3
−1/2
In situ characterizations: half cell performance experiments.— A
GDE ͑14 mm diameter disk͒ and a Nafion NRE 212 CS membrane
͑DuPont͒ were directly bonded together by a hot-pressing process
L
g
0
=
B ϫ C ϫ ␻−
1/2
͓1͔
0
͑
10 MPa at 150°C for 2 min and 30 s͒. The obtained membrane
where n is the number of electrons transferred in the ORR, F is the
electrode assembly ͑MEA͒ was inserted in the sample holder with
the GDL facing up ͑Fig. 2͒. The membrane is then in contact with
the liquid electrolyte ͑0.50 cm ͒ while the same surface area of the
GDL is fed with the reactive gas. A gold grid provides the electrical
contacts and a polytetrafluoroethylene ͑PTFE͒ gasket set on the
membrane side ensured the tightness of the device.
Faraday constant, A the is the geometric area of the RDE, D is the
g
17
O diffusion coefficient, v is the kinematic viscosity of the electro-
2
2
lyte, C is the O concentration in solution at the bulk level just
0
2
outside the nanoscopic diffusion layer developed on the catalyst
nanoparticle surface ͑see our atomistic-based nanoscale model in
Ref. 1, 14, and 15͒, and B is the Levich coefficient. The electrode
steady-state current j͑E,␻͒ ͑A m−2 geometric͒ at a given electrode
A three-electrode, two-compartment setup containing 40 mL of
H SO solution with continuous nitrogen bubbling was used. A
2
4
potential E and angular rotation rate ␻ is given by the following
equation
platinum mesh was used as a counter electrode, and the reference
electrode was in mercury sulfate ͑mercury sulfate electrode
Hg/Hg SO /0.5 M H SO radiometer͒. Electrochemical experi-
ments were conducted with a digital Bio-logic VMP2 potentiostat.
Polarization curves were captured following this test scheme: While
the oxygen flow was set at 2 nL h , six voltammetry cycles from
0
1
1
1
1
1
2
4
2
4
=
+
=
+
͓2͔
nB␻1/2
j͑E,␻͒ jk jL͑␻͒ jk
where j is the cathodic activation current density at a potential E
−1
k
and j ͑␻͒ is the transport-limited current density in the voltammo-
−1
L
.05 to 1.15 V
at 20 mV s and one voltammetry cycle at
RHE
−
2
gram ͑A m geometric͒ derived from Eq. 1.
−1
5
mV s were performed.
Results and Discussion
Microstructural characterization results on fresh electrodes and
The RDE voltammogram results presented in this paper display
current densities corrected ͑jcorrected͒ by the diffusion in solution
using the classical Levich method ͑for high ORR Butler–Volmer
overpotentials, i.e., for E Ͼ 0.8 VRHE
͒
aged electrodes.— Atomic Pt/Co volumetric ratios were calculated
from ICPAES measurements and compared to EDXS results ͑Table
I͒. Our DLI-MOCVD process allowed us to obtain very low metal
jL͑␻͒ ϫ j͑E,␻͒
jL͑␻͒ − j͑E,␻͒
jcorrected
=
͓3͔
2
loadings ͑Ͻ100 ␮g/cm ͒, making possible materials economies
Ex situ characterizations: RDE aging test.— Our RDE aging test is
and decreasing cost for possible technology commercialization.
Microstructural characterizations.— Figure 3 shows the FEGSEM
images of the four fresh materials prepared by DLI-MOCVD.
Atomic ratios were estimated by EDXS. On the pure Pt electrode,
analogous to Colón–Mercado and Popov’s accelerated durability
1
6
test where they used a three-electrode system with a Hg/Hg SO
2
4
electrode as the reference electrode. Our aging experiments were
Table I. Compositions of the prepared Pt Co model electrodes.
x
y
Pt ICPAES loadings
Co ICPAES loadings
2
2
͑
␮g/cm ͒
͑␮g/cm ͒
Pt/Co ratio calculated from ICPAES values
EDXS ͑PtL␣1/CoK␣1͒
Pt
Pt Co
PtCo
PtCo3
61
79
68
29
0
100/0
72/28
55/45
17/83
9.3
17
42
76/24
61/39
23/77
3
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

B946
Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
HRTEM images of Pt Co ͑Fig. 5a͒ show particle size of around
–10 nm. The nanoparticles do not have any specific shape. Further-
3
7
more, an amorphous oxide layer seems to be present on the nano-
particles ͑Fig. 5b͒.
Figure 6 presents the morphology change after the electrochemi-
cal aging described in the Electrochemical operation and postaging
results section. Before aging, the catalyst particles present a homo-
geneous distribution. After degradation, the electrode appears to be
more porous ͑higher roughness at the GDL surface͒, in agreement
with our modeling results in Ref. 1 and 14 where the Pt dissolution
rate is predicted to be considerably lower than the Co dissolution
rate when the two materials coexist at the catalyst ͑the Pt dissolution
rate increases as the Co content at the particle level decreases͒ and
the homogeneous catalyst become more and more like a skeleton-
structured particle ͑the transition element dissolves leaving “holes”
in a Pt matrix, like a “French emmental cheese”͒.
XRD structure characterization of the fresh materials.— XRD
characterizations provide information about the catalyst crystallo-
graphic structure. The diffractograms of all the fresh materials are
presented in Fig. 7. The first narrow peak corresponds to the PTFE
and the second peak to graphitic carbon, both materials constitutes
Figure 3. FEGSEM images of the four Pt Co materials prepared by DLI-
MOCVD.
x
y
the GDL. Pure Pt, Pt Co, and PtCo diffractograms show the main
3
peaks of the face-centered cubic ͑fcc͒ Pt structure. The diffracto-
gram of Pt Co also shows a slight shift to higher angles indicating
3
nanoparticles are well-defined and the carbon support is still visible
between them. When cobalt is added to platinum, depending on the
Pt/Co ratio, the aspect is closer to pure platinum or to pure cobalt,
the later growing with a “cauliflower” aspect, probably due to cobalt
oxide.
Co insertion inside the Pt crystal lattice. The cobalt atomic fraction
alloyed with platinum is calculated from Vegard’s law for lattice
parameters between a pure fcc platinum and a pure hexagonal close-
packed ͑hcp͒ cobalt
Figure 4 reports the coupled microtome EDXS analysis of the Pt
and Co contents inside the GDL. The deposit is present only in the
first 5 ␮m in the GDL thickness with a nonhomogeneous distribu-
tion. There is a “crust” at the GDL surface level with a composition
close to the expected atomic Pt–Co ratio. However, in depth, the Pt
loading appears to be higher.
asample − aPt,fcc
Co
x
=
͓4͔
sample
ͩ
ͪ
aCo,hcpͱ
2 − aPt,fcc
Where asample, aPt,fcc, and a
are the lattice parameters of the
Co,hcp
sample ͑3.893 nm͒, the pure fcc platinum ͑3.923 nm͒, and the pure
hcp cobalt ͑2.506 nm͒. We find that only about 8% of cobalt is
Figure 4. ͑Color online͒ ͓͑a͒, ͑b͒, and ͑c͔͒
EDXS results on fresh Pt Co catalysts
3
presented with their associated TEM im-
ages. The spectrum ͑d͒ corresponds to the
image ͑b͒. The Cu element appears be-
cause of the TEM grid.
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
B947
XPS surface characterization of the fresh electrodes.— The forma-
tion of an oxide layer on the Pt Co particles should be strongly
x
y
favored with respect to the conditions of preparation and the oxida-
tion state of the cobalt precursor. This hypothesis is confirmed by
1
8-21
the XPS analyses: metallic cobalt ͑778–778.5 eV͒
is not de-
tected on fresh Pt Co samples. Instead, a wide peak appears at 781
x
y
eV ͑Fig. 8a͒, showing the presence of a thick layer of Co oxides
2
2,23
͑
ϳ781 eV͒.
However, the abrasion of this oxide layer leads to
the apparition of a peak at 778.9 eV ͑considering the calculated
sputtering yields of Pt, Co, and O elements for 500 eV Ar ions,
2
4
preferential abrasion of oxygen is negligible͒ ͑Fig. 8b͒. This result is
coherent with our expectations and our model, which assumes initial
Figure 5. HRTEM images of Pt Co nanoparticles elaborated by DLI-
MOCVD on a GDL support.
3
oxide coverage in Ref. 1 and 14, as is the case for Pt Co in Ref. 25.
3
XPS surface characterization of the aged electrodes.— XPS analy-
sis on aged electrodes showed unambiguously that the Co disap-
pears from the nanoparticle composition after RDE aging experi-
ments, regardless of their initial composition. Figure 9 compares the
alloyed with platinum in the fresh Pt Co sample. The result is analo-
3
gous for the fresh PtCo sample. The diffractogram of fresh PtCo3
does not show the same main peaks. The peaks of the platinum
pattern are lost within the GDL background signals.
survey spectra recorded before and after aging for the PtCo sample.
3
After electrochemical aging, no Co peak ͑around 780 eV͒ is visible.
These characterization results provide a signature of homoge-
neous composition ͑bulk truncated or incomplete alloy͒ inside these
1
4
As presented in our previous paper, the intensity of the XPS Pt
signal after aging depends on the initial composition. For pure Pt,
the signal intensity is smaller meaning less metal is visible on the
surface after than before. For the three bimetallic catalysts, because
of the Co dissolution and of its disappearance from the surface, the
Pt signal intensity is greater after than before.
1
,14
catalysts.
In addition, the XPS analysis after abrasion confirms the hypoth-
esis of a homogeneous profile ͑bulk-truncated structure or incom-
plete alloy͒ in the particles.
Figure 10 shows the O 1s spectra for the three bimetallic mate-
rials. Before aging ͑Fig. 10a-c͒, two contributions are measured, and
after aging, only one is measured ͑Fig. 10d͒. Considering Fig. 10,
the Co is totally dissolved and no signal is observed. This informa-
tion confirms that on fresh materials, the Co is oxidized ͑the peak at
Figure 6. FEGSEM images of the Pt Co material elaborated by DLI-
MOCVD before and after aging.
3
5
30 eV, which corresponds to oxygen bonded to cobalt disappears
when the cobalt peaks disappear͒.
Electrochemical operation and post-aging results.— Chemical
leaching.— By dosing the Pt and Co dissolved in the acid solution
and by comparing these results with the initial loading values, we
are able to determine the quantity of catalyst dissolved from the
carbon support. The Pt dissolution is inferior to 1% in the four cases.
For the three Pt Co catalysts, the Co quantity dissolved is close to
x
y
5
0% ͑Pt Co: 58%, PtCo: 54%, and PtCo : 49%͒. Our analysis shows
3
3
that metal dissolution is caused by two different effects: acidic me-
dia and potential cycling. Thus, even for an initially well-alloyed
material, chemical leaching can possibly take place in the PEFC
environment that can lead to an electrode enriched in Pt on the
surface before polarization. This phenomenon, occurring without
polarization, is not explicitly accounted for in our model in Ref. 1
and 14 ͑focused on understanding the material degradation under
polarization conditions, including open-circuit voltage conditions,
i.e., zero total current͒ but has been recently modeled by ourselves
in Ref. 26.
Figure 7. ͑Color online͒ XRD spectra of the Pt Co materials deposited on
x
y
the GDL by DLI-MOCVD.
b
a
Figure 8. ͑Color online͒ XPS spectra of
Co2p 3/2 ͑a͒ before and ͑b͒ after Ar sput-
tering ͑PAr = 2.10− mbar, 2kV͒ the
7
PtCo sample surface.
3
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

B948
Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
Rigorously speaking, the “210” value in Eq. 5, derived from an
on-top’ adsorption on Pt atoms is unjustified with the Pt Co cata-
“
x
y
+
lysts ͑H can adsorb simultaneously on-top and in bridge between Pt
and Co atoms at the Pt Co catalyst surface and, thus, in a different
x
y
way compared to pure Pt surfaces͒. Detailed density functional
+
theory ͑DFT͒ studies on adsorption properties of H on Pt M sur-
x
y
faces are ongoing in our group and are the subject of an ongoing
3
0
publication.
In the zone of adsorption/desorption, some hydrogen was situ-
ated between 0.05 and 0.4 VRHE, three different peaks appeared for
all of the considered bimetallic electrodes. These peaks are repre-
sentative of the adsorption of H+ on the ͑110͒, ͑111͒, and ͑100͒
crystallographic faces of the platinum and they are very marked
3
1
͑
according to Paulus et al., the insertion of a transition metal leads
to poor definition of these peaks͒. This can indicate that the surface
of the particles is Pt-rich.
Figure 12b presents the current densities regarding the ORR as
function of the Pt loading for the four tested materials. The current
densities provided by the different systems can be ranked as
Pt Co Ͼ PtCo Ͼ Pt Ͼ PtCo . We emphasize here that the pure Pt
Figure 9. ͑Color online͒ XPS survey spectra of the PtCo3 sample, before
3
3
͑
green͒ and after ͑red͒ aging.
catalyst does not provide the highest current, in contrast to what it
could be expected from the ECSA measurement in Fig. 12a. This
clearly illustrates that Hupd is not an appropriate technique to evalu-
+
Ex situ characterizations: RDE performance experiments.— Figure
1 shows cyclic voltamperograms at the beginning of life ͑BoL͒ and
ate the ECSA of the PtxCoy catalysts, as only accounts for H ad-
1
sorption on Pt sites and as ORR elementary mechanisms on PtxCoy
1
,14
after the aging of the four tested electrocatalysts. At the BoL, these
involve both Pt and Co sites.
Pt Co materials exhibit very different effective ECSA lower to the
pure Pt catalyst ͑Fig. 12a͒.
The ECSA were calculated by measuring the peak area in the
voltammograms and using the following equation
x
y
Ex situ characterizations: RDE aging test.— Figure 11 also shows
the four cyclic voltamperometries after electrochemical aging of the
DLI-MOCVD electrodes at a high fixed potential ͑low current con-
ditions͒ in the presence of O . The highest current density loss is
2
QH A /␯
measured for the pure Pt catalyst ͑Fig. 13͒, and Pt3Co results to be
the most stable material, in agreement with our modeling predictions
H
ECSAHupd
=
=
͓5͔
qR
where AH is the hydrogen peak desorption area ͑C V s ͒, ␯ is the
qR
in Ref. 1 and 14. Pt Co and PtCo GDE are both still better than pure
3
−
1
Pt GDE after aging, in agreement with our model results for short
−
1
2
1
scan rate ͑mV s ͒, and q is the adsorption charge per cm of
electrode operation time ͑for longer times, Ϸ12 days, according to
R
2
platinum on a smooth Pt electrode ͓assumed to be 210 ␮C/cm for
our model, the Pt Co and PtCo performances could become lower
3
27-29
a hydrogen monolayer on Pt͑111͒
͔.
than pure Pt͒.
We emphasize that this methodology to determine the ECSA
based on Eq. 5 only provides a relative comparison of the materials.
The PtCo electrocatalyst has the lowest current density among
the four materials, and it shows the higher degradation of the three
3
Figure 10. ͑Color online͒ ͓͑a͒, ͑b͒, and
͑
c͔͒ XPS oxygen 1s spectra for the three
bimetallic electrocatalysts and ͑d͒ super-
position of the spectra of the PtCo mate-
rial before and after electrochemical ag-
ing.
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
B949
Figure 11. ͑Color online͒ Cyclic volta-
mperograms for the four materials elabo-
rated by DLI-MOCVD, before and after
aging at a potential of 0.85 VRHE during
60 h. Nitrogen bubbling, H SO 0.5 M,
1
1
2
4
00 mV/s, 900 rpm.
bimetallic materials, in agreement with our modeling predictions in
Ref. 1 where we demonstrated, on the basis of ab initio-based el-
ementary kinetics, that the Co dissolution rate is higher ͑more ther-
modynamically favorable in acidic media͒ than the Pt dissolution
rate and that the Co dissolution rate is higher for higher Co volume
contents. This explains why the stability decreases as the Co volume
content increases ͑i.e., Pt Co Ͼ PtCo Ͼ PtCo , ranking built here
The relatively high Pt ECSA suggests that the catalyst layer is
very thin ͑in agreement with the microtome analysis discussed in the
Microstructural characterization results on fresh electrodes and aged
electrodes section͒ and that a great part of the particles are electro-
chemically active.
Cyclic voltammetry experiments under O2 were performed on
each sample, prepared and tested in the same conditions to facilitate
a direct comparison of the performance of the various catalysts.
Electrochemical impedance spectroscopy ͑EIS͒ measurements
were also performed on the samples at four current densities: 0.1, 2,
3
3
on the basis of the “current loss” signature͒.
In situ characterizations: half cell performance experiments.— As
in RDE experiments, cyclic voltammetry was used here to evaluate
the ECSA of the catalysts. In this case, the ECSA are calculated
from the hydrogen desorption peak area and from the carbon oxide
−2
−2
1
0, and 20 mA cm , applying a 0.2 mA cm amplitude sinus
signal from 10 kHz to 0.2 Hz. Oxygen reduction current curves are
presented in Fig. 14. The potentials are corrected for the ohmic loss
with the values of the electrolyte resistance, R⍀ ͑cf. Fig. 16͒ previ-
ously determined by EIS measurements. A semilogarithmic repre-
sentation ͑Tafel plot͒ is given in Fig. 15 showing that the linear
evolution of the potential is the same for the three bimetallic cata-
͑
CO͒ stripping peak area. The voltammograms recorded in 0.5 M
−
1
H SO at 298 K and 20 mV s scan rate after CO adsorption on
2
4
pure Pt and illustrate the charges corresponding to desorption of
Hupd ͑Q ͒ and CO ͑Q ͒. The CO stripping ECSA is calculated
H
CO
2
from Eq. 5 with q = 420 ␮C/cm representing the charges re-
quired to oxidize a monolayer of CO on a Pt electrode.
R
2
8,32
As for
−1
lysts. The slopes are close to 130 mV dec whereas it comes near
−1
3
5
the Hupd methodology, we emphasize that this method only provides
a relative comparison of the materials. Detailed DFT studies on the
adsorption properties of CO ͑for the three compositions of PtCo͒
surfaces are ongoing in our group and are the subject of a future
to 120 mV dec for pure Pt.
The reduction current of the different catalysts at 0.9 VRHE are
given in Table III. Oxygen reduction current corrected with the Pt
loading shows similar values for the three PtCo catalysts. These
values are closely twice more than for the pure Pt catalyst, showing
3
0,33
publication.
From the calculated values reported in Table II, one can note the
quite good agreement of the results obtained by the two methods. It
the improvement on the ORR activity induced by the presence of
1,14
Co, as already demonstrated with our model.
Evolutions of the
also appears that the values of the roughness factor S /S
cat geo
transfer resistance for oxygen reduction vs current density, deter-
mined by impedance spectroscopy, are given in Fig. 16.
From these figures, it can be concluded that the best perfor-
mances are given by the highest Pt-loaded bimetallic samples, in
agreement with our modeled polarization curves in Ref. 1 and 14.
2
2
͑
cm Pt/cm ͒ are very low compared to typical values reported on
2
2
34
MEAs ͑200–300 cm Pt/cm ͒. However, the Pt specific surface
2
area APt ͑m /gPt͒ is of the same order of magnitude such as MEAs,
2
34
typically 50–80 m /gPt.
(a)12000
(b) 12
Pt
Pt Co
3
1
0000
10
8
PtCo
8
6
4
2
000
000
000
000
0
Pt Co
3
PtCo
Figure 12. ͑a͒ ECSA and ͑b͒ current den-
sity measured at a potential of 0.85 VRHE
plotted as function of Pt loadings for the
four prepared materials.
6
PtCo3
40
Pt
4
PtCo3
40
2
0
0
20
60
80
100
0
20
60
80
100
Platinum loading (µg/cm²)
Platinum loading (µg/cm²)
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

B950
Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
1
2
0
8
6
4
2
0
1.1
-
30 %
Fresh
Aged
1
1.05
1
b
= -130mV.dec-1
-
34 %
b
= -120mV.dec-1
0.95
0.9
0.85
0.8
-
47 %
-
42 %
Pt 61µg/cm²
PtCo 79-9µg/cm²
PtCo 68-17µg/cm²
PtCo 29-42µg/cm²
Pt : 61
µgPt/cm²
Pt3Co : 79
µgPt/c m²
PtCo : 68
µgPt/c m²
PtCo3 : 29
µgPt/cm²
0.75
0
.7
0
Figure 13. ͑Color online͒ Experimental aging results ͑using Levich equa-
tion͒ for Pt Co and elaborated by DLI-MOCVD and aged at 0 A during 160
.001
0.01
0.1
j / mA.cm
1
10
x
y
-2
h. The current densities presented here were measured at a potential of
.85 VRHE
0
.
Figure 15. ͑Color online͒ Semilogarithmic representation ͑Tafel plot͒ of the
iR-corrected oxygen reduction current.
Table II. ECSA and platinum surface area of the prepared cata-
lysts calculated by hydrogen adsorption (Hupd) and CO strip-
ping voltammograms.
Table III. Reduction current of the different catalysts at
0
.9 VRHE.
ECAHupd
ECACO
Pt
Pt Co
PtCo
PtCo3
3
Scat/Sgeo
Pt
A
Scat/Sgeo
Pt
A
Pt
͑m /g ͒
Pt
Jgeo at 0.9 V͑␮A/cm2͒
418
6.8
19
1057
13.4
37
1057
15.5
41
360
12.4
23
2
2
2
2
2
2
Catalyst
Pt
͑cm /cm ͒
͑m /g ͒
͑cm /cm ͒
Pt
Pt
Jm at 0.9 V͑A/gPt͒
2
Pt
JS at 0.9 V͑␮A/cm ͒
20
28
25
15
29
40
37
22
23
29
27
16
34
43
40
24
Pt Co
3
PtCo
PtCo3
Conclusions
In this paper, we report the experimental results on the prepara-
tion and the electrochemical characterization of Pt Co electrocata-
x
y
Compared to these catalysts, pure Pt shows a 30–40 mV shift to-
ward the anodic potential. Then, the low Pt loaded bimetallic sample
curve is still further shifted of approximately 40 mV. The following
activity rank can be built, in excellent agreement with our previ-
ously published modeling results: PtCo Ͼ Pt Co Ͼ Pt Ͼ PtCo .
lytic particles with homogeneous composition, which we previously
modeled in Ref. 1 and 14. We demonstrate that DLI-MOCVD elabo-
rated Pt Co catalysts supported on a GDL substrate are interesting
x
y
model systems to understand the electroactivity and stability prop-
erties of Pt Co bulk-truncated or incomplete alloy structures. Elec-
3
3
x
y
According to our model, the surface accumulation of HO species is
trochemical analysis of the samples has been carried out by using
RDE and half-cell experiments. Fresh catalysts and structural
changes were characterized by using TEM, XRD, and XPS. An op-
timal Co composition is identified for better ORR activity and cata-
lyst durability, validating our previously published multiscale model
results with kinetic parameters fully obtained by using ab initio DFT
data.
2
higher for PtCo compared to Pt Co because of the higher Co sur-
3
3
face composition; thus, PtCo has a lower overall ORR activity. The
3
number of possible surface transfers of ad-oxygenated species from
Pt to Co is maximum at a surface composition of 50–50%, and this
1
explains why homogeneous PtCo offers better ORR activity.
0
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1000
Pt 61µg.cm-2
-50
PtCo 79-9µg.cm-2
PtCo 68-17µg.cm-2
-
100
1
00
10
1
PtCo 29-42µg.cm-2
Pt 61µg/cm²
PtCo 79-9µg/cm²
PtCo 68-17µg/cm²
PtCo 29-42µg/cm²
-150
-200
-250
-300
0
5
10
15
20
25
EiRcorr / V/ENH
Current Density / mA.cm-2
Figure 14. ͑Color online͒ i -corrected voltammograms ͑0.5 M H SO ,
R
2
4
−
1
2
5°C, sweep rate of 20 mV s ͒. Oxygen reduction current density vs ap-
Figure 16. ͑Color online͒ Evolution of the transfer resistance for oxygen
plied potential.
reduction obtained by impedance spectroscopy vs current density.
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org

Journal of The Electrochemical Society, 157 ͑6͒ B943-B951 ͑2010͒
B951
8
9
. S. Mailley, P. Capron, S. Thollon, and T. Krebs, WO Pat. WO 2007/088292 A1
2007͒.
. S. Mailley, F. Sanchette, S. Thollon, and F. Emieux, WO Pat. 2007/088291 ͑2007͒.
These results provide a first indication on the performance and
durability of catalysts elaborated by DLI-MOCVD in conditions
͑
x
close to a PEFC cathode. Even if we noted degradation of the Pt Coy
10. J. P. Sénateur, R. Madar, F. Weiss, O. Thomas, and A. Abrutis, French Pat. 93/
under the RDE aging experiments, the Co element apparently slows
08838 ͑1993͒ and extended: European Pat. FR94/0000858 ͑1994͒.
1. F. Felten, J. P. Senateur, F. Weiss, R. Madar, and A. Abrutis, J. Phys. IV, C5, 1079
͑1995͒.
2. J. P. Senateur, F. Felten, S. Pignard, F. Weiss, A. Abrutis, V. Bigelyte, A. Teiserskis,
Z. Saltyte, and B. Vengalis, J. Alloys Compd., 251, 288 ͑1997͒.
13. S. Thollon, F. Sanchette, and J. Barrault, Abstract 87, The Electrochemical Society
1
down the mobility of the Pt on the carbon support, therefore de-
1
creasing the electrochemical ripening phenomena. Furthermore, the
1
degree of graphitization of the GDL support possibly plays a role on
the activity and stability properties of the prepared catalysts, and this
will be the focus of further studies in our group.
Meeting Abstracts, Vol. 501, Quebec City, Canada, May 15–20, 2005.
1
4. A. A. Franco, S. Passot, C. Anglade, E. Billy, L. Guetaz, N. Guillet, E. De Vito, P.
Fugier, and S. Mailley, ECS Trans., 13͑17͒, 29 ͑2008͒.
5. A. A. Franco, P. Schott, C. Jallut, and B. Maschke, J. Electrochem. Soc., 153,
A1053 ͑2006͒.
16. H. R. Colón-Mercado and B. N. Popov, J. Power Sources, 155, 253 ͑2006͒.
7. N. Guillet, L. Roué, S. Marcotte, D. Villers, J. P. Dodelet, N. Chhim, and S. Tré
Vin, J. Appl. Electrochem., 36, 863 ͑2006͒.
18. N. S. McIntyre and M. G. Cook, Anal. Chem., 47, 2208 ͑1975͒.
19. I. Alstrup, I. Chorkendorff, R. Candia, B. S. Clausen, and H. Topsøe, J. Catal., 77,
Moreover, DLI-MOCVD method can reveal an interesting cata-
lyst preparation technique to achieve ultralow catalyst loadings for
PEFC cathodes. To prepare electrochemically stable Pt Co cata-
1
x
y
lysts, this technique could be combined with thermal annealing ͑T
Ͼ 400°C͒ but other carbon supports than GDLs ͑not thermically
stable at T Ͼ 400°C͒ should be used. More generally, the prepara-
1
tion of model electrodes and
a
combined experimental
3
97 ͑1982͒.
electrochemistry/modeling approach provide interesting insights on
2
2
0. S. L. T. Andersson and R. F. Howe, J. Phys. Chem., 93, 4913 ͑1989͒.
1. A. B. Mandale, S. Badrinarayanan, S. K. Date, and A. P. B. Sinha, J. Electron
Spectrosc. Relat. Phenom., 33, 61 ͑1984͒.
the understanding of the Pt Co electrocatalyst nanostructure impact
x
y
on their activity and stability properties. Within this context, in an
36
incoming paper we are reporting the preparation of particles with a
nanostructure similar to the Monte Carlo-based one modeled in Ref.
22. M. Oku and K. Hirokawa, J. Electron Spectrosc. Relat. Phenom., 8, 475 ͑1976͒.
2
3. B. J. Tan, K. J. Klabunde, and P. M. A. Sherwood, J. Am. Chem. Soc., 113, 855
1991͒.
͑
1.
2
2
2
4. M. P. Seah, Thin Solid Films, 81, 279 ͑1981͒.
5. U. Bardi, B. C. Beard, and P. N. Ross, J. Catal., 124, 22 ͑1990͒.
6. A. A. Franco, Abstract 3016, The Electrochemical Society Meeting Abstracts, Vol.
Acknowledgments
9
02, Vienna, Austria, Oct 4–9, 2009.
This work was funded by the ANR PAN-H ͑French National
Research Agency͒ within the context of the OPTICAT project.
2
2
7. J. Wu, X. Z. Yuan, H. Wang, M. Blanco, J. J. Martin, and J. Zhang, Int. J. Hydro-
gen Energy, 33, 1735 ͑2008͒.
8. T. Vidakovic, M. Christov, and K. Sundmacher, Electrochim. Acta, 52, 5606
Commissariat à l’Energie Atomique et aux Energies Alternatives assisted
in meeting the publication costs of this article.
͑2007͒.
2
3
9. A. Lasia, J. Electroanal. Chem., 562, 23 ͑2004͒.
0. R. Ferreira de Morais, D. Loffreda, P. Sautet, and A. A. Franco, ECS Trans.,
References
2
5͑24͒, 167 ͑2010͒.
1
. A. A. Franco, S. Passot, C. Anglade, E. Billy, L. Guetaz, N. Guillet, E. De Vito, P.
Fugier, and S. Mailley, J. Electrochem. Soc., 156, B410 ͑2009͒.
. L. Xiong and A. Manthiram, J. Electrochem. Soc., 152, A697 ͑2005͒.
. P. Yu, M. Pemberton, and P. Plasse, J. Power Sources, 144, 11 ͑2005͒.
. S. Koh, C. Yu, P. Ma, I. R. Srivastava, and P. Strasser, J. Power Sources, 172, 50
31. U. A. Paulus, A. Wokaun, G. G. Scherer, T. J. Schmidt, V. Stamenkovic, N. M.
Markovic, and P. N. Ross, Electrochim. Acta, 47, 3787 ͑2002͒.
32. H. Gasteiger, N. Markovic, P. Ross, and E. Cairns, J. Phys. Chem., 98, 617 ͑1994͒.
33. S. K. Cheah, P. Gélin, O. Lemaire, and A. A. Franco, ECS Trans., 25͑35͒, 275
͑2010͒.
2
3
4
͑
2007͒.
34. H. A. Gasteiger, S. S. Kocha, B. Sompalli, and F. T. Wagner, Appl. Catal., B, 56,
9 ͑2005͒.
5
. X. Li, H. R. Colón-Mercado, G. Wu, J.-W. Lee, and B. N. Popov, Electrochem.
Solid-State Lett., 10, B201 ͑2007͒.
35. http://www.osti.gov/energycitations/servlets/purl/823267-aDW6iD/823267.PDF,
last accessed September 3, 2009.
36. E. Quenel, P. Fugier, R. Bouchmila, A. A. Franco, E. Paulliac-Vaujour, L. Guetaz,
O. Sicardy, and V. Muffato, In preparation.
6
7
. S. C. Zignani, E. Antolini, and E. R. Gonzalez, J. Power Sources, 182, 83 ͑2008͒.
. S. Thollon, F. Sanchette, S. Valange, E. Guelou, J. Barrault, and H. Guillon, WO
Pat. 2002/070130 ͑2002͒.
Downloaded on 2012-11-12 to IP 138.87.11.21 address. Redistribution subject to ECS license or copyright; see www.esltbd.org