Available online at www.sciencedirect.com
Electrochimica Acta 53 (2007) 1604–1617
Pd-Co carbon-nitride electrocatalysts for polymer electrolyte fuel cells Vito Di Noto a,b,∗,1 , Enrico Negro a , Sandra Lavina a , Silvia Gross b , Giuseppe Pace c a
b
Dipartimento di Scienze Chimiche, Universit`a di Padova, Via Marzolo 1, I-35131 Padova (Pd), Italy Istituto di Scienze e Tecnologie Molecolari, ISTM-CNR, Dipartimento di Scienze Chimiche, Via Marzolo 1, I-35131 Padova (Pd), Italy c Istituto di Scienze e Tecnologie Molecolari, ISTM-CNR c/o Dipartimento di Processi Chimici dell’Ingegneria, Via Marzolo 9, I-35131 Padova (Pd), Italy Received 30 December 2006; received in revised form 11 May 2007; accepted 13 May 2007 Available online 18 May 2007
Abstract In this paper is described the preparation of new platinum-free Pd-Co carbon-nitride electrocatalysts (Pd-Co-CNs) for application in lowtemperature fuel cells. Two groups of materials with formula Kn [Pdx Coy Cz Nl Hm ] were synthesized, which are grouped in two ensembles: the first is characterized by a molar ratio y/x > 1 (I), and the second by y/x < 1 (II). Kn [Pdx Coy Cz Nl Hm ] materials were prepared through a two-step synthesis protocol. The effect of the Pd/Co molar ratio and of the temperature of the thermal treatments on the structure and properties of the products were studied extensively by thermogravimetry, scanning electron microscopy, and vibrational (FT-IR and micro-Raman) and XPS spectroscopy. Vibrational studies revealed that I and II systems consist of two polymorphs of ␣- and graphitic-like carbon-nitride nanomaterials. The electrochemical activity towards the oxygen reduction reaction (ORR) and the hydrogen oxidation reaction (HOR) was measured by cyclic voltammetry measurements with thin-film rotating disk electrode (CV-TF-RDE). The electrochemical performance of Pd-Co-CNs of group I obtained at tf ≥ 700 ◦ C resulted higher than that measured for a platinum-based commercial electrocatalyst in terms of both the activity towards the ORR and the HOR and of the resistance towards the poisoning effect of methanol towards the ORR. © 2007 Elsevier Ltd. All rights reserved. Keywords: PEM fuel cells; Pd-Co carbon-nitride electrocatalysts; Vibrational spectroscopy; XPS; CV-TF-RDE method
1. Introduction Fuel cells are a class of devices used for the conversion of chemical energy into electrical energy. In recent years they have drawn a considerable amount of interest from both the scientific community and the industry. In particular, proton electrolyte membrane fuel cells (PEMFCs) operating at low temperatures (under 150 ◦ C) have shown great promise in the automotive industry [1] and for application in portable electronics [2]. In fact, after decades of improvements the internal combustion engines have almost reached their maximum theoretical efficiency (≈40%) [3]. In principle, fuel cells can: (a) reach much higher energetic yields (>65%) [3]; (b) damage much less the environment due to their reliance on clean, renewable fuels such as hydrogen and methanol. The two types of fuel cells which have drawn much interest are those fuelled with hydrogen and
∗ 1
Corresponding author. Tel.: +39 0498275229. E-mail address:
[email protected] (V. Di Noto). Active ECS member.
0013-4686/$ – see front matter © 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.electacta.2007.05.028
with methanol (direct methanol fuel cells, DMFCs). PEMFCs show the highest efficiency due to their high operation potential and the lack of any byproducts [3]. Nevertheless, they require bulky humidification systems and heat management modules for an optimal performance, thus increasing the weight and the cost of the final power plant [4,5]. DMFCs are characterized by a lack of these accessory modules and as such they weigh less, occupy a lower volume, have a simpler structure and in principle can be easily miniaturized [3,6,7]. In this paper are described the synthesis, the structural characterization and the electrochemical performance of platinum-free electrocatalysts for the oxygen reduction reaction (ORR) and the hydrogen oxidation reaction (HOR) in PEMFCs and DMFCs. Palladium was chosen as the active element due to: (a) its relatively low overpotential for the ORR [8]; (b) its lower cost, between three to four times lower than that of platinum [9]; (c) its tolerance towards the poisoning effect caused by the methanol crossover through the membrane [10]. In this report, two groups of electrocatalysts based on palladium-cobalt carbon-nitrides were prepared by thermal decomposition of metallorganic precursors obtained through alkoxide-free sol–gel and gel–plastic reactions [11–14].
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
Precursors were prepared according to a procedure previously proposed for the preparation of Pt-Ni carbon-nitride electrocatalysts for PEMFCs [15,16]. The crucial parameters to modulate in the synthesis protocol were: (a) the reagent stoichiometry in the preparation of the precursors and (b) the temperature of the final thermal treatments, tf . Two ensembles of electrocatalysts with formula Kn [Pdx Coy Cz Nl Hm ] were prepared, characterized by: (a) y/x > 1 for group I and (b) y/x < 1 for group II. The chemical composition of Kn [Pdx Coy Cz Nl Hm ] was carefully determined via ICP-AES and elemental analysis. The structure of the electrocatalysts was studied by mid- and far-Fourier transform infrared spectroscopy (FT-IR) and micro-Raman spectroscopy. Thermogravimetric (TG) analyses were carried out in both inert and oxidising atmospheres in order to obtain information on the thermal degradation process and the thermal stability of the electrocatalysts. Scanning electron microscopy (SEM) was used to inspect the surface morphology of the proposed materials. Xray photoelectron spectroscopy (XPS) was used to determine the chemical composition of the sample surface and to provide information about the oxidation states of the metal atoms. The electrochemical activity of the Pd-Co-based carbon-nitride (Pd-Co-CN) electrocatalysts was evaluated in both the oxygen reduction reaction and the hydrogen oxidation reaction by cyclic voltammetry with the thin-film rotating-disk electrode method (CV-TF-RDE) [17]. The influence of methanol on the activity of electrocatalysts towards the ORR was also evaluated. The electrochemical results thus obtained were compared with that of a standard commercial reference material containing only platinum as active element and tested in the same conditions. 2. Experimental 2.1. Reagents Potassium tetrachloropalladate(II) 99%, (K2 PdCl4 ), palladium(II) cyanide 98%, (Pd(CN)2 ), and H2 O2 36% vol., were supplied by ABCR GMBH. Potassium hexacyanocobaltate(III) 95%, (K3 Co(CN)6 ), was provided by Fluka. D(+)-sucrose, biochemical grade, was Acros reagent. HCl 37 wt.% and HNO3 65 wt.% were supplied by Riedel-de Ha¨en and Fluka, respectively. Potassium cyanide 97% (KCN) was Sigma–Aldrich reagent. All the reagents were used as received. Potassium tetracyanopalladate(II) hydrated (K2 Pd(CN)4 ·xH2 O), was synthesized reacting Pd(CN)2 with KCN according to a protocol reported elsewhere [18]. XC-72 R carbon black, provided as a courtesy by Carbocrom s.r.l., was washed with H2 O2 10% vol. prior to use. The commercial EC-20 product (ElectroChem Inc.) with a nominal Pt loading of 20 wt.% was used as a reference electrocatalyst without any further purification and washing procedure. 2.2. Preparation of the Kn [Pdx Coy Cz Nl Hm ] systems 2.2.1. Synthesis of I materials 652.8 mg (2 mmol) of potassium tetrachloropalladate(II) were dissolved in 5 ml of ultrapure water (Milli-Q, Millipore), yielding a deep brown solution A. 1246.2 mg (3.75 mmol) of
1605
potassium hexacyanocobaltate(III) were dissolved in 5 ml of ultrapure water yielding a clear solution B. 4.65 g of sucrose were dissolved in the minimum amount of ultrapure water. The sucrose solution was divided equally into two aliquots which were then added to A and B. After mixing together A and B, a sol–gel transition immediately occurred, which yielded a light green, transparent gel. After 2 days the gel was transformed into a plastic solid through a gel–plastic transition. The obtained plastic precursor was transferred into a quartz tube and heated under a dynamic vacuum of 10−3 mbar at 180 ◦ C for 16 h. The dried solid was thermally decomposed at 300 ◦ C for 2 h. The resulting deep brown powder was finely ground and then divided into four aliquots. Each aliquot underwent a different thermal treatment at tf = 400, 550, 700 and 900 ◦ C for 2 h under a dynamic vacuum. The resulting four black materials were then removed from the tube and finely ground with a ball mill for 3 h. The powders were washed three times with Milli-Q water, treated with H2 O2 , 10% vol. and finally dried under an infrared lamp until all the water was removed. The four prepared samples were singled out as Itf where I indicated group I and tf the temperature of the final thermal treatment. 2.2.2. Synthesis of II materials The reagent K2 Pd(CN)4 ·xH2 O was prepared as described elsewhere [18]. ICP-AES measurements performed on this latter material yielded palladium and potassium essays equal to 33.56 and 24.34%, respectively. These values are compatible with the preparation of a compound having the following formula: K2 Pd(CN)4 ·1.58H2 O. 391.7 mg (1.2 mmol) of K2 PdCl4 were dissolved in 5 ml of ultrapure water, yielding the deep brown solution A. 99.7 mg (0.3 mmol) of K3 Co(CN)6 were dissolved in 5 ml of ultrapure water together with 475.6 mg (1.5 mmol) of K2 Pd(CN)4 ·1.58H2 O, giving so rise to the clear solution B. 4.65 g of D(+) sucrose were dissolved in the minimum amount of ultrapure water and after division in two equal aliquots added to A and B solutions. The solution obtained mixing together A and B after 2 days was transformed into a plastic material through a sol–gel and a gel–plastic transition. The plastic material thus obtained was opaque and light yellow in colour. The subsequent steps in the preparation of II materials were identical to those of group I. In this case the precursor was treated for 2 h at tf = 500, 700 and 900 ◦ C. Three samples indicated as IItf were prepared, where II indicated that y/x < 1 and tf was the temperature of the final thermal treatment. 2.3. Instruments and methods The metal composition of the materials was determined by ICP-AES using the method of standard additions. The emission lines were: λ(Pd) = 340.458 nm, λ(Co) = 228.616 nm, λ(K) = 766.490 nm. The samples were mineralized as follows. A weighed aliquot of each sample was placed in a glass beaker filled with about 20 ml of ultrapure water, which was brought to the boiling point on an open flame. About 2 ml of HCl 37 wt.%, HNO3 65 wt.% and H2 O2 36 wt.% were subsequently added dropwise in a cyclic succession every time the volume of the boiling solution was too low (about 10 ml). This process was
1606
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
repeated for a few hours until the solution became transparent. The liquid was allowed to cool, transferred into a 25 ml volumetric flask, and brought to volume with bidistilled water. Further details on the experimental setup and the elemental analysis are reported in [16]. Thermal analyses were carried out with a HighRes modulated TGA 2950 thermogravimetric analyzer produced by TA Instruments, using an open platinum pan and a temperature range from 30 to 1000 ◦ C. The samples were analysed both in N2 and in oxidising air atmosphere. FT-IR measurements in both the medium- (MIR) and the far IR (FIR) were collected using a Nicolet FT-IR Nexus spectrometer, at a resolution of 4 cm−1 . MIR measurements were derived by averaging 500 scans. Samples were dispersed in anhydrous KBr pellets which were mounted in a cell filled with a N2 inert atmosphere. The farIR spectra were derived by averaging 2000 scans. Measurements were carried on samples spread as thin films on a polyethylene window using nujol. The baseline correction was performed with the Nicolet FT-IR Nexus spectrometer software. The microRaman spectra were collected using a home-made instrument, composed by: (a) a double Czerny-Turner monochromator at a focal distance of 400 mm and a grating of 1800 lines/mm blazed ˚ (b) a Spectra-Physics Stabilite 2016 Argon Ion Laser, at 5000 A; whose excitation line was set at 514.15 nm acting as the light source; (c) an Olympus BX-41 confocal microscope equipped with a 50x objective; (d) a Jobin-Yvon Symphony CCD detector, cooled with liquid nitrogen. The morphology of samples was inspected by using scanning electron microscopy. Standard images collected with backscattered electrons were recorded using a Cambridge Stereoscan 250 Mark 1 electron microscope at an acceleration voltage of 20 kV. XPS analyses were run on a Perkin-Elmer Φ5600ci spectrometer using standard Al-K␣ radiation (1486.6 eV) operating at 350 W. The working pressure was lower than 5 × 10−8 Pa. The spectrometer was calibrated by assuming the binding energy (BE) of the Au4f7/2 line at 83.9 eV with respect to the Fermi level. The standard deviation for the BE values was 0.15 eV. Survey scans (187.85 eV pass energy, 1 eV/step, 50 ms per step) were obtained in the 0–1300 eV range. Detailed scans (58.7 eV pass energy, 1 eV/step, 25 ms per step) were recorded for the O1s, C1s, N1s, Cl2p, K2p, Co2p and Pd3d regions. The atomic composition was evaluated after a Shirley type background subtraction [19] using sensitivity factors supplied by Perkin-Elmer [20]. Samples were introduced directly into the XPS analytical chamber by a fast entry lock system. Peaks were assigned using the values reported in [20] and in the NIST XPS Database [21]. 2.4. Electrode preparation and electrochemical measurements Porous electrodes were prepared according to a procedure described elsewhere [17]. A weighed aliquot of the material to be tested was added to a precise amount of XC-72 R carbon black to obtain a mixture where the sum of the active metals (palladium and cobalt, or platinum in the case of the EC-20 reference) was about 6% in weight. The mixture was ball-milled for 3 h yielding a homogeneous dispersion. A weighed aliquot of the resulting powder was placed into a test tube together with
a suitable amount of ultrapure water (Milli-Q, Millipore) and of a commercial Nafion solution (Aldrich, 5% weight), aiming to an overall loading on the glassy-carbon tip equal to 30 g of the mixture per square centimetre, covered by a Nafion layer about 150 nm thick. The resulting ink was sonicated for about 1 h to ensure a good homogeneity. Fifteen microliters of the ink were then transferred on the top of a freshly-polished glassycarbon electrode tip with an active diameter of 5 mm. Water was then removed from the ink by evaporation under an IR lamp. The tip was then mounted on an EDI 101 rotating electrode attached to a CTV 101 speed control unit (Radiometer Analytical). The resulting system was used as working electrode in order to test the specific ORR and HOR activities of the catalyst materials. A platinum counter-electrode was adopted. A Hg/HgSO4 reference electrode was used which was placed in a separate compartment connected to the main electrochemical cell by a salt bridge. The electrochemical cell was filled with a 0.1 M HClO4 solution. The electrochemical measurements were performed at 60 ◦ C by rotating the electrode at 4900 rpm. Data were collected with a Potentiostat/Galvanostat model 263A of EG&G Instruments. The materials were activated by cycling the working electrode between 0.05 and 1.15 V versus NHE until the voltammogramms became stable. High-purity nitrogen, oxygen and hydrogen (Air Liquide) were used to saturate the electrochemical cell and to test the electrocatalytic performances of the materials in different atmospheres. The cyclic voltammogramms were collected at 100 mV/s. The electrochemical activity per unit mass of palladium or platinum (expressed as A/g) was obtained by subtracting the cyclic voltammogramm (CV) of a given material collected in pure nitrogen from the cyclic voltammogramm of the same material collected in either a pure oxygen or hydrogen atmosphere. The poisoning effect of methanol in the ORR was evaluated as described elsewhere [22] by collecting linear sweep voltammetries (LSVs) at 100 mV/s at increasing methanol concentrations of 0, 0.1, 0.5 and 1 M. The LSV profiles thus determined were referred to the mass of palladium or platinum effectively present on the working electrode. 3. Results and discussion 3.1. Synthesis of Kn [Pdx Coy Cz Nl Hm ] materials The final chemical composition of the prepared materials after the washing and activation procedures is reported in Table 1. It should be highlighted that the modulation of the relative amounts of Pd and Co metals in materials resulted quite good. Nevertheless, some differences for I and II materials are observed. In I, the amount of cobalt slightly increases as tf is raised. This behaviour indicates that as tf increases a decreasing fraction of low molecular soluble species of cobalt is removed during the washing procedure of materials. In II, the Pd/Co metal ratio is practically constant on tf , thus witnessing that a complete incorporation of the original metal complexes in the bulk materials occurred. Despite the extensive washing procedure, a small fraction of potassium was measured in samples, indicating that anionic species centred on oxidised metal atom complexes are present in final materials. The amount of
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
1607
Table 1 Microanalytical data for Kn [Pdx Coy Cz Nl Hm ] materials of I and II groups tf (◦ C)
Compound
Atomic weight% aK
± ± ± ±
I (y/x > 1)
I400 I550 I700 I900
400 550 700 900
3.484 1.820 0.974 1.026
II (y/x < 1)
II500 II700 II900
500 700 900
1.760 ± 0.007 2.004 ± 0.003 1.443 ± 0.041
a b
Formula a Pd
0.009 0.012 0.013 0.056
6.161 7.15 7.59 9.17
a Co
± ± ± ±
0.025 0.10 0.14 0.45
16.213 ± 0.087 18.28 ± 0.46 20.82 ± 0.34
bC
bN
bH
0.028 0.046 0.15 0.55
49.85 56.66 59.74 62.42
4.95 4.58 3.85 1.66
2.10 1.15 0.53 0.06
K1.54 [PdCo1.43 C72 N6.1 H36 ] K0.69 [PdCo1.73 C70 N4.9 H17 ] K0.35 [PdCo1.78 C70 N3.9 H7.4 ] K0.30 [PdCo1.88 C60 N1.4 H0.69 ]
0.971 ± 0.010 1.102 ± 0.021 1.240 ± 0.039
61.05 62.50 60.7
1.98 1.87 1.49
1.83 – –
K0.30 [PdCo0.11 C30 N0.93 H12 ] K0.30 [PdCo0.11 C30 N0.78 ] K0.19 [PdCo0.11 C26 N0.54 ]
4.876 6.845 7.50 9.53
± ± ± ±
Determined by ICP-AES spectroscopy. By elemental analysis.
potassium decreases as tf increases owing to a progressive reduction to the (0) oxidation state of metal atom complexes. In accordance with other studies [15,16], potassium is removed as KCl either via sublimation during the thermal treatments or by washing processes. The amount of hydrogen and nitrogen in electrocatalysts decreases as tf is raised, thus indicating that: (a) the graphitic fraction in the materials increases; (b) the elimination of volatile low molecular nitrogen species during the thermal treatments occurs. In accordance with [15,16], the residual nitrogen amount present in the materials could be embedded in non-volatile bimetal clusters of carbon-nitride materials. In conclusion, the proposed synthesis protocol produces electrocatalysts characterized by a well-controlled molar ratio between Pd and Co metal atoms and results in materials exhibiting: (a) cluster sites supported in a carbon-nitride host matrix; (b) good electrical conductivity (data not reported). Indeed, the composition of the Kn [Pdx Coy Cz Nl Hm ] materials is consistent with that of a bimetal palladium-cobalt carbon-nitride compound having a low nitrogen content.
T < 120 ◦ C. This result indicates that the hydrophilic capability of materials here proposed are higher than that of the commercial EC-20 reference. The measurements performed under oxidising atmosphere confirmed the results reported above, showing an increase in the temperature of the main mass elimination as tf is raised (Table 2) and a higher thermal stability for the samples of II group. The mass residue measured at T > 950 ◦ C is slightly larger than the sum of palladium and cobalt atoms, thus pointing out that formation of oxides and/or carbon-nitrides derivatives occurred in the solid. The tiny mass elimination observed above 800 ◦ C, in accordance with [16] was attributed to the removal
3.2. Thermal and morphological analyses 3.2.1. Thermogravimetric studies The thermogravimetric profiles of I and II materials measured under N2 and air atmosphere are shown in Figs. 1 and 2, respectively. The thermal parameters determined on the TG curves measured under oxidizing atmosphere are summarized in Table 2. The comparison of the TG profiles of I and II measured under inert atmosphere shows that I materials are less stable towards thermal decomposition with respect to II materials. This result can be rationalized admitting that the metallorganic precursor of I was less reticulated than that of II. The lower thermal stability of I and II materials with respect to EC-20 is easily explained if we consider the larger concentration of heteroatoms such as nitrogen and oxygen embedded in their bulk structure (Table 1). Indeed, the thermally-generated heterocyclic moieties are decomposed as the temperature increases. Furthermore, as a general trend TG profiles of I and II indicate that the thermal stability of the materials increases as tf rises. On this basis, it is to be expected that the higher tf , the lower the number of heteroatoms of the bulk materials. Five to ten weight percent of loosely-adsorbed water was detected in TG profiles of I and II at
Fig. 1. Thermogravimetric profiles of Kn [Pdx Coy Cz Nl Hm ] materials measured under N2 atmosphere: (a) I; (b) II.
1608
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
achieve a better contrast between the different areas of samples. I and II products showed a very rough surface consisting of nanometric, metal-rich grains immersed in a metal-poor matrix. As a consequence, the overall surface area of these electrocatalysts seems to be significantly large. A careful inspection of the micrographs indicated that as tf is raised: (a) no clear dependence of the surface features of these nanoparticles was evidenced; (b) a slight increase of the size of metal-rich areas was revealed. The increase in particle size is more evident for materials of group II. Indeed, at tf = 900 ◦ C the metal-rich areas are larger and more interconnected with respect to those of their counterparts detected on the sample prepared at tf = 500 ◦ C. 3.3. Vibrational spectroscopy
Fig. 2. Thermogravimetric profiles of Kn [Pdx Coy Cz Nl Hm ] materials measured under oxidizing atmosphere: (a) I; (b) II.
via sublimation of KCl. The high-temperature residue of about 21.1 wt.% of EC-20 is consistent with the nominal platinum loading of 20% of this sample. 3.2.2. Morphology analyses I and II materials were inspected by SEM measurements in order to gain information on their surface morphology and homogeneity. The corresponding micrographs are reported in Fig. 3a–g and were collected with backscattered electrons to Table 2 Thermogravimetric analyses performed on I and II materials under oxidizing atmosphere Compound
tf (◦ C)
TME a (◦ C)
wt.%ME b
%Residuec
I (y/x > 1)
400 550 700 900
322 325 366 423
42.4 54.0 53.8 56.5
22.6 23.0 23.1 26.5
II (y/x < 1)
500 700 900
374 437 430
59.5 64.7 61.7
23.1 23.9 26.9
EC-20
n/a
403
63.2
21.1
a b c
Temperature of the mass elimination. Mass elimination at TME . Mass residue at T > 950 ◦ C.
Structural information of I and II materials was obtained by FIR, MIR and micro-Raman studies. The correlative attribution of vibrational peaks was performed in accordance with previous works [11,12,14,23–35] and is reported in Tables 3 and 4 for I and II materials, respectively. The vibrational spectra are diagnostic of the following four main structural characteristics (spectra not shown): (a) modes associated to carbon-nitride (CN) backbone components; (b) bands associated to the cyanide bridging and terminal groups; (c) peaks attributed to the vibrations of ether, phenol, alcohol and carboxyl groups present on the surface of CN grains; (d) metal-ligand modes associated to the active metal complexes supported on CN particles. Important spectral features typical of graphite-like materials are revealed in MIR and micro-Raman spectra in the region 1800–1000 cm−1 . Peaks in the Raman profiles at about 1575 and 1370 cm−1 were assigned, respectively to the Raman active E2g Graphitic (G) and to the A1g mode of lattice vibration. This latter vibrational mode becomes Raman active due to the finite size of disordered crystals (D) [25]. The G and D bands which are infrared inactive in typical crystalline carbons become active when nitrogen is substituted for carbon or when, owing to a considerable structural disorder, a breakdown of the k-selection rule in graphitic materials is observed [25,36]. The lower amount of nitrogen atoms in the bulk matrix accounts for the lower intensity of the band ascribable to the A1g mode in FT-IR spectra of: (a) II materials; (b) I and II materials at higher tf . Furthermore, in accordance with other studies, the shape of the G and D region in the infrared and Raman profiles are similar with those of ␣or graphitic-like carbon-nitride materials with a nitrogen percentage ranging from 2 to 9% [36,37]. These results suggest that in the graphitic microdomains of I and II materials: (a) C is present both as a distribution of sp3 and sp2 carbons as well as other functionalities like alcohols, ethers and phenols; (b) a fraction of the total N amount is replaced for sp2 carbons. In accordance with other FT-IR studies [28], the absorptions peaking in the region 1258–1020 cm−1 (Tables 3 and 4) are attributed to alcohols, ethers and phenols. As tf is raised, the overall increase in the intensity of these latter vibrational modes indicates that in I and II: (a) the G and D bands are decreasing owing to an inactivation process in the MIR spectra originated by a gradual increase of the graphitic ordering of the materials [25,36]; (b) the alcohol, ether and phenol functionalities
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
1609
Fig. 3. SEM images of Kn [Pdx Coy Cz Nl Hm ] materials collected with backscattered electrons. Group I: (a) I400 ; (b) I550 ; (c) I700 ; (d) I900 . Group II: (e) II500 ; (f) II700 ; (g) II900 .
of carbon-nitride-like materials are present in samples obtained at higher tf . The presence of ν(C N) vibrational modes in the 2000–2400 cm−1 region (Tables 3 and 4) [11,12,14] indicates that the nitrogen not implicated in the formation of carbonnitride particles gives rise to two types of cyanide groups. The
first C N group is bridging together Co and Pd atoms and the second is present as a terminal free group. As a general trend the intensity of the cyanide modes decreases as tf is raised, probably due to a progressive incorporation of these functional groups in the graphitic-like host matrix. A careful inspection
1610
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
Table 3 Vibrational assignments for I samples of Kn [Pdx Coy Cz Nl Hm ] materials
a Relative b ν:
intensities are reported in parentheses: vs: very strong; s: strong; m: medium; w: weak; vw: very weak; sh: shoulder. stretching; δ: bending; π: bridge deformation.
of the FIR (Tables 3 and 4) and Raman spectra at frequencies lower that 700 cm−1 shows that: (a) several metal-ligand stretching and bending vibrations are revealed in I and II materials [24], including: (a) ν(Pd-N), ν(Pd-CN), ν(Co-CN), ν(Pd-Cl), ν(PdN2 ), δ(Co-CN); (b) Ag lattice vibrations of both - and
graphitic-like carbon-nitride materials; (c) out-of-plane bending modes of graphitic-like carbon-nitride domains [34,35]. In addition, the particle size of the studied materials was determined by using G and D Raman bands following the procedure outlined in [25]. Results shown in Fig. 4 reveal that as tf increases a sig-
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
1611
Table 4 Vibrational assignments for II samples of Kn [Pdx Coy Cz Nl Hm ] materials
a Relative b ν:
intensities are reported in parentheses: vs: very strong; s: strong; m: medium; w: weak; vw: very weak; sh: shoulder. stretching; δ: bending; π: bridge deformation.
nificant decrease of particle size of the carbon-nitride grains is observed. Taken together, vibrational investigations witness that as tf raises an increase in the short-range structural order of the materials takes place owing to the graphitization process. This latter phenomenon is responsible of the formation of ␣- and graphitic-like carbon-nitride nanoparticles owing to the thermal degradation of the three-dimensional hybrid inorganic–organic network of the precursor. In summary, vibrational studies allowed us to conclude that I and II materials consist of ␣- and graphitic-like carbon-nitride materials supporting Pd-
Co bimetal clusters. These latter species are formed bonding together Pd and Co complexes by either bridges of cyanide groups or pyridinic rings of the graphitic-like carbon-nitride layers. 3.4. XPS studies Information on the oxidation states of the elements composing I and II materials was obtained by XPS studies. Two typical survey spectra are shown in Fig. 5. Carbon, oxygen and palladium were revealed in all the samples, while cobalt was clearly
1612
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
Fig. 4. Dependence of the particle size of carbon-nitride materials as a function of tf . The lines are guides for the eye.
Fig. 5. Survey XPS spectra of I900 and II900 materials.
detected only on samples of I group. In both I and II samples, the carbon peak is characterized by a BE of 284.4 eV. This value is in accordance with the typical figure reported for graphitic-like materials [20]. As expected, a substantial amount of nitrogen is revealed for the Pd-Co-CN materials, in particular for those of group I. This observation is consistent with the data of chemical composition of the samples shown in Table 5, which supports the hypothesis that nitrogen is included also in the outermost layers of the nanoparticles. In order to carry out quantitative analyses detailed scans were acquired for the C1s, O1s, N1s, Co2p, Pd3d, K2p and Cl2p regions. Results are summarized in Table 5. The
Fig. 6. Evolution of the Pd3d peak on tf for Kn [Pdx Coy Cz Nl Hm ] materials: (a) I; (b) II.
surface chemical composition of the samples belonging to the II group is almost independent on tf . Furthermore, in group I as tf rises a significant increase in the surface concentration of carbon together with a corresponding decrease in the amount of oxygen is observed. Particularly, at higher tf , oxygen concentration in I samples reached values very similar to that observed for the II900 material. This evidence is easily explained admitting that the degree of graphitization of the hybrid inorganic–organic precursors increases as tf is raised, consistently with the structural
Table 5 XPS surface chemical composition in atomic % of the Kn [Pdx Coy Cz Nl Hm ] materials Compound
tf (◦ C)
Formulaa
Element %Pdb
%Cob
%Kb
%Clb
%Cb
%Nb
%Ob
I
I400 I550 I700 I900
400 550 700 900
1.1 0.8 0.7 0.6
4.0 2.8 1.6 2.8
1.6 0.7 0.1 –
– 1.1 0.9 0.7
46.3 66.8 71.3 75.6
10.0 6.5 3.7 3.3
37.1 21.3 21.7 16.9
K1.54 [PdCo1.43 C72 N6.1 H36 ] K0.69 [PdCo1.73 C70 N4.9 H17 ] K0.35 [PdCo1.78 C70 N3.9 H7.4 ] K0.30 [PdCo1.88 C60 N1.4 H0.69 ]
II
II500 II700 II900
500 700 900
0.9 1.0 1.0
0.4 0.3 0.3
2.1 1.5 1.9
0.8 0.3 0.6
75.6 75.9 76.6
4.0 3.2 3.2
16.2 17.8 16.4
K0.30 [PdCo0.11 C30 N0.93 H12 ] K0.30 [PdCo0.11 C30 N0.78 ] K0.19 [PdCo0.11 C26 N0.54 ]
a b
Determined by ICP-AES and elemental analysis. Atomic%.
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
1613
Fig. 8. Dependence of the Pd(0) percentage (Pd(0)%) on tf for Kn [Pdx Coy Cz Nl Hm ] materials. Pd(0)% = APd(0) /(APd(0) + APd(II) ) × 100. The lines are guides for the eye.
limited to the samples of the group I due to the low concentration of this metal in the materials of the group II. Fig. 9(a and b) reports, respectively the evolution of the Co2p spectral region on tf and a typical spectral decomposition performed in this region. A careful inspection of XPS profiles in Co2p region indicates that: (a) no significant dependence of the oxidation state of cobalt on tf is revealed; (b) two components for Co2p located at about 781.3 and 786.3 eV are detected. The lower-energy component was ascribed to a Co(II) species [20,38]. This latter species is
Fig. 7. Typical decomposition of the Pd3d region in the XPS spectra of Kn [Pdx Coy Cz Nl Hm ] materials: (a) I900 ; (b) II900 .
information drawn from vibrational studies. The comparison between the XPS spectra of the analyzed materials shows that tf plays a crucial role on the oxidation state of the palladium atoms present on the materials. The evolution on tf of the Pd3d region in the XPS spectra of I and II materials are reported in Fig. 6(a and b), respectively. Two doublets of peaks are revealed, whose main components are peaking at about 335.4 and 337.8 eV. The lowerenergy component was ascribed to a Pd(0) species, while the higher-energy feature was attributed to a Pd(II) species [20]. It is expected that Pd metal atoms can be coordinated by nitrogenor carbon-based ligands of heterocyclic aromatic rings of the carbon-nitride matrix or other ligand species such as chlorine atoms. The semi-quantitative molar percentage of the Pd(0) and Pd(II) species in the Pd-Co-CN materials was determined by decomposing the XPS spectral region of Pd3d peaks. Two typical decompositions of I and II materials are shown in Fig. 7. Results showed a different dependence of Pd(0) concentration on tf for both I and II materials as shown in Fig. 8. In particular, as tf rises, the percentage of Pd(0) increases (see Fig. 8): (a) steeply, from 50 to about 80% in the samples of group II; (b) gradually, from 50 to 70% in the materials of group I. These evidences suggest that Pd(II) is easily reduced in samples with a low concentration of Co, thus confirming that Co acts as a stabilizing element for the higher oxidation states of the precious metal. It should be highlighted that the analysis of the oxidation states of Co was
Fig. 9. (a) Evolution of the Co2p peak on tf for the I materials; (b) decomposition of the Co2p region of the I900 material.
1614
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
ascribed to a Co(II) complex where the metal ion is coordinated by ligands such as chlorides, and oxygen- or nitrogen-based electron donor functionalities of heterocyclic aromatic rings of the carbon-nitride matrix. The high-energy component was associated with a shake-up satellite line of Co(II) [38]. 3.5. Electrochemical studies The electrochemical measurements reported in this paper allowed us to evaluate for I and II materials the performance in: (a) the oxygen reduction reaction; (b) the hydrogen oxidation reaction; (c) the influence of methanol poisoning in the oxygen reduction reaction. The commercial electrocatalyst EC-20 was tested in exactly the same conditions and was used as a reference material. Figs. 10 and 11 show the mass activities towards the ORR and the HOR for I and II electrocatalysts, respectively. The material tests were performed using the cyclic voltammetry with thin-film rotating-disk electrode (CV-TF-RDE) method as described elsewhere [17]. The most relevant figures describing the electrochemical performance of the materials are reported in Table 6. The maximum mass activity of each material was measured at potentials lower than 200 mV for the ORR and between 200 and 400 mV for the HOR. The values of the activation potential reported in Table 6 for the ORR were measured in the cathodic sweeps of CV-TF-RDE curves and were evaluated as
Fig. 11. CV-TF-RDE profiles of Kn [Pdx Coy Cz Nl Hm ] materials of group II. Profiles of EC-20 reference are shown: (a) HOR; (b) ORR.
the potential where the two CV-TF-RDE tangents centred at 750 and 1050 mV intersect each other. In order to compare the electrochemical efficiency of I and II materials towards the ORR with that of the EC-20 reference, a reasonable figure of merit is (1): V 1 Amax A0.6 V REff = (1) + + 3 VEC-20 Amax EC-20 A0.6 V EC-20 where V is the activation potential, Amax and A0.6 V are, respectively the maximum mass activity and the mass activity at 0.6 V versus NHE of the catalysts in the ORR. The dependence of REff
Fig. 10. CV-TF-RDE profiles of Kn [Pdx Coy Cz Nl Hm ] materials of group I. Profiles of EC-20 reference are shown: (a) HOR; (b) ORR.
Fig. 12. Dependence of the efficiency of the catalysts (REff ) on tf for the ORR. The lines are guides for the eye.
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
1615
Table 6 Parameters describing the electrochemical performance of I and II samples towards: (a) ORR; (b) HOR Compound
tf (◦ C)
Mass activity at 0.6 V (A/g Pd)
Maximum mass activity (A/g Pd)a
Maximum current flux (mA/cm2 g Pd)a
Activation potential (mV)
(a) ORR I I400 I550 I700 I900
400 550 700 900
78 691 1148 1997
−1489 −2693 −4402 −4214
−7.58 −13.70 −22.40 −21.45
704 853 842 833
II500 II700 II900
500 700 900
224 347 360
−1271 −1309 −1313
−6.47 −6.66 −6.68
796 796 805
EC-20b
n/a
1208
−2770
−14.10
861
II
Compound
tf (◦ C)
Maximum mass activity (A/g Pd)c
Maximum current flux (mA/cm2 g Pd)c
(b) HOR I I400 I550 I700 I900
400 550 700 900
1241 1992 2115 2382
6.32 10.15 10.77 12.13
II500 II700 II900
500 700 900
1259 1126 844
6.41 5.73 4.30
EC-20d
n/a
2162
11.01
II
Measurements were performed with a metal loading of a This value is determined at V < 200 mV vs. NHE. b The active metal for this material is platinum. c This value is determined at V ≈ 300 mV vs. NHE. d The active metal for this material is platinum.
1.85 g/cm2 ,
at 333 K and at 4900 rpm.
on tf is reported in Fig. 12. As a general trend, in I and II electrocatalysts as tf is raised REff increases. Particularly, the materials of group I outperform those of group II, reaching for samples obtained at tf ≥ 700 ◦ C REff values higher than that exhibited by the EC-20. In the HOR, the analysis of the performance of I and II materials clearly shows a maximum at about 300 mV versus NHE. This evidence indicates that at potentials lower than 300 mV molecular hydrogen is probably absorbed in the active sites of Pd-Co-CN materials where it is oxidised at about 300 mV [39]. It is interesting to note that in the HOR, the dependence on tf of the mass activities are very different for I and II materials (Table 2). Indeed, for I materials the mass efficiency increases as tf is raised reaching values very similar to that of the EC-20 reference, while for II materials the opposite behaviour is observed. In the former case it is proposed that as tf is raised the fraction of graphitized carbon-nitride moiety increases, thus originating active Pd-Co bimetal catalytic sites on the surface of I materials. In II, this latter phenomenon decreases the concentration of active Pd-Co complexes on carbon-nitride materials due to the coalescence event of the active sites with the Pd-rich domains as shown by SEM analyses and XPS investigations. All this information indicates that Pd-Co-CN materials having on the surface a large fraction of cobalt atoms gives rise to efficient electrocatalytic sites for both the ORR and the HOR reactions.
The methanol crossover from the anode compartment through the polymer electrolyte membrane is an important hindrance to the optimal performance of any DMFC [2,40–45]. Indeed, as methanol reaches the cathode it is oxidized by the catalyst, thus decreasing the overall output potential of the cell. This is especially troublesome for most of the traditional platinum-based catalysts which at the working potentials of fuel cell cathodes present a good efficiency in the methanol oxidation reaction. The dependence in the ORR of the shape of LSV curves on the concentration of methanol (0 ≤ [CH3 OH] ≤ 1 M) is shown in Fig. 13. It should be observed that the decrease in operating potential of Pd-Co-CN materials is relatively small if compared with that of the EC-20 reference, especially at low methanol concentrations (0.1 M). Similar results were obtained for I and II materials, both at high and low tf (data not shown). On this basis we can hypothesize that, with respect to platinum, Pd-Co clusters of both I and II groups are less efficient catalytic sites towards the methanol oxidation reaction [10]. Nevertheless, a significant drawback of the Pd-Co-CN systems is the limited durability of their electrochemical performance which prompted further studies in our laboratory in order to stabilize the proposed materials. In summary, the materials of group I obtained at tf > 700 ◦ C appear very promising for application as catalysts on the cathodic side of DMFCs owing to their activity, activation poten-
1616
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
I perform better than those of group II in the ORR reaction; (b) the intensity of the reduction current for I materials increases as tf is raised. 4. Conclusions
Fig. 13. Effect of the methanol concentration on the ORR: (a) EC-20 reference; (b) I900 .
tial and resistance towards methanol poisoning. In addition, it should be highlighted that the activity of Pd-Co-CN towards the HOR is quite good, thus making them viable candidates for applications as electrocatalysts at the anode of PEMFCs. 3.6. Structural and functional hypotheses It results that I and II materials consist of bimetal cluster complexes supported on tiny carbon-nitride-like nanoparticles. The bimetal clusters are formed by Pd square-planar coordination complexes bonded to octahedral Co(II) species through bridges of cyanide groups and pyridinic aromatic rings of graphitic layers of the carbon-nitride matrix. The coordination shell around Co(II) seems to be completed by ligands such as hydroxyl groups. The key to understand the features of the electrochemical activity exhibited by these materials lies in the evolution of the local structure of the metal sites. Indeed, as tf is raised: (a) palladium atoms are progressively reduced by carbon atoms to the (0) oxidation state; (b) surface contaminants such as chlorine atoms are removed; (c) cobalt atoms are always present quite mostly in the oxidised state (II). Now, it is reasonable to admit that the ORR is more easily catalysed by Pd(0) instead of Pd(II). This latter point is easily justified if we consider that: (a) Pd(II) is coordinated by electron donor ligands; (b) O2 molecules are easily adsorbed and reduced on Pd(0). These considerations allow us to suggest that bimetal clusters based on Pd(0) and Co(II) complexes coordinated by hydroxyl groups are very active catalytic sites towards the ORR reaction. Indeed, it is expected that the hydroxyls coordinated to Co(II) can readily protonate the products of the ORR process thus increasing the elimination rate of water molecules from the active site. In this way, in accordance with the ORR mechanisms proposed for platinum [46], an increase of the mass activity and a decrease of the ORR overpotential of Pd-Co-CNs is expected. This interpretation permits us to suppose that two ingredients are necessary in order to prepare electrocatalysts with high mass activity and a low ORR overpotential: (a) a significant fraction of active metal atoms such as Pd or Pt in the (0) oxidation state; (b) the active metal atom should be complexed with a co-catalyst complex which is able to enhance the rate of elimination of the ORR products. This hypothesis explains the evidences that: (a) the materials of group
Two groups of bimetal Pd and Co carbon-nitride electrocatalysts (I and II) with formula Kn [Pdx Coy Cz Nl Hm ] were prepared by a new synthesis protocol. Group I includes the electrocatalysts with molar ratio y/x > 1, while group II is made of the systems with y/x < 1. The main steps in the preparation of Kn [Pdx Coy Cz Nl Hm ] were: (a) the synthesis of metallorganic precursors through a sol → gel followed by a gel → plastic transition; (b) the thermal treatment of the plastic precursors to obtain bimetal carbon-nitride compounds. The precursors were obtained reacting together K2 PdCl4 with appropriate amounts of K3 Co(CN)6 and K2 Pd(CN)4 ·xH2 O in the presence of D(+) sucrose as an organic binder. The proposed synthesis protocol proved capable to lead to products featuring a very wellmodulated composition of the metal species. The temperature of the final thermal treatment tf played a crucial role in the control of the structural features of the materials. The electrocatalysts prepared at the highest tf featured a very good thermal stability. SEM studies showed that I and II electrocatalysts are biphasic materials of particles with nanometric sizes. MIR, FIR and micro-Raman studies revealed that I and II materials are a mixture of two polymorphic phases of nanoparticles with a structure similar to that of ␣- and graphitic-like carbon-nitride materials. In conclusion, the detailed assignment of the vibrational spectra allowed us to determine that I and II materials consist of a blend of ␣- and graphitic-like carbon-nitride materials supporting the active Pd and Co bimetal cluster sites. The results determined by XPS and electrochemical investigations evidenced that the best performance towards the ORR is achieved when a large number of clusters based on palladium atoms in the (0) oxidation state and cobalt atoms in the (II) oxidation states are present in materials supported on carbon-nitride nanoparticles. In addition, it was determined that a high solid state concentration of cobalt atoms hinders the growth and coalescence of Pd-rich domains, thus leading to a better mass utilization of the metals and to a higher mass activity both towards the ORR and the HOR. Both of these conditions are found in the materials belonging to group I prepared at tf > 700 ◦ C. It should be noticed that the latter materials exhibited an electrochemical performance higher than that shown by a commercial platinum-based electrocatalyst material. All these features, together with a better tolerance towards the poisoning effect of methanol for the ORR make this class of materials very promising for applications as electrocatalysts at both electrodes of PEMFCs and at the cathode of DMFCs. Further work is in progress in order to improve the electrochemical durability of Pd-Co-CN materials under operating conditions in fuel cells. Acknowledgements Research was funded by the Italian MURST project NUME of FISR2003, “Sviluppo di membrane protoniche composite e di
V. Di Noto et al. / Electrochimica Acta 53 (2007) 1604–1617
configurazioni elettrodiche innovative per celle a combustibile con elettrolita polimerico”. References [1] A. Lamm, J. M¨uller, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 4, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 64). [2] S.R. Narayanan, T.I. Valdez, N. Rohatgi, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 4, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 65). [3] W. Vielstich, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 1, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 4). [4] F. Barbir, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 4, John Wiley & Sons, Hoboken, New Jersey, 2003, p. 51 (Ch. 51). [5] G. Konrad, M. Sommer, B. Loschko, A. Schell, A. Docter, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 4, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 52). [6] S.R. Narayanan, T.I. Valdez, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 4, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 81). [7] A. Hamnett, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 1, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 18). [8] J.L. Fernandez, D.A. Walsh, A.J. Bard, J. Am. Chem. Soc. 127 (2005) 357. [9] Information retrieved on: http://www.advfn.com/p.php?pid=commodities &cb=1163783522&btn=allcomms&gid=PM. [10] M.H. Shao, K. Sasaki, R.R. Adzic, J. Am. Chem. Soc. 128 (2006) 3526. [11] V. Di Noto, J. Mater. Res. 12 (1997) 3393. [12] V. Di Noto, M. Vittadello, S.G. Greenbaum, S. Suarez, K. Kano, T. Furukawa, J. Phys. Chem. B 108 (2004) 18832. [13] V. Di Noto, M. Vittadello, S. Lavina, S. Biscazzo, M. Fauri, Electrochim. Acta 48 (2003) 2047. [14] V. Di Noto, J. Phys. Chem. B 104 (2000) 10116. [15] V. Di Noto, R. Gliubizzi, S. Lavina, E. Negro, G. Pace, Italian Patent no. PD2006A000141 (2006). [16] V. Di Noto, E. Negro, R. Gliubizzi, S. Lavina, G. Pace, ECS Trans. 2 (2007) 83. [17] T.J. Schmidt, GasteigerF H.A., in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 2, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 22). [18] J.H. Bigelow, B.S. Hopkins, J.L. Craven, in: W.C. Fernelius (Ed.), Inorganic Synthesis, vol. 2, Mc Graw-Hill, New York, 1946, p. 245.
1617
[19] D.A. Shirley, Phys. Rev. B 5 (1972) 4709. [20] J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben, in: J. Chastain (Ed.), Handbook of X-Ray Photoelectron Spectroscopy, Perkin-Elmer Corp., Eden Prairie, MN, 1992. [21] X-Ray Photoelectron Spectroscopy Database 20, Version 3.0, National Institute of Standards and Technology, Gaithersburg, MD, http://srdata.nist.gov/XPS. [22] V. Stamenkovic, N.M. Markovic, P.N. Ross, J. Electroanal. Chem. 500 (2001) 44. [23] V. Di Noto, D. Longo, V. M¨unchow, J. Phys. Chem. B 103 (1999) 2636. [24] D.M. Adams, Metal-Ligand and Related Vibrations, Edward Arnold Ltd., London, 1967. [25] F. Tuinstra, J.L. Koenig, J. Chem. Phys. 53 (1970) 1126. [26] M.P. Conrad, H.L. Strauss, Phys. Rev. 31 (1985) 6669. [27] S.A. Solin, Physica 99B (1980) 443. [28] P. Ramesh, S. Bhagyalakshmi, S. Sampath, J. Colloid Interface Sci. 274 (2004) 95. [29] D.W. Kim, J.H. Park, S.S. Choi, J. Cho, Mater. Sci. Eng. B64 (1999) 29. [30] T. Werninghaus, D.R.T. Zahn, E.G. Wang, Y. Chen, Diam. Relat. Mater. 7 (1998) 52. [31] S. Kumar, T.L. Tansley, Thin Solid Films 256 (1995) 44. [32] D.H. Lee, B. Park, D.B. Poker, L. Riester, Z.C. Feng, J.E.E. Baglin, J. Appl. Phys. 80 (1996) 1480. [33] F. Parmigiani, E. Kay, H. Seki, J. Appl. Phys. 64 (1988) 3031. [34] T.Y. Yen, C.P. Chou, Appl. Phys. Lett. 67 (1995) 2801. [35] T.Y. Yen, C.P. Chou, Solid State Commun. 95 (1995) 281. [36] J.H. Kaufman, S. Metin, D.D. Saperstein, Phys. Rev. B 39 (1989) 13053. [37] E.G. Wang, Prog. Mater. Sci. 41 (1997) 241. [38] A. Mekki, D. Holland, K. Ziq, C.F. McConville, J. Non-Cryst. Solids 220 (1997) 267. [39] F.C. Nart, W. Vielstich, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 2, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 21). [40] H.A. Gasteiger, S.S. Kocha, B. Sompalli, F.T. Wagner, Appl. Catal. BEnviron. 56 (2005) 9. [41] J.R.C. Salgado, E. Antolini, E.R. Gonzalez, Appl. Catal. B-Environ. 57 (2005) 283. [42] H. Yang, C. Coutanceau, J.M. Leger, N. Alonso-Vante, C. Lamy, J. Electroanal. Chem. 576 (2005) 305. [43] J.F. Drillet, A. Ee, J. Friedmann, R. K¨otz, B. Schnyder, V.M. Schmidt, Electrochim. Acta 47 (2002) 1983. [44] E. Antolini, J.R.C. Salgado, L.G.R.A. Santos, G. Garcia, E.A. Ticianelli, E. Pastor, E.R. Gonzalez, J. Appl. Electrochem. 36 (2006) 355. [45] U.A. Paulus, T.J. Schmidt, H.A. Gasteiger, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells—Fundamentals, Technology and Applications, vol. 2, John Wiley & Sons, Hoboken, New Jersey, 2003 (Ch. 38). [46] P.B. Balbuena, D. Altomare, N. Vadlamani, S. Bingi, L.A. Agapito, J.M. Seminario, J. Phys. Chem. A 108 (2004) 6378.
