Pt5Gd as a highly active and stable catalyst for oxygen electroreduction
Pt5Gd as a highly active and stable catalyst for oxygen electroreduction María Escudero-Escribano, Arnau Verdaguer-Casadevall, Paolo Malacrida, Ulrik Grønbjerg, Brian P. Knudsen, Anders K. Jepsen, Jan Rossmeisl, Ifan E. L. Stephens, and Ib Chorkendorff*
Supporting information Experimental details Each working electrode was a 5 mm in diameter disk supplied by Mateck GmbH, Germany. Before each electrochemical experiment, the Pt5Gd sample was sputtered using a 0.5 keV beam of Ar+ ions, until the surface was clean of adventitious contamination. The sputteredcleaned electrodes were protected with a drop of H2-saturated Milli-Q water and then transferred from UHV into a rotating ring-disk electrode (RRDE) assemble and immersed in the electrochemical cell.
X-ray diffraction (XRD) The XRD profile of the Pt5Gd electrode is shown in Figure S1. XRD peaks match with two different phases, a hexagonal phase and an orthorhombic phase. All the lines above 15% in relative intensity could be fitted with the orthorhombic phase, suggesting that this is the predominant phase. Both phases were reported previously for Pt5Gd [1-5]. The hexagonal phase forms a Cu5Ca-type structure, with the space group P6/mmm, and the fitted lattice parameters: a = b = 0.528 nm, c = 0.440 nm, consistent with the literature [4,5]. For the predominant phase, which was orthorhombic, the fitted lattice parameters are: a = 0.522 nm, b = 0.902, and c = 2.563, in perfect agreement with previous measurements by H.
S1
Lueken and co-workers [5]. The space group is unknown, although several works in the literature state that it should be similar to the Cu5Ca structure [1-5] (previously found for Pt5La, [6]) and the AuBe5 structure (previously found for Pt5Y, [7,8]). According to those investigations, the main difference between the orthorhombic structure and hexagonal Cu5Ca is the stacking of the different layers, resulting in a multiplication of the c axis. On Pt5La, which has the Cu5Ca structure, our previous calculations suggested that the strain in the Pt overlayer would be set by the lattice parameter a, which corresponds to the distance between adjacent lanthanide atoms along the (001) plane. Notably, on different Pt5X phases (where X = lanthanide), a is often constant between the orthorhombic and hexagonal phase of the same composition [5]. This indicates that the local environment of the lanthanide atoms is similar in both the orthorhombic and hexagonal phases. Consequently, for our DFT calculations, we assume that the strain on the Pt overlayer covering the orthorhombic bulk phase would be set by the lattice parameter, a.
Figure S1. XRD intensity profile of polycrystalline Pt5Gd. All the peaks with relative intensity above 15 % fit with the orthorhombic phase.
Angle-resolved X-ray photoelectron spectroscopy (AR-XPS) In-depth surface composition information of Pt5Gd was extracted from AR-XPS spectra recorded using a Theta Probe instrument (Thermo Scientific). The chamber has a base pressure of 5 x 10-10 mbar. The instrument uses monochromatised AlKα (1486.7 eV) X-rays, and the electron energy analyzer has an acceptance angle of 60°. It facilitates XPS spectra S2
recorded from within a diameter of 15 µm with a resolution corresponding to a Ag 3d5/2 full width half maximum (FWHM) smaller than 0.5 eV. The AR-XPS spectra were obtained in parallel, without tilting the sample. In consideration of the count statistics at the grazing angles, an X-ray beam size of 400 µm and an energy resolution corresponding to approximately 1 eV Ag 3d5/2 FWHM was used. The surface was sputter cleaned with a 0.5 keV beam of Ar+ ions, with a current of 1 µA, over a 6 x 6 mm2 area. This was typically continued for around 20 minutes, until the XPS measurement indicated that impurities were negligible. The XPS spectra were taken at several different locations over the metal surfaces. The calculation of the Pt and Gd XPS intensities for establishing the Pt:Gd concentration ratio was achieved by integration of the Pt4f and Gd4d peaks after background removal. A Shirley-type background was chosen for this purpose. The intensities were corrected for the transmission function of the analyzer, Wagner sensitivity factors and the electron mean free path. The surface sensitivity was minimized by selecting the spectra at an angle of 21° from the normal to the surface. Since the electron mean free path of the Pt4f and Gd4d photoelectrons are 11.6 Å and 11.1 Å (as estimated from the TPP-2M formula), at this angle the XPS signal originates from a depth of at least 10 Pt monolayers (~23 Å). Therefore the effects of contaminations and of differential sputtering are minimized. For both the sputter cleaned sample and the one transferred to UHV after ORR testing, the concentration ratio was taken as the average of three independent measurements. Only in the case of the sample analyzed after stability test I (see Figure 3 in the text), the Pt:Gd ratio was calculated from a single XPS measurement. For the depth profiles, the electrons emitted at angles between 20° and 80° to the surface normal were analysed in parallel and detected in 16 channels corresponding to 3.75° wideangle intervals. After XPS identification of the elements present at the surface, their main features were measured in detail with AR-XPS. The depth concentration profiles were obtained using the simulation tool, ARProcess (Thermo Avantage software), which uses a maximum entropy method combined with a genetic algorithm. In all cases, the simulations were based on the relative intensities between Pt 4f, O 1s and C 1s, and Gd 4d at each angle, up to 70.6°. The most grazing angles were omitted from the analysis to reduce the influence of diffraction effects and elastic scattering.
S3
Electrochemical measurements All glassware was cleaned for 24 h in a “piranha” solution consisting of a 3:1 mixture of 96% H2SO4 and 30% H2O2, followed by multiple runs of heating and rinsing with ultrapure water (Millipore Milli-Q, 18.2 MΩ cm) to remove sulphates. The electrochemical experiments were performed with a VMP2 potentiostat (Bio-Logic Instruments), controlled by a computer. The rotating ring-disk electrode (RRDE) assemblies were provided by Pine Instruments Corporation. A standard two-compartment glass cell was used, which was equipped with a water jacket attached to a hot water bath to control the temperature. All electrochemical experiments were carried out at 23 °C. We note the ring disk currents are not reported in this publication, as they were consistent with our earlier studies on similar systems [7]. A RRDE assembly was used in lieu of a rotating disk electrode (RDE) assembly, simply because it prevented the walls of the electrode from getting wet. The electrolyte, 0.1 M HClO4 (Merck Suprapur), was prepared with ultrapure water. The counter electrode was a Pt wire and the reference was an Hg/Hg2SO4 electrode, separated from the working electrode compartment using ceramic frits. Following each measurement, the potential of the reference electrode was checked against a reversible hydrogen electrode (RHE) in the same electrolyte. All the potentials in the text are referred to the reversible hydrogen electrode (RHE), and are corrected for ohmic losses. Following each measurement, 0 V vs. RHE was established by carrying out the hydrogen oxidation and hydrogen evolution reactions on Pt in the same electrolyte. The ohmic drop was measured by carrying out an impedance spectrum with a peak-to-peak amplitude of 10 mV, typically from 500 kHz down to 100 Hz. The target resistance was evaluated from the high-frequency intercept on the horizontal (real) axis of the Nyquist plot and further checked by fitting the impedance spectra by using EIS Spectrum Analyser software [9]. The uncompensated resistance came typically to approximately 30 Ω, and was independent of the potential, rotation speed and the presence of O2. This Ohmic resistance is consistent with the value reported by Koper, Markovic and coworkers [10].
S4
The RRDE was immersed into the cell under a potential control of 0.1 V vs. RHE into a N2 (N5, Air Products) saturated electrolyte. The Pt5Gd electrode was cycled in nitrogensaturated electrolytes until stable cyclic voltammograms (CVs) where obtained (100-200 cycles). A typical stable CV on sputtered-cleaned Pt5Gd is shown in Figure S2, and compared to the base CV on polycrystalline Pt. Figure S3 shows cycles 1, 10, 50 and 100 of sputtercleaned Pt5Gd. As shown in the Figure, the shape of the CV changes while cycling. We observed that after ca. 100 cycles the CV was stable. This can be attributed to two factors (a) a certain time frame is required for the surface to dealloy, reconstruct and reach the stable, active phase, and (b) inevitably, some impurities will accumulate on the surface during the transfer from the UHV chamber to the RRDE assembly. In addition to this, the Gd originally on the surface is dissolved in the electrolyte, and there might be slow surface changes before reaching a stable CV. The ORR activity measurements were conducted in an electrolyte saturated with O2 (N55, Air Products).
Figure S2. Cyclic voltammograms at 50 mV s-1 of Pt5Gd (red curve) and Pt (black curve) in a N2saturated 0.1 M HClO4 solution.
S5
Figure S3. Cyclic voltammograms at 50 mV s-1 (cycles 1 (black curve), 10 (dashed blue curve), 50 (dotted green curve), and 100 (red curve)) of sputter-cleaned Pt5Gd in a N2-saturated 0.1 M HClO4 solution.
Computational details The surface was simulated using the method previously developed for Pt5La [6]. It was modelled as a pure Pt overlayer under 6% compressive strain, relative to equilibrium Pt. The compressive strain of 6% was estimated by assuming that the Pt overlayer covered a Cu5Calike phase, with the lattice parameter a = 0.522 nm. The surface of the Pt/Pt5Gd was modelled by a Pt-slab with 6 % compression in the nearest-neighbour distance compared to equilibrium Pt. This allowed us to model the structure of water to approximate that on Pt(111). In the DFT calculations, the strained Pt (111) surface is modelled by a slab with 6 closepacked layers, where the three topmost layers and the adsorbates are allowed to relax. The
S6
ionic cores are described by PAW setups [11], and the Kohn-Sham valence states are described on a real-space grid with a spacing of 0.18 Å. Exchange and correlation effects are described by the RPBE functional [12]. The Kohn-Sham Hamiltonian is diagonalized iteratively using Pulay mixing of the density and Fermi-Dirac occupation of one-electron states with kBT = 0.1 eV. Total energies are extrapolated to kBT = 0 eV. The periodic images of the slab are separated by 20 Å of vacuum. All calculations have been carried out with the ASE and GPAW software packages [13-15]. The adsorption energies are calculated in a (3x2) surface unit cell, and the Brillouin zone is sampled by an (8 x 8 x 1) k-point grid. The slab is relaxed using the Quasi-Newton scheme until the maximum force component is less than 0.05 eV/Å-1. The effect of solvation is included for HO* by incorporating the adsorbate in an H2O*/HO* superstructure with 2/3 ML total coverage. The energy of HO* is calculated from the H2O*/HO* configuration which minimizes the average OH energy. The adsorption energy of H2O* is calculated in a similar superstructure with 2/3 ML total coverage. In the H2O*/HO* superstructure, half of the water molecules lie in a plane approximately parallel to the surface, and the other water molecules lie in a plane perpendicular to the surface with one hydrogen atom pointing away from the surface [16-21].
S7
Additional figures
Figure S4. Cyclic voltammograms at 1600 rpm and 50 mV s-1 in oxygen-saturated 0.1 M HClO4 on Pt5Gd before (red curve) and after: 10,000 cycles between 0.6 V and 1.0 V at 100 mV s-1 (orange curve), 100 (purple curve) and 200 (magenta curve) cycles between 0.05 V and 1.6 V at 50 mV s-1 (around 30 h of experiments). For comparison, the polarization curve for polycrystalline Pt is represented in black. Evidently, the polarization curve for the ORR in the kinetic-control region remains almost unchanged after 10,000 cycles between 0.6 V and 1.0 V. In addition, the hydrogen peroxide formation region does not change after 10,000 cycles between 0.6 V and 1.0 V vs. RHE under the conditions described above.
S8
Figure S5. Non-destructive AR-XPS profile of polycrystalline Pt5Gd after 10,000 cycles between 0.6 V and 1.0 V vs. RHE.
Figure S6. Pt4f and Gd4d XPS spectra of the sputter cleaned Pt5Gd sample. Measurements at 8 of the total 16 angles used for the profile calculation are shown. S9
Figure S7. Pt4f and Gd4d XPS spectra of Pt5Gd sample after ORR testing. Measurements at 8 of the total 16 angles used for the profile calculation are shown.
Figure S8. Pt to Gd concentration ratio as determined from AR-XPS spectra, as a function of the angle for a sputter cleaned sample. The absence of a specific trend is a good indication of the homogeneity of the sample and of the limited effects of differential sputtering.
S10
References (1) Bronger, W. J. Less-Common Metals 1967, 12, 63. (2) Lueken, V. H.; Bronger, W. Z. Anorg. Allg. Chem. 1973, 395, 203. (3) Zhang, K.; Chen, L. Acta Metall. Sin. 1988, 24, B65. (4) Ohm, V.; Raetz, S.; Sauer, M.; Merkens, M.; Schilder, H.; Lueken, H. J. Alloys Comp. 1996, 238, 95. (5) Sauer, M.; Raetz, S.; Ohm, V.; Merkens, M.; Sauer, C.; Schilder, H.; Lueken, H. J. Alloys Comp. 1997, 246, 147. (6) Stephens, I. E.L.; Bondarenko, A. S.; Grønbjerg, U.; Rossmeisl, J.; Chorkendorff, I. Energy Environ. Sci. 2012, 5, 6744. (7) Stephens, I. E.L.; Bondarenko, A.S.; Bech, L.; Chorkendorff, I. ChemCatChem 2012, 4, 341. (8) Johansson, T. P. PhD Thesis, Technical University of Denmark, 2012. (9) Bondarenko, A. S.; Ragoisha, G. A., in Progress in Chemmometrics Research, Nova Sci. Publishers, New York, 2005, pp. 89-102. (10) Van der Vliet, D.; Strmcnik, D. S.; Wang, C.; Stamenkovic, V. R.; Markovic, N. M.; Koper, M. T.M. J. Electroanal. Chem. 2010, 647, 29. (11) Blochl, P. E.; Forst, C. J.; Schimpl, J. Bull. Mater. Sci. 2003, 26, 33. (12) Hammer, B.; Hansen, L. B.; Norskov, J. K. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 7413. (13) Atomic Simulation Environment, https://wiki.fysik.dtu.dk/ase, accessed 14 December, 2011. (14) Mortensen, J. J.; Hansen, L. B.; Jacobsen, K. W. Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 71, 035109. (15) Bahn, S.; Jacobsen, K. Computing in Science & Engineering 2002, 4, 56. (16) Rossmeisl, J.; Karlberg, G. S.; Jaramillo, T.; Nørskov, J. K. Faraday Discuss. 2008, 140, 337. S11
(17) Ogasawara, H.; Brena, B.; Nordlund, D.; Nyberg, M.; Pelmenschikov, A.; Pettersson, L. G. M.; Nilsson, A. Phys. Rev. Lett. 2002, 89, 276102. (18) T. Schiros; Näslund, L.-Å.; Andersson, K.; Gyllenpalm, J.; Karlberg, G. S.; Odelius, M.; Ogasawara, H.; Pettersson, L. G. M.; Nilsson, A. J. Phys. Chem. C, 2007, 111, 15003. (19) Michaelides, A.; Hu, P. J. Chem. Phys. 2001, 114, 513. (20) Feibelman, P. J. Science 2002, 295, 99. (21) Clay, C.; Haq, S.; Hodgson, A. Phys. Rev. Lett. 2004, 92, 046102.
S12
Supporting information Experimental details Each working electrode was a 5 mm in diameter disk supplied by Mateck GmbH, Germany. Before each electrochemical experiment, the Pt5Gd sample was sputtered using a 0.5 keV beam of Ar+ ions, until the surface was clean of adventitious contamination. The sputteredcleaned electrodes were protected with a drop of H2-saturated Milli-Q water and then transferred from UHV into a rotating ring-disk electrode (RRDE) assemble and immersed in the electrochemical cell.
X-ray diffraction (XRD) The XRD profile of the Pt5Gd electrode is shown in Figure S1. XRD peaks match with two different phases, a hexagonal phase and an orthorhombic phase. All the lines above 15% in relative intensity could be fitted with the orthorhombic phase, suggesting that this is the predominant phase. Both phases were reported previously for Pt5Gd [1-5]. The hexagonal phase forms a Cu5Ca-type structure, with the space group P6/mmm, and the fitted lattice parameters: a = b = 0.528 nm, c = 0.440 nm, consistent with the literature [4,5]. For the predominant phase, which was orthorhombic, the fitted lattice parameters are: a = 0.522 nm, b = 0.902, and c = 2.563, in perfect agreement with previous measurements by H.
S1
Lueken and co-workers [5]. The space group is unknown, although several works in the literature state that it should be similar to the Cu5Ca structure [1-5] (previously found for Pt5La, [6]) and the AuBe5 structure (previously found for Pt5Y, [7,8]). According to those investigations, the main difference between the orthorhombic structure and hexagonal Cu5Ca is the stacking of the different layers, resulting in a multiplication of the c axis. On Pt5La, which has the Cu5Ca structure, our previous calculations suggested that the strain in the Pt overlayer would be set by the lattice parameter a, which corresponds to the distance between adjacent lanthanide atoms along the (001) plane. Notably, on different Pt5X phases (where X = lanthanide), a is often constant between the orthorhombic and hexagonal phase of the same composition [5]. This indicates that the local environment of the lanthanide atoms is similar in both the orthorhombic and hexagonal phases. Consequently, for our DFT calculations, we assume that the strain on the Pt overlayer covering the orthorhombic bulk phase would be set by the lattice parameter, a.
Figure S1. XRD intensity profile of polycrystalline Pt5Gd. All the peaks with relative intensity above 15 % fit with the orthorhombic phase.
Angle-resolved X-ray photoelectron spectroscopy (AR-XPS) In-depth surface composition information of Pt5Gd was extracted from AR-XPS spectra recorded using a Theta Probe instrument (Thermo Scientific). The chamber has a base pressure of 5 x 10-10 mbar. The instrument uses monochromatised AlKα (1486.7 eV) X-rays, and the electron energy analyzer has an acceptance angle of 60°. It facilitates XPS spectra S2
recorded from within a diameter of 15 µm with a resolution corresponding to a Ag 3d5/2 full width half maximum (FWHM) smaller than 0.5 eV. The AR-XPS spectra were obtained in parallel, without tilting the sample. In consideration of the count statistics at the grazing angles, an X-ray beam size of 400 µm and an energy resolution corresponding to approximately 1 eV Ag 3d5/2 FWHM was used. The surface was sputter cleaned with a 0.5 keV beam of Ar+ ions, with a current of 1 µA, over a 6 x 6 mm2 area. This was typically continued for around 20 minutes, until the XPS measurement indicated that impurities were negligible. The XPS spectra were taken at several different locations over the metal surfaces. The calculation of the Pt and Gd XPS intensities for establishing the Pt:Gd concentration ratio was achieved by integration of the Pt4f and Gd4d peaks after background removal. A Shirley-type background was chosen for this purpose. The intensities were corrected for the transmission function of the analyzer, Wagner sensitivity factors and the electron mean free path. The surface sensitivity was minimized by selecting the spectra at an angle of 21° from the normal to the surface. Since the electron mean free path of the Pt4f and Gd4d photoelectrons are 11.6 Å and 11.1 Å (as estimated from the TPP-2M formula), at this angle the XPS signal originates from a depth of at least 10 Pt monolayers (~23 Å). Therefore the effects of contaminations and of differential sputtering are minimized. For both the sputter cleaned sample and the one transferred to UHV after ORR testing, the concentration ratio was taken as the average of three independent measurements. Only in the case of the sample analyzed after stability test I (see Figure 3 in the text), the Pt:Gd ratio was calculated from a single XPS measurement. For the depth profiles, the electrons emitted at angles between 20° and 80° to the surface normal were analysed in parallel and detected in 16 channels corresponding to 3.75° wideangle intervals. After XPS identification of the elements present at the surface, their main features were measured in detail with AR-XPS. The depth concentration profiles were obtained using the simulation tool, ARProcess (Thermo Avantage software), which uses a maximum entropy method combined with a genetic algorithm. In all cases, the simulations were based on the relative intensities between Pt 4f, O 1s and C 1s, and Gd 4d at each angle, up to 70.6°. The most grazing angles were omitted from the analysis to reduce the influence of diffraction effects and elastic scattering.
S3
Electrochemical measurements All glassware was cleaned for 24 h in a “piranha” solution consisting of a 3:1 mixture of 96% H2SO4 and 30% H2O2, followed by multiple runs of heating and rinsing with ultrapure water (Millipore Milli-Q, 18.2 MΩ cm) to remove sulphates. The electrochemical experiments were performed with a VMP2 potentiostat (Bio-Logic Instruments), controlled by a computer. The rotating ring-disk electrode (RRDE) assemblies were provided by Pine Instruments Corporation. A standard two-compartment glass cell was used, which was equipped with a water jacket attached to a hot water bath to control the temperature. All electrochemical experiments were carried out at 23 °C. We note the ring disk currents are not reported in this publication, as they were consistent with our earlier studies on similar systems [7]. A RRDE assembly was used in lieu of a rotating disk electrode (RDE) assembly, simply because it prevented the walls of the electrode from getting wet. The electrolyte, 0.1 M HClO4 (Merck Suprapur), was prepared with ultrapure water. The counter electrode was a Pt wire and the reference was an Hg/Hg2SO4 electrode, separated from the working electrode compartment using ceramic frits. Following each measurement, the potential of the reference electrode was checked against a reversible hydrogen electrode (RHE) in the same electrolyte. All the potentials in the text are referred to the reversible hydrogen electrode (RHE), and are corrected for ohmic losses. Following each measurement, 0 V vs. RHE was established by carrying out the hydrogen oxidation and hydrogen evolution reactions on Pt in the same electrolyte. The ohmic drop was measured by carrying out an impedance spectrum with a peak-to-peak amplitude of 10 mV, typically from 500 kHz down to 100 Hz. The target resistance was evaluated from the high-frequency intercept on the horizontal (real) axis of the Nyquist plot and further checked by fitting the impedance spectra by using EIS Spectrum Analyser software [9]. The uncompensated resistance came typically to approximately 30 Ω, and was independent of the potential, rotation speed and the presence of O2. This Ohmic resistance is consistent with the value reported by Koper, Markovic and coworkers [10].
S4
The RRDE was immersed into the cell under a potential control of 0.1 V vs. RHE into a N2 (N5, Air Products) saturated electrolyte. The Pt5Gd electrode was cycled in nitrogensaturated electrolytes until stable cyclic voltammograms (CVs) where obtained (100-200 cycles). A typical stable CV on sputtered-cleaned Pt5Gd is shown in Figure S2, and compared to the base CV on polycrystalline Pt. Figure S3 shows cycles 1, 10, 50 and 100 of sputtercleaned Pt5Gd. As shown in the Figure, the shape of the CV changes while cycling. We observed that after ca. 100 cycles the CV was stable. This can be attributed to two factors (a) a certain time frame is required for the surface to dealloy, reconstruct and reach the stable, active phase, and (b) inevitably, some impurities will accumulate on the surface during the transfer from the UHV chamber to the RRDE assembly. In addition to this, the Gd originally on the surface is dissolved in the electrolyte, and there might be slow surface changes before reaching a stable CV. The ORR activity measurements were conducted in an electrolyte saturated with O2 (N55, Air Products).
Figure S2. Cyclic voltammograms at 50 mV s-1 of Pt5Gd (red curve) and Pt (black curve) in a N2saturated 0.1 M HClO4 solution.
S5
Figure S3. Cyclic voltammograms at 50 mV s-1 (cycles 1 (black curve), 10 (dashed blue curve), 50 (dotted green curve), and 100 (red curve)) of sputter-cleaned Pt5Gd in a N2-saturated 0.1 M HClO4 solution.
Computational details The surface was simulated using the method previously developed for Pt5La [6]. It was modelled as a pure Pt overlayer under 6% compressive strain, relative to equilibrium Pt. The compressive strain of 6% was estimated by assuming that the Pt overlayer covered a Cu5Calike phase, with the lattice parameter a = 0.522 nm. The surface of the Pt/Pt5Gd was modelled by a Pt-slab with 6 % compression in the nearest-neighbour distance compared to equilibrium Pt. This allowed us to model the structure of water to approximate that on Pt(111). In the DFT calculations, the strained Pt (111) surface is modelled by a slab with 6 closepacked layers, where the three topmost layers and the adsorbates are allowed to relax. The
S6
ionic cores are described by PAW setups [11], and the Kohn-Sham valence states are described on a real-space grid with a spacing of 0.18 Å. Exchange and correlation effects are described by the RPBE functional [12]. The Kohn-Sham Hamiltonian is diagonalized iteratively using Pulay mixing of the density and Fermi-Dirac occupation of one-electron states with kBT = 0.1 eV. Total energies are extrapolated to kBT = 0 eV. The periodic images of the slab are separated by 20 Å of vacuum. All calculations have been carried out with the ASE and GPAW software packages [13-15]. The adsorption energies are calculated in a (3x2) surface unit cell, and the Brillouin zone is sampled by an (8 x 8 x 1) k-point grid. The slab is relaxed using the Quasi-Newton scheme until the maximum force component is less than 0.05 eV/Å-1. The effect of solvation is included for HO* by incorporating the adsorbate in an H2O*/HO* superstructure with 2/3 ML total coverage. The energy of HO* is calculated from the H2O*/HO* configuration which minimizes the average OH energy. The adsorption energy of H2O* is calculated in a similar superstructure with 2/3 ML total coverage. In the H2O*/HO* superstructure, half of the water molecules lie in a plane approximately parallel to the surface, and the other water molecules lie in a plane perpendicular to the surface with one hydrogen atom pointing away from the surface [16-21].
S7
Additional figures
Figure S4. Cyclic voltammograms at 1600 rpm and 50 mV s-1 in oxygen-saturated 0.1 M HClO4 on Pt5Gd before (red curve) and after: 10,000 cycles between 0.6 V and 1.0 V at 100 mV s-1 (orange curve), 100 (purple curve) and 200 (magenta curve) cycles between 0.05 V and 1.6 V at 50 mV s-1 (around 30 h of experiments). For comparison, the polarization curve for polycrystalline Pt is represented in black. Evidently, the polarization curve for the ORR in the kinetic-control region remains almost unchanged after 10,000 cycles between 0.6 V and 1.0 V. In addition, the hydrogen peroxide formation region does not change after 10,000 cycles between 0.6 V and 1.0 V vs. RHE under the conditions described above.
S8
Figure S5. Non-destructive AR-XPS profile of polycrystalline Pt5Gd after 10,000 cycles between 0.6 V and 1.0 V vs. RHE.
Figure S6. Pt4f and Gd4d XPS spectra of the sputter cleaned Pt5Gd sample. Measurements at 8 of the total 16 angles used for the profile calculation are shown. S9
Figure S7. Pt4f and Gd4d XPS spectra of Pt5Gd sample after ORR testing. Measurements at 8 of the total 16 angles used for the profile calculation are shown.
Figure S8. Pt to Gd concentration ratio as determined from AR-XPS spectra, as a function of the angle for a sputter cleaned sample. The absence of a specific trend is a good indication of the homogeneity of the sample and of the limited effects of differential sputtering.
S10
References (1) Bronger, W. J. Less-Common Metals 1967, 12, 63. (2) Lueken, V. H.; Bronger, W. Z. Anorg. Allg. Chem. 1973, 395, 203. (3) Zhang, K.; Chen, L. Acta Metall. Sin. 1988, 24, B65. (4) Ohm, V.; Raetz, S.; Sauer, M.; Merkens, M.; Schilder, H.; Lueken, H. J. Alloys Comp. 1996, 238, 95. (5) Sauer, M.; Raetz, S.; Ohm, V.; Merkens, M.; Sauer, C.; Schilder, H.; Lueken, H. J. Alloys Comp. 1997, 246, 147. (6) Stephens, I. E.L.; Bondarenko, A. S.; Grønbjerg, U.; Rossmeisl, J.; Chorkendorff, I. Energy Environ. Sci. 2012, 5, 6744. (7) Stephens, I. E.L.; Bondarenko, A.S.; Bech, L.; Chorkendorff, I. ChemCatChem 2012, 4, 341. (8) Johansson, T. P. PhD Thesis, Technical University of Denmark, 2012. (9) Bondarenko, A. S.; Ragoisha, G. A., in Progress in Chemmometrics Research, Nova Sci. Publishers, New York, 2005, pp. 89-102. (10) Van der Vliet, D.; Strmcnik, D. S.; Wang, C.; Stamenkovic, V. R.; Markovic, N. M.; Koper, M. T.M. J. Electroanal. Chem. 2010, 647, 29. (11) Blochl, P. E.; Forst, C. J.; Schimpl, J. Bull. Mater. Sci. 2003, 26, 33. (12) Hammer, B.; Hansen, L. B.; Norskov, J. K. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 7413. (13) Atomic Simulation Environment, https://wiki.fysik.dtu.dk/ase, accessed 14 December, 2011. (14) Mortensen, J. J.; Hansen, L. B.; Jacobsen, K. W. Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 71, 035109. (15) Bahn, S.; Jacobsen, K. Computing in Science & Engineering 2002, 4, 56. (16) Rossmeisl, J.; Karlberg, G. S.; Jaramillo, T.; Nørskov, J. K. Faraday Discuss. 2008, 140, 337. S11
(17) Ogasawara, H.; Brena, B.; Nordlund, D.; Nyberg, M.; Pelmenschikov, A.; Pettersson, L. G. M.; Nilsson, A. Phys. Rev. Lett. 2002, 89, 276102. (18) T. Schiros; Näslund, L.-Å.; Andersson, K.; Gyllenpalm, J.; Karlberg, G. S.; Odelius, M.; Ogasawara, H.; Pettersson, L. G. M.; Nilsson, A. J. Phys. Chem. C, 2007, 111, 15003. (19) Michaelides, A.; Hu, P. J. Chem. Phys. 2001, 114, 513. (20) Feibelman, P. J. Science 2002, 295, 99. (21) Clay, C.; Haq, S.; Hodgson, A. Phys. Rev. Lett. 2004, 92, 046102.
S12
