6772
J. Phys. Chem. B 2007, 111, 6772-6775
Origin of Enhanced Activity in Palladium Alloy Electrocatalysts for Oxygen Reduction Reaction† Minhua Shao, Ping Liu, Junliang Zhang, and Radoslav Adzic* Chemistry Department, BrookhaVen National Laboratory, Upton, New York 11973 ReceiVed: December 28, 2006; In Final Form: March 14, 2007
We explored the origin of the enhanced activity of Pd-alloy electrocatalysts for the O2 reduction reaction by correlating the electrocatalytic activity of intrinsic Pd and Pt surfaces and Pd and Pt overlayers on several substrates with their electronic properties, and established the volcano-type dependence of O2 reduction activity on the binding energy of oxygen and the d-band center of the top metal layer. Intrinsic Pd and Pt surfaces bind oxygen too firmly to allow efficient removal of the adsorbed reaction intermediates. Therefore, they do not have the highest activity and are not on the top of the volcano plot. A Pd overlayer on a Pd3Fe(111) alloy, was predicted to lie on top of the volcano plot, and thus, it appears to be the most active catalyst among investigated ones because of its moderate interaction with oxygen. The results can help in designing better electrocatalysts for fuel cells and other applications.
1. Introduction Fuel cells, especially proton-exchange membrane fuel cells (PEMFCs), are expected soon to become a major source of clean energy.1-3 However, the sluggish kinetics of the oxygen reduction reaction (ORR) is one of the obstacles for their widespread application.4 A considerable ongoing research is focused on searching for non-Pt electrocatalysts that will lower the cost while maintaining high ORR activity. Recent reports from our group,5-7 as well as other published data,8-12 indicated that binary Pd-M alloys (M ) Cr, Co, Fe, Ni, Ti) and ternary ones were comparable to, or slightly better than, commercial Pt catalysts. These Pd-M catalysts have attracted considerable attention because they are comparatively inexpensive and have very high methanol tolerance, making them promising cathodic materials both for hydrogen PEMFC and direct methanol fuel cells (DMFCs). An understanding of the origin of their high activities may help us in designing inexpensive and more active catalysts. Theoretical calculations and experimental data demonstrated that, upon annealing at elevated temperatures, Pd-M alloys undergo phase segregations, in which the noble metal Pd migrates to the surface forming a pure Pd overlayer on the bulk alloys.13-15 The electronic structures of the metal overlayers can alter significantly upon bonding with the substrate metal, and, in turn, their catalytic properties can change.16,17 Nørskov et al. in their d-band center model correlated the electronic structure of the surface metal (represented as the energy center of the valence d-band density of states) and its catalytic activity; the model has been applied to explain the catalytic activity and electrochemical behavior of some strained surfaces and metal overlayers.18-23 Herein, we report our study on the relationship between the surface electronic properties of several Pd or Pt monolayer-modified noble metal- and Pd-based alloy-electrodes and their ORR activities. Volcano-type dependence is observed between the measured ORR activity of Pd or Pt monolayers supported on different noble metal single crystals and their †
Part of the special issue “Norman Sutin Festschrift”. * To whom correspondence should be addressed. E-mail: adzic@ bnl.gov. Phone: 1-631-344-4522. Fax: 1-631-344-5815.
d-band center, or oxygen-binding energy calculated using density functional theory (DFT). On the basis of such dependence and the value of d-band center of a Pd overlayer on an alloy substrate, we predict that Pd alloys should have an ORR activity at least comparable to that of pure Pt. 2. Experimental Section The synthesis of carbon-supported Pd-Fe nanoparticles and deposition of Pt or Pd monolayers on single-crystal electrodes by galvanic displacement of an underpotentially deposited (upd) Cu monolayer can be found elsewhere.6,7 Thin-film electrodes with nanoelectrocatalysts were prepared by placing a certain amount of nanoparticle suspension (1.0 mg mL-1) onto a rotating disk electrode (5 mm diameter, Pine Instrument). After drying in air, the electrode was covered with 5 µL of a dilute Nafion solution. Electrolytes were prepared from Optima perchloric acid obtained from Fisher and UV-plus water (Millipore). A leak-free (Ag/AgCl, 3 M Cl-) electrode with a double-junction chamber (Cypress) and a Pt flag were used as the reference and counter electrode, respectively. All the potentials are given with respect to a reversible hydrogen electrode (RHE). All the electrochemical measurements were carried out at the room temperature. Spin-polarized DFT calculations were performed using the CASTEP suite of programs,24 employing ultrasoft pseudopotentials, a plane wave basis set with a cutoff of 25Ry, and the GGA-RPBE description of the exchange and correlation.25 In order to understand the differences in electronic property between a Pd skin (overlayer) formed at elevated temperatures and a normal Pd surface, our calculations include a pure Pd overlayer supported on PdFe(111) and Pd3Fe(111) surfaces. The ordered Pd-Fe systems were chosen for this calculation because they are thermodynamically stable.26 PdFe(111) and Pd3Fe(111) adopt a tetragonal and a cubic structure, respectively. The calculated lattice constants (a ) b ) 3.74 Å, c ) 3.69 Å for PdFe, a ) b ) c ) 3.83 Å for Pd3Fe) are in reasonable agreement with XRD results. The compositions of the second layer were fixed as 50 atom % Fe for PdFe and 25 atom% Fe for Pd3Fe. The Fe content in the second layer may be higher
10.1021/jp0689971 CCC: $37.00 © 2007 American Chemical Society Published on Web 04/19/2007
Palladium Alloy Electrocatalysts
Figure 1. Top view of (a) Pd3Fe(111), (b) Pd overlayer on Pd3Fe(111), (c) PdFe(111), and (d) Pd overlayer on PdFe(111). The yellow and blue balls represent Fe and Pd atoms, respectively. The red dots indicate the most stable sites for O adsorption on each surface, and the corresponding adsorption energy is shown in Figure 2.
than its bulk molar volume due to the depletion of Pd. The effect of the variation of Fe content on the adsorption energy may be small; i.e., the trend in activity does not significantly depend on the composition of the second layer, as Stamenkovic et al. confirmed in their calculations on a Pt skin on Pt3M.22 Because there is a 2.3% lattice compression in bulk Pd3Fe alloy compared with pure Pd, we also extended our calculations to a 2.3%compressive-strained Pd(111) surface to understand the role of strain. In the present study, the oxygen adsorption on different surfaces was investigated including those at Pd(111), compressed Pd(111), Pt(111), Pd3Fe(111), PdFe(111), Pt overlayer on Pd(111), and Pd overlayer on Pd3Fe(111), PdFe(111), Ru(0001), Ir(111), Rh(111), Pt(111), Au(111). Each surface was constructed as a four-layer slab with a 2 × 2 unit cell and a 11 Å-thick vacuum separating the top and the bottom of the slabs. Brillouin-zone sampling was done on a 4 × 4 × 1 Monkhorst Pack k-point grid. The first two layers were allowed to relax in all dimensions, while the remaining metal atoms were fixed at the lattice position. The binding energy of atomic oxygen was expressed as BEo ) 2*E(O/surface)-E(O2)-2*E(surface). The d-band centers were calculated with an infinite cutoff radius. The d-band value for each surface was expressed as the average of metal atoms surrounding the hollow site where oxygen is adsorbed (see reference19 for details). 3. Results and Discussion The most stable oxygen adsorption configurations on bulkterminated Pd-Fe alloy surfaces, and Pd-Fe alloys-supported Pd overlayers, are presented in Figure 1. For Pd overlayer surfaces (Figure 1b,d), the Pd surrounded hollow sites are the most stable places for oxygen adsorption. However, in the case of Pd3Fe (111) and PdFe(111) with bulk terminations, the most stable adsorption takes place at the hollow sites surrounded by two Pd atoms and one Fe atom, and by one Pd atom and two Fe atoms, respectively. Figure 2 depicts the linear relationship between the calculated Pd or Pt d-band centers (d, relative to the Fermi level) and the binding energy of atomic oxygen (BEo), one of the important intermediates in the ORR. In the present study, the similar results were obtained when using the binding energies of other intermediates (OH or OOH). For all the systems shown in Figure 2, the surface species are either Pt or Pd and the atomic arrangements on surfaces have a close-packed structure. Thus, there is only electronic effect, which makes the binding energy of an adsorbate on one surface different from the other.
J. Phys. Chem. B, Vol. 111, No. 24, 2007 6773
Figure 2. Calculated oxygen-binding energies of Pd- and Ptoverlayers on various substrates as a function of the Pd d-band center (relative to the Fermi level). The energies plotted correspond to the most stable configuration for O adsorption at the Pd or Pt hollow sites on each surface.
Therefore, the variation of the binding energies to OH or OOH is expected to follow the similar trend as that one of atomic oxygen. This is corroborated by Nørskov et al.’s study.27 For oxygen adsorption, an up-shifting in d increases the interaction of the oxygen 2p states with the metal d states, i.e., forms a stronger metal-oxygen bond. Conversely, a downshift of d causes a weak interaction with oxygen. Pd overlayers on Ru(0001) and PdFe(111) lie at the upper left end of the plot with low d values, indicating a weak interaction between O and the Pd overlayers.28 On the other hand, Pd/Au(111) lies at the lower right end with a high d and, thus, provides a strong interaction to oxygen. The four-electron transfer in the ORR on metals of the Pt family occurs through either the direct or the parallel pathway, most likely with the first electron transfer to the adsorbed oxygen molecule being the rate-determining step.4,29 The stronger the Pd-O interaction, the easier is electron transfer and the O-O bond cleavage. The other controlling step is the removal of the adsorbed intermediates (mainly O and OH) formed after the O-O bond’s rupture that, on the other hand, favors a weak Pd-O bond. Thus, following Sabatier’s principle, a good ORR catalyst should exhibit moderate metal-O interaction. Pd overlayers on Ru(0001) and PdFe(111) are not expected to be very active because of their slow kinetics in the ratedetermining step, while Pd and Pd/Au(111) bind O too strongly, thereby hindering the subsequent reaction steps. Furthermore, the tardiness in removing O/OH also leads to the site-blocking in the surface and therefore impedes the adsorption of O2. As stated in the experimental section, the lattice constant of Pd3Fe(111) is smaller than Pd(111) by 2.3% due to the incorporation of Fe atoms into the Pd lattice. To understand the effect of this small compressive strain on the electronic properties of Pd surface, we calculated the d and BEo corresponding to a Pd(111) surface with 2.3% compressive strain and found that the strength of the Pd-O bond weakens by 0.1 eV compared with the unstrained one due to the downshifting of the d (Figure 2). This means that the compressive strain helps to destabilize the adsorption of oxygen. However, this small change in BEo is not sufficient to produce a good catalyst. As shown in Figure 2, the position of d is almost identical to that of Pd/Pt(111), which has been proved less active than Pt.7 For a Pd skin on a Pd3Fe substrate, the BEo further falls by ∼0.25 eV than that of a compressive-strained Pd(111) surface. Note that the lattice constants of both surfaces are 2.3%
6774 J. Phys. Chem. B, Vol. 111, No. 24, 2007
Shao et al.
Figure 4. Volcano dependence of the ORR activity (expressed as the kinetic current density at 0.8 V vs RHE; rotation rate 1600 rpm; room temperature) on the calculated oxygen-binding energy. The experimental data and the predicted current densities of Pd overlayer on Pd3Fe(111) and PdFe(111) based on the calculated oxygen-binding energy are shown as the solid and open squares, respectively.
Figure 3. (a) Comparison of polarization curves for the ORR on several carbon-supported Pd-based and Pt electrocatalysts in an oxygensaturated 0.1 M HClO4 solution obtained using a rotating disk electrode at 10 mV s-1; noble metal loading, 10 µg cm-2; room temperature. All the catalysts were heated in hydrogen at 500 °C for 2 h before electrochemical measurements. (b) Experimentally measured ORR activity (TOFs) of carbon-supported electrocatalysts at 0.85 V derived from part a as a function of the oxygen-binding energy.
smaller than an intrinsic Pd(111) surface. This result reflects the fact that the position of the d-band center depends not only on the amount of strain, but also on electron distribution between the substrate and the metal overlayer (the ligand effect).18 For the surfaces of Pd3Fe(111) and PdFe(111), their values of BEo corresponding to the most stable adsorption conformations are -3.38 and -3.92 eV, respectively. Both bind oxygen too strongly; thus, they are not good catalysts, and are not shown in Figure 2. Figure 3a shows the polarization curves for different carbonsupported nanostructured catalysts in an oxygen-saturated 0.1 M HClO4 solution obtained using a rotating disk electrode (RDE) technique. The activity of Pd was greatly enhanced when Pd and Fe were alloyed in certain ratios (Pd/Fe ) 3:1 is the best ratio according to our previous study6). The average particle size for Pt/C, Pd/C, PdFe/C, and Pd3Fe/C, obtained from TEM measurements, is 3.2 nm, 10.3, 7.6, and 9.7 nm, respectively. Figure 3b shows a comparison of turnover frequencies (TOFs)30 for different nanoelectrocatalysts. The number of active surface atoms used to calculate the TOFs was derived from the charges associated with the hydrogen adsorption in the Hupd region. In agreement with theoretical predictions, the Pd overlayer on Pd3Fe that has a moderate BEo value appears to be the best catalyst for ORR, apparently even better than Pt/C. While comparisons of the ORR activities derived from the nanoparticles data may not be very reliable because of the
possible effects of the particles’ size and shape,31 we verified this electrocatalytic trend with our findings obtained from welldefined single-crystal surfaces. Figure 4 shows the volcano relationship between the kinetic current and the binding energy of oxygen for Pd monolayers on different single-crystal surfaces, including Pt (111) and Pt monolayer on Pd(111). These data were obtained from refs 7 and 20. The activities of Pd overlayers on Pd alloys are predicted on the basis of the calculated binding energy and the corresponding activities read from the above volcano plot. On the left side of the plot, the ORR activity clearly increases almost linearly with weakening of metaloxygen binding. The corresponding kinetic rate of the ORR is limited by the rate of removal of the adsorbed reaction intermediates. However, because of insufficient data, the slope of the right branch, limited by the rate-determining step, is uncertain. For intrinsic Pt(111) surface, the Pt-O bond is still a little too strong for optimum results. A small weakening of it, such as found for a surface of Pt/Pd(111), can enhance the ORR activity remarkably.20 According to Stamenkovic et al.’s proposed model,22 the best catalyst was predicted to bind oxygen more weakly than pure Pt by 0.2 eV. The optimal BEo that corresponds to the top position of our volcano plot is around -1.86 eV (Figure 4). This is consistent with Stamenkovic et al.’s prediction, given that the BEo of Pt is -2.04 eV. We calculated that the BEo of the Pd overlayer on Pd3Fe(111) was -1.85 eV. Therefore, we predict that this surface is very close to the top of the volcano, and should have very high ORR activity. The surface of a Pd overlayer on PdFe(111) will be less active because it binds oxygen too strongly. The volcano plots can also explain our previous results,6 wherein the ORR activity of Pd-Fe nanoparticles showed a linear relationship with the Pd-Pd interatomic distance. To a very large extent, this phenomenon may reflect the linear dependence of the ORR activity on the d-band center, which also is proportional to the strain in the surface. The propensity of Pd to form PdOH, the reaction inhibitor that does not take part in the reaction, is another reason of its low activity for the ORR. It is caused by its high reactivity and thus, it is related to the d-band center position. Usually, the PdOH coverage decreases with alloying, Therefore, it is not expected that the behavior of PdOH can alter the d-band center-activity correlation. The prediction of
Palladium Alloy Electrocatalysts very high ORR activity of Pd overlayers needs to be further confirmed by studies of Pd alloy single-crystal surfaces, which is in progress. Two similar thermodynamic guidelines recently were proposed for designing non-Pt alloy electrocatalysts for ORR. Bard et al.10,32 suggested that for Pd-M alloys the metal M constitutes the site for breaking the O-O bonds. For Pd with fully occupied valence d-orbitals, Balbuena et al.33 proposed that alloying with transition metals, such as Co with unoccupied valence d-orbitals, reduces significantly the Gibbs free energy both for the first charge-transfer step, and for the steps involving the reduction of intermediates. However, these thermodynamic models did not take into account the strong surface segregation in many Pd-M systems. 4. Conclusions The experimental data and DFT calculations were used to correlate the electrocatalytic activity of modified extended surfaces and metal overlayers with their electronic properties. The volcano-type dependence of the ORR activity on the binding energy of oxygen and the d-band center of the noble metal overlayers was established. Intrinsic Pd and Pt surfaces bind oxygen too strongly to allow efficient removal of the adsorbed reaction intermediates, and, hence, have slow ORR kinetics. The Pd overlayer on Pd3Fe(111) formed during annealing was the most active, and correspondingly, lay on the top of the volcano plot. The reason for its good activity is that this overlayer affords a moderate bond with oxygen that balances well the two competing influences, i.e., the first electron transfer or O-O bond-breaking, and the removal of O and OH. Our model not only predicts the high ORR activity of Pd-Fe alloy, but offers guidance for designing better electrocatalysts than pure Pt. Consequently, in developing them, both Pt-containing and nonPt alloys should be considered first. In addition, designs should take into account the strong surface-segregation effect in alloys and mixed metal system. Finally, we note that alloying Pd with Pt and some transition metal should lower the d-band position of the noble-metal overlayer and, therefore, may modify the activity significantly by inducing strain and electron redistribution between the substrate and the overlayer. Although most of the systems we described are models, we believe that the results can help designing better catalysts for fuel cells, and can be of value for other applications. Acknowledgment. This work is supported by US Department of Energy, Divisions of Chemical and Material Sciences, under the Contract No. DE-AC02-98CH10886. M.H.S acknowledges partial support from Department of Materials Science and Engineering, State University of New York at Stony Brook. The authors thank K. Sasaki for conducting the XRD measurements.
J. Phys. Chem. B, Vol. 111, No. 24, 2007 6775 References and Notes (1) Dresselhaus, M. S.; Thomas, I. L. Nature 2001, 414, 332. (2) Jacobson, M. Z.; Colella, W. G.; Golden, D. M. Science 2005, 308, 1901. (3) Vielstich, W.; Lamm, A.; Gasteiger, H. A. Handbook of Fuel CellsFundamentals, Technology and Applications; John Wiley & Sons: Chichester, 2003; Vols. 1-4. (4) Adzic, R. R. In Electrocatalysis; Lipkowski, J., Ross, P. N., Eds.; Wiley: New York, 1998; p 197. (5) Adzic, R. R. In DOE Hydrogen and Fuel Cell ReView Meeting Philadelphia, PA, 2004. (6) Shao, M. H.; Sasaki, K.; Adzic, R. R. J. Am. Chem. Soc. 2006, 128, 3526. (7) Shao, M. H.; Huang, T.; Liu, P.; Zhang, J.; Sasaki, K.; Vukmirovic, M. B.; Adzic, R. R. Langmuir 2006, 22, 10409. (8) Savadogo, O.; Lee, K.; Oishi, K.; Mitsushimas, S.; Kamiya, N.; Ota, K.-I. Electrochem. Commun. 2004, 6, 105. (9) Lee, K.; Savadogo, O.; Ishihara, A.; Mitsushimas, S.; Kamiya, N.; Ota, K.-I. J. Electrochem. Soc. 2006, 153, A23. (10) Fernandez, J. L.; Walsh, D. A.; Bard, A. J. J. Am. Chem. Soc. 2005, 127, 357. (11) Fernandez, J. L.; Raghuveer, V.; Manthiram, A.; Bard, A. J. J. Am. Chem. Soc. 2005, 127, 13100. (12) Raghuveer, V.; Manthiram, A.; Bard, A. J. J. Phys. Chem. B 2005, 109, 22909. (13) Ruban, A. V.; Skriver, H. L.; Norskov, J. K. Phys. ReV. B 1999, 59, 15990. (14) Rousset, J. L.; Bertolini, J. C.; Miegge, P. Phys. ReV. B 1996, 53, 4947. (15) Bozzolo, G.; Noebe, R. D.; Khalil, J.; Morse, J. Appl. Surf. Sci. 2003, 219, 149. (16) Rodriguez, J. A. Surf. Sci. Rep. 1996, 223. (17) Hammer, B.; Norskov, J. K. Theoretical surface science and catalysis - Calculations and concepts. AdV. Catal. 2000, 45, 71. (18) Kitchin, J. R.; Norskov, J. K.; Barteau, M. A.; Chen, J. G. J. Chem. Phys. 2004, 120, 10240. (19) Xu, Y.; Ruban, A. V.; Mavrikakis, M. J. Am. Chem. Soc. 2004, 126, 4717. (20) Zhang, J. L.; Vukmirovic, M. B.; Xu, Y.; Mavrikakis, M.; Adzic, R. R. Angew. Chem., Int. Ed. 2005, 44, 2132. (21) Kibler, L. A.; El-Aziz, A. M.; Hoyer, R.; Kolb, D. M. Angew. Chem., Int. Ed. 2005, 44, 2080. (22) Stamenkovic, V.; Mun, B. S.; Mayrhofer, K. J. J.; Ross, P. N.; Markovic, N. M.; Rossmeisl, J.; Greeley, J.; Norskov, J. K. Angew. Chem., Int. Ed. 2006, 45, 2897. (23) Lima, F. H. B.; Zhang, J.; Shao, M. H.; Vukmirovic, M. B.; Ticianelli, E. A.; Adzic, R. R. J. Phys. Chem. C 2007, 111, 404. (24) Payne, M. C.; Allan, D. C.; Arias, T. A.; Johannopoulus, J. D. ReV. Mod. Phys. 1992, 64, 1045. (25) Hammer, B.; Hansen, L. B.; Nørskov, J. K. Phys. ReV. B 1999, 59, 7413. (26) Hansen, M. Constitution of Binary Alloys; McGraw-Hill: New York, 1958. (27) Norskov, J. K.; Rossmeisel, J.; Logadottir, A.; Lindqvist, L. R. K. J.; Bligaard, T.; Jonsson, H. J. Phys. Chem. B 2004, 108, 17886. (28) Greeley, J.; Norskov, J. K.; Mavrikakis, M. Annu. ReV. Phys. Chem. 2002, 53, 319. (29) Shao, M. H.; Liu, P.; Adzic, R. R. J. Am. Chem. Soc. 2006, 128, 7408. (30) Paulus, U. A.; Wokaun, A.; Scherer, G. G.; Schmidt, T. J.; Stamenkovic, V.; Markovic, N. M.; Ross, P. N. Electrochim. Acta 2002, 47, 3787. (31) Mayrhofer, K. J. J.; Blizanac, B. B.; Arenz, M.; Stamenkovic, V. R.; Ross, P. N.; Markovic, N. M. J. Phys. Chem. B 2005, 109, 14433. (32) Fernandez, J. L.; White, J. M.; Sun, Y.; Tang, W.; Henkelman, G.; Bard, A. J. Langmuir 2006, 22, 10426. (33) Wang, Y. X.; Balbuena, P. B. J. Phys. Chem. B 2005, 109, 18902.
