Report

Improved Oxygen Reduction Activity on Pt3Ni(111) via Increased Surface Site Availability

See allHide authors and affiliations

Science  26 Jan 2007:
Vol. 315, Issue 5811, pp. 493-497
DOI: 10.1126/science.1135941

Abstract

The slow rate of the oxygen reduction reaction (ORR) in the polymer electrolyte membrane fuel cell (PEMFC) is the main limitation for automotive applications. We demonstrated that the Pt3Ni(111) surface is 10-fold more active for the ORR than the corresponding Pt(111) surface and 90-fold more active than the current state-of-the-art Pt/C catalysts for PEMFC. The Pt3Ni(111) surface has an unusual electronic structure (d-band center position) and arrangement of surface atoms in the near-surface region. Under operating conditions relevant to fuel cells, its near-surface layer exhibits a highly structured compositional oscillation in the outermost and third layers, which are Pt-rich, and in the second atomic layer, which is Ni-rich. The weak interaction between the Pt surface atoms and nonreactive oxygenated species increases the number of active sites for O2 adsorption.

When a polymer electrolyte membrane fuel cell (PEMFC) is used in a demanding application such as an automobile, it must overcome the kinetic limitations on the oxygen reduction reaction (ORR), which have led to three fundamental problems (15). First, the substantial overpotential for the ORR (610) at practical operating current densities reduces the thermal efficiency well below the thermodynamic limits, typically to about 43% at 0.7 V [versus the theoretical thermal efficiency of 83% at the reversible potential for the ORR (1.23 V)]. Second, an approximately five-fold reduction of the amount of Pt (platinum-loading) in current PEMFC stacks is needed to meet the cost requirements for large-scale automotive applications (10). Finally, the dissolution and/or loss of Pt surface area in the cathode must be greatly reduced.

These limitations could be eliminated if stable cathode catalysts, with an order of magnitude increase in the specific activity over that of state-of-the-art Pt/C catalysts, can be developed (10). In the hope that a combination of different metals would have improved catalytic activity and stability relative to those of a pure metal, the ORR has been studied on numerous bi- or multimetallic alloys (6, 8, 1117). These studies have led to incremental improvements to catalyst performance, but large increases in activity have yet to be realized.

Rather than use a trial-and-error or combinatorial approach, we have examined selected cathode materials with well-characterized surfaces so that the mechanism of action can be attributed to a specific property (at the atomic and molecular level) of the surface under study. In this way, we can determine (i) whether the kinetics of the ORR are structure-sensitive, (ii) the composition of the topmost surface atomic layers (the segregation profile), and (iii) how alloying [usually described in terms of the ligand effect or/and ensemble effect (1820)] alters the chemical properties of the surfaces. Similar approaches are commonly used in gas-phase catalysis (21) under ultrahigh vacuum (UHV) and near-ambient conditions, but alloy surface chemistry on single-crystal surfaces at electrochemical interfaces is relatively unexplored. These aqueous interfaces are more complex in that they necessarily contain solvent and electronic/ionic charge, and (experimentally) it is very challenging but still tractable to use in situ surface-sensitive methods to characterize potential-induced changes in the surface properties and reactivity.

We have used a combination of ex situ and in situ surface-sensitive probes and density functional theory (DFT) calculations to study the ORR on Pt3Ni(hkl) single-crystal surfaces, identify which surface properties govern the variations in reactivity of PtNi catalysts, and determine how surface structures, surface segregation, and inter-metallic bonding affect the ORR kinetics. Well-characterized PtNi single-crystal electrode surfaces were formed and characterized with UHV methods for surface preparation and surface analysis. These surfaces were transferred into the electrochemical environment without airborne contamination, and the stability of the UHV-prepared surface was determined with a combination of in situ surface-sensitive probes with electrochemical methods so that activity relationships could be obtained in real time.

The results of the preparation and characterization of Pt3Ni(hkl) alloy surfaces in UHV are summarized in Fig. 1. The surface-sensitive techniques that were used included low-energy electron diffraction (LEED), Auger electron spectroscopy (AES), low-energy ion scattering (LEIS), and synchrotron-based high-resolution ultraviolet photoemission spectroscopy (UPS) (see methods in the supporting online material). Each of these methods has certain advantages, and they yield complementary information. The surface symmetry obtained from LEED analysis shows that, whereas the Pt3Ni(111) surface exhibits a (1 × 1) pattern (Fig. 1D) (i.e., that of the bulk termination), the atomically less dense Pt3Ni(100) surface shows a (1 × 5) reconstruction pattern (the so-called “hex” phase) in both the [011] and [01-1] directions (Fig. 1E). Analysis of the Pt3Ni(110) LEED data (Fig. 1F) indicates that this surface may exhibit a mixture of (1 × 1) and (1 × 2) periodicities, the latter being known as the (1 × 2) missing-row structure (22).

Fig. 1.

Surface characterization of the Pt3Ni single crystals in UHV and electrochemical environments. (A) AES spectra. (B) LEIS spectra. E/E0, where E is energy of scattered electrons, and E0 is the energy of the incident ion beam. (C) UPS spectra. (D to F) LEED patterns with corresponding ball models. (G to I) Cyclic voltammetry (CV) in HClO4 (0.1 M) as compared to the voltammetry of the corresponding Pt single crystal (gray curves). RHE, reversible hydrogen electrode.

The composition of the outermost atomic layer was obtained with LEIS, as previously shown for Pt3Ni polycrystalline alloys (15, 23); after a final anneal, the surface atomic layer of all three Pt3Ni(hkl) crystals is pure Pt (Fig. 1B) [i.e., all form the so-called Pt-skin structures (17, 20)]. Previous studies of PtNi single-crystal surfaces in UHV had indicated that this surface enrichment of Pt in the first layer is counterbalanced by its depletion in the next two to three atomic layers, so that the concentration profile oscillates around the bulk value (24, 25). We demonstrated that segregation-driven near-surface compositional changes result in distinctive electronic properties of the PtNi (hkl) alloys.

The surface electronic structures were obtained from the background-corrected UPS spectra (26). As summarized in Fig. 1C, the d-band density of states (DOS) is structure-sensitive, and the position of the d-band center shifts from –2.70 eVon Pt3Ni(110) to –3.10 eVon Pt3Ni(111) to –3.14 eV on Pt3Ni(100). Furthermore, the DOS of the alloy surfaces is quite different from that of corresponding pure Pt single crystals; that is, on the (110), (100), and (111) alloy surfaces, the d-band center is downshifted by about 0.16, 0.24, and 0.34 eV, respectively. An important yet largely unexplored issue in surface electrochemistry on alloys is understanding how chemisorption energies correlate with the average energy of the d-state on the surface atoms to which the adsorbate binds (i.e., the ligand effect). To do this, we compared the results on the (111) surfaces with the same composition and arrangement of surface atoms but with a different d-band center position. Although we are highlighting how the difference between the electronic surface structures of Pt(111) and Pt3Ni(111) can affect the adsorption of spectator species and the kinetics of the ORR, the findings are generally applicable to other single-crystal alloy surfaces.

We next determined the stability of a surface after transfer from UHV into the electrochemical environment. Surface x-ray scattering (SXS) was used to characterize both the potential range of stability as well as the near-surface composition of the alloy in situ. Only results for Pt3Ni(111) are presented because they provide the most useful information about the potential-induced changes in the surface structure and segregation profile; a brief description of results for (110) and (100) can be found in (27). The termination of the Pt3Ni(111) lattice at the surface in terms of elemental composition and surface relaxation was determined by measurement and analysis of the crystal truncation rods (CTRs) (28). Pt3Ni(111) has the face-centered cubic (fcc) lattice with random occupation of sites by Pt and Ni, and this lattice gives rise to SXS similar to that obtained from a monocrystalline fcc lattice. Sensitivity to atomic layer composition is enhanced by the use of anomalous x-ray scattering techniques in which the incident x-ray energy is tuned to an atomic adsorption edge of the material (29). Thus, from SXS, we obtained information about the positions and compositions, both in the surface and subsurface layers. The CTR analysis shows that, at 0.05 V, the first layer is composed entirely of Pt and, whereas the second atomic layer is Ni-rich (52% of Ni as compared to 25% of Ni in the bulk), the third layer is again Pt-enriched (87%) (Fig. 2A). Having determined the near-surface structure at 0.05 V, the potential was cycled while the scattered x-ray intensity was measured at a CTR position that is sensitive to surface relaxation and surface roughness (Fig. 2A). We find that both the Pt3Ni(111) surface structure as well as the segregation profile are completely stable over this potential range because the changes in the x-ray scattering signal are fully reversible, and we find that the decrease in intensity at positive potential is consistent with an inward relaxation (contraction) of the surface atomic layer (this result is confirmed by similar measurements at other reciprocal lattice positions). The contraction of Pt surface atoms is induced by the adsorption of oxygenated species, which is determined by the Ni-induced modification of the Pt-skin electronic structure. A direct consequence of contraction of the topmost layer of Pt-skin at the potentials higher than 0.8 V is the increased stability of this surface over corresponding pure Pt(111), which was additionally confirmed by prolonged cycling in the designated potential range.

Fig. 2.

In situ characterization of the Pt3Ni(111) surface in HClO4 (0.1 M) at 333 K. (A) SXS data and (A') concentration profile revealed from SXS measurements. at. %, atomic %. (B) Cyclic voltammetry in designated potential region (red curve) as compared to the voltammetry obtained from Pt(111) surface (blue curve). (C) Surface coverage calculated from cyclic voltammograms of Pt3Ni(111) (red curve) and Pt(111) (blue curve); polarization curves obtained from rotating ring disk electrode (RRDE) measurements. ΘOxide, surface coverage by adsorbed spectator oxygenated species. (D) Green scale refers to hydrogen peroxide production in designated potential region and (E) ORR currents measured on Pt3Ni(111) (red curve), Pt(111) (blue curve), and polycrystalline Pt (gray curve) surfaces. The arrows are showing the positive potential shift of 100 mV in electrode half-potential (ΔE112) between ORR polarization curves measured on Pt-poly and Pt3Ni(111) surfaces. I, II, and III represent potential region of Hapd adsorption/desorption processes, double-layer region, and region of OHad layer formation, respectively.

The relations between the surface electronic properties and the potential-dependent surface coverage by adsorbing species (the adsorption isotherms in Fig. 2C) were established by comparing the experimentally determined position of the d-band centers to the fractional coverages of adsorbed hydrogen (H+ +e = Hupd, where Hupd refers to the underpotentially deposited hydrogen) between 0.05 < E <0.4 V, where E is the applied potential, and hydroxyl species (2H2O = OHad +H3O+ +e, where OHad is the adsorbed hydroxyl layer) above 0.6 V (Fig. 2C). Careful inspection of the voltammograms in Fig. 2B revealed that on Pt(111)-skin, which consists of the same surface density of Pt atoms as Pt(111), a dramatic negative shift (≈0.15 V) in Hupd formation and positive shift (≈0.1 V) in OHad formation occurred relative to Pt(111). In agreement with the onset of adsorption, on Pt3Ni(111), the fractional coverages by Hupd and OHadHupd and ΘOHad) were dramatically reduced by 50% relative to Pt(111), which is in agreement with the large downshift (0.34 eV) of the d-band center position on the Pt-skin structure. As shown in Fig. 1, H and I, similar changes occurred for the other two single-crystal surfaces. On the Pt(100)-skin, ΘHupd was reduced by ≈15% relative to Pt(100), and ΘOHad was reduced by ≈25%. On Pt3Ni(110), small but clearly discernable decreases in ΘHupd (≈10%) and the fractional coverage by OHadOH) (≈20%) were observed relative to Pt(110).

To describe these effects quantitatively, we performed DFT calculations, using pseudopotentials and a plane-wave basis set, on the adsorption of OH and H2O at 0.25 monolayer coverage on modeled Pt(111) surfaces, with second atomic layers containing 0 or 50% Ni atoms. In acid solutions, OHad would react with H+ and form H2O on the catalyst surface. The change in the reversible potential ΔU ° of the above reaction on Pt(111) resulting from sublayer Ni atoms is Math(1) Math Here, Eads(OH)Pt, Eads(H2O)Pt, Eads(OH)PtNi, and Eads, (H2O)PtNi are the adsorption energies of OH and H2O on Pt(111) with or without sublayer Ni atoms, respectively, and F is the Faraday constant. The DFT calculations show a positive shift of ΔU° = 0.10 V when the sublayer has 50% Ni atoms. The experiment and theory thus reach an excellent and quantitative agreement in this case and clearly establish an electronic effect of subsurface Ni on the Pt-OH chemical bonding.

The ORR is a multielectron reaction (½O2 + 2H+ + 2e = H2O) that may include a number of elementary steps involving different reaction intermediates (7, 30). We proposed that the rate of the ORR can be expressed as Math(2) Math where i is the measured current; n, F, K, E, x, β, γ, and R are constants [(3); for definitions, see (31)]; cO2 is the concentration of O2 in the solution, Θad is the total surface coverage by adsorbed spectator species [hydroxyl and anions; for example, OHadOH) and specifically adsorbed anions (ΘAad)]; ΔGad is the Gibbs energy of adsorption of reactive intermediates, and T is temperature. In the derivation of Eq. 2, it is assumed that (i) the ORR takes place on electrodes that are modified by OHad, anions, etc., and (ii) the reactive intermediates (O 2 and H2O2) are adsorbed to low coverage (i.e., they are not a substantial part of Θad).

Based on these assumptions, the kinetics of O2 reduction are determined by the number of free Pt sites available for the adsorption of O2 (the 1 – Θad term in Eq. 2) and by the ΔGad of O2 and reaction intermediates (theΔGad term in Eq. 2) on metal surfaces precovered by OHad. We used this reaction pathway and rate expression first to analyze the effects of electronic properties on the kinetics of the ORR on Pt3Ni(111) and Pt(111) and then, by comparing activities on different Pt3Ni(hkl) surfaces, to establish structure sensitivity.

A characteristic set of polarization curves (the relation of i versus E) for the ORR on Pt-poly, Pt(111), and Pt3Ni(111) surfaces in HClO4 (0.1 M) at 333 Kis summarized in Fig. 2D. For all three surfaces, the polarization curves exhibit two distinguishable potential regions. By starting at 0.05 V and scanning the electrode potential positively, well-defined diffusion-limiting currents from 0.2 to 0.7 Vare followed by a mixed kinetic-diffusion control region between 0.8 < E < 1.0 V. Further inspection of Fig. 2D reveals that the ORR kinetic is accelerated on the Pt(111)-skin relative to Pt(111), causing the positive shift of 100 mV in the half-wave potential. Given that ΘOHad is attenuated on the Pt-skin structure, the key parameter that determines the unusually catalytic activity of Pt3Ni(111) is the low coverage by OHad [i.e., the (1 – ΘOHad) term in the kinetic equation for the ORR]. Additional confirmation that the fractional coverages by spectator species are indeed controlling the kinetics of the ORR was found by analyzing the results in the potential region where the adsorption of hydrogen is taking place (E < 0.2 V). Because of the lower coverage by Hupd, the production of peroxide is substantially attenuated on the Pt-skin surface. At the fuel cell relevant potentials (E > 0.8 V), the observed catalytic activity for the ORR on Pt3Ni(111) is the highest that has ever been observed on cathode catalysts, including the Pt3Ni(100) and Pt3Ni(110) surfaces.

Synergy between surface geometry and surface electronic structure for the ORR is illustrated in Fig. 3. As summarized, the different low-index surfaces have markedly different activity for this reaction [that is, Pt3Ni(100)-skin < Pt3Ni(110)-skin <<< Pt3Ni(111)-skin, with the change in activity between the least active (100) and the most active (111) surfaces being greater than an order of magnitude]. Structure sensitivity of the ORR on the Pt low-index single-crystal surfaces in perchloric acid has been well established (3), with activities increasing in the order Pt(100) ≪ Pt(111) < Pt(110) (Fig. 3). These differences have been attributed to the structure-sensitive adsorption of OHad on Pt(hkl) and its inhibiting (site blocking) effect on O2 adsorption. In the potential region of OH adsorption, the structure sensitivity of the Pt3Ni(hkl)-skin surfaces has the same origin.

Fig. 3.

Influence of the surface morphology and electronic surface properties on the kinetics of ORR. RRDE measurements for ORR in HClO4 (0.1 M) at 333 K with 1600 revolutions per minute on Pt3Ni(hkl) surfaces as compared to the corresponding Pt(hkl) surfaces (a horizontal dashed gray line marks specific activity of polycrystalline Pt) are shown. Specific activity is given as a kinetic current density ik, measured at 0.9 V versus RHE. Values of d-band center position obtained from UPS spectra are listed for each surface morphology and compared between corresponding Pt3Ni(hkl) and Pt(hkl) surfaces.

In order to reveal the role of the electronic structure onto ORR kinetic, we compared electrodes with the same surface morphology. The most pronounced effect was observed on the (111) surfaces. For the same p(1 × 1) arrangement of the topmost layer, the same surface atomic density, and the same surface composition (100% Pt) but for a different electronic structure (|Δd[111]| = 0.34 eV, where |Δd[hkl]| is the d-band center shift), the ORR is being enhanced by factor of 10 on Pt(111)-skin relative to that on Pt(111). Given that extended Pt surfaces have 5 to 10 times the activity per surface Pt atoms than the state-of-the-art Pt/C catalysts that are currently used in the PEMFC (∼0.2 mA/cm2 at 0.9 V) (10, 14), a total 90-fold enhancement of Pt(111)-skin versus that of Pt/C has been achieved.

Considering that the Pt3Ni(111)-skin surface exhibits the highest catalytic activity that has ever been detected, the challenge would be to create a nanocatalyst with electronic and morphological properties that mimic the Pt3Ni(111) surface. Therefore in the future, a way to reduce the current value (∼1.0 g of Pt per kW) of Pt-specific power density in a PEMFC without a loss in cell voltage, while also maintaining the maximum power density (W/cm2), would be the engineering of Pt3Ni(111)-skin–like nanocatalysts.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1135941/DC1

Materials and Methods

References and Notes

View Abstract

Navigate This Article