Report

High-performance transition metal–doped Pt3Ni octahedra for oxygen reduction reaction

See allHide authors and affiliations

Science  12 Jun 2015:
Vol. 348, Issue 6240, pp. 1230-1234
DOI: 10.1126/science.aaa8765

Molybdenum doping drives high activity

Platinum (Pt) is an effective catalyst of the oxygen reduction reaction in fuel cells but is scarce. One approach to extend Pt availability is to alloy it with more abundant metals such as nickel (Ni). Although these catalysts can be highly active, they are often not durable because of Ni loss. Huang et al. show that doping the surface of octahedral Pt3Ni nanocrystals with molybdenum not only leads to high activity (∼80 times that of a commercial catalyst) but enhances their stability.

Science, this issue p. 1230

Abstract

Bimetallic platinum-nickel (Pt-Ni) nanostructures represent an emerging class of electrocatalysts for oxygen reduction reaction (ORR) in fuel cells, but practical applications have been limited by catalytic activity and durability. We surface-doped Pt3Ni octahedra supported on carbon with transition metals, termed M‐Pt3Ni/C, where M is vanadium, chromium, manganese, iron, cobalt, molybdenum (Mo), tungsten, or rhenium. The Mo‐Pt3Ni/C showed the best ORR performance, with a specific activity of 10.3 mA/cm2 and mass activity of 6.98 A/mgPt, which are 81- and 73‐fold enhancements compared with the commercial Pt/C catalyst (0.127 mA/cm2 and 0.096 A/mgPt). Theoretical calculations suggest that Mo prefers subsurface positions near the particle edges in vacuum and surface vertex/edge sites in oxidizing conditions, where it enhances both the performance and the stability of the Pt3Ni catalyst.

Proton-exchange membrane (PEM) fuel cells use reactions between the fuel (such as hydrogen or alcohols) at the anode and the oxidant (molecular oxygen) at the cathode (13). Both cathode and anode reactions need catalysts to lower their electrochemical overpotential for high-voltage output, and so far, platinum (Pt) has been the universal choice (46). To fully realize the commercial viability of fuel cells, the following challenges, which may not be strictly independent of one another, need to be simultaneously addressed: the high cost of Pt, the sluggish kinetics of the oxygen reduction reaction (ORR), and the low durability of the catalysts (711).

Alloying Pt with a secondary metal reduces the usage of scarce Pt metal while at the same time improving performance as compared with that of pure Pt on mass activity (1215), which has led to the development of active and durable Pt-based electrocatalysts with a wide range of compositions (1620). However, although studies so far have led to a considerable increase in ORR activity, the champion activity as observed on bulk Pt3Ni(111) surface has not been matched in nanocatalyts (2125), indicating room for further improvement. At the same time, one noted major limitation of Pt-Ni nanostructures is their low durability. The Ni element in these nanostructures leaches away gradually under detrimental corrosive ORR conditions, resulting in rapid performance losses (2327). Thus, synthesizing Pt‐based nanostructures with simultaneously high catalytic activity and durability remains an important open challenge (28).

Because surface and near-surface features of a catalyst have a strong influence on its catalytic performance, we adopted a surface engineering strategy to further explore and enhance the performance of Pt3Ni(111) nanocatalysts. We specifically focused our efforts on Pt3Ni-based nanocatalysts because the bulk extended Pt3Ni(111) surface has been shown to be one of the most efficient catalytic surfaces for the ORR. On the basis of the control over dopant incorporation of various transition metals onto the surface of dispersive and octahedral Pt3Ni/C (termed as M‐Pt3Ni/C, where M = V, Cr, Mn, Fe, Co, Mo, W, or Re), we have developed ORR catalysts that exhibit both high activity and stability. In particular, our Mo‐Pt3Ni/C catalyst has high specific activity (10.3 mA/cm2), high mass activity (6.98 A/mgPt), and substantially improved stability for 8000 potential cycles.

We prepared highly dispersed Pt3Ni octahedra on commercial carbon black by means of an efficient one‐pot approach without using any bulky capping agents, which used platinum(II) acetylacetonate [Pt(acac)2] and nickel(II) acetylacetonate [Ni(acac)2] as metal precursors, carbon black as support, N,N-dimethylformamide (DMF) as solvent and reducing agent, and benzoic acid as the structure-directing agent (fig. S1A). The surface doping for the Pt3Ni/C catalyst was initiated by the addition of dopant precursors, Mo(CO)6, together with Pt(acac)2 and Ni(acac)2 into a suspension of Pt3Ni/C in DMF, and the subsequent reaction at 170°C for 48 hours (fig. S1B). The transmission electron microscopy (TEM) and high‐angle annular dark‐field scanning TEM (HAADF‐STEM) images of the Pt3Ni/C and Mo-Pt3Ni/C catalysts (Fig. 1, A and B, and fig. S2) revealed highly dispersive octahedral nanocrystals (NCs) in both samples, which were substantially uniform in size, averaging 4.2 ± 0.2 nm in edge length. High-resolution TEM (HRTEM) images taken from individual octahedra showed a single-crystal structure with well‐defined fringes (Fig. 1, C and D) and an edge lattice spacing of 0.22 nm, which is consistent with that expected for face-centered cubic (fcc) Pt3Ni.

Fig. 1 Schematic illustration of the fabrication process and the structure analyses for the transition metal–doped Pt3Ni/C catalysts.

(A and B) Representative HAADF‐STEM images of the (A) Pt3Ni/C and (B) Mo-Pt3Ni/C catalysts. (C and D) HRTEM images on individual octahedral (C) Pt3Ni/C and (D) Mo-Pt3Ni/C nanocrystals. (E and F) EDS line‐scanning profile across individual (E) Pt3Ni/C and (F) Mo-Pt3Ni/C octahedral nanocrystals. (G) Pt, Ni, and Mo XPS spectra for the octahedral Mo‐Pt3Ni/C catalyst.

For Pt3Ni, powder x‐ray diffraction (PXRD) patterns of the colloidal products displayed typical peaks that could be indexed as those of fcc Pt3Ni (fig. S3) (29, 30), and the Pt/Ni composition of 74/26 was confirmed by means of both inductively coupled plasma atomic emission spectroscopy (ICP‐AES) and TEM energy‐dispersive x‐ray spectroscopy (TEM‐EDS) (fig. S4 and table S1). Composition line-scan profiles across octahedra obtained by means of HAADF‐STEM‐EDS for Pt3Ni/C (Fig. 1E) and Mo-Pt3Ni/C (Fig. 1F) showed that all elements were distributed throughout the NCs (Fig. 1, E and F). For the doped NCs, x‐ray photoelectron spectroscopy (XPS) shows the presence of Pt, Ni, and Mo in the catalyst (Fig. 1G). The Ni 2p and Pt 4f XPS spectra of the Mo-Pt3Ni/C catalyst showed that the majority of the surface Ni was in the oxidized state and that the surface Pt was mainly in the metallic state, which were consistent with a recent Pt-Ni catalysts–based study (28). Mo exhibits mainly Mo6+ and Mo4+ states, which is in agreement with previous studies of PtMo nanoparticles (31). The overall molar ratio for Pt, Ni, and Mo obtained from ICP-AES was 73.4:25.0:1.6.

To assess ORR catalytic activity, we used cyclic voltammetry (CV) to evaluate the electrochemically active surface areas (ECSAs). Our catalysts were loaded (with the same Pt mass loading) onto glassy carbon electrodes. A commercial Pt/C catalyst [20 weight percent (wt %) Pt on carbon black; Pt particle size, 2 to 5 nm] obtained from Alfa-Aesar was used as a baseline catalyst for comparison (fig. S5). The CV curves on these different catalysts are compared in Fig. 2A. We calculated the ECSA by measuring the charge collected in the hydrogen adsorption/desorption region (between 0.05 and 0.35 V) after double‐layer correction and assuming a value of 210 μC/cm2 for the adsorption of a hydrogen monolayer. The octahedral Pt3Ni/C and Mo‐Pt3Ni/C catalysts display similar and high ECSAs of 66.6 and 67.5 m2/gPt, respectively, which is comparable with that of the commercial Pt/C catalyst (75.6 m2/gPt) (Fig. 2C, top).

Fig. 2 Electrocatalytic properties of high‐performance transition metal–doped octahedral Pt3Ni/C catalysts and a commercial Pt/C catalyst.

(A) Cyclic voltammograms of octahedral Mo‐Pt3Ni/C, octahedral Pt3Ni/C, and commercial Pt/C catalysts recorded at room temperature in N2‐purged 0.1 M HClO4 solution with a sweep rate of 100 mV/s. (B) ORR polarization curves of octahedral Mo‐Pt3Ni/C, octahedral Pt3Ni/C, and commercial Pt/C catalysts recorded at room temperature in an O2‐saturated 0.1 M HClO4 aqueous solution with a sweep rate of 10 mV/s and a rotation rate of 1600 rotations per min (rpm). (C) The electrochemically active surface area (ECSA, top), specific activity (middle), and mass activity (bottom) at 0.9 V versus RHE for these transition metal–doped Pt3Ni/C catalysts, which are given as kinetic current densities normalized to the ECSA and the loading amount of Pt, respectively. In (A) and (B), current densities were normalized in reference to the geometric area of the RDE (0.196 cm2).

The ORR polarization curves for the different catalysts, which were normalized by the area of the glassy carbon area (0.196 cm2), are shown in Fig. 2B. The polarization curves display two distinguishable potential regions: the diffusion‐limiting current region below 0.6 V and the mixed kinetic‐diffusion control region between 0.6 and 1.1 V. We calculated the kinetic currents from the ORR polarization curves by considering the mass‐transport correction (32). In order to compare the activity for different catalysts, the kinetic currents were normalized with respect to both ECSA and the loading amount of metal Pt. As shown in Fig. 2C, the octahedral Mo‐Pt3Ni/C exhibits a specific activity of 10.3 mA/cm2 at 0.9 V versus a reversible hydrogen electrode (RHE). In contrast, the specific activity of the undoped Pt3Ni/C catalyst is ~2.7 mA/cm2. On the basis of the mass loading of Pt, the mass activity of the Mo‐Pt3Ni/C catalyst was calculated to be 6.98 A/mgPt. The specific activity of the Mo‐Pt3Ni/C catalyst represents an improvement by a factor of 81 relative to the commercial Pt/C catalyst, whereas the mass activity of the Mo‐Pt3Ni/C catalyst achieved a 73‐fold enhancement. To compare the activities of our catalysts with the state-of-the-art reported Pt-Ni catalysts, we also calculated the catalytic activities of our catalysts at 0.95 V and with the ECSA calculated with the CO stripping method. Whether we calculated at 0.90 or 0.95 V or used the ECSA based on Hupd and/or CO stripping, both the specific activity and the mass activity of the Mo‐Pt3Ni/C (fig. S6) are higher than those of the state‐of‐the-art Pt-Ni catalysts (21, 24), including the recently reported Pt-Ni nanoframes catalyst (Table 1 and table S2) (28).

Table 1 Performance of Mo-Pt3Ni/C catalyst and several representative results with high performance from recent published works.

NA, not availlable.

View this table:

Because Mo‐Pt3Ni/C exhibited an exceptional activity toward ORR, we further examined the doping effects for Pt3Ni/C modified by other transition metals. Pt3Ni/C catalysts doped with seven other transition metals—V, Cr, Mn, Fe, Co, W, or Re—were synthesized in a similar fashion with metal carbonyls (figs. S7 and S8 and table S1; details are available in the supplementary materials), and their catalytic activity toward the ORR was tested under the same conditions (Fig. 2C; individual sample measurements are available in fig. S9). The ECSAs of these transition metal–doped Pt3Ni/C catalysts were all similar (Fig. 2C, top), but variable ORR activities were observed for differently doped Pt3Ni/C catalysts. None of the other dopants resulted in a catalyst with activity as high as that of the Mo‐Pt3Ni/C (Fig. 2C, middle). The change of mass activities in various M-doped Pt3Ni/C catalysts was also similar to that of the specific activities (Fig. 2C, bottom), with Mo-Pt3Ni/C showing the highest activity.

We further evaluated the electrochemical durability of the Mo‐Pt3Ni/C catalyst using the accelerated durability test (ADT) between 0.6 and 1.1 V (versus RHE, 4000 and 800 cycles) in O2‐saturated 0.1 M HClO4 at a scan rate of 50 mV/s. The Pt3Ni/C catalyst was used as a baseline catalyst for comparison. After 4000 and 8000 potential cycles, the Mo‐Pt3Ni/C catalyst largely retained its ECSA and activity (Fig. 3A), exhibiting only 1- and 3‐mV shifts for its half‐wave potential, respectively. And after 8000 cycles, the activity of the Mo‐Pt3Ni/C catalyst was still as high as 9.7 mA/cm2 and 6.6 A/mgPt (Fig. 3C), showing only 6.2 and 5.5% decreases from the initial specific activity and mass activity, respectively. On the other hand, the undoped Pt3Ni/C catalyst was unstable under the same reaction conditions. Its polarization curve showed a 33‐mV negative shift after durability tests (Fig. 3B), and the Pt3Ni/C retained only 33 and 41% of the initial specific activity and mass activity, respectively, after 8000 cycles (Fig. 3C). The morphology and the composition of the electrocatalysts after the durability change were further examined. As shown in fig. S4, although the size of the Pt3Ni/C octahedra were largely maintained, their morphologies became more spherical. This change of the morphology likely resulted from the Ni loss after the potential cycles, as confirmed by means of EDS and XPS analyses (the Pt/Ni composition ratio changed from 74.3/25.7 to 88.1/11.9) (figs. S4 and S10). In contrast, the corresponding morphology of the Mo‐Pt3Ni/C catalyst largely maintained the octahedral shape, and the composition change was negligible (from 73.4/25.0/1.6 to 74.5/24.0/1.5).

Fig. 3 Electrochemical durability of the high-performance octahedral Mo-Pt3NiCo/C catalyst and octahedral Pt3Ni/C catalyst.

(A and B) ORR polarization curves and (inset) corresponding cyclic voltammograms of (A) the octahedral Mo‐Pt3Ni/C catalyst and (B) the octahedral Pt3Ni/C catalyst before, after 4000, and after 8000 potential cycles between 0.6 and 1.1 V versus RHE. (C) The changes of ECSAs (left), specific activities (middle), and mass activities (right) of the octahedral Mo‐Pt3Ni/C catalyst and octahedral Pt3Ni/C catalyst before, after 4000, and after 8000 potential cycles. The durability tests were carried out at room temperature in O2‐saturated 0.1 M HClO4 at a scan rate of 50 mV/s.

To investigate the cause of the enhanced durability of the Mo‐Pt3Ni/C catalysts, cluster expansions of Pt‐Ni‐Mo NCs were used in Monte Carlo simulations (3335) to identify low‐energy NC and (111) surface structures for computational analysis (details of our calculations are provided in the supplementary materials). In vacuum, the equilibrium structures predicted by the cluster expansion have a Pt skin, with Mo atoms preferring sites in the second atomic layer along the edges connecting two different (111) facets (Fig. 4, A and B, and fig. S11). Density functional theory (DFT) (36) calculations indicate that in vacuum, the subsurface site is preferable to the lowest‐energy neighboring surface site, but in the presence of adsorbed oxygen, there is a strong driving force for Mo to segregate to the surface, where it was found to be most stable on a vertex site. This suggests the formation of surface Mo‐oxide species, which is consistent with our XPS measurements. Our calculations indicate that the formation of surface Mo-oxide species may contribute to improved stability by “crowding out” surface Ni. Our computational prediction that Mo favors sites near the particle edges and vertices is consistent with the dopant distributions for Fe shown in our STEM electron energy loss spectroscopy (EELS) line scan results (fig. S12).

Fig. 4 Computational results.

(A and B) The average site occupancies of the second layer of (A) the Ni1175Pt3398 NC and (B) the Mo73Ni1143Pt3357 NC at 170°C as determined by means of a Monte Carlo simulation. Occupancies are indicated by the color triangle on the right. Small spheres represent the atoms in the outer layer. (C) The calculated binding energies for a single oxygen atom on all fcc and hcp sites on the (111) facet of the Mo6Ni41Pt178 NC, relative to the lowest binding energy. Gray spheres represent Pt, and colored spheres represent oxygen sites. Three binding energies are provided for reference: the calculated binding energy on the fcc site of a pure Pt (111) surface, the binding energy corresponding to the peak of the Sabatier volcano (37), and the binding energy on a Pt3Ni(111) surface. (D) The change in binding energies when a Ni47Pt178 NC is transformed to a Mo6Ni41Pt178 NC by the substitution of Mo on its energetically favored sites in the second layer below the vertices.

Our calculations suggest that doping the NCs with Mo directly stabilizes both Ni and Pt atoms against dissolution and may inhibit diffusion through the formation of relatively strong Mo‐Pt and Mo‐Ni bonds. Calculations on a representative nanoparticle with dimensions and composition comparable with those observed experimentally (fig. S13) indicate that a Mo on an edge or vertex site increases the energy required to remove a Pt atom from a neighboring edge or vertex site by an average of 362 meV, with values ranging from 346 to 444 meV, and to remove a Ni atom by an average of 201 meV, with values ranging from 160 to 214 meV. These predictions are consistent with our ADT results (fig. S14). The evidence that Mo may have a stabilizing effect on undercoordinated sites suggests that Mo atoms may also pin step edges on the surface, inhibiting the dissolution process.

Although the exact mechanisms by which the surface-doped Pt3Ni shows exceptional catalytic performance demand more detailed studies, local changes in oxygen binding energies provide a possible explanation for some of the observed increase in specific activity. A Sabatier volcano of ORR catalysts predicts that ORR activity will be maximized when the oxygen binding energy is ~0.2 eV less than the binding energy on Pt(111) (37). Our calculations indicate that sites near the particle edge bind oxygenated species too strongly, such as in Pt(111), and sites near the facets of the particles bind oxygenated species too weakly, such as in Pt3Ni(111) (Fig. 4C). However, compared with the undoped NC, the oxygen binding energies in the doped NC near the Mo atoms are decreased by up to 154 meV, and binding energies at sites closer to the center of the (111) facet are increased by up to 102 meV (Fig. 4D). Thus, if Mo migrates to the thermodynamically favored sites near the particle edges, it may shift the oxygen binding energies at these sites closer to the peak of the volcano plot. Similarly, Mo doping may increase the oxygen binding energies at sites closer to the center of the (111) facet that bind oxygen too weakly. As a result of these shifts, some sites may become highly active for catalysis. Together, our studies demonstrate that by engineering the surface structure of the octahedral Pt3Ni nanocrystal, it is possible to fine-tune the chemical and electronic properties of the surface layer and hence modulate its catalytic activity.

Supplementary Materials

www.sciencemag.org/content/348/6240/1230/suppl/DC1

Materials and Methods

Figs. S1 to S17

Tables S1 to S2

References (3953)

References and Notes

  1. Acknowledgments: We acknowledge support from the National Science Foundation (NSF) through award DMR-1437263 on catalysis studies and the Office of Naval Research (ONR) under award N00014-15-1-2146 for synthesis efforts. Computational studies were supported by the NSF through award DMR-1352373 and using computational resources provided by Extreme Science and Engineering Development Environment (XSEDE) through awards DMR130056 and DMR140068. Atomic-scale structural images were generated by using VESTA (38). We thank the Director, Office of Energy Research, Office of Basic Energy Sciences, Materials Sciences and Engineering Division, of the U.S. Department of Energy under contract DE-AC02-05CH11231, under the sp2-bonded materials program, for TEM analytical measurements performed at the National Center for Electron Microscopy at the Lawrence Berkeley National Laboratory. X.D. acknowledges support from the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Materials Science and Engineering through award DE-SC0008055. The work at LLNL was performed under the auspices of the U.S. Department of Energy under contract DE-AC52-07NA27344. A.Y. and A.Z. received additional support from NSF grant EEC-083219 within the Center of Integrated Nanomechanical Systems. We also thank the Electron Imaging Center of Nanomachines at CNSI for TEM support.
View Abstract

Navigate This Article