ARTICLES PUBLISHED ONLINE: 11 AUGUST 2014 | DOI: 10.1038/NCHEM.2032

High-performance Ag–Co alloy catalysts for electrochemical oxygen reduction Adam Holewinski1, Juan-Carlos Idrobo2 and Suljo Linic1 * The electrochemical oxygen reduction reaction is the limiting half-reaction for low-temperature hydrogen fuel cells, and currently costly Pt-based electrocatalysts are used to generate adequate rates. Although most other metals are not stable in typical acid-mediated cells, alkaline environments permit the use of less costly electrodes, such as silver. Unfortunately, monometallic silver is not sufficiently active for economical fuel cells. Herein we demonstrate the design of low-cost Ag–Co surface alloy nanoparticle electrocatalysts for oxygen reduction. Their performance relative to that of Pt is potential dependent, but reaches over half the area-specific activity of Pt nanoparticle catalysts and is more than a fivefold improvement over pure silver nanoparticles at typical operating potentials. The Ag–Co electrocatalyst was initially identified with quantum chemical calculations and then synthesized using a novel technique that generates a surface alloy, despite bulk immiscibility of the constituent materials. Characterization studies support the hypothesis that the activity improvement comes from a ligand effect, in which cobalt atoms perturb surface silver sites.

E

ffective catalysis of the oxygen reduction reaction (ORR), O2 + 4H+ + 4e− → 2H2O in acid or O2 + 2H2O + 4e− → 4OH− in base, remains a major contemporary technological hurdle. Low-temperature proton-exchange membrane fuel cells (PEMFCs) are a promising technology for efficient power delivery in transportation and mobile devices. Commercialization of the technology is hampered by the slow rate of the ORR, which requires costly electrocatalysts with a high Pt content to achieve adequate power densities. Accounting for economies of scale, typical Pt nanoparticle catalysts represent about half the cost (projected by the US Department of Energy (DOE)) of an automotive PEMFC stack required to deliver ∼80 kW (ref. 1). Even current state-of-the-art Pt-alloy electrocatalysts, which achieve about twice the ORR activity per gram of Pt–group metal (PGM), will still contribute ∼35% to costs of equivalent-power stacks. Furthermore, the long-term stability of current state-of-the-art catalysts still falls short of the DOE 5,000 hour lifetime target (to 10% voltage loss) for most transportation applications1,2. With current projections, including balance-ofplant costs, it has been determined that a viable Pt-based catalyst needs to exhibit at least a fourfold enhancement over pure Pt in PGM mass activity, with similar or better long-term stability1–4. Alternatively, if a Pt-free material—a so-called costless catalyst— can be utilized, activity targets can be decreased by about an order of magnitude (on a volumetric basis) before reaching electrodethickness limits for effective mass transport and a break-even point in cost because of the need for larger cell components2–5. One of the fundamental reasons for the general lack of success in identifying viable alternatives to Pt-based ORR catalysts is the issue of stability. As the majority of metals are unstable in the acidic oxidative environment of a PEMFC cathode, very few materials can be considered. However, in alkaline environments a number of alternative electrode-material candidates are stable. This is illustrated in the Pourbaix diagram in Fig. 1, in which equilibrium potentials for the most thermodynamically favourable oxidation process of a variety of metals are plotted as a function of pH6. In base, Pt retains its superior ORR activity compared with that of other metals and shows a quantitatively similar performance as in acid,

although common contaminants in basic electrolytes can suppress this activity over time7,8. The Pourbaix diagram shows that Ag stands out as having a superior electrochemical stability in base. Another appealing feature of Ag is that at, a cost of ∼$1 g−1 (roughly 50 times cheaper than Pt)9, it is a comparatively very inexpensive catalyst. The main deficiency of Ag is that it exhibits an area-specific activity roughly an order of magnitude lower than that of Pt; that is, the surface Ag atoms are approximately ten times less active than Pt atoms10–14. Assuming a similar balance of plant, at current Ag performance levels an Ag-based anionexchange membrane fuel cell (AEMFC) stack that matches the performance of Pt-based PEMFC stacks would be a break-even exchange at best—the savings that result from the cheaper catalyst would be offset by the higher cost of larger cell components. However, any improvements in the activity of Ag-based electrocatalysts, along with continuing advances in the design of more-stable hydroxide-exchange membranes15–17 and improved anode electrocatalysts for alkaline hydrogen oxidation18, have the potential to make AEMFCs competitive with current state-ofthe-art PEMFCs. In this contribution we demonstrate the design of Pt-free Ag–Co alloy electrocatalysts for the ORR. We show that the Ag alloy exhibits a more than fivefold improved performance in area-specific alkaline ORR activity compared to pure Ag electrocatalysts at 0.8 V versus the reversible hydrogen electrode (RHE), and over half the specific activity of Pt nanoparticles at the same potential. The discovery of the alloy electrocatalysts was guided by quantum-chemical density functional theory (DFT) calculations. A novel bimetallic-precursor synthesis technique, which yielded a near-surface alloy of Ag and Co, was employed to synthesize the materials. Cyclic voltammetry (CV) surface characterization and cobalt-selective electrochemical leaching provided evidence that Co atoms act through a ligand mechanism to perturb the Ag surface sites and improve their inherent electrochemical activity. The atomistic surface structure of the materials was characterized further with aberration-corrected scanning transmission electron microscopy (STEM), coupled with atomic resolution

1

Department of Chemical Engineering, University of Michigan, Ann Arbor, Michigan 48109, USA, 2 Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA. * e-mail: [email protected]

828

NATURE CHEMISTRY | VOL 6 | SEPTEMBER 2014 | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved

NATURE CHEMISTRY

ARTICLES

DOI: 10.1038/NCHEM.2032 2.5

1.4

OOH* + OH– O2 /H

2O

Ag/Ag+

0.6

Pt/PtO Pd/Pd O

Cu/Cu++

2.0

2 O3

1.5

Ag/Ag

2O

ΔG (eV)

Potential (VNHE)

1.0

Au/Au

0.2 Cu/Cu H2 /H +

–0.6

OH)

2

Co/Co

η = 0.43 V

O* + 2OH–

2

4

6

8

10

12

OH* + 3OH– 4OH–

O2 0.0

–1.0 2

0.80 VRHE

0.5

Ni/Ni(

(OH)

0

AgNi AgCo AgFe Pt

2O

Ni/Ni++ Co/Co++

–0.2

1.0

Ag AgCu

14

1.23 VRHE 4e–

3e–

2e–

1e–

0e–

–0.5

pH

Reaction coordinate

Figure 1 | Pourbaix diagram for the oxidation of various metals. Potential–pH boundaries for the most thermodynamically favourable oxidation processes of various metals6 are shown. All phase boundaries assume the relevant metal-ion activities of either unity or the saturated equilibrium concentration if less. At high pH, the metallic state is generally preserved at potentials closer to the oxygen reduction equilibrium potential. Metal-cation solubilities also generally decrease with pH, and oxidative dissolution gives way to oxidative passivation, which makes the electrode less prone to dissolution.

electron-energy loss spectroscopy (EELS) mapping that showed the presence of Co in the catalyst surface layers.

Results and discussion DFT calculations of elementary step energetics for ORR. The ORR involves four electron transfers to the O2 molecule. The elementary step mechanism has been studied extensively, and it has been shown that on the metals of interest (that is, Pt and Ag), the kinetically limiting step involves an initial reduction of O2 to form an adsorbed OOH intermediate19–22. Subsequent generation of adsorbed O and OH, and the formation of H2O (or OH− in base, thermodynamically equivalent on the RHE scale), are rapid and effectively equilibrated. On Pt electrodes it has been postulated that the most-abundant reaction intermediate under operating potentials above ∼0.8 VRHE is adsorbed OH, which limits the reaction by effectively blocking sites (a high equilibrium coverage)22–24. In comparison, Ag is relatively more noble and only forms a surface oxide at higher potentials25–27. This suggests that, under the conditions of interest (0.7–0.9 VRHE), the rate of the reaction on Ag is governed almost completely by the initial rate of OOH formation. The mechanistic details discussed above are consistent with data in Fig. 2, which shows free-energy paths obtained using DFT calculations (see Methods)28 for the relevant reaction intermediates involved in the ORR on pure Pt(111) and Ag(111) surfaces, as well as the (111) surface of several model Ag alloys. The freeenergy diagram illustrates that, in contrast to Pt, for which high OH coverage thermodynamically limits the performance, the performance of Ag electrocatalysts is limited by the reduction of O2 to form OOH. More-active Ag-based electrocatalysts should bind the OOH intermediate more strongly than pure Ag, and thereby accelerate the rate-limiting step. The data in Fig. 2 indicate that this criterion is satisfied on alloys of Ag and the late 3d transition metals (Co, Fe, Ni and Cu). The Ag-alloy materials also bind other O-containing species, including OH, more strongly than pure Ag, but not to a degree at which their removal would be limiting. Such stronger binding characteristics are actually rather counterintuitive when bond-order conservation is considered. One would expect the reactive 3d element to interact strongly with surface Ag atoms, and so reduce their affinity for other species, as

Figure 2 | DFT-calculated free-energy diagram for ORR on alloy catalyst surfaces. Free energies of ORR surface intermediates on (111) surfaces of Ag, various Ag alloys and Pt at 0.8 V with respect to the RHE. The model system for Ag alloys (inset) is a periodic slab with four atomic layers that contain 50% guest metal (violet) in the first subsurface layer of Ag (blue), and 25% in the bottom layers. The pure Ag lattice constant was used in all cases (except pure Pt), and all free energies were calculated at a coverage of 0.25 monolayer adsorbate. η, electrical overpotential.

is the case for alloys of Pt and 3d metals. Recently, we developed a general framework for understanding these counterintuitive adsorbate-binding trends in terms of electronic structure, described elsewhere29,30. Although a multitude of different alloy structural configurations are conceivable, the trend in activity reported in Fig. 2 is identical for all alternative models investigated. Irrespective of the alloy composition (metal ratios, bulk mixing versus surface mixing only, etc.), any perturbation of Ag surface atoms by adjacent 3d metal atoms results in a stronger interaction of OOH with the Ag site. ORR free-energy diagrams for alloys with different geometries (location of the 3d metals and degree of lattice contraction) are given in Supplementary Fig. 1. In each of these systems we have assumed that the outermost layer of the alloy is occupied by Ag atoms because of the low surface energy of Ag. Although we acknowledge that functioning systems contain a number of complexities not present in our models, such as the presence of spectator species, there is no physical basis to expect these factors to impact the relative trends in the activity. Synthesis. Based on the near-optimal characteristics for Ag–Co alloys shown in Fig. 2, we chose to focus on these materials. A major obstacle to the synthesis of Ag–3d alloys is that these systems are immiscible in the bulk phase. In addition, the large differences in the reduction potentials of Ag and 3d metals generally cause Ag to reduce much faster than the 3d metals, which makes even a kinetic entrapment of 3d-metal atoms in the Ag matrix challenging. It has been shown that this issue can be addressed by reducing precursors rapidly under extreme conditions, such as laser ablation or γ-ray irradiation31. Utilizing the idea of rapid reduction, we developed a more-scalable synthesis approach based on a modified incipient-wetness technique (depicted schematically in Fig. 3a), adapted from concepts in several works on different metal alloys32–35. The water-soluble precursors K3[Co(CN)6] and AgNO3 were added sequentially to a Vulcan XC72 carbon support (1:3 Co:Ag molar ratio, 20 wt% metals) and formed an insoluble salt, Ag3[Co(CN)6], which achieved atomic-scale mixing of the two metals. After rinsing away excess precursors, an almost pure phase of Ag3[Co(CN)6] nanoparticles remained, confirmed by X-ray diffraction (XRD) (see Supplementary Fig. 2). The bimetallic salt particles were then introduced into a preheated furnace at 500 °C under dilute hydrogen for ten minutes and

NATURE CHEMISTRY | VOL 6 | SEPTEMBER 2014 | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved

829

ARTICLES

NATURE CHEMISTRY

a i. K3[Co(CN)6] ii. AgNO3

Rapid heating

Vulcan XC72

500 °C, H2/Ar

d

DOI: 10.1038/NCHEM.2032

Ag AgCo (heat/quench) Ag/Co (slow heating) Pt AgCo leached (heat/quench) 0.90

b

Current (mA cm–2)

1.5

Potential (VRHE)

2.0 Cycle 1 Cycle 2 Cycle 5

1.0 0.5

0.85

2 µA

0.80

0.75 VRHE

0.0

0.2

0.3

0.4

0.5

0.70 –0.5 –0.2

1

0.1 –0.1

0.0

0.1

0.2

10

–2 ) Log current (mass trans corrected) (mA cmAg

0.3

Potential (VRHE) 1

e 10 Ag Co AgCo (heat/quench) AgCo leached (heat/quench) Ag/Co (slow heating) Pt

–1

Specific activity (mA

Current (mA cm–2)

0

–2 cmAg )

c

–2 –3 –4

3 Pt AgCo leached Ag

8

AgCo slow reduction AgCo heat/quench

2 6

4 1 2

0

0

–5 0.0

0.2

0.4

0.6

0.8

1.0

0.80 VRHE

0.85 VRHE

Potential (VRHE)

Figure 3 | Synthesis scheme and electrochemical oxygen-reduction performance of Ag–Co surface alloy materials. a, Schematic of Ag3[Co(CN)6] precursor formation and rapid reduction to form Ag–Co surface alloys. Ag, blue, Co, violet; C, green; N, black. b, Leaching CV in Ar-purged 0.001 M H2SO4 + 0.1 M Na2SO4 , which shows dissolution of Co on the first cycle, followed by featureless scans afterwards. c, Rotating disk electrode I–V polarization curves (potentials relative to the RHE) taken in O2-saturated 0.1 M NaOH at 25 °C and 900 revolutions per minute for the Ag–Co surface alloy, as well as a leached alloy, a segregated Ag/Co sample and pure Ag and Co. d, Mass-transport corrected kinetic current Tafel plots for the samples. The inset shows Pb-stripping voltammograms used to measure the electrochemical surface area at 10 mV s−1 in Ar-purged electrolyte with 125 µM Pb(NO3)2 added after the activity measurements. e, Summary of kinetic current densities for oxygen reduction on each sample, measured at 0.8 VRHE and 0.85 VRHE. Error bars represent the standard deviation for three or more separately synthesized samples (except commercial Pt, which varied mainly because of a higher electrolyte sensitivity).

quenched. Subsequent characterization of the material, discussed below, revealed that the synthetic method cannot produce complete bulk alloying, but can be optimized to isolate enough Co in the near-surface region of Ag to modify its activity. Although we are aware that the utilization of Ag–3d-metal combinations for ORR has been attempted in a few prior reports, we emphasize that these works used conventional synthesis approaches that did not produce alloys or show large differences in activity36–38. Our rapid-heating procedure was found to be critical to achieving surface alloying, and slowly heated samples exhibited minimal activity enhancements. Electrochemical performance. Figure 3c shows the electrochemical polarization behaviour of the Ag–Co catalyst, prepared as discussed above using rapid heating and cooling, and referred to here as ‘asprepared’ alloy. Measurements were performed in the thinfilm rotating disc electrode (RDE) configuration39. Also shown in Fig. 3c are the performance for Ag–Co prepared by slower heating (10 °C min−1), a ‘preleached’ Ag–Co (any Co exposed to electrolyte was removed from the sample by controlled 830

electrochemical dissolution from the as-prepared Ag–Co (see Fig. 3b and the Supplementary Information)), elemental Ag and Co electrocatalysts (prepared under the same conditions), as well as a commercial Pt-nanoparticle catalyst (20 wt%, 5 nm Pt on Vulcan carbon). The active electrochemical surface area of Ag for each sample was measured using Pb-stripping voltammetry40,41 (inset of Fig. 3d). The polarization curves in Fig. 3c show that both as-prepared and leached Ag–Co alloy catalyst particles have substantially higher ORR activity than pure Ag or slowly heated Ag–Co, rising to over half the activity of Pt at 0.8 VRHE. The slowly heated Ag–Co showed a performance comparable to that of pure Ag, which is not surprising because the mixing between Co and Ag in these samples is negligible. As a result of two-electron peroxide production, pure Co exhibits very poor activity and a much lower limiting current than the Ag materials. It can also be seen that the asprepared and leached alloy materials actually start off with relatively similar activity to Ag at high potentials and then separate at lower potentials—that is, the performance enhancement is highest at a larger overpotential. Figure 3e shows that the enhancement in Ag NATURE CHEMISTRY | VOL 6 | SEPTEMBER 2014 | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved

ARTICLES

DOI: 10.1038/NCHEM.2032

Electrochemical surface analysis. To help understand the origin of the observed ORR activity enhancement we probed the surface and subsurface structure of Ag–Co particles with CV, which can provide direct chemical identification of species exposed to the electrolyte. In Fig. 4, we show CVs for the as-prepared and leached Ag–Co catalyst particles, as well as for pure Ag and Co samples, with upper cycling potential limits of 1.00 VRHE and 1.15 VRHE. These potential limits were selected to investigate differences between non-destructive surface oxidation, observed for potentials below 1 VRHE , and destructive surface oxidation at a higher potential, which induces roughening and exposes atoms from the subsurface layers of the particles to the electrolyte. Most critically, insight into the distribution of cobalt in Ag–Co can be gained by comparing the presence or absence of cobalt oxidation peaks across samples and potential limits. In Fig. 4, two overlapping peaks in the 0.1–0.3 V range, which correspond to Co(OH)2 and CoO formation6,45,46, are clearly visible on pure Co particles and the as-prepared Ag–Co, whereas they are absent on the preleached material when kept below 1 VRHE. This is consistent with the hypothesis that leaching removes any Co atoms that are accessible to the electrolyte. However, when the alloy surface is roughened

a

0.2 Ag Co

Current (mA cm–2)

area-specific activity reaches a factor of ∼6 at 0.8 VRHE , whereas it is ∼2.5 at 0.85 VRHE. The pure Ag and Ag–Co samples exhibit similar size distributions (average diameters of roughly 8 nm and 10 nm, respectively), and samples of Ag with larger and smaller sizes (controlled with changes to reduction temperature and metal loading) did not show appreciable differences in area-specific activity. Size histograms and activity summaries for all samples are shown in Supplementary Fig. 3 and Supplementary Table 1. We also assessed selectivity with Koutecky–Levich analyses, shown in Supplementary Fig. 4, and found that the number of electrons transferred for all Agbased materials is close to four over a wide range of potentials, which indicates a virtually complete selectivity to H2O. Long-term stability of the Ag and Ag–Co ORR catalysts was assessed by a DOE protocol for accelerated ageing in the RDE three-electrode cell configuration1. The electrodes were aged by repeated voltage cycling between 0.6 VRHE and 1.0 VRHE at 50 mV s−1 in O2-saturated electrolyte. ORR activity was assessed every 500 (2.5 hours) or 1,000 (five hours) cycles by measuring polarization curves across the voltage range of –0.1 VRHE to 1.0 VRHE at 10 mV s−1. The performance of the Ag and Ag–Co catalysts over the course of 10,000 cycles is provided in Supplementary Fig. 5. We found that both Ag and Ag–Co exhibit similar ageing behaviour, with losses in kinetic current at 0.8 VRHE plateauing at roughly 50% of the initial kinetic currents. The aged Ag–Co electrocatalysts exhibited rates that are still 2–3 times higher than that of pristine Ag before any ageing. Activity losses also closely scaled with measured losses in the surface area for both materials. Although factors such as area loss through particle growth likely contribute to the degradation process, experiments showed that periodically pushing the potential below 0 VRHE led to partial activity regeneration. Thus, we hypothesize that a major reason for the similar performance degradation of both materials is a gradual accumulation of trace anions, some of which can be removed by reduction. Sulfate, carbonate and silicates were present at roughly ten parts per 109 levels in the electrolyte solution and are known to affect Ag electrodes. We discuss this in more detail in Supplementary Fig. 5. Although a long-term stability comparison with Pt is desirable, we found that Pt electrodes exhibit an even higher sensitivity towards basic electrolyte contaminants than Ag (Supplementary Fig. 6). Commercial Pt particles in acidic electrolytes exhibit somewhat better stability than our Ag-based samples in base, losing about 50% of their initial activity with extended cycling to 30,000 cycles, although losses are generally asymptotic with most degradation in the first 10,000 cycles42–44.

AgCo AgCo leached

0.1

0.0

–0.1

–0.2 –0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.8

1.0

1.2

Potential (VRHE)

b

Current (mA cm–2)

NATURE CHEMISTRY

0.2

0.1

0.0

–0.1

–0.2 –0.2

0.0

0.2

0.4

0.6

Potential (VRHE)

Figure 4 | Comparative CV of the Ag–Co surface alloy. a,b, CV was performed with potential limits to yield non-destructive (a) and destructive (b) surface oxidation. This revealed the presence of subsurface Co, identified by the emergence of a Co oxidation peak near 0.1 V on destructive oxidation of preleached Ag–Co. Scans were recorded in Ar-purged 0.1 M NaOH for the as-prepared and leached Ag–Co catalyst particles, as well as for pure Ag and Co control samples, with upper potential limits of 1.00 VRHE (a) and 1.15 VRHE (b). Arrows indicate the scan direction.

by cycling the preleached sample to 1.15 V, the low-potential cobalt oxidation peak returns, which supports the notion that appreciable Co is retained after leaching, but that it is trapped in the subsurface layers of Ag. Cycling to 1.15 V essentially exposes the Co atoms buried in the particles to the electrolyte. The identification of subsurface Co, taken in concert with the poor activity of physical mixtures of Ag and Co formed by slow heating, supports the hypothesis that the activity enhancement of the rapidly reduced Ag–Co is imparted by a ligand effect whereby Co coordinated to Ag perturbs the reactivity of Ag sites. Also, the Ag–Co material exhibits features (Fig. 4) that are not seen on either pure Ag or pure Co. As-prepared and preleached Ag–Co surface alloys show unique oxidation peaks at ∼0.6 V and ∼0.8 V. These peaks may correspond to oxygen species that bind to modified Ag surface sites on the alloy, which is consistent with the hypothesis that Ag–Co alloys should be more reactive than Ag towards oxygen-containing species (O, OH, OOH). Also consistent with a stronger oxygen-binding energy is the observation that the sharp current increase near 1.1 V sets in at a lower potential on Ag–Co compared with that on Ag. Ex situ surface characterization. In Supplementary Fig. 2 we show a STEM image of the as-prepared Ag–Co alloy catalyst particles and corresponding Ag and Co elemental energy-dispersive X-ray spectroscopy (EDS) maps based on their Kα fluorescence signals.

NATURE CHEMISTRY | VOL 6 | SEPTEMBER 2014 | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved

831

ARTICLES

NATURE CHEMISTRY

a BF

d As-prepared

DOI: 10.1038/NCHEM.2032

e Leached

1 nm

2 nm

20 nm

b HAADF ADF

ADF

Ag EELS

Ag EELS Co EELS

ii iii

i

c

f OK

OK

Co L2,3

v vi

Ag M3 M2

M1

600

700

Co L2,3

vi v Intensity (a.u.)

Ag M3 M2

Intensity (a.u.)

Ag M4,5

iv

Co EELS

M1

iv iii ii i

300

500

700

900

Energy loss (eV)

500

800

900

Energy loss (eV)

Figure 5 | Electron microscopy and spectroscopy of Ag–Co materials. Atomically resolved EELS imaging was used to characterize the composition of the near-surface region of representative catalyst particles. a,b, Bright field (BF) (a) and high-angle annular dark field (HAADF) (b) (Z-contrast) STEM images of an Ag–Co surface alloy particle. c, Spectra of electron-energy loss acquired at several points on the particle surface. d,e, Reference images for as-prepared (d) and leached (e) samples with insets showing spectral images of the marked regions for Ag (green) and Co (red), as well as a simultaneously acquired Z-contrast image signal. f, Corresponding sample spectra that show the oxygen and cobalt ionization-edge regions at marked points (i–vi).

The elemental maps point to a significant segregation between Co and Ag. Supplementary Fig. 2 also shows XRD scans taken for a representative sample of the Ag3[Co(CN)6] precursor before and after rapid reduction at 500 °C. The diffraction pattern of the Ag– Co particles, formed in the reduction of this precursor, closely resembles that of pure Ag. Control samples of metallic cobalt particles also reduced at 500 °C, as well as several cobalt oxide standards (not shown), indicate that there are no detectable oxides, but a phase of very small, pure face-centred cubic (fcc) Co nanoparticles exists in the bimetallic product. These techniques point to catalyst particles that are not significantly alloyed in the bulk phase. Although a uniform bulk alloy was not formed, we sought a more-direct characterization of the near-surface region of the Ag–Co catalyst particles using EELS in an aberration-corrected STEM. Figure 5a–c shows a Z-contrast STEM image of a typical Ag–Co alloy catalyst particle, with energy-loss spectra taken with the electron beam scattering through several different points at the edge of the particle. The spectra from the edge of the particle gave the electronic fingerprint of primarily near-surface atoms and showed a distinct Co L2,3 absorption edge at 780 eV, as well as the Ag M4,5 edge with onset just below 400 eV. These spectra indicate that Co atoms are present near the particle surface. The Co was more prevalent near the surfaces of larger particles (above the mean diameter), but because of a log-normal distribution the vast majority of the surface area is concentrated in these larger particles. To gain a more detailed picture of the local spatial distribution of Co in the Ag particle surface, we also performed atomic-resolution 832

EELS imaging. Figure 5d,e shows reference images of representative catalyst particles from as-prepared and leached samples, respectively. A Z-contrast image (top insets in Fig. 5d,e) was collected simultaneously with the EELS to visualize any specimen drift over the course of the acquisition. The inset images in Fig. 5d,e just below the Z-contrast images show maps of the integrated intensities for the Ag and Co ionization edge signals (green is the Ag M4,5 region and red is the Co L2,3 region) averaged over each acquisition region of the Z-contrast image. EEL spectra that correspond to several regions (numerically labelled in Fig. 5d,e) are shown in Fig. 5f for the as-prepared and leached samples. The spectra are presented with background subtraction in a window that precedes the oxygen K edge. Figure 5 provides clear evidence that there is mixing of Ag and Co close to the surface of the Ag–Co particles, and that after the leaching process Co is retained in the particles. We also find that the Co signal is generally diminished and more frequently accompanied by oxygen in the leached samples. One could argue that the presence of oxygen results from oxidized Co, highly dispersed at the particle surface. This argument could temper the conclusion that activity increases are related to a ligand effect as opposed to bimetallic/oxide interface sites. We thus recall that physical mixtures of Ag and Co formed by a slower heating in the synthesis are poor catalysts; that is, Ag and Co oxide in close proximity (at the surface of a nanoparticle) are no better than pure Ag. This argument points to the subsurface Co-ligand effect as the primary source of enhanced activity. Thus, it may be concluded that a greater degree of alloying between Ag and Co, or other 3d elements, will likely yield even more robust and active ORR catalysts. NATURE CHEMISTRY | VOL 6 | SEPTEMBER 2014 | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved

NATURE CHEMISTRY

ARTICLES

DOI: 10.1038/NCHEM.2032

Methods Catalyst synthesis. The carbon-supported Ag–Co nanoparticles were synthesized by multistage incipient-wetness impregnation. In a typical synthesis 0.25 mmol K3[Co(CN)6] and 1.125 mmol AgNO3 (both Sigma Aldrich) were dissolved in separate 2 ml solutions of deionized water. The solutions were added dropwise, alternating 1 ml of each precursor at a time, to 440 mg Vulcan XC72 (Cabot, pretreated for two hours in H2 at 400 °C), with drying at 60 °C between each millilitre of water added, for a total of four impregnation/drying cycles. This formed Ag3[Co(CN)6] particles with excess AgNO3 and KNO3, removed by triply rinsing and centrifuging the powder with deionized water, followed by drying again. The Ag/C and Co/C particles were prepared to comparable molar loadings by single-step incipient-wetness impregnation using the same precursors, with the modification of postreduction rinsing for Co to remove KCN. For the reduction, precursor powders were inserted into a preheated tube furnace at 500 °C under 4% H2/Ar for ten minutes and then quenched with flowing air on the tube exterior. Pt samples were commercially available, 20 wt% 5 nm Pt on Vulcan carbon (Sigma Aldrich). Electrochemical testing. Electrochemical measurements were performed at room temperature in an all-Teflon three-electrode cell (made in house) with a Gamry Instruments Reference 3000 potentiostat/galvanostat/FRA. The reference (Hg/HgO in 1 M KOH, Radiometer Analytical) and counter electrodes (Pt wire, Alfa Aesar) were both in isolated compartments with long diffusion paths to the working electrode chamber. Working electrodes were prepared by sonicating the catalyst powders in 99.9% acetone (Fisher) at 0.75 mg ml−1 for more than an hour and dispersing two 12 µl droplets onto a 5 mm glassy carbon electrode insert (Pine Instruments), which was polished with 0.05 µl alumina paste and sonicated in ultrapure water prior to use. These electrodes were put into a Teflon RDE housing, attached to a rotator (Pine Instruments) and inserted into the cell under a potential control using a Pt-wire dummy electrode, temporarily connected to the working lead. Electrolyte solutions of 0.1 M NaOH were prepared from ultrapure water (18.2 MΩ cm (Millipore)) and 99.9995% sodium hydroxide (Fluka TraceSelect). A consistent uncompensated resistance of ∼40 Ω was measured via high-frequency impedance, and was corrected for in the polarization curves. All potentials are reported relative to the RHE (VRHE = VNHE + 0.0591 pH (NHE, normal hydrogen electrode)), calibrated by H2 oxidation equilibrium. Cyclic voltammograms were taken in Ar-purged solution, and rotating disc polarization curves were taken in an O2-saturated solution. Leaching, where applicable, was performed in 0.001 M H2SO4 + 0.1 M Na2SO4 using CV from –0.3 to 0.1 VNHE to dissolve Co and leave Ag intact. Pb-stripping voltammetry was performed immediately after polarization measurements by purging the solution and adding 125 µM Pb(NO3)2 , then repeatedly holding the potential at 0.20 VRHE (which caused Pb underpotential deposition) and sweeping to 0.6 VRHE at 10 mV s−1. The stable voltammograms were integrated assuming 280 µC cm−2 (refs 40,41). Pt voltammograms were taken at 100 mV s−1, and the hydrogen underpotential region (0.05 VRHE to 0.45 VRHE) was integrated assuming 210 µC cm−2 for area estimates. Electrochemical methods are provided in more detail in the Supplementary Information. Characterization. STEM with EDS was carried out in a JEOL 2010F 200 keV field-emission electron microscope with an EDAX X-ray detector and Genesis software package. Aberration-corrected imaging and atomic-resolution EELS were performed at Oak Ridge National Laboratory with a Nion UltraSTEM 100 operated at a 100 kV accelerating voltage47. The device is equipped with a cold fieldemission electron source and is capable of third- and fifth-order aberration correction. EEL spectra were collected using a Gatan Enfina spectrometer, with an EELS collection semiangle of 48 mrad. Energy dispersion for the spectrometer was 0.5 eV per pixel. XRD measurements were conducted in a Rigaku rotating anode diffractometer with a monochromated Cu Kα X-ray source at a scan rate of 1° min−1. Theoretical methods. All DFT calculations were carried out with the Dacapo pseudopotential plane wave code (www.camd.dtu.dk) using the generalized gradient approximation and revised Perdew, Burke and Ernzerhof exchange correlation functional. All surface calculations used a four-layer 2 × 2 fcc [111] periodic unit cell separated by 12 Å of vacuum space in the [111] direction and a dipole layer to decouple the slabs electrostatically. The bottom two layers were fixed and the top two layers and all adsorbates were relaxed until the sum of forces was below 0.05 eV Å−1. Ultrasoft pseudopotentials were used to represent the ionic cores, with the valence electron density determined through iterative diagonalization of the Kohn–Sham Hamiltonian using Pulay mixing. Unit cells were sampled with a 4 × 4 × 1 Monkhorst–Pack k-point grid, and the plane-wave basis-set energy cutoff was 350 eV. An electronic temperature of 0.1 kBT was used, with final energies extrapolated to 0 K. Potential-dependent free-energy diagrams were calculated by using the relation ΔEadsorption = Esystem – Eslab – Eadsorbates , and then converting into Gibbs free energy with the relationship ΔG = ΔE – TΔS + ΔZPE + ne(U – U0). Entropy changes (ΔS) were obtained from tabulated values (webbook.nist. gov/chemistry/). Zero point energies (ZPE) were taken from prior DFT work28. The contribution of potential was computed assuming equilibrium between H2 and H+/e− at 0.0 VNHE and shifting energies by the number of electrons (n) times the elementary charge (e) and the potential change (U – U0). On an RHE scale, all the

intermediates considered shifted in energy by equivalent amounts on a change in pH, so the reaction energies between these states are independent of pH to a good approximation. Additional corrections for stabilization effects by the aqueous interface were also made in accordance with prior studies48,49.

Received 11 March 2014; accepted 10 July 2014; published online 11 August 2014; corrected after print 11 August 2014

References 1. The US Department of Energy (DOE). Energy Efficiency and Renewable Energy; www1.eere.energy.gov/hydrogenandfuelcells/mypp/ 2. Debe, M. K. Electrocatalyst approaches and challenges for automotive fuel cells. Nature 486, 43–51 (2012). 3. Borup, R. et al. Scientific aspects of polymer electrolyte fuel cell durability and degradation. Chem. Rev. 107, 3904–3951 (2007). 4. Gasteiger, H. A., Kocha, S. S., Sompalli, B. & Wagner, F. T. Activity benchmarks and requirements for Pt, Pt-alloy, and non-Pt oxygen reduction catalysts for PEMFCs. Appl. Catal. B 56, 9–35 (2005). 5. Wagner, F. T., Lakshmanan, B. & Mathias, M. F. Electrochemistry and the future of the automobile. J. Phys. Chem. Lett. 1, 2204–2219 (2010). 6. Pourbaix, M. Atlas of Electrochemical Equilibria in Aqueous Solutions (National Association of Corrosion Engineers, 1974). 7. Oezaslan, M., Hasché, F. & Strasser, P. Oxygen electroreduction on PtCo3 , PtCo and Pt3Co alloy nanoparticles for alkaline and acidic PEM fuel cells. J. Electrochem. Soc. 159, B394–B405 (2012). 8. Subbaraman, R. et al. Origin of anomalous activities for electrocatalysts in alkaline electrolytes. J. Phys. Chem. C 116, 22231–22237 (2012). 9. US Geological Survey. Mineral Commodity Summaries 2010; http://minerals. usgs.gov/minerals/pubs/mcs/ (2010). 10. Spendelow, J. S. & Wieckowski, A. Electrocatalysis of oxygen reduction and small alcohol oxidation in alkaline media. Phys. Chem. Chem. Phys. 9, 2654–2675 (2007). 11. Blizanac, B. B., Ross, P. N. & Markovic, N. M. Oxygen reduction on silver low-index single-crystal surfaces in alkaline solution: rotating ring disk (Ag(hkl)) studies. J. Phys. Chem. B 110, 4735–4741 (2006). 12. Coutanceau, C., Demarconnay, L., Lamy, C. & Leger, J. M. Development of electrocatalysts for solid alkaline fuel cell (SAFC). J. Power Sources 156, 14–19 (2006). 13. Wiberg, G. K. H., Mayrhofer, K. J. J. & Arenz, M. Investigation of the oxygen reduction activity on silver – a rotating disc electrode study. Fuel Cells 10, 575–581 (2010). 14. Singh, P. & Buttry, D. A. Comparison of oxygen reduction reaction at silver nanoparticles and polycrystalline silver electrodes in alkaline solution. J. Phys. Chem. C 116, 10656–10663 (2012). 15. Varcoe, J. R. & Slade, R. C. T. Prospects for alkaline anion-exchange membranes in low temperature fuel cells. Fuel Cells 5, 187–200 (2005). 16. Varcoe, J. R., Slade, R. C. T., Wright, G. L. & Chen, Y. L. Steady-state dc and impedance investigations of H2/O2 alkaline membrane fuel cells with commercial Pt/C, Ag/C, and Au/C cathodes. J. Phys. Chem. B 110, 21041–21049 (2006). 17. Adams, L., Poynton, S. & Tamain, C. A carbon dioxide tolerant aqueouselectrolyte-free anion-exchange membrane alkaline fuel cell. ChemSusChem 1, 79–81 (2008). 18. Strmcnik, D. et al. Improving the hydrogen oxidation reaction rate by promotion of hydroxyl adsorption. Nature Chem. 5, 300–306 (2013). 19. Blizanac, B. B., Ross, P. N. & Markovic, N. M. Oxygen electroreduction on Ag(111): the pH effect. Electrochim. Acta 52, 2264–2271 (2007). 20. Stamenkovic, V. R. et al. Improved oxygen reduction activity on Pt3Ni(111) via increased surface site availability. Science 315, 493–497 (2007). 21. Stephens, I. E. L., Bondarenko, A. S., Grønbjerg, U., Rossmeisl, J. & Chorkendorff, I. Understanding the electrocatalysis of oxygen reduction on platinum and its alloys. Energy Environ. Sci. 5, 6744–6762 (2012). 22. Holewinski, A. & Linic, S. Elementary mechanisms in electrocatalysis: revisiting the ORR Tafel slope. J. Electrochem. Soc. 159, H864–H870 (2012). 23. Markovic, N. M. & Ross, P. N. Surface science studies of model fuel cell electrocatalysts. Surf. Sci. Rep. 45, 117–229 (2002). 24. Tripković, V., Skúlason, E., Siahrostamia, S., Nørskov, J. K. & Rossmeisl, J. The oxygen reduction reaction mechanism on Pt(111) from density functional theory calculations. Electrochim. Acta 55, 7975–7981 (2010). 25. Savinova, E. R. et al. Structure and dynamics of the interface between a Ag single crystal electrode and an aqueous electrolyte. Discuss. Faraday Soc. 121, 181–198 (2002). 26. Horswell, S. L. et al. A comparative study of hydroxide adsorption on the (111), (110), and (100) faces of silver with cyclic voltammetry, ex situ electron diffraction, and in situ second harmonic generation. Langmuir 20, 10970–10981 (2004). 27. Hansen, H. A., Rossmeisl, J. & Norskov, J. K. Surface Pourbaix diagrams and oxygen reduction activity of Pt, Ag and Ni(111) surfaces studied by DFT. Phys. Chem. Chem. Phys. 10, 3722–3730 (2008).

NATURE CHEMISTRY | VOL 6 | SEPTEMBER 2014 | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved

833

ARTICLES

NATURE CHEMISTRY

28. Nørskov, J. K. et al. Origin of the overpotential for oxygen reduction at a fuel-cell cathode. J. Phys. Chem. B 108, 17886–17892 (2004). 29. Xin, H. L. & Linic, S. Communications: exceptions to the d-band model of chemisorption on metal surfaces: the dominant role of repulsion between adsorbate states and metal d-states. J. Chem. Phys. 132, 221101 (2010). 30. Xin, H., Holewinski, A. & Linic, S. Predictive structure–reactivity models for rapid screening of Pt-based multimetallic electrocatalysts for the oxygen reduction reaction. ACS Catalysis 2, 12–16 (2012). 31. Zhang, Z. Y., Nenoff, T. M., Huang, J. Y., Berry, D. T. & Provencio, P. P. Room temperature synthesis of thermally immiscible Ag-Ni nanoalloys. J. Phys. Chem. C 113, 1155–1159 (2009). 32. Wang, D. & Li, Y. Bimetallic nanocrystals: liquid-phase synthesis and catalytic applications. Adv. Mater. 23, 1044–1060 (2011). 33. Gong, K., Su, D. & Adzic, R. R. Platinum-monolayer shell on AuNi0.5Fe nanoparticle core electrocatalyst with high activity and stability for the oxygen reduction reaction. J. Am. Chem. Soc. 132, 14364–14366 (2010). 34. Goodwin, A. L. et al. Colossal positive and negative thermal expansion in the framework material Ag3[Co(CN)6]. Science 319, 794–797 (2008). 35. Ludi, A., Gudel, H. U. & Dvorak, V. The structure of H3Co(CN)6 and Ag3Co (CN)6. Helv. Chim. Acta 7, 2035–2039 (1967). 36. Fernandez, J., Walsh, D. & Bard, A. Thermodynamic guidelines for the design of bimetallic catalysts for oxygen electroreduction and rapid screening by scanning electrochemical microscopy. M–Co (M: Pd, Ag, Au). J. Am. Chem. Soc. 357–365 (2005). 37. Lima, F. H. B., de Castro, J. F. R. & Ticianelli, E. A. Silver–cobalt bimetallic particles for oxygen reduction in alkaline media. J. Power Sources 161, 806–812 (2006). 38. Loglio, F. et al. Cobalt monolayer islands on Ag(111) for ORR catalysis. ChemSusChem 4, 1112–7 (2011). 39. Schmidt, T. J. et al. Characterization of high-surface area electrocatalysts using a rotating disk electrode configuration. J. Electrochem. Soc. 145, 2354–2358 (1998). 40. Kirowa-Eisner, E., Bonfil, Y., Tzur, D. & Gileadi, E. Thermodynamics and kinetics of upd of lead on polycrystalline silver and gold. J. Electroanal. Chem. 552, 171–183 (2003). 41. Herrero, E., Buller, L. J. & Abruna, H. D. Underpotential deposition at single crystal surfaces of Au, Pt, Ag and other materials. Chem. Rev. 101, 1897–1930 (2001). 42. Zhang, J., Sasaki, K., Sutter, E. & Adzic, R. R. Stabilization of platinum oxygenreduction electrocatalysts using gold clusters. Science 315, 220–222 (2007).

834

DOI: 10.1038/NCHEM.2032

43. Koenigsmann, C. et al. Enhanced electrocatalytic performance of processed, ultrathin, supported Pd–Pt core–shell nanowire catalysts for the oxygen reduction reaction. J. Am. Chem. Soc. 133, 9783–9795 (2011). 44. Wang, C. et al. Multimetallic Au/FePt3 nanoparticles as highly durable electrocatalyst. Nano Lett. 11, 919–926 (2011). 45. Bratsch, S. Standard electrode potentials and temperature coefficients in water at 298.15K. J. Phys. Chem. Ref. Data 18, 1–21 (1989). 46. Gerken, J. B. et al. Electrochemical water oxidation with cobalt-based electrocatalysts from pH 0–14: the thermodynamic basis for catalyst structure, stability, and activity. J. Am. Chem. Soc. 133, 14431–14442 (2011). 47. Krivanek, O. L. et al. An electron microscope for the aberration-corrected era. Ultramicroscopy 108, 179–195 (2008). 48. Greeley, J. et al. Alloys of platinum and early transition metals as oxygen reduction electrocatalysts. Nature Chem. 1, 552–556 (2009). 49. Viswanathan, V., Hansen, H. A., Rossmeisl, J. & Nørskov, J. K. Universality in oxygen reduction electrocatalysis on metal surfaces. ACS Catalysis 2, 1654–1660 (2012).

Acknowledgements We acknowledge support from the US DOE Office of Basic Energy Sciences, Division of Chemical Sciences (FG-02-05ER15686). We also acknowledge the University of Michigan Electron Microbeam Analysis Laboratory for use of the microscopy facilities. This research is also supported as part of a user project by Oak Ridge National Laboratory (ORNL)’s Center for Nanophase Materials Sciences, which is sponsored by the Scientific User Facilities Division, Office of Basic Energy Sciences, US DOE (J-C.I.). Finally, we acknowledge H. Xin and T. van Cleve for helpful discussions and experimental assistance.

Author contributions A.H. and S.L. devised and developed the project. A.H. carried out experimental work, theoretical calculations and data analysis. J-C.I. performed STEM and EELS imaging at ORNL. All the authors wrote the manuscript.

Additional information Supplementary information is available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to S.L.

Competing financial interests

The authors declare no competing financial interests.

NATURE CHEMISTRY | VOL 6 | SEPTEMBER 2014 | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved

ERRATUM

High-performance Ag–Co alloy catalysts for electrochemical oxygen reduction Adam Holewinski, Juan-Carlos Idrobo and Suljo Linic Nature Chemistry 6, 828–834 (2014); published online 11 August 2014; corrected after print 11 August 2014. Technical issues with our online publication processes resulted in this Article being published the day after that referred to in the print version. The official date of publication is 11 August 2014.

© 2014 Macmillan Publishers Limited. All rights reserved