Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts

Article

Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts Liang Cao, Zipeng Zhao, Zeyan Liu, ..., Xiaoqing Pan, Tim Mueller, Yu Huang [email protected] (T.M.) [email protected] (Y.H.)

HIGHLIGHTS Adding Cu to PtNi nanoparticles greatly improves catalytic activity and stability A novel combination of experimental growth tracking and simulations was used Simulations show particles with Cu benefited from larger initial surface Pt content The higher reduction potential of Cu likely led to a more Pt-rich surface structure

The introduction of Cu into PtNi nanoparticles has improved the catalytic activity and stability of the particles for the oxygen reduction reaction. To explain this improvement, we ran kinetic Monte Carlo (KMC) simulations initialized using layerby-layer compositions deduced from the experimental growth trajectory of the particles. The KMC runs revealed that enhanced initial surface Pt composition in the particles containing Cu suppressed the dissolution of subsurface Cu and Ni atoms, leading to improved stability and activity.

Cao et al., Matter 1, 1–14 December 4, 2019 ª 2019 Elsevier Inc. https://doi.org/10.1016/j.matt.2019.07.015

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

Article

Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts Liang Cao,1,8 Zipeng Zhao,2,8 Zeyan Liu,2,8 Wenpei Gao,3 Sheng Dai,3 Joonho Gha,2 Wang Xue,4 Hongtao Sun,4,7 Xiangfeng Duan,4,5 Xiaoqing Pan,3,6 Tim Mueller,1,* and Yu Huang2,5,9,*

SUMMARY

Progress and Potential

PtNi-based nanomaterials represent an emerging class of highly active catalysts for the oxygen reduction reaction (ORR) in fuel cells. However, they suffer from poor stability in operating conditions, which is the key challenge in maintaining their activity advantage over Pt in practice. We report significantly enhanced stability and activity of octahedral PtNi nanoparticles by tuning their surface elemental distribution through the introduction of a third element (Cu) during synthesis. To uncover the mechanism behind this observation, we performed kinetic Monte Carlo (KMC) simulations initialized using growth-tracking experiments and demonstrated that PtNiCu has improved Ni and Cu retention compared with PtNi, in agreement with experiments. The tracked movement of individual atoms in KMC reveals that the enhanced stability can be attributed to increased surface Pt composition in as-synthesized catalysts, which reduces the generation of surface vacancies and suppresses the surface migration and subsequent dissolution of subsurface Cu and Ni atoms.

One of the main challenges in rationally designing highperformance nanoscale catalysts is to understand the atomic-scale mechanisms that contribute to the properties and structural evolution of the catalyst. This is primarily due to the limitations in atomic-scale experimental characterizations and the lack of computational models to describe the realistic in situ evolution of atomic arrangements. We have used a novel combination of experimental characterization and atomic-scale simulations to explain the improvement in the stability and activity of PtNi nanoparticles for the oxygen reduction reaction when Cu is added. Our simulations reveal that the improved catalytic activity and stability are primarily due to enhanced initial surface Pt composition in the nanoparticles that contain Cu, which we attribute to the higher reduction potential of Cu relative to Ni. This study opens the door to new design strategies for highperformance nanoscale catalysts.

INTRODUCTION Fuel cells generate power by fuel oxidation at the anode and oxygen reduction at the cathode. Powered by renewable fuel with high energy-conversion efficiency, fuel cells hold the potential for replacing internal combustion engines for powering automotive vehicles in the future.1 Currently the broad adoption of fuel cells is mainly limited by their prohibitive cost. Reactions at both the anode and cathode, especially the oxygen reduction reaction (ORR) at the cathode, need catalysts to lower their electrochemical overpotential and increase power output.2 In proton exchange membrane (PEM) fuel cells, platinum (Pt)-based catalysts are widely used.2,3 To realize the widespread adoption of PEM fuel cells, several challenges need to be addressed, including increasing the activity and durability of the catalysts and reducing the amount of Pt required. With highly active and durable catalysts, the operation cost of fuel cells can be substantially reduced. Alloying Pt with transition metals is a widely adopted approach to address the performance challenge of Pt-based electrochemical catalysts. A variety of structures and compositions of Pt-based alloy catalysts has been studied.4–8 To date, the best ORR specific activity (SA; activity normalized by electrochemical surface area [ECSA]) is achieved on a Pt3Ni(111) single-crystal surface, which shows about 18 mA/cm2Pt and is about 90-fold more active than commercial Pt/C.9 Stimulated by this finding, intensive research has been focused on developing nanocatalysts that approach the specific activity established on the Pt3Ni(111) single-crystal

Matter 1, 1–14, December 4, 2019 ª 2019 Elsevier Inc.

1

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

surface, as nanoscale catalysts hold the advantage of high mass activity (activity normalized by Pt mass loading) due to their high ECSA.10–18 With exposed (111) facets, PtNi octahedral nanoparticles were able to reach drastically improved activity compared with commercial Pt/C catalysts, although challenges remained with regard to the poor stability.14,15 To further improve the activity as well as the stability of octahedral PtNi nanomaterials, introducing a third element to form a ternary alloy or surface doping modification has been explored.19–24 Reported Mo-Pt3Ni, PtRhNi, PtNiCu, Ga-PtNi, and PtNiCo octahedral nanoparticles showed enhanced stability and/or activity compared with binary octahedral PtNi catalysts.19–23,25,26 Studies to date have suggested possible mechanisms that the third element contributes to the enhanced stability, including lowering surface Pt diffusion/mobility,20,27 and stabilizing Ni by suppressing the Ni dissolution.26 Here, we report the synthesis of octahedral PtNiCu nanoparticles with wellcontrolled octahedral morphology and uniform dispersity on carbon support in solution (details about synthesis are included in Supplemental Information). To the best of our knowledge, this is the first time octahedral PtNiCu nanoparticles are prepared via the solution-phase synthesis route. These PtNiCu nanocatalysts demonstrated improved activity and stability compared with PtNi of similar size and similar Pt composition, as ORR catalysts. The kinetic growth pathway of the PtNi and PtNiCu nanoparticles was recorded, and their respective size and composition at several time points were carefully analyzed. It was clear from the experiments that adding Cu changes the growth behavior of these nano-octahedral particles. In addition, a strong correlation between the initial differential Pt/Ni/Cu element distribution and the resultant varied stability was observed by comparing PtNi and PtNiCu in ORR performance. To understand the reason for the enhanced properties of PtNiCu nanoparticles, we have used atomic-scale computational modeling. As the cores of the nanoparticles are likely kinetically trapped in their as-synthesized structures,25 it is necessary to use kinetic modeling to determine the atomic-scale structures of the particles after electrochemical cyclic voltammetry (CV) activation. Here we go beyond previous efforts to model the thermodynamic equilibrium21,28,29 or kinetic evolution25,30,31 of alloy nanoparticles by using time-tracking experiments to determine the initial composition profile of the particles and using a cluster-expansion32,33 energy model trained on density functional theory (DFT)34 calculations in kinetic Monte Carlo (KMC)35,36 simulations to determine the structures of the nanoparticles after activation. Evaluation of the rationally initialized KMC simulation for PtNi and PtNiCu particles reveals the origins of the highly enhanced durability of the PtNiCu particles at the atomic level: the higher Pt surface fraction reduces the number of surface vacancies created in the early stage of CV activation, which in turn reduces the opportunity for atoms in subsurface layers to move to the surface and dissolve.

1Department

of Materials Science and Engineering, Johns Hopkins University, Baltimore, MD 21218, USA

2Department

of Materials Science and Engineering, University of California, Los Angeles, CA 90095, USA

3Department

of Chemical Engineering and Materials Science, University of California, Irvine, CA 92697, USA

4Department

of Chemistry and Biochemistry, University of California, Los Angeles, CA 90095, USA

5California

NanoSystems Institute, University of California, Los Angeles, CA 90095, USA

RESULTS AND DISCUSSION The synthesized PtNi and PtNiCu nanoparticles showed well-controlled octahedral morphology and uniform distribution on carbon black support, as shown in transmission electron microscopy (TEM) images (Figure S1). The edge lengths of the resultant octahedral PtNi and PtNiCu are comparable (5.0 G 0.6 nm for PtNi and 4.9 G 0.6 nm for PtNiCu, as shown in Figure S1). Powder X-ray diffraction (XRD) spectra show that the atomic packing for these PtNi and PtNiCu alloys is face center cubic (fcc) packing with a lattice parameter of 0.379 nm (Figure 1A), indicating the same Pt ratio in these two alloys. Atomic-resolution scanning TEM (STEM) images revealed the (111)

2

Matter 1, 1–14, December 4, 2019

6Department

of Physics and Astronomy, University of California, Irvine, CA 92697, USA

7Present

Address: Department of Industrial and Manufacturing Engineering; Materials Research Institute; Pennsylvania State University, University Park, PA 16802, USA

8These 9Lead

authors contributed equally

Contact

*Correspondence: [email protected] (T.M.), [email protected] (Y.H.) https://doi.org/10.1016/j.matt.2019.07.015

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

Figure 1. Spectroscopic Characterization and Electrochemical Performance of Octahedral PtNi and PtNiCu Nanoparticles (A–F) XRD spectra (A) of octahedral PtNi/C, and PtNiCu/C (vertical lines represent standard XRD peak positions: black for Pt [PDF #04-0802], green for Ni [PDF #04-0850], and blue for Cu [PDF #04-0836]). Atomic-resolution STEM images of octahedral (B) PtNi and (C) PtNiCu nanoparticles. Comparison of Pt/C, octahedral PtNi/C, and PtNiCu/C: (D) CV curves measured in N 2 -saturated 0.1 M HClO4 with scan rate 100 mV/s, (E) ORR polarization curves measured in O 2 -saturated 0.1 M HClO 4 with scan rate 20 mV/s, and (F) SA and MA along with the error bars representing standard deviations. (G–I) Comparison of MA retention between PtNi/C and PtNiCu/C (G). EDS composition analysis of the atomic fraction of Ni and Cu at initial stage, after CV activation, and after ADT for (H) PtNi/C and (I) PtNiCu/C. Octahedral PtNi/C and PtNiCu/C are noted in figures as PtNi and PtNiCu due to limited space.

interplanar distances of the alloy PtNi and PtNiCu nanoparticles were 0.219 nm (Figures 1B and 1C). The lattice parameters based on STEM match well with those based on XRD spectra. These observations confirm that PtNi and PtNiCu were prepared with comparable morphology, size, Pt ratio, and lattice, leaving the only difference

Matter 1, 1–14, December 4, 2019

3

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

between octahedral PtNi and PtNiCu to the Cu and Ni content. The electrochemical performance of the octahedral PtNiCu was studied in comparison with PtNi. CV curves were recorded for commercial Pt/C, and octahedral PtNi and PtNiCu (Figure 1D), in N2-saturated 0.1 M HClO4 after CV activation (30 CV cycles for PtNi/C and PtNiCu/C, and 60 CV cycles for Pt/C). Electrochemical surface area (ECSA) was determined by integrating hydrogen underpotential deposition (Hupd) charge (Hupd charge to surface area conversion constant: 210 mC/cm2, Table S2). CO stripping was also employed to evaluate the ECSA (Figure S2; Tables S1 and S2). ORR polarization curves were recorded in O2-saturated 0.1 M HClO4 (Figure 1E). The comparison of these tests showed that the mass activity (MA), and SA of octahedral PtNiCu/C were higher than those of octahedral PtNi (Figure 1F). The octahedral PtNiCu showed 15.9-fold SA and 13.2-fold MA compared with commercial Pt/C. Several results from recent studies are also included in Table S3 for comparison, confirming the high performance of the prepared PtNiCu/C. To study the stability of catalysts, we tested these octahedral nanoparticles in O2-saturated 0.1 M HClO4 for 30,000 CV cycles for an accelerated durability test (ADT). Octahedral PtNiCu showed 69.3% MA retention, which was significantly enhanced compared with 49.4% MA retention of octahedral PtNi (Figures 1G and S3; Table S3). The stability of PtNiCu/C exceeds the US Department of the Environment37 target of fuel cell catalyst stability (less than 40% MA loss after 30,000 CV cycles). The morphological change of nanocatalysts after ADT was investigated by TEM (Figure S4). Significant aggregation was observed in commercial Pt/C (Figure S4), attributing the low MA retention to the loss of the ECSA due to aggregation. For octahedral PtNi and PtNiCu, the activity loss can be explained by the loss of the octahedral morphology to some extent and the leaching of (Ni and Cu) (Figure S4 and Table S4). Interestingly, significant composition change was observed for octahedral PtNi and PtNiCu after CV activation and ADT based on energy-dispersive spectroscopy (EDS) analysis. After CV activation, the atomic Ni ratio in octahedral PtNi decreased from 34.5% to 15.2% (Figure 1H and Table S4). In octahedral PtNiCu, Ni was reduced from 16.8% to 12.7%, Cu was reduced from 16.9% to 14.6%, and the total (Ni + Cu) was reduced from 33.7% to 27.3% (Figure 1I and Table S4). It was found that relatively more Cu was retained within the octahedral PtNiCu compared with Ni, consistent with the fact that Cu is more inert than Ni based on their reduction potentials38 (Tables S5 and S6). The overall fraction of Cu and Ni after activation for PtNiCu is 27.3%, which was significantly higher than 15.2% for PtNi. As the electrochemical cycling continues, after ADT (30,000 CV cycles) octahedral PtNiCu still maintained 11.9% Cu and 5.5% Ni (total 17.4% Cu and Ni) while octahedral PtNi only maintained 6.3% Ni (a stability comparison among several representative studies is provided in Table 1). These experimental observations suggest that Ni dissolution can be significantly reduced by the presence of Cu. It was also clear that with the presence of Cu, the electrochemical performance of octahedral PtNiCu/C was significantly improved compared with octahedral PtNi/C. To understand the fundamental mechanism of these improvements, we carried out theoretical simulations on octahedral PtNi and PtNiCu nanoparticles with the experimentally observed Pt/Ni/Cu compositions and particle size (with edge lengths of approximating 4.8 nm). As the elemental distribution in a Pt-alloy nanocatalyst may differ significantly from a typical particle in thermodynamic equilibrium,25 we have used ab initio kinetic models initialized according to experimental growth-tracking analysis to closely determine the distributions of the elements within the PtNi and PtNiCu nanoparticles.

4

Matter 1, 1–14, December 4, 2019

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

Table 1. Stability Test Condition Comparison of Our Octahedral PtNiCu/C and Several Pt-Alloy Octahedral Nanoparticle Catalysts in the Literature Reference

Octahedral Catalyst

SA (mA/cm2Pt) Based on Hupd

CO

Choi et al.14

Pt2.5Ni

15.7a

10.6a

Cui et al.

15

b

MA (mA/mgPt)

3.3

5,000 b

PtNi

3.8

NA

1.7

This work

PtNi

5.2 G 0.3

4.6 G 0.3

2.6 G 0.2

Huang et al.16

Pt71Ni25Co4

3.88

NA

2.33

22

Ara´n-Ais et al.22

Ara´n-Ais et al.

Zhao et al.

19

Zhang et al.23 Beermann et al.

20

ADT Cycles

b

ADT Cycle Potential Range versus RHE (V)

Loss of SA (%)

Loss of MA (%)

0.65–1.0 (square wave)

NA

40

b

66

4,000

0.6–1.0

66

15,000

0.6–1.0

47.7

40.2

30,000

0.6–1.0

51.7

50.6

6,000

0.6–1.1

NA

57

4,000

0.5–1.0

51

51

4,000

0.5–1.0

34

47 51

Pt48Ni27Co25

b

3.5

NA

0.45

Pt30Ni51Co19

1.5b

NA

0.5b

PtNi0.55Co0.1

6.01

5.05

2.8

6,000

0.6–1.1

58

Pt2CuNi

6.65

NA

2.35

10,000

0.6–1.0

28.1 b

31.9

PtNiRh

2.75

NA

0.82

30,000

0.6–1.0

56

Lim et al.26

PtNiGa

2.53

2.34a

1.24

30,000

0.6–1.1

61.6a

61 65.5

This work

PtNiCu

6.2 G 0.4

5.7 G 0.4

3.7 G 0.2

15,000

0.6–1.0

26.8

18.4

30,000

0.6–1.0

37.6

30.7

NA, not available; RHE, reversible hydrogen electrode. a Calculated based on available electrochemical surface area. b Extracted from plots in literature.

The growth tracking of the nanoparticle from 6 to 60 h was used as the basis to calculate the initial layer-by-layer compositions of PtNi and PtNiCu nanoparticles for KMC simulations. With the aid of EDS as well as TEM, growth-tracking experiments unveiled the composition evolution accompanied by octahedral size growth. For PtNi nanoparticles, the atomic percent of Ni continuously increased from 28.7% to 34.5% (all compositions are atomic percent if without a specific note), while the atomic percent of Pt decreased from 71.3% to 65.5% from 6 h to 60 h (Figure 2A). In contrast to PtNi, PtNiCu nanoparticles comprised 64.1% Pt, 31.2% Cu, and only 4.7% Ni, after 6 h of reaction. At 12 h, the Cu fraction decreased to 21.4% while the Ni fraction increased to 16.1%. After 24 h, Cu and Ni were nearly equal (Figure 2B). The growth tracking of the composition indicated that Cu played a vital role at the early stage of the nanoparticle nucleation and growth. The reduction rate of Cu was much faster than that of Ni during the reaction, resulting in a much higher atomic fraction of Cu (31.2%) than that of Ni (4.7%) at 6 h. Both octahedral PtNi (4.0 G 0.6 to 5.0 G 0.6 nm) and PtNiCu (3.2 G 0.5 to 4.9 G 0.6 nm) showed continuous nanoparticle size growth from 6 to 60 h (Figures 2A and 2B), with the difference that PtNiCu nanoparticle showed a smaller size at early growth stage compared with PtNi (Figure S5), possibly due to faster nucleation rates in the presence of Cu. Moreover, for PtNi the growth-tracking experiment showed that the overall Ni fraction continued to increase during the growth process, while in PtNiCu the overall (Ni + Cu) fraction slightly decreased from 12 h to 60 h. This observation indicates that PtNi nanoparticles may show a lower Pt ratio on the surface than PtNiCu, given that both have similar overall Pt ratios at 60 h. To account for this significant difference of the surface elemental distributions at different reaction times during the synthesis, we calculated the layer-by-layer composition profiles from the core to the surface of particles following the timeline of the growth process (Figures 2A and 2B; details are provided in Supplemental Information). We found that the Pt fraction in the PtNiCu particle is about 14.3% higher in the first layer, and 8.5% higher in the second layer, than in PtNi (Table S7). To

Matter 1, 1–14, December 4, 2019

5

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

6

Matter 1, 1–14, December 4, 2019

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

Figure 2. Growth-Tracking Study and Simulation Model Construction Growth tracking of atomic ratio based on EDS and octahedral edge length based on TEM image for (A) PtNi and (B) PtNiCu particles. Error bars represent standard deviations. Layer-by-layer nano-octahedral model construction based on growth tracking for (C) PtNi and (D) PtNiCu. Schemes for the growth of randomly initialized (E) PtNi and (F) PtNiCu nanoparticles based on (C) and (D) to mimic the increasing octahedral size and various Pt/Ni/Cu compositions during growth. EDS map of octahedral nanoparticles of two representative PtNi (G and I) and two representative PtNiCu (H and J) nanoparticles. The left side of each panel shows the map of all elements overlapped while the right side of the panel exhibits the map of a single element or Ni + Cu overlapped. The inset panel shows the STEM image of the mapped nanoparticle.

further validate our approach, we generated EDS maps for PtNi and PtNiCu nanooctahedra. The obtained EDS maps show observable amounts of Ni over Pt on the surface for both PtNi and PtNiCu octahedra (Figures 2G–2J). The EDS maps further reveal that the distribution of Cu is more concentrated in the core of the nanoparticles than Ni is (Figures 2H and 2J), which is consistent with our composition results obtained via growth tracking and the layer-by-layer composition profiles we constructed (Figures 2B, 2D, and 2F). In contrast, Metropolis Monte Carlo39 simulations predict that in thermodynamic equilibrium almost all Cu atoms (99.6%) in the PtNiCu nanoparticle segregate to the second layer, the first and third layers are nearly pure (>96%) Pt, and most of Ni atoms are in the core below the third layer (Figure S6 and Table S8). These results suggest that due to the relatively low temperature (150 C–170 C) of the particle growth process, the particles are not in equilibrium and the cores are trapped in a metastable state (Figure 2), similar to our observations on other PtNi-based nanoparticles.25 Thus, for consistency with experimental characterization, we initialized the nanoparticle structures for our KMC simulations by creating nearly octahedral nanoparticles (with six vertex atoms removed) with 21 (111) layers, and randomly distributing Ni and Cu atoms in each layer based on the derived layer-by-layer composition profiles (Figures 2E and 2F). To determine the structure of particles after activation, we performed KMC simulations of particle evolution in oxidizing conditions (0.95 V versus reversible hydrogen electrode [RHE]) based on a cluster expansion trained on DFT calculations. The KMC-derived composition values discussed in this work are averages over ten independently initialized KMC runs unless otherwise noted. Additional details of our computational methods are provided in Supplemental Experimental Procedures and Supplemental Information. The KMC simulations show a rapid loss of Ni and Cu from the near-surface sites in the early stage, followed by a steadier concentration profile (Figure 3). We stopped the KMC runs when Ni and Cu compositions reached plateaus (insets of Figure 3). Snapshots of the resultant particles after Ni and Cu dissolution are shown in Figures 4B and 4E, respectively. The octahedral shape of both PtNi and PtNiCu particles still holds after KMC runs, in agreement with the experiments (Figures S1M–S1P). After the KMC runs, the Ni (Cu) fraction in the PtNiCu particle (Figure 4F and Table S10) drops from 16.8% (16.9%) to 12.2% (14.7%), which is close to the experimental values of 12.7% (14.6%) for activated particles. In the PtNi particle (Figure 4C), the Ni fraction drops from 34.5% to 21.4%, compared with 15.2% Ni in experiments. In both simulations and experiments, there was considerably less Ni and (Ni + Cu) loss in the PtNiCu particle than in the PtNi particle. In both particles, the Ni and Cu atoms in the outermost layer dissolve from the particles early in the KMC simulation (Figure 3), consistent with the experimental observation of no Ni and Cu signals on the near-surface of particles after activation according to the EDS line scans (Figures S1Q and S1R). For the PtNi particle, about 64.4% of Ni in the second layer is lost to dissolution, while for PtNiCu much less (Ni + Cu) content (only about 14.2%) is lost. This leads to a larger fraction of 3d transition metals in the second layer of the PtNiCu nanoparticle (33.7%) compared with

Matter 1, 1–14, December 4, 2019

7

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

Figure 3. Ni and Cu Compositions of PtNi and PtNiCu during KMC Runs The large graphs show Ni and Cu compositions of (A) PtNi and (B) PtNiCu nanoparticles as a function of KMC time for representative KMC runs. Snapshots of the nanoparticle structures are shown on the bottom row. The insets are the Ni and Cu compositions over the course of the entire KMC simulation.

the PtNi nanoparticle (25.6%, Table S11). Both computational (Figure 3 and Table S11) and experimental results (Figures S1Q and S1R) indicate that the amount of near-surface dissolution is more significant in PtNi nanoparticles than in PtNiCu nanoparticles, suggesting that introduced Cu suppresses the dissolution of subsurface Ni and Cu. The decreased platinum content in the second layer likely

8

Matter 1, 1–14, December 4, 2019

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

Figure 4. Composition Evolution of PtNi and PtNiCu after KMC Simulations/CV Activation Cross-sections of (A and B) a representative PtNi particle and (D and E) a representative PtNiCu nanoparticle before (A and D) and after (B and E) KMC simulations at a temperature of 25  C (298 K). The insets at bottom right show snapshots of the particle surfaces. The snapshots were taken when Ni composition reached plateaus (21.4% Ni for the PtNi particle, and 12.2% Ni and 14.7% Cu for the PtNiCu particle), close to the experimental compositions after CV activation. The analysis of the average overall compositions of PtNi (C) and PtNiCu (F) particles after KMC simulations, compared with experimental data after CV activation (EDS in Table S4). Error bars represent standard deviations.

contributes to the enhanced activity of the PtNiCu nanoparticles, as previous work has indicated that PtNi nanoparticles bind oxygen too strongly on average25,40 and the oxygen-binding energy can be reduced by decreasing the Pt content in the second layer.41 The strain effect also likely contributes to the observed differences in catalytic activity, as the greater fraction of relatively small Ni and Cu atoms in the activated PtNiCu particles will act to reduce surface Pt-Pt bond length, further contributing to relatively weaker oxygen-binding energy and higher ORR activity.25,41–43 These conclusions are supported by DFT calculations of oxygen adsorption energies on PtNiCu(111) and PtNi(111) slabs built to mimic the layer-by-layer compositions and lattice parameters of particles after activation (see Figure S8 and Tables S12–S15). To better understand the reason for the enhanced stability of PtNiCu nanoparticles, we tracked the movement of individual atoms in KMC simulations (Figure 5; a breakdown by each element is provided in Figure S10). It was found that in both PtNi and PtNiCu, there is almost no diffusion of atoms that were deeper than the fourth layer, consistent with a kinetically trapped structure. As our KMC simulations likely overestimate the relative rates at which highly coordinated atoms hop (Section 2.4 in Supplemental Information), it is likely that there is even less interlayer migration for atoms in subsurface layers than our simulations show. Overall, more atoms are retained in their initial layers in the PtNiCu nanoparticle (e.g., 54.9% in the first layer) than in the PtNi nanoparticle (e.g., 33.1% in the

Matter 1, 1–14, December 4, 2019

9

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

Figure 5. The Tracking of the Movement of Individual Atoms during KMC Simulations for PtNi and PtNiCu Atomic tracking results for (A–D) PtNi and (E–H) PtNiCu nanoparticles. In (A) and (E), the x axis gives the initial layers before KMC, the y axis gives the final locations after KMC for each atom, and ‘‘sol’’ indicates the solution. The color scheme indicates average percentages of atoms in the initial layer moved to the final layer (from red indicating 100% to navy indicating 0%). It is clear that a higher percentage of atoms stayed in the initial layers (i.e., same initial and final layers, along the diagonal of the table) in PtNiCu than in PtNi. (B–D) and (F–H) show the initial layers of atoms that ended up in the first (B and F), second (C and G), and third (D and H) layers of representative snapshots after KMC runs. The black spheres mean that the atoms were created through growth events. See also Figures S9–S11.

first layer), suggesting decreased diffusivity. It is also observed that more atoms initially in subsurface layers are exposed to the surface in PtNi nanoparticles than in PtNiCu nanoparticles, which can be attributed to more surface vacancies created in PtNi than in PtNiCu (Figure S12A), because in part of the fewer vacancies created in PtNiCu, there are 50% fewer Pt hops per unit time in the PtNiCu nanoparticle than in the PtNi nanoparticle (Figure S13). When we run KMC simulations on PtNi and PtNiCu nanoparticles initialized with the same completely random distribution of Pt and Ni/Cu atoms, the observed differences in the composition and migration of atoms largely disappear (Figures S14– S16; for more details see Supplemental Information). These results indicate that although the presence of Cu somewhat reduces the tendency of Ni to dissolve, the enhanced stability is primarily due to the different initial elemental distributions in the PtNi and PtNiCu nanoparticles; i.e., the PtNi nanoparticles start with a significantly lower surface fraction of Pt than PtNiCu nanoparticles (shown in the experimental EDS maps in Figures 2G–2J, and the deduced layer-by-layer composition profiles in Figures 2C and 2D; Table S7). As a result, there are more vacancies created on the surface by dissolution in the early stage of KMC runs/CV activation in PtNi, resulting in a substantial initial loss of Ni and allowing more opportunity for the atoms in subsurface layers to dissolve. It has previously been observed that the stability of PtNi nanoparticles can be enhanced by doping them with Rh20 or Mo,21,25 where the dopants are primarily located on the

10

Matter 1, 1–14, December 4, 2019

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

particle surface. In these studies, the surface dopant is believed to displace surface Ni and/or reduce the mobility of surface Pt atoms, inhibiting shape loss and reducing exposure of subsurface Ni to the surface. In PtNiCu nanoparticles, our simulations indicate that the decrease in Ni dissolution is a primary result of the decreased surface vacancy formation rather than the preferential segregation of the dopant to the surface, but the end result is similar. Decreased exposure of subsurface Ni to the electrolyte results in less Ni dissolution, consistent with experimental results. Conclusion We reported that octahedral PtNiCu nanoparticles synthesized with a solution-phase method showed significantly enhanced stability and activity compared with octahedral PtNi. It was found that the introduction of the Cu precursor affected the kinetic growth of the nanoparticles, wherein the Cu ratio is higher than that of Ni at the nucleation stage, followed by the deposition of more Ni. The increased concentration of Cu in early stages of particle growth, relative to Ni, is likely due to the higher reduction potential of the Cu(Ac)2 precursor relative to Ni(Ac)2, providing a design guideline for future efforts. Such PtNiCu nanoparticles were found to retain more Cu and Ni during electrochemical cycling compared with PtNi, leading to improved activity and stability. KMC simulations were rationally initialized to reflect a realistic elemental distribution within the nanoparticles, resulting in good agreement between simulations and experimental characterizations of the compositions and structures of the PtNiCu and PtNi nanoparticles after activation. The novel integration between simulation and experiment revealed that the enhanced stability and activity in PtNiCu likely resulted from an increase in the fraction of Pt atoms on the surface of the preactivated PtNiCu nanoparticles compared with PtNi, which reduces surface dissolution of the 3d transition metals (Ni and Cu in this work) and the creation of surface vacancies, thus limiting the dissolution of atoms from subsurface layers.

EXPERIMENTAL PROCEDURES Synthesis of Octahedral PtNiCu/C The typical octahedral PtNiCu/C was synthesized by utilizing 9 mg of platinum(II) acetylacetonate (Pt(acac)2), 4.5 of mg nickel(II) acetate tetrahydrate (Ni(Ac)2,4H2O), and 1.5 mg of copper(II) acetate monohydrate (Cu(Ac)2,H2O) as metal precursors in a 25-mL vial. Benzoic acid (65 mg) was used for morphology control and 10 mL of N,N-dimethylformamide (DMF) was used for solvent and a reducing agent, as demonstrated by our previous studies.16 The vial was then heated in a 140 C oil bath and slowly heated to 160 C. Two-step synthesis was designed for controlling octahedral size. In the second step, 1 mg of Pt(acac)2, 0.5 mg Ni(Ac)2$4H2O, and 0.5 mg Cu(Ac)2$H2O were dissolved in 0.5 mL of DMF and added into the vial after a 12-h reaction. The reaction temperature then increased to 170 C and was kept at that temperature for 48 h. After the reaction finished the catalysts were collected by centrifugation, then dispersed and washed with isopropanol and acetone mixture. Synthesis of Octahedral PtNi/C Vulcan XC-72 carbon black (20 mg) was dispersed in 9 mL of DMF under ultrasonication for 30 min in a 25-mL vial. Thereafter, 9 mg of Pt(acac)2, 7.2 mg of nickel(II) acetylacetonate (Ni(acac)2), and 85 mg of benzoic acid were dissolved in 1 mL of DMF and were also added into the 25-mL vial with carbon black dispersion. After ultrasonication for 5 min, the vial with the well-mixed solution was directly put into a 140 C oil bath and then slowly heated to 150 C. The vial was kept at 150 C for 12 h. Next, 1 mg of Pt(acac)2 and 0.8 mg of Ni(acac)2 were dissolved in 0.5 mL of DMF and added into the vial. The vial was then kept in a 150 C oil bath for another 48 h. After the reaction finished, the catalysts were collected by centrifugation, then

Matter 1, 1–14, December 4, 2019

11

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

dispersed and washed with isopropanol and acetone mixture. The catalysts were then dried in a vacuum at room temperature, making them ready for characterization and electrochemistry tests. Electrochemistry Test of Prepared Catalysts A three-electrode cell was used to carry out the electrochemical measurements. The working electrode was a catalyst-coated glassy carbon electrode. An Ag/AgCl electrode was used as the reference electrode. Pt wire was used as the counter electrode. CV measurements were conducted in an N2-saturated 0.1 M HClO4 solution between 0.05 and 1.1 V versus RHE (for RHE, the same potential scale is used in the following discussion unless otherwise specified) at a sweep rate of 100 mV/s. ORR measurements were conducted in an O2-saturated 0.1 M HClO4 solution between 0.05 and 1.1 V versus RHE at a sweep rate of 20 mV/s. ADT was performed in O2-saturated 0.1 M HClO4 solution by applying CV sweeps between 0.6 and 1.0 V versus RHE at a sweep rate of 100 mV/s. For the CO-stripping voltammetry measurements, working electrodes coated with different catalysts were firstly immersed in a CO-saturated 0.1 M HClO4 solution for up to 1.5 min, then the CO-stripping voltammetry was recorded respectively in N2-saturated 0.1 M HClO4 between 0.05 and 1.1 V versus RHE at a sweep rate of 25 mV/s. Computational Methods We have developed a quaternary Pt-Ni-Cu-Vacancy cluster expansion,32,33 in the same way as we previously generated a Pt-Ni-Mo-Vacancy cluster expansion.21,25,29,44 All the training structures for the cluster expansion were generated using DFT calculations34 as implemented in the Vienna Ab-initio Software Package (VASP)45 with the revised Perdew-Burke-Eznerhof (RPBE) exchange-correlation functional.46–48 All DFT calculations were run with spin polarization enabled. The Cu_pv_GW, Ni, Pt_pv_GW, O_GW, and H_GW Perdew-Burke-Eznerhof (PBE) projector-augmented wave (PAW)49 potentials provided with VASP were used, and all VASP calculations were run with high precision. A single k-point at the center of the Brillouin zone was used for each nanoparticle. For bulk materials and slabs, the Brillouin zone was sampled using grids generated by the k-point grid server50 with a minimum distance of 46.5 A˚ between real-space lattice points. Second-order Methfessel-Paxton smearing51 with a width of 0.2 eV was used to set partial occupancies. Real-space projectors were used to evaluate the non-local part of the PAW potential. The convergence criteria for the electronic self-consistent iteration and the ionic relaxation loop were set to 0.1 meV and 1 meV per unit cell, respectively. We used the cluster expansion in Metropolis Monte Carlo39 simulations at 170 C to predict the thermodynamic equilibrium of PtNi particles, and in KMC simulations at 25 C, using a standard rejection KMC algorithm,35,36 to simulate Ni and Cu dissolution and investigate how the introduction of Cu atoms stabilizes the surface and subsurface Ni and Cu atoms. More details of the cluster expansion and Monte Carlo simulations (Figures S6 and S7; Tables S8 and S9) are provided in Supplemental Information.

DATA AND CODE AVAILABILITY Data supporting the findings of this manuscript are available from the corresponding authors upon reasonable request. A reporting summary for this article is available as a Supplemental Information file.

SUPPLEMENTAL INFORMATION Supplemental Information can be found online at https://doi.org/10.1016/j.matt. 2019.07.015.

12

Matter 1, 1–14, December 4, 2019

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

ACKNOWLEDGMENTS Y.H., X.D., Z.Z., and Z.L. acknowledge support from Office of Naval Research by the grant number N000141812155. T.M. and L.C. acknowledge the support from National Science Foundation (NSF) through award no. DMR-1409765. X.D. acknowledges the support from the US Department of Energy, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering through award DESC0008055. T.M. and L.C. acknowledge the computational resources provided by XSEDE through NSF award DMR-140068 and by the Maryland Advanced Research Computing Center (MARCC). Calculations were run on the Stampede2 supercomputer at the Texas Advanced Computing Center and the BlueCrab supercomputer at MARCC. Atomic-scale structural images were generated using VESTA.52 The work at UC Irvine is supported by the National Science Foundation with the grant numbers CBET 1159240, DMR-1420620, and DMR- 1506535. TEM work on JEM Grand ARM was conducted using the facilities in the Irvine Materials Research Institute at the University of California, Irvine. We also thank the Electron Imaging Center of Nanomachines at CNSI for the TEM support.

AUTHOR CONTRIBUTIONS Z.Z., L.C., and Z.L. contributed equally to this work. Z.Z. conceived the idea, designed and performed the experiments, and wrote the manuscript; L.C. designed and performed the simulations and wrote the manuscript; Z.L. contributed to the development of catalysts and electrochemical tests; W.G. and S.D. performed STEM and EDS line-scan studies, supervised by X.P.; J.G., W.X., and H.S. contributed to the preparation of catalysts and electrochemical tests; T.M. designed the simulations and wrote the manuscript; Y.H. conceived the idea, designed the study, oversaw the project, and wrote the manuscript.

DECLARATION OF INTERESTS T.M. discloses unrelated research funding and support from Toyota Motor Corporation. The other authors declare no competing interests. Received: May 5, 2019 Revised: July 1, 2019 Accepted: July 22, 2019 Published: October 9, 2019

REFERENCES 1. Cano, Z.P., Banham, D., Ye, S., Hintennach, A., Lu, J., Fowler, M., and Chen, Z. (2018). Batteries and fuel cells for emerging electric vehicle markets. Nat. Energy 3, 279–289. 2. Debe, M.K. (2012). Electrocatalyst approaches and challenges for automotive fuel cells. Nature 486, 43–51. 3. Holton, O.T., and Stevenson, J.W. (2013). The role of platinum in proton exchange membrane fuel cells. Platinum Met. Rev. 57, 259–271.

multimetallic nanoframes with threedimensional electrocatalytic surfaces. Science 343, 1339–1343. 6. Zhang, L., Roling, L.T., Wang, X., Vara, M., Chi, M., Liu, J., Choi, S.I., Park, J., Herron, J.A., Xie, Z., et al. (2015). Platinum-based nanocages with subnanometer-thick walls and welldefined, controllable facets. Science 349, 412–416.

4. Lim, B., Jiang, M., Camargo, P.H., Cho, E.C., Tao, J., Lu, X., Zhu, Y., and Xia, Y. (2009). Pd-Pt bimetallic nanodendrites with high activity for oxygen reduction. Science 324, 1302–1305.

7. Li, M., Zhao, Z., Cheng, T., Fortunelli, A., Chen, C.Y., Yu, R., Zhang, Q., Gu, L., Merinov, B.V., Lin, Z., et al. (2016). Ultrafine jagged platinum nanowires enable ultrahigh mass activity for the oxygen reduction reaction. Science 354, 1414– 1419.

5. Chen, C., Kang, Y., Huo, Z., Zhu, Z., Huang, W., Xin, H.L., Snyder, J.D., Li, D., Herron, J.A., Mavrikakis, M., et al. (2014). Highly crystalline

8. Bu, L., Zhang, N., Guo, S., Zhang, X., Li, J., Yao, J., Wu, T., Lu, G., Ma, J.Y., Su, D., and Huang, X. (2016). Biaxially strained PtPb/Pt core/shell

nanoplate boosts oxygen reduction catalysis. Science 354, 1410–1414. 9. Stamenkovic, V.R., Fowler, B., Mun, B.S., Wang, G., Ross, P.N., Lucas, C.A., and Markovic, N.M. (2007). Improved oxygen reduction activity on Pt3Ni(111) via increased surface site availability. Science 315, 493–497. 10. Wu, J., Zhang, J., Peng, Z., Yang, S., Wagner, F.T., and Yang, H. (2010). Truncated octahedral Pt3Ni oxygen reduction reaction electrocatalysts. J. Am. Chem. Soc. 132, 4984– 4985. 11. Zhang, J., Yang, H., Fang, J., and Zou, S. (2010). Synthesis and oxygen reduction activity of shape-controlled Pt3Ni nanopolyhedra. Nano Lett. 10, 638–644. 12. Carpenter, M.K., Moylan, T.E., Kukreja, R.S., Atwan, M.H., and Tessema, M.M. (2012).

Matter 1, 1–14, December 4, 2019

13

Please cite this article in press as: Cao et al., Differential Surface Elemental Distribution Leads to Significantly Enhanced Stability of PtNi-Based ORR Catalysts, Matter (2019), https://doi.org/10.1016/j.matt.2019.07.015

Solvothermal synthesis of platinum alloy nanoparticles for oxygen reduction electrocatalysis. J. Am. Chem. Soc. 134, 8535– 8542. 13. Cui, C., Gan, L., Li, H.H., Yu, S.H., Heggen, M., and Strasser, P. (2012). Octahedral PtNi nanoparticle catalysts: exceptional oxygen reduction activity by tuning the alloy particle surface composition. Nano Lett. 12, 5885–5889. 14. Choi, S.-I., Xie, S., Shao, M., Odell, J.H., Lu, N., Peng, H.C., Protsailo, L., Guerrero, S., Park, J., Xia, X., et al. (2013). Synthesis and characterization of 9 nm Pt-Ni octahedra with a record high activity of 3.3 A/mgPt for the oxygen reduction reaction. Nano Lett. 13, 3420–3425. 15. Cui, C., Gan, L., Heggen, M., Rudi, S., and Strasser, P. (2013). Compositional segregation in shaped Pt alloy nanoparticles and their structural behaviour during electrocatalysis. Nat. Mater. 12, 765–771. 16. Huang, X., Zhao, Z., Chen, Y., Zhu, E., Li, M., Duan, X., and Huang, Y. (2014). A rational design of carbon-supported dispersive Ptbased octahedra as efficient oxygen reduction reaction catalysts. Energy Environ. Sci. 7, 2957– 2962. 17. Niu, G., Zhou, M., Yang, X., Park, J., Lu, N., Wang, J., Kim, M.J., Wang, L., and Xia, Y. (2016). Synthesis of Pt-Ni octahedra in continuous-flow droplet reactors for the scalable production of highly active catalysts toward oxygen reduction. Nano Lett. 16, 3850– 3857. 18. Zhang, C., Hwang, S.Y., Trout, A., and Peng, Z. (2014). Solid-state chemistry-enabled scalable production of octahedral Pt-Ni alloy electrocatalyst for oxygen reduction reaction. J. Am. Chem. Soc. 136, 7805–7808. 19. Zhao, Z., Feng, M., Zhou, J., Liu, Z., Li, M., Fan, Z., Tsen, O., Miao, J., Duan, X., and Huang, Y. (2016). Composition tunable ternary Pt-Ni-Co octahedra for optimized oxygen reduction activity. Chem. Commun. (Camb.) 52, 11215– 11218. 20. Beermann, V., Gocyla, M., Willinger, E., Rudi, S., Heggen, M., Dunin-Borkowski, R.E., Willinger, M.G., and Strasser, P. (2016). RhDoped Pt-Ni octahedral nanoparticles: understanding the correlation between elemental distribution, oxygen reduction reaction, and shape stability. Nano Lett. 16, 1719–1725. 21. Huang, X., Zhao, Z., Cao, L., Chen, Y., Zhu, E., Lin, Z., Li, M., Yan, A., Zettl, A., Wang, Y.M., et al. (2015). High-performance transition metal-doped Pt3Ni octahedra for oxygen reduction reaction. Science 348, 1230–1234. 22. Ara´n-Ais, R.M., Dionigi, F., Merzdorf, T., Gocyla, M., Heggen, M., Dunin-Borkowski, R.E., Gliech, M., Solla-Gullo´n, J., Herrero, E., Feliu, J.M., and Strasser, P. (2015). Elemental anisotropic growth and atomic-scale structure of shape-controlled octahedral Pt-Ni-Co alloy nanocatalysts. Nano Lett. 15, 7473–7480. 23. Zhang, C., Sandorf, W., and Peng, Z. (2015). Octahedral Pt2CuNi uniform alloy nanoparticle catalyst with high activity and promising

14

Matter 1, 1–14, December 4, 2019

stability for oxygen reduction reaction. ACS Catal. 5, 2296–2300. 24. He, D., Zhang, L., He, D., Zhou, G., Lin, Y., Deng, Z., Hong, X., Wu, Y., Chen, C., and Li, Y. (2016). Amorphous nickel boride membrane on a platinum-nickel alloy surface for enhanced oxygen reduction reaction. Nat. Commun. 7, 12362. 25. Jia, Q., Zhao, Z., Cao, L., Li, J., Ghoshal, S., Davies, V., Stavitski, E., Attenkofer, K., Liu, Z., Li, M., et al. (2018). Roles of Mo surface dopants in enhancing the ORR performance of octahedral PtNi nanoparticles. Nano Lett. 18, 798–804. 26. Lim, J., Shin, H., Kim, M., Lee, H., Lee, K.S., Kwon, Y., Song, D., Oh, S., Kim, H., and Cho, E. (2018). Ga-doped Pt-Ni octahedral nanoparticles as a highly active and durable electrocatalyst for oxygen reduction reaction. Nano Lett. 18, 2450–2458. 27. Zhang, Y., Duan, Z., Xiao, C., and Wang, G. (2011). Density functional theory calculation of platinum surface segregation energy in Pt3Ni (111) surface doped with a third transition metal. Surf. Sci. 605, 1577–1582. 28. Wang, G., Van Hove, M.A., Ross, P.N., and Baskes, M.I. (2005). Monte Carlo simulations of segregation in Pt-Ni catalyst nanoparticles. J. Chem. Phys. 122, 024706.

downloads/fuel-cell-technologies-office-multiyear-research-development-and-22). 38. Arning, M.D., and Minteer, S.D. (2007). Electrode potentials. In Handbook of Electrochemistry, C. Zoski, ed. (Elsevier), pp. 813–827. 39. Metropolis, N., Rosenbluth, A.W., Rosenbluth, M.N., Teller, A.H., and Teller, E. (1953). Equation of state calculations by fast computing machines. J. Chem. Phys. 21, 1087– 1092. 40. Shao, M., Odell, J.H., Peles, A., and Su, D. (2014). The role of transition metals in the catalytic activity of Pt alloys: quantification of strain and ligand effects. Chem. Commun. (Camb.) 50, 2173–2176. 41. Cao, L., and Mueller, T. (2015). Rational design of Pt3Ni surface structures for the oxygen reduction reaction. J. Phys. Chem. C 119, 17735–17747. 42. Jia, Q., Liang, W., Bates, M.K., Mani, P., Lee, W., and Mukerjee, S. (2015). Activity descriptor identification for oxygen reduction on platinum-based bimetallic nanoparticles: in situ observation of the linear compositionstrain-activity relationship. ACS Nano 9, 387–400.

29. Cao, L., and Mueller, T. (2016). Theoretical insights into the effects of oxidation and Modoping on the structure and stability of Pt-Ni nanoparticles. Nano Lett. 16, 7748–7754.

43. Strasser, P., Koh, S., Anniyev, T., Greeley, J., More, K., Yu, C., Liu, Z., Kaya, S., Nordlund, D., Ogasawara, H., et al. (2010). Lattice-strain control of the activity in dealloyed core-shell fuel cell catalysts. Nat. Chem. 2, 454.

30. Callejas-Tovar, R., Diaz, C.A., de la Hoz, J.M.M., and Balbuena, P.B. (2013). Dealloying of platinum-based alloy catalysts: kinetic Monte Carlo simulations. Electrochim. Acta 101, 326–333.

44. Mueller, T., and Ceder, G. (2009). Bayesian approach to cluster expansions. Phys. Rev. B 80, 024103.

31. Wang, C., Chi, M., Wang, G., van der Vliet, D., Li, D., More, K., Wang, H.–H., Schlueter, J.A., Markovic, N.M., and Stamenkovic, V.R. (2011). Correlation between surface chemistry and electrocatalytic properties of monodisperse PtxNi1-x nanoparticles. Adv. Funct. Mater. 21, 147–152. 32. Sanchez, J.M., Ducastelle, F., and Gratias, D. (1984). Generalized cluster description of multicomponent systems. Physica A 128, 334–350. 33. Cao, L., Li, C., and Mueller, T. (2018). The use of cluster expansions to predict the structures and properties of surfaces and nanostructured materials. J. Chem. Inf. Model. 58, 2401–2413. 34. Kohn, W., and Sham, L.J. (1965). Self-consistent equations including exchange and correlation effects. Phys. Rev. 140, A1133–A1138. 35. Schulze, T.P. (2008). Efficient kinetic Monte Carlo simulation. J. Comput. Phys. 227, 2455– 2462. 36. Serebrinsky, S.A. (2011). Physical time scale in kinetic Monte Carlo simulations of continuoustime Markov chains. Phys. Rev. E 83, 037701. 37. US Department of Energy Fuel Cell Technologies Office Multi-Year Research, Development, and Demonstration Plan. (2016). (Updated May 2017) (Department of Energy https://www.energy.gov/eere/fuelcells/

45. Kresse, G., and Furthmu¨ller, J. (1996). Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186. 46. Zhang, Y., and Yang, W. (1998). Comment on ‘‘generalized gradient approximation made simple’’. Phys. Rev. Lett. 80, 890. 47. Perdew, J.P., Burke, K., and Ernzerhof, M. (1996). Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865–3868. 48. Hammer, B., Hansen, L.B., and Nørskov, J.K. (1999). Improved adsorption energetics within density-functional theory using revised Perdew-Burke-Ernzerhof functionals. Phys. Rev. B 59, 7413–7421. 49. Blo¨chl, P.E. (1994). Projector augmented-wave method. Phys. Rev. B 50, 17953–17979. 50. Wisesa, P., McGill, K.A., and Mueller, T. (2016). Efficient generation of generalized MonkhorstPack grids through the use of informatics. Phys. Rev. B 93, 155109. 51. Methfessel, M., and Paxton, A.T. (1989). Highprecision sampling for Brillouin-zone integration in metals. Phys. Rev. B 40, 3616– 3621. 52. Momma, K., and Izumi, F. (2008). VESTA: a three-dimensional visualization system for electronic and structural analysis. J. Appl. Crystallogr. 41, 653–658.