Crystal structure engineering in multimetallic high-index facet nanocatalysts

Significance

Crystal structure engineering in nanoparticles represents an important strategy in catalyst design, as catalytic performance is highly dependent on atomic arrangement. However, realizing crystal structure engineering in high-index facet nanocatalysts has been a synthetic challenge for decades. Here, we first employed density functional theory calculations to determine whether surface modifications can stabilize high-index facets during crystal structure transformations and then synthesized a library of multimetallic high-index facet tetrahexahedral (THH) nanoparticles with controllable crystal structures. Importantly, the crystal structural transition between the intermetallic and chemically disordered states is reversable, and the THH morphology of nanocatalysts is maintained during this transformation. This approach broadens the synthetic scope of nanocatalysts available for use in diverse chemical processes.

Abstract

In the context of metal particle catalysts, composition, shape, exposed facets, crystal structure, and atom distribution dictate activity. While techniques have been developed to control each of these parameters, there is no general method that allows one to optimize all parameters in the context of polyelemental systems. Herein, by combining a solid-state, Bi-influenced, high-index facet shape regulation strategy with thermal annealing, we achieve control over crystal structure and atom distribution on the exposed high-index facets, resulting in an unprecedentedly diverse library of chemically disordered and ordered multimetallic (Pt, Co, Ni, Cu, Fe, and Mn) tetrahexahedral (THH) nanoparticles. Density functional theory calculations show that surface Bi modification stabilizes the {210} high-index facets of the nanoparticles, regardless of their internal atomic ordering. Moreover, we find that the ordering transition temperatures for the nanoparticles are dependent on their composition, and, in the case of Pt₃Fe₁ THH nanoparticles, increasing Ni substitution leads to an order-to-disorder transition at 900 °C. Finally, we have discovered that ordered intermetallic THH Pt₁Co₁ nanocatalysts exhibit a catalytic performance superior to disordered THH Pt₁Co₁ nanoparticles and commercial Pt/C catalysts toward methanol electrooxidation, highlighting the importance of crystal structure and atom distribution control on high-index facets in nanoscale catalysts.

Internal crystal structure is an important design parameter for multimetallic nanoparticle (NP) catalysts, as their performance is highly dependent on atomic arrangement (1–5). Additionally, high-index facet NPs generally exhibit superior activity and selectivity compared to their low-index facet counterparts due to the high density of catalytically favorable low-coordination steps and edges on their surfaces (6–10). As such, a method to engineer the crystal structure of multimetallic high-index facet NPs is highly desirable. However, due to the structural complexity of these NPs, such strategies are lacking to date. While high-index facet nanostructures with specific crystal structures have been prepared by solution-phase methods, the complex reaction kinetics involved in these syntheses limit their generalizability to various compositions and crystal structures (11, 12). On the other hand, thermal annealing treatments are commonly used to modify the internal atomic ordering of as-synthesized NPs, since distinct crystal structures are thermodynamically favored at different annealing temperatures (13–15). This method, however, is not compatible with conventional high-index facet NPs because the organic ligands that stabilize their low-coordination facets decompose upon thermal annealing, resulting in the reconstruction of their facets into low-index ones. One way to circumvent this challenge is to synthesize high-index facet NPs that are stable under high temperatures.

Recently, we reported a highly generalizable strategy for synthesizing a library of multimetallic tetrahexahedral (THH) NPs enclosed by {210} high-index facets through alloying/dealloying with foreign metals (Bi, Sb, Pb, or Te) (16–18). Importantly, these high-index facet NPs are prepared at high temperatures (∼900 °C), offering a wide temperature window to further modify their crystal structures. Here, we demonstrate through density functional theory (DFT) calculations that the {210} high-index facets of multimetallic THH NPs can be stabilized with foreign metals (e.g., Bi), regardless of their internal crystal structure. We further confirm this result by experimentally realizing a library of 16 multimetallic (with components Pt, Co, Ni, Cu, Fe, and Mn) high-index facet THH NPs with either chemically disordered or ordered (intermetallic) crystal structures through Bi modification. Moreover, we show that the crystal structure transition temperatures of these NPs can be modulated by tuning their elemental ratios, and by increasing Ni substitution, the crystal structure of Pt₃Fe₁ THH NPs changes at 900 °C from ordered L1₂ to disordered A1. Finally, we evaluate both chemically disordered and intermetallic Pt₁Co₁ THH NPs as electrocatalysts for the methanol oxidation reaction and find that the intermetallic NPs exhibit superior activity and stability.

Results and Discussion

As an initial proof-of-concept, we targeted the Pt–Co system and evaluated the impact of crystal structure and surface Bi modification on NP morphology using surface-energy calculations. While PtCo NPs can form a chemically disordered A1 structure with any Pt/Co elemental ratio, intermetallic ones can only form over narrow ranges around certain Pt/Co ratios, including L1₀ Pt₁Co₁, L1₂ Pt₃Co₁, and L1₂ Pt₁Co₃ (19, 20). Importantly, L1₀ Pt₁Co₁ has a tetragonal structure with ordered, alternating Pt and Co atomic layers along the c axis, which is markedly different from the face-centered cubic (fcc) structure of A1 Pt₁Co₁, where Pt and Co atoms are randomly distributed within the lattice (14). In previous reports (16), we discovered that surface Bi modification can stabilize the {210} facets of NPs with chemically disordered fcc structures. However, it remains uncertain whether this strategy can be extended to ordered tetragonal structures. To answer this question, we used DFT simulations to calculate the specific surface energies of the (210), (102), and (201) high-index facets, as well as all the low-index facets [including the (111), (100), (001), (101), and (110) facets] of L1₀ Pt₁Co₁ NPs before and after Bi modification (Fig. 1A and SI Appendix, Fig. S1). The calculation results show that Bi modification can significantly change the specific surface energies of these facets. Specifically, without Bi modification, the specific surface energies of low-index facets [e.g., 1.70 J/m² for the (111) facet and 1.42 J/m² for the (001) facet] are smaller than those of high-index facets [e.g., 2.27 J/m² for the (210) facet and 2.44 J/m² for the (201) facet], resulting in a truncated octahedron equilibrium Wulff shape capped by (111), (100), (001), and (101) facets, consistent with literature reports (21). However, with increasing Bi coverage, the specific surface energies of high-index facets decrease drastically. For instance, the (210), (102), and (201) facets all have an ∼90% drop in specific surface energies when the Bi coverage increases from 0 to 100%. On the other hand, Bi modification has the opposite effect on low-index facets, making them energetically unfavorable. Consequently, with 100% Bi coverage, high-index facets exhibit much lower surface energies than low-index ones, and the equilibrium Wulff shape for L1₀ Pt₁Co₁ becomes a tetrahexahedron (Fig. 1C).

Fig. 1.

(A and B) Specific surface energies of Bi-modified facets of L1₀ Pt₁Co₁ [A; Upper shows the surface energies of (111), (100) and (001) facets, while Lower shows those of other facets evaluated] and L1₂ Pt₃Co₁ as a function of the surface Bi coverage (B). (C and D) Wulff structure evolution of L1₀ Pt₁Co₁ (C) and L1₂ Pt₃Co₁ (D) as a function of surface Bi coverage.

To confirm the DFT results, we experimentally explored the behavior of Pt₁Co₁ NPs upon Bi surface modification. To achieve compositionally uniform NPs, 8 (±1) nm A1-structured quasispherical Pt₁Co₁ NPs were synthesized via a solution-phase method to serve as precursor seeds (SI Appendix, Fig. S2) and then transferred to a tube furnace for thermal treatment in a Bi atmosphere. Based on the phase diagram, Pt₁Co₁ alloys tend to take on the A1 structure at high temperatures and the L1₀ structure at low temperatures, with a structural transition at ∼820 °C (20). After 1-h annealing at 900 °C, the quasispherical Pt₁Co₁ NPs were successfully transformed into THH NPs, as confirmed by scanning electron microscopy (SEM), scanning transmission electron microscopy (STEM), and energy-dispersive X-ray spectroscopy (EDS; Fig. 2A and SI Appendix, Figs. S3 and S4). Furthermore, X-ray diffraction (XRD) data confirm that these THH NPs take on the A1 structure (Fig. 2B), and the Miller indices of their exposed facets were indexed to be {210} (Fig. 2 E and F), consistent with previous results (16). Further annealing of the A1 Pt₁Co₁ THH NPs at 700 °C for 0.5 to 2 h induced a transition to the L1₀ phase, as indicated by the appearance of (001) and (110) peaks at 24.2° and 33.4°, respectively, in the XRD patterns, which are forbidden in A1 structures (Fig. 2B). Additionally, due to the difference in intrinsic magnetism between L1₀ Pt₁Co₁ (ferromagnetic) and A1 Pt₁Co₁ (superparamagnetic) at room temperature (20), the crystal structure transition can be further confirmed by measuring their hysteresis loops. Specifically, with increasing annealing time at 700 °C, the coercivity of the resulting Pt₁Co₁ NPs increased and reached a maximum value of 5.2 kilooersteds (kOe) after 2 h of annealing. Annealing these NPs for another 2 h did not significantly change their coercivity, indicating that Pt₁Co₁ NPs were fully transformed into L1₀ structures (Fig. 2 C and D). To unambiguously confirm the formation of L1₀ Pt₁Co₁ THH NPs, a selected-area electron diffraction (SAED) pattern was collected from the [001] direction of a single Pt₁Co₁ THH NP annealed at 700 °C for 2 h. The SAED pattern exhibited (100) and (110) diffraction spots, indicative of L1₀ Pt₁Co₁ (Fig. 2 H and I). Additionally, the L1₀ Pt₁Co₁ THH NP lattice was directly imaged by using aberration-corrected STEM. Due to differences in their Z-contrast, Pt atoms are brighter than Co atoms in dark-field STEM imaging. In the A1 Pt₁Co₁ THH NP, no obvious brightness difference was observed due to the random distribution of Pt and Co atoms (Fig. 2G), while alternating Pt and Co atomic layers can be clearly identified in the L1₀ Pt₁Co₁ THH NP (Fig. 2J). Finally, the transition between the A1 and L1₀ structures was confirmed to be reversible. Specifically, the L1₀ Pt₁Co₁ THH NPs were transformed back into A1 structures by further annealing them at 900 °C for 1 h (SI Appendix, Fig. S5).

Fig. 2.

Morphology and crystal structure characterization of A1 and L1₀ Pt₁Co₁ THH NPs. (A) SEM image of as-prepared Pt₁Co₁ THH NPs at 900 °C. (B) XRD patterns of A1 and L1₀ Pt₁Co₁ THH NPs obtained at 900 and 700 °C, respectively. Standard diffraction patterns of A1 (Powder Diffraction File [PDF]: 04-003-4659) and L1₀ Pt₁Co₁ (PDF: 03-065-8969) are provided for comparison. A.u., arbitrary units. (C) Hysteresis loops of Pt₁Co₁ THH NPs annealed at 700 °C for different times. (D) Three-dimensional model of an L1₀ Pt₁Co₁ THH NP, where the gray atoms are Pt and the blue atoms are Co. (E, F, H, and I) TEM images and corresponding SAED patterns of A1 Pt₁Co₁ (E and F) and L1₀ Pt₁Co₁ (H and I) THH NPs. (Scale bars, 25 nm.) (E and H, Insets) Ideal model of a THH NP projected along the [001] direction. (G and J) Aberration-corrected STEM images of an A1 Pt₁Co₁ THH NP imaged along the [111] direction (G) and an L1₀ Pt₁Co₁ THH NP imaged along the [110] direction (J). (J, Inset) An atomic model of the L1₀ Pt₁Co₁ structure, where Pt is in red and Co is in blue.

Motivated by the successful crystal structure engineering of Pt₁Co₁ THH NPs, we further tested this method on Pt₃Co₁ and Pt₁Co₃ NPs, which can take on either A1 or intermetallic L1₂ structures (SI Appendix, Fig. S6). DFT calculations predict that in the L1₂ Pt₃Co₁ system with 100% Bi coverage, the specific surface energy of {210} high-index facets (0.11 J/m²) is much smaller than those of low-index facets (0.56 J/m² for {110}, 1.72 J/m² for {100}, and 3.94 J/m² for {111} facets; Fig. 1B). Therefore, L1₂ Pt₃Co₁ NPs should form THH rather than truncated octahedron, which is the preferred morphology when there is 0% Bi coverage (Fig. 1D). To test this prediction, we synthesized A1 Pt₃Co₁ THH NPs by heating corresponding quasispherical ones at 900 °C for 1 h in a Bi atmosphere. Since Pt₃Co₁ has a structural transition temperature at 735 °C (20), L1₂ Pt₃Co₁ NPs were synthesized by further annealing the A1 THH NPs at 700 °C for 2 h (Fig. 3 A and B). As with the Pt₁Co₁ system, Pt and Co ordering within L1₂ Pt₃Co₁ THH NPs resulted in (100) and (110) diffractions. This long-range ordering was characterized by aberration-corrected STEM along the [001] axis, which shows alternating layers of pure Pt and mixed Pt–Co (Fig. 3G). Similarly, A1 and L1₂ Pt₁Co₃ THH NPs can be obtained at 900 °C and 600 °C, respectively [structure transition temperature: 655 °C (20); Fig. 3 C, D, and H]. Indeed, the generalizability of this strategy to engineer the crystal structure of high-index facet NPs is unprecedented, and, in this work, we further extended its scope to other systems, including Pt–Ni, Pt–Cu, Pt–Fe, and Pt–Mn (SI Appendix, Figs. S6–S9). For example, A1 Pt₁Ni₁ THH NPs and Pt₁Cu₁ THH NPs can be synthesized at 900 °C, and by postannealing them at 600 °C for 12 h and 700 °C for 2 h, respectively, L1₀ Pt₁Ni₁ THH NPs and L1₁ Pt₁Cu₁ THH NPs can be obtained (Fig. 3 E, F, and I and SI Appendix, Fig. S8).

Fig. 3.

(A–F) TEM images (Left) and corresponding SAED patterns along the [001] direction (Right) of A1 Pt₃Co₁ (A), L1₂ Pt₃Co₁ (B), A1 Pt₁Co₃ (C), L1₂ Pt₁Co₃ (D), A1 Pt₁Ni₁ (E), and L1₀ Pt₁Ni₁ (F) THH NPs. (G) Aberration-corrected STEM image of a L1₂ Pt₃Co THH NP. (G, Inset) Atomic model of the L1₂ Pt₃Co₁ structure, where Pt is in red and Co is in blue. (H and I) STEM images and EDS maps of Pt₁Co₃ (H) and Pt₁Ni₁ (I) THH NPs. (Scale bars, 30 nm.)

Due to the high structural transition temperatures of Pt–Fe and Pt–Mn (1,350 °C for bulk Pt₃Fe₁ and 995 °C for bulk Pt₃Mn₁) (22), intermetallic L1₂ Pt₃Fe₁ and L1₂ Pt₃Mn₁ THH NPs were directly obtained after heating the corresponding quasispherical ones at 900 °C for 1 h in a Bi atmosphere (Fig. 4A and SI Appendix, Fig. S9). While further heating these NPs at a sufficiently high temperature (above their structural transition temperatures) can, in principle, produce the corresponding A1 THH NPs, here, we explored elemental substitution as an alternative strategy for synthesizing A1 THH NPs under milder annealing conditions. Taking the Pt–Fe system as an example, incorporating increasing amounts of Ni into the Pt₃Fe₁ NPs resulted in more disordered structures after annealing at 900 °C for 1 h. With 10% Ni substitution, the (100) and (110) diffraction peaks disappeared in the XRD pattern of the resulting Pt₃Fe_0.6Ni_0.4 THH NPs, indicating the formation of A1 structures (Fig. 4 B and C). Moreover, the 10% Ni-substituted Pt–Fe THH NPs can be reconverted into L1₂ NPs by annealing at 600 °C for 2 h (SI Appendix, Fig. S10). Taken together, these results indicate that elemental substitution can be used as an additional tool to fine-tune the crystal structure transition process of multimetallic THH NPs. Since the solid-state THH shape regulation strategy works for a vast library of multimetallic NPs (e.g., Pt, Pd, Rh, Au, Cu, Ni, Co, Fe, and Mn) (17), it offers great flexibility in the choice of elemental compositions for engineering the crystal structure of multimetallic high-index facet NPs, including the potential to form high-entropy alloys and intermetallics (23, 24). Indeed, as a proof-of-concept, we prepared quinary Pt₄₅Fe₁₅Co₁₃Ni₁₂Cu₁₅ THH NPs with both A1 and L1₀ crystal structures at 900 °C and 600 °C, respectively (Fig. 4D and SI Appendix, Fig. S11).

Fig. 4.

(A and B) SEM images, STEM images, and corresponding EDS maps of Pt₃Fe₁ (A) and Pt₃Fe_0.6Ni_0.4 (B) THH NPs. (C) XRD patterns of PtFe THH NPs with 0% (Pt₃Fe₁), 5% (Pt₃Fe_0.8Ni_0.2), and 10% (Pt₃Fe_0.6Ni_0.4) Ni substitution synthesized at 900 °C for 1 h. A.u., arbitrary units. (D) SEM image, STEM image, and corresponding EDS maps of a quinary PtFeCoNiCu THH NP. (Scale bars, 50 nm.)

We tested the effect of crystal structure on the catalytic performance of THH NPs toward methanol electrooxidation, an important anode reaction in direct-methanol fuel cells (25–29), using A1 and L1₀ Pt₁Co₁ THH NPs (Fig. 5). To reveal their intrinsic catalytic activity, we normalized the measured current to the electrochemical surface area of Pt. According to the literature (25), a typical methanol-oxidation polarization curve consists of two sweeps. The current peak at ∼0.9 V in the forward sweep originates from methanol oxidation into CO₂, CO, and/or other carbonaceous intermediates, while the one at ∼0.65 V in the backward sweep corresponds to the oxidation of the carbonaceous intermediates into CO₂. Current densities during the forward sweep are typically used to evaluate the catalytic efficiency of NPs. These measurements indicate that both L1₀ and A1 Pt₁Co₁ THH nanocatalysts are more active than the commercial Pt/C catalyst. At an overpotential of 0.8 V, the current densities of L1₀ and A1 Pt₁Co₁ THH nanocatalysts are 3.4 and 2.3 mA/cm², which are 14 and 9 times higher than that of the commercial Pt/C catalyst (0.25 mA/cm²), respectively. This indicates the favorable effect of high-index facets on catalytic activity. Additionally, the L1₀ Pt₁Co₁THH nanocatalyst exhibited higher activity than the A1 Pt₁Co₁ THH nanocatalyst, indicating that atomic ordering within these NPs can further improve catalytic activity (Fig. 5C). We also compared the performance of the THH nanocatalysts to Johnson–Matthey PtRu black, an industry-standard catalyst used in direct methanol-oxidation fuel cells and some recently reported nanocatalysts (SI Appendix, Fig. S12 and Table S1). The L1₀ and A1 Pt₁Co₁ THH nanocatalysts exhibited sixfold and fourfold higher activity, respectively, than PtRu black. Finally, amperometric i–t curves revealed that the L1₀ Pt₁Co₁ THH nanocatalyst exhibits superior catalytic stability compared to both A1 Pt₁Co₁ THH and the commercial Pt/C catalysts. After continuous operation for 3,600 s at 0.8 V, the measured current density for L1₀ Pt₁Co₁ THH was 0.41 mA/cm², which is 3 times and 21 times higher than that from A1 Pt₁Co₁ THH (0.15 mA/cm²) and the commercial Pt/C catalyst (0.02 mA/cm²; Fig. 5D), respectively.

Fig. 5.

Catalytic properties of A1 and L1₀ Pt₁Co₁ THH NPs. (A) Cyclic voltammograms of different catalysts in an Ar-saturated 0.5 M H₂SO₄ solution. (B and C) Polarization curves (B) and histograms (C) of the specific activities at 0.8 V of methanol electrooxidation in Ar-saturated 0.5 M H₂SO₄ + 1 M CH₃OH for different catalysts. RHE, reversible hydrogen electrode.(D) Chronoamperometry curves of methanol electrooxidation for the different catalysts at 0.8 V. Error bars in C represent the SD.

Conclusions

To summarize, we have prepared a library of multimetallic (Pt, Co, Ni, Cu, Fe, and Mn) THH NPs with chemically disordered and intermetallic structures. Bi atoms stabilize the {210} facets of these NPs at high temperatures, which provides a unique opportunity to deliberately control their crystal structure through annealing temperature and chemical composition while keeping THH morphology. Testing of these catalysts for the methanol-oxidation reaction showed that the L1₀ Pt₁Co₁ THH nanocatalysts exhibit superior activity and stability over their A1-structured counterparts. Given the generalizability of this synthesis strategy, these findings highlight the vast potential for the design and industrial-scale manufacturing of cost-effective highly optimized catalysts. Additionally, since THH formation is independent of initial NP shape, and the structural transformation between the intermetallic and chemically disordered state of THH NPs is reversable, this strategy may play a role in catalyst recycling and reactivation for diverse chemical processes.

Materials and Methods

Materials.

Platinum acetylacetonate (97%), manganese acetylacetonate (97%), iron acetylacetonate (97%), cobalt acetylacetonate (97%), nickel acetylacetonate (95%), copper acetylacetonate (97%), oleylamine (70%), Pt/C catalyst (20 wt% Pt loading), Nafion solution (5 wt%), ethanol, and hexane were purchased from Sigma-Aldrich, Inc., and utilized without further purification. Carbon black powder (Vulcan XC-72) was purchased from Cabot. Johnson–Matthey PtRu black (HiSPEC 6000) was purchased from Alfa Aesar. Transmission electron microscopy (TEM) grids with silicon nitride support films were purchased from Ted Pella, Inc.

Synthesis of A1-Structured THH NPs.

Colloidal syntheses of uniform Pt-based NP seeds are described in SI Appendix. Then, a piece of silicon wafer or a silicon nitride TEM grid was loaded with colloidally synthesized spherical NPs of interest and placed in the center of a tube furnace. Approximately 1 mg of Bi powder was loaded in a combustion boat and placed in the tube, upstream of the NPs. The thermal treatment was programmed as follows: Under H₂ gas flow (flow rate: 200 standard cubic cm/min [sccm]), ramp to 600 °C within 12 min, hold at 600 °C for 10 min, ramp to 900 °C in 10 min, and hold at 900 °C for 1 h. Upon completion of the synthesis, the tube was quenched to room temperature.

Synthesis of Intermetallic PtCo, PtNi, or PtCu THH NPs.

A piece of silicon wafer loaded with the above synthesized A1 THH NPs was placed in a tube furnace for thermal treatment, which differed for different NPs. To synthesize L1₀ Pt₁Co₁, L1₁ Pt₁Cu₁, and L1₂ Pt₃Co₁ THH NPs, ramp to 700 °C in 23 min, and hold at 700 °C for 2 h. To synthesize L1₂ Pt₁Co₃, and L1₀ Pt₁Ni₁ THH NPs, ramp to 600 °C in 20 min, and hold at 600 °C for 12 h. The thermal treatments were all performed under H₂ gas flow (flow rate: 200 sccm).

Synthesis of Intermetallic Pt₃Mn₁ or Pt₃Fe₁ THH NPs.

Intermetallic Pt₃Mn₁ or Pt₃Fe₁ THH NPs were synthesized by thermal annealing the corresponding colloidally synthesized quasispherical Pt₃Fe₁ or Pt₃Mn₁ NPs on a piece of silicon wafer. The thermal treatment condition was the same as that used for the synthesis of A1 Pt₁Co₁ THH NPs.

Synthesis of Multimetallic THH NPs.

L1₂ Pt₃Fe_0.8Ni_0.2, A1 Pt₃Fe_0.6Ni_0.4, and A1 Pt₄₅Fe₁₅Co₁₃Ni₁₂Cu₁₅ THH NPs were obtained by annealing Pt₃Fe_0.8Ni_0.2, Pt₃Fe_0.6Ni_0.4, or PtFeCoNiCu colloidal NPs according to the same synthesis procedure as that of A1 Pt₁Co₁ THH NPs. To synthesize L1₂ Pt₃Fe_0.6Ni_0.4 and L1₀ Pt₄₅Fe₁₅Co₁₃Ni₁₂Cu₁₅ THH NPs, further thermal treatments were programmed as follows: Under H₂ gas flow (flow rate: 200 sccm), ramp to 600 °C in 20 min, hold at 600 °C for 2 h, and quench to room temperature.

DFT Simulations.

All DFT calculations in this work were performed by using the Vienna ab initio simulation package (30–33) with the projector-augmented wave potentials and the Perdew–Burke–Ernzerhof (34–36) formulation of the generalized gradient approximation. The plane-wave cutoff energy was set to be 400 eV, and the structures were fully relaxed until the energy was converged within 10⁻⁵ eV. At least 4,000 k-points per reciprocal atom gamma-centered k-point meshes were generated to sample the Brillouin zone, and spin polarization was included for all DFT calculations in this work. The surface models were constructed by following the procedure used by Huang et al. (16), where a crystal slab containing 11 atomic layers and a vacuum region of 15 Å were stacked in the direction perpendicular to the surface. For slabs with two possible terminations [e.g., (001) and (110) in the L1₀ structure and (100) and (110) in the L1₂ structure], only models with lower energies were considered. In the presence of foreign metal (Bi) absorbed on the surface, Bi atoms were only attached to the topmost layer for all surface models. When performing DFT calculations on these surface models, the middle three layers were fixed, while the other layers could relax. The surface energy was calculated by γ = (E_slab − Σn_iμ_i)/2A_surface, where E_slab is the total energy of the surface model, n_i is the number of atoms, μ_i is the chemical potential of the element i, and A_surface is the surface area. The total energy per atom of the bulk structure was used as the chemical potential for Pt and Bi, while the chemical potential of Co was determined by subtracting the energy of Pt atoms from the total energy of Co–Pt binary compounds (i.e., ${\mu }_{Co}^{CoPt}=E_{tot}^{CoPt}-{\mu }_{Pt}$).

Characterization.

SEM images were taken with a Hitachi SU-8030 field-emission SEM with an acceleration voltage of 30 kV and a current of 20 µA. STEM images were taken with a Hitachi HD-2300 STEM at an acceleration voltage of 200 kV. The composition was studied by using the dual EDS detectors on the HD-2300 STEM. The Lα peaks of Pt and the Kα peaks of Mn, Fe, Co, Ni, and Cu in the EDS spectra were used for elemental mapping and composition quantification. Atomic compositions of the NPs were calculated based on EDS spectral analysis with the Cliff–Lorimer correction method. The atomic composition measured by EDS had an inherent error of less than 5% due to X-ray absorption and fluorescence. Thermo Scientific NSS software was used for background subtraction in the EDS maps. High-resolution TEM images and SAED patterns were taken with a JEOL ARM 200F GrandARM TEM at an acceleration voltage of 300 kV. XRD spectra were collected on a Rigaku Ultima with a Cu Kα source. Magnetic properties were measured by using a vibrating sample magnetometer under a maximum applied field of 15 kOe. The electrochemical measurements were performed in a three-electrode glass cell at 298 K using an Epsilon Eclipse Workstation. A coiled Pt wire and an Ag/AgCl electrode were used as the counter and reference electrodes, respectively. All potentials were calibrated versus a reversible hydrogen electrode. The electrochemical surface area was determined electrochemically by the adsorption–desorption of hydrogen between 0.05 and 0.4 V and assuming 210 μC⋅cm⁻² for a monolayer of adsorbed hydrogen on the Pt surface. The cyclic voltammetry measurements were carried out in 0.5 M H₂SO₄ under a flow of Ar gas at a sweep rate of 50 mV⋅s⁻¹. Methanol electrooxidation reactions were measured in 0.5 M H₂SO₄ + 1 M CH₃OH at a sweep rate of 20 mV⋅s⁻¹.

Data Availability

All study data are included in the article and SI Appendix.