Unveiling the Origin of Morphological Instability in Topologically Complex Electrocatalytic Nanostructures

Abstract

Coarsening and degradation phenomena in metals have largely focused on thermally driven processes, such as bulk and surface diffusion. However, dramatic coarsening has been reported in high-surface-area, nanometer-sized Pt-based catalysts during potential cycling in an electrolyte at room temperaturea temperature too low for the process to be explained purely by surface mobility values measured in both vacuum and electrolytes (∼10^–22 and ∼10^–18 cm²/s, respectively). This morphological evolution must be due to a different mechanism for mass transport that is sensitive to electrochemical conditions (e.g., electrolyte composition, potential limits, and scan rate). However, there have been no notable studies of electrochemically induced coarsening in nanometer-sized electrocatalysts. Here, we unveil the origins of coarsening in an electrolyte through coupled in situ experiments and atomistic kinetic Monte Carlo (kMC) simulations. Our work demonstrates electrochemical coarsening is driven by two concurrent mechanisms that can be explained at the atomistic level: (i) dissolution/redeposition during the reduction of an oxidized species and (ii) rapid surface diffusion of undercoordinated atoms.

Introduction

Electrochemical energy conversion technologies, such as electrolyzers and fuel cells, are critical to a renewable carbon-neutral energy portfolio. , The growth of their integration into consumer and industrial applications is directly tied to economic descriptors that depend on their efficiency and operational lifetimes, ultimately determined by the performance of the electrocatalysts that make up the anodic and cathodic electrodes. , While significant research efforts have been underway to develop earth-abundant electrocatalytic materials, necessary performance metrics, such as energy density and energy efficiency, have only reliably been met with platinum group metal (PGM)-incorporated materials. , However, the high cost and relative scarcity of PGM necessitate significant improvements in their active area and mass-normalized activities to lower their mass loadings. ,− For example, the widespread commercialization of polymer electrolyte membrane fuel cells remains limited by the high loading of platinum (Pt) to compensate for the slow oxygen reduction reaction (ORR) kinetics. ,

To overcome these limitations, recent electrocatalyst development has focused on the design and synthesis of nanoarchitectured catalysts (NACs). − We define NACs as a broad class of catalystsconsisting of tailored surface facet ratios, feature sizes, surface site distributions, and compositionwhere the activity and surface fraction of catalytically useful precious metal have been maximized. NACs possess higher surface-area-to-volume ratios than spherical nanoparticles due to more complex shapessuch as nanocages, , porous networks, − and jagged nanowires , and as a result have continuously set new records for specific and mass activities over the last ten years. These catalysts possess activities 4–20× over the industry standard Pt/C but suffer electrochemical active surface area (ECSA) losses of 8–40% during accelerated stability testing (AST), as shown in Figure . ,,,,− Perhaps more concerning, these ECSA losses become more severe with increasing upper potential limit (UPL) during load cycling. ,,

1.

Summary of representative studies on ECSA change with various upper potential limits after AST for nanoarchitectured (open-frameworks, nanowires, anisotropic shapes, etc.) Pt-based electrocatalysts. ,,,− Solid circles represent AST protocols of 10,000 cycles, and open circles are for protocols with a number of cycles other than 10,000. Green stars represent data from this article for both electrochemically and thermally driven coarsening. The DOE 2030 target for ECSA retention (90,000 cycles) (black triangle) and representative Pt/Vulcan data from literature (black squares) − are also included.

This ECSA degradation is distinctly different from the activity loss observed in traditional spherical nanoparticle electrocatalysts during load cycling. For instance, carbon-supported solid Pt and Pt-alloy (both homogeneous and core–shell , ) nanoparticle electrocatalysts under load cycling have two primary active area loss mechanisms: (1) sintering/agglomeration due to weak physisorption onto the carbon support as well as oxidative loss of carbon support and (2) Ostwald ripening driven by the dissolution of the metallic catalyst. In contrast, NACs have been shown to degrade (i.e., lose active area per mass) without significant mass exchange between particles. ,,, Thus, unveiling the atomic-scale processes that govern the electrochemical degradation of these nanostructured materials at the fundamental level is critical for the development of next-generation electrocatalysts that are active and stable.

In this study, we examine the coarsening behavior of a model NAC material and highlight the limiting atomic processes that govern changes in the surface area and facet ratios during electrochemical cycling and thermal annealing. Through the combination of experimental and computational metrics, we demonstrate that surface dopants with either high reduction potentials or low surface mobility can limit the loss in the total electrochemically active surface area over an AST by a factor of 5. This work is the first instance where electrochemical coarsening is deconvoluted into distinct surface diffusion and dissolution/redeposition events. Insights from this work will have a measurable impact on the integration of active and morphologically stable materials into electrochemical energy conversion and storage devices.

Results and Discussion

We use nanoporous NiPt nanoparticles (np-NiPt) as representative NACs to assess morphological evolution (defined as changes in surface area, surface facet ratios, composition, and concentration of surface sites) during load cycling. NACs, especially nanoporous metals, have intrinsically metastable morphologies, which provide a large thermodynamic driving force to reduce their surface energy and thus surface area. , We fabricate np-NiPt nanoparticles by electrochemically dealloying precursor Ni₈₀Pt₂₀ alloy nanoparticles and then deposit Ir or Au on the surface of the nanoporous nanoparticles, Figure A,B (see Methods for further details). The resulting surfaces of the nanoparticles are uniformly decorated with a fraction of ∼5 atom % Au (np-NiPt + Au) and Ir (np-NiPt + Ir) (Figure C and Table S1), respectively.

2.

(A) TEM image, HAADF-STEM image, and corresponding Pt and Ni EELS mapping of nanoporous NiPt nanoparticles (np-NiPt) examined in this study. (B) Schematic illustration of the surface doping process. Cu is first underpotentially deposited onto np-NiPt nanoparticles and then decorated with either Au or Ir via galvanic displacement. (C) STEM-EDS analysis of the Ir, Pt, and Au L_α1 peaks. (D) Reduction potential and surface diffusivity comparison between Au, Pt, and Ir.

The benefit of controlling the introduction of dopants is that the initial surface conditions are identical for all elemental additions. Au and Ir are chosen for both their differing surface mobilities and standard reduction potentials, as demonstrated in Figure D. Surface diffusion rates for metallic species scale roughly with the melting point of the metal. , As such, we expect Ir dopants to have a lower surface mobility than Pt atoms and Au dopants to have a higher surface mobility. , The standard reduction potentials of Ir³⁺/Ir, Pt²⁺/Pt, and Au³⁺/Au are 1.0, 1.188, and 1.52 V vs SHE, respectively. For both surface mobility and standard reduction potential, Pt values sit about halfway between those of Ir and Au. The expected result is an impact on both the mobility of surface atoms and the average “nobility” of the surface following dopant incorporation as a function of dopant identity.

To characterize the morphological evolution of np-NiPt during thermal coarsening, where morphology evolution is limited to temperature-induced surface mobility, we applied in situ heating transmission electron microscopy (TEM), shown in Figure A. Starting from room temperature, we do not observe substantial morphological changes until about 300 °C, and the nanoporous structure coarsens into a solid particle after holding at 450 °C for 10 min. In situ heating TEM of np-NiPt + Ir and np-NiPt + Au (Figure A) nanoparticles under identical thermal conditions yields distinctly different time-dependent morphology profiles. In the case of capillary-driven coarsening, morphological evolution will be limited by the slowest diffusing species. For nanoporous metals, it has been demonstrated that the rate-limiting behavior is step-edge flow. Given Au’s high surface mobility, we expect Pt diffusion to be the rate-limiting behavior for both the np-NiPt and np-NiPt + Au samples, and both compositions will exhibit similar degrees of coarsening. However, the presence of Ir, likely concentrated at step edges, , will substantially reduce the surface mobility and lead to a negligible change in nanoporous morphology, confirmed in Figures A and C.

3.

(A) Thermal coarsening of np-NiPt, np-NiPt + Au, and np-NiPt + Pt nanoparticles through in situ heating TEM at 450 °C for 10 min. The scale bar is 50 nm. (B) Experimental summary (markers) of ECSA as a function of time for np-NiPt + Au, np-NiPt, and np-NiPt + Ir at a constant temperature of 450 °C; kMC simulation results (colored bands) were incorporated where the upper bound refers to total surface area and the lower bound refers to active surface area. The inset shows the change in the ratio of 111/100 facet area after 60 simulated seconds with the indicated surface decoration. (C) In situ heating TEM series of np-NiPt, np-NiPt + Ir, and np-NiPt + Au at a temperature of 450 °C (left) and corresponding kMC simulations of thermal coarsening (right) after the indicated time duration.

Figure B contains a summary of time-dependent ECSA for np-NiPt, np-NiPt + Ir, and np-NiPt + Au at 450 °C. For these experiments, catalyst films on glassy carbon (GC) substrates are exposed to elevated temperatures in a furnace under an inert environment for a fixed time. The ECSA is then measured electrochemically (CO stripping) in a hanging meniscus rotating disk electrode (RDE) configuration (details in the Methods section). In agreement with our assessment, the ECSA data indicate that a partial monolayer (ML) of Au has a negligible effect on thermal coarsening, but the partial ML of Ir helps retain much of the high aspect ratio morphology and area (Figure B,C). This morphological trend for each dopant is mirrored in our atomistic kinetic Monte Carlo (kMC) simulations of thermal coarsening in nanoporous nanoparticles via surface diffusion, as shown in Figures B and C. While the total surface area of the simulated nanoparticles (lower bound, Figure B) decreases more than in the experiments, the estimated active surface area (upper bound, Figure B), which accounts for changes in facet distribution, decreases less than in the experiments; we expect the experimentally measured ECSA to fall between these boundaries. In agreement with our above assessment, we find that the slow-diffusing Ir atoms collect at step edges (Figures S3 and S4) and inhibit mass transport. In contrast, fast-diffusing Au atoms do not impact Pt transport across the surface.

We now turn our attention to coarsening in electrochemical environments during an AST protocol with an UPL of 1.1 V vs reversible hydrogen electrode (RHE) at room temperature (Figure ). Similar to thermal coarsening, the presence of an Ir partial ML results in a negligible morphological change after AST in an electrolyte. We note that at the UPLs used, surface Ir is likely to be oxidized to higher-valent IrO _x species. As transition metal oxides possess lower surface mobility than their constituent metals, this does not change our analysis or conclusions as in all cases the Ir-based component, regardless of its oxidation state, is the slowest-moving component. IrO _x species are also sufficiently stable at the potentials tested here. However, unlike the thermal environment, electrochemical coarsening is also inhibited by the presence of a partial ML of Au, as shown in Figures and . It has been previously demonstrated that Au on or near a Pt surface increases the “nobility” of neighboring Pt atoms, which we expect to limit the degree of Pt dissolution during a full redox cycle by increasing the onset potential for Pt oxidation. , The stabilizing effect of Au during electrochemical coarsening suggests that under electrochemical cycling there must be an additional mechanism for mass transport: Pt dissolution followed by redeposition.

4.

(Left) Identical location (IL)-TEM series and corresponding kMC simulation (with dissolution/redeposition integrated into the coarsening mechanism) of np-NiPt, np-NiPt + Ir, and np-NiPt + Au during AST in Ar-purged 0.1 M HClO₄ at room temperature between 0.6 and 1.1 V vs RHE with a sweep rate of 50 mV s^–1. The electrolyte is static. The scale bar is 10 nm. (Right) IL-TEM and corresponding kMC simulation of np-NiPt, np-NiPt + Ir, and np-NiPt + Au after 10,000 cycles with a stirred electrolyte to induce Pt²⁺ ion concentration gradients.

5.

(A) Experimental and (B) kMC-simulated percent ECSA retained as a function of AST cycle number for np-NiPt + Au, np-NiPt, and np-NiPt + Ir with (dashed line) and without (solid line) electrolyte stirring. AST was assessed in Ar-purged 0.1 M HClO₄ at room temperature between 0.6 and 1.1 V vs RHE with a sweep rate of 50 mV s^–1.

This mechanism is distinct from Ostwald ripening because mass is not transferred between individual particles. Unlike solid nanoparticles, NACs have local regions of both positive and negative curvature and a highly tortuous morphology, which inhibits mass transport between particles until they coarsen and their morphologies resemble that of a dense nanoparticle. − Prior to this shape transition, coarsening is confined to individual NAC particles, by which material will dissolve from regions of high positive curvature (during an anodic/cathodic potential sweeps , ) and be redistributed along the surface via electrodeposition during the cathodic sweep. This event will couple with surface diffusion and result in the coarsening of the porous structure. Furthermore, any formation and reduction of surface oxides will lead to a disordered surface structure with a high density of adatoms, which will rapidly incorporate into the nearest step edge to reduce the energy of the system. , The experimental degradation trends are confirmed in the kMC simulations by implementing a simple dissolution/redeposition mechanism using an atomistic description of the Butler–Volmer equation (details in the Methods section). Without this mechanism, a model driven purely by surface diffusion exhibits limited coarsening in an electrolyte at RT for np-NiPt nanoparticles.

What accentuates dissolution and redeposition in NACs is their tortuosity: the interconnected solid network increases the probability of a cationic species contacting the surface prior to elution into the bulk electrolyte. To further demonstrate the impact of dissolution/redeposition, we compared the ECSA retained during AST with and without convective stirring in the electrolyte, shown in Figures , A, and B. The stirring condition is analogous to that of a fast-flowing electrolyte, which will ensure that a Pt ion concentration gradient is established radially in the NAC and increase the probability of Pt ion elution into the bulk solution. This will also potentially increase the diffusion distance along the surface prior to redeposition. There is a substantial difference for both experimental and simulated coarsening, shown in Figures A and B, across all three compositions with and without convection. The undoped np-NiPt and np-NiPt + Ir samples behave in a similar manner, showing a much higher ECSA loss with convection in comparison to that for np-NiPt + Au. The origin of these differing responses is associated with the dopant metal-induced changes in the oxophilicity of the surface, which directly impacts the degree of surface oxidation and consequently the amount of solubilized Pt ions formed during a potential sweep.

In the absence of convective Pt ion removal, dissolution/redeposition during repetitive redox cycles introduces low-coordinated surface atoms, which then undergo surface diffusion in an analogous mechanism to porosity evolution during dealloying. The relative concentration of these surface defects (e.g., adatoms versus kink atoms) will be dependent on the UPL, potential sweep rate, and the load cycling profile (e.g., a triangular versus square wave). A summary of our proposed mechanisms is listed in Figure . Au surface atoms reduce the oxophilicity of neighboring Pt atoms, reducing the quantity of locally oxidized and subsequently dissolved Pt per potential cycle. In addition to electronic effects, Au at the step edges could also decrease the degree of Pt oxidation/dissolution per cycle by passivation of the lower coordinated Pt atoms.

6.

Illustration demonstrating the key differences between Au and Ir dopants on electrochemical coarsening: Au dopants decrease the frequency of Pt dissolution events, whereas Ir dopants decrease the surface mobility of Pt atoms.

Our kMC simulations, Figure S3, demonstrate the Au-doped particles have a higher probability for the dopant to be preferentially located at step and edge sites. Both the decrease in the degree of Pt surface oxidation, corresponding to a lower degree of roughening following reduction, and the subsequent reduction in Pt elution and redeposition reduce step-edge flow during cycling. Convection increases the diffusion distance, but the frequency of dissolution/redeposition events is lower than that of undoped np-NiPt. The behavior of Ir dopants is more nuanced in that Ir is not expected to affect the propensity of neighboring Pt atoms to dissolve. However, Ir dopants are expected to pin rate-limiting sites for coarsening (here, steps , ), slowing surface diffusion and potentially protecting lower coordinated Pt atoms.

The reduction in surface mobility afforded by Ir, however, becomes negligible under convective conditions when mass can be transported across the surface via nonsurface diffusional processes and material can be lost to the bulk solution. This observation explains why Ir dopants have no substantial effect on electrochemical coarsening under stirring conditions, whereas Au dopants are protective under both static and convective electrochemical conditions. As noted in the beginning, the morphology of nanoporous metals affords intrinsic protection to this electrochemical degradation mechanism: the tortuosity of the porous network makes it more challenging (compared to a thin film) for species to be swept into the electrolyte, and the high fraction of saddle points increases the average coordination number of surface atoms to prevent dissolution during potential cycling. Unfortunately, these morphological factors are universal across nanoporous metals and are thus challenging to probe in our study. However, this degradation mechanism is expected to be more substantial in NACs that primarily have positive curvature such as dense nanoparticles and nanowires.

To further characterize the impact of metal surface dopants on Pt dissolution, in situ ICP-MS is used to simultaneously (i) measure the dissolution rate of Pt during potential cycling and (ii) confirm the suppression of Pt dissolution with the presence of partial ML of Au, Figures A and B. Electrolyte flow is used to acquire the in situ sample and transfer it to the ICP-MS, mimicking the convective conditions. The mass loss of Pt during a single cathodic–anodic redox cycle is found to decrease with increasing Au coverage (Figure A and B). As has been demonstrated previously, , the majority of the Pt dissolution occurs during the cathodic sweep through reduction of the oxide. Figures C and D show in situ ICP-MS measurements of Pt dissolution rates from np-NiPt, np-NiPt + Ir, and np-NiPt + Au. There is a clear reduction in the rate of Pt dissolution, during the cathodic sweep, in the presence of the Ir and Au dopants. The reduction in Pt dissolution with Ir is lower than that observed for Au. The partial ML of Ir can passivate some of the low-coordinated steps, but not all, while the same coverage of Au can both passivate and enhance the nobility of neighboring exposed Pt atoms, which results in a smaller amount of Pt dissolved. This is supported by the positive shift in the onset of Pt dissolution for the np-NiPt + Au in comparison to np-NiPt and np-NiPt + Ir, Figure D. Previous work has shown that subsurface Au stabilizes surface Pt through electronic and structural modification; this is in addition to the stabilization of lower coordinated Pt atoms through passivation by surface Au atoms. , Additionally, we also see a larger degree of Ir dissolution at the higher UPLs in comparison to Au, which could influence the comparative result, as shown in Figure S6. The total amount of metal ions lost to the bulk solution during a full AST protocol (10,000 cycles) is also quantified through inductively coupled plasma optical emission spectroscopy (ICP-OES) analysis of the electrolyte post-AST (Figure S7). After 10,000 AST potential cycles, both Pt and Ni losses are detected for all cases.

7.

(A) Cyclic voltammetry curves (top) and corresponding in situ Pt dissolution rates (bottom) for RF sputter-deposited 40 nm Pt thin film on glassy carbon (black) and partially covered with a Au overlayer with varying coverage (Θ_Au = 0.32 ML in red and Θ_Au = 0.43 ML in blue). (B) Comparison of Pt dissolution for all three films. (C) Pt dissolution rates for nanoporous nanoparticles (np-NiPt (orange), np-NiPt + Ir (green), and np-NiPt + Au (blue)) during triangular wave potential cycling for three upper potential limits 1.0, 1.2, and 1.5 V vs RHE. (D) Higher-resolution view of Pt dissolution rates with a cycling upper potential limit of 1.2 V vs RHE. All experiments were performed in 0.1 M HClO₄ and a voltammetry sweep rate of 50 mV s^–1 at room temperature.

Surface and subsurface Ni lost during AST are mainly due to Pt transport (dissolution or surface diffusion) and exposure of subsurface Ni as the structure coarsens. In the presence of partial monolayers of both Ir and Au, the total loss of Ni and Pt to the solution decreases. This is not unexpected as both Au and Ir slow coarsening, which reduces surface mobility-driven exposure of underlying Ni, limiting its dissolution. The tangible impact of subsurface Ni loss is a decay in the intrinsic activity of the catalyst (Figure S8). In the presence of Ir dopants, there is no loss in specific activity (SA) as the Ir dopant acts to retain subsurface Ni and its impact on the ORR activity. The SA for np-NiPt + Ir even improves over the course of the ORR, likely due to either surface alloying of Ir with Pt or redistribution of Ir on the surface, increasing exposure of the more reactive Pt. The opposite is observed for the Au dopant, Figures S5 and S8. As the NAC coarsens during the AST, Au will remain on the surface as the UPL is below the potential for significant Au oxidation/dissolution, see Figure S6. The result will be an increasing surface coverage of Au as the particle surface area decreases, blocking an increasing fraction of the surface with inactive Au and leading to reduced ORR activity. Similar effects have been observed with Au coatings on Pt thin films, where ORR performance significantly reduced beyond a Au surface coverage of 0.2 ML and in PtNiAu alloy nanoparticles beyond a Au content of 3 at. %. The increase in the onset potential for CO stripping, Figure S5, following the AST protocol is also indicative of a surface smoothening and passivation by inactive species, i.e., Au. High-resolution TEM with EDS mapping, Figures S9 and S10, of the beginning of life (BOL) and end of life (EOL) shows significant aggregation of Au for np-NiPt + Au and a moderate loss of Ir for np-NiPt + Ir. These in situ ICP-MS results provide further support for the link between electrochemical coarsening and Pt dissolution/redeposition.

Conclusions

For the first time, we effectively deconvoluted the distinct surface diffusion and dissolution atomic processes that govern electrochemical coarsening. Contrary to current strategies, future catalysts need to be designed to inhibit both Pt dissolution and mass transport across multiple length scales (from within the electrode structure to within the nanoporous particle itself). In this regard, a new challenge emerges: balancing a high ECSA without negatively impacting the intrinsic activity. Au and Ir are used here for demonstrative and investigative purposes, and their use in commercial catalyst materials must be judiciously evaluated by considering their cost and the negative impact on ORR activity. The most promising dopants are likely elements that are known to form stable, inert oxide species at high UPLs (e.g., Ti or Ta); these elements could be beneficial from a surface-mobility perspective (as compounds with high cohesive energies will have sluggish diffusion), but first-principles calculations would be necessary to screen their effect on the electrochemical stability of adjacent Pt species. In addition, these elements are challenging to electrodeposit at room temperature in aqueous electrolytes and would thus require specialized approaches in bespoke ionic liquids. Ultimately, the development and integration of these morphologically stable NACs will yield significant improvements in both the precious metal utilization and operational longevity for electrochemical energy conversion and storage devices.

Methods

Nanoparticle and Catalyst Synthesis

The precursor Ni₈₀Pt₂₀ alloy nanoparticles were synthesized through an organic solvothermal reduction method. , Ni­(acac)₂ (0.80 mmol), oleylamine (4 mL), and 1,2-tetradecadeniol (TDD, 0.5 mmol) were initially introduced into 10 mL of diphenyl ether (DPE) at 100 °C and then heated to 180 °C under an Ar atmosphere. Following several vacuum and Ar purging cycles, a solution of 0.20 mmol Pt­(acac)₂ and 3 mmol adamantanecarboxylic acid (ACA) in 3 mL of dichlorobenzene (DCB) was injected into the stirring solution at 180 °C, and then the solution was heated to 225 °C. After the temperature remained for 1 h at 225 °C, the solution was cooled to room temperature under flowing Ar. The formed nanoparticles were centrifuged at 8000 rpm, washed with hexane/ethanol, and finally deposited onto carbon support (Vulcan XC-72) through a colloidal deposition process. After centrifugation and three washing cycles with hexane/ethanol, the as-made catalyst was annealed in a tube furnace at 180 °C in air for 1 h, followed by 400 °C in H₂/Ar for 2 h. The metal loading, determined through thermogravimetric analysis (TGA), was found to be 20 wt % metal on carbon for the as-synthesized Ni₈₀Pt₂₀ alloy particles and 13 wt % metal on carbon for the dealloyed np-NiPt/C catalyst.

Dealloying and Electrochemical Measurements

The as-annealed nanoparticle catalysts were dealloyed, electrochemically characterized, aged, and assessed for oxygen reduction reaction (ORR) activity in a three-electrode cell using a rotating disk electrode (RDE) setup from Pine Instruments controlled by a Metrohm Autolab potentiostat (PGSTAT302N). The counter electrode was Pt mesh (99.9%, Alfa Aesar) bonded to the end of a Pt wire (99.9%, Alfa Aesar). The Ag/AgCl (BASi) reference electrode was calibrated against a hydrogen reference and found to have an offset of 0.27 V at 25 °C for 0.1 M HClO₄. All potentials listed are referenced to the reversible hydrogen electrode (RHE). Prior to any electrochemical experiments, all glassware was cleaned by soaking in a solution of concentrated 1:1 H₂SO₄/HNO₃ for at least 8 h, followed by rinsing and boiling in Millipore (Milli-Q Synthesis A10) water for 3 times.

The thin-film catalyst layers, with 15 μg_Pt cm^–2 loading, were formed on glassy carbon (GC) disks (5 mm diameter, 0.196 cm²) by drop casting from a catalyst ink and drying under a flow of Ar. The catalyst ink was composed of a 4:1 H₂O/IPA volume ratio solvent solution with a solid catalyst concentration of 1 mg_catalyst mL^–1. Additionally, 0.5 μL of Nafion 5 wt % solution (Ion Power LQ-1105 1100 EW) per milligram of catalyst was added to the ink (ionomer/carbon mass ratio ≈1:37.2) to aid in dispersion and adhesion of the catalyst particles to the GC substrate.

Electrochemical dealloying of catalysts, to form np-Ni₃₀Pt₇₀, was accomplished in Ar-purged 0.1 M HClO₄ in a standard three-electrode electrochemical cell by cycling the potential between 0.05 and 1.2 V vs RHE at 250 mV s^–1 for 65 cycles. The accelerated stability test (AST) consisted of 10,000 triangular wave potential cycles between two specified potential limits in 200 mL of Ar-purged 0.1 M HClO₄ with a sweep rate of 50 mV s^–1 at room temperature. The starting composition of np-NiPt is confirmed by inductively coupled plasma atomic emission spectroscopy (ICP-OES) and energy-dispersive X-ray spectroscopy (EDS) analysis. The evolution of the catalyst ECSA was determined through the integration of the current in the hydrogen underpotential deposition (H _UPD) region of the CV curves and the current in the CO oxidation region of the CO stripping curve following the procedure outlined by van der Vliet et al. For the ORR activity measurement, the dealloyed catalyst, either before or after AST, was transferred to a three-electrode electrochemical cell containing fresh O₂-saturated 0.1 M HClO₄ at 25 °C. Anodic ORR polarization curves were recorded at 1600 rpm while running linear sweep voltammetry between 0.1 and 1.1 V vs RHE at 20 mV s^–1. Experiments were repeated more than three times in order to confirm that the results were repeatable. Kinetic current densities for ORR were calculated using the Koutecky–Levich equation to adjust for mass transport limitations:

$$\frac{1}{i}=\frac{1}{i_{\mathrm{k}}}+\frac{1}{i_{\mathrm{d}}}$$ 1

where i is the measured current density, i _d is the diffusion-limited current density, and i _k is the kinetic current density. Specific activities were obtained through normalizing the i _k at 0.9 V versus RHE by Pt surface area. All potentials are corrected for iR drop within the electrochemical cell.

Partial Monolayer of Ir/Au Deposition

Partial monolayers (ML) of Ir and Au were deposited on the surface of the np-NiPt nanoparticles through the surface-limited redox replacement. , First, a partial ML of Cu was underpotentially deposited on the surface of the nanoporous nanoparticles in a 1 mM CuSO₄ + 0.1 M H₂SO₄ solution at a constant potential of 0.30 V vs RHE. After formation of the partial Cu ML, the catalyst layer was immersed in a solution of 0.025 mM IrCl₃ or AuCl₃ at open-circuit potential to drive the galvanic displacement of Cu with Ir or Au, respectively. After galvanic displacement of Cu, the Ir- or Au-doped nanoporous nanoparticles were washed with DI water prior to further testing.

Thermal Coarsening ECSA Measurement

A series of catalyst-coated GC disk electrodes was prepared using the same batch of catalyst ink, and each electrode was assigned for only one time point. Prior to thermal coarsening, electrodes were transferred into the quartz tube, which was then purged with Ar (10 psi pressure) at room temperature for 15 min to ensure complete removal of oxygen. Subsequently, the tube was quickly placed inside the tube furnace already heated to a constant temperature of 450 °C. After a certain period, the tube was quickly taken out and transferred into an ice bath to let it cool down for 10 min while Ar was purging continuously. Finally, the electrodes were installed into the hanging meniscus RDE configuration for further electrochemical ECSA measurements. The evolution of the catalyst ECSA under thermal conditions as a function of time was determined through the integration of the current in the H _UPD region. Five electrodes for each time point were tested to confirm that the results were repeatable.

Electrolyte Stirring Setup for Electrochemical Coarsening

A magnetic stir bar with 1.27 cm length (Fisherbrand PTFE) was placed in a standard three-electrode electrochemical cell, which was on a magnetic-stirrer device (Fisherbrand). Standard AST was accomplished in 200 mL of Ar-purged 0.1 M HClO₄ with a stir bar rotating at 400 rpm.

Reynolds number in a stirred vessel is defined as

$$Re=\frac{\rho N{D}^2}{\mu }=\frac{N{D}^2}{\nu }$$ 2

where ρ is the density of the electrolyte, N is the rotational speed, D is the diameter of the agitator, and μ and ν are the dynamic viscosity and the kinematic viscosity of the electrolyte, respectively. In our case, N is 41.8879 rad/s, D is 1.27 cm, and ν is 0.00893 cm²/s in 0.1 M HClO₄. By taking account of these parameters, the Reynolds number can be approximately estimated to be 7565.62, which is higher than 4000, indicating a turbulent flow.

Morphological and Compositional Characterization

Transmission electron microscopy (TEM) (JEOL JEM-2100) and scanning TEM (STEM) (JEOL JEM-2100F with a Schottky source) were performed at 200 keV to visually characterize the microstructure of the nanoparticles. Scanning TEM (STEM) electron energy loss spectroscopy (EELS) (JEOL 2100F with a Quantum Gatan Imaging Filter) was used to measure Ni and Pt fractions and generate elemental maps. STEM analysis was conducted with a probe size of approximately 1 nm and a high-angle annular dark field (ADF) detector with inner and outer detection semiangles of 27 and 54 mrad, respectively. The Ni and Pt atomic maps in Figure of the main text were generated from Ni L-edge and Pt M-edge EELS measurements. The STEM EDS data was acquired with an Oxford X-Max^N 80T EDS system with an 80 mm² SSD. The presented X-ray spectra area was spatially averaged across individual nanoparticles. The amount of metal, Pt and Ni, transferred to the electrolyte during AST through catalyst dissolution was quantified with post-mortem inductively coupled plasma atomic emission spectroscopy (ICP-OES) (Thermo Scientific iCAP 7000) testing of the electrolyte. Identical location TEM (IL-TEM) was used to qualitatively track the change in nanoporous nanoparticle morphology as a function of cycle number during AST. A gold TEM grid (Pacific Grid Tech) with a carbon supportive film was used as both the working electrode for AST and the material support for TEM analysis.

In situ heating TEM experiments were performed within a JEOL 2100F microscope, equipped with a Schottky source and a high-resolution pole piece with Cs = 1.0 mm. The TEM was operated in bright-field imaging mode. Annealing was performed with two holders, the Gatan 626 hot stage (main text Figure A) and the DENSsolutions Lightning D9+ sample holder, using heating-only nanochips. High-frame-rate data was collected during annealing using a Gatan K2 IS direct detection camera, operated in IS mode.

In Situ ICP-MS Characterization

Thin metal films of Pt and Au were deposited by magnetron sputter deposition on a mirror-polished glassy carbon substrate (base vacuum 1 × 10^–9 Torr). The deposition rate was calibrated by using a quartz-crystal microbalance. The deposition rate of Pt film was set to 0.3 Å s^–1 for ∼30 s, creating a 5 nm average nanograin size. Deposition of the Au thin film was done at 0.75 Å s^–1, exposing the glassy carbon surface during approximately 5–10 min, making a ∼20–40 nm thick film. Simultaneous electrochemical and metal dissolution rate measurements were done by attaching a stationary probe to the RDE (SPRDE method), and the fraction of electrolyte pumped from the electrode surface was analyzed directly into an ICP-MS (PerkinElmer, NexION 350S). Plasma, auxiliary, and nebulization flow rates were 15.6 L min^–1, 1 L min^–1, and 1 L min^–1, respectively, and plasma RF power was set to 1600 W. Pt (195 a.m.u) and Au (197 a.m.u) ions were simultaneously measured in the ICP-MS at a 4 points per second total, while the working electrode was controlled by a Metrohm Autolab potentiostat (PGSTAT302N). The electrodes were rotating at 100 rpm, with a reproducible SPRDE collection efficiency at 0.25.

Operando measurement of Pt dissolution rates from the np-NiPt, np-NiPt + Ir, and np-NiPt + Au catalysts was performed with online ICP-MS coupled with a Teflon flow cell. During the ICP-MS test, the glassy carbon electrode with a catalyst loading of 0.612 mg/cm² was immersed into the flowing electrolyte, a Au wire was used as the counter electrode, and SCE was used as the reference electrode. The Ar-saturated 0.1 M HClO₄ solution was the electrolyte, with a flow rate of 0.085 mL/s. A CV protocol triggered the Pt dissolution at a sweep speed of 50 mV/s in the potential ranges of 0.0 ∼ 1.0 V, 0.0 ∼ 1.2 V, and 0.0 ∼ 1.5 V.

Kinetic Monte Carlo (kMC) Simulation, Kinetic Rate Equations, and Model Parameters

We used a lattice-based kMC algorithm, MESOSIM, to simulate coarsening in np-NiPt nanoparticles. Our model is based on the approach laid out in ref , which uses a first nearest neighbor, bond-breaking model to capture the morphological evolution physics during dealloying and subsequent coarsening occurring over experimental time scales.

The time evolution of the system is governed by the kMC algorithm with the following steps:

• (1)Tabulate all N possible transitions, each indexed by i and possessing a rate k _i.
• (2)Calculate the cumulative function K _i = ∑ _j = 1 k _j for i = i, ...N.
• (3)Select a random number ζ = (0,1].
• (4)Determine the event from K _i–1 < rK _N ≤ K _i , where K _N = ∑ _j = 1 k _j = k _total.
• (5)Update the simulation time by the time interval Δt = −ln­(ζ)/k _total.
• (6)Go to step 1.

Specifically, rates for surface diffusion and dissolution in our model are governed by the expressions k _diff = ν₁ exp­[−nE _B/k _B T] and k _diss = ν₂ exp­[(−nE _B – ϕ)/(k _B T)], where ν₁ and ν₂ are attempt frequencies, n is the number of first nearest neighbors, E _B is the bond energy, ϕ is the applied potential, T is the temperature, and k _B is the Boltzmann constant. These kinetic expressions have been shown to accurately model dissolution current versus potential as well as coarsening behavior in nanoporous metals. ,−

In previous kMC studies on nanoporous gold (formed via dealloying Ag out of AgAu alloys), the surface diffusion of gold and silver atoms was captured by using E _B = E _B = 0.15 eV, and dissolution of silver atoms was captured using E _B = 0.15 eV. In order to capture the mobility of Pt and Ir in this study, bond energies were scaled under the assumption that the activation energy for surface diffusion scales with the homologous temperature (within ±10%): E _B = 0.23 eV, E _B = 0.31 eV, E _B = 0.19 eV, and E _B = 0.19 eV. The exponential prefactor, ν₁ = 10¹² s^–1, is the surface Debye frequency for a Pt surface. , In order to capture the activation energy for dissolution/redeposition of Au, Pt, and Ir in this study, the bond energy was scaled by their standard reduction potentials: E _B = 0.19 eV, E _B = 0.154 eV, E _B = 0.22 eV, E _B = 0.172 eV, and E _B = 0.205 eV. The exponential prefactor ν₂ = 2 s^–1 was chosen based on the exchange current density for Pt oxidation, in close agreement with a recent study by Erlebacher et al. on the evolution of Pt surfaces during cyclic voltammetry.

Simulated Thermal Coarsening of Individual Nanoporous Nanoparticles

To study coarsening trends in np-NiPt nanoparticles, we first generated fully dense Ni₈₀Pt₂₀ nanoparticles, 80 atoms in diameter (∼22 nm in diameter). Simulations were initialized by placing atoms on a three-dimensional fcc lattice, and the type of atom (e.g., Pt or Ni) was assigned at random based on a weighted probability determined by the composition of the alloy (80 at. % Ni, 20 at. % Pt). Nanoporous particles were generated by dealloying Ni out of the fully dense particles under a constant potential of 1.2 eV at 300 K for a total of 10⁴ simulated seconds, or ∼10⁸ iterations. Figure S2 shows a simulated nanoparticle before and after dealloying; the nanoparticles have a slightly smaller diameter, ∼70–75 atoms (∼19 nm in diameter), in close agreement with the experimental np-NiPt particles. We generated 10 unique dealloyed nanoparticles, which were used as “seeds” for studying the impact of individual surface dopants on thermal and electrochemical coarsening. Dopants were introduced to the np-NiPt nanoparticles by depositing 3500 atoms (∼5 at. %) on the surface at locations with first nearest neighbor coordination numbers n ≤ 5; dopant deposition and density are visualized in Figure B of the main text.

A total of 30 porous nanoparticles (10 np-NiPt, 10 np-NiPt + Ir, and 10 np-NiPt + Au) underwent simulated thermal coarsening at 450 °C for 60 s. The dealloyed structure was output at regular intervals and analyzed using a hybrid meshing and fairing algorithm for topologically complex materials, outlined in ref . The hybrid method enables relevant structural features to be extracted from the simulations, such as the topological genus, interfacial shape distribution, surface area distribution, density of crystallographic facets, and surface defect density. For this study, we primarily focused on the surface area and facet distribution.

The data presented as bands in Figure are a combination of the total surface area (the lower bound in Figure B) and an estimate of the active surface area (the upper bound in Figure B). The total surface area is the raw surface area calculated from every surface atom. The active surface area is the total surface area normalized by the area fraction of terrace atoms on (111) facets. This normalization accounts for integrated charge variations from H _upd adsorption due to alloy chemistry and facet distribution.

Simulated Electrochemical Coarsening of Individual Nanoporous Nanoparticles

Simulated electrochemical coarsening was carried out on the same 30 nanoparticles described above. Several kMC studies have implemented multistep oxidation/reduction mechanisms to capture the atomistic behavior of a Pt surface in acid and basic electrolytes during electrochemical cycling. , These mechanisms are based on experimental LEED and STM observations, which reported a periodic length scale associated with roughening of Pt surfaces following Pt oxidation and reduction. , This morphology is associated with a kinetic competition between surface roughening (oxide reduction) and surface smoothing from mobile Pt and Pt oxide species. We defined a potential-dependent “swapping” mechanism that encompasses atomic movement resulting from Pt oxidation/reduction. When this roughening event is chosen, the Pt atom is dissolved and redeposited at an unoccupied first nearest neighbor location. For the experimental condition of convection via electrolyte stirring, the roughening mechanism is modified to redeposit the dissolved Pt atom at a random, unoccupied surface site on the np-NiPt nanoparticle. An illustration of these mechanisms is shown in Figure S1. The simulated AST cycling conditions were chosen to mimic the experimental study, sweeping from ϕ = 0.6 to 1.1 eV at a ramp rate of 50 meV/s; the attempt frequencies and bond energies used are defined above. ,,−

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c07842.

• Particle TEM images; compositional data; kMC simulation outputs; performance data; and ICP-MS analysis results (PDF)

The authors declare no competing financial interest.