Low‐Pt NiNC‐Supported PtNi Nanoalloy Oxygen Reduction Reaction Electrocatalysts—In Situ Tracking of the Atomic Alloying Process

Abstract We report and analyze a synthetic strategy toward low‐Pt platinum‐nickel (Pt‐Ni) alloy nanoparticle (NP) cathode catalysts for the catalytic electroreduction of molecular oxygen to water. The synthesis involves the pyrolysis and leaching of Ni‐organic polymers, subsequent Pt NP deposition, followed by thermal alloying, resulting in single Ni atom site (NiNC)‐supported PtNi alloy NPs at low Pt weight loadings of only 3–5 wt %. Despite low Pt weight loading, the catalysts exhibit more favorable Pt‐mass activities compared to conventional 20–30 wt % benchmark PtNi catalysts. Using in situ microscopic techniques, we track and unravel the key stages of the PtNi alloy formation process directly at the atomic scale. Surprisingly, we find that carbon‐encapsulated metallic Ni@C structures, rather than NiN x sites, act as the Ni source during alloy formation. Our materials concepts offer a pathway to further decrease the overall Pt content in hydrogen fuel cell cathodes.

The metal contents of the different catalysts were determined by inductively coupled plasma mass spectroscopy (ICP-MS) using a 715-ES-ICP analysis system (Varian). The samples were prepared by dissolving the catalysts powders in a mixture of 2 mL H2SO4, 2 mL HNO3 and 6 mL HCl. The solutions were heated from RT to 180 °C in 10 min using a Microwave Discover SP-D (CEM corporation), and the temperature was held at 180 °C for 20 min. After cooling down to RT, the solutions were filtered and diluted with ultrapure water to a known volume (50 mL). To estimate the concentration of the solution, 4 standard solutions of Pt and Ni with a known concentration were prepared. The concentrations of Pt were 0, 5, 10 and 20 mg/L, whereas the concentrations of Ni were 0, 10, 20 and 40 mg/L. The selected wavelengths for Pt measurement were 203. 646 Transmission electron microscopy (TEM) images were recorded on a FEI Tecnai G 2 20 S-TWIN with a LaB6 cathode operating with 200 kV acceleration voltage and a resolution limit of 0.24 nm. Samples were dispersed in ethanol with an ultrasonic horn (~15 min), drop casted on a Cu grid (400 mesh), and dried in air at 60 °C for 10 min.
X-ray photoelectron spectroscopy: XPS measurements were carried out in an ultrahigh vacuum (UHV) chamber at room temperature using a non-monochromatized X-ray source (Al Kα, 1486.6 eV) and a hemispherical analyzer (Phoibos 150, SPECS). The Casa XPS software was used for analyzing the obtained spectra. All spectra were aligned according to the C1s main peak which was assigned with a binding energy of 284.3 eV. The Ni/Pt atomic ratios were obtained from the area under the Ni 2p and Pt 4f peaks after normalization by their corresponding relative sensitivity factors (RSF) of 21.1 and 15.86, respectively. High resolution TEM: HRTEM was performed using a FEI Titan 80-300 TEM with a Cs corrector for the objective lens (CEOS GmbH). The microscope was operated at 300 kV.
Scanning transmission electron microscopy (STEM) was performed using a FEI Titan 80-200 ("ChemiSTEM") electron microscope with a Cs-probe corrector (CEOS GmbH) and an HAADF detector. The microscope was operated at 200 kV. In order to achieve "Z-Contrast" conditions, a probe semi-angle of 25 mrad and an inner collection semi-angle of the detector of 88 mrad were used. Compositional maps were obtained with energy-dispersive X-ray spectroscopy (EDX) using four large-solid-angle symmetrical Si drift detectors. For EDX elemental mapping, Pt L and Ni K peaks were used. The error of the EDX composition measurement for individual particles after a typical investigation of 15 minutes is about ±2 at.%.
In situ heating TEM study was performed at a FEI Tecnai G 2 20 S-TWIN transmission electron microscope with a LaB6 cathode operated at an accelerating voltage of 200 kV (ZELMI Centrum, Technical University of Berlin). For in situ TEM experiments, a heating holder (DENS solutions B.V.) was applied. The catalyst powder was first dispersed into isopropanol solution with ultrasonication. Then this catalyst solution was drop cast onto a MEMS chip. The precise control of temperature was realized by fourpoint measurement integrated on the chip. After complete drying under ambient condition, the heating chip was mounted on a TEM holder.

Electrochemical measurements
5 mg catalyst was added to a solution of 3.98 mL ultrapure water, 10 µL 5 wt% Nafion ionomer solution and 1 mL isopropanol. The suspension was then sonicated with an ultrasonic horn sonifier for 30-40 min while immersed in an ice water bath. An aliquot of the ink was drop casted by a pipette onto a glassy carbon (GC) rotating disk electrode (RDE, diameter: 5 mm, geometrical surf ace area: 0.196 cm 2 ). The film was dried at 60 °C for 8 min.
For electrochemical characterizations a conventional three electrode cell with a graphite rod as counter electrode, a mercury/mercury(I) sulfate electrode (MMS, Hg/Hg2SO4) in a Luggin capillary as reference electrode and a catalyst-coated GC-RDE as working electrode was used. MMS was calibrated regularly by a homemade reversible hydrogen electrode (RHE) using a polycrystalline Pt disk and continuous hydrogen bubbling. All potentials are referred to RHE. Freshly prepared 0.1 M HClO4 diluted from concentrated HClO4 (99.999% trace metal bases, Sigma Aldrich) with ultrapure water was used as electrolyte. The catalystcoated RDE was controlled by a rotator from Pine Research instrument. All measurements were performed with a BioLogic SP-200 potentiostat. Gases for purging of electrolyte were H2 (99.999%), N2 (99.999%) and O2 (99.998%).
To avoid bubbles to be trapped on the dry catalyst film, it was thoroughly wetted until an uniform thin water film could emer ged. Then the catalyst-coated RDE was immersed into the N2-saturated electrolyte under potential control at 0.05 VRHE. For conditioning of the catalyst film, 50 potential cycles from 0.05 to 0.925 VRHE at 100 mV/s were carried out. 3 potential cycles from 0.05 to 0.925 VRHE at 20 mV/s were performed subsequently.
Electrochemical impedance spectroscopy (EIS): the as-measured results were iR-corrected manually after measurements, where the value of R was determined by EIS technique. A potential of 0.5 VRHE was applied for 1 min, followed by an impedance measurement from 100,000 Hz to 10 Hz with an AC amplitude of 10 mV. The resistance R was then determined from the Im (Z) vs. Re (Z) plot.
ORR testing: For background correction, measurements at 1600 rpm were first carried out in N2-saturated electrolyte. A potential of 0.05 VRHE was applied for at least 30 s. Then LSV was measured between 0.05 and 1.0 VRHE at 20 mV/s and repeated once. The electrode was then raised above the electrolyte. After purging of pure oxygen through the electrolyte for 10 min, the electrode was reinserted under an applied potential of 0.05 VRHE, and another three LSV measurements at 20 mV/s and one LSV measurement at 5 mV/s were conducted. In the case of a noticeable decrease in diffusion limited currents, oxygen was bubbled in between the scans.
Analysis of ORR activity: First the potentials were iR-corrected. Then capacitance correction was applied. LSV curve performed in N2-saturated electrolyte was used as background and subtracted from the LSV curve performed in O2-saturated electrolyte at the same scan rate. ORR current (iORR, at 0.9 VRHE) and diffusion limiting current (iL, at 0.4 VRHE) were obtained from the iR-and capacitance-corrected LSV curve. Finally the kinetic current ik was calculated by Koutecky-Levich equation. The mass activity (MA) was calculated by dividing ik by the Pt mass loading on the electrode. And the specific activity (SA) was was calculated by dividing ik by the Pt area estimated from CO stripping method.
Stability test (RDE protocol): The stability test was performed after carrying out the activation, impedance measurement and activity measurements. The rotation was turned off, the electrode was raised from the electrolyte under potential control and N2 was bubbled for 10 min. Then the electrode was immersed and the potential was cycled between 0.6 and 0.925 VRHE at 100 mV/s for 10,000 cycles, while venting above the electrolyte with N2. After that the potential was cycled as described in "activation", but only 3 cycles at 100 mV/s and 3 cycles at 20 mV/s. Then impedance spectroscopy and ORR activity test after stability test were also carried out.
CO stripping: initial CO stripping measurement was performed after initial ORR test. After ORR test, N2 was bubbled for 10 min. The electrode was reinserted under an applied potential of 0.05 VRHE and the rotation was set to 400 rpm. Then CO was bubbled for another 1 min at the same rotation speed. The gas was switched back to N2 and the electrolyte was bubbled for another 10 min to remove excess non-adsorbed CO in the electrolyte. Then the rotation was turned off and three CV cycles were recorded at 50 mV/s between 0.05 VRHE and 1 VRHE, while venting N2 above the electrolyte.
Analysis of CO stripping: The CO stripping data were analyzed to calculate the CO-ECSA and the ratio of CO-ECSA/Hupd-ECSA. For the CO-ECSA, the positive scan of the second cycle was used as background and subtracted from the positive scan of the first cycle. Then the integral of the peak was calculated in the iR-corrected potential range from the crossing of the curves to the highest potential (~1 VRHE). The integral area was normalized by the scan rate and the Pt mass loading on the electrode. A value of 420 µC/cm 2 was used for the electrooxidation of a monolayer of adsorbed CO. The Hupd-ECSA after CO stripping was calculated using the cycle after CO stripping. Hupd-ECSA was determined by integrating the current in the negative going sweep. Due to the low content of Pt, it is quite hard to determine the accurate potential range of Hupd region. For consistency and better comparison, we selected 0.4 VRHE as the upper potential limit, whereas the lowest potential was chosen as lower potential limit (~0.05 VRHE). Subtraction of the capacitive background current was applied. The integral area was normalized by the scan rate and the Pt mass loading on the electrode. A value of 210 µC/cm 2 was used for the adsorption of a hydrogen monolayer.
CO stripping in 0.1 M KOH alkaline electrolyte: for CO stripping in alkaline electrolyte, electrochemical pre-treatments in 0.1 M HClO4 acidic electrolyte were first performed. The potential was cycled between 0.05 and 0.6 VRHE at 100 mV/s in deaerated 0.1 M HClO4 acidic electrolyte. Catalyst that underwent different potential cycles in acidic electrolyte were denoted as "UL600acid -XXC". UL600 means that the upper potential limit during cycling is 600 mVRHE, and XXC denotes the actual number of potential cycles. After electrochemical pre-treatment in acidic solution, the electrode was rinsed by ultrapure water. CO stripping was then performed following the same procedure as described in the section of "CO stripping", except that 0.1 M KOH alkaline electrolyte was used.
CV in 0.1 M KOH alkaline electrolyte: for CV test in alkaline electrolyte, electrochemical pre-treatments in 0.1 M HClO4 acidic electrolyte were first performed. The potential was cycled between 0.05 and 0.925 VRHE at 100 mV/s in deaerated 0.1 M HClO4 acidic electrolyte. Catalyst that underwent different potential cycles in acidic electrolyte were denoted as "UL925acid -XXC". Herein, UL925 means that the upper potential limit during cycling is 925 mVRHE, and XXC denotes the actual number of potential cycles. After electrochemical pre-treatment in acidic solution, the electrode was rinsed by ultrapure water. CV was performed in N2-saturated 0.1 M KOH alkaline electrolyte between 0.05 and 1.6 VRHE at 100 mV/s.  Note S1. Importance of the green and scalable synthetic route towards active cathode catalysts for the development of PEMFC technologies.
Although so many remarkable achievements have been achieved in the development of cathode ORR electrocatalysts, approaching to a commercially viable PEMFC technology that is feasible for mass production still requires some further improvements. [4] From a holistic perspective, energy-consuming, toxic and environmentally unfavorable synthetic procedures for the production of catalysts will largely undermine the advantages and attractiveness of this promising and green technology. [5] Focusing solely on the performance of Pt-based ORR catalysts, but disregarding both green chemistry and scalability of their synthetic routes will not suffice to realize the large-scale production of cathode catalysts, an indispensably key constituent at the core of automotive fuel cell. At a research level, several inconspicuous and usually overlooked shortcomings involved in the small-scale "test-tube" synthesis in lab will not be negligible in the large-scale industrial production, and are not minor issues anymore. These shortcomings often include the use of toxic, expensive and hazardous chemical reagents as metal precursors, capping agents, reducing agents and solvents when preparing Pt-based ORR catalysts. Besides, elaborate synthetic strategies, which are sensitive to various experimentally synthetic parameters, make the large-scale preparation even more difficult, if not impossible. [4c, 4d, 6] Consequently, designing a simple, green, robust and up-scalable synthetic route for the large-scale production of highly efficient Pt-based ORR catalyst is an important prerequisite for the genuinely commercial application of automotive fuel cells. As shown in Figures S2e and 20, although restricted by the limited volume of reaction flasks used in our lab synthesis, our PLDA method are still able to achieve the gram-scale synthesis in one single batch. Therefore, in terms of cost and green chemistry, our method shows great suitability for mass production.         regions, which neither were exposed to electron beam during in situ heating experiment. Herein, similar changes are observed as in main text: after heating, these Ni NPs marked by red arrows disappear completely. From this beam control experiment, we could conclude that all changes observed in our in situ heating experiment result from heating, rather than electron beam damage.
Note S2. Experimental verification of absence of beam damage effect (relating to Figure S10 for Pt/NiNC and Figure S18 for Pt/NiNC-1) In order to check for the influence of electron beam damage, we recorded TEM images of two other catalyst regions only briefly before and after completion of our in situ heating experiment ( Figure S10 for Pt/NiNC and Figure S18 for Pt/NiNC-1). As very similar morphological changes to Pt and Ni NPs were observed on the non-irradiated regions, we are confident that the diffusion and shrinkage of Ni atoms and subsequent alloy formation originate from the thermal treatment rather than from beam effects.

S11-19)
To verify the dominant role of metallic Ni NPs as the source of mobile Ni atoms alloying with Pt particles during the thermal formation of PtNi nanoalloy NPs, we prepared a reference material in analogy to Pt/NiNC (synthetic process see Figures S11, 12), denoted as "Pt/NiNC-1", which had a lower Ni content of only 1 wt% and contained no metallic Ni NPs present ( Figure S11b). Key difference compared to the synthesis of Figure 1 was the preparation and use of the lower Ni-wt% support, NiNC-1. Pt deposition on NiNC-1 and further thermal annealing of Pt/NiNC-1 was identical.
After thermal annealing of the Pt/NiNC-1 catalyst, the right shift of the Pt(111) reflection in Figure S13 suggested the formation of a bimetallic PtM alloy phase with a second metallic component "M". But due to the fact that both Zn (3.6 wt%) and Ni (1.2 wt%) coexist in the pyrolyzed "NiNC-1" support material, it is unclear which metal element took part in the alloy formation and at what proportions if both do. This is the reason why we denote the annealed catalyst as "PtMx/NiNC-1", in which M represents either Zn or Ni. Importantly, the voltammetric CO stripping and associated CV measurements in alkaline electrolyte (Figure S19a and b) unambiguously indicate the absence of "Ni"-type CO stripping peaks and Ni 2+ /Ni 3+ redox peaks. This implies that Ni cannot be involved in the formation of the PtMx alloy phase. Overall, this control experiment confirms that Ni NPs act as the main source of Ni species in the process of alloy formation. Owing to a mass loss caused by unstable volatile species during the annealing process, the contents of Pt and Ni in annealed samples are slightly increased.        other two regions, which neither were exposed to electron beam during in situ heating experiment.
Herein, similar changes are observed as in Figure S16. From this beam control experiment, we could conclude that all changes observed in our in situ heating experiment result from heating, rather than electron beam damage.       First of all, from a qualitative standpoint, due to the much lower intrinsic ORR activity of NiNx sites than Pt, it is expected that these ORR-inactive NiNC sites barely contribute to the overall activity of the Pt/NiNC catalyst. Previous DFT studies suggest that this poor ORR activity results from the weak O* adsorption energy on Ni site. [8] Then such a hypothesis is corroborated by our quantitative estimation result as shown below.
The geometric catalyst loading of Pt/NiNC on the GC electrode is And the geometric mass loading of accessible NiNx sites is Considering that most of Ni in NiNC support is Ni NPs, which are completely encapsulated by thick carbon layer, we prefer to use the Ni content derived from XPS technique, rather than that derived from ICP technique, as accessible Ni site, ωNi site. We would like to point to the difference in the distinct probing characteristics of both techniques: XPS is surface-sensitive with detection depth at specific photo energy, but ICP is a bulk method probing the total content of Ni species both from surface and buried under the thick carbon layer.

And the geometric molar amount of accessible NiNx site is
Where MNi is the molar mass of nickel.
Thus, the geometric kinetic current density from accessible NiNx sites is Where TOF is the turnover frequency, NA is the Avogadro's constant, e is the electric charge of one electron.
The overall kinetic current density, jK, was calculated according to Koutecky-Levich equation: Where j0.8 is the current density at 0.8 VRHE, and jD is the diffusion current density.
Finally, the percentage contribution of accessible NiNx sites to the overall activity can be calculated as: The the NiNC support, NiNC support could ensure very uniform coverage of the ionomer. [2a] Second, carbon supports with a mesoporous structure (with a pore opening of 4-7 nm) promise excellent ORR activity and transport properties simultaneously. [9] Third, the highly graphitized carbon is more resistant to carbon corrosion.
For Pt/NiNC catalyst, due to the difficulty in obtaining an uniform catalyst film without the addition of Nafion ionomer, we only measured its performance with the addition of Nafion ionomer. As shown in Figure 5a and d, for Pt/NiNC, albeit with ionomer, it still exhibits measurable CO signal and Hupd features, indicating its advantages over conventional Vulcan carbon.
To evaluate the effect of ionomer distribution on the electrochemical performance of Pt/Vul-C catalyst, we measure its ORR performance with and without the addition of Nafion ionomer. When ionomer is absent, Pt surface is free of ionomer and the poisoning effect is absent. As shown in Figure S26, upon the addition of Nafion ionomer, both CO signal and Hupd features of Pt/Vul-C were inhibited due to the poisoning effect on Pt surface. Carbon monoxide strongly adsorbs on metallic Pt surface. The weakening of CO signal demonstrated that a faction of Pt surface was fully covered by the thick ionomer layer and was not electrochemically-accessible anymore.
As shown in Figure S26b, both kinetic current at 0.9 VRHE and diffusion current became smaller when ionomer was added. In the diffusion limiting regime, Pt particles hidden under the thick ionomer layer exhibit restricted protonic and reactant gas access, thus showing lower diffusion current density at low potential. In the kinetic regime, sulfonate groups in the perfluorosulfonic acid (PFSA) ionomer could also poison Pt surface, thus lowering its intrinsic activity.   From XRD pattern as shown in Figure 2a and Figure S28, after ex situ thermal annealing under ambient pressure for 6 h, the PtNi/NiNC catalyst after chemical dealloying continued to show weak diffraction peaks corresponding to crystalline Ni NPs. With an even shorter heating protocol, the in situ thermal annealing in the TEM holder appears to induce more severe changes on our catalyst material compared with ex situ annealing. Therefore, we conclude that in situ heating condition is much harsher  [11] PtNi@Pt/C Cui et al.
(2013) [12] Activated PtNi We have compared the ORR performance with other reported PtNi bimetallic catalysts. As shown in Table S8, our catalysts in this present study, although not having the highest mass activity, still exhibit one of the highest reproducible ORR performances, which exceeds those of other existing PtNi bimetallic catalysts in terms of mass activities when normalized by the mass of Pt, validating the advantages of our strategy for fabricating high-performance Pt-based catalysts.

Supporting Videos
Video S1 Image sequences of Pt/NiNC-7 catalyst during in situ heating TEM.

Video S2
Image sequences of Pt/NiNC-1 catalyst during in situ heating TEM.