Chemical Structure of Fe-Ni Nanoparticles for Efficient Oxygen Evolution Reaction Electrocatalysis.

Bimetallic iron-nickel-based nanocatalysts are perhaps the most active for the oxygen evolution reaction (OER) in alkaline electrolytes. Recent developments in literature have suggested that the ratio of iron and nickel in Fe-Ni thin films plays an essential role in the performance and stability of the catalysts. In this work, the metallic ratio of iron to nickel was tested in alloy bimetallic nanoparticles. Similar to thin films, nanoparticles with iron-nickel atomic compositions where the atomic iron percentage is ≤50% outperformed nanoparticles with iron-nickel ratios of >50%. Nanoparticles of Fe20Ni80, Fe50Ni50, and Fe80Ni20 compositions were evaluated and demonstrated to have overpotentials of 313, 327,, and 364 mV, respectively, at a current density of 10 mA/cm2. While the Fe20Ni80 composition might be considered to have the best OER performance at low current densities, Fe50Ni50 was found to have the best current density performance at higher current densities, making this composition particularly relevant for electrolysis conditions. However, when stability was evaluated through chronoamperometry and chronopotentiometry, the Fe80Ni20 composition resulted in the lowest degradation rates of 2.9 μA/h and 17.2 μV/h, respectively. These results suggest that nanoparticles with higher iron and lower nickel content, such as the Fe80Ni20 composition, should be still taken into consideration while optimizing these bimetallic OER catalysts for overall electrocatalytic performance. Characterization by electron microscopy, diffraction, and X-ray spectroscopy provides detailed chemical and structural information on as-synthesized nanoparticle materials.

Introduction

The global en class="Chemical">conomy remains heavily reliant on fossil-fuel-based non-renewable sources of energy.[1] Fossil fuels are a major contributor to greenhouse gases, as the burning of fossil fuels leads to the release of carbon dioxide (CO2) in the atmosphere.[2] In 2013 alone, an estimated 9.78 billion metric tons of CO2 were released into the atmosphere because of the use of fossil fuels.[3] Having fuel alternatives to fossil fuels thus continues to be urgent as the environmental impacts brought on by global warming become increasingly severe.[4] Hydrogen as a fossil fuel replacement has the potential to reconfigure the energy sector’s use of fuel sources and usher in an era of clean energy.[5−7] However, currently, almost 95% of the estimated 55 million metric tons of hydrogen produced annually worldwide are done so using non-renewable sources (i.e., through coal or natural gas steam reforming).[7,8]

Electrolysis of water is considered to be a more envn class="Chemical">ironmentally sustainable approach to produce hydrogen (i.e., CO2-emissions free when coupled with renewable energy input).[8−10] Besides, electrolysis may also play a crucial role in the storage of energy from renewable sources such as sunlight and wind.[6,11,12] The splitting of water results from two half-reactions: the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER).[13] In an alkaline environment, the half reactions are as follows

The efficiency of water electrolysis depends on the efficiency of both HER and OER.[14] OER is the slower of the two half-reactions, as it is kineticn class="Chemical">ally hindered by a multi-step-proton-coupled-electron transfer mechanism.[15−17]

Electrocatalysts increase the kinetics of OER and lower the overpotentin class="Chemical">al (i.e., the applied potential required beyond the theoretical potential, E0 = 1.23 V). Rare transition metals such as RuO2 and IrO2 are considered the most advanced commercially available catalysts for OER. However, RuO2 and IrO2 are expensive and scarce precious metals, thus making these catalyst materials non-viable candidates for large-scale industrial manufacturing of electrolysis technology.[18,19] First-row late transition metals (e.g., Fe, Co, and Ni) have recently been shown to have excellent promise as alternatives to precious metals. These first-row transition metals are less costly, widely available, stable, and insoluble in alkaline electrolytes. Besides, these metals result in catalyst materials that are highly active for OER, where the metals typically exist as oxides or hydroxides.[20−22]

Recent studies have shown that monometallic n class="Chemical">oxide and hydroxide catalysts of iron, nickel, and cobalt are not as active as their sister bimetallic or trimetallic oxide and hydroxide catalysts.[23−25] Further, bimetallic catalysts at certain ratios have been shown to complement the drawbacks of monometallic catalysts. Corrigan found that by precipitating Fe (10–50 at. %) in a composite iron–nickel hydrous oxide, the Tafel slope was greatly lowered (by ∼65%), indicating faster reaction kinetics for OER.[26] The Tafel slope can also offer insight on the rate-limiting step, and the decrease in the Tafel slope can be associated with the surface adsorbed species formed during the initial phase of the OER remaining predominant.[27] Fe0.25Ni0.75OOH films resulted in an increase of OER activity by 500-fold over pure Fe and Ni films,[28] while Stevens et al. were able to show an increase in activity of 150-fold for 10% Fe-containing NiOH films.[29] Finally, Louie and Bell observed optimal OER activity for iron–nickel oxide thin films with 40 at. % iron composition.[30] Generally, measured overpotential also appears to be directly correlated to the FeNi atomic composition.[28,31] The lowest OER onset potential (i.e., overpotential) was observed by Steimecke et al. for Ni/Fe films containing 15% Fe, while Fidelsky et al. used density functional theory (DFT) calculations to theorize that Fe-doped NiOOH are capable of lowering the overpotential by more than 40% compared to pure NiOOH.[32,33] These studies demonstrate the importance of iron incorporation into nickel-based hydroxide catalysts and suggest that the ratio of iron to nickel plays a significant role in the eventual OER activity. Several mechanisms for the improvement in OER performance have been suggested. Fidelsky and Toroker suggested that a band-like charge mechanism for charge transport may be present in Fe-doped NiOOH materials, which positively influences both the catalytic performance and electronic conductivity.[33] Iron doping into NiOOH has also been suggested to enhance the probability of hydrogen transfer while simultaneously requiring less activation energy.[34] DFT calculations have shown that the addition of iron dopant increases the predicted OER activity by weakening the bonds of OER intermediates.[35]

In related work, Burke et al. demonstrated that n class="Chemical">Fe incorporation in CoOOH (Co1–FeOOH; x = 0.6–0.7) enhanced the OER activity by almost 100-fold and hypothesized that Fe is the active site for OER catalysis, while CoOOH affords chemical stability, conductivity, and electrolyte permeability for Fe active sites.[20] In the case of Ni(OH)2/NiOOH, Trotochaud et al. reported that Fe incorporation in Ni(OH)2/NiOOH increased the conductivity of the catalyst by more than 30-fold and indicated that Fe induces a partial-charge transfer mechanism that initiates Ni centers throughout the catalyst film, thereby increasing the OER activity of the catalyst.[31] Friebel et al. applied operando X-ray absorption spectroscopy (XAS), along with high energy resolution fluorescence detection and computational methods, to suggest that Fe3+ incorporated in Ni1–FeOOH increases the OER activity and that Fe sites, and not Ni sites, are the active catalytic sites in Ni1–FeOOH.[28] As the discussion of the nature of the active site of these Ni1–FeOOH materials continues in the literature, thus far, it appears clear that Ni and Fe play synergistic roles and that the active site is likely composed of a Ni–Fe–O coordinated chemical structure. Recently, Shin et al. revealed through DFT studies that Fe4+ and Ni4+ in Ni1–FeOOH function as cocatalysts for OER as high-spin d4 Fe4+ stabilizes the radical character on the O of M–O bond which subsequently facilitates O–O coupling on low-spin d6 Ni4+.[36] Surface metal sites (M) are largely considered to be where OER occurs, including facilitating the adsorption and reaction of a series of intermediates such as M–OH, M–O, M–OOH, and M–OO.[37] However, to further complicate understanding of Ni1–FeOOH materials, it was recently suggested by Doyle et al.[35] that Ni1–FeOOH materials may have a bulk contribution to the overall observed OER activity. This computational study was based on the more ordered β-Ni(OH)2, rather than the disordered alpha phase that is a closer match to the disordered Ni1–FeOOH materials,[31] but the results are compelling and require further inquiry.

The majority of studies thus far have focused on n class="Chemical">iron incorporated into homogeneous thin films. However, Burke et al. point to the need for high-surface-area catalysts to be developed to enable a scalable catalyst and electrode development with low mass transport limitations and high mass activity performance.[37] Görlin et al. synthesized nano-sized Ni–Fe catalysts using a microwave-assisted, surfactant-free solvothermal route and showed that Fe incorporation of around 50% resulted in the highest OER activity.[38] Further, Liu et al. managed to develop a nanostructured FeNi alloy where they incorporated S and N-doped carbon which drastically improved OER catalysis with a very low overpotential of 230 mV to achieve the current density of 10 mA cm–2.[39] Similarly, Liu et al. were also involved in the fabrication of FeNi3 nanoparticles incorporated on carbon doped with multiple nonmetal elements (FeNi3/M–C) that demonstrated an overpotential of 246 mV at 10 mA cm–2.[40] Meng et al. claimed that the excellent OER performance shown by a stereo film on carbon cloth comprising FeNi3 nanosheet covered FeOOH was due to factors such as the synergistic effects of iron and nickel ratio and the availability of exposed catalytic sites for OER derived from the oxidation of the FeNi3 nanosheets during the anodic oxidation.[41] Interestingly, Du et al. were able to engineer N-doped carbon-coated FeNiP nanoparticles, which turned out to be a good bifunctional catalyst for alkaline water electrolysis.[42] There is now an opportunity to develop nanostructured catalysts that have similar, or better, performance metrics for OER but are amenable to scalable electrode design.

In this work, results are presented for a suite of bimetallic n class="Chemical">iron–nickel (Fe–Ni) alloy nanoparticles that are active for OER in alkaline electrolytes. To explore the role of Fe content on OER activity, Fe–Ni alloy nanoparticles were synthesized at three different molar ratios of iron to nickel. The ratios studied are mole Fe/mole Ni—1:4, 1:1, and 4:1 and will be denoted as Fe20Ni80, Fe50Ni50, and Fe80Ni20. The synthesis of these Fe–Ni bimetallic nanoparticles involved a series of precisely-timed synthesis steps in aqueous solution. Fe–Ni nanoparticles were characterized subsequently using energy-dispersive X-ray spectroscopy (EDX), inductively coupled plasma mass spectrometry (ICP–MS), X-ray diffraction (XRD), and X-ray photoelectron spectroscopy (XPS). Detailed characterization was performed via synchrotron-based hard XAS. Transmission electron microscopy (TEM) was used for imaging morphology and diffraction analysis. From characterization results, we report new findings on the chemical and structural nature of these complex alloy nanoparticle materials. The electrochemical analysis was conducted on the Fe–Ni bimetallic catalysts using cyclic voltammetry (CV), and stability testing was performed using chronoamperometry (CA) and chronopotentiometry (CP). The results from this work demonstrate that the ratio of iron to nickel plays a critical role in the OER activity of synthesized nanoparticles, similar to results reported for the thin films work outlined above. Activity and stability, however, must both be evaluated before a catalyst design direction is determined. Furthermore, the results demonstrate that highly active nanocatalysts can be synthesized by a scalable synthesis process, with the potential for scale-up and industrial production of the Fe–Ni bimetallic catalysts at gram to kilogram levels.

Results and Discussion

Nanoparticle morphology is shown in Figure . The images suggest that the nanoparticles of n class="Chemical">all three ratios tend to agglomerate together when deposited onto the TEM grid; from these results, it may be likely that the nanoparticles also similarly agglomerate on the working electrode surface. The nanoparticles appear to be roughly spherical but structurally disorganized and heterogeneous. The nanoparticles of all three ratios appear to have a mixture of both crystalline and amorphous regions. Fe20Ni80 and Fe50Ni50 may be more crystalline, compared to Fe80Ni20, as a higher content of lattice planes is observed. Fe80Ni20 had a d-spacing value of 0.21 ± 0.02 nm which was indexed to the lattice plane (111) of FeNi alloy by Li et al.[43] Fe50Ni50 had a d-spacing value of 0.20 ± 0.01 nm, and Fe20Ni80 had a d-spacing value of 0.20 ± 0.02 nm, which Ding et al. described to be approximately close to (110) planes of bcc FeNi alloy while Xia et al. indexed the d-spacing value to (111) planes of face-centered cubic (fcc) γ-FeNi phase.[43−45] For all three samples, the nanoparticles have an apparent core–shell-like morphology (additional images presented in Figure S1 of the Supporting Information), but instead of one core being enclosed by a shell, multiple cores are surrounded by the same shell. From these results, we postulate that the synthesis approach causes a mixed metallic phase to form, with regions of crystallinity, where the increasing iron content causes an expansion of the lattice structure, with possible shifts in the phase present and insertion of iron into the nickel lattice likely.[31] During synthesis, the nanoparticles are exposed to water, and once synthesized, the nanoparticles are also exposed to the ambient environment. Both iron and nickel are expected to oxidize upon contact with water and oxygen, and the outer shell visible as a lighter phase contrast region in TEM imaging is likely an oxidized layer that forms on the bimetallic core nanoparticles. Line scans and point measurements of nanoparticles (Figure S1) identify both iron and nickel in the shell and support this conclusion.

Figure 1

TEM and HRTEM images of synthesized alloy nanoparticles: (a,d) Fe20Ni80, (b,e) Fe50Ni50, and (c,f) Fe80Ni20.

TEM and HRTEM images of synthesized alloy nanoparticles: (a,d) n class="Chemical">Fe20Ni80, (b,e) Fe50Ni50, and (c,f) Fe80Ni20.

All three ratios of n class="Chemical">iron–nickel nanoparticles appear to have a relatively similar size, thus indicating that the ratio between the two metals in the bimetallic composition does not control the overall size of the nanoparticles during the synthesis process. Often, the ratio between the metal precursors and the stabilizers used [e.g., amino tris(methylene phosphonic acid) (ATMP), polyvinylpyrrolidone (PVP)] plays a bigger role in controlling the particle size of nanoparticles synthesized through solution-phase chemistry techniques, such as our aqueous-based chemical reduction technique.[46]

Because of the n class="Chemical">complexity of the nanoparticle morphology, a suite of characterization tools was used to investigate elemental composition (summarized in Table ). The purpose of the suite of measurements was to determine possible differences between the observed core and shell structures, as well as to compare methods that provide bulk versus surface-sensitive measurements. Scanning electron microscopy (SEM) EDX was used to obtain elemental information, where the spectra represent an overall average measurement from a large population of nanoparticles. TEM EDX point measurements were made on individual nanoparticles for both the observed darker core region and the observed lighter phase contrast shell encapsulating the cores. While EDX, in general, is a highly useful characterization tool, all EDX measurements are at most semi-quantitative and thus should be used to compare within a sample set and should always be compared to more quantitative measurements. Both ICP–MS and XPS characterization were used to compare to EDX measurements. ICP–MS is a precise and quantitative measurement tool, but is a bulk measurement of the metals in the sample and cannot measure oxygen content. Finally, XPS is surface-sensitive and measures a large particle population, but is considered to be a more precise quantification of composition within a comparative sample set, as compared to EDX. XRD was also used to characterize the samples, and the result is presented in the Supporting Information. Thus, the combination of these techniques can provide an overall understanding of the nanoparticle bulk and surface composition, the relative amount of oxygen in the sample (which can include contributions from ligand stabilizers), the atomic ratio of iron to nickel, and any discrepancies between individual versus populations of particles.

Table 1

Comparison of Elemental Composition Analysis Results from Bulk and Surface-Sensitive Techniquesa

	SEM EDX atomic %			TEM EDX atomic %			TEM EDX atomic %			ICP atomic %		XPS atomic %
	multiple particle average			individual particle core			individual particle shell			bulk sample average		surface sample average
sample	Fe	Ni	O	Fe	Ni	O	Fe	Ni	O	Fe	Ni	Fe	Ni	O
Fe20Ni80	9	42	49	9	38	53	4	21	75	18	82	16	44	40
Fe50Ni50	25	25	50	18	19	63	9	11	80	50	50	25	29	46
Fe80Ni20	45	9	46	38	5	57	11	1	88	83	17	40	9	51

All ratios are given as Fe/Ni.

All ratios are given as n class="Chemical">Fe/Ni.

The elemental ann class="Chemical">alysis from SEM EDX of the three ratios of iron–nickel nanoparticles is summarized in Table , for the relative amounts of iron and nickel in each sample. In the data shown in Table , iron and nickel atomic % are reported out of 100%; other minor components (e.g., salt contaminants from synthesis such as sodium and chloride) and oxygen content are not reported. As a semi-quantitative measurement, EDX is primarily used herein to compare relative ratios of Fe/Ni between samples synthesized in this work. Our EDX measurements during SEM imaging verify that the ratios measured in the synthesized nanoparticle samples match the theoretical target ratios of precursors used in the synthesis procedure. The Fe/Ni ratios reported in Table are below the expected EDX margin of error of 10%, compared to the theoretical target molar ratios (i.e., 5:1, 1:1, and 1:5) for all three nanoparticle samples. EDX measurements thus corroborate the initial calculations used for the synthesis of different molar ratios of nanoparticles and demonstrate that the samples tested contained the target Fe concentrations. A more detailed table of SEM EDX for both atomic and weight % is reported in the (Supporting Information).

TEM EDX analysis was performed as a series of point measurements. TEM dark field images of the point measurement locations are reported in the Supporting Information, and the elementn class="Chemical">al composition data are summarized in Table . Several interesting observations can be made from these data, as well as when these data are compared to SEM EDX data. First, there are measurable differences in the oxygen content of the core versus the shell, where the shell of all three samples has significantly higher oxygen atomic % than the core. These results are consistent with the observable phase contrast between the core and shell for all three samples in Figures and S1, where a darker phase contrast typically indicates a larger metal content. Second, the Fe50Ni50 composition does not result in measurable differences in the iron to nickel ratio in the core versus the shell, whereas Fe20Ni80 has a slightly higher content of nickel in the shell than the core. Interestingly, the iron content in the shell compared to the core for Fe80Ni20 was much larger, and both measured ratios of Fe/Ni were greater than the bulk ratios obtained from SEM EDX and ICP, as well as greater than the surface ratio of Fe/Ni as measured by XPS. This discrepancy may indicate particle-to-particle variability. Finally, the measured iron to nickel ratios are, overall, generally consistent with the target theoretical ratios intended during synthesis. As mentioned above, the aqueous synthesis method does allow some exposure to oxygen as well as direct exposure to water molecules during synthesis, and the synthesized nanoparticles are also exposed to the ambient atmosphere after synthesis. Our characterization data presented in Table further support the initial conclusion based on TEM imaging that exposure to oxygen and water likely enables the formation of a more oxygen-rich shell, with associated oxide or hydroxide formation as well as metal migration and restructuring. We do not currently understand when the shell structure forms or exactly how the morphology changes during or after synthesis. However, the result is a complex and heterogeneous morphology with direct implications for OER performance and catalyst stability.

Analysis of the data n class="Chemical">collected by ICP–MS is used to quantitatively account for the metal compositions (iron and nickel) of the samples. The raw data provided by the instrument gave counts per second for iron and nickel, which was then converted into concentration at parts per billion (ppb) (reported in Supporting Information). The concentrations in μM for iron and nickel were obtained by dividing concentrations in ppb by the molecular weight of their respective isotopes (57 Fe for iron and 60 Ni for nickel). The molar ratio of Fe/Ni was obtained by dividing the concentration of iron in μM by that of nickel in μM. The molar ratios of Fe20Ni80, Fe50Ni50, and Fe80Ni20 obtained from ICP–MS were below a margin of error of 10% in comparison to the targeted theoretical molar ratios during synthesis for all three nanoparticle samples. Also, the ratios measured by ICP–MS are the same as those measured by SEM EDX. Along with EDX results, ICP–MS results thus provide an additional verification that the composition of iron and nickel in all three nanoparticle samples was reflective of the desired molar ratios from a bulk particle perspective.

Surface elemental n class="Chemical">composition of the nanoparticle samples was determined by XPS, which is a highly surface-sensitive tool that analyzes to a depth of <10 nm of a surface.[47] For the nanoparticle morphology and size scale shown in Figure , XPS measurements likely probed both the surface oxide shell and the core. For a nanoparticle sample, XPS measures a population of nanoparticles in a nanoparticle film, and one might expect some variability in nanoparticle packing with resulting variability in the microscopic topology of the film. This variability can result in variability in the actual depth of the XPS measurement into the sample, and thus, our results may represent an average of surface and subsurface composition. Survey spectra lines (reported in Supporting Information) were analyzed to calculate relative elemental concentration for Fe, Ni, O, C, and N. All trace elements present in the samples were discarded from the summation presented in Table , but are reported in the Supporting Information. All three nanoparticle samples contained a substantial amount of oxygen, as was also suggested by the TEM EDX results. However, the oxygen content estimated by XPS measurements is less relative to iron and nickel, as compared to TEM EDX point measurements of the shell regions and is rather more similar to the oxygen content measured for the cores of the nanoparticles. This result could suggest that there may be a potential overestimate in relative oxygen content by the TEM EDX measurements and also that there is reduced metal content in the subsurface of the nanoparticles, which is not surprising given that the synthesis technique relies on the chemical reduction of both metal cation precursors by borohydride to form particles. However, we propose that the TEM EDX measurements are not an overestimate and that the XPS measurements are, in this case, not solely measuring the surface of the nanoparticles. Given that XPS will probe approximately 3–5 nm into the surface of the nanoparticles, and the shell morphologies observed in Figures and S1 are on the order of ∼2 nm, it is likely that the XPS data are reflective of a combined surface and subsurface chemical composition, while the TEM EDX data of the shell are more reflective of just the shell composition. If we make this assumption, we can evaluate the metal to oxygen ratio and begin to develop a picture of the surface chemistry. For the Fe20Ni80 composition, the ratio of metal to oxygen is 25–75, while for the Fe50Ni50 composition, the ratio of metal to oxygen is 20–80 and for Fe80Ni20, the ratio is 12–88. The two compositions with lower Fe content thus have a metal (M) to oxygen (O) ratio, M/O, of ∼1:3–∼1:4. These ratios are higher than what would be expected for a nominal oxide or hydroxide based on an iron-incorporated nickel host phase, but might suggest that the phase is more likely to be similar to an iron-incorporated nickel hydroxide [FeNi100–(OH)2] rather than a nickel oxide (NiO) or an iron oxide (e.g., Fe2O3). Because we believe that the shell measurement for the Fe80Ni20 composition is slightly off based on our other measurements, we assume at this point in the analysis that the higher iron content composition has a similar initial surface phase as the other as-synthesized nanoparticle compositions. Overall, these results likely point to some extent of overestimation of oxygen content by TEM EDX, where the analysis of low atomic weight elements such as oxygen is less reliable than the analysis of metal species. The combination of our TEM EDX and XPS results can be used as complementary techniques to probe the surface, subsurface, and bulk chemical character of as-synthesized nanoparticles. The results also illustrate the importance of using and understanding multiple characterization tools to characterize morphologically complex nanomaterials.

The oxygen present in the surface of nanoparticle samples can thus likely be attributed to mn class="Gene">etal hydroxides, with a minor contribution from the oxygen atoms attached to the organic ligands (ATMP and PVP) that were used as stabilizers during nanoparticle synthesis.[48] It is not expected that there would be a significant contribution from ambient organic contamination to the oxygen signal because the samples were cleaned and stored in methanol until prepped for XPS measurements. The amount of carbon is approximately equal in the three nanoparticle samples, while nitrogen accounts for a smaller percentage. Both carbon and nitrogen can be traced back to the use of organic compounds (ATMP and PVP) that could not be completely rinsed off after synthesis. The XPS-measured iron and nickel content in Fe80Ni20 and Fe50Ni50 match relatively closely with the SEM EDX results (Table ) for the same sample, while the amount of nickel in Fe20Ni80 is 40% lower in comparison to the SEM EDX results.

High-resolution XPS was performed on both iron and n class="Chemical">nickel regions to further understand the oxidation state of the metal species in the nanoparticles (Figure ). The Fe 2p region for the three nanoparticles shows a shift in Fe 2p3/2 peaks, but the Fe 2p1/2 peak positions are similar for all three samples. The Fe 2p3/2 peak was observed at 710.8 eV for Fe80Ni20 (Figure a), which can be attributed to both iron oxide and hydroxide species. For Fe50Ni50, the Fe 2p3/2 peak (Figure c) was at 711.4 eV, which suggests that iron is likely present as iron hydroxide in either the alpha or gamma phase. Similarly, Fe20Ni80 has a Fe 2p3/2 peak at 711.8 eV (Figure e), which also points toward the presence of an alpha or gamma hydroxide species. The 2p1/2 peak positions for Fe80Ni20, Fe50Ni50, and Fe20Ni80 were located at 724.2, 723.9,, and 724.0 eV, respectively, which collectively can be attributed to both iron oxide and hydroxide species.[49−51] For Fe80Ni20, there appears to be a small peak around 707 eV which indicates the presence of some iron metal species at an oxidation state of 0. Neither Fe50Ni50 nor Fe20Ni80 composition had a distinct iron metal peak, but both spectra have a broadening of the Fe 2p3/2 peak between 708 and 703 eV, suggesting that there is a metallic contribution to the chemical environment of the iron atoms. The lack of a distinct metal peak may result from the inherent variability of the nanoparticle film and the probing depth of the XPS technique, which includes both the shell and a portion of the bulk underneath the visible shell. The nanoparticles are also exposed to the atmospheric conditions which could result in oxidized surfaces, and because XPS is a surface-sensitive technique, the spectra could not account for the metallic region that might be present underneath the oxidized surfaces. Because the lack of a distinct metal peak occurred in the two nanoparticles with higher nickel content, Fe50Ni50 and Fe20Ni80, the iron metal signal could have been suppressed by the greater amount of nickel that is present in the surface of the two nanoparticle films.

Figure 2

XPS spectra of (a) Fe 2p for Fe20Ni80, (b) Ni 2p Fe20Ni80, (c) Fe 2p for Fe50Ni50, (d) Ni 2p for Fe50Ni50, (e) Fe 2p for Fe80Ni20, and (f) Ni 2p for Fe80Ni20.

XPS spectra of (a) Fe 2p for n class="Chemical">Fe20Ni80, (b) Ni 2p Fe20Ni80, (c) Fe 2p for Fe50Ni50, (d) Ni 2p for Fe50Ni50, (e) Fe 2p for Fe80Ni20, and (f) Ni 2p for Fe80Ni20.

In the Ni 2p region, n class="Chemical">all three nanoparticles had bulk metal content, which supports the compositional analysis discussed above. Unlike the iron spectra, the nickel spectra all display a sharp, distinct peak for the metallic nickel. The peaks at 852.2, 852.3,, and 852.0 eV for Fe80Ni20, Fe50Ni50, and Fe20Ni80, respectively, can be classified as nickel metal peaks. The Ni 2p3/2 peak at 855.6 eV was observed for Fe80Ni20, but the Ni 2p1/2 peak was not distinguishable from the noise. Ni 2p3/2 peaks for Fe50Ni50 and Fe20Ni80 were observed at 855.6 and 855.4 eV, respectively, and the nearly identical peak locations suggest that the nickel chemical environment is similar in all three samples. The Ni 2p3/2 peaks for all three nanoparticles suggest the presence of nickel hydroxide. The Ni 2p1/2 peak for Fe50Ni50 was located at 873.0 eV and similarly for Fe20Ni80 at 873.2 eV, which suggests the presence of nickel oxide.[25,48,52,53] The peak positions for both iron and nickel spectra for all three nanoparticles are within a similar range of one another which suggests that the speciation and oxidation state are similar between the three samples. With the XPS results for the three nanoparticles, a case can be made for iron to be primarily in a +3 oxidation state, with a minor contribution from Fe0, whereas nickel is primarily in a +2 oxidation state, with some of the nickel content in the 0 oxidation state.

To further understand the chemistry of the nanoparticles, XAS was performed. Unlike XPS, which is a surface characterization tool, XAS is a bulk characterization technique. XAS data can be classified into three different regions: pre-edge region, X-ray near edge spectrosn class="Chemical">copy (XANES), and extended X-ray absorption fine structure (EXAFS). XANES data are reported in Figure . XANES data for the iron K edge region show that all three samples qualitatively appear to be similar with only minor differences in the spectra. The Fe K edge spectra for Fe80Ni20 and Fe50Ni50 virtually overlap on top of each other. The Fe K edge of Fe20Ni80 has higher intensity at the main absorption peak and slightly lower intensity at the pre-edge, compared to the Fe80Ni20 and Fe50Ni50. Out of the three standards shown, only the pre-edge feature of metallic iron is slightly similar to that of the three nanoparticle samples, while the other two standards have much more distinct pre-edge features (additional standards are reported in the Supporting Information). Generally, the weak pre-edge feature, which corresponds to 1s → 3d electronic transitions, can be seen in both the iron and nickel regions for the three nanoparticle samples and is potentially indicative of octahedral coordination,[54−57] but a more in-depth analysis suggests, particularly for iron, that the XAS measurement is a combination of multiple iron chemical environments and, potentially, multiple coordination environments.

Figure 3

XANES region of the (a) Fe K-edge and the (b) Ni K-edge.

XANES region of the (a) n class="Chemical">Fe K-edge and the (b) Ni K-edge.

For iron specificn class="Chemical">ally, the pre-edge shape and position suggest contributions from metallic iron and Fe3+, where the position matches the pre-edge peak position of the Fe(OH)3 reference, but the shape of the pre-edge is more reflective of a lower-intensity iron metal pre-edge. The Fe metal reference material was an iron foil, with the bulk bcc crystal structure and iron atoms in a tetrahedral geometry. The ferric hydroxide reference, also known as goethite, had a closed packed hexagonal structure, where iron atoms occupy octahedral positions within the hydroxide structure. The nickel ferrite (NiFe2O4) reference is known to have an inverse spinel fcc crystal structure, where the divalent nickel occupies only octahedral sites, while trivalent iron equally occupies both octahedral and tetrahedral interstitial sites within the crystal structure.[58] In addition to the oxidation state, the pre-edge feature can also provide information about the metal coordination geometry[59] of a material. Comparison to these reference spectra pre-edge peak shape and positions suggests that iron atoms that are in the +3 oxidation state may be chemically situated in an octahedral coordination environment, while a population of the iron atoms may be in a zero valent metallic state. However, the coordination environment of the zero valent iron atoms is unlikely to be a simple bcc coordination like that of bulk iron metal. Given that the lattice spacing results from TEM imaging and the possibility that internal iron and nickel atoms are in a metallic alloyed phase, we consider the impact of iron composition on the crystal structure of iron–nickel alloys. Mckeehan reported in 1923[60] on the effect of composition on the FeNi alloy crystal structure, where for iron atomic compositions of 70% or less, the alloy takes an fcc crystal structure with octahedral coordination, while for iron compositions of greater than 75%, the alloy phase takes on an iron-like bcc crystal structure with tetrahedral coordination. Based on this work, the iron atoms in a metallic oxidation state are likely to be coordinated more similarly to octahedral coordination chemistry for the Fe20Ni80 and Fe50Ni50 compositions but may be more likely to be closer to tetrahedral geometry for the Fe80Ni20 composition. The challenge with our nanoparticles is that XAS analysis is a bulk characterization technique and averages morphologically and chemically distinct regions of our nanoparticles into one single spectrum. While other studies[28,38,59] can model XAS data based on a singular phase, we know that our nanoparticles have a multiregion morphology with distinct phases likely. Thus, our analysis remains qualitative, but characterization through XAS does provide additional detail about potential phases and metal coordination that is not possible with the other techniques presented herein.

The main Fe K edge absorption peak of n class="Chemical">Fe20Ni80 aligns well with that of Fe(OH)3, whereas the primary Fe K absorption peak for NiFe2O4 is quite similar in shape to that of Fe50Ni50 and Fe80Ni20. This difference is perhaps the most significant when the three FeNi nanoparticle samples are compared. Both reference materials have iron in the 3+ oxidation state; none of the three experimental sample Fe K edges are at all similar to the Fe K edge of the iron foil reference. From the XANES region of Fe K edge, a strong case can be made for the majority of the iron atoms in our nanoparticles to be primarily in +3 oxidation with low contributions from metallic iron, with perhaps a small contribution of iron atoms in the 0 oxidation state.

Similarly, the XANES data for the n class="Chemical">nickel region of the three samples have almost identical spectra with a slight exception of Fe20Ni80 having a moderately higher normalized intensity. The Ni K edge spectra of Fe50Ni50 and Fe80Ni20 are effectively the same. In comparison to the pre-edge of the three standards, only the metallic nickel feature had resemblance with the three nanoparticles. The pre-edge features of the other two standards, β-Ni(OH)2 and NiFe2O4, are very distinct in comparison to that of the three nanoparticles and the Ni metal foil reference. The samples show a metal-like pre-edge feature, but interestingly very high white-line intensity which expected in oxidized Ni. The main absorption peaks for NiFe2O4 and β-Ni(OH)2 have absorption energies in the same range as the three nanoparticle samples but with higher normalized intensity.

Nickel foil was used as a ren class="Chemical">ference for the nickel metal and had an fcc crystal structure. β-Ni(OH)2 has a hexagonally closed packed-structure of Ni2+ and OH– and is the more stable polymorph of Ni(OH)2.[61] β-Ni(OH)2 is found naturally as the mineral theophrastite and is isostructural with brucite [Mg(OH)2].[62] In β-Ni(OH)2, each Ni atom is surrounded by six O atoms to form an octahedral structure. The symmetry of the metal–ligand cluster is the determining factor for the existence and intensity of the pre-edge features.[63] McBreen et al. further suggested that in a pure octahedral configuration, no pre-edge feature should be present at all owing to the center of inversion of the octahedron. Thus, weak pre-edge features only arise with the distortion of the octahedral environment, which leads to the removal of the center of inversion of the symmetry.[63] The presence of broad pre-edge peaks for the nanoparticles could be indicative of Ni distorted octahedral coordination.[64] Kim et al. also suggested that small magnitude of pre-edge features can be emblematic of distorted octahedral coordination.[57] Similar to the alpha phase of Ni(OH)2, β-Ni(OH)2 can also be classified as an insulator and bulk NiFe2O4 also behaves as an insulator.[62,65,66] Unlike in metals, where the screening of the core-hole electron is usually complete, the core-hole electron is only partially screened in insulators. Because of the photo-electron core hole interactions in these nickel oxides (insulators), pre-edge features can be influenced by molecular orbitals of the cluster formed by absorbing and backscattering atoms. Therefore, Mansour et al. cautioned against the direct comparison of pre-edge regions in nickel metal and nickel oxides.[64] Thus, for all three nanoparticle samples, which lacked a distinct pre-edge feature and had only weak and broad spectra, we conclude that the pre-edge shape is the result of the nanoparticles having nickel in distorted octahedral configuration. The pre-edge shape and position strongly align with metallic nickel, but judging by the main absorption peak position, both β-Ni(OH)2 and NiFe2O4, which have an oxidation state of +2 for Ni, closely match with the three nanoparticle samples. However, the intensities of the main absorption peaks of the three nanoparticle samples are lower than both β-Ni(OH)2 and NiFe2O4 references. Numerous variables, which include charge density, ligand symmetry, and spin density, can affect the XANES spectra, but the absorption edge position can be a reliable indicator of the oxidation state in nickel.[67,68] Therefore, taking into account the pre-edge and XANES spectra points to a mixed oxidation state of +2 and 0 for nickel, where nickel atoms in the +2 oxidation are likely the majority species at both the bulk and also the surface of the nanoparticles, as seen through XPS results, while some amount of metallic nickel is prominent in the bulk. Given the known difference in the standard reduction potentials for both Fe2+ (E0 = −0.44 V) and Ni2+ (E0 = −0.23 V), the latter is more likely, as the reduced iron species will preferentially donate electrons to the nickel cations present, causing a reduction to the zero oxidation state. This thermodynamically favorable reaction thus results in the oxidation of the iron metal to a ferrous iron cation. The iron and nickel spectra of three nanoparticle samples having closely related features could be explained by the fact that the synthesis process used was identical regarding salt precursors, ligand stabilizers, reducing agents, and general procedure. XANES data thus give a great insight into the atomic level chemistry of the nanoparticles and thus give a basis for further investigation of bimetallic iron–nickel nanoparticles for enhanced OER performance.

The results of the electrochemical ann class="Chemical">alysis done on the iron–nickel nanoparticles using CV are shown in Figure a. Fe20Ni80 had the lowest onset potential, followed by Fe50Ni50 and Fe80Ni20. The same trend held for the potential required to achieve a benchmark current density (j) of 10 mA/cm2. The overpotentials for Fe20Ni80, Fe50Ni50, and Fe80Ni20 were 313, 327,, and 364 mV, respectively, at 10 mA/cm2 (Figure b). Another benchmark used to test the catalysts in OER is to measure the current density at an overpotential (η) of 300 mV. Fe20Ni80 yet again outperformed Fe50Ni50 and Fe80Ni20. Fe20Ni80 had a current density of 4.92 mA/cm2, which was followed by Fe50Ni50 with 3.43 mA/cm2 and Fe80Ni20 with 0.54 mA/cm2. However, at an overpotential (η) of 450 mV, Fe50Ni50 outperformed the other two catalysts as it reached current density (j) of 207 mA/cm2, while Fe20Ni80 and Fe80Ni20 managed to have current densities (j) of only 150 and 131 mA/cm2, respectively. Based on those three standards (j = 10 mA/cm2, η = 300 mV, and η = 450 mV), it can be inferred that the iron–nickel nanoparticle with lowest iron proportion of iron is the most active at lower overpotential and the nanoparticle having iron quantity of 50% is only slightly less active at lower overpotential but had the highest current density at greater overpotentials. This switch in the best-performing composition is particularly important when one considers that an electrolysis cell would be operating at elevated current densities beyond 10 mA/cm2.[69] Fe80Ni20, with the highest amount of iron, performed significantly worse than the other two nanoparticles. The result obtained in this study closely mirrors the ones in the literature for nickel-based thin films. Corrigan et al. observed that a sweet spot existed for the iron addition in nickel thin films which occurred between 10 and 50%, while Louie et al., Friebel et al., and Trotochaud et al. independently came to the same conclusion.[26,28,30,31] We find it fascinating that the relationship between the ratio of iron to nickel and the OER catalytic performance that exists for the FeNi hydroxide thin films also holds in this study for nanoparticles, but it is also compelling that our results suggest that the recent literature benchmarks for OER (i.e., measurement of overpotential at 10 mA/cm2 and measurement of current at 300 mV of overpotential) are perhaps not appropriate for selecting optimal OER catalysts nor for necessarily selecting directions for catalyst development and design. Instead, we recommend considering a second set of metrics at potentials relevant to commercial electrolysis, such as those tested in the study by Speck et al.[69] At a current density of 100 mA/cm2, our nanoparticle catalysts resulted in overpotentials of 415, 391,, and 436 mV for Fe20Ni80, Fe50Ni50, and Fe80Ni20, respectively. Interestingly, the gap between Fe20Ni80 and Fe80Ni20 is even lower when the catalysts reached a current density of 150 mA/cm2. At a current density of 150 mA/cm2, our nanoparticle catalysts resulted in overpotentials of 450, 415, and 457 mV for Fe20Ni80, Fe50Ni50, and Fe80Ni20, respectively. Prior work by our group has evaluated the electrochemical surface area.[70] Turnover frequency was calculated for each sample, and the result is shown in Table S5. A CV graph of IrO2, a reference catalyst for OER, is shown in Figure S8.

Figure 4

(a,b) CV scans of Fe–Ni NPS in 1 M KOH electrolyte solution after 20 cycles.

(a,b) CV scans of Fe–n class="Chemical">Ni NPS in 1 M KOH electrolyte solution after 20 cycles.

Our performance data generally suggest that the nanoparticles synthesized herein may be in some respects a close replica of advanced n class="Chemical">iron–nickel hydroxide-based thin films, which is a similar conclusion to our prior work.[48] An interesting feature of nickel-based catalysts for OER is the presence of the nickel redox peak before the OER onset potential. Fe20Ni80 had a large nickel redox peak, while Fe50Ni50 had a notably smaller peak and Fe80Ni20 had virtually nonexistent nickel redox peak. The size of the peak is proportional to the nickel content in the nanoparticles. In Fe20Ni80, which had the earliest OER onset potential among the three nanoparticles, the nickel redox peak occurred at the lowest potential. The second most active catalyst Fe50Ni50 had a nickel redox peak at a slightly higher potential than Fe20Ni80. The trend observed in our work suggests that an increase in nickel content (decrease in iron content) shifts the nickel redox peak to occur at a lower potential. The difference in oxidation peaks between Fe20Ni80 and Fe50Ni50 was around 40 mV. Similarly, Louie and Bell noted that the incorporation of iron into nickel hydroxide films shifted the nickel redox peak toward higher potentials, and the amount of iron was linearly proportional to the shift in potential and inversely proportional to the redox area, that is, increase in the iron content in the film directly corresponded to the higher potentials at which the redox peaks occurred.[30] Louie and Bell. implied that increasing iron content led to the shift in nickel redox peak toward higher potential, which in turn resulted in the suppression of electrochemical oxidation of Ni(OH)2 to NiOOH by iron.[30] The result obtained by Louie et al. confirmed the trend observed in our results. In Corrigan’s work, between 0% Fe containing nickel oxide film and 10% Fe containing nickel oxide film, the nickel redox couple for iron-containing film increased by almost 50 mV and the redox couple area was visibly decreased.[26] Lu et al. observed the same trend for NiFe-layered double hydroxide where Ni(OH)2/NiOOH redox couple shifted to higher potential and the redox area decreased compared to just a Ni(OH)2 film which they attributed to inhibition of Ni2+ to Ni3+ conversion by highly charged Fe3+ ions which occupied the neighboring spaces.[71] Hu and Wu claimed that the OER activity for Fe–Ni materials could be reliably predicted by the nickel redox peak that occurs before OER and that the nickel redox peak is strongly dependent on the composition of Fe–Ni which also controls the overall activities of OER. Similar to our work, Hu et al. observed that the nickel redox peak shifted toward more positive potential with increasing iron content.[72] Li et al. also observed the nickel redox peak shift toward more positive potential for Ni/Fe(OH)2 film in comparison to Ni(OH)2 film and asserted that the oxidation potential in a mixed Ni/Fe(OH)2 is actually determined by the potential of Fe(III)/Fe(IV) conversion as Fe(OH)3 is a poor conductor until it is converted into the higher oxidation state.[73] Similar to our work where Fe80Ni20, which contained a high amount of iron did not show any redox peak, Bates et al. also observed a muted redox peak for their Fe/Raney Ni and suggested that the iron-rich surfaces do not increase OER activity as much as other iron–nickel based catalysts.[23] To address the phenomenon of iron affecting nickel redox properties, Görlin et al. theorized that negatively charged oxygen ligands are formed in iron centers which suggests a transformation from a two proton-two electron process to a two proton-one electron transfer.[74]

CA and CP data are presented in Figure , with a summary of performance metrics in Table . CA was conducted at a n class="Chemical">constant potential of 1.6 V versus reversible hydrogen electrode (RHE), and CP was performed at a constant current density of 10 mA/cm2. Observing the CA data, Fe80Ni20 is the most stable catalyst throughout the 12 h with a degradation rate of −0.003 mA/h. The overall degradation rate of Fe50Ni50 was −0.025 mA/h, while Fe20Ni80 was the most unstable catalyst out of the three tested, with a degradation rate of −0.026 mA/h. Interestingly, the CA results are inverse in trend compared to the OER activity results obtained through the CV. Both Fe20Ni80 and Fe50Ni50 compositions, which had the lowest onset OER potentials, were much more unstable than Fe80Ni20, which had the largest onset OER potential among the three. CP data in Figure also indicates the same stability trend of the nanoparticle catalysts. Fe80Ni20 was the most stable with a degradation rate of 0.017 mV/h, followed by Fe50Ni50 with a 0.951 mV/h degradation rate, and Fe20Ni80 with a 1.288 mV/h degradation rate. Tafel slopes for the three catalysts are plotted in Figure c. Fe50Ni50 had the lowest Tafel slope of 44.2 mV dec–1 followed by Fe80Ni20 with a Tafel slope of 48.6 mV dec–1, and Fe20Ni80 had the highest Tafel slope of 62.5 mV dec–1. Tafel slopes for all three nanoparticle catalysts being less than 120 mV dec–1 rules out the prevalence of surface species that were formed a step earlier than the rate determining step.[27] A Tafel slope value of 40 mV dec–1 demonstrates that there is a pre-equilibrium where a one-electron electrochemical step is followed by another one-electron electrochemical rate-determining step.[71,75] Typically, Tafel slopes close to 40 mV dec–1 are indicative of a second-electron transfer step as the rate determining step.[76,77] In comparison, a Tafel slope value of 60 mV dec–1 occurs when the rate-determining step involves a chemical step after the first electron transfer step.[78] Judging by the Tafel slope values for the three nanoparticles, there arises a distinct possibility that the OER kinetics for Fe50Ni50 and Fe80Ni20 was controlled by a second-electron transfer step as the rate determining step and the kinetics for Fe20Ni80 controlled by the chemical step following the first electron transfer step. In their work on electrodeposited nickel–iron alloy thin films, Singh et al. calculated Tafel slopes for thin films with a differing composition of iron and nickel at 40 ± 5 mV dec–1. Singh et al. proposed the reaction order to be approximately 2 concerning OH– concentration. Singh et al. also suggested that OER initiates at the Ni3+ sites in the mixed iron–nickel films and goes through a fast electrochemical step where surface adsorbed OH– is formed and followed by a slow step, where desorption takes place.[79] Lyons and Brandon measured the Tafel slope value of around 40 mV dec–1 for both nickel oxide and iron oxide electrodes in their work. Lyons and Brandon proposed that the rate-determining step is caused by the formation of an −OOH species for iron oxides, while the mechanism for nickel oxides was similar to that of iron oxides.[78]

Figure 5

(a) CA of Fe–Ni NPs at 1.6 V vs RHE. (b) CP of Fe–Ni NPs at 10 mA/cm2. (c) Tafel slopes of Fe–Ni NPs.

Table 2

Electrochemical Performance of the Catalysts for OER

				degradation rate
sample	Tafel slope (mV dec–1)	current density@overpotential of 300 mV (mA cm–2)	overpotential@current density of 10 mA cm–2 (mV)	mA/h	mV/h
Fe80Ni20	48.6	0.45	363	–0.003	0.017
Fe50Ni50	44.2	3.21	326	–0.025	0.951
Fe20Ni80	62.5	5.53	313	–0.026	1.288

(a) CA of Fe–n class="Chemical">Ni NPs at 1.6 V vs RHE. (b) CP of Fe–Ni NPs at 10 mA/cm2. (c) Tafel slopes of Fe–Ni NPs.

Conclusions

Three nanoparticles with varying iron to n class="Chemical">nickel ratios were synthesized to test whether the ratio between iron and nickel in these catalysts impacted the OER activity and performance. TEM, ICP, and EDX were used to observe the morphology and verify the composition of the nanoparticles. Characterization tools such as XPS and XANES helped in exploring and understanding the catalysts at the atomic level. Existence of different phases was found in the catalysts with iron mostly in the +3 oxidation state and nickel primarily in the +2 oxidation state. Electrochemistry tests (CV, CA, and CP) were performed to assess the activity and stability of the synthesized catalysts. Fe20Ni80 showed the best performance with lower overpotential and higher current density, thus indicating that bimetallic iron–nickel nanoparticles with iron around 20% and nickel 80% are the most active one. However, the stability of Fe20Ni80 underperformed compared to the other two catalysts, Fe50Ni50 and Fe80Ni20. The results of this study indicate that there is a trend with catalysts containing a lower amount of iron having greater OER activity. However, stability is also a critical factor in the design of a catalyst and the nanoparticle with the higher iron content (Fe80Ni20) was much more stable compared to the low iron ones (Fe50Ni50 and Fe20Ni80). Thus, future research into bimetallic iron–nickel based catalysts should help in a further understanding of the effects that ratio plays in the electrocatalysis of OER to ultimately design low-cost and effective OER catalysts.

Experimental Section

Materials

Chemicals were obtained as n class="Gene">ACS grade commercial products and used without further purification unless specified. Iron(II) sulfate heptahydrate (FeSO4·7H2O), nickel(II) chloride hexahydrate (NiCl2·6H2O), amino tris(methylene phosphonic acid) (ATMP), polyvinylpyrrolidone (PVP40000), sodium borohydride (NaBH4), potassium hydroxide (KOH), methanol, concentrated nitric acid (HNO3), and concentrated sulfuric acid (H2SO4) were purchased from commercial vendors. Ultrapure water (18.2 MΩ·H2O) was obtained from a Milli-Q integral system installed in the laboratory. Cationic ionomer was obtained from Prof. E. Bryan Coughlin at the University of Massachusetts, Amherst and used in solution to make catalyst inks.

Nanoparticle Synthesis

Fe–n class="Chemical">Ni nanoparticles were synthesized under room temperature and atmospheric pressure conditions. All of the solutions used ultrapure deionized water as the solvent. Solutions of 29.79 g/L ATMP and 4.982 g/L of FeSO4·7H2O were mixed at a molar ratio of 0.05:1 in water and hand-mixed for a short time. The ATMP compound stabilizes the iron cations in solution. Separately, PVP40000 (molar ratio of Ni/PVP40000 = 1:0.005) and NiCl2·6H2O solutions were hand-mixed together for a short time. The amount of NiCl2·6H2O added depends on the molar ratio of Fe and Ni desired. The first mixture of Fe/ATMP solution, along with the second mixture of Ni/PVP solution, were then transferred to a 250 mL three-neck borosilicate flask and placed on an orbital shaker. The solution inside the three-neck flask was mixed under argon gas for 15 min and at 100 rpm. Argon bubbling of the iron–nickel solution is performed to prevent the unwanted oxidation of the iron and nickel precursors and to control oxidation during nanoparticle formation. At approximately 13 min of argon bubbling, NaBH4 [molar ratio of metal/BH4– = (1:2.2)] aqueous solution was prepared to minimize the reaction time of NaBH4 with water before adding to the iron–nickel precursor solution. The aqueous solution of NaBH4 was then added into the metal precursor solution in the three-neck flask dropwise via a syringe at a rate of approximately 30 μL/s while hand mixing the solution. Borohydride ions (BH4–) reduce the stabilized Fe2+ and Ni2+ ions into Fe0 and Ni0, respectively, forming nanoparticles during the reduction reaction. The solution in the three-neck flask was then mixed under vacuum for 15 min on an orbital shaker at 100 rpm. After 15 min of mixing, the solution in the three-neck flask was transferred to a 50 mL test tube and centrifuged for approximately 3 min. The supernatant from the centrifuged test tube was then decanted. The nanoparticles remaining in the test tube were mixed with 20 mL of methanol in a vortex shaker for about 30 s. The test tube containing the solution was again centrifuged for about 3 min, and the supernatant was decanted. The nanoparticles were then mixed with 20 mL of methanol and resuspended. Because the nanoparticles are exposed to water and air during and after synthesis, respectively, and the nanoparticles are not kept in a completely anoxide environment, we expect that the surface of the nanoparticles will oxidize.

Characterization

Imaging was performed on an FEI Titan 80-300 transmission electron microsn class="Chemical">cope operating at 300 kV. The imaging was carried out in TEM mode. The TEM samples were prepared on TEM grids (Ted Pella Inc Formvar/Carbon 200M Cu) by diluting the nanoparticles from a concentration of 1–0.01 g/L in methanol, and 2 μL of the diluted nanoparticle solution was dropped cast on to the grid, and the methanol present in the solution was allowed to evaporate completely. Analysis of TEM images was done using Gatan software. EDX was performed with a Nova Nanolab 200.

ICP–MS was performed using IcapQ ICP–MS (Thermo Scientific) on the Fe–n class="Chemical">Ni nanoparticles to confirm the stoichiometry in the as-synthesized materials. Each nanoparticle sample was diluted to a concentration of 50 ppb in a 2.5% nitric acid matrix. The standard curves for iron and nickel were generated using commercial ICP standards (Aristar, BDH) for each metal and diluting standard solutions to concentrations of 5, 20, 40, 60, 80, and 100 ppb. A 2.5% nitric acid matrix was used as the lab blank.

X-ray photoelectron spectroscopy (XPS) was performed on a PHI Versaprobe 5000 for each n class="Chemical">Fe–Ni nanoparticle sample to obtain a detailed chemical and elemental analysis, and the data analysis was done with the PHI MultiPack software. A monochromated Al Kα beam, along with Ar and C-60 cluster ion guns, was used. An initial scan of 0–1400 eV was carried out on the samples. This initial analysis was accompanied by detailed scans of iron (700–740 eV) and nickel (844–894 eV).

XAS was performed on the Fe–n class="Chemical">Ni nanoparticles at a synchrotron facility in Advanced Photon Source (12-BM-B) which is a user facility branch of the Argonne National Laboratory. The samples and the standards tested for XAS were dried and suspended in a solid matrix (Kapton tape). The measurements were done around the K-edge of Fe (7112 eV) and K-edge of Ni (8333 eV). Data analysis was done using the software program Athena.

Electrochemical Measurements

All electrochemicn class="Chemical">al experiments were carried out in a conventional cell at room temperature using a potentiostat (PINE WaveNow 50) in three stationary electrodes (working electrode, counter electrode, and reference electrode) system. AfterMath software was used for data collection purposes. A gold electrode (1.6 mm Au BASi) was used as the working electrode, 3 M Ag/AgCl was used as the reference electrode of the cell, and a graphite rod was used as the counter electrode. The electrolyte solution for all experiments was 1 M KOH. The reference electrode was placed in a salt bridge consisting of 3 M NaCl solution to avoid the dissolution of the frit, as well as to prevent the formation of Ag2O due to contact with KOH solution.

Catalyst inks were prepared for n class="Chemical">all three samples by combining the nanoparticles with a cationic ionomer at a ratio of iron to ionomer of 6:1 in methanol. The ink was then sonicated for 45 min in a cold-water bath to ensure a homogeneous mixture of the nanoparticle and the ionomer, and 1 μL of the ink was then drop-cast on the surface of a clean Au electrode (surface area of 0.02 cm2), and the methanol was allowed to dry for 15 min at room temperature. The mass loading onto the electrode surface was ∼50 μg/cm2. The cleaning procedure involved the polishing of the electrode surface with alumina solutions (5, 0.05 μm) on a polishing pad, followed by sonication of the electrode in deionized (DI) water for 5 min. Next, the gold working electrode was placed in a three-electrode cell setup consisting of 0.5 M H2SO4 as the electrolyte solution. CV was carried out in the cell for 50 cycles at 100 mV/s between −0.3 and 1.7 V versus Ag/AgCl. The working electrode was then taken out and then rinsed with 18.2 MΩ DI water.

All solutions of the 1 M KOH electrolyte were purified to remove trace n class="Chemical">iron impurities by following a procedure developed by Trotochaud et al.[31] In essence, to purify KOH, first, all the glassware and polypropylene centrifuge tubes that were used in the procedure were cleaned with 10% sulfuric acid (H2SO4) solution. Then, 2 g of nickel nitrate hexahydrate [Ni(NO3)2·6H2O] was dissolved with 4 mL of 18.2 MΩ DI H2O in a tube, and 20 mL of 1 M KOH was added together and mixed to obtain nickel hydroxide [Ni(OH)2] as a precipitate. The mixture was then shaken on a vortex shaker for approximately 1 min and centrifuged. The supernatant was emptied from the tube, and three more cycles of washing Ni(OH)2 precipitate with 20 mL H2O and 2 mL of 1 M KOH were employed with each wash cycle followed by re-dispersion, centrifugation, and supernatant decantation successively. After the 3rd cycle was completed, 50 mL of the 1 M KOH electrolyte to be purified was added into the tube and shaken with a vortex shaker for approximately 10 min, and the mixture was allowed to rest for 3 h for iron impurity removal, after which the solution was centrifuged, and the purified 1 M KOH supernatant was decanted into an H2SO4-cleaned tube.

CV was conducted on the samples in a 3-electrode setup as described above. The n class="Chemical">Fe-free KOH (electrolyte solution) was bubbled with Ar gas for 30 min before running CVs to remove dissolved O2. During CV experiments, a continuous flow of Ar gas was maintained in the headspace. CVs were taken between 0 and 0.8 V versus the reference electrode. Five CV cycles were run with the clean Au electrode, and 15 CV cycles were run with the catalyst deposited on the Au working electrode. The scan rate was 20 mV/s for all CV experiments. Also, the measured potential versus Ag/AgCl was converted to potential versus RHE, ERHE, using the following equation

The measured pH of 1 M KOH was ∼14. EAg/n class="Chemical">AgCl0 is 0.21 V for the Ag/AgCl reference electrode in 3 M NaCl. The data were then adjusted for iRu correction, where i is the current and Ru is the uncompensated resistance. Ru was measured using potentiostatic electrochemical impedance spectroscopy, and the values for Ru were taken at a frequency of 100 KHz. iRu values were then subtracted from the measured potential versus RHE. Overpotential was calculated by subtracting the theoretical potential for OER, 1.23 V, from the measured potential versus RHE. Current measurements are reported as current density (j), where current is normalized to the geometric surface area of the Au electrode (0.02 cm2). CP and CA were performed on the sample using the same electrochemical setup as was used to obtain CV data. CP was conducted at a constant current density of 10 mA/cm2 for 12 h and CA at a constant potential of 1.6 V versus RHE for 12 h.