Introduction

Dimensional reduction of first row metal oxides gives rise to metallate oxygen evolving catalysts (M-OECs) that exhibit high activity for the oxygen evolution reaction (OER)1,2,3,4,5. Electrodeposition of oxidic cobaltates and nickelates in the presence of phosphate and borate (CoPi6,7, CoBi8,9, NiBi10,11, MnPi12,13) results in clusters of 10–60 metal atoms, as determined from in situ pair distribution functional analysis9,14,15,16,17,18,19. The self-healing property of the M-OECs2,20,21,22 allows them to promote water-splitting under benign conditions. Under such conditions, the catalysts may be easily interfaced with materials for direct conversion of water to oxygen and hydrogen at high efficiency23,24,25,26, as well as interfaced with biological organisms to perform artificial photosynthesis27 at efficiencies greatly exceeding natural photosynthesis28,29. The metallate clusters possess a high edge-to-area ratio that engenders high activity, as revealed by isotopic labelling studies30 that show the critical O–O bond formation step to occur by proton-coupled electron transfer (PCET) at cluster edge sites8,31,32,33,34,35,36. Moreover, the electronic charge in M-OECs can delocalize within the clusters37,38 giving rise to electron/hole transport39 that can maximally couple to the ion transport needed to support the OER40,41.

Iron doping of metal oxide films has long been known to increase overall OER activity of metal oxide OER catalysts42,43. The behaviour of Fe in Ni-OECs has been revisited44, and the role of Fe has been ascribed to various factors, including active site Fe4+ or higher valent species45,46,47, near neighbour Fe effects on Ni resulting from strain on the oxide lattice48,49,50,51, active oxygen intermediates at Ni–Fe sites52,53,54, Fe induced partial-charge-transfer to Ni sites55,56, and Fe acting as a Lewis acid that promotes charge transfer character and favourable energetics for Ni oxyl formation57,58. Quizzically, though detected by Mössbauer spectroscopy, the presence of Fe4+ does not correlate with the observed catalytic activity59. Iron loading has also been shown to affect the OER activity of Co-OECs60,61,62, but at very different alloying loads. Whereas Ni-OECs show maximal activity with Fe loadings of ~5 mol% Fe42,55, the maximal activity of Co-OECs is observed for Fe loadings of >40 mol% Fe61. These higher loadings suggest different roles for Fe in enhancing M-OEC activity at high versus low alloying.

We now report the zero-field 57Fe Mössbauer and X-ray absorption spectroscopy (XAS) of Co-OEC alloyed with Fe from 0 to 100% and show that, unlike Fe alloyed in Ni-OECs, the presence of Fe4+ tracks OER activity, suggesting that Fe4+ is intimately involved as a redox activator of OER. The results suggest different roles for Fe in alloyed M-OEC catalysts. At low loadings such as in (Ni:Fe)-OECs, OER is performed by Ni active sites and Fe promotes the PCET activation of OER. At high loadings, as is observed here for Fe-alloyed Co-OEC catalysts, the redox properties of Fe appear to play a prominent and more direct role in promoting OER.

Results

A series of CoFeOx films with varying Fe content were prepared by cathodic deposition upon the reduction of nitrate to induce a high local basic pH near the electrode, resulting in the electrodeposition of a Co:Fe hydroxide film13, which was then converted to CoFeOx with the application of an anodic potential. Metal elemental compositions were determined by inductively coupled mass spectrometry (ICP-MS) of digested films after electrochemical measurements. Previous studies have shown that Fe and Co deposit homogeneously as detected by SEM/EDS analysis63. As Fe content increases, both the cathodic and anodic features of the Co2+/3+ couple are shifted towards higher potentials (Supplementary Fig. 1). CoFeOx films with 40–80 mol% Fe in 1 M KOH that was scrubbed of trace metal contaminants display Tafel slopes of ~30 mV/dec (Supplementary Fig. 2), similar to previously published results61. We note that the lowest Tafel slopes in CoFeOx films are obtained at much higher Fe:Co ratios than observed for NiFeOx films.

The electronic structure of Fe centres in CoFeOx films was probed with zero-field 57Fe Mössbauer spectroscopy. A representative 57Fe Mössbauer spectrum is given in Fig. 1; the spectra of all CoFeOx film samples are given in Supplementary Fig. 3 and Supplementary Fig. 4. The spectra are reproducible and sensitive to Fe population changes between samples with 10 mol% Fe differences (Supplementary Fig. 5). Two species of Fe are detected in the 57Fe Mössbauer spectra. Fits of the spectra furnish corresponding isomer shifts (δ) and quadrupole splittings (|ΔEQ|) for one species with δ ~ 0.3 mm/s and |ΔEQ| ~ 0.7 mm/s and the other species with δ ~ –0.2 mm/s and |ΔEQ| less than the resolved linewidth (~0.3 mm/s). These values, which are somewhat sensitive to total Fe alloying concentrations and fittings (Supplementary Fig. 7), do not correspond to either Fe2O364,65 or metallic Fe66,67. One species matches the Mössbauer parameters of high spin (HS) Fe3+ in the oxide ligand field of NiFeOx5,68,69 and FeOOH70. The Mössbauer parameters of the second Fe species correspond to those observed previously for Fe4+ in NiFeOx70, and is consistent with theoretical calculations showing the persistence of Fe4+ in NiFeOx71. The Fe3+:Fe4+ ratio, which may be determined from the Mössbauer parameters, shifts towards Fe3+ at low and high Fe concentrations (Supplementary Fig. 6) with a maximal absolute Fe4+ concentration observed between 40 and 60% Fe loading (Fig. 2). Strikingly, as Fig. 2 illustrates, the population of Fe4+ in CoFeOx films tracks OER activity as reflected in Tafel slopes. We observe a direct correlation between absolute Fe4+ content and low Tafel slopes (30 mV/dec), implicating the important role of Fe4+ in enhancing OER activity in CoFeOx films at high Fe alloying concentrations. The maximum in activity is likely a result of Fe becoming the dominant compositional metal. Unary Fe oxide films are inferior OER catalysts even as ultrathin sub-monolayer films72. Thus the observed maximum in activity is consistent with the active site for OER becoming dominated by an Fe-only composition at high iron loadings in excess of 50% loading.

Fig. 1: Mössbauer spectra of CoFeOx.
figure 1

Zero-field 57Fe Mössbauer spectra for CoFeOx films with the composition 40% Fe:60% Co. Raw data (black circle), fit for Fe3+ species (blue line), Fe4+ species (green line), and overall fit (red line).

Full size image
Fig. 2: Correlation of Tafel slope with Fe4+ composition.
figure 2

Overlay of Tafel slope (red circle) with absolute Fe4+ (green triangle) population in CoFeOx films with increasing Fe content. Tafel measurements were run in triplicate, and the average value is shown on the graph at 95% confidence limits.

Full size image

The assignment of Fe3+ and Fe4+ as deduced by Mössbauer spectroscopy is supported by Fe K-edge XAS of CoFeOx samples with varying Fe content. The X-ray absorption near edge structure (XANES) spectra and the k-space and R-space data are given in Fig. 3 for CoFeOx with a 50% Fe:50% Co (black line) composition. Analogous data for several other compositions are given in Supplementary Fig. 8. As the proportions of Fe3+ and Fe4+ are known from Mössbauer, the individual XANES spectrum for each Fe3+/Fe4+ species may be ascertained from linear combination fitting. The resulting Fe3+ and Fe4+ spectra are reproducible for specific linear combinations (Supplementary Figs. 9 and  10); the spectrum for the 50% Fe:50% Co sample (black line) and corresponding linear combinations for the Fe3+ and Fe4+ (blue and green line, respectively) contributions are given in Fig. 3. Several important observations can be made between the Fe3+ and Fe4+ species. There is a large edge shift from ~7124 to ~7128 eV between the Fe3+ and Fe4+ species. Additionally, the 1s → 3d pre-edge intensity of the Fe3+ species is significantly higher than that for Fe4+ (inset of Fig. 3a). The R-space amplitude of the Fe3+ species is significantly lower than the Fe4+ species (Fig. 3c). This decreased amplitude suggests a lower coordination number. There is also a clear contraction of the first and second shell scattering distances for Fe4+. The pre-edge region reflects transitions to the many-electron excited states of the metal centre. The spectral intensity of the pre-edge derives from both electric quadrupole and electric dipole mechanisms. In a centrosymmetric ligand field (e.g., Oh), the electric dipole contribution is parity forbidden, and only the quadrupole intensity is present. Conversely, deviation from centrosymmetry (e.g., Td) results in a significant increase in the pre-edge intensity. This increase in intensity derives from electric dipole allowedness, which tracks with the amount of 3d–4p mixing in a noncentrosymmetric ligand field73. These observations, together with a low coordination number from the low amplitude R-space data, suggest that Fe3+ is present in a distorted ligand field lacking inversion symmetry—either Td or square pyramidal ligand field geometries are likely possiblities.73 Along similar lines for Fe4+, the low pre-edge intensity suggests a more symmetric Oh ligand field, which will largely exhibit electric quadrupole intensity. A more symmetric ligand field is also consistent with the higher amplitude R-space data and small |ΔEQ|.

Fig. 3: X-ray absorbance spectra of CoFeOx.
figure 3

a Fe K-edge X-ray absorbance spectra and corresponding b k-space and c R-space for CoFeOx with the composition 50% Fe:50% Co (black line), and calculated Fe3+ (blue line) and Fe4+ spectra (green line). Inset of a highlights the pre-edge region.

Full size image

The combination of Mössbauer, XAS, and EXAFS data support the assignment of a high-valent Fe4+ species in a symmetric Oh ligand field. An alternative scenario to consider is a low-spin Fe3+ centre, though it must be in a strongly electron withdrawing environment. For instance, negative isomer shifts can be obtained in low-spin Fe3+ complexes in the presence of strong back-bonding (e.g., K3[Fe3+(CN)6] or Na2[Fe3+(CN)5NO])74. Additionally, when [Fe3+(CN)6]3– is coordinated in supramolecular assemblies (e.g., Prussian blue analogues) involving metal–metal interactions. Such second-sphere coordination of the Fe–CN bonds by another metal ion can shift the Fe δ by ~–0.1 mm/s. Thus, metal–metal or charge transfer interactions in CoFeOx could effectively decrease the Fe-based s electron density and give rise to a negative δ and high Fe K-edge energy. However, this scenario would result in an isomer shift that gradually shifts more negative as this Fe species is surrounded by more Co centres. This is not observed here; the isomer shift of the Fe species remains relatively constant at δ ~ –0.2 mm/s and in fact becomes slightly less negative with increasing Co concentration (Supplementary Fig. 7). Similarly, the Fe4+ XAS spectra obtained from linear combination fits using different Co:Fe ratios are very similar (Supplementary Fig. 9). These considerations, together with the weak ligand field imposed by oxide coordination and the consistency between the Mössbauer and Fe K-edge XAS data, suggest that the Fe species observed here can be assigned to a high-valent Fe4+ centre in an Oh ligand field.

Discussion

Iron activates M-OECs for OER but its role appears to differ with the nature of the M-OEC and the condition under which it operates. Although most OER is performed in concentrated base, the large-scale deployment of renewable energy storage has prompted interest in performing OER in neutral water sources75,76. For this line of investigation, M-OECs excel owing to their stability arising from their self-healing properties22. At neutral pHs, Fe3+ plays a role in OER that appears to be derived from non-redox properties. PCET activation of water is impaired since water is a poor proton acceptor; Fe may act as a Lewis acid77 to increase the acidity of OHx (aqua/hydroxo) moieties that are coordinated to M-OECs and thereby lower the reduction potential for the M3+/4+ couple and lead to a greater population of M4+ in the Fe-doped catalysts. This in turn gives rise to increased oxyl character (M(IV)⋯ O ↔ M(III)–O•). This Lewis acidity behaviour is supported by the observation that Fe doping in NiPbOx shows no enhancement in OER at solution pH values commensurate with the pKa of Fe3+. Moreover, OER enhancement may be replicated by non-redox active, Lewis acidic cations in Fe-free Ni-OECs78. In concentrated base, OH– can adequately serve the role of a proton acceptor and the influence of the Fe3+ is diminished. When Fe4+ is implicated in OER, as has been proposed in numerous studies, OER appears to occur at the M (Co or Ni) metal centre with Fe4+ promoting the activation of OER at the M of the M-OEC. Such proposals are consistent with the electronic structure of first row transition metal centres confronting the “oxo-wall”79. Moving to the right in the periodic table, the d-electron count for the M4+ formal oxidation state increases and in a tetragonal oxide ligand field, the dxz and dyz orbitals are populated, preventing electron donation from terminal oxygen to the metal centre. Consequently, the M–O bond strength is much weaker for Co4+ than for Fe4+, which formally accommodates a kinetically more inert metal-oxo double bond. From a kinetics perspective, we believe that Fe4+ is not the active site from which OER occurs but rather OER occurs from Co centres with the Fe4+ participating as a redox cooperative centre where Fe4+ enhances the oxidizing power of a Co:Fe active site (Supplementary Fig. 1) versus a Co4+-only active site. Thus, we believe that for both NiFeOx and CoFeOx systems, Fe3+ functions as a Lewis acid in promoting PCET reactivity for the OER. However, unlike NiFeOx, OER activity in CoFeOx tracks the Fe4+ alloying concentration, suggesting that the redox properties of the Co4+ centre is further enhanced by the presence of redox active Fe4+ centres.

In conclusion, we have spectroscopically detected and characterized a high-valent Fe4+ centre in CoFeOx thin film OECs. Spectroscopic data suggest that this Fe4+ centre is located in a symmetric Oh oxide ligand field. The correlation between Fe4+ content and OER activity in CoFeOx thin films suggests an important role of this high-valent state in the mechanism of O–O bond formation and oxygen evolution and supports the merits of exploring mixed-metallate oxygen evolution catalysts.

Methods

Materials

Catalysts with specific Fe:Co ratios were prepared by electrodeposition from metal nitrate salt solutions that were degassed. After deposition, the film was rinsed briefly in Type I water and then submerged in KOH buffer. Films were held at a constant potential of 0.84 V in 1 M KOH pH 14 or 1.0 V in 0.1 M KOH pH 13 for 3 h to convert the film to the oxyhydroxide form before further electrochemical analysis. To obtain films of various thicknesses, the total deposition time was altered between 30 and 120s and the current held during deposition was changed between 0.5, 1.0 and 5.0 mA/s. The exact film loading was obtained from ICP-MS analysis of the films.

Electrochemistry

All electrochemical experiments were conducted at room temperature (23 ± 1 °C). Electrode potentials were converted to the NHE scale using E(NHE) = E(Ag/AgCl) + 0.197 V. Overpotentials for the OER from water were computed using η = E(NHE) − (1.23 V − 0.059 V × pH).

Spectroscopy

CoFeOx catalyst with natural 57Fe abundance were prepared for Mössbauer spectroscopy at 77 K. The data were calibrated and fit to linear combinations of symmetric pairs of Lorentzian peaks. Fe K-edge XANES spectra were collected at beamline 12BM-B at the Advanced Photon Source at Argonne National Laboratory. Reconstructed spectra of the pure Fe3+ and Fe4+ species were obtained through linear combinations of the XAS spectra of the various CoFeOx films.