Elucidation of Structure–Activity Correlations in a Nickel Manganese Oxide Oxygen Evolution Reaction Catalyst by Operando Ni L-Edge X-ray Absorption Spectroscopy and 2p3d Resonant Inelastic X-ray Scattering

Herein, we report the synthesis and electrochemical oxygen evolution experiments for a graphene-supported Ni3MnO4 catalyst. The changes that occur at the Ni active sites during the electrocatalyic oxygen evolution reaction (OER) were elucidated by a combination of operando Ni L-edge X-ray absorption spectroscopy (XAS) and Ni 2p3d resonant inelastic X-ray scattering (RIXS). These data are compared to reference measurements on NiO, β-Ni(OH)2, β-NiOOH, and γ-NiOOH. Through this comparative analysis, we are able to show that under alkaline conditions (0.1 M KOH), the oxides of the Ni3MnO4 catalyst are converted to hydroxides. At the onset of catalysis (1.47 V), the β-Ni(OH)2-like phase is oxidized and converted to a dominantly γ-NiOOH phase. The present study thus challenges the notion that the β-NiOOH phase is the active phase in OER and provides further evidence that the γ-NiOOH phase is catalytically active. The ability to use Ni L-edge XAS and 2p3d RIXS to provide a rational basis for structure–activity c...


■ INTRODUCTION
The ability to use sunlight to split water into H 2 and O 2 is an attractive target toward realizing a renewable energy economy. The major challenge is to have catalysts that are able to efficiently enable H 2 and O 2 evolution. In terms of electrocatalysis, the oxygen evolution reaction (OER) is by far the most demanding of the two half reactions. It is perhaps for this reason that there has been an increasing interest in the development of earth-abundant transition-metal oxides as electrocatalysts for OER. 1−6 In this context, nickel oxides and (oxy)-hydroxide materials are promising targets due to their low cost, high catalytic activity, and relatively lower overpotentials. 7−10 Obtaining a detailed understanding of the electrocatalytic activity and mechanism of nickel oxide-based OER catalysts has been challenging due to uncertainties in the NiO x H y phase that is present under operating conditions. Figure 1 summarizes the well-known Bode scheme for the NiO x H y phases, which may be present under various conditions. 11 The depicted structures are well described elsewhere by Strasser and Dionigi. 12 In general, the γ-NiOOH phase is either obtained via the oxidation of the α-Ni(OH) 2 phase or by the overcharging of the β-NiOOH phase in concentrated alkaline solution. The Bode scheme assumes a reversible cycle where the α-Ni(OH) 2 oxidizes and transforms into γ-NiOOH upon charging (Figure 1), transitioning from a Ni(II) precursor to a higher valent Ni compound that has been proposed to have an average oxidation state of Ni ∼ + 3.6. 11 Upon aging, however, α-Ni(OH) 2 is transformed to a β-Ni(OH) 2 , and once charged allows for the formation of a +3 β-NiOOH. Furthermore, upon overcharging, it can again produce a γ-NiOOH. The cycle is completed by the discharge of the γ-NiOOH into the α-Ni(OH) 2 . A question of longstanding interest has been whether the β-NiOOH or γ-NiOOH phase is more active toward OER. While it is has been argued in the past that the β-NiOOH phase is the most efficient for OER catalysis, 13,14 the majority of the recent studies and reviews in this field have seemingly reached consensus that γ-NiOOH with its higher oxidation state (>3+) is active species for this reaction. 12,15−17 While general agreement has been reached on the phase of Ni−OOH in pure nickel oxide materials, the nature of the catalytically active nickel species in the presence of other transition metals is not well understood. In particular, it has been shown that nickel oxyhydroxides are compromised by the presence of transition-metal contamination, for example, Fe, Co, and Mn in the electrolyte, 18 which can influence the catalytic activity by several orders of magnitude. 19,20 Several studies in which Mn was intentionally introduced showed contradictory reports regarding its effect. For instance, studies of electrodeposited nickel oxide films have reported that the addition of Mn does not significantly alter OER activity. 19,20 Furthermore, the presence of Mn has been shown to have a negligible impact on the Tafel slope. 20 In contrast, a recent study by Lu et al. suggests that Mn doping modifies the electronic structure of the catalyst and improves the electrocatalytic OER performance. 21 Studies by Driess and coworkers have shown that Ni-rich manganese oxides are more active for photocatalytic and oxidant-driven OER than pure NiO. However, they showed that pure NiO is more active for OER during electrocatalysis than any Mn-containing nickel oxide. These discrepancies in the literature suggest that it would be very useful to better understand the active phase of manganese-containing NiO materials during catalysis.
In the present study, we focus on a Ni 3 MnO 4 catalyst, with the goal of assessing the presence of the active phase of Ni− OOH. Specifically, we wish to assess whether a β-NiOOH or γ-NiOOH is formed under operando conditions. Herein, the synthesis and electrochemical oxygen evolution experiments for a graphene-supported Ni 3 MnO 4 catalyst are reported. We then utilize a combination of Ni L-edge X-ray absorption spectroscopy (XAS) and Ni 2p3d resonant inelastic X-ray scattering (RIXS) to obtain detailed insights into the Ni 3 MnO 4 catalyst under operating conditions. This was achieved by a spectroelectrochemical cell, where a potentiostat continuously applies a defined redox potential while monitoring the catalytic current during X-ray spectroscopic measurements. We note that ideally one would also like to follow the changes that occur at Mn. However, due to the to the close proximity of the Mn L-edge to the O K-edge, operando measurements in aqueous solution are prohibitive at the Mn L-edge due to the strong O K-edge background. Hence, in the present study, we focus only on the Ni active site.
As 2p3d RIXS is still not broadly applied in catalysis research, we first briefly describe the physical phenomenon. In 2p3d RIXS, one excites a transition metal from a 2p 6 3d n ground state to a 2p 5 3d n+1 intermediate state ( Figure 2). Dipole allowed emission to give rise to a 2p 6 3d n ′ final state, which may either be at the same energy as the ground state (thus giving rise to an elastic scattering) or at a different energy (corresponding to an inelastic scattering). The energy difference between the ground and final state is referred to as the energy transfer, where the elastic event will appear at 0 eV and d−d transitions will appear at nonzero energy transfer values. By tuning the energy of the incident beam to different excitation energies, one may preferentially populate different intermediate states, thus giving rise to different final state spectra. This is particularly useful in studying catalysts under operando conditions, as it means that one can tune the energy of the excitation to a specific XAS L-edge energy and thus preferentially select for RIXS spectra from an absorber of a specific oxidation state. This point will be further elaborated on in the studies presented herein.
To the best of our knowledge, the reported experiment is among the first soft X-ray RIXS studies of OER catalysis in transition-metal-based systems. 22 While numerous studies have been reported in the hard X-ray regime, 23−25 the experimental challenges presented when utilizing low-energy X-rays for operando measurements in water have made such studies relatively limited. 17,26−29 For the present work, this was achieved by the controlled deposition of the sample on a 150 nm-thick Si 3 N 4 membrane and the use of the specially designed vacuum-compatible liquid electrochemical cell. 30 By utilizing the low energy Ni 2p excitation for these experiments, the resultant data greatly benefit from reduced core-hole lifetime broadening. 31 By employing 2p3d RIXS, we are able to experimentally map out the energy of the low-lying excited states. 32−34 As the relative energy of these states is key to understanding the catalytic activity, these measurements provide unique quantitative insights into the electronic structure perturbations that occur in the Ni 3 MnO 4 catalyst relative to NiO. Furthermore, by systematically comparing the spectra of the Ni 3 MnO 4 catalyst to reference measurements on NiO, β-Ni(OH) 2 , β-NiOOH, and γ-NiOOH, we are able to provide experimental evidence for the nickel phase, which is present during electrocatalysis.

■ EXPERIMENTAL SECTION
Synthesis. For the synthesis of exfoliated graphite (EG) nanosheets, graphite powder (Merck, >99.5%) was added to N,Ndimethyl formamide (DMF) to achieve a concentration of 0.75 mg mL −1 and then the mixture was treated in an ultrasonication bath (Sonorex Super RK 510, power = 400 W, frequency = 35 kHz) for 3 h. To prevent overheating, the water in the sonication bath was maintained at a temperature lower than 30°C. The resulting suspension was centrifuged at 5000 rpm for 10 min to remove the nonexfoliated graphite agglomerates. The supernatant was a stable suspension, which was recovered for further use in the synthesis of the EG composites.
For the synthesis of nanoparticles, a modified version of the reverse micellar method based on Driess et al. and Fukuzumi et al. was selected. 35,36 To increase the sample conductivity and stability of the

ACS Applied Materials & Interfaces
Research Article active phase on the electrode, the catalyst particles were supported on exfoliated graphene. The selection of this support material was based on studies by Song et al. where additionally, the doping effect of transition metals to the NiOOH phase was studied. 37,38 For a 10 wt % Ni 3 MnO 4 /graphene sample, three identical microemulsions were prepared by mixing 0.017 g of cetyltrimethyl ammonium bromide (CTAB) in 290 mL of 1-hexanol and 180 mL of hexane solution. In this mixture, CTAB, 1-hexanol, and hexane served as the surfactant, cosurfactant, and hydrophobic phase, respectively. The microemulsion solutions were prepared by addition of each of the surfactant-containing mixtures to an aqueous solution of 0.1 M nickel acetate, 0.1 M manganese acetate, and 0.1 M ammonium oxalate. The three microemulsions were then mixed and stirred for 16 h at room temperature. Then, while stirring, the 0.2 g graphene support powder was gradually added to the resulting green emulsion. To precipitate the nanoparticles on the graphene, 4 mL of tetrahydrofuran (THF, Sigma 99.99%) was added to the sample and left stirring for 16 h. Finally, the black precipitate was centrifuged (at 5000 rpm), washed with a 1:1 mixture of chloroform and methanol, and subsequently dried at 60°C for 16 h.
The sample was dispersed in ethanol and drop-casted directly on the Cr/Au-coated Si 3 N 4 membrane electrode. Finally, the loaded electrode was heated to 310°C at a rate of 2°C/min in N 2 , kept at 310°C for 16 h in a tubular oven to calcine the samples, and then cooled down to room temperature. γ-NiOOH was prepared in the same manner using NiO/graphene. Prior to measurements, the electrode was poised to 1.47 V (vs RHE) for 20 min. In addition, the reference samples used in this work were β-NiOOH (Nanoshel, 99.99%), NiO/graphene (Reverse Micellar), MnO/graphene (Reverse Micellar), and β-Ni(OH) 2 (Sigma Aldrich). All reference samples were measured as powders on a carbon tape.
(S)TEM Characterization. High-resolution transmission electron microscopy (HR-TEM) was performed to determine the morphology, size, and distribution of the graphene-supported nanoparticles by using an FEI microscope with an acceleration voltage of 200 kV. The samples were prepared by immersing the Cu-coated TEM grid into the catalyst powder.
The local chemical composition and elemental distribution were investigated by scanning transmission electron microscopy coupled to energy-dispersive X-ray spectroscopy (STEM-EDX). This microscope has a double Cs-corrected JEOL JEM-ARM200CF scanning transmission electron microscope with a cold field emission gun (FEG). The instruments were operated at 200 kV. EDX maps were recorded by using a silicon drift EDX detector.
XRD. The X-ray diffraction (XRD) measurements were performed in Bragg−Brentano geometry on a Bruker AXS D8 Advance II theta/ theta diffractometer, using Ni-filtered Cu Kα radiation and a positionsensitive energy-dispersive LynxEye silicon strip detector. The sample powder was filled into the recess of a cup-shaped sample holder, the surface of the powder bed being flush with the sample holder. The XRD patterns were recorded in the order of 10°< 2θ < 140°. N 2 Physisorption. The surface area and pore volume determination were carried out in a volumetric N 2 physisorption setup (Autosorb-6-B Quantachrome). Nitrogen adsorption/desorption isotherms were determined at −196°C after degassing the sample at 200°C for 2 h prior to the measurement.
Electrochemistry. The electrochemical measurements were performed on an Autolab PGSTAT204 potentiostat equipped with a three-electrode setup. A glassy carbon disk (4 mm diameter) was used as the working electrode and polished to a mirror finish using 1 mm alumina (Buheler), a platinum wire was used as the counter electrode, and the reference electrode was a Hg/HgO 1 M KOH (CH Instruments). The electrode rotation rate was controlled with an Autolab rotator. Metal traces were removed from KOH solutions using Chelex 100 chelating resin (Bio Rad). The solutions were stirred overnight and then filtered. The resulting KOH solutions did not show any detectable Fe by inductively coupled plasma-optical emission spectrometry (ICP-OES) analysis. All glassware, electrochemical cell, and electrodes were cleaned with aqua regia and rinsed with Milli-Q water before each measurement. All potentials in this paper are reported versus the reversible hydrogen electrode (RHE) at the working pH, and no iR correction has been applied. The potentials were converted following this equation E RHE = E Hg/HgO + E Hg/HgO 0 + 0.059pH, where E Hg/HgO is the experimentally applied potential and E Hg/HgO 0 is the standard potential of the reference electrode. The standard potential of the electrode was determined experimentally by measuring the open circuit potential (OCP) on a Pt-working electrode using the same electrochemical setup and bubbling the 0.1 M NaOH solution with H 2 at 1 bar ( Figure S1). An OCP of −0.882 V versus Hg/HgO was determined, which gave E Hg/HgO 0 = 0.115 V. Glassy carbon electrodes were prepared by dropcasting 5 μL of a freshly sonicated suspension of 2.5 mg of calcined graphene/metal oxide particles in 1 mL of Milli-Q water. The electrode was left overnight to dry. We also note that due to the high activity of the catalyst, there were challenges with the losing catalyst from the support over the long time course of the measurement, which is mainly caused by the formation of O 2 bubbles in both the internal and the external structure of the catalyst. Hence, while the currents were measured during the RIXS measurements, these cannot be used in a quantitative fashion and simply demonstrate that we are taking measurements on a catalyst that is under operating conditions.
Ex Situ X-Ray Absorption Spectroscopy (XAS) and Operando 2p3d RIXS Spectroscopy. Both ex situ reference and operando Ni L-edge XAS and 2p3d RIXS measurements were performed at the beamline BL07LSU HORNET end station in SPring-8 synchrotron facilities using an ultrahigh resolution soft X-ray Emission (SXE) spectrometer. 39,40 Furthermore, the incident RIXS excitation energies for both, ex situ and operando, were selected based on an initial XAS measurement. The XAS spectra were the first baseline corrected and then normalized by the L 3 -edge peak maxima. The energy calibration was the same for the ex situ and in situ measurements.
Ex Situ Measurements. The 2p3d RIXS spectra for the NiO reference sample were collected at a scattering angle of 90°with the samples oriented at 45°relative to the incident beam. For all ex situ XAS and RIXS measurements, the NiO and Ni 3 MnO 4 samples are spread as a loose powder on a double coated carbon tape, which was fixed on a Cu holder as reported in previous studies. 34,41,42 Both measurements XAS and RIXS were done at room temperature and in ultrahigh vacuum (10 −8 mbar). The energy transfer (ET) is derived from the emission spectra by subtracting the energy of the incident photons from the energy of the emitted photons. The total energy resolving power (E/ΔE) was more than 2843 (∼330 meV) at 852.9 eV as determined by the full width at half-maximum (fwhm) of a Gaussian fit of the elastic scattering features of NiO. The incident energy of the monochromator was calibrated by setting the Ni L 3edge maximum of the NiO reference to 852.8 eV and collected by a total fluorescence detector. The emission energy was calibrated by using a series of the elastic scattering peaks at corresponding excitation energies (and fitted by a polynomial regression to correct the optical X-ray aberration). In this context, the energy transfer (ET) is derived from the emission spectra by subtracting the energy of the incident photons from the energy of the emitted photons. A linear horizontal polarized beam was selected and focused to a spot size of To avoid chances of sample damage due to beam effects, a raster-scanning protocol was followed as previously used to obtain damage-free RIXS spectra of vanadium and iron complexes. 34,41,42 To be consistent in the method for molecular 2p3d RIXS spectroscopy and avoid chances of sample damage due to the high flux of BL07LSU HORNET, all the reference samples as well as the Ni 3 MnO 4 catalyst were intensively tested for signs of beam damage. 43 However, RIXS measurements for scan speeds of 0, 50, 150, 250, and 300 μm/s have not shown an experimental difference in spectra measured at the L 3 maxima. Consequently, NiO and Ni 3 MnO 4 samples were measured on the same spot for one selected incident energy.
Operando Measurements. The operando measurements were performed by a vacuum compatible membrane electrode assembly (MEA) unit, which consists of the stainless steel manipulator and the electrochemical cell. The detailed mechanical layout of this unit is

ACS Applied Materials & Interfaces
Research Article designed by Harada and co-workers and shown in Figure S2. 40,44 The employed liquid cell has been designed to the flow electrolyte and accommodates a counter electrode (Pt wire) and a reference electrode (Ag/AgCl 3 M). 30 A silicon nitride (Si 3 N 4 ) membrane was chosen as a window material of the liquid cell. 44 The vacuum environment (∼10 −6 mbar) of the spectrometer is separated by a 150 nm-thick Si 3 N 4 membrane (NTT Advanced Technology Co.) from the ambient pressure of the liquid electrochemical cell, while ensuring high transmittance of soft X-rays. Incident and emitted photons at ∼853 eV oblique to the 150 nmthick Si 3 N 4 membrane surface have 85% transmission. However, the high flux requires to measure pointwise, otherwise the beam will burn a hole in the Si 3 N 4 window and break the vacuum. The penetration time per spot was 4 min.
In more detail, an O-ring, a 150 nm-thin Au-coated Si 3 N 4 membrane, and PTFE gasket sheets are sandwiched between a stainless steel vacuum flange and the electrochemical cell made of an acrylic resin. The cell has two gas/liquid inlet/outlet channels and two guides for the electrodes. The PTFE gasket sheets individually seal the liquid phase, and the deposited Au on the electrolyte side of the Si 3 N 4 membrane operates as a working electrode. The redox potential between both electrodes was regulated and kept constant during measurements in the operando OER experiment, and the catalytic current was simultaneously measured by a potentiostat/galvanostat (P/G stat: VersaSTAT4-400). The cell mounted on the flange is attached to a three-dimensional adjustable manipulator to locate the sample on the focal point. The detailed optical layout of the SXE station is described elsewhere. 40,44 The Ni 3 MnO 4 catalyst sample was first dispersed in ethanol solution and following a sonication step drop-casted on the Au-coated Si 3 N 4 membrane. The deposited layer was analyzed by a Keyence VHX-5000 digital microscope. Before starting the 2p3d RIXS experiment, the L-edge X-ray absorption spectra of the dry/ nonactivated Ni 3 MnO 4 /graphene and NiO/graphene samples in the electrochemical cell were collected.
During the subsequent steps of the experiment, the liquid cell was filled with 0.1 M KOH solution and the changes in the electronic structure of the Ni site was measured by Ni L-edge total fluorescence yield (TFY) XAS and Ni 2p3d RIXS. The absorption was measured in the 850−877 eV energy range, whereas for the RIXS part, the emission from the L 3 -edge main absorption features was collected. During this in situ experiment, first, the reactor cell was filled with the basic solution and left at the open circuit potential (OCP). Next, a CV was measured and the sample was treated at three selected potentials: (I) 1.25 V, (II) 1.47 V, and (III) 1.55 V (vs RHE). In addition to the absorption spectra, Ni 2p3d RIXS planes were also collected for the aforementioned voltages during the electrochemical OER experiment. The Ni absorption spectra were normalized to the maximum of the L 3 -edge peak. SEM/TEM. The morphology of the graphene-supported Ni 3 MnO 4 catalyst was characterized by scanning electron microscopy (SEM) and transmission electron microscopy (TEM) as shown in Figure 4A,B. The recorded images show the macrostructure made of dense spherical/elliptical nanoparticles with sizes of about 4−6 nm. For the pure Ni 3 MnO 4 particle in this range, the estimated X-ray transmission for particles at 800 eV is >90% [CXRO]. Therefore, the incident X-rays effectively probe the entire particle. The d-spacing is slightly larger than that of a pure NiO crystal, which is attributed to the doping of manganese in the crystal structure. 45 The enhancement of diffraction spots indicates the existence of the structural texture in the material. The small nanoparticles within the mesocrystal show oriented attachment. By using higher magnification, the lattice spacings were calculated to be consistent with the d-spacings of bunsenite [JCPDS 71-1179].
Both SEM and TEM images confirmed a homogeneous dispersion on the surface of the graphene support. Also, the corresponding energy-dispersive X-ray (EDX) analysis on various particles indicated a general ratio of ∼3:1 for Ni/Mn. This data also confirms a homogeneous distribution of both Ni and Mn on the surface of the graphene support.
In addition, the high-resolution transmission electron microscopy (HRTEM) images ( Figure 4C,D) also reveals that the particles have oriented attachment.
Subsequently, electron energy loss spectroscopy (EELS) measurements were also performed to define and confirm the elemental ratios obtained by EDX data. By calculating the L 3 / L 2 ratios (area/area) from EELS data ( Figure S3), the

ACS Applied Materials & Interfaces
Research Article oxidation state of the Ni and Mn was determined. 46 The L 3 /L 2 ratio for Mn is ∼2.76, which is in between the 2.5 (Mn 2 O 3 ) and 2.9 (Mn 3 O 4 ) values for the reference samples. The interval of Mn-L 3 and Mn-L 2 is about 11 eV, which fits to that of Mn 2 O 3 and Mn 3 O 4 . 47 These results indicate that the average oxidation state of manganese in this catalyst is between 2.7+ and 3+. These data confirm that the manganese oxide compound is a mixed valence 2+ and 3+ state, consistent with a hausmannite-like (Mn 2+ Mn 3+ 2 O 4 ) structure. Based on these data, we propose a general formula for this compound as Ni 2+ 9 (Mn 3+ 2 Mn 2+ )O 23− 12.5 , which in turn is simplified and written as Ni 3 MnO 4.1 , which for simplicity is written as Ni 3 MnO 4 throughout. Electrocatalytic Characterization. The activity was analyzed by cyclic voltammetry (CV). CV shows a redox couple at 1.4 V, which has been previously proposed as the NiO oxidation to form the γ-NiOOH (NiO + OH − + e − → γ-NiOOH), 18,48 an observation that will be confirmed by the spectroelectrochemical measurements presented in this manuscript (see below). Following this oxidation, a catalytic oxidation current appears at E > 1.5 V, corresponding to the O 2 evolution reaction ( Figure 5 and Figure S5). A control electrode where 0.1 mg/mL graphene was deposited on the electrode did not show any catalytic currents in the measured potential range. It has been reported that the catalytic current increases upon repetitive cycling 18 or after anodization treatments. 15 To obtain consistent comparable results, we conditioned the electrodes by running 50 cycles between 1 and 1.55 V at 100 mV/s, which resulted in an average 20% increase of the measured catalytic current at 1.7 V for the Ni 3 MnO 4 particles, while for pure NiO particles, there was barely any catalytic current increase (when using purified "Fe-free" KOH). The stability of the electrodes was adequate for the spectroelectrochemical experiments and very similar for both oxides, with less than a 10% decay of the catalytic current at E = 1.7 V during 2 h. It is important to note that electrodes with higher material loading led to much less stable catalytic currents, probably due to poor adhesion of the particles to the electrode surface.
From the CVs shown in Figure 5, Ni 3 MnO 4 and NiO show very similar catalytic currents, while in this potential range, MnO does not display any significant catalytic current. When using unpurified KOH, Ni 3 MnO 4 and NiO display almost identical catalytic behavior, with lower overpotential and an approximate 20−30% increase of catalytic current (not shown). This observation is consistent with previous reports that have indicated that doping NiO with Fe results in a much more efficient catalyst. 49,50 This is evidenced by increased Tafel slopes for our measurements in comparison to what has been previously reported for similar catalysts ( Figure S5). 35 The intensity of the precatalytic redox signal is significantly lower for NiO, which suggests a lower electrode coverage for this material. Integration of the charge under this signal can be used to obtain an estimation of the solvent accessible to Ni centers on the material. This gave a value of 2 × 10 −8 mol cm −2 for Ni 3 MnO 4 and 7 × 10 −9 mol cm −2 for NiO. If we consider this number as the total surface electroactive Ni atoms participating in electrocatalysis, we can estimate the turnover frequency (TOF) for each Ni atom using the following expression 51 where i is the catalytic current, n = 4 electrons for H 2 O oxidation to O 2 , F is Faraday's constant, Γ is the concentration of electroactive Ni, and A is the electrode area. Figure S4 shows a plot of the TOF versus E for both materials, which highlights the lower catalytic performance of Ni 3 MnO 4 with respect to NiO, with almost one-third of the TOF for

ACS Applied Materials & Interfaces
Research Article charge-transfer (LMCT) processes. 52−59 Therefore, the 2p to 3d dipole transition transforms the ground state, which consists of the 2p 6 3d 8 + 2p 6 3d 9 LMCT configuration to final state configurations with 2p 5 3d 9 + 2p 5 3d 10 LMCT according to Ishii et al. 52 There are subtle differences on the high-energy side of the L 3 -edge, suggesting a modulation in the LMCT contributions of Ni 3 MnO 4 relative to NiO. In addition, the L 2edge spectra of Ni 3 MnO 4 show an inverted intensity ratio of the two absorption features at 869.9 and 871.1 eV relative to NiO. In the case of the dry Ni 3 MnO 4 , the 871 eV peak is more intense, whereas for the NiO, the 869.9 eV feature is higher in intensity. These data thus indicate that although the Ni sites are similar in Ni 3 MnO 4 and NiO, the presence of the Mn has subtly altered the electronic structure at the Ni site.
Ni 2p3d RIXS of NiO versus Fresh Ni 3 MnO 4 / Graphene. To obtain more detailed insights into the changes in an electronic structure upon incorporation of Mn into the nickel oxide-based material, we also measured Ni 2p3d RIXS. In this experiment, the features, which appear on the energy transfer (ET) axis, correspond to d−d excitations (at energies <∼3 eV) as well as charge-transfer states (to higher energies) and thus provide a detailed map of the changes in the electronic structure and local coordination environment. The Ni 2p3d resonant inelastic X-ray scattering (RIXS) spectra of Ni 3 MnO 4 and NiO were collected at the L 3 -edge main absorption features, that is, 852.8, 854.7, 856.0, and 858.8 eV. The RIXS emission slices from the predetermined XAS peaks were converted to their respective RIXS energy transfer spectra and are plotted in  Figure 7). The peak at ET = 0 eV corresponds to the elastic recombination. The dominant features on the energy transfer axis appear at 1.1 eV (A) with a shoulder at 1.6 eV (B), a less intense feature at 3.0 eV (C), and two broad chargetransfer peaks between 4 and 10 eV (D and E). Based on a previous study, feature A is assigned to the 3 A 2g → 3 T 2g transition. 60 Feature B was assigned as 3 A 2g → 3 T 1g and 3 A 2g → 1 E g spin-flip transitions. Feature C mainly originates from the triplet to singlet spin-flip ( 3 A 2g → 1 T 2g ) transition with a small contribution from the 3 A 2g → 3 T 1g transition. 52,60 For both samples, feature A has its highest intensity at the 852.8 eV excitation energy, while upon increasing the incident energy to 854.7 eV, feature A is suppressed and feature C gains intensity. This change in intensity of the Raman features indicates that the lowest energy feature A is dominated by the triplet spin state and feature C is dominated by the singlet spin state. This implies that feature C likely derives from a spin-flip transition, which is enhanced by increasing the excitation energy. These assignments are fully consistent with the reports of Ghiringhelli et al. 60 and Braicovich et al. 61 Upon increasing the excitation energy to 856 and 858.8 eV (excitation into the CT region of the L-edge spectrum), features D and E increase in intensity, consistent with these features being assigned at LMCT features.
The similarity in the ET spectra for Ni 3 MnO 4 and NiO is consistent with the very similar L-edge XAS spectra and the XRD diffraction patterns. There are again only subtle differences in the spectra, namely, the low energy ET features in Ni 3 MnO 4 , which are attributed to d-to-d transitions, are shifted by ∼0.1 eV to higher energies with respect to NiO. Further, we note that the LMCT features have moved to lower energy (by up to ∼0.65 eV) in the Ni 3 MnO 4 catalysts relative to NiO. This may suggest a slightly less covalent interaction of Ni with the surrounding oxygens in Ni 3 MnO 4 and is also consistent with the slight decrease in the L 3 -edge XAS energy of Ni 3 MnO 4 relative to NiO. 62 Operando OER of Ni 3 MnO 4 . Having identified subtle differences between Ni 3 MnO 4 and NiO in the solid state, we now turn our focus to operando study of Ni 3 MnO 4 . We first focus on the changes in the Ni L-edge XAS and then proceed to the 2p3d RIXS.
Ni L-Edge XAS of Operando OER. To prepare for operando catalysis, a 0.1 M KOH solution was added to the dry Ni 3 MnO 4 catalyst (Figure 8, top). It is evident from the spectrum (in purple) that upon wetting, the 854.6 eV feature shifts by +0.25 to 854.85 eV and increases dramatically in intensity. In addition, the feature at ∼853 eV shifts up slightly (by +0.1 eV) in energy. Based on previous Ni L-edge XAS studies of nickel oxides, the feature at ∼855 eV likely results from changes in the local ligand field parameters and may also have the partial ligand to metal charge-transfer contributions. 52,58,60,61 Hence, this feature will be sensitive to changes in the Ni coordination environment. We hypothesize that upon wetting in 0.1 M KOH solution, a partial transformation of NiO to β-Ni(OH) 2 occurs. Figure 8 (bottom panel) shows the reference spectrum for β-Ni(OH) 2 , which also exhibits a prominent absorption feature at ∼855 eV. We note that the observed spectrum for Ni 3 MnO 4 in KOH is readily simulated by a 50:50 mixture of NiO and β-Ni(OH) 2 ( Figure S6), thus supporting this interpretation. The simulated spectrum was obtained by taking a linear combination of the NiO and the β-Ni(OH) 2 L-edge spectra.
In the next step, 1.25 V was applied, which represents the potential regime before the appearance of the anionic peak. At

ACS Applied Materials & Interfaces
Research Article this voltage, the changes in the L 3 -edge spectrum (in green) are subtle, with only slight shifts of the L 3 -edge features (by ∼0.04−0.1 eV) to higher energy.
Upon increasing the potential to 1.47 V and subsequently 1.55 V (orange and red spectra, respectively), striking changes in both the Ni L 3 -and L 2 -edges, the spectral shape and position are observed. At 1.47 V, the applied potential is already higher than the redox peak and the oxidation of the Ni(II) site to either Ni(III) or Ni(IV) is predicted. The L 3edge spectrum in Figure 8 (top) has shifted significantly up in energy by +2.45 eV with a dominant peak at 855.25 eV and a shoulder at 853.2 eV. At the L 2 -edge, however, the two-peak structure has collapsed into a single broad structure and is shifted by +1.0 to 872.25 eV. These shifts are consistent with the formation of Ni(III) and/or a mixture of Ni(III) and Ni(IV). We note that although formal Ni(IV) is often invoked in the literature, 30 in the oxide material, it may be the case that the hole character resides on the oxygen. By comparing both the L 3 -and L 2 -edge spectra for the 1.47 and 1.55 V treated samples with that of reference spectra (Figure 8 bottom), it appears that the catalyst is now dominantly in a γ-NiOOH phase, thus confirming the presence of this phase under the applied OER conditions. There is, however, a more intense shoulder at ∼853 eV in the Ni 3 MnO 4 catalyst at 1.55 V as compared to the γ-NiOOH reference. This may correspond to the catalytic wet precursor that has not been fully converted. To test this hypothesis, different ratios of wet Ni 3 MnO 4 and γ-NiOOH spectra were linearly averaged and compared to the catalyst data at 1.55 V ( Figure S7). The best agreement with the experiment was obtained with a 20:80 mixture of the wet Ni 3 MnO 4 and γ-NiOOH ( Figure S8). We note however that based on the Ni L-edge XAS data alone, one could also argue that an admixture of β-NiOOH and γ-NiOOH is present. The RIXS analysis (vide infra), however, supports the presence of the unconverted Ni(II) precursor. Importantly, the Ni L-edge XAS data clearly indicate that the γ-NiOOH phase dominates upon the onset of the catalytic oxidation current.
Ni 2p3d RIXS of Operando OER. In addition to the operando Ni L-edge XAS measurements, operando 2p3d RIXS data were also obtained for the Ni 3 MnO 4 catalyst. ET spectra were measured at excitation energies corresponding to the main L 3 -edge absorption positions: 852.8 ( Figure 9A), 854.7 ( Figure 9B), and 856.0 ( Figure 9C). At the lowest excitation energy (852.8 eV), the spectra are dominated by d−d

ACS Applied Materials & Interfaces
Research Article transitions. 53 However, relative intensities between elastic and inelastic features changed dramatically under working conditions. In this context, the elastic line gains intensity and becomes broader with increasing potential. Upon wetting Ni 3 MnO 4 , the lowest energy ET features all move up in energy (only by ∼0.05 eV), suggesting a slight increase in the local ligand field at Ni. Upon applying a potential of 1.25 V, essentially no changes in the ET spectra are observed, consistent with the nearly identical L-edge spectra. However, ongoing to 1.47 and 1.55 V again, very similar ET spectra to the wet Ni 3 MnO 4 spectrum are observed. This may suggest at this excitation energy (852.8 eV) that we are largely selecting for unconverted Ni(II) and hence the spectrum remains dominated by the same features as the precursor.
In contrast to the ET spectra obtained at 852.8 eV, the emission slices from the 854.7 eV ( Figure 9B) and 856 eV ( Figure 9C) incident energies show more dramatic changes as a function of applied potential. The reason for this is twofold: (1) the higher resonant energies select dominantly for the higher-valent Ni species, and (2) higher incident excitation energies also increase the charge-transfer contributions to the ET spectra.
To understand the extent to which 2p3d RIXS can select for a given oxidation state, it is illustrative to examine the changes that occur in the low-energy d−d transitions for the highpotential (at 1.55 V) case as a function of excitation energy (Figure 10). At 1.55 V applied potential, the 852.8 eV excitation gives rise to ET peaks that are almost identical to the wet Ni 3 MnO 4 (1.1, 1.6, and 1.8 eV), suggesting that the observed species may be due to the presence of the unconverted Ni(II) precursor. However, upon increasing the excitation energy to 854.7 eV, the peak at 1.6 eV is enhanced and new features to lower (1.5 eV) and higher energy (1.7 eV) are observed ( Figure 10). Upon further increasing the excitation energy to 856.0 eV, the low-energy features remain constant in energy, consistent with these features arising from d−d excitation of the high-valent Ni species. These data thus show that the γ-NiOOH phase electronic structure can be uniquely probed by increasing the excitation energy to its specific L-edge XAS peak position. We note, however, that as γ-NiOOH is generally believed to be a mixture of Ni(III) and Ni(IV), a clear assignment of the multiplets is not possible. It is nonetheless interesting to note that the d−d transitions for the Ni 3 MnO 4 precursor and the high-valent species generated during electrocatalysis are in close energetic proximity. As the low-lying states govern reactivity, their energetic accessibility may be a key feature in optimizing a metal oxide for catalysis.
Finally, we note that there are clear changes in the energy of the LMCT features in the ∼4−5 eV range under operando conditions. The decrease in the energy of the charge-transfer features upon increasing potential is consistent with increasing covalency upon oxidation of the Ni site.

■ CONCLUSIONS
Herein, the synthesis and structural characterization of a Ni 3 MnO 4 /graphene OER catalyst has been reported. The electrocatalytic characterization shows that the Ni 3 MnO 4 catalyst shows an ∼20−30% decrease in catalytic current and almost one-third of the TOF on NiO. Operando Ni L-edge XAS and 2p3d RIXS were utilized to follow the evolution of the Ni site during the activation and OER conditions.
The proposed transformations at the Ni site are best illustrated schematically in Figure 11. Namely, we propose that first, the partial hydration of the NiO-like sites occurs during the wetting in 0.1 M KOH.
The comparison of the Ni L-edge XAS and 2p3d RIXS data of NiO and the dry Ni 3 MnO 4 shows only modest perturbations in the electronic structure, suggesting that electronic structural perturbations are likely not the primary driving force for the decreased catalytic activity. Operando Ni L-edge XAS and 2p3d RIXS allow for the electronic structural changes, which occur during electrocatalysis to be revealed. The comparison of the Ni L-edge XAS data to the measured reference complexes supports the formation of γ-NiOOH during catalysis, with some (∼20%) unconverted Ni(II) precursor present, which manifests as a low energy shoulder in the Ni L-edge XAS spectrum. Furthermore, 2p3d RIXS data were key in identifying the nature of the low-energy shoulder in the L-edge XAS spectra. By resonantly exciting into this feature, energy transfer spectra were obtained that were essentially identical to that of the Ni(II) catalytic precursor. We note that while previous Ni XAS studies provided evidence for the formation of Ni(IV) during electrocatalysis, 51,53 based on the XAS data alone, one cannot rule out that an admixture of β-NiOOH and γ-NiOOH is present. Hence, the 2p3d RIXS data presented here are key in establishing that the γ-NiOOH is mechanistically relevant. Further, the present data support the notion that a high-valent nickel population is required for the efficient mediation of OER.
Finally, the 2p3d RIXS data show that enhanced selectivity for the d−d excited states of the high-valent active phase may be achieved by tuning the excitation energy to higher energies. Interestingly, the d−d transitions for the Ni(II) and Ni(III)/ Ni(IV) species are in close energetic proximity, perhaps providing a mechanism for facile redox chemistry and catalysis.

ACS Applied Materials & Interfaces
Research Article ■ ASSOCIATED CONTENT

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsami.9b06752.
(1) The determination of the standard potential, (2) STEM-EELS analysis, (3) electrochemical OER of NiO vs Ni 3 MnO 4 , and (4) Ni speciation following the wetting step and after the onset of catalysis (PDF)