Analyzing the Local Electronic Structure of Co3O4 Using 2p3d Resonant Inelastic X-ray Scattering.

We present the cobalt 2p3d resonant inelastic X-ray scattering (RIXS) spectra of Co3O4. Guided by multiplet simulation, the excited states at 0.5 and 1.3 eV can be identified as the 4 T 2 excited state of the tetrahedral Co2+ and the 3 T 2g excited state of the octahedral Co3+, respectively. The ground states of Co2+ and Co3+ sites are determined to be high-spin 4 A 2(T d ) and low-spin 1 A 1g (Oh ), respectively. It indicates that the high-spin Co2+ is the magnetically active site in Co3O4. Additionally, the ligand-to-metal charge transfer analysis shows strong orbital hybridization between the cobalt and oxygen ions at the Co3+ site, while the hybridization is weak at the Co2+ site.

Introduction

Electronic properties of 3d transition-metal (TM) oxides are governed by the interactions between the TM ions and the neighboring contributions. Those interactions are determined by Hund’s coupling, crystal-field splitting, spin–orbit coupling, and the interatomic hybridization (covalent bonding). The determination of the local electronic structure is a relatively straightforward task for localized materials, but it faces an experimental challenge for a mixed-valence TM compound, especially when the constituting TM ions have a rich overlapping multiplet structure, which is the case for Co3O4.

Co3O4 crystallizes in a normal spinel structure (AB2O4) below 850 K.[1,2] The Co ions in the A sites with tetrahedral (T) symmetry are divalent (Co2+), while those in the B sites with octahedral (O) symmetry are trivalent (Co3+) with a small trigonal distortion. An experimental determination of the electronic structure of two inequivalent Co sites is of fundamental importance to elucidate the magnetic coupling responsible for the antiferromagnetic order (below ∼40 K).[3,4] The Co-site dependence of covalency is crucial for large exchange anisotropies achieved by a substitution of Ni or Fe for Co ions and its application to magneto-optical information storage.[5−7] The covalency also influences the charge capacity. Co3O4 shows a high charge capacity (700 mAh/g) and a good cycle performance of nanosize negative electrodes, which makes Co3O4 an interesting compound for the lithium battery applications.[8,9] Moreover, the role of Co2+ ions in the oxygen evolution reaction step of water oxidation is debated.[10−17] A small band gap was found in Co3O4 that is important for the application in a photovoltaic cell,[18−20] where the electronic structures of individual sites potentially plays an important role.

The electronic structure of Co3O4 has been characterized by the ultraviolet/visible (UV/vis)[21−25] and near-infrared (NIR) absorption.[4,26] These absorption spectra show excitations at 0.82, 0.93, 1.64, and 2.81 eV,[22] while diverse interpretations for these excitations are proposed,[4,10,21−25] partly related to the fact that the dd excitations are usually dipole forbidden.[27] Hibberd et al. revealed distinct absorption peaks for the Co valence (2+ or 3+) and local environment (O or T) of Co3O4 using Co L2,3-edge (2p) X-ray absorption spectroscopy (XAS).[28] However, due to possible overlap of the multiple Co-site signals and the lifetime broadening (∼200 meV), the details of the local electronic structure are limited.

In this work, we study the Co-site resolved low-energy local excitations in Co3O4 using Co 2p3d resonant inelastic X-ray scattering (RIXS). By setting the resonant photon energy to the distinct Co L3-XAS features of the Co2+ and Co3+ sites,[28] the site-selective local excitations are measured. The site-selectivity of RIXS has been discussed in literature.[29−31] The (photon-in–photon-out) transitions in RIXS brings a broad sensitivity to dd excitations of the system that provides a better determination of the local electronic structure including the orbital hybridization effects.[32−36] This has also been demonstrated using 1s3d RIXS on Co3O4,[37,38] but the intensity was limited by the absorption cross-section of the quadrupole 1s3d transition. We show that the Co 2p3d RIXS is a good chemical site-selective approach to study accurate details of the ground state and the core excited states.

Methodology

Experimental Details and Sample Preparation

The experiments were performed on 99.9985% Co3O4 powder produced by Alfa Aesar. A 10 mm × 0.5 mm (diameter × height) cylindrical pellet Co3O4 was prepared for the measurements. Co 2p XAS spectra were acquired at the soft X-ray WERA beamline of the Karlsruhe Research Accelerator (KARA) synchrotron in Germany. The instrumental resolution was calibrated to be ∼280 meV full width at half-maximum (fwhm) at the Co L2,3 edge (∼780 eV). Both the total electron yield (TEY) and the fluorescence yield (FY) methods were employed. The Co 2p3d RIXS measurements with linearly vertical (V) and horizontal (H) polarized incident X-rays were performed at the Taiwan Light Source (TLS) beamline 05A of National Synchrotron Radiation Research Center (NSRRC) in Taiwan.[39] The experimental energy resolution of the incident photon was ∼700 meV and the combined resolution of RIXS was ∼90 meV. A grazing incident geometry (∼20°) with the spectrometer at 90° was used. Given the energy-compensation principle, the RIXS resolution is much better than the incident photon resolution.[39,40] For a perfect monochromator-spectrometer pair, the photons will be focused at the mirror position of the photon source. As soon as the photon lost its energy in the path, the focus will not be kept at the same position. Thus, using a position sensitive detector combined with the active monochromator-spectrometer system, the RIXS resolution can be decoupled from the incident photon resolution. To compare the XAS spectra measured at different facilities, the background signals were subtracted from the spectra, as described in the Supporting Information (SI). The subtracted spectra were normalized to the maximum of the Co L3 edge. We calibrated the photon energy of the RIXS to the absorption spectra acquired at the WERA beamline. The RIXS spectra measured with horizontal (H) polarization were normalized to the exposure time. Then the spectra with vertical (V) polarization were normalized to the H-polarization spectra according to the profile at high energies (above 2.5 eV). The measurements at the KARA-WERA and TLS-05A beamlines were carried out at 50 and 150 K, respectively. We note that the simulated intensity is given as absolutely cross-section, while the experimental intensity is given in normalized units.

Theory

The results were analyzed using a cluster model that includes the Coulomb multiplet interaction, the crystal-field splitting, the spin–orbit coupling, and the charge transfer between Co 3d and O 2p orbitals. The calculation was carried out with the program Quanty, which implements a configuration-interaction scheme.[41,42] To study the mixed-valence spectra of Co3O4, the Co2+(T) site and the Co3+(O) site were calculated separately. We took the initial parameters from literature for a high-spin Co2+ ground state (4B1 in D2 symmetry) and a low-spin Co3+ ground state (1A1 in O symmetry).[43−45] The values were refined by a detailed comparison with the high-resolution RIXS data and its polarization dependence (section 3.2). To evaluate the used values, we compared the ones estimated by an ab initio calculation for Co3O4. The ab initio calculation was based on a density-functional calculation with the local-density approximation (LDA). The parameters (crystal field and hopping parameters) were extracted from the tight-binding model which spans the Co 3d and O 2p bands by a Wannier projection.

Results

2p3d RIXS Experimental Results

The experimental Co 2p XAS and 2p3d RIXS results are presented in Figure . The TEY spectrum shows sharp features at 778.5 and 780.2 eV, which are the characteristic signals of the Co2+ and the Co3+ sites in Co3O4, respectively.[28] The FY spectrum, on the other hand, is rather broad, and the features are obscure. This is due to strong saturation and self-absorption effects for a bulk sample. In the 2p3d RIXS result (Figure b), a fluorescence-like signal increases with incident photon energies as guided by a red line. Sharp features at ∼0.0, 0.5, 0.9, 1.2, and 1.9 eV are observed and labeled from A to E, respectively. The features A–C are resonantly enhanced at 778 eV, while feature D is enhanced at 780 eV. Feature E resonates at 778.7 eV. These features are attributed to local excitations of the Co2+ and Co3+ sites, as we will discuss later.

Figure 1

Experimental results. (a) The H-polarization 2p XAS spectra of Co3O4. (b) The H-polarization 2p3d RIXS energy map of Co3O4.

To gain more insight, Figure a compares the H- and V-polarization energy loss spectra at three different incident energies. At 780.0 eV, feature C (at ∼0.9 eV) is enhanced by the V-polarization, while feature E (at ∼1.9 eV) shows opposite behavior. This polarization dependence resembles the 3T1 and 1T1 excited states in LaCoO3 at 20 K, where the Co3+ ions have the 1A1 ground state, see Figure a,b. The energy differences indicate that the crystal field splitting varies between the two. Feature B (at ∼0.5 eV) is identified as the 4T2 excited state of the Co2+(T) site.[5] As a reference, divalent 4B1(D2 symmetry) Co2+ RIXS at the L3 edge of K5H[CoW12O40]·xH2O is shown in Figure b,[44] where the local structure is a distorted tetrahedral Co2+. The distortion changes the ground state symmetry from the 4A2(T) to the 4B1(D2) that gives rise to a strong polarization dependence at the 0.5 eV peak in the RIXS spectra. However, the feature B shows no dichroism in Co3O4, which suggests that the distortion is negligibly small on the Co2+ site and supports the 4A2(T) symmetry of the ground state in the calculation.

Figure 2

Comparison of H- and V-polarization RIXS spectra of (a) the Co3O4 and (b) the 4B1(D2) and 1A1(O) ground states from refs (44 and 45). The blue (green) arrows indicate the characteristic features of the Co2+(Co3+) site. (c) The calculated 2p3d RIXS polarization comparison of distorted and nondistorted tetrahedral Co2+ using parameters in ref (44) without considering the ligand-to-model charge transfer.

Multiplet Simulation Results

Guided by the experimental results, the low-energy local excitations characterize the two Co sites in Co3O4 and the crystal-field parameters can be determined. In the cluster-model analysis, these parameters identify the symmetry of the local site and the crystal-field energy (10Dq) is the key factor to identify the energy splitting of the Co 3d orbitals (t2g(t2) and eg(e) orbitals in O(T) symmetry). For a system involving charge transfer effects, the crystal-field energy can be decomposed into two components:[36] the ionic crystal field energy (10Dqionic) and the effective crystal field energy induced by the charge transfer effect (10DqCT). The combination of the two (10Dqtot) gives the final splitting between the e/eg and t/t2g orbitals. Figure b,c shows the excited states energies as a function of the 10Dqionic for the Co3+ and Co2+ site, respectively. The 2p3d RIXS probed at 780.0 eV shows three features at 0.9, 1.3, and 1.9 eV, indicated with the green arrows in Figure a, which are attributed to the signals at the Co3+ site. The feature at 0.5 eV shows no polarization dependence, which indicates that the feature is likely the tail contribution of the 4T2 feature of the Co2+ site (cf Figure ). As seen in Figure b, the matching of the three features gives 10Dqionic ∼ 1.15 eV at the Co3+ site. We note that this value is smaller than the 10Dqtot value due to the missing of 10DqCT. The 10Dqtot values will be given in Table . Further discussion about the 10Dqtot can be found in the SI. The 10Dqionic value is much larger than the spin-state transition point (10Dqionic ∼ 0.6 eV), suggesting that the ground state on the Co3+ site in Co3O4 is a robust low-spin singlet 1A1. For the Co2+ site, the 10Dqionic ∼ −0.1 eV reproduces the energy position of the characteristic features at 0.5, 0.9, and 1.9 eV in the RIXS data probed at 778.5 eV, which are indicated in Figure a,c. The negative 10Dq value indicates an inversion of the e and t2 manifolds in the T symmetry.

Figure 3

(a) The H-polarization 2p3d RIXS spectra of Co3O4 excited at 778.5 and 780.0 eV. The calculated energy diagrams of (b) the Co3+ ion and (c) the Co2+ ion as a function of ionic crystal field energy (10Dqionic), including the charge transfer.

Table 1

Model Parameters Used in the Simulation (in eV), Which Are the Crystal Field Energy, Hopping Integrals, Charge Transfer Energy, Udd and Upda

	10Dqionic	10Dqtot	Δ	Ve(eg)	Vt2(t2g)	Udd	Upd
Coi2+	–0.10	–0.55	4.5	1.0	2.0	4.5
Com2+	–0.02	–0.47	4.5	1.0	2.0	4.5	6.0
Coi3+	1.15	1.90	1.5	3.12	1.8	6.5
Com3+	0.84	1.59	1.5	3.12	1.8	6.5	7.5

The i and m stand for the configurations of initial ground state and intermediate state, respectively.

Table summarizes the key parameters, and more values can be found in SI. The charge transfer energy Δ is an energy related to electron transfer from a ligand to the Co site and Ve/Vt are the values for electron hopping. The Udd and Upd values parametrize the Coulomb interaction, which are set to reference values.[36,43] Since the ligand-to-metal charge transfer also contributes to the energy splitting of Co 3d states,[36] we provide the 10Dqtot value that is calculated by the resultant energy splitting by including the charge transfer. To simulate a contraction of Co 3d wave functions by the presence of the core hole,[46] the 10Dqtot value in the intermediate state was reduced from that in the ground state by ∼15%, see also SI.

To evaluate the parameters in Table , the values estimated by the LDA calculation are provided. For the Co2+(T) site, the estimated values are −0.10 eV for 10Dqionic and 1.29 (1.82) eV for Ve (Vt). For the Co3+(O) site, the estimated values are 0.7 eV for 10Dqionic and 3.03 (1.74) eV for Ve (Vt, where we omitted the energy splitting within the t2g manifolds due to a small trigonal distortion and estimated the Vt value by averaging the hopping integrals over the two. The Dσ value, which measures the trigonal distortion, is estimated as 0.05 eV. This value is much smaller than the required value (>0.5 eV) for changing the ground state symmetry (singlet 1A1) of the Co3+ site in Co3O4, but gives a minor correction to the multiplet energies. Because the trigonal distortion gives no visible effect on the spectra, we neglect it in our simulation and the site is referred to as O for simplicity. We conclude that the optimized values agree well with the ab initio estimates except for a small discrepancy in the 10Dqionic value at the Co3+ site, which is possibly due to an underestimate of the covalency in the LDA scheme.[42,47]

Figure a,b shows the RIXS calculation of the Co2+ and Co3+ sites with the values listed in Table . Since the low-spin (1A1) ground state on the Co3+ site is angular isotropic, its intensity is strongly suppressed when the scattering angle is about 90° with the H-polarization condition. Thus, the intensity at zero energy loss (elastic line) is mainly due to the Co2+ site with the 4A2 ground state. Figure c shows the sum of the calculated intensities which was corrected by a combination ratio of stoichiometry (1:2) and the number of holes (3:4), where the ratio was also applied to weight the area normalized XAS spectra. The incident energies of Co2+ and Co3+ spectra were adjusted by a 1.8 eV difference in between the main L3 feature to reproduce the experimental Co L3 XAS (cf Figure a). This value agrees with the finding in literature (1.7 eV).[28] The energy shift is not only depending on the oxidation state but also on the spin-state, symmetry (O, respectively, T) and hybridization (charge transfer effect). Some intensity discrepancies are observed, that is, the calculated RIXS at 780.5 eV shows much stronger 1.3, 1.7, and 2.5 eV excited features. These discrepancies can be explained by saturation and self-absorption effects (see section ) and further related to the differential broadening at different site (see section ).

Figure 4

Simulated H-polarization 2p3d RIXS energy maps. (a) Co2+4A2 (T) ground state. (b) Co3+1A1 (O) ground state. (c) The simulation of Co3O4. (d) Experiment of Co3O4.

Figure 5

Comparison of XAS spectra. (a) The comparison between the experimental TEY spectrum and the simulations. The inset panel zooms in the pre-edge region and indicates a small feature before the edge. This small feature matches to the first feature indicated in the simulation. (b) The comparison of the PFY spectra correspond to the excited states A–E.

Discussion

Site Selectivity of Partial Fluorescence Yield Spectra

Partial fluorescence yield (PFY) spectra were acquired for energy transfer of the dd excitations (A–E) from the RIXS map to discuss the local excitations in detail. The experimental data are obtained by fitting the RIXS intensities of these dd excitations (cf Figure d), and the theoretical spectra are obtained by summing simulated intensities in an energy window larger than the width of these excitations in Figure c. Since the features are well separated in the simulation, this summation represents the peak intensity properly.

Before the discussion on the PFY spectra, the TEY result is compared ahead with the simulated spectra. In Figure a, the sum of the simulated XAS of Co2+ (blue) and Co3+ (green) sites well matches with the experimental TEY data (gray bars and red line). The contributions of two sites are separated in the incident energies that enables to observe site-resolved local excitations by RIXS, as we show in Figure . The site selectivity can be better understood from the PFY spectra. In the PFY spectra (Figure b), the features A–B and D are unambiguously attributed to the Co2+ site and the Co3+ site excitations, respectively, while the features C and E show an overlap with both Co2+ and Co3+ regions. We stress that the overlap is essential since, according to our simulation in Figure a,b, both sites have excitations at around 0.9 and 1.9 eV. A systematic discrepancy is observed that the intensity at about 780 eV is always overestimated in the simulation, particularly, of the PFY spectra of the features C and E. To analyze the two features, the saturation and self-absorption effects need to be applied and the correction can be written as[48,49]Here, X, α, and θ refer to the emission edge of element, the scattering angle, and the sample rotation angle, respectively. In this work, the experimental geometry was set to α = 90° and θ = 20°. The μtot(ωin) and μtot(ωout) are the absorption factor of the photon-in and photon-out channel in the RIXS process. We multiplied a self-absorption coefficient to the simulated RIXS result (multiplication of S(ωin, ωout) and μ(ωin)). The formula implies that the saturation effect is stronger for a large absorption factor. Thus, the features at ∼780 eV in PFY spectra are to be suppressed. Consequently, the PFY weights on the Co3+ region is largely suppressed for the feature C (cf. Figure b). It also enhances the intensity at ∼782 eV of PFY spectra for the feature D and shows better agreement with the experiment. However, there is still an intensity discrepancy of the feature at 782 eV for excitation E. This overestimation of 782 eV feature is due to the model of oxygen 2p orbitals, where its bandwidth is neglected. In our simulations, the energy broadening was assumed to be the same for both sites, which yields rather sharp features in the Co3+ site contribution compared to the experiment. A larger energy broadening for the Co3+ site is caused by the strong ligand–metal hybridization on the Co3+ site. To reproduce the broad XAS/PFY structure above the L3 edge, the band formation of ligand 2p states must be taken into accounted, which is beyond the description of the cluster model used in this study.

The present analysis shows that the excited states at 0.5 and 1.3 eV are the unique features to identify the 4T2(T) excited state on the Co2+ site and 3T1(O) excited state on the Co3+ site. In comparison with the infrared[4] and 1s3d RIXS,[37,38] those individual site can be resolved much better. Our result confirms that the Co3O4 is mainly composed by the magnetically active high-spin Co2+(T) and the diamagnetic low-spin Co3+(O).

Ligand–Metal Hybridization Influence of Different Co Sites in Co3O4

We discuss here the question whether the ligand–metal hybridization influences the local electronic structure based on the ligand-to-metal charge transfer model. Hibberd et al. have shown in their ionic-model analysis that the Slater integrals of the ionic Coulomb multiplet need to be reduced substantially to reproduce the Co L3 XAS spectra, which implies that the ligand–metal hybridization is strong in Co3O4.[18,28,50] The sensitivity of 2p3d RIXS to the dd excitations allows us to address the question, and furthermore study the site-dependence of the covalency in Co3O4. Figure compares the energy diagrams obtained with three different models: (i) the ionic model with the bare (ionic) values of the Slater integrals, (ii) the one with reduced Slater integrals, and (iii) the cluster model including the ligand-to-metal charge transfer channel explicitly. Apparently, the ionic model (i) overestimates energies of observed excitations. For the Co2+ site, the ionic model (ii) with reduced Slater integrals (80% from the ionic values) yields good agreement with the experimental data. Note that both model (i) and (ii) show nice agreement on the 4T2 and 4T1 excited states, which indicates that those features are less sensitive to the hybridization change. On the other hand, for the Co3+(O) site, the Slater integrals are reduced to 55% (80%) from the ionic F2dd (F4dd) value to fit the experimental data. These unconventional reduction rates for the Co3+(O) site imply that the differential screening effect via a ligand-to-metal charge transfer channel is not negligible. The cluster model including the ligand-to-metal charge transfer channel shows good agreement to the experimental results.

Figure 6

Energy diagrams calculated using the values of the ionic Slater integrals (ionic SI), the reduced Slater integrals (reduced SI), and the ligand-to-metal charge transfer (LMCT) effect.

To obtain further information about the ligand–metal hybridization, Table shows the configuration weights in the ground state of the cluster model using the optimized parameters in Table . The ligand-to-metal charge transfer channel mixes the ionic configuration (3d) with ones with ligand holes (3dL1 and 3dL2). In the Co3+ site, the ligand-hole configurations (3d7L1 and 3d8L2) show large weights, which indicates that the Co3+ site is strongly hybridized with the ligand states. In contrast, the ligand-hole configuration (3d8L1) contributes only ∼20% to the ground state in the Co2+ site. This observation suggests that the Co3+ site is highly covalent, while the Co2+ site is rather ionic in Co3O4. The difference also affects the RIXS profile: the ionic Co2+ site shows sharp local dd excitations; the covalent Co3+ site exhibits a broad intense fluorescence-like feature (cf. Figure ).

Table 2

Weight of Configurations and Cation Orbital Covalency in the Ground State (unit in %)a

	weight of configurations			orbital covalency
	|3dn⟩	|3dn+1L1⟩	|3dn+2L2⟩	e(eg)	t2(t2g)
Co2+(3d7)	79	20	1	100	80
Co3+(3d6)	40	50	10	50	100

Although the number of ligand holes is considered up to two in the spectral simulations, the covalency is estimated only using the configurations up to one ligand hole.

The orbital covalency of the Co2+ (Co3+) cation was analyzed using the approach described in the SI(36) and is summarized in Table . The orbital covalencies of e and t2 orbitals on the tetrahedral Co2+ cation in Co3O4 are 100% and 80%, respectively. The e orbital is fully occupied and cannot participate in the ligand–metal hybridization, thus 100% orbital covalency is found. A high value of the cation orbital covalency for t2 orbital indicates that it less contributes to the ligand-hole configuration d8L, which is consistent with the ionic character of the Co2+ site. For the Co3+ site, the orbital covalencies of eg and t2g orbitals are 50% and 100%. This indicates that the ligand–metal hybridization mainly influences to the eg orbital of the Co3+ ions. Although the Co3+ cation is the singlet ground state (t2g6eg0 state in the ionic picture), the eg orbital forms a strong bonding with neighboring oxygen 2p orbitals. This bonding is represented by the d7L configuration in an eg symmetry, yielding the reduced value (50%) of the orbital covalency.

Conclusions

We present the Co 2p XAS and 2p3d RIXS experimental results in comparison with cluster model simulations on Co3O4. The 2p3d RIXS provides good chemical site selectivity to the local electronic structure, from which we can identify orbital covalencies of different ions in the compound. The polarization-dependent analysis indicates the symmetry character of the dd excitations, which provides a solid guide to analyze the local electronic structure. By selecting characteristic excitations, the PFY spectra are able to give additional site dependent information. The result shows the 4T2 excited state of the tetrahedral Co2+ site at 0.5 eV, which is beyond the discriminative power of optical absorption. In addition, the 1A1 to 3T2 excitation of the octahedral Co3+ site at 1.3 eV can be uniquely identified. The ground state electronic structure of the Co2+ ions and the Co3+ ions are respectively high-spin 4A2(T) and low-spin 1A1(O), where the high-spin Co2+ must be the magnetically active site. Our result also shows strong ligand–metal hybridization on the Co3+ site, which indicates that the Co3+ site in Co3O4 is rather covalent. In contrast, the Co2+ site shows weak hybridization implying that Co2+ is more ionic. This chemical site selectivity will help the further understanding on the site-dependent catalytic activity and magnetic activity of the spinel oxides.