Electronic structure of hole-doped delafossite oxides CuCr{}_{1-x}Mg{}_{x}O{}_{2}

# Electronic structure of hole-doped delafossite oxides CuCr1−xMgxO2

T. Yokobori    M. Okawa    K. Konishi    R. Takei    K. Katayama Department of Applied Physics, Tokyo University of Science, Shinjuku, Tokyo 162-8601, Japan    S. Oozono    T. Shinmura    T. Okuda Department of Electrical and Electronics Engineering, Kagoshima University, Kagoshima, Kagoshima 890-0065, Japan    H. Wadati Department of Applied Physics and Quantum-Phase Electronics Center, University of Tokyo, Bunkyo, Tokyo 113-8656 Japan    E. Sakai    K. Ono Photon Factory, KEK, Tsukuba, Ibaraki 305-0801, Japan    H. Kumigashira Photon Factory, KEK, Tsukuba, Ibaraki 305-0801, Japan PRESTO, Japan Science and Technology Agency, Chiyoda, Tokyo 102-0076, Japan    M. Oshima Department of Applied Chemistry, University of Tokyo, Bunkyo, Tokyo 113-8656, Japan    T. Sugiyama    E. Ikenaga Japan Synchrotron Radiation Research Institute, Sayo, Hyogo 679-5198, Japan    N. Hamada Department of Physics, Tokyo University of Science, Noda, Chiba 278-8510, Japan    T. Saitoh Department of Applied Physics, Tokyo University of Science, Shinjuku, Tokyo 162-8601, Japan
July 14, 2019
###### Abstract

We report the detailed electronic structure of a hole-doped delafossite oxide CuCrMgO () studied by photoemission spectroscopy (PES), soft x-ray absorption spectroscopy (XAS), and band-structure calculations within the local-density approximation + (LDA+) scheme. Cr/Cu - resonant PES reveals that the near-Fermi-level leading structure has primarily the Cr character with a minor contribution from the Cu through Cu –O –Cr hybridization, having good agreement with the band-structure calculations. This indicates that a doped hole will have primarily the Cr character. Cr PES and -edge XAS spectra exhibit typical Cr features for all , while the Cu -edge XAS spectra exhibited a systematic change with . This indicates now that the Cu valence is monovalent at and the doped hole should have Cu character. Nevertheless, we surprisingly observed two types of charge-transfer satellites that should be attributed to Cu () and Cu () like initial states in Cu - resonant PES spectrum of at , while Cu PES spectra with no doubt shows the Cu character even for the lightly doped samples. We propose that these contradictory results can be understood by introducing not only the Cu state, but also finite Cu , –Cr charge transfer via O states in the ground-state electronic configuration.

###### pacs:
79.60.-i, 71.20.Ps, 78.70.Dm
preprint: Journal ref.: Phys. Rev. B 87, 195124 (2013)

## I Introduction

The search for new sustainable energy resources, including new innovations, is an urgent issue in modern societies. Thermoelectricity is one of the promising candidates because there exists so much waste heat that could be recovered without sacrificing environmental costs. Delafossite-type oxides CuO ( = trivalent cation) have considerable potential for thermoelectric materials Okuda et al. (2005) because of their layered structure of edge-shared O octahedrons that is very similar to the one in thermoelectric NaCoO.Terasaki et al. (1997) Hole-doped CuCrMgO is a member of this family, being a candidate for a future thermoelectrode. In CuCrO, electrons of the Cu ions under the pseudo- local symmetry fill up the narrow Cr band, which is the conterpart of the Co band filled by six electrons in NaCoO. Hence, as in NaCoO, a rapid change in the density of states (DOS) at the Fermi level () (Ref. Takeuchi et al., 2004) may be realized near the band edge in the hole-doped system CuCrMgO,Note1 () because the Cr band is expected to be at the top of the valence band in terms of a comparison of the charge-transfer energy of the Cr ion and that of the Cu ion.Iwasawa et al. (2006) More precisely, in -resolved electronic structure, this situation would correspond to the pudding-mold band structure that yields a large thermopower in NaCoO.Kuroki and Arita (2007) As a consequence, a combination of a large and the highest electrical conductivity among delafossite oxidesNagarajan et al. (2001) may be able to produce a large thermoelectric figure of merit (: thermal conductivity) in the present system.

Aside from thermoelectricity, CuO has various interesting physical properties both in fundamental and applicational terms. A former example is multiferroic oxides CuFeO (Ref. Kimura et al., 2006) and the present compound CuCrO (Ref. Seki et al., 2008) as well, whereas an important finding for the latter was a -type transparent conducting oxide (TCO); the -type TCO’s such as InO, SnO, or ZnO based ones had been realized earlier,Hamberg and Granqvist (1986) yet the -type counterpart was more difficult. A delafossite CuAlO was the first -type TCO with high carrier mobility and a wide band gap.Kawazoe et al. (1997) From the view point of the near- electronic structure, this was accomplished by hole doping into a wide gap Cu oxide, which has the closed shell.Kawazoe et al. (1997) Hence, the top of the valence band was expected to have the Cu character with some O one due to hybridization.

The electronic structure of CuCrO has been investigated both theoretically and experimentally in the context of TCO,Scanlon et al. (2009); Arnold et al. (2009); Hiraga et al. (2011) or of thermoelectric/multiferroic materials.Maignan et al. (2009) Along the conventional strategy for TCO, the top of the valence band is expected to have mainly the Cu character, whereas it would be desirable to have mainly the Cr character for better thermoelectric properties as mentioned before. On this point, reported first-principles band-structure calculations are still controversial; Scanlon et al. reported that the Cr partial DOS has the maximum peak at the same energy as the maximum peak of the Cu partial DOS and negligibly small Cr partial DOS at the top of the valence band.Scanlon et al. (2009) In contrast, Maignan et al. reported considerable Cr partial DOS at the top of the valence band,Maignan et al. (2009) and a recent study by Hiraga et al. showed the Cr partial DOS in a much deeper energy.Hiraga et al. (2011) Experimental electronic structure of CuCrO has been investigated by photoemission spectroscopy (PES), x-ray absorption spectroscopy (XAS), and x-ray emission spectroscopy. In these studies, Scanlon et al. and Arnold et al. interpreted the development of the upper part of the valence band with in CuAlCrO as a reconstruction of the Cu bands in stead of a development of the Cr states, and concluded that the Cr DOS minimally contributed to the top of the valence band.Scanlon et al. (2009); Arnold et al. (2009) However, magnetic and transport studies reported a close coupling of the doped holes by Mg substitution and the spin of the Cr ions that suggested the mixed-valences state Cr/Cr,Okuda et al. (2005); Ono et al. (2007) which in turn implies Cr character at the top of the valence band in the parent compound CuCrO.

From the above overview, the electronic structure of CuCrMgO, particularly near , has not been established yet. In this paper, we performed a comprehensive study on the electronic structure of lightly hole-doped CuCrMgO ( = 0–0.03) by photoemission spectroscopy with various photon energies, soft x-ray absorption spectroscopy, and band structure calculations using the local density approximation + (LDA+) method.

## Ii Experiment and Calculation

Polycrystalline samples of CuCrMgO (=0, 0.02, 0.03) were prepared by the standard solid-state reaction.Okuda et al. (2005) Vacuum ultraviolet (VUV)-PES measurements in the range of the Cr/Cu 3- resonance (–90 eV) were performed at BL-28A of the Photon Factory, KEK, using a SCIENTA SES-2000 electron analyzer. Hard x-ray PES (HX-PES) spectra taken with eV were measured at BL47XU of SPring-8 using a SCIENTA R4000 electron analyzer. XAS spectra of the Cr and Cu edge regions and Cu - resonant soft x-ray PES (SX-PES) spectra were measured at BL-2C of the Photon Factory, KEK, using a SCIENTA SES-2000 electron analyzer. In order to obtain clean surface, we fractured the samples in situ right before the measurements. The fracturings and the measurements were done in ultrahigh vacuum, namely, about Pa (VUV-PES, SX-PES, and XAS), about Pa (fractureing for HX-PES), and about Pa (measurement for HX-PES), all at 300 K. The intensity of the resonant PES spectra was normalized using photon current of the exit mirror. The energy resolution was 30 meV (VUV-PES), 140 meV (SX-PES), and 250 meV (HX-PES). All the Fermi-level () positions in the experiments were calibrated with Au spectra.

We also performed band-structure calculations with the full potential linearized augmented plane-wave (FLAPW) methodAndersen (1975); *Takeda79 in the LDA+ scheme.Hohenberg and W. Kohn (1964); *Kohn65; *Vosko80; Anisimov et al. (1991); *Solovyev96; *Anisimov97 For the effective Coulomb repulsion , relatively small values (2.0 eV for Cu and Cr) were adopted. The rhombic lattice parameters ( Å, Å) were taken from Ref. Poienar et al., 2009. The plane-wave cut off energy was 653 eV for the wave function. We took 1313 points in the irreducible Brillouin zone for the rhombohedral Brillouin zone.Maignan et al. (2009) Although the system is known to be antiferromagnetic,Okuda et al. (2005) the magnetic structure was assumed to be ferromagneticMaignan et al. (2009) because the detailed magnetic structure is not experimentally well-determined.Poienar et al. (2009)

## Iii Results

### iii.1 Experimental valence-band electronic structure compared with band structure calculations

When the photon energy comes near the 3-3 (or 2-3) excitation threshold, resonant behaviors appear in intensity of the valence-band photoemission due to the interference between the direct () and indirect [ (or ) ] processes. This is called 3-3 (or 2-3) resonant photoemission, which can be used to extract the 3 contribution of a specific element to the valence band. Figure 1 shows the valence-band spectra of CuCrMgO taken with a series of photon energy around the Cr 3- [Fig. 1(a)] and the Cu 3- [Fig. 1(b)] resonance. One can easily observe that the intensity of the near- leading structures, namely, the shoulder at 1.4 eV and the peak at 2.3 eV, systematically varies with incident photon energy. This intensity evolution is displayed in Figs. 1(c) and 1(d) as the constant initial state (CIS) spectra at the binding energy () of 1.4 eV [Fig. 1(c)] and 2.3 eV [Fig. 1(d)], respectively. To remove the background intensity from the CIS spectra as taken (filled symbols),Li et al. (1992) we also show the CIS spectra after subtracting the background by the Shirley method (open symbols).Shirley (1972)

Figure 1(c) shows that the 1.4-eV shoulder exhibits a distinct resonance-type line shapeFano (1961) with the maximum intensity at 50.0 eV, which is the Cr 3- resonance energy.Li et al. (1992) In contrast, the 2.3-eV peak shows a typical weak anti resonance-type line shape with a dipFano (1961) at the Cu 3- resonance energy 74.0 eV, as shown in Fig. 1(d).Thuler et al. (1982) However, one also notices that a weak resonance of the 2.3-eV peak does exist at 50.0 eV and a tiny antiresonance of the 1.4 eV at 74.0 eV. These observations are clearly demonstrating that (1) the 1.4-eV shoulder includes a major contribution of the Cr states with a minor contribution of the Cu states and vice versa for the 2.3-eV peak, and (2) nevertheless there exists sizable hybridization between the Cr and Cu states via O states. The major contribution of the Cr states in the 1.4-eV shoulder is also confirmed by a - resonant PES measurement as shown in Fig. 2. Figure 1(a) demonstrates that the 1.4-eV shoulder at eV (off resonance) rapidly grows to an intense peak at eV (on resonance) with increasing photon energy. Accordingly, the on-off difference spectrum, representing the Cr partial DOS, has a sharp peak at 1.4 eV (Panel (b)).

From the above results, the schematic energy diagram is that the Cr is at the top of the valence band, the next is Cu , and then O states come in the order of binding energy. This conclusion is different from recent PESArnold et al. (2009) or opticalHiraga et al. (2011) studies, both of which concluded that the Cu states are located at the top of the valence band. The origin of this difference will be discussed later in relation to band structure calculations. The present result is reasonable also from the viewpoint of the O TM charge transfer energy because the location of the Cr states and the Cu states, hybridizing with each other via O states in this compound, is governed by the difference of and ,Iwasawa et al. (2006) and would be larger than even considering the different valence and local configurations.Fujimori et al. (1993); Saitoh et al. (1995)

In order to analyze the valence-band electronic structure in more detail, we performed LDA+ band-structure calculations. Figure 3 shows the result of our LDA+ calculations. The Cu partial DOS has intense peaks between and eV with small Cr partial DOS in this range, whereas the Cr partial DOS exhibits a considerably large peak centered at about eV, distributed from the top of the valence band to eV with small Cu partial DOS in this range. Here, it is noted that the calculated Cr partial DOS has good agreement with the experimental Cr spectral weight in Fig. 2(b). The O bands are mainly located below the Cr and Cu bands, from to eV. All the states, Cu , Cr and O 2, show very small DOS in the entire energy range. The present calculation, particularly on the location of the Cu/Cr partial DOS, agrees well with the experimental result shown in Fig. 1 and the interpretation/prediction using the difference of and as well.Iwasawa et al. (2006)

The agreement between our experiment and calculation is demonstrated more clearly in Fig. 4, which shows a comparison between the experimental spectrum of CuCrMgO taken at eV and the calculated DOS.Note2 () A theoretical simulation curve has been constructed by broadening the cross-section-weighed total DOS with an energy dependent Lorentzian function due to the lifetime effect and a Gaussian due to the experimental resolution.Saitoh et al. (1997); Iwasawa et al. (2009); Yeh and Lindau (1985); Note3 () This theoretical specctrum shows that the leading structure at the top of the valence band (labeled as ) is dominated by the Cr states with a minor contribution of the Cu states whereas the most intense peak (labeled as ) primarily originates from the Cu states. In both structures, appreciable O DOS exist as well because of large photoionization cross section.Yeh and Lindau (1985) One can see that the theoretical spectrum satisfactorily reproduces the experimental one and thus the experimental structures A to F can be assigned to the theoretical structures to , respectively.

Our calculation agrees well with the calculation by Maignan et al.Maignan et al. (2009) while it is different from Scanlon et al.Scanlon et al. (2009) or Hiraga et al.Hiraga et al. (2011) However, we note that the spectrum by Scanlon et al. and Arnold et al. can simply be interpreted by our calculation as a development of the Cr states by Cr substitution for Al.Scanlon et al. (2009); Arnold et al. (2009) Hence, we consider that their experiment is actually consistent with ours. On the other hand, Hiraga et al. consistently interpreted their optical absoption spectra using their band structure calculations.Hiraga et al. (2011) However, optical absorption spectroscopy is indirect to probe the valence-band electronic structure because it gives the joint DOS. While we (and Maignan et al.) have assumed the ferromagnetic state in the calculations, we believe that the different Cr partial DOS does not come from the different magnetic structures because both Scanlon et al. and Hiraga et al. have calculated antiferromagnetic states by the same generalized gradient approximation + (GGA+) method, resulting in the quite different Cr partial DOS’s. The differences in the two calculations probably originate from the fact that Scanlon et al. adopted theoretically optimized lattice parameters and Hiraga et al. set the value for the Cu states to be zero. Our result is also supported by another band structure calculation of CuAlCrO that reported the same energetic order of the Cr and the Cu states as ours, namely, the Cr states come to the top of the valence band by Cr doping.Kizaki et al. (2005)

### iii.2 Cu and Cr valence

Figure 5(a) shows the valence-band photoemission spectra of CuCrMgO taken across the resonant energies of the Cu satellite structures. There can be observed two distinct satellite peaks at the binding energy of 13 and 15 eV, which have their maximums at the photon energy of 74 and 77 eV, respectively, as shown in Fig. 5(b). These numbers are in very good agreement with the reported satellite peaks in CuO (12.5–12.9 eV) and CuO (15.3 eV), which have mainly and final-state character, respectively.Thuler et al. (1982); Ghijsen et al. (1990); Shen et al. (1990) The 15-eV satellite peak has also been observed in Al XPS spectra of CuAlO and CuCrO.Arnold et al. (2009) Figure 5(c) shows the CIS spectra of these two satellite peaks. The CIS profiles of the satellites again well reproduce those of CuO and CuO, respectively, including the two-peak structure due to 3 and 3 splitting.Thuler et al. (1982); Ghijsen et al. (1990) All these results indicate that the doped hole in CuCrMgO produces Cu ions, namely, holes will be doped into the Cu sites. However, this observation seems to be incompatible with the result that the top of the valence band has mainly the Cr character, demonstrated in Figs. 14. Moreover, the 13-eV satellite due to Cu seems to be too intense for only 3% doping of Mg, which corresponds to 3% Cu ions.

To confirm this observation, we performed Cu - resonant photoemission spectroscopy measurements, as shown in Fig. 6. The excitation energies were determined by Cu XAS spectra shown in Fig. 6(d).Note4 () Figure 6(a) shows the valence-band spectra of the sample taken in the Cu - resonance region. The giant resonance peak at 15 eV is due to Cu ions as seen in Fig. 5 and as reported for CuO.Tjeng et al. (1992) Figures 6(b) and 6(c) show the spectra taken at the photon energies before the giant resonance develops. In Fig. 6(c), the spectrum shows the distinct 13-eV resonant peak of Fig. 5 at eV that corresponds to the photon energy of the pre-peak structure in Fig. 6(d). This hump has been observed in some CuO (Refs. Hulbert et al., 1984; Grioni et al., 1989), CuAlO (Ref. Aston et al., 2005) and CuCrO (Ref. Arnold et al., 2009) but has not been observed in pure CuO,Grioni et al. (1992) and it is accordingly interpreted as final state due to Cu impurity,Grioni et al. (1992); Aston et al. (2005); Arnold et al. (2009) where denotes a core hole of the Cu level. Therefore, both the Cu - resonant photoemission and the Cu XAS spectra of the sample clearly demonstrate the Cu nature of the doped holes observed in Fig. 5.

Surprisingly, however, Fig. 6(b) shows that the =0 sample, too, has the 13-eV satellite. This can never be due to Cu impurity because the Cu XAS spectrum has no appreciable prepeak [see Fig. 6(d)]. Here, we noted that the very slight modulation from the baseline at the prepeak of the =0 spectrum cannot explain the large 13-eV resonance peak because the Cu impurity concentration in the sample, if exists, can be estimated to be a few percent at most by a comparison with the reported relation between the concentration and the prepeak intensity in CuAlO.Aston et al. (2005) Therefore, it can be undoubtedly concluded that some kind of state that does not originate from Cu impurities, should exist even in the pure CuCrO, and based on this fact, one may further go beyond the case, and arrive at the idea that the whole portion of a doped hole may not necessarily go into a Cu site even because the 13-eV satellite is observed.

Figure 7 shows Cu core-level spectra of CuCrO and CuCrMgO. The spectrum in Panel (a) is almost identical to the reported spectra of CuCrO (Refs. Arnold et al., 2009; Le et al., 2011) and also CuAlO.Aston et al. (2005) There is no trace of structures at 934 eV due to the Cu state that, if exist, can easily be identified as is the case of oxidized CuAlZnO or CuRhMgO.Aston et al. (2005); Le et al. (2011) A reported energy shift of the Cu peak due to Mg dopingArnold et al. (2009) was not observed and the spectrum is almost identical to that of , which is very similar to what was observed in CuAlZnO.Aston et al. (2005) This fact raises doubt about the Cu nature of a doped hole. Nevertheless, a small but important change due to Mg doping can be observed in Fig. 7(b); the Cu line shape becomes asymmetrically broad. A Doniach-Šunjić lineshape analysisDoniach and Šunjić (1970) has confirmed a large increase in asymmetry with hole doping, which is reflecting an increase in metallicity of the system, particularly on the Cu sites.Note5 () Hence, this small change suggests the Cu nature of a doped hole again.

Figure 8 shows Cr core-level spectra of CuCrO and CuCrMgO. The double-peak structure observed in the Cr peak of the both samples is characteristic of Cr compound. Both the spectra in Panel (a) are indeed very similar to those of CrO and CrN.Biesinger et al. (2004); Bhobe et al. (2010) The and 0.02 spectra are very similar to each other, displaying Cr nature. However, the Cr peak shows a remarkable change due to Mg doping; the first peak at 575 eV obviously decreases in intensity with Mg doping. A very similar change has recently been observed in CrN across its insulator-metal transition, which has been explained by the screening effects due to mobile carriers.Bhobe et al. (2010) Therefore, the observed change is likely an evidence that doped holes move around the Cr sites, suggesting the Cr nature of a doped hole. This result is consistent with the valence-band satellite analyses in Figs. 5 and 6. Nevertheless, all the three Cr XAS spectra in Fig. 8(b) are very similar to the reported spectra of LaCrO and CrO,Sarma et al. (1996); Matsubara et al. (2002) indicating that the Cr ions are trivalent. Unlike the Cu edge, Cr XAS spectra show no detectable changes with hole doping that were observed for LaSrCrO with .Sarma et al. (1996)

## Iv Discussion

It is already established now that the ground-state electron configuration of CuO is not a simple , but , while that of CuO is described as , where denotes an O ligand hole; the configuration of the Cu ion should be spherical, but it was long ago pointed out that the charge distribution in CuO can be non-spherical due to the hybridization between the orbital ( axis along the Cu-O bonding) and the orbital,Orgel (1958) and has been discussed theoretically later.Marksteiner et al. (1986) This hybridization yields a hole and hence the ground state of CuO should have the component. The hole state has recently been directly observed,Zuo et al. (1999) confirming the interpretation of the satellite structures at 15 eV (the process) in CuO and at 13 eV (the process) in CuO.Thuler et al. (1982); Ghijsen et al. (1990); Shen et al. (1990)

The situation in CuCrO is quite analogous to CuO because the local environment around Cu is the same O-Cu-O dumb-bell structure, and therefore it is not surprising that the ground state has the component. What is striking in our results is that even the sample with no Cu impurity centers has shown a weak but detectable 13-eV satellite (Fig. 6). This inevitablly indicates that not a “virtual” state (), but the “real” state () has to exist in CuCrO. However, the Cu core-level spectra do not show any trace of such a configuration even for hole-doped samples, either. Nevertheless, the development of the Cu pre-peak structure with , again, undoubtedly demonstrates that this configuration increases with . On the other hand, the doped hole should have the Cr character from the Cr HX-PES spectra while the Cr -edge XAS spectra show no detectable changes.

To understand the above contradictory results, we reconsider the local electronic structure of the Cu site beyond the nearest-neighbor oxygens, namely, consider the two metal sites, Cu and Cr, because their wave functions are actually connected via the O wave functions.

Within a metal-oxygen single cluster model (CuO and CrO for the Cu and the Cr sites, respectively), the local electronic configuration of Cu can be described as , whereas that of Cr will be .Saitoh et al. (1995); Uozumi et al. (1997) Although the molecular orbitals of the Cu and the Cr sites have in fact different symmetries, there should be sizable overlap between some of them as discussed in Fig. 1. Hence, we consider the Cu-O-Cr cluster and re-define as an O ligand hole in a molecular orbital of this cluster. In this model, the combination of the configuration of Cu and the configuration of Cr can produce the and configurations at the Cu and Cr sites, respectively, because of the extended nature of the state. Hence, the ground state can be described as

 |g⟩=α|d10d3⟩+β|d9sd3⟩+γ|d9d4⟩+δ|d10L––d4⟩,

where the left and denote the Cu states, denotes the Cu state, and the right and denote the Cr states. is the main configuration, corresponds to the hole state, is the Cu –to–Cr charge-transfer state, and finally originates from the O –to–Cr charge-transfer state, which is the second main configuration. The configuration is not included because this is the origin of the configuration.

The final state of the valence-band photoemission by Cu emission is

 |fCuv⟩=a|d10L––d3⟩+b|d9d3⟩+c|d8sd3⟩+d|d8d4⟩.

Here, is neglected because this configuration will easily transform into due to the combination of one extra electron at the Cr site and the lack of one electron at the Cu site.

For the Cu core-level photoemission, the final state will be

 |fCuc⟩=a|c–d10d3⟩+b|c–d9sd3⟩+c|c–d10L––d4⟩+d|c–d9d4⟩,

and for the Cu -edge XAS, the final state will be

 |fCuL⟩=a|c–d10sd3⟩+b|c–d10d4⟩+c|c–d10sL––d4⟩,

where denotes a Cu core hole.

Within this framework, the Cu - and - resonant photoemission spectra can have both the (at 15 eV) and (at 13 eV) final-state satellites due to the processes of and , respectively. This scenario even predicts that CuAlO will not have the 13-eV satellite because there are no available Al states in the valence band, and indeed, an XPS spectrum of CuAlO shows a dip around 13 eV, while that of CuCrO has extra spectral weight,Arnold et al. (2009) supporting the scenario. The absence of the final-state satellite in the Cu core-level spectra can be explained by strong screening effects due to the presence of a core hole at the Cu site: The large - Coulomb attraction increases the number of electrons and accordingly it makes the and weight negligible even for the lightly hole-doped samples. Likewise, the lack of the pre-peak structure in the Cu -edge XAS specturm of the =0 sample can also be explained by the core-hole screening effects that reduce the weight of the (and the ) configuration(s) in (Fig. 6(b)).

From the above consideration, there must be weak but finite Cu spectral weight at the top of the valence band, and this can actually be observed; Figure 9 shows a HX-PES valence-band spectra of and 0.03 samples. Considering that the photoionization cross section of states of this energy range is largely enhanced,Yeh and Lindau (1985) the small enhancement at very near (see the inset) can be interpreted as an increase in the Cu emission due to hole doping, namely, supporting finite Cu spectral weight at the top of the valence band. It is accordingly revealed that the very top of the valence band has actually the Cu character in addition to the Cr character. This Cu –Cr duality of a doped hole can explain the observed magnetic and transport properties of CuCrMgO; the doped holes moving in the Cr-O network can lift the magnetic frustration in the Cr triangular spin lattice, resulting in an increase in the magnetic susceptibility with .Okuda08 () The holes that are not restricted in the Cu-O network also explain a higher electric conductivity compared with other hole-doped Cu delafossites such as CuAlO.Nagarajan et al. (2001) In particular, the highest conductivity by selecting Cr is strikingly demonstrating the importance of the Cu-Cr combination.Nagarajan et al. (2001) From the viewpoint of the electronic structure, this can be interpreted as a consequence of an “appropriate” combination in terms of the difference of and of CuO.Iwasawa et al. (2006)

## V Conclusions

We have studied the electronic structure of hole-doped delafossite oxides CuCrMgO by high-resolution photoemission spectroscopy, x-ray absorption spectroscopy, and LDA+ band-structure calculations. The Cr and Cu - resonant PES spectra demonstrated that the leading structure of the valence band near the has primarily the Cr character with a minor contribution from the Cu due to hybridization with the O states, in good agreement with the band-structure calculation. This result indicates that a doped hole will primarily have the Cr character. The Cr PES and -edge XAS spectra of CuCrMgO showed typical Cr features, whereas the Cu -edge XAS spectra exhibited a systematic change with . This result, by contrast, indicates that the Cu valence is monovalent at =0 and the holes will be doped into the Cu sites, which contradicts the Cr and Cu - resonant PES. Nevertheless, the Cu - resonant PES spectra display the two types of charge-transfer satellites that should be attributed to Cu () and Cu () like initial states, while the Cu PES with no doubt shows the Cu character even for .

We have proposed that the above apparently contradictory results can consistently be understood by introducing not only the Cu state as traditionally, but also newly finite Cu Cr charge transfer via O states in the ground-state electronic configuration. We found that this model can explain well some of the characteristic magnetic and transport properties of this compound.

###### Acknowledgements.
The authors would like to thank T. Mizokawa for enlightening discussions. The synchrotron radiation experiments at the Photon Factory and SPring-8 were performed under the approval of the Photon Factory (Proposal Numbers 2008G688, 2010G655, 2009S2-005, and 2011S2-003) and of the Japan Synchrotron Radiation Research Institute (Proposal Numbers 2011A1624, 2011B1710, and 2012B1003), respectively. This work was supported by JSPS KAKENHI Grants No. 22560786 and No. 23840039. This work was also granted by JSPS the “Funding Program for World-Leading Innovative R&D on Science and Technology (FIRST Program)”, initiated by the Council for Science and Technology Policy (CSTP).

## References

• Okuda et al. (2005) T. Okuda, N. Jufuku, S. Hidaka, and N. Terada, Phys. Rev. B 72, 144403 (2005).
• Terasaki et al. (1997) I. Terasaki, Y. Sasago, and K. Uchinokura, Phys. Rev. B 56, R12685 (1997).
• Takeuchi et al. (2004) T. Takeuchi, T. Kondo, T. Takami, H. Takahashi, H. Ikuta, U. Mizutani, K. Soda, R. Funahashi, M. Shikano, M. Mikami, S. Tsuda, T. Yokoya, S. Shin, and T. Muro, Phys. Rev. B 69, 125410 (2004).
• (4) The doped hole amount is .
• Iwasawa et al. (2006) H. Iwasawa, K. Yamakawa, T. Saitoh, J. Inaba, T. Katsufuji, M. Higashiguchi, K. Shimada, H. Namatame, and M. Taniguchi, Phys. Rev. Lett. 96, 067203 (2006).
• Kuroki and Arita (2007) K. Kuroki and R. Arita, J. Phys. Soc. Jpn. 76, 083707 (2007).
• Nagarajan et al. (2001) R. Nagarajan, A. Draeseke, A. Sleight,  J. Tate, J. Appl. Phys. 89, 8022 (2001).
• Kimura et al. (2006) T. Kimura, J. C. Lashley, and A. P. Ramirez, Phys. Rev. B 73, 220401(R) (2006).
• Seki et al. (2008) S. Seki, Y. Onose, and Y. Tokura, Phys. Rev. Lett. 101, 067204 (2008).
• Hamberg and Granqvist (1986) I. Hamberg and C. G. Granqvist, J. Appl. Phys. 60, R123 (1986).
• Kawazoe et al. (1997) H. Kawazoe, M. Yasukawa, H. Hyodo, M. Kurita, H. Yanagi, and H. Hosono, Nature 389, 939 (1997).
• Scanlon et al. (2009) D. O. Scanlon, A. Walsh, B. J. Morgan, G. W. Watson, D. J. Payne, and R. G. Egdell, Phys. Rev. B 79, 035101 (2009).
• Arnold et al. (2009) T. Arnold, D. J. Payne, A. Bourlange, J. P. Hu, R. G. Egdell, L. F. J. Piper, L. Colakerol, A. De Masi, P.-A. Glans, T. Learmonth, K. E. Smith, J. Guo, D. O. Scanlon, A. Walsh, B. J. Morgan, and G. W. Watson, Phys. Rev. B 79, 075102 (2009).
• Hiraga et al. (2011) H. Hiraga, T. Makino, T. Fukumura, H. Weng, and M. Kawasaki, Phys. Rev. B 84, 041411(R) (2011).
• Maignan et al. (2009) A. Maignan, C. Martin, R. Frésard, V. Eyert, E. Guilmeau, S. Hébert, M. Poienar, and D. Pelloquin, Solid State Commun. 149, 962 (2009).
• Ono et al. (2007) Y. Ono, K. Satoh, T. Nozaki, and T. Kajitani, Jpn. J. Appl. Phys. 46, 1071 (2007).
• Andersen (1975) O. K. Andersen, Phys. Rev. B 12, 3060 (1975).
• Takeda and Kubler (1979) T. Takeda and J. Kubler, J. Phys. F: Met. Phys. 9, 661 (1979).
• Hohenberg and W. Kohn (1964) P. Hohenberg and W. W. Kohn, Phys. Rev. 136, B864 (1964).
• Kohn and Sham (1965) W. Kohn and L. J. Sham, Phys. Rev. 140, A1133 (1965).
• Vosko et al. (1980) S. H. Vosko, L. Wilk, and M. Nusair, Can. J. Phys. 58, 1200 (1980).
• Anisimov et al. (1991) V. I. Anisimov, J. Zaanen, and O. K. Andersen, Phys. Rev. B 44, 943 (1991).
• Solovyev et al. (1996) I. Solovyev, N. Hamada, and K. Terakura, Phys. Rev. B 53, 7158 (1996).
• Anisimov et al. (1997) V. I. Anisimov, F. Aryasetiawan, and A. I. Linchtenstein, J. Phys.: Condens. Matter 9, 767 (1997).
• Poienar et al. (2009) M. Poienar, F. Damay, C. Martin, V. Hardy, A. Maignan, and G. André, Phys. Rev. B 79, 014412 (2009).
• Li et al. (1992) X. Li, L. Liu, and V. E. Henrich, Solid State Commun. 84, 1103 (1992).
• Shirley (1972) D. A. Shirley, Phys. Rev. B 5, 4709 (1972).
• Fano (1961) U. Fano, Phys. Rev. 124, 1866 (1961).
• Thuler et al. (1982) M. R. Thuler, R. L. Benbow, and Z. Hurych, Phys. Rev. B 26, 669 (1982).
• Fujimori et al. (1993) A. Fujimori, A. E. Bocquet, T. Saitoh, and T. Mizokawa, J. Electron Spectrosc. Relat. Phenomen. 62, 141 (1993).
• Saitoh et al. (1995) T. Saitoh, A. E. Bocquet, T. Mizokawa, and A. Fujimori, Phys. Rev. B 52, 7934 (1995).
• (32) A spectrum is compared with theory because of a lower quality of spectra in this photon energy range.
• Saitoh et al. (1997) T. Saitoh, T. Mizokawa, A. Fujimori, M. Abbate, Y. Takeda, and M. Takano, Phys. Rev. B 55, 4257 (1997).
• Iwasawa et al. (2009) H. Iwasawa, S. Kaneyoshi, K. Kurahashi, T. Saitoh, I. Hase, T. Katsufuji, K. Shimada, H. Namatame, and M. Taniguchi, Phys. Rev. B 80, 125122 (2009).
• Yeh and Lindau (1985) J. J. Yeh and I. Lindau, At. Data Nucl. Data Tables 32, 1 (1985).
• (36) The binding energy () dependent Lorentzian FWHM was set to be (eV).
• Kizaki et al. (2005) H. Kizaki, K. Sato, A. Yanase, and H. Katayama-Yoshida, Jpn. J. Appl. Phys. 44, L1187 (2005).
• Ghijsen et al. (1990) J. Ghijsen, L. H. Tjeng, H. Eskes, G. A. Sawatzky, and R. L. Johnson, Phys. Rev. B 42, 2268 (1990).
• Shen et al. (1990) Z.-X. Shen, R. S. List, D. S. Dessau, F. Parmigiani, A. J. Arko, R. Bartlett, B. O. Wells, I. Lindau, and W. E. Spicer, Phys. Rev. B 42, 8081 (1990).
• (40) Because of a problem of energy calibration, our photon energy of the Cu XAS spectra is a little different from the reported values for CuCrO or CuO (Refs. \rev@citealpnumArnold09,Grioni89-1,Grioni92,Tjeng92).
• Tjeng et al. (1992) L. H. Tjeng, C. T. Chen, and S.-W. Cheong, Phys. Rev. B 45, 8205 (1992).
• Hulbert et al. (1984) S. L. Hulbert, B. A. Bunker, F. C. Brown, and P. Pianetta, Phys. Rev. B 30, 2120 (1984).
• Grioni et al. (1989) M. Grioni, J. B. Goedkoop, R. Schoorl, F. M. F. de Groot, J. C. Fuggle, F. Schäfers, E. E. Koch, G. Rossi, J.-M. Esteva, and R. C. Karnatak, Phys. Rev. B 39, 1541 (1989).
• Aston et al. (2005) D. J. Aston, D. J. Payne, A. J. H. Green, R. G. Egdell, D. S. L. Law, J. Guo, P. A. Glans, T. Learmonth, and K. E. Smith, Phys. Rev. B 72, 195115 (2005).
• Grioni et al. (1992) M. Grioni, J. F. van Acker, M. T. Czyzyk, and J. C. Fuggle, Phys. Rev. B 45, 3309 (1992).
• Le et al. (2011) T. K. Le, D. Flahaut, H. Martinez, N. Andreu, D. Gonbeau, E. Pachoud, D. Pelloquin, and A. Maignan, J. Solid State Chem. 184, 2387 (2011).
• Doniach and Šunjić (1970) S. Doniach and M. Šunjić, J. Phys. C 3, 285 (1970).
• (48) The derived asymmetric parameter was () and ().
• Biesinger et al. (2004) M. C. Biesinger, C. Brown, J. R. Mycroft, R. D. Davidson, and N. S. McIntyre, Surf. Interface Anal. 36, 1550 (2004).
• Bhobe et al. (2010) P. A. Bhobe, A. Chainani, M. Taguchi, T. Takeuchi, R. Eguchi, M. Matsunami, K. Ishizaka, Y. Takata, M. Oura, Y. Senba, H. Ohashi, Y. Nishino, M. Yabashi, K. Tamasaku, T. Ishikawa, K. Takenaka, H. Takagi, and S. Shin, Phys. Rev. Lett. 104, 236404 (2010).
• Sarma et al. (1996) D. D. Sarma, K. Maiti, E. Vescovo, C. Carbone, W. Eberhardt, O. Rader, and W. Gudat, Phys. Rev. B 53, 13369 (1996).
• Matsubara et al. (2002) M. Matsubara, T. Uozumi, A. Kotani, Y. Harada, and S. Shin, J. Phys. Soc. Jpn. 71, 347 (2002).
• Orgel (1958) L. E. Orgel, J. Chem. Soc. 4186 (1958).
• Marksteiner et al. (1986) P. Marksteiner, P. Blaha, and K. Schwarz, Z. Phys. B 64, 119 (1986).
• Zuo et al. (1999) J. M. Zuo, M. Kim, M. O’Keeffe, and J. C. H. Spence, Nature 401, 49 (1999).
• Uozumi et al. (1997) T. Uozumi, K. Okada, A. Kotani, R. Zimmermann, P. Steiner, S. Hüfner, Y. Tezuka, and S. Shin, J. Electron Spectrsc. Relat. Phenom. 83, 9 (1997).
• (57) T. Okuda, Y. Beppu, Y. Fujii, T. Onoe, N .Terada, and S. Miyasaka, Phys. Rev. B 77, 134423 (2008); T. Okuda, R. Kajimoto, M. Okawa, and T. Saitoh, Int. J. Mod. Phys. B 27, 1330002 (2013).
You are adding the first comment!
How to quickly get a good reply:
• Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
• Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
• Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters