Introduction

Iron ions in oxides usually show the +2 and +3 oxidation states typically seen in wüstite (Fe2+O)1,2, magnetite (Fe2+Fe3+2O4)3,4,5 and hematite (Fe3+2O3)6,7. A few oxides, such as SrFeO3 and CaFeO3, contain unusual high-oxidation-state iron ions like Fe4+ and the behaviors of such high-valence iron ions have been attracting much attention for a long time8,9,10,11,12. The cubic perovskite SrFeO3 shows a metallic conductivity down to low temperatures because the linear Fe4+-O-Fe4+ bonds stabilize broad conduction bands. CaFeO3, on the other hand, has a distorted perovskite structure with a Fe-O-Fe bond angle of ≈160°13,14. The unusual oxidation state of the Fe4+ in CaFeO3 cannot be maintained at low temperatures and at 290 K its instability is relieved by charge disproportionation (CD): 2Fe4+ → Fe3+ + Fe5+ 11,12. Charge disproportionation is also seen in some perovskite-related-structure compounds like Sr3Fe2O7 and La1−xBaxFeO3−y, relieving the instability of their unusual valence states of iron15,16,17,18.

More than five decades after SrFeO3 was discovered, a new Fe4+-containing material was found, which has the A-site-ordered double-perovskite structure (see the crystal structure in Fig. 1). CaCu3Fe4O12 is obtained by synthesis under high-pressure and high-temperature conditions and the high-valence Fe4+ is stabilized at room temperature19. At 210 K the compound shows B-site charge disproportionation (4Fe4+ → 2Fe3+ + 2Fe5+) similar to the charge disproportionation in the simple perovskite CaFeO3 and it changes from a high-temperature paramagnetic-and-metallic phase to a low-temperature ferromagnetic-and-insulating phase (a charge-disproportionated phase). High-pressure synthesis techniques can be used to produce the analogous compound LaCu3Fe4O12, in which La3+ instead of Ca2+ occupies the A site in the A-site-ordered perovskite structure20. At 393 K LaCu3Fe4O12 shows not the CD behavior seen in CaCu3Fe4O12 but instead exhibits A'-B intersite charge transfer (CT), 3Cu2+ + 4Fe3.75+ → 3Cu3+ + 4Fe3+ and changes from a high-temperature paramagnetic-and-metallic phase to a low-temperature antiferromagnetic-and-insulating phase (a charge-transferred phase). Thus the instabilities of the unusual oxidation states of iron in these two A-site-ordered perovskites, Fe4+ in CaCu3Fe4O12 and Fe3.75+ in LaCu3Fe4O12, are relieved by completely different ways. To find out how the instabilities of the unusual oxidation states of the transition-metal ions in oxides are relieved, we made solid solution of CaCu3Fe4O12 and LaCu3Fe4O12 and investigated their temperature-dependent transitions.

Figure 1
figure 1

Crystal structure of the A-site-ordered double-perovskite AA'3B4O12.

The A, A', B, and O atoms are respectively represented by green, purple, blue and red spheres. The atom positions in the cubic Im-3 (No. 204) symmetry are A at the 2a site (0, 0, 0), A' at the 6b site (0, 1/2, 1/2), B at the 8c site (1/4, 1/4, 1/4) and at O at the 24g site (x, y, 0), where x ≈ 0.30 and y ≈ 0.17.

Results

Each solid-solution sample was confirmed by synchrotron X-ray diffraction (XRD) data (see Supplementary Fig. S1) to be a single phase at high temperatures and to be crystallized with a cubic Im-3 A-site-ordered double-perovskite structure. Oxygen off-stoichiometry was not detected in the Rietveld structure refinements and the refined occupancies for Ca and La at the A site were within 2% of those corresponding to the designed composition Ca1−xLaxCu3Fe4O12 (see Supplementary Fig. S1 and Supplementary Table S1 for the refinement results). When changing the A-site composition x the lattice constant at 450 K changes linearly in accordance with Vegard's law (Fig. 2a). No superlattice reflection was observed in the diffraction patterns, suggesting the absence of any extra ordering in the solid solution. Each Mössbauer spectrum at high temperatures showed a paramagnetic singlet component (Fig. 3), further confirming that each of the samples consisted of a single-phase solid solution. Note that the isomer shift values of the Mössbauer spectra of the paramagnetic states at 400 K gradually increase with increasing x (see Supplementary Fig. S2), suggesting that the Fe oxidation state decreases slightly. Furthermore, the bond valence sums (BVS) of Fe at 450 K, which are obtained from the structure refinements, gradually decrease with increasing x while those of Cu remain unchanged (Fig. 2b). The results suggest that electrons are doped into the Fe site instead of the Cu site by the La3+ substitution for Ca2+ at the A site. Thus, as we expected from the end compositions, the ionic formula of a solid-solution sample at high temperature can be described as (Ca2+1−xLa3+x)Cu2+3Fe(4−x/4)+4O12.

Figure 2
figure 2

Composition and temperature dependences of the lattice parameter of the Ca1−xLaxCu3Fe4O12 solid solution.

(a) Lattice parameters of the high-temperature phases at 450 K plotted as a function of composition x. The linear change follows Vegard's law, confirming that the samples are solid solutions. (b) BVS changes for Fe (blue) at the B site and Cu (green) at the A' site in the high-temperature phase of the Ca1−xLaxCu3Fe4O12 solid solution. BVS was calculated from the results of the structure refinements for the synchrotron XRD data obtained at 450 K. (c) Temperature dependence of lattice parameters of solid-solution samples. Data shown by closed symbols were obtained from synchrotron XRD and data shown by open symbols were obtained from laboratory XRD with Mo and Cu sources. The square, circle and triangle markers represent the high-temperature, charge-transferred and charge-disproportionated phases, respectively. Negative (Δa < 0 with increasing T) and positive (Δa > 0 with increasing T) thermal-expansion-like changes respectively indicate intersite CT and the CD transitions.

Figure 3
figure 3

Mössbauer spectra of the Ca1−xLaxCu3Fe4O12 solid solution at selected temperatures.

The observed spectra and fitting curves are represented by dots and solid lines, respectively. The high-temperature spectra shown in red have a single component due to paramagnetic high-valence Fe and each the low-temperature spectrum shown in blue is a magnetically ordered Fe3+ sextet originating from the phase due to CT. The spectra shown in orange and green are respectively those of charge-disproportionated Fe3+ and Fe5+. Note that the spectrum weights of Fe3+ and Fe5+ in the charge-disproportionated phase are always close to 1:1.

For CaCu3Fe4O12 (x = 0.0) a phase transition at 210 K is evident in the temperature dependence of the lattice parameter (Fig. 2c) and below that temperature very weak superstructure peaks, indicating rock-salt-type B-site ordering, were observed in the synchrotron XRD patterns. The ferrimagnetic transition (Fig. 4) and the metal-to-insulator transition (see Supplementary Fig. S3), together with the change in the Mössbauer spectra (Fig. 3a), confirm that the B-site CD occurs at 210 K, as we reported previously19,21. In the other end compound LaCu3Fe4O12 (x = 1.0), a first-order isostructural phase transition takes place at 393 K, as shown by the large increase of the lattice parameter with decreasing temperature (Fig. 2c). At the transition temperature the Fe-O bond length increases significantly whereas the Cu-O bond length decreases, decreasing the BVS for Fe and increasing it for Cu. From the Mössbauer spectra shown in Fig. 3b, one can infer that above the transition temperature there is a paramagnetic component of unusual high-valence Fe that has an isomer shift of ≈0.17 mm s−1 and that at 300 K there is a single component of magnetically ordered Fe3+. In addition, the compound changes from a high-temperature paramagnetic metal to a low-temperature antiferromagnetic insulator at the phase transition (Fig. 4 and Supplementary Fig. S3). Thus it is concluded that the compound changes from a high-temperature La3+Cu2+3Fe3.75+4O12 phase to a low-temperature La3+Cu3+3Fe3+4O12 phase as a result of the intersite CT between the A'-site Cu and B-site Fe. No CD behavior, either that shown by CaCu3Fe4O12 or the more complicated one shown by La1−xSrxFeO3−δ22,23,24, is seen in LaCu3Fe4O1220,25.

Figure 4
figure 4

Temperature dependence of the magnetic susceptibility of the Ca1−xLaxCu3Fe4O12 solid solution.

The samples were zero-field cooled and the measurements were made under a 1 T external magnetic field. The large increase in magnetization at 210 K indicates the ferrimagnetic transition accompanying the CD transition. The inset shows the magnified view of temperature dependence of the magnetic susceptibility of samples with x = 1/2, 3/4 and 1.0. The decrease in magnetic susceptibility is the result of antiferromagnetism due to the intersite CT transition.

The temperature dependence of the XRD patterns of the Ca3/4La1/4Cu3Fe4O12 sample (x = 1/4) shows electronic phase separation below 210 K (Supplementary Fig. S4). The large increase in the lattice parameter at 210 K with decreasing temperature indicates the appearance of the CT phase (Fig. 2c). On the other hand, the temperature dependence of the magnetic susceptibility shows a behavior similar to that shown by CaCu3Fe4O12, namely a sharp ferrimagnetic increase below ≈210 K but with a lower magnetization (Fig. 4) indicative of the CD transition. These are consistent with the Mössbauer spectrum at 4 K shown in Fig. 3c, which consists of a pair of Fe3+/Fe5+ (29.6%/28.5%) components for the charge-disproportionated phase and a magnetic ordered Fe3+ component (41.9%) for the phase due to CT. It is thus clear that below 210 K phases due to CD and CT coexist.

When the further La-doped Ca1/2La1/2Cu3Fe4O12 (x = 1/2) is cooled its lattice parameter increases sharply at 280 K (Fig. 2c), suggesting the occurrence of CT. Note that at low temperatures its diffraction peaks are rather broad (Supplementary Fig. S4). In the magnetic susceptibility data, a sharp decrease is found at 280 K, also indicating a CT transition like that in LaCu3Fe4O12. When the sample was further cooled, a CD-like increase was seen near 210 K, although the magnetization is more than an order of magnitude lower than that of CaCu3Fe4O12 at low temperatures (Fig. 4). The Mössbauer spectra change accordingly with decreasing temperature (Fig. 3d). A single paramagnetic component is seen at high temperatures and in the spectrum at 228 K a sextet (82%) originating from magnetically ordered Fe3+ is seen in addition to the high-temperature singlet (18%). At 4 K the high-temperature singlet changes to a Fe3+/Fe5+ sextet pair (refined area of each component: 12.8%/11.9%), confirming that CD occurs in roughly 25% of the sample.

For Ca1/4La3/4Cu3Fe4O12 (x = 3/4) the changes in the temperature-dependence of the XRD patterns (Supplementary Fig. S4) and the lattice parameters derived from them (Fig. 2c) indicate that intersite CT transition occurs around 330 K. In the Mössbauer spectra a paramagnetic component of high-valence Fe with an isomer shift 0.12 mm s−1 is seen at 400 K, while a single component from magnetically ordered Fe3+ is seen at 4 K, suggesting that most of the sample undergoes a CT transition (Fig. 3e). These CT behaviors are consistent with the observed sharp decrease in the magnetic susceptibility (Fig. 4) and the large increase in the resistivity below 330 K (Supplementary Fig. S3). Although a very minor CD-like transition around 210 K is seen in the magnetic susceptibility measurement, the charge-disproportionated phase is not as evident in the Mössbauer spectra and the XRD patterns.

A compositional phase diagram of the Ca1−xLaxCu3Fe4O12 solid solution is derived from all the experimental results described above and is shown in Fig. 5. For CaCu3Fe4O12 (x = 0.0) a CD transition is seen at 210 K. For x = 1/4 both CD and CT transitions are seen at almost the same temperature, ≈210 K, where about 60% of the sample shows CD while the other 40% shows CT. For x = 1/2, around 280 K CT occurs on cooling in about 75% of the sample and the other 25% remains in the high-temperature state with unusually high-valence Fe. The remaining high-temperature phase then undergoes CD at 210 K, which is the same as the CD transition temperature for the x = 0.0 sample. Below this temperature, phases due to CD and CT coexist. For x = 3/4 the CT transition occurs around 330 K in most of the sample and a very minor charge-disproportionated phase appears below 210 K. And for LaCu3Fe4O12 (x = 1.0), only CT transition is observed at 393 K. In summary, the intersite CT transition temperatures (TCT) in the solid solution samples increase with increasing La doping (i.e., with increasing x), while the CD transition temperatures (TCD) do not change from that of CaCu3Fe4O12.

Figure 5
figure 5

Compositional phase diagram for the Ca1−xLaxCu3Fe4O12 solid solution.

At high temperatures the whole solid solution is a single phase. Different charge behaviors are seen in the end-composition compounds: CD in CaCu3Fe4O12 (x = 0.0) and intersite CT in LaCu3Fe4O12 (x = 1.0). In the intermediate-composition samples, phases due to both CD and CT coexist at low temperatures. With increasing ligand-hole concentration δ (decreasing La substitution at the A site), TCT decreases but TCD remains constant.

Discussion

Note again that at high temperatures each solid-solution sample is a single phase and that the A-site substitution of La3+ for Ca2+ causes electron doping at the B-site Fe. Nevertheless, the electronic phase separation in the samples with the intermediate compositions is clearly seen at low temperatures. For x = 1/2, for example, the high-temperature phase can be described as ACu2+3Fe3.875+4O12 and the CT transition at 330 K changes some portion of the sample to ACu3+3Fe3+4O12, leaving the remaining portion of the sample to be a phase with unusually high-valence Fe. In a simple ionic model, this change can be described as ACu2+3Fe3.875+4O12 → 50%ACu3+3Fe3+4O12 + 50%ACu2+3Fe4+4O12. Although such an electronic phase-separation behavior seems to be uncommon, we could never observe a transition of the whole sample, such as a CT-like ACu2+3Fe3.875+4O12ACu3+3Fe3.125+4O12 transition or a CD-like ACu2+3Fe3.875+4O12ACu2+3Fe3+2.25Fe5+1.75O12 transition. Further decreasing temperature induces the CD transition for the ACu2+3Fe4+4O12 portion at 210 K, where ACu2+3Fe3+2Fe5+2O12 is stabilized. Although the observed fractions of the CT and CD phases (75%/25%) determined from the areas of the Mössbauer spectra measured at 4 K are rather different from the 50%/50% fractions predicted by the simple ionic model, they are reasonably close to the predicted ones. An important point is that the Fe3+:Fe5+ ratios in the CD phases are always close to 1:1. Furthermore, none of the TCD of the samples changes with composition x, suggesting that the observed CD transitions are essentially the same in the entire solid solution. The XRD peak broadening seen at low temperatures strongly suggests that the domains of the phases due to CD and CT coexist on a microscopic scale.

Why are the instabilities of these unusual oxidation states of Fe relieved in different ways? And why do we see both the CD and CT transitions in a single-phase sample? As discussed in previous reports on some specific oxides, high-oxidation-state transition-metal ions like Fe, Co, Ni and Cu have very low-lying 3d levels and the covalent electronic states due to the strong hybridization of 3d and oxygen 2p orbitals produce oxygen p holes (ligand holes)26,27,28,29,30,31,32,33,34,35. Thus, realistic electronic pictures of the unusual Fe4+, Fe5+ and Cu3+ states can respectively be described as d5L, d5L2 and d9L, where L represent a ligand hole. Indeed, unlike the isoelectric Mn3+ (t2g3eg1), Fe4+ (d4) with octahedral oxygen coordination does not show Jahn-Teller distortion. With the ligand-hole picture, CD in CaCu3Fe4O12 is expressed as (4d5L → 2d5 + 2d5L2), similar to that in CaFeO3. The transition is regarded as a redistribution of the ligand holes in the Fe sites, making the Fe-O bonds alternately shorter and longer in a rock-salt-type manner (Fig. 6a). The rock-salt-type ordering should contribute to minimizing the lattice energy and stabilizing the CD phase. Because at high temperatures the ligand holes are itinerant, as we see in the metallic conductivity, the CD transition in CaCu3Fe4O12 can be regarded as the localization of the ligand holes at the Fe sites, or in other words, as a charge ordering of the ligand holes. In LaCu3Fe4O12, on the other hand, the CT between the A'-site Cu ions and the B-site Fe ions is mediated by the transfer of ligand holes (3d9 + 4d5L0.75 → 3d9L + 4d5) from the Fe site to the Cu site (Fig. 6c). This ligand-hole transfer is also linked to the lattice change with the isotropic volume expansion. Since the charge transferred phase is insulating, the intersite CT transition can also be regarded as the localization of the ligand holes at the Cu site. Since the observed metal-to-insulator transition is caused by the localization of an odd number of itinerant ligand holes without breaking the cubic structural symmetry, it can be regarded as a Mott transition of the igand holes. It is clear that the difference between the CD and CT transitions is only the localization site of the ligand holes, so the energy difference between them should not be significant. This also explains why the charge-disproportionated and the charge-transferred phases coexist in the solid solution.

Figure 6
figure 6

Ligand-hole localization model of charge disproportionation and intersite charge-transfer transition behaviors.

At high temperatures the ligand holes L are homogeneously distributed at the Fe sites and here the ligand-hole concentration increases from (c) to (a). (a) The CD (4Fe4+ → 2Fe3+ + 2Fe5+) at 210 K in CaCu3Fe4O12 is described as 4d5L → 2d5 + 2d5L2 and can be regarded as the localization of the ligand holes at the Fe site (charge ordering of L). (c) The intersite CT (3Cu2+ + 4Fe3.75+ → 3Cu3+ + 4Fe3+) at 393 K in LaCu3Fe4O12 is described as 3d9 + 4d5L0.75 → 3d9L + 4d5 and can be regarded as the localization of the ligand holes at the Cu site (a Mott transition of L). (b) In the intermediate-composition samples (e.g., x = 1/2), with decreasing temperature the ligand holes are first localized at the Cu site by making CT-phase domains (about 75% of the sample). When temperature decreases further, the ligand holes in the other portion (25%) of the sample are localized at the Fe site, making CD-phase domains.

At high temperatures the ligand holes in the solid solution are homogeneously distributed at the Fe sites like d5Lδ (δ = 1−x/4) and they are itinerant. When temperature decreases, the ligand holes lose kinetic energy and become unstable, resulting in the localization. The instability of the itinerant ligand holes is first relieved by transferring the ligand holes from the Fe site to the Cu site and the localization at the Cu site produces the d9L state. This also explains the change of the TCT in the solid solution. Because the ligand-hole concentration δ increases with decreasing x, the high-temperature state with a higher concentration of the itinerant holes is more stable over a wider temperature range and thus the TCT decreases. The Cu2+ counter cation in the A-site-ordered perovskite-structure oxide plays a crucial role in accepting the ligand holes. It is also interesting that the CT transition in the intermediate compositions does not occur in the whole sample but leaves some portion (domains) with itinerant ligand holes at the Fe site, causing an electronic phase separation. With further decreasing temperature, the instability of the remaining itinerant ligand holes is relieved by CD in which the d5/d5L2 states are ordered alternately (Fig. 6b). In the higher ligand-hole-concentration region (0.0 ≤ x ≤ 1/4), TCT is lower than (or almost the same as) TCD and thus the CT is not observed.

In conclusion, the unusual high-valence Fe ions are stabilized in the high-pressure synthesized A-site-ordered perovskite-structure Ca1−xLaxCu3Fe4O12 solid-solution samples and the end-composition compounds CaCu3Fe4O12 and LaCu3Fe4O12 display distinct charge behaviors, respectively charge disproportionation (CD) and intersite charge transfer (CT). Compounds with intermediate compositions first show intersite CT transition in part of the sample and then show CD in the rest of the sample. In this system, d orbitals of Fe at the B site and Cu at the A' site strongly hybridize with p orbitals of oxygen, producing ligand holes and the distinct charge behaviors can be explained by the localization of the itinerant ligand holes at low temperatures. In the charge-disproportionated phase the ligand holes are localized at the Fe site and the transition is regarded as one to the rock-salt-type charge ordering of the ligand holes. The CD transition is essentially the same in all solid-solution samples, so the TCD is same regardless of their La content. In the intersite CT, on the other hand, the ligand holes are localized at the Cu site and the transition can be regarded as a Mott transition of the ligand holes. TCT decreases with increasing concentration of the ligand-hole carriers. In the A-site-ordered perovskite-structure oxides, transition-metals at both A' and B sites mediate A'-A', A'-B and B-B interactions that lead to intriguing physical properties36,37,38,39,40,41. The ligand holes produced by the strong hybridization of transition-metal cation d orbitals and oxygen p orbitals also play important roles in giving rise to various electronic and structural properties. The present A-site-ordered perovskite-structure Ca1−xLaxCu3Fe4O12 solid solution is a novel example exhibiting interplay of the interactions mediated by the ligand holes.

Methods

Polycrystalline samples of the A-site-ordered double-perovskite solid solution Ca1-xLaxCu3Fe4O12 (x = 0.0, 1/4, 1/2, 3/4 and 1.0) were prepared, at 15 GPa and 1300 K, from stoichiometric amounts of Ca2Fe2O5, La2O3, CuO, Fe2O3 and the oxidizing agent KClO4 by using a multianvil press. Synchrotron XRD patterns at temperatures between 100 and 450 K were collected at beamline BL02B2, SPring-8, Japan, with wavelength 0.7737405 Å and the profiles were analyzed with Rietveld method by using the General Structure Analysis System (GSAS) software package42,43. The XRD patterns at temperatures between 80 and 400 K were also collected using a Rigaku RINT diffractometer with a Mo and a Cu source. Magnetic susceptibility, magnetization and electric conductivity were measured using a Quantum Design Magnetic Properties Measurement System (MPMS) and Quantum Design Physical Properties Measurement System (PPMS). The 57Fe Mössbauer spectra were obtained in transmission geometry in combination with a constant-acceleration spectrometer using 57Co/Rh as a radiation source. α-Fe was used as a control for velocity calibration and isomer shift. The obtained spectra were fitted by a least-squares method with Lorentzian functions.