Co site preference and site-selective substitution in La–Co co-substituted magnetoplumbite-type strontium ferrites probed by 59Co nuclear magnetic resonance

In the La–Co co-substituted magnetoplumbite-type strontium (Sr–La–Co) ferrite, the base materials of the high-performance hard ferrite magnet, Co tends to occupy more than two crystallographically inequivalent sites. To reveal the Co site preference and the function of each site, 59Co nuclear magnetic resonance (NMR) spectrum was measured for the Sr–La–Co ferrites with various Co compositions. The intimate correlation between the anisotropy field and the relative intensity of the strongest NMR line clearly indicates that Co occupying the tetrahedrally coordinated 4f1 site is only responsible for the enhancement in the uniaxial anisotropy. In the samples with the Co concentrations much larger than commercial magnets, most Co selectively occupies the 4f1 site, indicating possible tuning of the magnetic performance with limited Co content in the future development. The close correlation between orbital moments and local strain is also demonstrated.


Introduction
The economical and chemically stable ferrite magnet is one of industrially important permanent magnets along with rare-earth magnets [1]. The development of high-performance ferrite magnets leads to a large social impact. A recent milestone in this field is the improvement in magnetic performance by the substitution of a small amount of Co for Fe [2]. A typical example is the Sr-La-Co system, in which Fe 3+ in the Sr magnetoplumbite-type (M-type) ferrite, SrFe 12 O 19 , is substituted by a small amount of Co 2+ and a part of Sr 2+ is replaced by La 3+ to compensate electric charge [2]. Another example is the Ca-La-Co system, in which Sr in the Sr-La-Co system is replaced by Ca [3]. In the M-type ferrite, there are five crystallographically inequivalent Fe sites ( a b f f 2 , 2 , 4 , 4 1 2 , and 12k by Wyckoff notations in the space group P mmc 6 3 ). Oxygen coordination is octahedral for 2a, 4f 2 , and 12k, tetrahedral for 4f 1 , and bipyramidal for 2b. It is ferrimagnetic with majority-spin at 2a, 2b, and 12k sites, and minority spin at 4f 1 and 4f 2 sites. The improvement of the magnetic performance by the Co substitution is mainly ascribed to the increase in the coercivity, namely the enhancement in the uniaxial anisotropy associated with the unquenched orbital magnetic moment of Co 2+ occupying at particular Fe sites. In both Sr-La-Co and Ca-La-Co systems, Co tends to occupy more than two Fe sites [4][5][6][7]. It is likely, however, that Co at particular sites only contributes to the uniaxial anisotropy, while the other Co is useless or even harmful in enhancing magnetic performance. Furthermore, since Co occupation at the majority-spin sites reduces the net magnetization, the Co occupation at the minority-spin sites is preferable in the viewpoint of the saturation magnetization. In order to improve magnetic performance with limited Co content, site-selective Co substitution is highly required; commercially it is important to reduce the usage of Co because of its unstable price and supply. For the Sr-La-Co system, although it was widely believed that Co mainly occupies the octahedrally coordinated 4f 2 site [8], it is now gradually accepted that Co mainly occupies the tetrahedrally coordinated 4f 1 site and partly octahedrally coordinated 2a and/or 12k sites [5,6,9,10] been revealed that, in the samples synthesized under ambient atmosphere, La 3+ content, x, and Co 2+ content, y, are not necessarily identical but in general x>y, and their difference is compensated by the reduction of Fe 3+ to Fe 2+ [4,[11][12][13]. In order to get information on Co occupation sites, we have previously reported the result of 59 Co nuclear magnetic resonance (NMR) experiment for the Sr-La-Co sample synthesized under ambient atmosphere [7]. As principal components, three kinds of 59 Co-NMR lines, S1, S2, and S3, were observed at ∼90, 310, and 390MHz, respectively. However, the origin of the anomalously small internal field (∼9 T) for the strongest S1 line is yet to be explained. Namely, there are two possibilities; one is the compensation between spin and orbital contributions of high-spin Co 2+ to the internal field, and the other is the possible presence of octahedrally coordinated Co 3+ in the low-spin state with S=0. Since we found the Fe valence instability in the Sr-La-Co system, we could not exclude Co valence instability, i.e. the possible coexistence of Co 3+ . Recently, we have shown that a higher oxygen potential during sample synthesis extends the Co solubility limit and increases the anisotropy field almost linearly to the Co concentration [14].
In this study, in order to explain the origin of the S1 line and extract information on the site preference of Co, we performed 59 Co-NMR experiment for the Sr-La-Co ferrites with x≈y and x=0 prepared under oxygen and ambient atmosphere, respectively, and with much larger y than commercial magnets synthesized under a high oxygen pressure using hot isostatic pressing (HIP). We reveal the function of Co at each site and show a guiding principle to develop high-performance hard ferrite magnet.

Experimental procedures
Four samples were measured in the present study. One is Sr 1−x La x Fe 12−y Co y O 12 with x=0.32 and y=0.28, and another with x=0 and y=0.07. The former and latter single crystals were grown by the Na 2 O flux method under an oxygen gas stream (oxygen partial pressure p 1 O 2 = atm) and ambient atmosphere p 0.2 atm O 2 = ( ), respectively. The compositions of metallic elements, x and y, were determined by wavelength-dispersive x-ray spectroscopy (WDX). See [13] for details of the crystal growth and characterization. In synthesizing Sr-La-Co samples under ambient atmosphere, the Co content, y, is generally smaller than that of the La concentration, x. The first sample was synthesized under an oxygen gas stream to obtain crystals with x≈y. The higher oxygen potential can suppress the appearance of Fe 2+ , which is the origin of the discrepancy between x and y. The second is a La-free sample, in which Co is expected to be trivalent because of the lack of degrees of freedom in the electric charge of transition-metal elements. The Co concentration y=0.07 was the maximum limit for the Lafree single crystals prepared by the Na 2 O flux method under ambient atmosphere. The single crystals were crushed to fine powder for the NMR experiment. The others are two Sr-La-Co samples in powder form synthesized using a HIP apparatus. They were synthesized under an oxygen partial pressure of p 387 O 2 = atm. Their WDX-analyzed compositions are x=0.70, y=0.72 and x=0.92, y=0.93. See [14] for the details of the HIP-synthesized samples. For the samples treated in the present study, lattice parameters at room temperature and the anisotropy field, H A , at 5 K [13,14] are listed in table 1. The lattice parameters were estimated from powder x-ray diffraction data measured by X'Pert PRO Alpha-1 PANalytical with use of Cu-K α1 radiation. H A was estimated from the area surrounded by easy-and hard-axis magnetization curves, K 1 , and saturation magnetization, M s , using the relation, The 59 Co-NMR experiment was done at IPCMS using a nontuned wide-band spectrometer, which ensures good reproducibility of the frequency spectrum. The nuclear quantities for 59 Co are the nuclear spin I=7/2, the gyromagnetic ratioγ/2π=10.05 MHz T −1 , the quadrupole moment Q=42.0×10 −30 m 2 , and a natural abundance of 100%. The 59 Co NMR spectrum was measured at 2 K under zero external field in a frequency range of 30-650 MHz by measuring the spin-echo intensity. The radio-frequency power dependence of the spin-echo intensity was accumulated at each frequency and integrated to obtain the frequency spectrum. In the same process, the frequency-dependent enhancement

Results and discussion
3.1. Site assignments of 59 Co-NMR signals Figure 1 shows 59 Co-NMR spectra for the Sr-La-Co samples synthesized under the oxygen gas stream (x=0.32, y=0.28) and by HIP (x=0.70, y=0.72 and x=0.92, y=0.93), together with that synthesized under ambient atmosphere (x=0.29, y=0.15) reported previously [7]. These samples are labeled as A15, O28, H72, and H93 in order of Co concentration, y (see table 1). Three main lines, which are labeled as S1, S2, and S3, were observed at around 89-90, 310-330, and 390-400 MHz, respectively, for all four samples. We consider that S2 and S3 are independent lines corresponding to Co occupying different sites because the lines are wellseparated at least for A15 and O28. In fact, the spectra measured with a weaker radio-frequency field, S2 and S3 are overlapped due to the anisotropic broadening characteristic of the domain wall signal. In contrast, the spectra shown in figure 1 were observed in much stronger radio-frequency pulse conditions, implying that we mainly observe magnetic domains. The presence of the S1 and S2/S3 resonances was also reported by Pieper et al [16] and Kouril [17] respectively. The resonance frequencies and their intensity ratios, namely Co occupation ratios, are listed in table 2. A faint Co signal, S4, was also observed at ∼530 MHz for A15 as reported in [7] and for O28. S4 is not visible in the present scale and is reasonably assigned to high-spin Co 3+ . The high-spin Co 3+ is generally unstable in the La-Co co-substituted system, but may be included slightly, depending on the sample preparation condition. In the following, we will not discuss the high-spin Co 3+ . In the spectrum measured for the La-free sample (x=0, y=0.07, labeled as N07), which is not included in figure 1, none of S1, S2, or S3 was observed.
The values of internal fields corresponding to S2 and S3 (∼30-40 T) are reasonable as those for high-spin Co 2+ when the empirical value of the spin part of hyperfine coupling for Co, A spin ∼−10 T/μ B , is applied. In the following calculations, we use A spin =−12 T/μ B estimated theoretically for Co 2+ [18]. In contrast, the resonance frequency of S1 looks anomalously small as that of high-spin Co 2+ . In [7], we pointed out two possible interpretations. One interpretation is the compensation between spin and orbital contributions to the internal field, H int , which is given by H A m A m int spin spin orb orb = + where m spin and m orb are spin and orbital parts of magnetic moment, respectively. A reported value of the orbital hyperfine coupling, A orb , for Co is, for example, +65 T/μ B [19], the magnitude of which is much larger than A spin . This means that the net internal field can be small depending on the compensation between spin and orbital contributions. A similar interpretation Figure 1. 59 Co-NMR spectra for samples A15, O28, H70, and H90 measured at zero applied field and at 2 K. Intensities were normalized to directly represent the amount of Co in such a way that the total integrated intensity of each material is in proportion to the Co concentration, y. The inset shows a magnified view of the S2 and S3 part. has already been done by Morel et al [8] for the data reported by Pieper et al [16], although their Co site assignment is essentially different from ours, as discussed below. The other possibility is the presence of low-spin (i.e. nonmagnetic) Co 3+ at some of the octahedrally coordinated sites if Co valence were unstable like that of Fe. We do not consider the presence of low-spin Co 2+ , which is generally unrealistic. To conclude this problem, we measured O28 with x≈y≈0.3, together with N07 for x=0. Since the charge compensation between La 3+ and Co 2+ roughly holds in O28 owing to the suppression of Fe 2+ , we expect that O28 includes Co 2+ only. In contrast, we expect that N07 includes Co 3+ only. From the fact that S1 is still strongest in O28, we can assign S1 to Co 2+ . The fact that none of S1, S2, and S3 was observed in N07 strongly suggests that all of S1, S2, and S3 come from Co 2+ . In fact, for N07, another sharp 59 Co signal, which can be assigned to low-spin Co 3+ , has been observed at ∼50 MHz in a completely different radio-frequency pulse condition [9]. This low-spin Co 3+ signal has not been observed for A15 and O28. Thus, all S1, S2, and S3 are reasonably assigned to high-spin Co 2+ , and their different resonance frequencies are ascribed to the difference in m orb . The divalent high-spin state of Co is also supported by Co K β x-ray emission spectroscopy [10].
Recently, Kobayashi et al [5] claimed that Co occupies 4f 1 , 2a, and 12k sites from the results of x-ray absorption spectroscopy (XAS) combined with neutron diffraction Rietveld refinement analysis. In addition, they claimed from the results of hard x-ray magnetic circular dichroism (MCD) experiment that Co occupies the tetrahedrally coordinated, namely 4f 1 , site. Our independent XAS and MCD experiments corroborated by firstprinciple calculations confirmed their conclusion (unpublished). Analyses of 57 Fe NMR and Mössbauer spectra also support the conclusion [9,10]. Thus, we assign the strongest S1 line to 4f 1 -Co and S2/S3 to 2a/12k-Co. So far, we have not succeeded in distinctively assigning 2a-and 12k-Co.

Co site preference and its correlation with anisotropy
As seen in figure 1 and table 2, for the samples synthesized under high oxygen partial pressure, the relative intensity of S1 is markedly enhanced compared with those of S2 and S3. The result indicates that, in the HIPsynthesized samples, not only that Co preferentially occupies the 4f 1 site, but also that the Co occupation at the site corresponding to S2 tends to be suppressed, contrary to no appreciable variation in S3. Thus, the higher oxygen potential not only expands the Co solubility limit, but also enhances the Co site selectivity. This result demonstrates that, by controlling appropriate parameters, the magnetic performance of the ferrite magnet can be tuned with limited Co content. As listed in table 1, the anisotropy field (at 5 K), H A , increases almost linearly to the amount of 4f 1 -Co, demonstrating intimate correlation between the uniaxial anisotropy and, particularly, tetrahedrally coordinated 4f 1 -Co.
One of the important characteristics of NMR for ferromagnetic (or ferrimagnetic) materials is the effect of signal enhancement. The effective oscillating field at the nuclear site, H ⊥ , is amplified by the oscillation of the internal field, H int , produced by electronic moments following the operating radio-frequency field, H 1 . Therefore, the enhancement factor H H 1 h =^gives the information on magnetic stiffness of the corresponding electronic moment, namely local anisotropy at the corresponding site. The local stiffness is usually represented by the restoring field, H r , which is the necessary field to recover electronic moments to the equilibrium and defined as H H int r h = . H r is given from optimum H 1 by H r =β H 1 , where β is a constant depending only on nuclide and experimental setup [15]. Figure 2 shows H r which gives optimal intensity of S1, namely for 4f 1 -Co, plotted as a function of Co concentration, y. H r for 4f 1 -Co increases appreciably and almost linearly to y. This behavior is in good accordance with the trend of H A (see the broken line in figure 2) but in good contrast to H r for S2 and S3 (inset of figure 2), which rather decrease with y. This result clearly demonstrates that 4f 1 -Co correlates with the anisotropy enhancement, while 2a/12k-Co acts negatively on the uniaxial anisotropy; 2a/12k-Co may prefer to planar or cone anisotropy. When crystal field potentials are much larger than the spin-orbit interacton, we expect that d orbitals split to the e g doublet and the t 2g triplet in cubic crystal fields. In the octahedral coordination, t 2g is lower, whereas, in the tetrahedral case, vice versa. The additional trigonal symmetry lowering splits the t 2g triplet to the d z 2 singlet and the (d x y 2 2 -+d xy ) doublet, where we take the principal z axis to be parallel to the c axis of the hexagonal lattice. In the octahedral case for d 7 (with the Hund coupling stronger than the crystal field energy), if the d z 2 singlet is lower than the (d x y 2 2 -+d xy ) doublet, the doublet with one hole results in double degeneracies of the ground state with an unquenched orbital moment [20,21]. This is the main reason why the 4f 2 was assigned to the principal Co occupation site. In fact, for the 4f 2 site, the simple symmetry consideration is not feasible due to its relatively low site symmetry (3m.). For the octahedrally coordinated 2a/12k sites corresponding to S2/S3, a similar argument may be applicable. In the tetrahedral case for 4f 1 , the (d x y 2 2 -+d xy ) doublet is expected to be lower than the d z 2 singlet; the 4f 1 site is coordinated by a nearly regular tetrahedron compressed slightly along c (site symmetry m 3 . ). In the ground state, full occupation in lower e g and half occupation in t 2g result in no orbital degeneracy. However, the mixing of the e g doublet with the (d x y 2 2 -+d xy ) doublet, possibly via the spin-orbit interaction, can revive orbital moment.
The low resonance frequency for S1 is reasonable provided that the orbital moment at 4f 1 -Co is smaller than those at 2a/12k-Co. Assuming a constant spin moment regardless of the occupation site and the magnitude of the orbital moment, and applying the coupling constants A spin =−12 T/μ B and A orb =65 T/μ B , relatively large orbital moments of 0.4 μ B and 1.0-1.2 μ B are estimated for 4f 1 -and 2a/12k-Co, respectively. We assumed that signs of the observed internal fields are all positive for S1, S2, and S3. The applied field dependence of the resonance frequency for S1 showed that the sign of the net internal field is positive for S1 [9]. Positive internal fields can be reasonably assumed also for S2 and S3 because larger orbital moments are expected for octahedrally coordinated 2a/12k-Co.

Correlation between orbital moments and local strain
As seen in figure 1, with increasing Co concentration, y, S1 slightly shifts to lower frequencies, whereas S2 and S3 shift to higher frequencies. The central frequencies of S1, S2, and S3 are plotted in figure 3 against y, where they all show good linearity with y. As mentioned above, the net internal fields are determined from the sum of spin and orbital contributions. Applying the same assumptions above (positive internal fields for all Co and the yand site-independent spin component), the variation is ascribed to the change in the orbital component. The negative y dependence of S1 implies that the 4f 1 -Co orbital field decreases with increasing y, whereas the positive variation for S2 and S3 indicates opposite tendencies for 2a/12k-Co.
The inset of figure 3 shows the c/a ratio plotted against y, where a and c are lattice parameters. Despite that the lattice parameter a itself does not necessarily vary systematically against y (see table 1), c/a shows an excellent linearity to y just like the resonance frequencies in figure 3. This indicates that the frequency shifts in figure 3 is dominated by the lattice deformation, suggesting strongly that the orbital contribution principally depends on local strain at the respective site, as generally expected. In the Sr-La-Co system, the decrease in c/a implies that the coordination polyhedra are shrunk along c. At the octahedrally coordinated 2a and 12k sites, we expect that the d z 2 singlet and the (d x y 2 2 + +d xy ) doublet are more separated, resulting in a higher population in the doubly degenerate ground state with the unquenched orbital moment. In contrast, at the tetrahedrally coordinated 4f 1 site, the decrease in c/a makes the split between the d z 2 singlet and the (d x y 2 2 + +d xy ) doublet narrower, reducing excitations to the t 2g -derived orbital, consequently leading to smaller orbital moment. These tendencies are in good agreement with our experimental observations, and, at the same time, support strongly our site assignments in section 3.1. The reduction in the orbital contribution at 4f 1 -Co is rather unfavorable for enhancing the uniaxial anisotropy, but seems to be overwhelmed by the appreciable increase in the Co occupation at the 4f 1 site. It is likely that 2a/12k-Co atoms do not contribute to the anisotropy enhancement, even if they have larger orbital moments, possibly because they prefer to planar or cone anisotropy.
The Co site selectivity may also be related to the local strain because the Ca-La-Co system with smaller c/a ratios tends to show larger anisotropy at smaller Co contents [3].

Conclusions
We have performed 59 Co-NMR experiments for Sr-La-Co ferrites with different Co compositions prepared under different oxygen potentials. Assigning the S1 resonance at the lowest frequency to high-spin Co 2+ at the tetrahedrally coordinated 4f 1 site, and S2 and S3 at higher frequencies to high-spin Co 2+ at the octahedrally coordinated 2a and 12k sites, intimate correlation between tetrahedrally coordinated 4f 1 -Co and the anisotropy field is demonstrated. The application of high oxygen pressure leads to site-selective Co occupation at the tetrahedrally coordinated 4f 1 site. This result indicates that, if Co can be concentrated in the 4f 1 site by controlling appropriate parameters, the magnetic performance of the hard ferrite magnet can be tuned more efficiently with limited Co content in the future development. Our result also suggests strongly that the orbital contribution to the magnetic moment, which is closely related to the enhancement in the magnetic anisotropy, is dominated by the local strain at the respective Co site.