Microstructure and coercivity in La-coated Nd₂Fe₁₄B thin films

The demand for Nd-Fe-B permanent magnets in renewable energy applications, such as electric motors for electric vehicles and wind power generators, is increasing rapidly. Increasing the coercivity of Nd-Fe-B magnets without using rare metals is desirable for high-temperature (around 200 °C) applications. Therefore, the coercivity mechanism in Nd-Fe-B magnets is attracting much attention.

Vial et al.¹ and Shinba et al.² found that there is a thin, amorphous, Nd-rich grain boundary (GB) phase at the interface between Nd₂Fe₁₄B grains and the morphology of the Nd-rich GB phase contributes to the resistance to demagnetization. Fukagawa et al.^3–5 showed that partial loss of the Nd-rich GB phase on the surface Nd₂Fe₁₄B grains leads to low coercivity, and the surface coercivity is recovered by the formation of the face-centered cubic (fcc) phase, which is derived from metastable C-Nd₂O₃ (⁠$Ia3¯$ with a ≈ 1.1 nm). Matsuura et al.^6,7 reported that in Nd₂Fe₁₄B/Nd(-O) model interfaces, appearance of the metastable C-Nd₂O₃ phase increases the fluidity of the Nd-rich GB phase. Mo et al.⁸ showed that the crystal structure of the Nd-rich phase mainly depends on the O content, and the Nd-rich phase containing 11–43 at.% O has an fcc structure with a = 0.52–0.55 nm. Thus, the fcc Nd-rich phase is derived from a CaF₂ ordered structure or C-Nd₂O₃ (⁠$Ia3¯$⁠), in which the O vacancies are arranged symmetrically. Because the exact atomic positions and occupancy of the fcc Nd-rich phase is still unclear, some researchers have described the fcc phase as NdO_x or NdO_2-δ.^5,7,9–13 The crystal structure of the fcc Nd-rich phase may also become amorphous in the form of a thin film along the GB. Watanabe et al.^9,12,14,15 demonstrated that even in hot-deformed magnets, the coercivity is increased by the formation of a smooth, continuous GB phase. From these microstructural studies, the following coercivity mechanisms have been proposed. (1) Magnetic isolation of Nd₂Fe₁₄B grains by surrounding with the thin Nd-rich phase. (2) Restraint of the reversed magnetic domain by restoration of the magnetocrystalline anisotropy of Nd at the surface of the Nd₂Fe₁₄B grain. (3) Restraint of the reversed magnetic domain by the relaxation of the lattice strain at the surface of the Nd₂Fe₁₄B grain.

Furthermore, the coercivity mechanism has been investigated by first-principles calculations.^16–20 Sepehri-Amin et al.^21–24 showed experimentally that the thin GB phase contains a high amount of Fe of more than 60 at.% and exhibits ferromagnetic features. Therefore, the coercivity mechanism is more controversial than before. The effect of the thin GB phase on the coercivity has been investigated by using thin-film model specimens.^25–30 In most of these studies, Nd thin films were deposited on the surface of Nd₂Fe₁₄B and oxidized by annealing, and then the effects of the crystal structures of the Nd-oxide layer on the coercivity were investigated by microstructure analyses. However, Koike et al.²⁷ found that the coercivity in the Nd₂Fe₁₄B film magnet was increased by coating with La, which has no 4f electrons, and was a source of the magnetocrystalline anisotropy.

In this study, as-deposited and post-annealed La-coated Nd₂Fe₁₄B thin film magnets were prepared by ultra-high vacuum (UHV) sputtering, and the microstructure changes of the magnets were investigated by spherical aberration (Cs)-corrected scanning/transmission electron microscopy (S/TEM). From these results, we discuss the effect of the surface conditions on the coercivity of Nd₂Fe₁₄B.

La-coated Nd₂Fe₁₄B films were fabricated by UHV helicon sputtering on Al₂O₃(0001) substrates, with the stacking sequence Al₂O₃ substrate/Mo buffer (10 nm)/Nd₂Fe₁₄B (20 nm)/La (0, 20 nm)/Mo cap (5 nm), to obtain as-deposited film magnets. Some of the as-deposited films were heat-treated at 470 °C for 40 min to obtain the post-annealed film magnets. Details of the film fabrication methods are described elsewhere.²⁷ 

Magnetization curves of the films were measured by a superconducting quantum interface device magnetometer in fields of up to 50 kOe. Thin foil specimens were prepared by Ga⁺ focused ion beam (FIB) milling (Versa 3D DualBeam, FEI) with a lift-off technique, followed by low-angle Ar⁺ ion milling (1050, Fischione Instruments) to remove surface damage caused by the FIB milling. The S/TEM observations and energy-dispersive spectroscopy (EDS) analysis were carried out by S/TEM (Titan Cubed, FEI) with double Cs correctors and a four windowless Si drift detector-EDS system with a high-brightness field emission electron gun (X-FEG, FEI).

Figures 1(a) and (b) show the demagnetization curves of the as-deposited and post-annealed La-coated Nd₂Fe₁₄B thin films. Data for uncoated Nd₂Fe₁₄B thin film is shown as a dashed curve for comparison (Fig. 1(c)). The coercivity of the as-deposited La-coated film is about 8.5 kOe, which is similar to that for uncoated films. After post-annealing, the coercivity of the La-coated film increases substantially to about 15 kOe.

Figure 2 shows a cross-sectional bright-field (BF) TEM image and nanobeam diffraction (NBD) patterns of a typical microstructure for the as-deposited La-coated Nd₂Fe₁₄B film. A plate-like Nd₂Fe₁₄B grain about 30 nm thick and 150 nm wide is formed on the Mo buffer layer on the Al₂O₃ substrate, and a La-O layer covers the plate-like Nd₂Fe₁₄B grain with good adhesion. The NBD pattern in Fig. 2(e) indicates that the La-O layer has an fcc structure with a lattice parameter a ≈ 0.564 nm. The lattice parameters were estimated by using the camera constants, which were calculated by considering the bulk Nd₂Fe₁₄B lattice constants of a = 0.8795 nm and c = 1.2188 nm. This La-O phase probably corresponds to NdO_x, which has often been observed in Nd-Fe-B magnets.^5,7,9–13 The formation of the La-O phase is considered to be due to a very small amount of residual oxygen in the sputtering target and/or that adsorbed on the inner wall of the deposition chamber. The as-deposited film is fully epitaxially grown with a crystallographic orientation relationship of Al₂O₃ $(001)[010]$ || Mo $(111)[011¯]$ || Nd₂Fe₁₄B $(001)[110]$ || LaO_x $(111)[112¯]$⁠.

Similar TEM analysis was performed for the post-annealed La-coated film, and a typical result is shown in Fig. 3. The Nd₂Fe₁₄B grain appears to be thicker, but this change is not essential. The upper region of the Nd₂Fe₁₄B grain appears damaged by FIB milling, resulting in an amorphous structure. The Nd₂Fe₁₄B crystal lattice is known to be relatively weak against electron beam irradiation and easily tends to be amorphous during a normal electron microscopic observation. The Nd₂Fe₁₄B/LaO_x interface is a metal/oxide one, whereas the Nd₂Fe₁₄B/Mo interface is the metal/metal one. Therefore, there is a possibility that the Nd₂Fe₁₄B lattice being adjacent to the LaO_x was selectively damaged by the Ga ion beam during the FIB processing owing to the differences in thermal conductivity, electric conductivity and so on. In Fig. 3(a), we see that a plate-like Nd₂Fe₁₄B grain covered smoothly with a La-O layer is formed. The NBD analysis shows that the La-O layer keeps its crystal structure for LaO_x, and the lattice parameter increases slightly to a ≈ 0.566 nm, which is slightly (0.4%) larger than that in the as-deposited film. The crystallographic orientation relationship is the same as that in the as-deposited film given in Fig. 2. To understand the interface structure, high-resolution TEM observations were obtained for the other post-annealed specimen, which was prepared by the ion-milling method. Figure 4(a) shows the high-resolution TEM image of the interface between Nd₂Fe₁₄B grain and LaO_x layer having two variants. The NBD patterns of Fig. 4(b) show that variants 1 and 2 have a twin relationship to the $(111¯)$ plane. In Fig. 4(a), there is a ledge on the side of Nd₂Fe₁₄B in this interface and the LaO_x variants change at the ledge. The height of the ledge is half of the c-axis of Nd₂Fe₁₄B, and the upper and lower atomic planes of the ledge correspond to the RE planes at z = 0 and z = 1/2. This suggests that the LaO_x phase is epitaxially grown under the influence of the atomic arrangement of the Nd ions at the (001) surface of the Nd₂Fe₁₄B grain. Therefore, the crystal structure and the epitaxial orientation relationship of the LaO_x layer are not changed by annealing.

STEM-EDS analysis was performed for the Nd₂Fe₁₄B/LaO_x interfaces in the as-deposited and the post-annealed films. The composition profiles of Fe, Nd, La, and O are shown in Fig. 5(a). The profiles of Fe and Nd are similar for both films. However, the O content in the LaO_x layer increases substantially from 15 to 35 at.% O with annealing, accompanied by a reduction in the La content. The revised Fe, Nd, and La profiles reevaluated by excluding the O content remain completely unchanged before and after the post-annealing (Fig. 5(b)). This result indicates that the self-diffusion of Fe and the inter-diffusion of Nd and La do not occur, even near the interfaces. Consequently, the structural and compositional changes caused by annealing are only in the O content in the LaO_x layer and no diffusion of the other elements occurs, at least over several atomic planes.

We found that the crystal structure and the epitaxial orientation relationship of the LaO_x coating layer was unchanged and only the O content was increased by annealing. These facts indicate that we can eliminate a possibility of exchange-decoupling change between the Nd₂Fe₁₄B grains being an origin of the observed coercivity increase by the annealing. We also confirmed that the degree of the c-axis alignment of the Nd₂Fe₁₄B grain does not change significantly by the annealing. Therefore, we examined the three factors for the coercivity enhancement by the annealing as shown in Fig. 6: (1) Changes in the lattice strain at the interface caused by the difference in O content. (2) The contribution of O to the crystalline field of the Nd ions in the (001) surface of Nd₂Fe₁₄B. (3) Substitution of La for Nd at the Nd₂Fe₁₄B surface.

The lattice parameters of the Nd₂Fe₁₄B and the LaO_x layers were measured from their NBD patterns, and the lattice strains were estimated roughly at the interfaces. The lattice parameter of the Nd₂Fe₁₄B layer was the same in both films, whereas that of the LaO_x layer was increased slightly from a = 0.564 to 0.566 nm by annealing. From the crystallographic orientation relationship of Nd₂Fe₁₄B of $(001)[110]$ || LaO_x $(111)[112¯]$⁠, the most likely matching planes of Nd₂Fe₁₄B and LaO_x are shown in Fig. 7. In the projected crystal structures, the atomic positions in the matching planes or the surfaces of Nd₂Fe₁₄B and LaO_x are marked with red circles. Because the rhombohedral atomic arrangement of the Nd ions in the (001) plane (z = 0) of Nd₂Fe₁₄B affects the similar rhombohedral arrangement of La ions in the (111) plane of LaO_x, the atomic arrangements surrounded by yellow-rectangular should be connected with each other. Assuming that only the Nd₂Fe₁₄B lattice is strained by connecting with the rigid LaO_x, it can be estimated that the compressive strain along the $[110]$ direction from –3.8% to –3.5% and tensile strain along the $[1¯10]$ direction from +11.1% to +11.5% are generated before and after post-annealing, (Fig. 8). Although there is another domain in which the LaO_x cell is rotated by 90° in a (111) plane, the connection between the (111) plane of LaO_x in this rotated domain and the (001) plane (z = 1/2) of Nd₂Fe₁₄B is equivalent to the situation in Fig. 7. Therefore, the changes in lattice strain caused by annealing are less than 0.5% at any of the interfaces. Sakuma et al.³¹ reported that, based on their first-principle calculations, the sign of the crystal-field parameter, A₂⁰, changes from positive to negative, affecting the magnetic anisotropy constant, K_u, when a large compressive strain of over 5% is generated at the Nd(g)-site, or over 10% at the Nd(f)-site. Because the expected strain change caused by annealing is small, we conclude that the effect of lattice strain on the coercivity increase is negligible in the present films.

Because the O content in the LaO_x layer is increased by annealing, some of the O atoms could be located around the Nd ions in the (001) surface of Nd₂Fe₁₄B (Fig. 6(b)). Moriya et al.¹⁶ showed by first-principles calculations that the crystal-field parameter, A₂⁰, at the Nd ion has a negative value when the Nd ion in the (001) plane is exposed to a vacuum. However, Toga et al.²⁰ predicted that the sign of the A₂⁰ parameter of the Nd ions at the (001) surface remains positive when the O atom is located near the surface with appropriate distance and direction ranges. Although the exact coordination of O atoms around the Nd ions at the interface is still not clear, we suggest that such a crystal-field recovery scenario occurs in the present La-coated films and in the Nd-coated films,^25,30 explaining the increase in coercivity after annealing.

The substitution of La for Nd in the surface of the Nd₂Fe₁₄B grain probably recovers the coercivity if the Nd ions at the surface are substituted by La ions without 4f electrons or magnetocrystalline anisotropy. The substitution of La for Nd in the (001) surface means that all the Nd ions in the Nd₂Fe₁₄B grains are surrounded by Fe atoms (σ layers) (Fig. 6(c)), and the Nd ions retain the magnetocrystalline anisotropy in the Nd₂Fe₁₄B grain. Therefore, dissolving the magnetization reversal nuclei can recover the coercivity. Figure 9 shows the STEM-EDS elemental map and line profiles for the uncoated Nd₂Fe₁₄B thin film, which exhibits the coercivity of about 10 kOe. The film structure is Al₂O₃ substrate/Mo (10 nm)/Nd₂Fe₁₄B (20 nm)/Mo (5 nm) and the whole of the Nd₂Fe₁₄B grain should be covered with a Mo layer. However, the STEM-EDS analysis indicates that a thin Fe layer (5 nm thick) is formed at the interface between the Nd₂Fe₁₄B grain and the Mo layer; thus, most of the surface of the Nd₂Fe₁₄B grain is surrounded by a thin Fe layer. Tsuchiura et al.³² recently reported that the exchange coupling between the Nd 5d and the nearest-neighboring Fe 3d electrons produces the magnetocrystalline anisotropy of the Nd ions in the Nd₂Fe₁₄B system. This mechanism would also work at the interface of the Nd₂Fe₁₄B/Fe system where Nd ions are exposed at the outermost surface of the Nd₂Fe₁₄B grains. If we assume that the Nd atoms are exposed on a bare surface of the Nd₂Fe₁₄B grain, the local anisotropy would be reduced owing to the partial missing of the neighboring Fe atoms. Since these Nd atoms would act as a nucleation site, the coericity would be very small as reported by Moriya et al.¹⁶ If extra Fe atoms are located near to such Nd atoms due to the segregation, then the reduced anisotropy might be locally restored. When the thickness of such segregated Fe layer is as thin as monatomic size, we would presume to have a coercivity much larger than 10 kOe. According to our knowledge of hard/soft nanocomposite magnets, the coercivity decreases with increasing thickness of the soft phase.³³ As described above, the thickness of the soft magnetic Fe layer is about 5 nm. We therefore consider that the magnitude of the observed coercivity of 10 kOe is reasonable as a result of a reduction due to the presence of the 5-nm soft phase.

It should be noted that the thin Fe layer surrounding the Nd₂Fe₁₄B grain was observed only at the interface with Mo and was not present at the interface with the LaO_x layer. Since the α-Fe phase is mainly formed on the surface of the Nd₂Fe₁₄B grains by an oxidation,³⁴ it is considered that the surface of the Nd₂Fe₁₄B grain at the interface with Mo was oxidized to form the thin Fe layer by the reaction with residual oxygen in the vacuum chamber. On the other hand, at the interface with LaO_x, the La atoms would absorb oxygen due to their higher oxygen affinity and then the oxidative decomposition of Nd₂Fe₁₄B does not occur. Therefore, the O and the Fe atoms are considered to be the key elements for improving the coercivity at the LaO_x/Nd₂Fe₁₄B and the Mo/Nd₂Fe₁₄B interfaces, respectively.

Detailed microstructure analysis of La-coated Nd₂Fe₁₄B thin films was performed with Cs-corrected S/TEM, to investigate an origin of the coercivity enhancement by annealing. Fine Nd₂Fe₁₄B grains were covered with an epitaxial-grown LaO_x layer and the O content of the LaO_x layer increased from 15 to 35 at.%, although the microstructure was not changed by annealing. We suggest that the O and the Fe atoms would be the key elements for improving the coercivity at the LaO_x/Nd₂Fe₁₄B and the Mo/Nd₂Fe₁₄B interfaces, respectively.

This work was supported by the Industry-Academia Collaborative R&D Programs from the Japan Science and Technology Agency, and was performed as an extension of collaborative research supported by the Nanotechnology Platform Project from the Ministry of Education, Culture, Sports, Science and Technology, Japan.