Magnetics Hysteresis Properties and Microstructure of High-Energy (Nd,Dy)–Fe–B Magnets with Low Oxygen Content

Figure 1 shows the dependence of magnetic hysteresis properties of magnets on the total rare-earth element (R) content in initial alloys. A Dy-free magnet has the lowest Nd content of 29.2 wt %, which results in the highest B_r and (BH)_max, and the lowest coercivity. With increasing Nd content, the coercivity _MH_c increases up to 12–13 kOe; however, B_r and (BH)_max monotonously decrease. To stabilize the coercivity of magnets, 0.5 and 1.0 wt % Dy were added to alloys. This Dy addition stabilizes _MH_c at a value no less than 13 kOe for R = Nd + Dy > 29.6 wt %. In this case, (BH)_max of magnets with 0.5 wt % Dy and 29.6–30.1 wt % R is in the range of 48.5–49.5 MG Oe, and that of magnets with 1.0 wt % Dy decreases down to 47 MG Oe. The magnetization reversal curves of magnets with various Dy contents are shown in Fig. 2. Thus, in the alloys with 30 wt % R, the 0.5 wt % Dy addition is a convenient way to stabilize high hysteresis properties of high-energy magnets.

The total chemical composition, phase composition, and microstructure significantly affect structure-sensitive hysteresis properties of Nd–Fe–B magnets. In the course of sintering, oxygen adsorbed by powder upon preparation with a considerable part of excess Nd forms oxides between Nd₂Fe₁₄B grains. Figure 3 shows the effects of oxygen concentration in magnets on their properties. Similarly, B_r, H_c, and (BH)_max tend to decrease in the Dy-free and 0.5/1.0 wt % Dy-containing magnets. On the one hand, this trend is caused by an increase in the R content of magnets (Fig. 1); on the other hand, this result indicates that the higher R content, the higher oxygen content in magnets.

As is well known, the Nd-rich phases localized between Nd₂Fe₁₄B grains have various crystal structures and have different oxygen contents [10]. In order to find a correlation between magnetic properties, phase composition of the magnets, and their oxygen content, some magnets were studied by X-ray analysis and electron microscopy. Figure 4 shows diffraction patterns of the magnets prepared from alloys 3 (Dy 0%), 6 (Dy 0.5%), and 11 (Dy 1.0%). The phase compositions of these magnets are listed in Table 2. In addition to the main Nd₂Fe₁₄B phase, all magnets contain NdO_x with the fcc structure of NaCl (Space group \(Fm\bar {3}m\)).

Additionally, in magnets prepared from alloys 6 and 11, the stable Nd₂O₃ oxide has been found; however, its quantity is too small for X-ray determination. Despite the fact that the amount of oxides is determined with insufficient accuracy, it qualitatively shows that their weight fraction grows with oxygen content in magnets.

The microstructure of magnets whose phase composition was studied by X-ray diffraction analysis (XRD) was investigated. For instance, Fig. 5 demonstrates the microstructure of an unetched polished and etched surfaces of the magnet prepared from alloy 6. In the microstructure study, the plane of surfaces was parallel to the alignment axis which is vertical in the figure. The microphotograph of the etched surface (Fig. 5a) has clear boundaries between the Nd₂Fe₁₄B grains, which allows correct estimation of grain sizes. In the magnets under consideration with 0, 0.5, and 1.0 wt % Dy, the average grain size D_av of the Nd₂Fe₁₄B phase weakly depends on the Dy content and falls into the range of 3.5–3.8 μm. The electron-microscopy micrographs of the unetched surfaces (Fig. 5b) contain several phases different by the image contrast. The grains of the main Nd₂Fe₁₄B phase (A) have uniform dark-grey contrast. In the triple junctions of the Nd₂Fe₁₄B grains, there are Nd-rich phases. The largest volume of triple junctions is occupied by bright-light inclusions (В). According to the XRD, inclusions (B) are the NdO_x phase with a fcc structure. It can be persistently found that, in the vicinity of sharp edges of the triple junctions of the Nd₂Fe₁₄B grains, phase (В) changes to light-gray phase (C). The crystal structure determination of this phase requires additional investigation. Rarely, the Nd₂O₃ phase in the shape of dark-gray circular inclusions (D) (absent in Fig. 5b) can be found in the micrographs. The phase contrast of the observed regions depends on the content of main elements and oxygen. The results of phase microanalysis of magnets prepared from alloys 2 (Dy 0%), 6 (Dy 0.5%), and 11 (Dy 1%) are listed in Table 3. According to the microanalysis (M/A), in phases, the oxygen content is overestimated by 0.8–1.0 wt %. This conclusion derives from the fact that, first, there is no oxygen in the (Nd, Dy)₂Fe₁₄B grains, and, second, the M/A data on the oxygen content determined on a large surface area are an order greater than the results of chemical analysis (C/A) of the same magnets. Such discrepancy can be caused by oxidation of the polished surfaces which are employed for microanalysis. Taking into account this discrepancy, the composition formula was corrected to decrease oxygen concentration in each phase by a value equal to the erroneously determined oxygen in the (Nd, Dy)₂Fe₁₄B grains of each magnet.

For phase (B) (NdO_x) with the fcc structure, the ratio х = O/Nd is approximately ~ 0.7. In addition to Nd and O, this phase contains up to 10 wt % Fe, as well as negligible amount of doping elements М, i.e., Co, Cu, Ga; therefore, in Table 3, phase (B) is designated by the generalized composition formula (Nd,Fe,M)O_х. Let us note that Li et al. [34] showed by the 3D atom-probe tomography (3DAP) that the NdO_x phase had a large Fe content. According to their results, the composition of NdO_x corresponds to Nd_30.8Fe₄₅Cu_1.9B_2.8O_19.5.

In phase (С) (in Table 3, its composition is designated by the generalized formula (Nd, Fe, M)O_х), the oxygen content is negligible (x ≤ 0.05). However, Fe content of this phase is high; the ratio r = Fe/(Nd + Dy) can reach ~3. Besides, Cu and/or Ga are primarily localized in phase (С); generally, their contents in this phase exceed their total concentrations in a magnet. Phase (С) localized in the triple junctions is connected through its acute angles with thin light-gray Nd-rich intergranular boundaries (IGB), which separate the grains and are clearly visible in Fig. 5b. In the course of additional annealing of sintered Nd–Fe–B–M (M – Cu, Ga, Co) magnets at 520–600°С, phase (С) acts as a source for the liquid-phase diffusion of elements М, which spread over the IGB due to capillary effects [18, 35, 36]. It is worth noting the anisotropy of the microstructure morphology. Nd-rich inclusions are mainly localized at the Nd₂Fe₁₄B grain faces perpendicular to the с axis of easy magnetization (alignment axis is vertical in the figure plane); whereas, the IGBs separate the grain faces which form small angles with the с axis. Structural and chemical compositions of the IGBs were studied in detail by the high-resolution transmission electron microscopy and 3DAP in a series of works performed in Japan [9, 11, 12, 15, 16, 37, 38]. In these works, it was shown that after annealing at 520–600°С, the width of the IGBs is around 3 nm. The IGBs with planes mainly perpendicular to the с axis of the Nd₂Fe₁₄B grains are crystalline and enriched in Nd; whereas, boundaries parallel to the с axis are amorphous and enriched in Fe [15]. The chemical composition of amorphous IGBs determined by the 3DAP was Nd₃₀Fe₆₆B₃Cu₁ [12, 39]. There is no oxygen in this phase. Copper and gallium introduced into the initial alloy in content of around 0.1% are mainly localized in this phase in the shape of segregations at the IGB interface with the Nd₂Fe₁₄B grains. It was revealed by special methods, i.e., spin-polarized scanning electron microscopy [40] and X‑ray magnetic dichroism [19] that this phase is ferromagnetic at room temperature with the saturation magnetization around 80 G cm³/g. As could be expected from the amorphous structure, this phase is soft magnetic. The nucleation of the reverse magnetization can occur in this amorphous IGB phase. However, the magnetization reversal does not proceed easily in these interfaces, because the Nd₂Fe₁₄B grains and soft magnetic IGB layers are exchange-coupled. Due to the great difference in magnetic anisotropy constants of the Nd₂Fe₁₄B and IGB phases, domain walls of the emerging reverse nuclei will be pinned at the interfaces with the high gradient of domain wall energy, until the strength of the demagnetizing magnetic field reaches the values close to the coercivity, which is followed by the avalanche-like growth of the reverse nuclei and complete magnetization reversal of the magnet [22, 41]. Thus, it has been concluded that the main mechanism of magnetization reversal of the sintered Nd–Fe–B magnets with the ferromagnetic IGBs is pinning, rather than nucleation. Such view on magnetization reversal of the sintered Nd–Fe–B magnets agrees with the angular dependence of coercivity of these magnets, which has been studied by Matsuura et al. [42, 43].

The angular dependence of coercivity is an important tool to identify processes controlling magnetization reversal in hard magnetic materials. In order to determine the main processes of magnetization reversal of the magnets under consideration, we measured angular dependence of H_с for spherical samples cut out of magnets with 0.5 and 1.0 wt % Dy. Prior to measurement, samples were magnetized along the alignment direction in the pulsed magnetic field of a amplitude of 70 kOe. After placing a sample into a vibrating sample magnetometer, the alignment axis of the sample was turned by the angle ϕ off the direction of the applied field; then, the magnetization reversal was recorded. Figure 6 shows angular dependences of the relative coercivity h = H_с(ϕ)/H_с(0), where H_с(ϕ) designates values measured at the angle ϕ between the alignment axis and applied field; H_с(0), along the alignment axis.

Experimental dependences are compared with two theoretical models of angular dependence of coercivity. The Stoner–Wohlfarth model (S–W) considers coherent rotation of magnetization vectors. According to the S–W model, the critical field is equal to:where H_A = 2K₁/µ₀M_s is the anisotropy field, K₁ is the constant of uniaxial magnetic anisotropy, and M_s is the saturation magnetization. This model is used to qualitatively estimate a contribution from nucleation to the magnetization reversal. The Kondorsky model described by the following expression:

was initially developed to consider the pinning of domain walls (DW) at internal defects. Here, H_p is the pinning field required to move the DW from defects (H_p \( \ll \) H_A) upon the magnetization reversal along the easy axis of magnetization. Figure 6 shows that none of these theoretical models adequately describes the experimental results. The inset in the figure shows that for small angles ϕ < 30°, h < 1; i.e., the magnetization reversal is mainly controlled by the nucleation. Contrary, at 30° < ϕ < 70°, h > 1, but they are considerably lower than the 1/cosϕ dependence. This evidences that together with the main pinning mechanism of magnetization reversal, the hysteresis is also controlled by the nucleation. The curve h(ϕ) of the magnet with 1 wt % Dy has a maximum at ϕ = 60°; it appears because at ϕ > 60° the projection of the applied field on the alignment axis is not sufficient to demagnetize a considerable volume of the magnet. The higher Н_с(0), the smaller angles ϕ at which the maximum is observed.

Thus, the angular dependences of Н_с of the high-energy sintered Nd–Fe–B magnets show that their magnetization reversal is controlled by both the nucleation and pinning. Similar deviations of the angular dependences of Н_с from the Stoner–Wohlfarth and Kondorsky models were studied in detail by Bance et al. [44] by finite-element micromagnetic calculations. It has been shown that this deviation is caused by a number of reasons, including presence of soft magnetic defects at the grain surface and width of the IGBs.