2020 Volume 61 Issue 3 Pages 438-443
We have investigated the microstructure in a Nd–Fe–B sintered magnet subjected to the grain-boundary diffusion (GBD) process with Dy2O3. It was found that thin (Nd,Dy)2Fe14B shells with a thickness of at least 10 nm and an average Dy concentration of approximately 0.7 at% were formed even at the center of the specimen, at a depth of approximately 1 mm from the specimen surface. However, the area fraction of the fcc NdO phase remained almost constant at approximately 3% throughout the magnet irrespective of depth, and the grain-boundary wetting phase hardly changed except in the vicinity of the surface. Furthermore, the thickness (ca. 3 nm) and structure (amorphous) of the thin grain boundary phase existing between Nd2Fe14B grains remained almost unchanged by the Dy-GBD process, despite the formation of the core–shell structure. Therefore, the Dy concentration in the thin Dy-substituted shell is important for improving the coercivity via the GBD process.
Fig. 6 STEM-EDS elemental mapping images near a typical grain boundary in the center (ca. 1.0 mm) of the Dy-GBD-treated Nd–Fe–B sintered magnet: (a) Nd-Lα, (b) Fe-Kα, (c) Al-Kα, (d) Pr-Lα, (e) Dy-Mα, (f) Co-Kβ, (g) Cu-Kα, and (h) Zr-Lα. White arrows indicate the width of the TGB phase and the depth of Dy diffusion.
The rapidly increasing demand for electric vehicles and various types of robots necessitates the development of high-performance Nd–Fe–B sintered magnets for the drive motors. However, as the coercivity of Nd–Fe–B magnets decreases drastically with increasing temperature, it is necessary to improve the coercivity at room temperature to ensure sufficient heat resistance. The most common approach for increasing the coercivity involves replacing a portion of the Nd with a heavy rare-earth (HRE) element such as Dy or Tb.1–3) However, it remains difficult to achieve both high heat resistance and high remanence because the remanence decreases with increasing HRE substitution. Furthermore, from the perspective of sustainability, it is also desirable to reduce the consumption of HRE elements.
Recently, HRE-free Nd–Fe–B sintered magnets with improved coercivity have also been developed.4) For certain applications, however, the effective use of HRE elements remains essential for realizing stable high coercivity. In this respect, Nakamura et al.5–7) developed the grain-boundary diffusion (GBD) process, in which the HRE elements become enriched along the grain boundaries of Nd–Fe–B sintered magnets to form thin HRE-rich shells around the core Nd2Fe14B grains. The key advantage of the GBD process is that the coercivity can be greatly improved without substantially decreasing the remanence, because the amount of HRE elements incorporated into the magnet is extremely small (approximately 0.2–0.3 mass%). However, it remains unclear how the HRE elements incorporated during the GBD process affect the microstructure and lead to the coercivity improvement. In particular, the thickness and composition of the HRE-rich shell inside the magnet are important, but there is no report that the formation of the HRE-rich shell was confirmed inside the magnets with a depth of >100 µm by scanning electron microscope (SEM) and electron probe micro-analyzer (EPMA) measurements.8–11)
In this study, the diffusion behavior of Dy in the interior of Dy-GBD-processed Nd–Fe–B sintered magnets was investigated via scanning electron microscopy (SEM) and scanning transmission electron microscopy (STEM) energy-dispersive X-ray spectrometry (EDS).
Sintered magnets with a nominal composition of Nd11.4Pr2.8Febal.Co1.1B6.1 in at% containing small amounts of Al, Cu, and Zr (0.1–0.2 at%) were used in this study. Plates of the sintered magnet with dimensions of 13 × 13 × 2 (easy axis) mm3 were immersed briefly in a mixture of Dy2O3 powder and ethyl alcohol and immediately dried in hot air. The powder-coated magnets were annealed at 1123 K for 5 h under an argon atmosphere with subsequent aging at approximately 773 K for 1 h. The magnetic properties of the original and the Dy-GBD-treated magnets were measured using a BH tracer. The specimens for the measurement were cut with dimensions of 10 × 10 × 2 mm3 from the center of the Dy-GBD-treated magnet. The demagnetization curves are shown in Fig. 1. The Dy-GBD treatment increased the intrinsic coercivity by approximately 30% (Hcj = 1.15 → 1.48 MA/m) without decreasing the remanence (Br = 1.40 T).
The demagnetization curves of the original and the Dy-GBD-treated Nd–Fe–B sintered magnets.
Microstructural analyses were conducted on SEM (Zeiss Ultra 55) and STEM (FEI Tecnai F20 and JEOL JEM-ARM200F) instruments equipped with EDS systems. The SEM specimen was prepared using a cross section polisher (JEOL SM-09020CP). The STEM specimen was prepared via Ga+ focused ion beam (FIB) milling (Hitachi FB-2000K) using the lift-off technique followed by low-angle Ar+ ion milling (Fischione 1010A) to remove the surface damage caused by the FIB milling.
SEM backscattered electron (BSE) images showing the typical textures at four depths from the surface to the center of the specimen are presented in Fig. 2, where the image contrast is nearly proportional to the average atomic mass or composition.12) Three kinds of Nd-rich phases with different contrasts in addition to the Nd2Fe14B main phase can be observed in all of the BSE images. Although it is difficult to distinguish the types of Nd-rich phases solely on the basis of the BSE images, they can be identified in combination with surface-sensitive secondary electron (SE) images. An example of the phase identification and color-coded classification is presented in Fig. 3, where the images were taken at a depth of approximately 0.1 mm from the magnet surface. From the SEM-SE and SEM-BSE images of the same area, it can be seen that at least three types of Nd-rich phases were generated, as circled in Fig. 3(a) and (b). These three phases were identified as NdO, Nd2O3, and grain-boundary wetting (GBW) phases via electron diffraction experiments. The NdO and Nd2O3 phases possessed fcc ($a \simeq 0.51$ nm) and hcp ($a \simeq 0.38$ nm, $c \simeq 0.60$ nm) structures, respectively, and were formed via oxidation during sintering. The GBW phase displayed numerous broad diffraction spots corresponding to an fcc ($a \simeq 0.53\unicode{x2013}0.55$ nm) structure and was formed by wetting and spreading to the grain boundaries of Nd2Fe14B mainly during the post-annealing process. It has been reported that the crystal structure of this wetting phase varies with the amount of oxygen13–15) and the presence of additive elements such as Cu,16,17) although the details are not yet well understood. At the moment, the nomenclature used for the wetting phase varies depending on the researchers;17–20) we refer to the spreading type as the GBW phase in this paper. A representative example of a color-coded SEM image prepared using an image processing program (ImageJ,21) Java-based public-domain software) is shown in Fig. 3(c), where the NdO and GBW phases are colored in red and green, respectively, and the small uncolored areas that are slightly lighter than the Nd2Fe14B main phase correspond to the Nd2O3 phase. Figure 4 shows the variation of the area fractions of the NdO and GBW phases with distance from the specimen surface, as determined by image analysis in a similar manner to that shown in Fig. 3. In the original magnet prior to the GBD process (closed symbols), the area fractions of both the NdO and GBW phases were approximately 3% and remained almost constant irrespective of depth. Even in the GBD-processed magnet (open symbols), the area fraction of the NdO phase remained almost constant at approximately 3%. On the other hand, the area fraction of the GBW phase increased toward the surface from a depth of 0.3 mm and was approximately 1.5 times higher in the vicinity of the surface. However, the area fraction of the GBW phase hardly changed at depths exceeding 0.5 mm. Heat treatment at approximately 773 K is known to cause the GBW phase to become liquid and then spread to the grain boundaries in the Nd2Fe14B phase.22,23) In the annealing step of the GBD process, therefore, the liquid GBW phase reacts with the Dy2O3 powder and the Dy-containing liquid phase then diffuses inside the magnet along the grain boundaries. The reason for the increase in the GBW phase only near the surface is considered to be that the internal liquid GBW phase migrates toward the surface Dy2O3 owing to the chemical potential for the reduction of Dy2O3. As can be seen from Fig. 1, the Nd2Fe14B grain size has hardly changed as the GBW phase increased. This is thought to be due to vacancy diffusion rather than body diffusion (interstitial diffusion), and is a feature of GBD processing that introduces Dy as an oxide.24)
SEM-BSE images showing the typical textures at four depths of the Dy-GBD-treated Nd–Fe–B sintered magnet: (a) 0.1 mm, (b) 0.3 mm, (c) 0.5 mm, and (d) 1.0 mm from the magnet surface (i.e., Dy diffusion depths).
Representative example of phase identification and color-coded classification of Nd-rich phases at a depth of approximately 0.1 mm from the surface of the Dy-GBD-treated Nd–Fe–B sintered magnet: (a) SEM-SE image, (b) SEM-BSE image, and (c) color-coded SEM-BSE image.
Variation of the area fractions of the NdO and GBW phases with distance from the magnet surface.
Figure 5(a) presents an SEM-BSE image showing the typical microstructure in the center (ca. 1.0 mm) of the magnet. It can be observed that the NdO and GBW phases formed at the triple junctions of the Nd2Fe14B grains, and further that the thin grain boundary (TGB) phase was formed continuously in between Nd2Fe14B grains. A TEM specimen was extracted from the center of the magnet using the FIB apparatus via the lift-off technique, and a detailed STEM analysis of the vicinity of the grain boundaries of Nd2Fe14B was conducted. A high-resolution TEM image of the TGB phase is presented in Fig. 5(b). The TGB phase was amorphous, possessed a thickness of approximately 3 nm, and adhered well to the surfaces of the Nd2Fe14B grains on both sides. The characteristic microstructures near the grain boundaries were almost identical to those in the original sintered magnet,25) and no structural changes due to the GBD treatment were observed. We investigated the behavior of the constituent elements near the grain boundaries in detail via STEM-EDS analysis.
(a) High-resolution SEM-BSE image showing the typical texture and (b) high-resolution TEM image near the TGB phase in the center (ca. 1.0 mm) of the Dy-GBD-treated Nd–Fe–B sintered magnet.
Figure 6 presents STEM-EDS elemental mapping images near a typical grain boundary in the center part of the Nd–Fe–B sintered magnet after the GBD treatment. The TGB phase in between two Nd2Fe14B grains was enriched in Nd, Pr, Dy, Al, and Cu, whereas the Fe concentration in the TGB phase had decreased. Furthermore, the Dy distribution was clearly wider than the thickness of the TGB phase in the other elemental mapping images, while the Nd concentration had decreased locally on both sides of the TGB phase. In other words, substitution of Nd with Dy had occurred in the surface areas of the Nd2Fe14B grains to form thin (Nd,Dy)2Fe14B shells with a width of at least 10 nm. Assuming that the Nd2Fe14B grains were spherical with a diameter of 5 µm, the (Nd,Dy)2Fe14B shells with a thickness of 10 nm accounted for only approximately 1.2% of the sphere volume, which indicates that Dy was concentrated in the shell part. Givord and Rossignol26) theoretically determined the temperature dependence of the domain wall width δw and reported that all of the domain wall widths in Nd2Fe14B, Pr2Fe14B, and Dy2Fe14B were $\delta _{\text{w}} \simeq 2\unicode{x2013}4$ nm at temperatures below 200°C. They also reported the thermal variation of the activation volume va by plotting va/δw3 for Nd2Fe14B and Pr2Fe14B sintered magnets,26) as shown in Fig. 7(a). The activation volume va for magnetization reversal is proportional to δw3. If va refers to a cube with side length l, the thermal variation of l against δw (l/δw) is approximately 2–3 times in the range from room temperature to approximately 150°C, as shown in Fig. 7(b). In the case of $\delta _{\text{w}} \simeq 2\unicode{x2013}4$ nm, therefore, the Dy-substituted shell with a width of approximately 10 nm is sufficient for suppressing the nucleation of reversed magnetization. Our previous atomic-resolution EDS analysis of a hot-deformed magnet27) revealed that the substitution with Dy was limited to a surface layer 2–3 unit cells thick in Nd2Fe14B grains (approximately 2.5–3.7 nm). This also suggests that a substitution shell with a thickness of 10 nm is sufficient for increasing the coercivity.
STEM-EDS elemental mapping images near a typical grain boundary in the center (ca. 1.0 mm) of the Dy-GBD-treated Nd–Fe–B sintered magnet: (a) Nd-Lα, (b) Fe-Kα, (c) Al-Kα, (d) Pr-Lα, (e) Dy-Mα, (f) Co-Kβ, (g) Cu-Kα, and (h) Zr-Lα. White arrows indicate the width of the TGB phase and the depth of Dy diffusion.
Thermal variations of (a) va/δw3,26) and (b) l/δw for Nd2Fe14B and Pr2Fe14B sintered magnets.
From the EDS analyses, the average chemical composition of the TGB phase was quantitatively estimated to be approximately Nd15.6Pr5.0Dy1.1Fe73.1Co2.3Cu1.4Al1.5 in at%. The TGB phase of this magnet had a relatively low rare-earth (RE) element concentration and a high Fe concentration compared with those reported for other sintered magnets.28,29) On the other hand, the overall average composition of the thin (Nd,Dy)2Fe14B shell was estimated to be approximately Nd11.7Pr4.6Dy0.7Fe79.7Co3.3 $ \simeq $ (Nd69Pr27Dy4)17(Fe96Co4)83 in at%. On the basis of STEM-EDS analysis of Tb4O7-GBD-processed Nd–Fe–B sintered magnets, Seelam et al.30) reported that the width of the shell near the center of the magnet was less than 100 nm with an average Tb concentration of approximately 0.66 at%. Although the width of the shell was somewhat different in that study, the HRE concentration introduced via the GBD process was generally consistent with our present result of approximately 0.7 at% Dy. This amount of Dy substitution corresponds to a total amount of RE elements of approximately 4 at% or 4.6 mass%. Upon alloying with Dy, the coercivity of conventional magnets was reported to increase at the rate of approximately 165 kA/m per 1 mass% Dy.31) In this study, the actual coercivity increase was 340 kA/m, which corresponds to alloying with approximately 2 mass% Dy. As conventional magnets contain approximately 31 mass% of RE elements, the Dy addition of 2 mass% corresponds to about 6.5 mass% (= 2/31) of the total substitution amount of RE elements. The EDS value of 4.6 mass% corresponds to about 70% of 6.5 mass% above, indicating that the formation of thin Dy-substituted shells even at the center effectively contributes to improving the coercivity.
As the eutectic temperatures32) of Nd–CuNd, Pr–PrCu, Nd–Nd3Al, and Pr–Pr3Al are much lower than those of Dy–DyCu and Dy–Dy2Al, the mass diffusivity of the TGB phase is considered to deteriorate with Dy substitution. Moreover, the ferromagnetic character should be stronger because the GBD process increases the Fe concentration in the TGB phase to above 70 at%. Therefore, the pinning force or the magnetic separation ability of the TGB phase would be expected to decline. In fact, however, the GBD process greatly increases the coercivity, which suggests that the contribution of the Dy-substituted shell to the coercivity improvement is very large. In addition, the Dy-substituted shell formed is very thin, while the quantitative change of the nonmagnetic Nd-rich phase is also small, and this change is limited to within a few hundred nanometers from the magnet surface. This is consistent with the fact that the GBD process hardly changes the magnetization.
The microstructures and local elemental distributions in a Dy-GBD-processed Nd–Fe–B sintered magnet were investigated via SEM and STEM. The results confirmed that the substitution of Nd with Dy near the surface of the Nd2Fe14B grains occurred to form thin (Nd,Dy)2Fe14B shells with a thickness of approximately 10 nm, in which approximately 4 at% of the RE elements were substituted by Dy, even at the center of the magnet. Owing to the very thin Dy-substituted shell and the small change in the area fraction of the nonmagnetic Nd-rich phase, the GBD process hardly changed the magnetization. Therefore, the GBD process represents an effective strategy for increasing the coercivity of Nd–Fe–B sintered magnets without reducing the magnetization, by effectively using small amounts of Dy.
