Effects of Grain Boundary Phase on Coercivity of Dysprosium-Free Rare Earth Magnet
2016 Volume 57 Issue 11 Pages 1960-1965
Details
2016 Volume 57 Issue 11 Pages 1960-1965
As a part of the challenge of reducing the use of scarce rare-earth elements in magnets, a Dy-free Nd-Fe-B magnet with the remanence and coercivity of 1370 mT and 1830 kA/m, respectively, was investigated. The grain boundary was composed of mainly two phases, R6T13M and R-rich phases. The R6T13M phase formed at around 750 to 1000 K, and in that temperature range, coercivity improved and remanence decreased. By increasing the amount of grain boundary phases to 17.3 at% R addition, coercivity higher than 1990 kA/m (25 kOe) was realized.
Neodymium-iron-boron (Nd-Fe-B) sintered magnets were invented by Dr. Masato Sagawa et al.1) in 1983 and have been widely used worldwide because of their high magnetic properties. Unlike ferrite magnets, the coercivity of Nd-Fe-B magnets decreases at high temperature. Therefore, high coercivity is required in Nd-Fe-B magnets used under high temperature conditions. For that purpose, Nd or Pr is often substituted by heavy rare earth elements such as Dy and Tb. Generally, Nd-Fe-B magnets used in automobiles contain around 3 to 10 mass% of Heavy Rare-Earth Elements (HREEs), mainly Dy. As the use of Electric Vehicles (EVs) and Hybrid Electric Vehicles (HEVs) has expanded steadily in recent years, the consumption of HREE has also increased. Since the so-called “rare-earth shock” in 2010, a reduction of HREE usage has been an unavoidable issue for all magnet manufacturers and users.
Non-ferromagnetic phases such as a Nd-rich grain boundary phase have been considered to play an important role in increasing the coercivity of Dy-free Nd-Fe-B magnets.2) Because a Nd-rich grain boundary phase decouples the exchange interaction between Nd2Fe14B grains, a homogeneously distributed Nd-rich phase around each grain of Nd2Fe14B can magnetically isolate the Nd2Fe14B phase effectively. Many approaches to forming a thick and continuous Nd-rich grain boundary phase, such as Cu addition, have been examined.3) Ga addition is also known to heighten coercivity. Although this was considered to be the effect of wettability improvement4,5) rather than improvement of intrinsic magnetic properties6), the role of Ga in achieving high coercivity is still not clear7). Moreover, it is difficult to control the thickness of the grain boundary phase because it tends to accumulate at triple junctions.
While the formation of R6T13M by a small quantity of an additive such as Sn, Bi and Ga has been known for many years,8,9) annealing conditions affect the formation of the phase. However, the role of the phase has not fully understood.
The authors attempted to form the R6T13M phase at grain boundaries in order to control decoupling, and succeeded in making Nd-Fe-B magnets with a coercivity of 1830 kA/m without using Dy.10) In this paper, we report on the influence of grain boundary phases on the coercivity of a Dy-free Nd-Fe-B magnet, focusing on R6T13M phase formation.
All raw alloys were produced by the strip casting method and had the composition of (14–14.5) at% TRE (Total Rare Earth)-Co-B-M-bal. Fe (M = Cu, Al, Ga, Zr). The rare-earth elements contained in the alloys were Nd and Pr, and the amount of others was negligibly small.
The raw alloys were coarsely pulverized by a hydrogen grinding method followed by fine pulverization by a jet milling method to obtain fine powders of 2.5 to 3.5 μm in average diameter. The powders were pressed in a magnetic field of 1600 kA/m to obtain compacts. Those compacts were sintered for 4 hours at around 1273 to 1333 K in a vacuum, and were then annealed for 1 hour. The pulverizing, pressing and sintering processes were carried under a low oxygen atmosphere of less than 50 ppm.
The magnetic properties of the specimens were measured with a BH tracer (Tamakawa Co., Ltd.). Microstructural analysis was performed by FIB-SEM, EPMA, TEM and 3DAP (three-dimensional atom probe). For 3DAP, needle-like specimens with a diameter of 100 nm, which contained the grain boundary phase, were prepared. The specimens were biased by laser pulsing, and the evaporated atoms were detected by a position sensitive detector to obtain a 3D mapping of the elements. A line profile perpendicular to the grain boundary plane was taken from the 3D mapping in order to roughly estimate the grain boundary width.
The XRD measurement for the phase analysis was performed with synchrotron radiation. The measurement was carried out on beamline BL02B2 at SPring-8 in Hyogo, Japan. A sintered specimen was ground to the size of 0.24 × 0.24 × 15 mm and then sealed in a quartz tube capillary filled with inert gas. For the high temperature measurement, the specimen was heated to a desired temperature at a heating rate of 50 K/min, after which XRD measurements were performed twice within 4 min, and the specimen was then heated repeatedly to another desired temperature. The XRD profiles were measured with a wavelength of 0.412452(2) Å.
In order to evaluate the microstructure, the coverage ratio of grains by the grain boundary phase was estimated. Grains and grain boundaries were recognized automatically on SEM images by using an in-house image analysis software. Because it was not possible to recognize grain boundaries smaller than 20 nm, those grain boundaries were manually complemented on the image. The ratio of the length of the grain boundary which was automatically detected to the length of the grain boundary which was manually complemented was defined as the coverage ratio of grains by grain boundary phases with thicknesses of more than 20 nm.
A developed specimen with a Ga addition of 0.6 mass% was prepared. Its magnet composition and heating conditions were optimized to achieve the best balance of residual magnetic flux density, coercivity and squareness ratio. The developed specimen showed magnetic properties of Br = 1370 mT, HcJ = 1830 kA/m and Hk/HcJ = 97%, as shown in Fig. 1.10) Most of the analyses discussed below were performed with this developed specimen. For comparison, a conventional specimen with the composition of 14 at% Nd-Co-B-M-bal. Fe (M = Cu, Al, Zr), which showed a coercivity of 1080 kA/m, was also prepared.
Demagnetization curve of developed specimen.
The microstructures of the conventional and developed specimens were observed by SEM, as shown in Fig. 2. As shown in Fig. 2 (a), in the conventional specimen, a grain boundary is hardly observed between the grains but triple junctions are present. In contrast, in the developed specimen, a grain boundary is easily observed and is continuously connected to the triple junctions, as shown in Fig. 2 (b). Judging from the SEM contrast, several grain boundary phases coexist. The grain boundary phases were roughly identified by EPMA as R(Nd, Pr)-rich, Fe-rich and a small amount of ROx phase. Most of the Ga content existed at the grain boundary.
Cross-sectional images; (a) Conventional specimen, (b) Developed specimen.
In the SEM image, the R(Nd, Pr)-rich phase was observed as a bright area and the Fe-rich phase, as a dark area. The composition of these phases at the grain boundary was precisely investigated by using 3DAP. Line profiles perpendicular to the grain boundary plane were extracted from the 3D mapping of the elements.
A typical line profile across the grain boundary in the conventional specimen was measured, as shown in Fig. 3. The width of the grain boundary is less than 5 nm, and the Fe content is slightly lower than that in the grains. The width estimated from 3DAP almost coincides with the observational result by FE-SEM (field emission SEM) without correction. Although the concentration of R(Nd, Pr) was not constant between grain boundaries, most grain boundaries contained less than 30 at% R(Nd, Pr).
Compositional distribution at grain boundary of conventional specimen; (a) 3DAP mapping of Nd, (b) Line profiles of Nd and Fe.
The line profiles of the bright and dark phases at the grain boundary in the developed specimen were measured. Both the grain boundaries were thicker than that in the conventional specimen.
The bright phase was composed of more than 95 at% R (Nd and Pr) and less than 5 at% Fe, as shown in Fig. 4(a). The concentration of R (Nd and Pr) varied with the observation point, and was roughly from 70 to 100 at%. In all cases, the concentration was much higher than that of the conventional specimen.
Compositional distribution at grain boundary of developed specimen; (a) Bright phase, (b) Dark phase.
Next, an attempt was made to identify the Fe-rich phase, which is observed as a dark area at the grain boundary. A composition profile of the dark grain boundary is shown in Fig. 4(b). The concentration of each element in the grain boundary phase is almost constant, being 30 at% R (Nd and Pr), 65 at% Fe and 5 at% Ga. The Fe-rich phase at the grain boundary was identified as the R6T13M phase, which has a La6Co11Ga3 crystal structure, as shown by its diffractogram in Fig. 5. From this, the dark grain boundary phase is considered to be the R6T13M phase. Both R-rich and R6T13M phases are considered to be important for high coercivity, but the formation of the latter phase is unique to Ga added specimens. Therefore, the R6T13M phase was studied in detail.
Diffractogram of Fe-rich phase at grain boundary; (a) TEM image of Fe-rich phase, (b) Diffractogram taken at rectangular area in (a).
First, the relationship between the annealing temperature and magnetic properties was investigated. Developed specimens were prepared without annealing, and were then annealed at various temperatures for 1 hour, after which Br and HcJ were measured, as shown in Fig. 6. Coercivity drastically increases in the temperature range of 750 to 1000 K, and simultaneously with this, remanence decreases slightly. Generally, the densification and phase formation of Nd2Fe14B grains is completed in the sintering stage. Since the annealing procedure is performed at much lower temperatures than the sintering procedure, the Nd2Fe14B main phase, which decides magnetization, does not change in the annealing stage. However, both the increase of coercivity and the decrease of magnetization occurred in the annealing stage. This indicates that decomposition of the Nd2Fe14B main phase and formation of grain boundary phases occurred simultaneously in this temperature range. The decrease of remanence is not observed below 700 K or above 1100 K. This means the grain boundary phase which affects coercivity is formed only in the limited temperature range between 700 K and 1100 K. The cross sections of the developed specimens annealed at various temperatures are shown in Fig. 7. A thick, continuous grain boundary is observed in the specimens annealed at 773 K and 973 K, but this grain boundary becomes thinner and discontinuous at 1073 K.
Annealing temperature dependence of magnetic properties of developed specimens.
Cross sections of developed specimens annealed at various temperatures.
Next, the temperature range of grain boundary phase formation during annealing was studied more directly, focusing on the R6T13M phase. The temperature dependence of the XRD spectra in the developed specimen was studied in detail. Due to the low intensity of the R6T13M phase, synchrotron radiation was used.
The XRD profiles of the developed specimen without annealing were measured at high temperatures, as shown in Fig. 8. At lower temperatures, there are no peaks which indicate the R6T13M phase. The phase appears between 800 K and 850 K, but substantially disappears at 1075 K, probably because of decomposition. On the other hand, the peak of Nd-metal existing at room temperature disappears above 850 K, probably because of melting by a eutectic reaction with Cu.
Temperature dependence of XRD spectra of developed specimen during heating.
After sintering, the specimens were exposed once to the same temperature range as that used during annealing. However, it was thought that the rapid cooling after sintering might make formation of the R6T13M phase difficult. In order to confirm this, the specimen was once annealed to a high temperature at which the R6T13M phase disappears, and then rapid-cooled at the cooling rate of −100 K/min. The XRD profiles are shown in Fig. 9. When rapid cooling was applied, the R6T13M phase did not form.
Temperature dependence of XRD spectra of developed specimen during rapid cooling.
These results indicate that the formation of the R6T13M phase occurs in a certain limited temperature range and also requires a certain amount of time. Most importantly, the high coercivity temperature range shows good agreement with the R6T13M phase formation range. The slight difference in the temperature ranges for phase formation and for high coercivity could be caused by the difference in annealing time. Specifically, for measurement of magnetic properties, the specimens were annealed for 1 hour, while for XRD, annealing was performed for only 4 min at each temperature.
From the above, the R6T13M phase is considered to form during annealing by consuming Fe from Nd2Fe14B grains and the R-rich phase. Decomposition of Nd2Fe14B grains lowers remanence, while the formation of the R6T13M phase and lower-Fe R-rich phase improves coercivity. Formation of the R6T13M phase requires Ga as M. Thus, formation of the R6T13M phase occurs around the grain boundaries which have the highest contents of Ga, and it accelerates the surface decomposition of Nd2Fe14B grains. As a result, thicker grain boundaries form during annealing.
3.3 Magnetism of grain boundary phaseFor a high coercivity Dy-free Nd-Fe-B magnet, a non-ferromagnetic or low magnetic grain boundary is considered to be desirable. From this viewpoint, the magnetization of a grain boundary phase was estimated at a triple junction.
First, an electron holographic method was applied. The R6T13M phase was cut away and thinned for TEM observations by using a focused ion beam (FIB). A magnetic field was applied to this specimen, and its magnetization was estimated from the width of the interference fringes and the sample thickness. An image of this specimen is shown in Fig. 10. Since a small part of the Nd2Fe14B phase remains in the prepared specimen, non-uniform interference fringes are observed. From the difference of the interference fringes inside and outside the specimen, magnetization was estimated to be 0.05 T. Although the R6T13M phase contains as much as 65 at% of T (mainly Fe), it showed quite low magnetization.
Electron holographic measurement of R6T13M phase at triple junction; (a) Sample image for electron holography, (b) Electron holographic image.
To study magnetization more accurately, a STEM-Lorentz method was applied. Magnetization was calculated from the Lorentz polarization angle and the sample thickness.11) The line profile of magnetization across the R6T13M phase is shown in Fig. 11. By using this method, the magnetization of the R6T13M phase at the triple junction was estimated to be 0.04–0.06 T, which showed good agreement with the holographic result.
STEM-Lorentz measurement of R6T13M phase; (a) STEM image, (b) Magnetization map measured by STEM-Lorentz method, (c) Line profile of magnetization extracted from map (b).
Another grain boundary phase which is important for the coercivity of Dy-free Nd-Fe-B magnets is the R-rich phase. Recent studies on the magnetism of grain boundaries have revealed that some grain boundaries show ferromagnetic behavior. Hono et al. reported that the grain boundary phase of a Nd-Fe-B magnet contained 35 at% R (Nd and Pr) with μ0M of 0.4 T12). While the R-rich phase in the developed specimen contains 70 to 100 at% R, the magnetization of the phase is considered to be much lower.
From these results, both the R-rich and R6T13M phases in the developed specimen are considered to have much lower magnetization than the 1.6 T of the Nd2Fe14B phase. These phases can exist at a grain boundary and decouple the exchange interaction between Nd2Fe14B grains effectively.
3.4 Increase of grain boundary phaseThe introduction of the R6T13M phase suggests the possibility of forming a thick grain boundary with low magnetization. On the other hand, another simple method for forming a thick grain boundary is merely increasing the total content of the grain boundary phase.
Specimens with different TRE concentrations of 14.5 to 17.3 at% were prepared. The magnetic properties of these specimens are shown in Fig. 12. Higher coercivity was obtained with higher TRE concentration. The specimen with 17.3 at% R(Nd, Pr) showed magnetic properties of Br = 1181 mT and HcJ = 1998 kA/m. Coercivity of higher than 1990 kA/m (25 kOe) without using heavy rare-earth elements is remarkably good, considering the fact that a general Nd-Fe-B magnet requires from 6 to 8 mass% Dy to meet a similar requirement. A cross-sectional image of the specimen is shown in Fig. 13. Most grains are clearly separated by thick grain boundary phases.
Magnetic properties of developed Dy-free Nd-Fe-B magnets with increased amount of grain boundary phase.
Cross section of specimen with 17.3 at% TRE.
The increase in the grain boundary phase does not necessarily mean a thicker grain boundary. The excess R-rich phase tends to accumulate at triple junctions, where it does not contribute to thicker grain boundaries, although the existence of the R6T13M phase at grain boundaries could help to impede the accumulation at triple junctions. For decoupling of the exchange interaction between Nd2Fe14B grains, both grain boundary thickness and grain boundary continuity are considered to be important. Therefore, the coverage rate of Nd2Fe14B grains by grain boundary phases containing both R-rich and R6T13M phases was studied by image analysis. The sample composition examined here is (14.0–17.3) at% TRE-Co-B-M-bal. Fe (M = Cu, Al, Ga, Zr). The results are shown in Fig. 14. A good correlation between the coverage ratio and coercivity is observed. Moreover, as coercivity is still not saturated at a coverage ratio of 89%, further improvement may be expected.
Relationship between coverage ratio, ratio of grains covered by grain boundary phase and coercivity.
Coercivity of 1998 kA/m was achieved in a Dy-free Nd-Fe-B sintered magnet. In order to decouple the exchange interaction between Nd2Fe14B grains, a thick and continuous grain boundary phase with low magnetization is important. A R6T13M phase was introduced by adding Ga. The R6T13M phase is considered to form in a limited temperature range by consuming Fe from Nd2Fe14B grains and the R-rich phase. The magnetization of the R6T13M phase was estimated to be 0.05 T, which is much smaller than that of the Nd2Fe14B main phase. The coverage ratio of Nd2Fe14B grains by grain boundary phases affects the coercivity of the Dy-free Nd-Fe-B sintered magnet. The introduction of the R6T13M phase is considered to contribute to the reduction of the magnetization of the R-rich phase. Thus, both grain boundary phases play important roles in achieving high coercivity in Dy-free Nd-Fe-B magnets.
The authors are grateful to Dr. C. Moriyoshi and Dr. Y. Kuroiwa, Professors of Hiroshima University, for their valuable discussion of the XRD measurements at SPring-8.
