Paper
Influence of Cooling Rate on Constituent Phases and Distribution of Elements in (Nd,Dy)-Fe-B Magnet Alloys
2016 Volume 63 Issue 7 Pages 630-635
Details
2016 Volume 63 Issue 7 Pages 630-635
Substitution of Dy for Nd is a widely used method to improve the thermal stability of sintered NdFeB magnets. However, there are serious concerns over the stable supply of Dy due to its looming shortage and rising prices. Therefore, it is essential to develop (Nd,Dy)FeB magnets with reduced Dy content. In this study, we investigated the influence of casting conditions of Nd10.85Dy3.21Fe80.12B5.82 alloys prepared by mold casting and strip casting on the constituent phases and elemental distribution to improve the magnet coercivity. Microstructural observations revealed the formation of primary crystals (α-Fe) at all selected positions in the mold cast samples. In contrast, no primary crystals were found in the strip cast samples. To obtain sintered (Nd,Dy)FeB magnets with high coercivity, the optimum thickness of the starting alloy was found to be 0.3 mm. The melting behavior was studied in a temperature range from 773 K to 1173 K and the melting temperature increased after pre-heating at 973 K and 1073 K for the strip cast samples. These results indicate that cooling below 900 K during strip casting influenced the melting behavior of the alloy. Furthermore, it could affect the densification during sintering of the magnets.
The NdFeB sintered magnets developed by Sagawa et al. in 19821) have attracted considerable attention due to their excellent magnetic properties and reduced environmental impact. The various application fields of NdFeB magnets have progressively advanced from voice coil motor (VCM) for PC to industrial motors to recent main motors for electric vehicles (EVs) and hybrid EVs (HEVs) and electric power steering (EPS) motors with high energy efficiency2). However, coercivity of NdFeB magnets needs to be significantly increased to improve their heat resistance because of the rather low Curie temperature of Nd2Fe14B (585 K)1). The emerging technologies used for the development of next-generation electric motors also need heat-resist magnets with lower Dy content.
As a result, great efforts have been dedicated to the development of R-Fe-B (R = rare earth) magnets with high coercivity. The following four methods are commonly used for the mass production of sintered NdFeB magnets: (1) Partial substitution of Dy for Nd; however, ores containing Dy often contain small amount of radioactive elements and removing radioactive materials from the ore is complex and expensive. In addition, high-quality ore deposits with low radioactive elements are distributed unevenly around the world limiting the stable supply of Dy. Therefore, it is important to reduce the Dy content in the (Nd,Dy)FeB magnets; (2) Grain size reduction of sintered magnets3,4); (3) Grain boundary diffusion treatment with Dy or Dy containing compounds on sintered magnets5,6); and (4) Optimization of dopant elements such as Dy, Ga, Cu, and Al and heat treatment conditions7).
On the other hands, to the best of our knowledge, studies on optimizing the microstructure of starting alloy to improve the coercivity are quite limited. In general, the manufacturing process of sintered NdFeB magnets includes the following steps: First, a starting alloy is prepared. It is then crushed and pulverized into a fine powder. Next, the powder is press formed in magnetic field for particle alignment followed by sintering, aging, and cutting8).
Hirose reported that the microstructure of starting alloys, especially the primary α-Fe crystals and the dispersion of R-rich phase affect the degree of pulverizing and sintering9). However, microstructural effects of the starting alloys on coercivity enhancement of sintered NdFeB magnets, have not been completely studied.
The objective of this study is to investigate the relation between microstructure and thickness of the starting alloys and determine the optimum casting conditions to prepare a suitable starting alloy for obtaining sintered magnets with high coercive force.
In this study, Nd10.85Dy3.21Fe80.12B5.82 was selected as the desired alloy composition, which is a typical composition of heat-resistant magnets for high temperature applications. Alloy samples were prepared by vacuum induction melting in an Ar atmosphere and casted using mold-casting method (hereinafter referred to as “M.C.”) or strip-casting method (hereinafter referred to as “S.C.”). The mold used for M.C. samples is shown in Fig. 1. This mold was designed to obtain several samples with the same composition and different cooling rates at a given time. S.C. sample was prepared using a casting system as shown in Fig. 2.
A schematic image of the mold used for M.C. samples.
A schematic image of the S.C. process.
Three test samples were cut out of the middle of the ingot along the direction of dotted line in Fig. 3 (along the heat flow direction during solidification) at 90 mm, 20 mm, and 5 mm from the bottom of the M.C. samples for microstructural evaluation.
Sampling positions of M.C. samples.
In the case of S.C. sample, microstructure in the middle of the strips along the thickness direction was examined. The backscattered electron images and element distribution maps were obtained using an electron probe micro analyzer (EPMA).
Quantitative analysis was performed at four different spots around the center of each type of R2Fe14B matrix grains, and an average value of the four measurements was used as the composition of the R2Fe14B phase.
The mean dendritic size was determined as follows: We first acquired five backscattered electron images for each position, followed by drawing a line of length L on each image perpendicular to the heat flow direction. For each image, we counted the number of dendrites (N) intersected by the line and calculated the L/N ratio. The average of all the five L/N ratios was considered as the mean dendritic size.
2.3 Differential thermal analysis (DTA)Differential thermal analysis (DTA) was performed to investigate the melting and solidification behavior of alloys for S.C. samples. The first DTA study was performed from 773 K to 1173 K at a rate of 3 K/min as shown in Fig. 4 (a). The next DTA study was carried out after heating the sample up to 973 K and 1073 K, as shown in Fig. 4 (b). It was held for 5 minutes at 773 K, 973 K, 1073 K, and 1173 K.
Differential thermal analysis procedures of (a) standard measurements and (b) pre-heating measurements.
Fig. 5 shows backscattering electron images (left), and elemental maps of Nd (center) and Dy (right) for I-(1), (2), and (3) of M.C. samples (from top to bottom) and the S.C. sample. The examined sample surfaces were along the heat flow direction during solidification. The respective thickness of the four abovementioned samples was 31.7, 7.1, 1.7, and 0.3 mm. Elemental mapping revealed the distribution of elements within each sample. The gray region corresponds to main phase R2Fe14B, the black region corresponds to primary α-Fe crystals, and the white region corresponds to R-rich phase in the elemental maps shown in Fig. 5 (a), (b), and (c). The sample thickness directly controlled the amount of constituent phases, which might affect the cooling rate during solidification. As a result, no primary α-Fe crystals were found in the S.C. sample, as shown in Fig. 5 (d).
Backscattered electron images (left column) and concentration maps (center and right columns are Nd, Dy respectively) of M.C. and S.C. samples by EPMA, (a) I-(1), (b) I-(2), (c) I-(3), (d) S.C. sample.
Fig. 6 shows the relationship between the mean dendritic size and thickness of I-(1), (2), and (3) of M.C. samples and the S.C. sample. It was observed that the mean dendritic size got finer with decreasing sample thickness. The mean dendritic size was reduced to approximately 2.9 μm for the thinnest S.C. sample.
Mean dendritic size (μm) as a function of alloy thickness (mm).
It is evident from Fig. 5 (a), (b), and (c) that Dy was uniformly dispersed in the main phase R2Fe14B and was not detected in the R-rich phase. Fig. 7 shows relative Dy content (Dy/(Nd + Dy)) at the center of the R2Fe14B matrix grains for each samples. The relative Dy content in the R2Fe14B phase reduced with decreasing sample thickness.
Relative Dy content (Dy/(Nd + Dy)) at the center of (Nd,Dy)2Fe14B matrix grains for I-(1), (2), (3) and S.C. samples.
A unique trend was observed in the case of S.C. sample (see Fig. 5 (d)). No Dy was detected in the R-rich phase. The relative Dy content at the center of R2Fe14B matrix grains for the S.C. sample was lower than those of the M.C. samples and Dy segregated near the interface of R2Fe14B and the R-rich phase. This nonuniform distribution of Dy (core-shell type structure) is clearly observed in Fig. 5 (d) (bottom-right).
More detailed studies were carried out along the surface perpendicular to the heat flow direction in S.C. sample. Fig. 8 shows (a) backscattered electron image and element distribution maps of (b) Dy, (c) Nd, and (d) Fe for the S.C. sample. The R-rich phase is observed around R2Fe14B phase in Fig. 8 (a). Nd concentration is higher in the R-rich phase and Dy is preferentially distributed within the R2Fe14B phase, as seen in Fig. 8 (b) and (c). In addition, the concentration of Dy in the shell region of the R2Fe14B matrix grains was relatively higher than that in the core region.
Backscattered electron image and a concentration map of S.C. sample by EPMA, (a) backscattered electron image, (b), (c), and (d) concentration maps of Dy, Nd, Fe, respectively.
The nonuniformity of Dy distribution for the S.C. sample increased with increasing cooling rate compared with those of the M.C. samples based on the abovementioned results. The R2Fe14B phase formed under rapid solidification of the S.C. sample might be in a metastable state, which could affect the subsequent melting behavior of the S.C. sample. The melting behavior of the S.C. sample was investigated using differential thermal analysis.
Fig. 9 shows the DTA curve of the S.C. sample for first cycle (under condition of Fig. 4 (a)). Endothermic reaction was observed during heating from 945 K to 1046 K up to 1173 K (blue line) and exothermic reaction was observed during cooling from 983 K to 900 K (red line). The onset melting temperature (endothermic peak) of S.C. sample was approximately 1020 K and the solidification of R-rich phase (exothermic peak) was at 979 K. The final solidification temperature was 900 K.
DTA curve obtained under condition of Fig. 4 (a) for S.C. sample.
Fig. 10 shows the DTA curves of first cycle (black curve) and two second cycles for the heating up process (red: first cycle was up to 973 K and green: first cycle was up to 1073 K as shown in Fig. 4 (b)). After comparing the DTA curves of first cycle with two second cycles, it is clear that the endothermic peak attributed to the sample melting was lower in the case of first cycle than in the two second cycles.
DTA curves obtained under condition of Fig. 4 (b) for S.C. sample.
Fig. 5 shows that a decrease in the M.C. sample thickness and/ or an increase in the alloy cooling rate results in smaller primary α-Fe crystals. Herman et al.10) and Umeda et al.11) investigated the relationship between cooling rate and growth of primary γ-Fe crystals for the NdFeB alloy. These studies demonstrated that quenching faster than a certain critical cooling rate could prevent the formation of primary crystals in peritectic reactions. The starting alloy should not contain α-Fe phase for obtaining sintered magnets with high coercivity. Based on these results, thickness of the starting alloy should be less than 1.7 mm.
Generally, starting alloys for the sintered magnets are crushed and pulverized into powders with 3–5 μm particle size and pressed in the magnetic field for particle alignment followed by sintering for densification.
Therefore, starting alloys with following two characteristics should be obtained the proper magnetic properties of sintered magnets: 1) All pulverized powders are monocrystalline 2) R-rich phase (as grain boundary phase) is dispersed finely and uniformly. To this end, the desirable dendritic size is approximately 3–5 μm, i.e., the optimum thickness of starting alloy is approximately 0.3 mm, as seen in Fig. 6. Ozawa et al.12), Sugiyama et al.13), and Strohmenger et al.14) studied the metastable phase formation in NdFeB ternary alloy. They were able to form Nd2Fe17 phase under rapid solidification condition but we could not obtain such metastable Nd2Fe17 phases in this study.
4.2 Effect of cooling rate during solidification on elemental distributionThe result of section 3.3 indicated that Dy distribution in the R2Fe14B phase was dependent on the cooling rates, which was consistent with its nonuniform (core-shell type structure) distribution observed in the S.C. sample. The Dy content was lower around the core solidifying at faster cooling rate in the initial stage of solidification. In contrast, Dy-rich shell was formed at a slower cooling rate towards the end of solidification. To conclude, the distribution ratio of Nd to Dy in the R2Fe14B phase was affected by the cooling rates.
4.3 Effect of heating on melting behavior of the S.C. sampleThe DTA curves of second cycles shifted to higher temperatures by heating at 973 K and 1073 K (Fig. 4 (b)), based on the results of Section 3.4. H. Sepehri-Amin et al. investigated the compositional change in R-rich phase by aging treatment for sintered magnets15). The compositional change in R-rich phase of S.C. sample was studied by heating the sample close to its melting point. It indicated that cooling below the final solidification temperature (900 K) during strip casting has influenced the melting behavior of the alloy. Moreover, it might affect the densification during green compact sintering of magnets.
4.4 Optimum microstructure of starting alloy for sinteringThe starting (Nd,Dy)FeB alloys prepared by fast cooling rates as in the S.C. method showed a decline in the melting temperature. This decline in the melting temperature could provide an effective densification of a green compact during sintering. The homogeneous and fine grained R2Fe14B phase could allow to obtain sintered magnets with high coercive force. As a result, this decline in the melting temperature might enhance the coercivity by preventing grain growth and accelerating densification at lower sintering temperature.
The optimum casting conditions for obtaining suitable starting alloy for sintered magnets with high coercivity are determined based on the following results of microstructural evaluation (size, constituent phase, and elemental distribution) and differential thermal analysis.
Three phases were observed (primary α-Fe crystal phase, main phase R2Fe14B and R-rich phase) in the case of M.C. samples, whereas, the S.C. sample showed only two phases, the main phase and R-rich phase with no α-Fe phase.
Mean dendritic size of M.C. and S.C. samples was strongly affected by thickness of the samples, which might be due to the thickness dependence of the cooling rate during solidification. The optimum thickness of the starting alloy with a suitable dispersion of R-rich phase is about 0.3 mm.
Dy content at the center of R2Fe14B matrix grains decreased with decreasing sample thickness. On the other hand, Dy enriched shells around the R2Fe14B matrix grains were observed in the case of S.C. sample.
The compositional change in the R-rich phase of S.C. sample was studied by heating the sample close to its melting point. This suggested that the starting alloy quenched below the final solidification temperature of 900 K in the case of S.C. sample could provide an effective densification at a lower sintering temperature.
