MATERIALS TRANSACTIONS
Online ISSN : 1347-5320
Print ISSN : 1345-9678
ISSN-L : 1345-9678
Influence of Cooling Roll Roughness on Nucleation and Growth of Primary Crystals in Strip Cast NdFeB Alloy
Kazuhiko YamamotoShinya TabataTakuya Onimura
Author information
JOURNAL FREE ACCESS FULL-TEXT HTML

2016 Volume 57 Issue 10 Pages 1789-1793

Details
Abstract

In general, the manufacturing process of NdFeB sintered magnets is as follows: preparing the starting alloys, crushing and pulverizing into fine powders, press-forming in magnetic fields for alignments, sintering, aging, crushing and cutting. It is well known that microstructure of starting alloys affect the ease of pulverization, press-forming for alignments in the magnetic field and sintering. A single roll casting process (hereinafter referred to as strip-casting) is a beneficial method for obtaining starting alloys with suitable microstructure. For this reason, strip-casting is regarded as a de-facto standard process for the production of starting alloys for NdFeB magnets. With this method, strips of 100–500 μm thickness are formed by pulling up from a melt by a single roll. For this reason, it is important to control primary crystal growth. In this study, we investigated primary crystal growth in starting alloys. First, Nd15.66Febal.B5.52 (atomic %) alloys were prepared using a strip-casting furnace. A batch size was 500 kg and roll diameter was 500 mm. Microstructures of obtained strip-casting samples were observed by electron probe micro-analyzer (EPMA). In this measurement, we focused on the morphologies of primary crystal growth and the distribution of constituent elements. Next we investigated the relation between the surface roughness of the rolls and frequencies of nucleation at the surface contacting a roll. The alloys were prepared using a strip-casting furnace. A batch scale was 5 kg and roll diameter was 300 mm. The alloy composition was Nd10.35Dy3.88Febal.B5.96 (atomic %). In this study, we also reviewed the method of reducing segregations of chill crystals existing in the microstructures of the starting alloys.

 

This Paper was Originally Published in Japanese in J. JFS 88 (2016) 154–159.

1. Introduction

In the early 1980s, Sagawa et al. developed a new type of sintered NdFeB magnets.1,2) Sintered NdFeB magnets are widely used in voice coil motors and their application has increased with the expanding market for hard disk drives in personal computers (PCs). Since then, the application of such magnets for high energy efficiency motors has been increasing gradually in various fields to achieve energy saving. Recently, sintered NdFeB magnets have also been used in home electrical appliances, electric vehicles, hybrid electric vehicles, and electronic power steering systems. These days, NdFeB magnets are an important component of many technologies that facilitate a more sustainable society.

The rare-earth and B contents of NdFeB magnets are slightly higher than those of the Nd2Fe14B intermetallic compound.2) The solidification process of these compositions occurs over a wide temperature range via solidification and peritectic reactions. The primary γ-Fe crystal precipitates around 1200℃, Nd2Fe14B phases form by the peritectic reaction at temperatures lower than approximately 1100℃, and the solidification finishes around 650℃.3) As a result, remaining primary α-Fe crystals and a huge segregation of rare-earth elements are observed in solidified ingots.

In the early days of producing sintered NdFeB magnets, the starting alloys were often prepared using the mold casting method. Generally speaking, the cooling rate of the mold cast method is relatively low due to the high thickness of the ingot. Therefore starting alloys produced using the mold cast method generally includes primary α-Fe crystals with segregation of the rare-earth elements.

According to a general fabrication process for RFeB magnets, starting alloys for magnets are crushed and pulverized using a jet-mill into powders with an average particle size around 3 μm and pressed while exposed to a magnetic field to align with the easy magnetization axis of the particles.2) After forming, the compact is sintered to achieve a dense alloy; the rare-earth rich melt phase acts as a sintering aid. The desired magnetic properties can be obtained by good alignment of easy magnetization axis of Nd2Fe14B and having a microstructure where the grains are magnetically isolated from each other by rare-earth rich (non-magnetic phases) phases. Hence, suitable starting alloys should satisfy the three following conditions in order to produce magnets with appropriate properties after sintering: 1) No remaining primary α-Fe crystals, 2) the pulverized powders are mono-crystalline, and 3) the R-rich phase (as the grain boundary phase) is finely and uniformly dispersed. Hirose et al. reported suppressing precipitations of primary α-Fe crystals and improving the magnetic properties of NdFeB magnets fabricated using strip-casting (SC) methods.4) Sugimoto et al. also reported specific morphologies around the area in contact with the cooling roll of SC samples.5) However, to the best of our knowledge, a thorough investigation of the relationship between SC conditions and the resulting microstructures of the fabricated alloys has not yet been published. Therefore, clarifying the nucleation mechanisms is critically important for optimizing the alloy microstructures.

In this study, we investigated the nucleation of Nd2Fe14B phases in starting alloys prepared using strip casting. First, we investigated the morphologies of the Nd2Fe14B phases and the distribution of the constituent elements in the SC samples. Next, we investigated the relationship between the surface roughness of the cooling rolls and the nucleation frequency at the surface in contact with the roll.

2. Experimental

2.1 Sample preparation

2.1.1 Samples for microstructural observation

SC samples were prepared in scaled mass-production batches. The Nd15.66Febal.B5.52 (atomic %) composition was selected as a suitable alloy composition for typical sintered magnets. A crucible was filled with 500 kg of raw materials including Nd metal where (Nd/(total rare-earth)) = 99.9 mass%, Fe-21.3 mass%B alloy, and 99.9 mass% pure-iron. The crucible was placed in a furnace chamber and evacuated down to a pressure of 2 Pa. Then, Ar was introduced into the chamber and the alloy was melted by induction heating at 1550℃. The samples were then cast by the SC method illustrated in Fig. 1. The melt was poured onto the tundish (W = 250 mm h = 20 mm) and guided through the cooling roll. A copper cooling roll with a diameter (D) of 500 mm was used. This roll was cooled with water and had a surface velocity (Vs) of 1.0 m/s. The obtained SC samples were 250 mm in width and 300 μm in thickness on average.

Fig. 1

A schematic of the strip-casting process (a) side view and (b) top view.

2.1.2 Samples for investigating the surface roughness of the roll

SC samples were prepared using a laboratory-scale process which could easily obtain uniform microstructures throughout the batch. The Nd10.35Dy3.88Febal.B5.96 (atomic %) composition was selected as this alloy is sensitive to changes in the microstructures. SC samples were prepared by vacuum induction melting as well as using the melting process described in the previous section. These melts were cast in 5 kg batches (a laboratory scale process) as illustrated in Fig. 1. The melt was poured onto the tundish (W = 40 mm, h = 10 mm) and guided through the cooling roll (D = 300 mm). Similar cooling rolls were used as for the previous experiments, but using a smaller diameter and a lower Vs = 0.9 m/s. The obtained SC samples were 40 mm in width and 300 μm in thickness on average.

Three different cooling rolls were used (A, B, and C) with different surface roughness values (as shown in Table 1). The cooling rolls labeled A, B, and C were polished using P60, P120, and P300 emery papers both parallel and normal to the rotation direction, respectively.

Table 1 Surface roughness Ra and average interval RSm values for cooling rolls A, B, and C.
  Roll A Roll B Roll C
Ra (μm) 4 6 11
Rsm (μm) 164 207 273

2.2 Microstructural observation

The microstructures of the chilled surfaces in contact with the cooling rolls and of cross-sectional SC samples were analyzed using optical microscopy (OM) and scanning electron microscopy (SEM). For the SEM analysis, the sample surfaces were highly polished to a mirror-like finish using 1 μm diamond slurry. For optical microscopy, these mirror polished surfaces were further etched using 3% HNO3-H2O. Back-scattered electron imaging and elemental mapping were carried out using an electron probe microanalysis (EPMA) instrument (JXA-8530F, JEOL).

2.3 Nucleation frequency

As shown in Fig. 2, the mean nucleation frequency (N) was determined from nave2/l2 where nave is the average number of grains crossed by three lines of length l drawn randomly on ten different optical microscopy images (×100 magnification) of the chilled surface in contact with a cooling roll.

Fig. 2

An illustration of the measurement method for counting nucleation sites.

2.4 Roughness measurements

The roughness values of the roll surfaces were measured using a Surftest SJ-301, surface roughness meter (Mitsutoyo Corporation), with a contact needle and an assessment length of 8 mm. The average arithmetic surface roughness Ra and an average interval RSm were calculated using the following equations.   

\[Ra = \frac{1}{l} \times \int_0^1 |f(x)| {\rm dx}\]
where l is the length of the measurement and f(x) is a function representing the ordinates of the measured roughness profile.   
\[Rsm = \frac{1}{n} \times \sum_{\rm i = 1}^{\rm n} Smi\]
where n is the number of peaks and valleys in the roughness profile and Smi gives the ith widths of the profile elements of the roughness profile.

3. Results and Discussion

3.1 Microstructure of Nd15.66Febal.B5.52 SC samples

Figure 3 shows optical microscopy images of Nd15.66Febal.B5.52 SC samples prepared in the scaled mass-production batches. Figure 3 (a) shows an image of the surface in contact with a cooling roll. Figure 3 (b) and (c) show cross-sections parallel to the thermal flow direction (as indicated by the arrows on the images) during solidification. SC samples prepared in the scaled mass-production batches had a tendency for uneven solidification due to the wide strips (width = 300 mm). Figure 3 (b) shows an area with a relatively fine dendrite structure and Fig. 3 (c) shows an area of relatively large dendrite structure. The downwards direction is the thermal flow direction for Fig. 3 (b) and (c). As seen in Fig. 3 (a), we observed dark circles with diameters <10 μm surrounded by dendritic structures. It is proposed that the dark areas are the nucleation sites of the dendrite crystal. Prates and Biloni reported similar microstructures for an Al-Cu alloy.6) Figure 3 (b) shows ultra-fine dendrite structures grown in various directions and small dimples (indicated by the black triangle) at the chilled surface. In previous studies,710) the dendrite size d has been represented as   

\[d = c/(G^a \times R^b),\]
where G is the temperature gradient, R is the growth rate and a, b, and c are constants.
Fig. 3

Optical microscopy images of Nd15.66Febal.B5.52 alloy samples (a) the surface in contact with the cooling roll (b) microstructures of the discs and fine particulate crystals and (c) microstructures of the dendrite crystals.

In the case of the strip-casting method, the temperature gradient can be assumed constant because the thin solidification phase moves in a molten puddle with a much larger volume. Hence, the growth rate of the dendrites shown in Fig. 3 (b) is higher than that of those shown in Fig. 3 (c). It can be seen that the ultra-fine dendrites occurred at the vicinity of the super-cooled area.

Figure 3 (c) shows the dendrite structure growing parallel to the thermal flow. These dendrites are relatively large in comparison with those shown in Fig. 3 (b) and hence we can conclude that these dendrites grew relatively slowly. The small dimples (indicated by the white triangles in Fig. 3 (b)) were observed in details using electron microscopy.

Figure 4 (a), (b), and (c) shows electron microscopy images of the areas around the small dimples. The gray areas are the Nd2Fe14B phase and the white areas are rare-earth rich phases. Figure 4 (a) shows white half circles (the dimples) surrounded by fine particles, with dendrites growing outwards in all directions from the area containing the fine particles. Figure 4 (b) shows the fine particles transforming to the dendrite structure along the thermal flow direction. The preferentially direction of growth for the Nd2Fe14B crystal is “a” axis of the tetragonal structure.11) Since the crystalline directions of fine Nd2Fe14B particles can be seen to be isotropic, these particles are thought to prevent the crystalline alignments of jet-milled powders during press-forming in a magnetic field for magnet fabrication. Figure 4 (c) shows a lower magnification image of the ultra-fine dendrite structures growing in various directions, where some small dimples are visible at the bottom of the image. It is clear that these dimples are acting as nucleation sites for the dendritic structures. Biloni and Chalmer reported similar nucleation areas for an Al-Cu alloy in 1965.12) It seems that the white circles and fine particles are the so-called “disc” and “pre-dendrite” structures, respectively, as reported by Biloni and Chalmer.

Fig. 4

Electron microscopy images showing the microstructure of Nd15.66Febal.B5.52 alloy samples (a) small disc, (b) large disc, and (c) discs and fine particulate crystals.

Figure 5 (a) and (b) shows a secondary electron image and a back-scattered electron image, respectively, of the area around disc. Figure 5 (c) and (d) shows elemental distribution maps of Nd and Fe, respectively, of the same area shown by the SEM images. It was found that within the disc microstructure, the content of Nd was higher and that of Fe was lower, than those in the Nd2Fe14B phase. From these results, we concluded that the nucleation occurred at the surface in contact with the cooling roll and the dendrites grew along the thermal flow direction from these disc-shaped nucleation sites. The disc and pre-dendrite microstructures were found at the super-cooled area of the surface in contact with the cooling roll. The disc was regarded as a phase with segregated rare-earths due to their high rare-earth content. It can be considered that the segregation prevented uniform sintering during magnet processing. Hence, the disc and pre-dendrite structures observed in the super-cooled area might result in poor magnetic properties.

Fig. 5

Elemental distribution around a disc feature in a Nd15.66Febal.B5.52 SC sample. (a) secondary electron image, (b) back-scattered electron image, (c) distribution of Fe and (d) distribution of Nd.

3.2 Influence of roll surface roughness on Nd10.35Dy3.88Febal.B5.96 samples

Figure 6 (a), (b), and (c) show optical microscopy images of the surfaces in contact with cooling rolls with Ra = 4, 6, and 11 μm, respectively. The number of nucleation sites decreased and their average diameter increased with increasing surface roughness. Figure 7 shows the relationship between Ra and the nucleation frequency. The nucleation frequency decreased from 940 to 230 sites/mm2 with increasing Ra from 4 to 11 μm, respectively. This suggests that the nucleation frequency was strongly dependent on the surface roughness of the cooling roll. Prates and Biloni reported that the nucleation of an Al-Cu alloy occurred at the rugosity of the mold when using thin plate casting.6) In this study, it seemed that the nucleation sites for the NdFeB alloy also occurred at the points of rugosity on the cooling roll surface.

Fig. 6

Optical microscopy images of the surfaces in contact with a roll for Nd10.35 Dy3.88Febal.B5.96 SC samples produced with rolls with average roughness; (a) Ra = 4 μm (b) Ra = 6 μm, and (c) Ra = 11 μm.

Fig. 7

Nucleation frequency N at the surface of Nd10.35 Dy3.88Febal.B5.96 samples in contact with the cooling roll as a function of the surface roughness Ra of the roll.

Based on the above results, we might be able to comprehend the morphology of nucleation and growth of the primary crystals, as depicted in the schematics in Fig. 8. The nucleation occurs at the points of rugosity on the surface of the cooling roll. When Ra = 4 μm (Fig. 8 (a)), the disc and chill grain sizes are smaller than those when Ra = 11 μm (Fig. 8 (b)). This is due to the higher density of contact points with the cooling roll with lower rugosity (Ra = 4 μm). The higher density of contact points results in a relatively lower thermal flow of each contact point. Hence, the surface roughness of the cooling roll surface seems to have a significant influence on the disc and chill grain sizes.

Fig. 8

Illustration of relationship between roughness of roll surface and chill crystals; (a) Ra = 4 μm and (b) Ra = 11 μm.

4. Conclusion

The following conclusions have been drawn from the microstructural characteristics, analysis of the distribution of the alloy elements, and investigation of the relationship between the surface roughness of the cooling rolls and the nucleation frequency of grains at the surface of the metal in contact with the cooling roll.

Disc and pre-dendrite features were found in the area close to the super-cooled surface of the SC samples. The discs were surrounded by pre-dendrite particles, which transformed into the dendrite structure along the thermal flow direction. The composition of the disc structures was higher in Nd, and lower in Fe, than the stoichiometric Nd2Fe14B phase. In SC samples where discs were observed, we also found super-fine dendritic structures growing in various directions.

The nucleation frequency decreased from 940 to 230 sites/mm2, and the disc and chill grain also become coarser, with increasing the surface roughness of the cooling roll from 4 to 11 μm.

Hence, we can conclude that the nucleation morphology of NdFeB alloys can be easily controlled by varying the surface roughness of the cooling roll. As the morphology of such alloy materials is critical in determining their magnetic properties, this technique is useful for obtaining optimized magnetic alloys for various applications.

REFERENCE
 
© 2016 Japan Foundry Engineering Society
feedback
Top