Journal of the Korean Physical Society, Vol. 63, No. 6, September 2013, pp. 1203∼1207

Effects of the Cooling Rate on the Microstructure of Nd-Fe-B AlloysFabricated by Strip Casting

Kyoung-Mook Lim∗ and Seung-Yeon Park

Liquid Processing & Casting Technology R&D Department, KITECH, Incheon 406-840, Korea

Sung-Yong Ahn

Corporate R&D institute, Samsung Electro-Mechanics, Suwon 443-743, Korea

Chul-Sung Kim

Department of Nano and Electronic Physics, Kookmin University, Seoul 136-702, Korea

(Received 22 April 2013)

The effects of the cooling rate on the microstructure of Nd-Fe-B alloys fabricated by using a stripcasting process have been investigated. In the cast strips prepared under a proper cooling rate, thegrowth of well-arranged Nd2Fe14B columnar grains occurs and the columnar grains exhibit apparentalignment along [00l]. The alignment coefficient of Nd2Fe14B columnar grain is highest at wheelspeed V = 4 m/s. The Nd-rich phase is well distributed enclosing the finest columnar grains of theNd2Fe14B phase in the strips prepared at wheel speed V = 4 m/s.

PACS numbers: 81.05.-tKeywords: Nd-Fe-B, Strip casting, Cooling rate, Microstructure, Grain alignment, Lamellar spacingDOI: 10.3938/jkps.63.1203

I. INTRODUCTION

Sintered Nd-Fe-B magnets, which were introduced bySagawa et al., are the representative high performancemagnets [1]. They are widely used as core materials inmany magnetic applications such as electric vehicles, ITdevices and hard disk drives. Although the recent de-veloped sintered Nd-FeB magnets have the similar max-imum energy product ((BH)max) to their ideal values,their coercivity values are only 12% of the anisotropyfield of Nd2Fe14B main magnetic phase [2,3]. Since theDy2Fe14B and TB2Fe14B phases have higher anisotropyfield than that of the Nd2Fe14B phase, some heavy rare-earth elements such as the Dy and Tb are now added intothe sintered magnets. However, these elements are veryexpensive due to their small amount of natural reservesand the addition of these elements reduces the satura-tion magnetization and the (BH)max. of the sinteredmagnets. Therefore, it is necessary to increase coerciv-ity of the sintered magnets without the addition of heavyrare-earth elements. Two methods to increase the co-ercivity without heavy rare-earth elements have becomeknown. The first one is a uniform distribution of the Nd-rich phase, which surrounds the Nd2Fe14B grains and re-duces the nucleation frequency of reverse domains. The

∗E-mail: mook@kitech.re.kr; Fax: +82-32-850-0430

second one is controlling grain size to be less than thesingle-domain size [3,4]. The manufacturing process ofthe sintered magnets consists of many production stepssuch as melting, casting, hydrogen decrepitation (HD)treatment, jet milling, magnetic pressing, sintering andheat treatments [5]. Therefore, the microstructure andmagnetic properties of the sintered magnets are affectedby each production process.

The strip casting (SC) is a very suitable method forpreparing Nd-Fe-B ingots in the melting and casting pro-cess. Since it is possible to obtain the high cooling rateduring the solidification in the strip casting, the forma-tion of α-Fe is prevented and a very fine lamella mi-crostructure composed of Nd2Fe14B and Nd-rich phases.The lamella microstructure of the SC alloys is effectivefor improving the magnetic properties of the sinteredmagnets. Also, there is an optimum cooling rate to ob-tain ideal microstructures and magnetic properties of theSC [6–8]. Therefore, in the present study, the effects ofthe cooling rate on the microstructure, grain alignmentand phase evolution of Nd-Fe-B alloys fabricated by stripcasting were investigated systematically.

-1203-

-1204- Journal of the Korean Physical Society, Vol. 63, No. 6, September 2013

Fig. 1. (Color online) Cross-sectional optical micrographsof cast strips for wheel speeds of (a) 2, (b) 3, (c) 4, and (d) 8m/s . The upper part is the free side, and the lower part isthe wheel side of the strips.

II. EXPERIMENTAL

Nd and Fe with purities of 99.9wt% and a FeB alloyconsisting of 16 wt% B and 84 wt% Fe were used as start-ing materials. Alloys with a composition of Nd33Fe66B1

(wt.%) were prepared by using a strip casting techniqueat wheel speeds being 2 - 8 m/s. The microstructureof the cast strips were investigated by using optical mi-croscopy (OM) and scanning electron microscopy (SEM).The distribution of Nd-rich phase in cast strip was evalu-ated by using the point counting method [9]. The averageinter-lamellar spacing of the Nd-rich phase was measurednear the wheel side of the cast strip. As the distribu-tion of the Nd-rich phase is improved, the average inter-lamellar spacing of the Nd-rich phase decreases. Also,the grain growth direction was evaluated by countingthe frequency of intersection points between the Nd-richphase and a line perpendicular to the wheel surface. Thefrequency of intersections was also measured near thewheel side of the cast strip and increases with increasingdifference between the crystalline growth direction andthe direction of the thermal gradient. The phase and thecrystalline orientation of the cast strips were character-ized by using X-ray diffraction (XRD) with Cu-Kα ra-diation and a Mossbauer spectrometer of the electrome-chanical type with a 50 mCi57Co source in an Rh matrixin the constant-acceleration mode.

III. RESULTS AND DISCUSSION

The obtained cast strips were about 460, 400, 350,and 150 µm in thickness for the increasing wheel speed

Fig. 2. (Color online) SEM backscattered electron micro-graphs of the cross section of cast strips for wheel speed of(a) 2, (b) 3, and (c) 4 m/s. The upper part is the free side,and the lower part is the wheel side of the strips.

(V) from 2 m/s to 8 m/s respectively. The optical mi-crographs of the cast strips are shown in Fig. 1. Asshown in Fig. 1(a), the volume fraction of the colum-nar grains is less than 70% in the cast strip preparedat V = 2 m/s and the columnar grains grow with ran-dom orientations. Well-arranged columnar structure isobserved in cast strips prepared at V = 3 m/s and 4 m/sin Figs. 1(a) and (b), where the columnar grains growalong the direction perpendicular to wheel surface andthe volume fraction of the columnar grains is over 80%.Figure 1(d) shows two kinds of microstructure in the caststrip prepared at V = 8 m/s. There is relative uniformappearance near the free surface side, not only for thecolumnar main phase grains but also for Nd-rich phaseat grain boundaries. Close to the wheel side, no typicalcrystalline microstructure can be observed by using OMafter etching treatment.

Figure 2 shows the cross-sectional back scattered elec-tron images of the cast strips, the corresponding wheelspeeds are V = (a) 2, (b) 3, (c) 4, and (d) 8 m/s, re-spectively. The grey part is the Nd2Fe14B phase and thewhite part is the Nd-rich phase in the cast strips. Thecooling rate has a significant effect on the microstructureof the cast strips. For the cast strip prepared with V = 2m/s, α-Fe dendrites appear near the free surface (circledregion in Fig. 2 (a)) due to the slow cooling rate and theorientations of Nd-rich are randomly distributed. Forthe cast strips prepared with V = 3 m/s and 4 m/s, theuniform distributed Nd-rich phase with a width about0.1 µm divides the Nd2Fe14B phase with width rangingfrom 2 to 5 µm. The main phase lamella grains showparallel orientations. The Nd-rich phase is distributeduniformly and α-Fe is not present. For the cast strips

Effects of the Cooling Rate on the Microstructure· · · – Kyoung-Mook Lim et al. -1205-

Table 1. Temperature-dependent Mossbauer parameters for cast strips.

Sites

Mossbauer parameter j1 j2 k1 k2 c e

Hhf 278.7 339.9 283.2 301.8 281.2 249.5

∆EQ 0.09 0.35 0.15 0.08 –0.39 0.07

293kδ –0.25 –0.06 –0.13 –0.25 –0.14 –0.22

A(%) 14 14 28 29 7 8

Hhf 311.8 375.3 330.3 340.8 307.6 301.1

∆EQ 0.30 0.24 0.11 –0.02 –0.02 –0.024.2k

δ 0.19 0.09 –0.08 –0.09 0.004 –0.14

A(%) 11 12 29 30 9 9

Fig. 3. (Color online) Variations of (a) average inter-lamellar spacing and (b) frequency of intersection with thewheel speed.

prepared with v = 8 m/s, as shown in Fig 1. (d), thereis a region of very fine crystalline microstructure nearthe wheel surface (circled region in Fig. 2 (d)). Also,Nd-rich phase is randomly distributed inside grains andat the grain boundaries.

Since the cooling rate of strip casting process is fasterthan that of conventional casting, the temperature gra-dient in the strips higher than that in the ingots. Whenthe cooling rate is high enough, the nucleation of γ-Fephase is suppressed, and the peritectic reaction is ob-structed; thus, the Nd2Fe14B phase forms directly fromliquid phase, and no α-Fe forms [10]. For a wheel speed ofV = 2 m/s, the solidification rate in the cast strips is stillnot high enough to suppress the formation of α-Fe effec-tively (Fig. 2(a)). On the contrary, a very high coolingrate enhances the nucleation events at the wheel surfaceof the cast strip, which causes many fine Nd2Fe14B grainsto nucleate and grow. That is the reason for the forma-tion of the fine crystalline layer in the strip prepared at awheel speed of V = 8 m/s (Fig. 2(d)). Generally, solidifi-cation starts at a nucleation site on the wheel side of thestrips during the strip casting process. After nucleationevents, a columnar structure of the Nd2Fe14B phase isformed due to directional solidification, and the growthdirection of the grains is typically parallel to the direc-tion of heat flow during solidification [11]. Also, a proper

Fig. 4. (Color online) X-ray diffraction spectra of the pol-ished surface near the free sides of the cast strips prepared atvarious wheel speeds.

solidification rate and temperature gradient, which areboth closely related to the wheel speed, are needed forthe formation of a columnar grain structure [12]. In thepresent study, wheel speeds of V = 3 m/s and 4 m/s weresuitable for providing a proper cooling rate that could re-fine and homogenize the columnar microstructures anddistribute the Nd-rich phase uniformly along the grainboundaries.

Figure 3(a) shows the average inter-lamellar spacingof the Nd-rich phase measured from the cast strips. Theinter-lamellar spacing of the cast strip decreases with in-creasing wheel speed. A lower wheel speed induces aslow heat flow from the melt to the wheel, which causesa thicker lamellar structure. Figure 3(b) shows the fre-quency of intersection of the Nd-rich phase accordingto the wheel speed. As mentioned above, the lamellastructure is consists of the Nd2Fe14B phase and Nd-rich phase parallel to the growth direction at a propercooling rate, and the heat flux is perpendicular to thewheel surface. Therefore, the frequency of intersection in

-1206- Journal of the Korean Physical Society, Vol. 63, No. 6, September 2013

Fig. 5. (Color online) Mossbauer spectra of the cast stripsprepared at the wheel speed of V = 4 m/s and measured at(a) 4.2 K and (b) 293 K.

Fig. 3(b) denotes a mismatch between the heat flux andgrowth direction of lamella during solidification. FromFig. 3(b), frequency of intersection decreases graduallywith increasing wheel speed. The random grown colum-nar microstructure of the cast strip prepared at a wheelspeed of V = 2 m/s induces a large mismatch between theheat flux and the growth direction, and the well-orientedcolumnar microstructures of the cast strip prepared atwheel a speed of V = 3 m/s and 4 m/s have a smallmismatch.

Figure 4 shows the X-ray diffraction spectra on thewheel side of the cast strips prepared at various wheelspeeds. Although the XRD pattern of the cast strip pre-pared at a wheel speed of V = 2 m/s shows an α-Fephase peak, no clear diffraction peaks of the α-Fe phasecan be observed in the XRD patterns of the cast stripsprepared at wheel speeds of V = 3 m/s and 4 m/s. Thesephase analysis results coincide with the microstructureobserved in Fig. 2. In order to confirm the absent of thenon-magnetic α-Fe phase in the cast strips, we performeda Mossbauer analysis of the cast strip. Figure 5 show theMossbauer spectra of the cast strip prepared at a wheelspeed of V = 4 m/s measured at 4.2 and 293 K. Usinga least-squares computer program, we fitted six sets ofsix Lorentzian lines corresponding to the occupancy ofthe Fe-ion probability distribution 8j1, 8j2, 16k1, 16k2,4c, and 4e sites to the Mossbauer spectra. The obtained

Fig. 6. (Color online) Effect of the wheel speed on thealignment coefficient of the grains in the cast strips.

Mossbauer parameters for the absorbers are listed in Ta-ble 1. The resulting area ratios of the sub-spectra 8j1,8j2, 16k1, 16k2, 4c, and 4e in various temperature rangesare 14, 14, 28, 29, 7, and 7%, respectively, at 293 K. Thevalues of the electric quadrupole splitting of each siteare found to be –0.39 mm/s < ∆EQA < 0.35 mm/s at293 K and –0.02 mm/s < ∆EQA < 0.30 mm/s at 4.2K. The temperature dependent isomer shifts (δ) valuefor the 8j2, 16k2, 16k1, 4c, 8j, and 4e sites of were –0.25mm/s < δ < 0.19 mm/s. Therefore, the Fe valance stateswere Fe+0 of in the alloy states. The magnetic hyper-fine fields for the Fe sites decrease in the order Hhf (8j2)> Hhf (16k2) > Hhf (16k1) > Hhf (4c) > Hhf (16j1) >Hhf (4e) at 293 K. These results show that only the hardmagnetic Nd2Fe14B phase exits and that other Fe-relatedphases are absent.

In Fig. 4, the cast strips consist of Nd2Fe14B andNd-rich phases. The indexed strong peaks indicatethe Nd2Fe14B phase. A clear texture of the tetrago-nal Nd2Fe14B phase in the prominent (004), (006) and(008) peaks appears, which shows that a directional crys-talline growth is obtained during solidification. Theseresults show that the growth direction of the Nd2Fe14Bphase is perpendicular to the wheel surface. The de-gree of the columnar growth can be expressed by thealignment coefficient [12]. The alignment of Nd2Fe14Bgrains along the (00l) direction can be calculated by us-ing the formula ΣI(00l)/ΣI(hkl). The wheel sides of thecast strips prepared at different wheel speeds were an-alyzed by using XRD. Figure 6 shows the calculatedalignment coefficients of the cast strips. With increas-ing wheel speed, the alignment coefficient of the caststrip increases. These results are well matched with theresults of the frequency-of-intersection analysis in Fig.3(b). The growth of the columnar microstructure is re-lated to the solidification rate determined by the tem-perature gradient. The temperature gradient is mainlydetermined by the wheel speed. Therefore, the align-

Effects of the Cooling Rate on the Microstructure· · · – Kyoung-Mook Lim et al. -1207-

ment coefficient of the Nd2Fe14B phase is closely relatedto the wheel speed. When the wheel speed is low, thealignment coefficient is low due to the low solidificationrate induced by the slow heat flow from the melt to thewheel. When the wheel speed is too high, the greatly-increased nucleation rate causes some randomly-orientedgrains, so that the alignment coefficient of the cast stripis low. The highest alignment coefficient can be obtainedby adjusting the wheel speed. In the present study, thealignment coefficient of the Nd2Fe14B phase was highestwhen the wheel speed was V = 4 m/s, which means thatthe wheel speed provided a very proper solidification ratefor a columnar growth of the Nd2Fe14B phase along thepreferred orientation.

IV. CONCLUSION

The microstructures of NdFeB cast strips were an-alyzed by using cross-sectional OM, SEM images andgrain alignment analyses. Under a proper cooling rate,the growth of well-arranged Nd2Fe14B columnar grainsoccurs during strip casting process. The microstruc-ture of Nd33Fe66B1 cast strips is mainly composed ofNd2Fe14B grains that are apparent aligned along the[00l] direction. The Nd2Fe14B phase is surrounded bya Nd-rich phase, and no α-Fe phase is observed. Theaverage inter-lamellar spacing decreases and the align-ment coefficient increases with increasing wheel speed.The cast strip prepared at wheel speed V = 4 m/s showsthe highest alignment coefficient of the Nd2Fe14B phaseand a homogenous distribution of the Nd-rich phase withfine lamellar spacing. These analyses show that a deter-mination of the optimum cooling rate for the formationof a well-defined microstructure of NdFeB cast strips isneeded and that this optimum cooling rate depends onother process conditions, such as amount of melt and

composition.

ACKNOWLEDGMENTS

This work has been supported by the Korea Insti-tute of Energy Technology Evaluation and Planning(KETEP). The authors would like to thank the KETEPfor its financial support .

REFERENCES

[1] M. Sagawa, S. Fujimura, N. Togawa, H. Yamamoto andY. Matsuura, J. Appl. Phys. 55, 2083 (1984).

[2] Y. Matsuura, J. Magn. Magn. Mater. 303, 344 (2006).[3] M. Sagawa, Basic Principles of Nd Magnets, (Agne,

Tokyo, 2011).[4] J. D. Livingston J. Appl. Phys. 57, 4137 (1985).

[5] H. Q. Liu, B. Wang, G. B. Han and R. W. Gao, J. Magn.Magn. Mater. 285, 23 (2005).

[6] H. Yamamoto, M. Nagakura, T. Katsuno and T. Ya-mamoto, IEEE Trans. Magn. 26, 2595 (1009).

[7] I. Ahmad, H. A. Davies and R. A. Buckley, J. Magn.Magn. Mater, 157, 31 (1996).

[8] W. C. Chang, S. H. Wu, B. M. Ma, C. O. Bounds, J.Magn. Magn. Mater. 167, 65 (1997).

[9] S. E. Park, T. H. Kim, S. R. Lee, D. H. Kim, S. N.Namkung and T. S. Jang, J. Kor. Powder Metall. 18, 24(2011).

[10] X. Jing, Y. Shihong, Y. Dunbo, L. Zongan, L. Shipeingand L. hongwei, J. Rare Earths 24, 306 (2006).

[11] J. Bernardi, J. Fidler, M. Sagawa and Y. Hirose, J. Appl.Phys. 83, 6396 (1998).

[12] H. Q. Liu, B. Wang, G. B. Han and R. W. Gao, J. Magn.Magn. Mater. 285, 23 (2005).