Journal of Crystal Growth 255 (2003) 286–292

Effect of steady ampoule rotation on axial segregation invertical Bridgman growth of Terfenol-D

Jong Chul Kima,*, Won Je Parka, Zin Hyoung Leea, Byung Joon Yeb

a Department of Materials Science and Engineering, Korea Advanced Institute of Science and Technology, 373-1 Kusong-dong,

Yusong-gu, Taejon 305-701, South Koreab Department of Metallurgical Engineering, Kyungpook National University, Teagu 702-701, South Korea

Received 24 January 2003; accepted 1 April 2003

Communicated by Dr. K.W. Benz

Abstract

In Bridgman growth of Terfenol-D, the axial macrosegregation can occur due to the compositional in-homogeneity

of initial melt. Since this melt segregation was hard to diminish by buoyant or solutal convection, we have

introduced the steady ampoule rotation technique to change the convection pattern and investigated the effect of

rotation speed and growth rate. The results showed that ampoule rotation changed the axial segregation and effectively

diminished the segregation at 60 rpm and 25 mm/s growth rate. The degree of the axial segregation was affected both by

the growth rate and rotation speed, and could be described as a function of a combined variable of rotation speed/

growth rate.

r 2003 Elsevier Science B.V. All rights reserved.

Keywords: A1. Steady rotation; A1. Convection; A1. Segregation; A2. Bridgman technique; B1. Rare earth compounds; B2. Magnetic

materials

1. Introduction

The giant magnetostrictive material, rare earthiron compound of Tb0.3Dy0.7Fe2, known asTerfenol-D, has significant technological interestsand wide application field due to its hugemagnetostrictive strains, high energy intensityand high response speed at low frequency. It waswell known that RFe2 phase (Laves structure, R

means the solid solution of rare earth constituents)

has a magnetic easy axis of /1 1 1S and further-more l1 1 1 is the largest of all the anisotropicmagnetostrains [1]. Therefore, Terfenol-D hasbeen produced as a single crystal or a grainaligned polycrystal to increase the efficiency ofmaterial. Especially, /1 1 1S oriented single crys-tals have the best energy efficiency and strainmagnitude; however, it is hard to grow along/1 1 1S direction because the RFe2 dendritesstrongly tend to grow along the preferred growthdirections, /1 1 0S or /1 1 2S [2]. Fortunately, the/1 1 2S direction make only an angle of 19.5� with/1 1 1S and the magnetostrictive properties of/1 1 2S are not significantly inferior than that

ARTICLE IN PRESS

*Corresponding author. Tel.: +82-53-950-5567; fax: +82-

53-950-7504.

E-mail address: jckim@casting.knu.ac.kr (J.C. Kim).

0022-0248/03/$ - see front matter r 2003 Elsevier Science B.V. All rights reserved.

doi:10.1016/S0022-0248(03)01305-8

of /1 1 1S, so that it is common to grow along/1 1 2S in commercial use.

Several techniques such as Bridgman [3], zonemelting method [4] and Czochralski [5] have beenused for unidirectional solidification of Terfenol-D. In a previous work of Bridgman growth [6], wereported that the axial macrosegregation was aserious problem in Bridgman-grown specimensespecially when a pre-alloyed ingot was remelted.In addition, this macrosegregation cause theinhomogeneous properties of specimen alonggrowth direction and deteriorate its quality be-cause the magnetostrictive properties stronglydepend on the composition. Therefore, the com-position should be controlled as well as thegrain orientation to obtain optimal properties ofTerfenol-D.

To reduce the macrosegregation, many techni-ques such as magnetic fields, the acceleratedcrucible rotation technique (ACRT) and vibrationhave been studied. In addition, a steady ampoulerotation technique was proposed and found useful[7–9]. These techniques are adopted to reduce thenatural convection and obtain the diffusion-controlled limit in vertical Bridgman configura-tion. However, our purpose of ampoule rotationis to diminish the initial melt segregation that hasoccurred during melting of charge rod. It wasreported that steady ampoule rotation couldsignificantly affect the flows and the dopantmixing [8]; hence, we investigated the effect ofsteady ampoule rotation on axial segregationof Terfenol-D in vertical Bridgman growth. Be-sides, the influence of both parameters, growthrate and rotation speed, on the segregation wasstudied.

2. Experimental procedure

2.1. Charge rod preparation

Tb0.3Dy0.7Fe1.80 alloys were prepared from rawmaterials, Tb, Dy and Fe of 99.9% purity. Theraw materials were alloyed using vacuum arcmelting furnace. The alloyed buttons were re-melted to cast a charge rod by vacuum suctioncasting. Fig. 1 illustrates the process with the

vacuum suction casting equipment. A high-qualityfused quartz crucible (2 mm thick, 20 mm innerdiameter) was positioned on the pedestal in thechamber. Chamber wall was made of also quartzglass to observe inside. The pedestal was movableto adjust the position of the crucible inside of theinduction coil. A quartz tube (+8 mm ID)penetrating the upper cap was connected to thepressure reservoir, which was connected to va-cuum pump and Ar gas bottle. The Ar pressure inthe reservoir was set to 750 torr, while rawmaterials were melted in the chamber under760 torr Ar. After the melt was mixed completelyby electromagnetic force, the quartz tube wasinserted into the melt and the melt was sucked upthrough the quartz tube due to the pressuredifference between the chamber and the reservoir.The filled tube was pulled upward from theresidual melt and cooled down. By this methods,charge rods of +8� 200 mm were obtained.Specimens were etched by 3% Nital solution (3%nitric acid in ethanol) for optical microstructuralexamination. For chemical composition analysis,ICP method was carried out.

ARTICLE IN PRESS

Fig. 1. Illustration of vacuum suction casting process.

J.C. Kim et al. / Journal of Crystal Growth 255 (2003) 286–292 287

2.2. Directional solidification

The surface of the rod obtained by the vacuumsuction casting was ground to remove the oxidefilm and rinsed with acetone and dried. Aftercleaning, the rod was put in a quartz tube of 9 mmID and sealed under 400 torr of high purity Aratmosphere. Directional solidification was carriedout using an apparatus shown in Fig. 2. Theapparatus consisted of an electric resistancefurnace, transportation part and rotation part.The furnace is elevated for directional solidifica-tion and quartz crucible was rotated around thevertical axis.

The specimen was heated to 1400�C, held for30 min and rotated with selected speed varied from25 to 60 rpm. The grown specimen was cuttransversely at intervals of 5 mm from bottomalong growth direction and the microstructureobservation was performed to investigate thevolume fraction of rare earth phase using imageanalysis program.

3. Results and discussion

3.1. Charge rod and abnormal segregation

Data in Table 1 refer to analyzed composition ofthe charge rod that was prepared for Bridgmangrowth. Compositional difference between lowerand upper part was small enough to say that thecomposition of charge rod was uniform alongthe rod axis. However, there were some losses ofrare earth elements during preparation of thecharge rod. The losses of Tb and Dy were 0.42 and0.19 wt% on average, respectively.

The optical microstructures of transverse sectionof grown specimen at a growth rate of 42.5 mm/swithout rotation are shown in Fig. 3. In bottompart (a), rare earth and eutectic RFe2 phase wereobserved between primary Laves phase (RFe2,bright phase), while the eutectic RFe2 phase wasdiminished in upper part (c). This is abnormalsegregation caused by initial solute content differ-ence in the melt along growth axis [10]. The RE-rich eutectic phase melts first in the quartz tubeand leaks out and flows down in the gap betweenthe quartz tube and the charge rod. Therefore, thelower part is rich in RE after melting. In thevertical Bridgman furnace, the temperature dis-tribution in the melt is stable and the density of theRE-rich melt in the bottom is higher than that ofresidual melt. Therefore, the initial RE-rich zoneat the bottom of the melt can be hardly mixed bynatural convection.

3.2. Effect of rotation speed

Fig. 4 shows the microstructures of transversesections grown at 42.5 mm/s and a rotation speedof 25 rpm. The tendency of macrosegregation with25 rpm was similar to that of no rotation, i.e.higher RE fraction in lower part and lower REfraction in upper part. In the case of 60 rpm and25 mm/s growth rate (shown in Fig. 5), the fractionof RE phase of upper part was nearly similar tothat of the lower part. The variation of atomicratio of Fe, y in RFey along growth axis is shownin Fig. 6. The atomic ratio of Fe was estimatedfrom the volume fraction of RE phase of themicrostructure [11]. Without rotation the RE

ARTICLE IN PRESS

Fig. 2. Schematic diagram of vertical Bridgman configuration.

The furnace is translated for moving solid/liquid interface.

J.C. Kim et al. / Journal of Crystal Growth 255 (2003) 286–292288

phase was the highest at the bottom of the crucibleand sharply decreased with the sample height. Thiskind of distribution would be similar to that of the

initial melt composition and was not changedremarkably at 25 rpm (Fig. 6a). When the rotationspeed was increased to 60 rpm at the same growth

ARTICLE IN PRESS

Table 1

Analyzed composition of charge rod and losses of rare earth elements (based on wt%)

Weighing composition (%) Analyzed composition (%) Loss

Tb Dy Formula Tb Dy Formula Tb Dy

Upper part 17.83 43.14 Tb0.3Dy0.7Fe1.85 0.37 0.29

18.20 43.43 Tb0.3Dy0.7Fe1.8

Lower part 17.74 43.34 Tb0.3Dy0.7Fe1.85 0.46 0.08

Fig. 3. Optical microstructures of transverse sections from

different axial position of grown specimen at a growth rate of

42.5mm/s without rotation: (a) 5 mm, (b) 15 mm, and (c) 25 mm

from the bottom of crystal. Total specimen length was 40 mm.

Fig. 4. Optical microstructures of transverse section, grown at

42.5mm/s and 25 rpm: (a) 5 mm, (b) 15 mm, and (c) 35 mm from

the bottom of crystal.

J.C. Kim et al. / Journal of Crystal Growth 255 (2003) 286–292 289

rate, 42.5 mm/s, the uniformity of y along growthaxis was significantly improved. Furthermore, theeffect of rotation on axial segregation wasconspicuous at a lower growth rate of 25 mm/s.The y value was nearly the same along growthdirection at 60 rpm as shown in Fig. 6b.

These results indicate that the ampoule rotationaffects the convection flow, which changes theaxial segregation. In vertical Bridgman growthsystem, when the solid/liquid interface is concave,the buoyancy flows typically consist of severalvortices that are driven by radial thermal gradients

as shown in Fig. 7a. A main vortex develops at thecentral region of the melt, where the temperature ishotter at the wall than at the center and the warmfluid rises along the ampoule wall and comes downalong the center. The lower vortex induced at thegrowth front rotates in the opposite direction, i.e.hotter in the center and cooler at the wall due tothe latent heat released by solidification [7,8].These two flow cells are independent and do notmix effectively each other; therefore, it is difficultto homogenize the initial melt composition, whichis significantly different between bottom and top.As a result, the initial melt composition wasmostly maintained during growth.

As the rotation speed is increased, the flowpatterns are expectably changed. The rotation of

ARTICLE IN PRESS

Fig. 5. Optical microstructures of transverse section, grown at

25.0mm/s and 60 rpm: (a) 8 mm, (b) 20 mm, and (c) 40 mm from

the bottom of crystal.

Fig. 6. Variation of atomic ratio of Fe, y in RFey, along the

growth axis: (a) 42.5mm/s and (b) 25 mm/s.

J.C. Kim et al. / Journal of Crystal Growth 255 (2003) 286–292290

ampoule introduces a centrifugal force that pushesthe heavier melt to the radial direction near thesolid/liquid interface. As a consequence, thedirection of forced convection is opposite to thatof buoyant flow; hence, the buoyant convectioncell on the interface is suppressed and pushedtoward the ampoule wall. The effect of ampoulerotation on the flow and solute segregation hasbeen studied by several researchers [7–9] and ourresults were in a good agreement. Yeckel et al. [7]reported that with increasing rotation speed theintensity of lower vortex decreases more rapidlythan that of main vortex so that the primary celldominates the flow intensity at higher rotationspeed. Lan [8] showed that the rotation pushed theflow cell toward the ampoule wall and stretched itaxially. The primary flow cell extends far into theliquid and fills the whole melt as shown in Fig. 7b.As a result, the change of flow patterns bringseffective mixing deeply into the bulk melt so thatthe axial segregation can be reduced. When thedifference of solute composition is large alonggrowth axis, elongated cell is much effective tohomogenize the whole melt.

Therefore, the axial segregation could be dimin-ished due to the deeper solute mixing into the bulkmelt with increasing rotation speed.

3.3. Effect of growth rate

In a vertical Bridgman growth, the growth rateaffects the interface shape, which influences con-vection pattern and intensity. As the growth rate isincreased, the growth interface is more concave tothe melt because the more latent heat is released atthe growth interface [12]. As a result, the radialthermal gradient is increased and the intensity oflower flow cell is increased.

As shown in Fig. 6, in the cases of 60 rpm, theaxial composition is uniform at 25 mm/s but theaxial segregation increases with increasing growthrate at 42.5 mm/s. In addition, as the growthinterface is more concave, the intensity of solutalconvection is increased due to the solutal bound-ary layer. The density of the rejected elements (Tband Dy) is higher than that of iron and thedirection of solutal convection flow is in the samesense with that of the thermal flow cell so that itenhances the intensity of the lower cell. Therefore,the increase in the growth rate causes the increaseof lower cell intensity and the effect of rotation isdiminished relatively.

Fig. 8 shows that how overall segregation ratio,ymax=ymin; depends on the combined parameter,the rotation speed divided by the growth rate. Thesegregation ratio decreases with not only increas-ing the rotation speed, but also decreasing thegrowth rate and shows a clear dependency on the

ARTICLE IN PRESS

Fig. 7. Schematic diagrams of convection flow in vertical

Bridgman configuration with a concave interface: (a) basic

buoyancy-driven convection without ampoule rotation and (b)

with ampoule rotation [7,8].

Fig. 8. Overall axial segregation ratio, ymax=ymin; vs. ratio of

rotation speed to growth rate.

J.C. Kim et al. / Journal of Crystal Growth 255 (2003) 286–292 291

combined parameter, the ratio of rotation speedto growth rate. Therefore, it is possible to estimatethe optimal rotation speed for a given growthrate.

4. Conclusions

The effect of steady ampoule rotation on axialsegregation in a vertical Bridgman growth ofTb0.3Dy0.7Fe1.8 alloys was studied at growth ratesof 25 and 42.5 mm/s. The results have demon-strated that:

(1) Forced convection by steady ampoule rota-tion changed the flow pattern in the Bridgmancrucible and enhanced an effective mixing ofthe solute boundary layer with the bulk melt.The axial segregation was effectively dimin-ished in 60 rpm at 25 mm/s growth rate.

(2) The axial segregation was also affected by thegrowth rate. As the growth rate is increased,the segregation was increased because theeffect of rotation became weak as comparedwith the convection cell intensity at theconcave interface.

(3) The axial overall segregation ratio, ymax=ymin;was diminished as the combined parameter,the ratio of rotation speed to growth rate,increased.

Acknowledgements

The authors wish to thank the Korea BasicScience Institute (KBSI) for the ICP analysis. Thiswork was supported by the Ministry of Scienceand Technology of Korea through the NRLProgram.

References

[1] M. Al-Jiboory, D.G. Lord, IEEE Trans. Magn. 26 (5)

(1990) 2583.

[2] J.D. Verhoeven, I.E. Ostenson, E.D. Gibson, O.D.

McMasters, J. Appl. Phys. 66 (2) (1989) 772.

[3] J.D. Verhoeven, E.D. Gibson, O.D. McMasters, H.H.

Baker, Metall. Trans. A 18A (1987) 223.

[4] O.D. McMasters, J.D. Verhoeven, E.D. Gibson, J. Magn.

Magn. Mater. 54–57 (1986) 849.

[5] Y.J. Bi, J.S. Abell, D. Fort, J. Crystal Growth 166 (1996)

298.

[6] W.J. Park, J.C. Kim, B.J. Ye, Z.H. Lee, J. Crystal Growth

212 (2000) 283.

[7] A. Yeckel, F.P. Doty, J.J. Derby, J. Crystal Growth 203

(1999) 87.

[8] C.W. Lan, J. Crystal Growth 197 (1999) 983.

[9] H. Lee, A.J. Pearlstein, J. Crystal Growth 240 (2002) 581.

[10] Z.H. Lee, B.J. Lee, M.S. Kang, S.M. Chung, S.R. Coriell,

J. Crystal Growth 141 (1994) 209.

[11] P. Westwood, J.S. Abell, I.H. Clarke, J. Appl. Phys. 64

(1988) 5414.

[12] C.E. Chang, W.R. Wilcox, J. Crystal Growth 21 (1974)

135.

ARTICLE IN PRESS

J.C. Kim et al. / Journal of Crystal Growth 255 (2003) 286–292292