Accepted Manuscript

Mechanism of texture enhancement in nanocomposite magnets during processof die upsetting coupled with Nd-Cu grain boundary diffusion

Xin Tang, Renjie Chen, Wenzong Yin, Jinzhi Wang, Xu Tang, Don Lee, AruYan

PII: S0925-8388(14)02748-0DOI: http://dx.doi.org/10.1016/j.jallcom.2014.11.100Reference: JALCOM 32654

To appear in: Journal of Alloys and Compounds

Received Date: 6 October 2014Revised Date: 12 November 2014Accepted Date: 13 November 2014

Please cite this article as: X. Tang, R. Chen, W. Yin, J. Wang, X. Tang, D. Lee, A. Yan, Mechanism of textureenhancement in nanocomposite magnets during process of die upsetting coupled with Nd-Cu grain boundarydiffusion, Journal of Alloys and Compounds (2014), doi: http://dx.doi.org/10.1016/j.jallcom.2014.11.100

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customerswe are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, andreview of the resulting proof before it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1

Mechanism of texture enhancement in nanocomposite magnets during process of die

upsetting coupled with Nd-Cu grain boundary diffusion

Xin Tanga, Renjie Chena∗, Wenzong Yina, Jinzhi Wangb, Xu Tanga, Don Lee c, Aru Yana* aKey Laboratory of Magnetic Materials and Devices, Ningbo Institute of Material Technology and

Engineering, Chinese Academy of Sciences, Ningbo 315201, People’s Republic of China bNingbo University of Technology, Ningbo 315211, People’s Republic of China cUniversity of Dayton, Dayton, Ohio 45469, USA

Abstract

The microstructure of die-upset nanocomposite magnets with different height reduction was

studied to investigate the mechanism of texture development in processes of die upsetting coupled

with Nd-Cu grain boundary diffusion. At beginning stage of die upsetting process, inhomogeneous

Nd-Cu grain boundary diffusion results in two kinds of different grain boundary microstructures. As

for adjacent grains with grain boundary phase, the stress-induced solution-precipitation underlies the

formation of the platelet-shaped grains and texture enhancement. Whereas there are two types of

texture enhancement for the neighboring grains without grain boundary phase: grain boundary sliding

and grain devouring. As die upsetting further proceeds, the stress-induced solution precipitation is the

dominant mechanism for texture enhancement due to fuller Nd-Cu grain boundary diffusion.

Key words: Grain boundary diffusion; Die upsetting; Nd2Fe14B/α-Fe nanocomposite

magnet; Texture

1. Introduction

Nd2Fe14B/α-Fe nanocomposite magnets have attracted much attention for their theoretically

excellent performance enhanced by the exchange coupling between the magnetically hard and soft

phases [1-6]. However, the magnetic properties (~23 MGOe) of nanocomposite melt-spun magnets are

by far less than the theoretical performance [7-10]. One of the main reasons is considered as the

isotropy of hard phase in melt-spun magnets. In fact, up to now, the nanocomposite magnets are

∗ Corresponding authors at: Key Laboratory of Magnetic Materials and Devices, Ningbo Institute of Material Technology and Engineering, Chinese Academy of Sciences, Ningbo 315201, People’s Republic of China. E-mail addresses: chenrj@nimte.ac.cn (Renjie Chen); aruyan@nimte.ac.cn (Aru Yan).

2

challenging to be processed into bulk anisotropic nanocomposite magnets with full density.

Die upsetting is an effective method to produce bulk anisotropic Nd-Fe-B magnets with

nanostructure and full density. It has been shown that the slightly under-stoichiometric precursors are

successfully prepared into anisotropic nanocomposite magnets with acceptable texture by hot pressing

and die upsetting, other relative studies show that the texture was further enhanced by adding Cu and

Ga but only trace volume fraction (< 5%) for soft phase was found contained in these anisotropic

nanocomposite magnets, and further reduction of RE content (where RE is rare earth) in precursors

resulted in sharp deterioration in the texture [11-14]. Some investigators have demonstrated that a

weak texture can be obtained in the highly under-stoichiometric amorphous precursors by severe

plastic deformation [15]. These studies unambiguously indicate that the texture in

under-stoichiometric alloy is improved to some extent by applying effective techniques, it is, however,

still far poorer than that obtained in the over-stoichiometric counterparts.

To improve the texture in nanocomposite magnet, it is necessary to get better understanding on

mechanism of texture enhancement. As for over-stoichiometric precursors, the solution-precipitation is

mainly accountable for the texture enhancement [16-18]. Specifically, the anisotropy of strain energy

of individual Nd2Fe14B grains under compressive stress underlies the texture development. In the die

upsetting process, the grains with their c-axes parallel to the press direction have low strain energy,

leading to a preferential growth. Meanwhile, the grains with their c-axes out of the pressing direction

have high strain energy and thus are inclined to dissolve into the grain boundary phase. Besides, the

grain boundary phase is believed to facilitate the grain boundary sliding and grain boundary migration,

as a result, misoriented grains rotate towards the preferential direction (i.e., the press direction) and

texture is further enhanced [19-21]. As for under-stoichiometric precursors, the mechanism of texture

enhancement is still far from well understood. Relative reports on weak texture development in

amorphous precursors demonstrate that the weak texture is attributed to a preferential nucleation of

Nd2Fe14B nanograins with a (00l) orientation during die upsetting process [15]. The other works have

shown that the mechanism of texture development in under-stoichiometric precursors is dominated by

slip of basal plane in Nd2Fe14B. The modulus of elasticity reflects bonding strength between atoms,

meaning the bonding strength between atoms in the basal plane is much larger than that along the easy

magnetization c-axis, during the die upsetting process, the slip along the basal plane accompanied with

rotation of basal plane will occur, consequently, easy magnetization c-axes of hard grains orient

3

parallel to press direction and texture is thus developed [12]. However, this kind of mechanism of

texture enhancement in under-stoichiometric alloys is short of convincing evidence.

Recently, we have fabricated the anisotropic nanocomposite magnets by hot pressing and die

upsetting coupled with Nd-Cu grain boundary diffusion [22]. The texture was improved to a large

extent in the die upsetting process. To investigate mechanism of texture development in this kind of

textured nanocomposite magnets, in this study, we observed microstructure of die-upset

nanocomposite magnets with different height reduction.

2. Experimental procedures

The low melting point eutectic alloy ingot with nominal composition Nd90Cu10（wt. %）was

produced by arc melting of 99.9 wt. % pure elements and subsequently subjected to melt-spin at a

speed of 40m/s under high purified argon protection. The as-melt-spun ribbons were ground in an

agate mortar to fine powder with size of 80-150 µm in a glove box with a controlled atmosphere. And

Nd90Cu10 powder mixed with the commercial nanocomposite powder MQP-15-7 with a size of 40~250

µm at a mass fraction of 8 wt. %. The as-mixed powders were hot pressed at 700oC under 270 MPa in

vacuum, then the hot-pressed precursors were deformed at strain rate of ~ 0.0013 s-1 in an oversized

die at 850oC under 105 MPa in high purified argon atmosphere until their different height reduction ε

(ε=10, 30, 50, 70, 80%) achieved. The crystal structure and phase analysis were carried out by x-ray

diffraction (XRD) with Cu Kα radiation. Detailed microstructure was identified by scanning electron

microscopy (SEM) and transmission electron microscopy/high-resolution transmission electron

microscopy (TEM/HRTEM) on a Tecnai-F20 system. The samples were pre-magnetized in a pulsed

field of ~70 kOe and then magnetic properties of all samples were measured along the direction

parallel to the pressing direction using a closed circuit BH apparatus at room temperature.

3. Results and discussion

Figure 1 presents the XRD patterns of the hot-pressed and die-upset magnets with different

degree of deformation. With the increasing deformation degree, the relative intensity of peaks for the

(004), (105) and (006) of Nd2Fe14B ascends, indicating the texture is enhanced gradually in die

upsetting process. Considering the addition of the low melting point Nd-Cu alloy, the diffusion of

Nd-Cu should be fuller and it will provide a liquid phase environment under a higher deformation

degree, which will facilitate stress-induced solution-precipitation and enhance texture. (These details

will be demonstrated later.) In the meanwhile, given consideration that the (110) reflection of α-Fe and

4

the (006) reflection of Nd2Fe14B nearly overlap, the intensity of α-Fe diffraction peak becomes

weakened with an increase of deformation degree, and when height reduction is over 50%, the (110)

reflection of α-Fe becomes unidentifiable.

The dependency of magnetic properties on height reduction is also studied. Fig. 2 shows the

demagnetization curves and magnetic properties of the magnet hot pressed at 700℃ and the die-upset

magnets with different height reduction. As a result of texture development in die upsetting process,

the remanence indicated in Fig. 2(b) monotonically increases from 7.71 kGs of hot-pressed magnet to

13.01 kGs of die-upset magnet with 80% height reduction, which is in a good agreement with XRD

patterns. The coercivity increases firstly from 8.65 kOe of hot-pressed magnet to 14.56 kOe of magnet

with 30% height reduction, and then decreases sharply to 9.17 kOe of die-upset magnet with 80%

height reduction. This was discussed in detail elsewhere [23]. The most noteworthy thing is that the

coercivity and remanence increase simultaneously when the height reduction ≤30%, which is

different from the dependence of the magnetic performances on the hot deformation degree [24].

Moreover, the squareness of demagnetization curve is becoming superior with increasing ε, higher

height reduction leads to better texture and more homogeneous microstructure, thus the magnetization

reversal is more homogeneous and consistent [24-28]. When ε ascends to 80%, the maximum energy

product (BH)max increases to 37 MGOe. The remanence enhancement resulted from intergrain

exchange coupling is generally accountable for a high (BH)max in nanocomposite magnets. Here, the

high maximum energy product is attributed to not only the remanence enhancement, but even more

importantly to high remanence [29] and excellent squareness of demagnetization curves caused by the

superior hard grain texture.

Fig. 3 shows the microstructure of die-upset magnets with ε=10%, 30%. In the beginning stage of

die upsetting process, as illustrated in Figs. 3(a)~(b), most grains are the equiaxed shape with size of <

100 nm and a small proportion of grains has developed into elongated grains. In the upper-left of Fig.

3(a), there is a dark area where Nd-Cu has not fully diffused into and the grains are not clear enough to

discern from each other. This dark area is similar with the microstructure of die-upset nanocomposite

magnet with 2 wt. % Nd-Cu addition [22]. When ε arrives 30%, the percentage of platelet-like grains

increases dramatically while the equiaxed grains still exist in this sample and the size of them ascends.

The difference in morphology of grains may result from the inhomogeneous Nd-Cu diffusion.

Specifically, in the Nd-Cu diffused area, the grains are likely to develop into platelet-shaped grains

5

with help of high pressure and temperature, whereas the grains tend to remain exquiaxed shape in the

area without Nd-Cu diffusion after die upsetting process, which will be confirmed by further

investigation of microstructure.

Figure 4 shows the TEM/HRTEM images of die-upset magnet with ε=10%. The composition

analysis corresponding to different positions in Fig. 4(a) is detailed in the table 1, the energy

dispersive X-ray spectrometer (EDS) analysis illustrates that the composition of position 1 is close to

that of Nd-Cu alloy. Position 2 should be the Nd-Cu liquid phase, inter-diffusing with nanocomposite

particles under the effect of pressure at high temperature. It can be observed that the grains around

position 2 are developed into elongated shape, which demonstrates that the grain boundary phase is

critical for formation of platelet-shaped grains. While the composition of position 3 corresponds to that

of the traditional MQIII magnet, where the total content of rare earth exceeds that of raw

nanocomposite powder, suggesting Nd-Cu liquid phase has diffused into this area.

From figure 4(b), die-upset magnet with ε=10% is consisting of equiaxed grains with a size of

50-200 nm and platelet-shaped grains with size of 50-200 nm thickness and 200-800 nm length. The

random distribution and remarkable misalignment of grains result from the deficient deformation

degree. Further investigation of HRTEM shown in figure 4(c) illustrates that there is no grain

boundary phase between the equiaxed grains, but there is continuous and clear grain boundary phase

with 1-2 nm width existing between the platelet-shaped grains as shown in figure 4(d). According to

our previous work [22], these two kinds of grain boundary microstructures are similar to that of

die-upset nanocomposite magnets without and with Nd-Cu diffusion, respectively. Due to the

inhomogeneous grain boundary diffusion, the presence of these two kinds of grain boundary

microstructure gives rise to different mechanism of texture enhancement in these die-upset

nanocomposite magnets.

Figure 5 shows the TEM/HRTEM and Fast Fourier Transformation (FFT) images for the sample

with ε= 30%. In contrast to die-upset sample with ε=10%, the amount of the equiaxed grains with a

size of 100-200 nm diminishes while that of the platelet-shaped grains increases in this sample as

shown in figures 5(a) and (b). The low deformation level creates favorable conditions for observation

of intermediate state of grain growth and texture development. HRTEM, as shown in figures 5(c)-(e),

carried out for the grain boundary between adjacent equiaxed grains reveals that the absence of grain

boundary phase available for the solution-precipitation leads to the equiaxed grains maintaining the

6

same morphology after die upsetting process. Microstructure shown in figures 5(c)-(e) illustrates that

the scenarios of neighboring grains in these areas without Nd-Cu grain boundary diffusion includes

three following types:

(i); In figure 5(c), according to the inset FFT images, the (00l) planes of Nd2Fe14B phase is

easily marked in these images and the (001) plane of Nd2Fe14B with d-spacing of ~about 12.20 Å is

shown in this figure as well, hence the easy magnetization c axes of two grains perpendicular to (00l)

planes are indicated in Fig. 5(c). It shows their c axes are parallel with each other and their lattices

perfectly match each other’s. Assuming that the angle between c axis of grain and press direction (PD)

represents the amount of strain energy of this grain under pressure, two grains possess the same strain

energy and neither of them are preferentially energy favored. Therefore, these two grains would retain

as two separate grains with clear grain boundary after die upsetting as shown in Fig. 5(c). Nevertheless,

the texture in this case would not be enhanced, because neither of their axes become closer to press

direction.

(ii); In contrast, the two grains with similar crystal orientation, which are indicated in inset FFT

images of figure 5(d), have distinctive zigzag-shaped grain boundary. The lattice mismatch of the

adjacent grains is very small, the angle between c1 and c2 axes of two Nd2Fe14B grains is only 5

degrees. Reasonably, the gap of strain energy of two grains appears but is still insignificant. Due to the

fact that the angle between c1 and PD is a little smaller than that between c2 and PD, compared to the

grain with c2 axis, the grain with c1 is a little more energetically favorable. Considering c2 make only a

small angel with PD, the driving force is too small to glide these grains by grain boundary sliding.

However, it should be noted that the stress needed for the same strain rate is about several times higher

for the adjacent grains without grain boundary phase than that for the neighboring grains with grain

boundary phase [30]. Taking this factor into account, the local stress may be enough to drive grain

boundary slide. Hence, with the die upsetting process proceeding, it is more likely to happen that the

grain with c2 would slightly glide accompanied with rotation under resolved shear stress, so that the

strain energy could reduce to some extent. As a result, the c2 axis is closer to press direction and then

texture enhances. When the orientation of two grains is close enough, the two grains will almost merge

into one grain. As shown in figure 5(d), the lattice of grains almost join with each other. This finding is

consistent with Kwon’s model of texture development in under-stoichiometric alloy [12], and thus

7

supports Kwon’s model in microstructural perspective. The similar mechanism of texture formation

was also proposed in the over-stoichiometric alloy [31].

(iii); The angle between c1 and c2 of these two Nd2Fe14B grains shown in figure 5(e) amounts to

25 degrees, this suggests that two grains are orientated in quite different directions. In the bottom-left

of figure 5(e), the lattices of these two grains are separated by the clear grain boundary, while the grain

boundary becomes unidentifiable in the upper-right of this image, and a transition area (marked by

white dotted line), where lattices of these two grains overlap with each other, is documented in figure

5(e). The formation of this transition area is largely due to that the lattice of one grain diffuses into the

other one. Specifically, the lattice of one grain is gradually extending into the other one from

bottom-left end to upper-right end. Consequently, the width of transition area is rising with the

increasing distance far from bottom-left end where the grain boundary is clearly existent. This scenario

seemingly shows that one grain is devouring the other one.

These two grains orientate in different directions and the angle between c1 and c2 amounts to 25

degrees, this indicates that gap of strain energy between two grains is quite remarkable. Due to the

anisotropy of the elastic properties and strain energy of individual Nd2Fe14B grain under compressive

stress, strain energy of grain with c-axis far from pressure direction is much higher than that of grain

with c-axis close to pressure direction [17]. Therefore, it is more likely to happen that one grain with

low strain energy grows and the other grain with high energy will be eliminated gradually under

pressure. In this case, the grain with c2 axis close to the press direction would grow at the expense of

disappearance of the grain with c1 and texture thus develops. It is clearly observed in figure 5(e) that

one grain is devouring the other one by extending lattice into the other grain.

It’s necessary to mention that all these three types for grain growth and texture enhancement are

believed to be driven by reduction of strain energy stored in these grains, these three types of grain

growth kinetic are remarkably different from that proceeding with grain boundary energy as the

driving force for grain growth, where the grain growth occurs via grain boundary migrating from one

side with lower curvature to the other side with higher curvature [32, 33]. Therefore, the grain

boundary energy is excluded as a driving force when we discuss the grain growth and texture

enhancement.

8

Fig. 5(f) shows the existent grain boundary phase between two elongated grains. Nd-Cu diffusion

into grain boundary forming grain boundary phase facilitate the formation of platelet-shaped grains

and texture enhancement through the stress-induced solution-precipitation.

With the height reduction further increasing, texture has developed to a great extent accompanied

with grain growth as shown in figure 6, which is in a good agreement with the XRD and magnetic

performance analyses. In contrast to samples with ε=50%, in the die-upset magnet with ε≥70%, with

Nd-Cu relatively totally diffusing into the grain boundary, misaligned grains and equiaxed grains

remarkably diminish. The microstructure of magnet with ε=70% shown in figure 6(b) remains the

same as that of die-upset MQⅢ type magnets. Figure 6(c) indicates microstructure of die-upset

magnet with 80% height reduction, apart from grains with thinner dimension parallel to the pressure,

there is negligible difference in morphology between the sample with ε=70% and the one with ε=80%.

Figure 7 shows the further TEM analysis of these die-upset magnets with high height reduction

(ε=70%). In Fig. 7(a), after height reduction reaches 70%, the platelet-shaped grains with dimension

80–200 nm parallel to the compressive stress and 200–1000 nm perpendicular to the compressive

stress exist in die-upset nanocomposite magnets and the equiaxed grains almost completely disappear.

It reveals that Nd-Cu has diffused into grain boundary thoroughly and plays a critical role in the grain

growth and texture formation during die upsetting. Moreover, the clear and continuous grain boundary

phase between two elongated grains is observed. The thickness of grain boundary phase is generally

claimed to be below 5 nm as shown in Fig. 4(d) and Fig. 5(f), but in Nd-Cu over-diffused area, the

grain boundary phase is as thick as 10-20 nm as indicated in Fig. 7(b). Unlike die-upset magnet with

low ε (ε<50%), considering that Nd-Cu diffuses into grain boundary relatively fully, the grain

boundary without grain boundary phase is unlikely to be observed by TEM in die-upset sample with

high ε (ε≥50%). Hence, the texture development in the die-upset sample with high ε is dominated by

the mechanism of stress-induced solution precipitation.

4. Conclusions

In summary, in the die-upset nanocomposite magnet with low height reduction, inhomogeneous

Nd-Cu grain boundary diffusion gives rise to two kinds of grain boundary microstructures: grain

boundary with and without grain boundary phase, resulting in different mechanism of texture

enhancement. In the area with Nd-Cu grain boundary diffusion, Nd-Cu grain boundary diffusion

provides a liquid phase environment, which is critical for the formation of the platelet-shaped grains

9

and texture through stress-induced solution-precipitation. While in the area without Nd-Cu grain

boundary diffusion, there are two modes, depending on mismatch degree of adjacent grains, for texture

enhancement: grain boundary sliding and grain devouring. After height reduction increases to ~ above

50%, Nd-Cu grain boundary diffusion is relatively complete and the dominant mechanism for texture

development is stress-induced solution-precipitation. As a result, with die upsetting process

proceeding, the remanence rises from 7.71 kGs of hot-pressed magnet to 13.01 kGs of die-upset

magnet with 80% height reduction.

Acknowlegements

This work is supported by the Natural Science Foundation of China (No. 51101167, No.

60901047), National Science and Technology Major Project (No. 2012ZX02702006-005), the

Program of International Science and Technology Cooperation of China (No. 2010DFB53770), Local

Cooperation Project of CAS (DBSH-2011-013), the State Key Program of National Natural Science of

China (No. 50931001). Acknowlegement is also extended to Dr. Mingming Gong for fruitful

discussion.

10

Reference

[1] R. Coehoorn, D. B. Demooij, C. Dewaard, J. Magn. Magn. Mater. 80 (1989) 101.

[2] E. F. Kneller, R. Hawig, IEEE Trans. Magn. 27 (1991) 3588.

[3] A. Manaf, R. A. Buckley, H. A. Davies, J. Magn. Magn. Mater. 128 (1993) 302.

[4] T. Schrefl, J. Fidler, H. Kronmüller, Phys. Rev. B 49 (1994) 6100.

[5] H. Fukunaga, H. Nakamura, IEEE Trans. Magn. 36 (2000) 3285.

[6] H. Fukunaga, H. Nakamura, Scr. Mater. 44 (2001) 1341.

[7] R. Skomski, J. Coey, Phys. Rev. B 48 (1993) 15812.

[8] J. Bauer, M. Seeger, A. Zern, H. Kronmüller, J. Appl. Phys. 80 (1996) 1667.

[9] R. J. Chen, S. Guo, J. Z. Wang, D. Lee, A. R. Yan, J. Appl. Phys. 109 (2011) 07A756.

[10] D. Goll, M. Seeger, H. Kronmüller, J. Magn. Magn. Mater. 185 (1998) 49.

[11] A. M. Gabay, Y. Zhang, G. C. Hadjipanayis, J. Magn. Magn. Mater. 302 (2006) 244.

[12] H. W. Kwon, J. H. Yu, IEEE Trans. Magn. 45 (2009) 4435.

[13] D. Lee, J. S. Hilton, C. H. Chen, M. Q. Huang, Y. Zhang, G. C. Hadjipanayis, S. Liu, IEEE Trans.

Magn. 40 (2004) 2904.

[14] D. Lee, J. S. Hilton, S. Liu, Y. Zhang, G. C. Hadjipanayis, C. H. Chen, IEEE Trans. Magn. 39 (2003)

2947.

[15] Y. G. Liu, L. Xu, Q. F. Wang, W. Li, X. Y. Zhang, Appl. Phys. Lett. 94 (2009) 172502.

[16] L. Li, C.D. Graham Jr., J. Appl. Phys. 67 (1990) 4756.

[17] L. Li, C.D. Graham Jr., IEEE Trans. Magn. 28 (1992) 2130.

[18] W. Grünberger, Proc. 15th Int. Workshop on R.E. Magnets and their Applications, Dresden, 1998, p.

333.

[19] R. K. Mishra, J. Appl. Phys. 62 (1987) 967.

[20] R. K. Mishra, E. G. Brewer, R. W. Lee, J. Appl. Phys. 63 (1988) 3528.

[21] R. K. Mishra, T. Y. Chu, L. K. Rabenberg, J. Magn. Magn. Mater. 84 (1990) 88.

[22] X. Tang, R. J. Chen, W. Z. Yin, J. Z. Wang, X. Tang, D. Lee, A. R. Yan, Appl. Phys. Lett. 102 (2013)

072409.

[23] X. Tang, R. J. Chen, W. Z. Yin, J. Z. Wang, X. Tang, D. Lee, A. R. Yan, Scr. Mater. 88 (2014) 49.

[24] K. Khlopkov, O. Gutfleisch, D. Hinz, K-H. Müller, L. Schultz , J. Appl. Phys. 102 (2007) 023912.

[25] O. Gutfleisch, K-H. Müller, K. Khlopkov, M. Wolf, A. Yan, R. Schäfer, T. Gemming, L. Schultz, Acta

Mater. 54 (2006) 997.

[26] H. Kronmüller, D. Goll, Physica B: Cond. Matter. 319 (2002)122.

[27] E. A. Perigo, H. Takiishi, C. C. Motta, R. N. Faria, IEEE Trans. Magn. 45 (2009) 4431.

[28] R. Ramesh, K. Srikrishna, J. Appl. Phys. 64 (1988) 6406.

[29] Y. L. Ma, Y. Liu, J. Li, C. Y. Li, L. H. Chu, J. Magn. Magn. Mater. 322 (2010) 2419.

[30] W. Grünberger, D. Hinz, A. Kirchner, K-H. Müller, L. Schultz, J. Alloys Compd. 257 (1997) 293.

[31] Y. Luo, L. Zhang, Proc. 10th Int. Workshop on R.E. Magnets and their Applications, Kyoto, 1989, p.

275.

[32] M. Hillert, Acta Metall. 13 (1965) 227.

[33] J. E. Burke, D. Turnbull, Prog. Metal Phys. 3 (1952) 220.

11

Table’s Caption

Table 1. EDS analysis of positions marked in figure 4(a).

Figures’ Caption

Figure 1. XRD patterns of magnet hot pressed at 700℃ and die-upset magnets with different

height reduction.

Figure 2. (a) Demagnetization curves of magnet hot pressed at 700℃ and die-upset magnets

with different height reduction, (b) magnetic properties of hot-pressed magnet and die-upset

magnets with different height reduction.

Figure 3. Field emission SEM images for the cross section parallel to pressure direction of

die-upset magnets with (a), (b) ε=10% and (c), (d) ε=30%.

Figure 4. TEM/HRTEM images of die-upset magnets with ε=10%.

Figure 5. TEM/HRTEM and inset FFT images of die-upset magnets with ε=30%.

Figure 6. Field emission SEM micrographs parallel to pressure direction of die-upset magnets

with (a) ε=50%, (b) ε=70% and (c) ε=80%.

Figure 7. TEM/HRTEM images of die-upset magnets with ε=70%.

12

13

14

15

16

18

19

20

Position Pr(at.%) Nd(at.%) Cu(at.%) Fe(at.%) RE(at.%)

1 0 83.41 11.61 4.98 83.41

2 21.23 61.94 11.21 5.61 83.17

3 3.14 11.52 3.61 81.73 14.66

21

Highlights

• Inhomogeneous Nd-Cu diffusion in die-upset nanocomposite magnets

results in two kinds of grain boundary microstructures: grain boundary

with and without grain boundary phase.

• These two types of grain boundary microstructures underlie different

mechanism of texture enhancement in die upsetting process.

• Grain boundary sliding and grain devouring are accountable for texture

development in the two adjacent grains without grain boundary phase.