- Email: [email protected]
Temperature-dependent Dy diffusion processes in Nd–Fe–B permanent magnets
Available online at www.sciencedirect.com
ScienceDirect Acta Materialia 83 (2015) 248–255 www.elsevier.com/locate/actamat
Temperature-dependent Dy diffusion processes in Nd–Fe–B permanent magnets ⇑
K. Lo¨ewe,a, C. Brombacher,b M. Katterb and O. Gutfleischa,c a
Material Science, TU Darmstadt, Alarich-Weiss-Str. 16, 64287 Darmstadt, Germany b Vacuumschmelze GmbH & Co. KG, 63412 Hanau, Germany c Fraunhofer IWKS Project Group for Materials Cycles and Resource Strategy, 63450 Hanau, Germany Received 2 June 2014; revised 19 September 2014; accepted 21 September 2014
Abstract—Nd–Fe–B permanent magnets have been coated with 0.6 wt.% dysprosium and annealed at various temperatures to study the impact of the temperature-dependent Dy diffusion processes on both the magnetic properties and the microstructure. When optimum annealing conditions are applied the Dy processed magnets with initial coercivity of 1100 kA m1 yield coercivity increases which can exceed 400 kA m1 without a significant reduction of the remanent magnetic polarization. The improved stability against opposing magnetic fields can be observed up to a depth of 3 mm along the diffusion direction, restricting the application of the Dy diffusion process to either thin magnets or magnets with tailored coercivity gradients. While in the proximity of the Dy-coated surface, each grain has a Dy-enriched shell with a Dy content of 6 at.%; the Dy concentration decreases exponentially to 1.8 at.% after a diffusion depth of 400 lm and to 1 at.% after a diffusion depth of 1500 lm, as was found with wavelength dispersive X-ray spectroscopy and scanning transmission electron microscopy–energy dispersive X-ray spectroscopy, respectively. In the vicinity of the Dy-coated surface, the mechanism of the Dy-shell formation is attributed to the melting/solidification of a heavy-rare-earth-rich intermediate phase during high-temperature annealing. This is based on the observation that a constant Dy concentration over the width of the shells was found. Also an epitaxial relation between the Dy-poor core and the Dy-rich shell was observed by electron backscattered diffraction, which is supported by results obtained with Kerr microscopy. Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Nd–Fe–B; Permanent magnet; Coercivity enhancement; Grain boundary diffusion process; Microstructural characterization
1. Introduction Nd–Fe–B permanent magnets exhibit excellent magnetic properties and are widely used in industry as essential components of energy applications such as motors, generators, transformers and actuators [1,2]. Depending on the grain size, in commercial Dy-free sintered magnets a coercive field l0Hc of 1–1.5 MA m1 at room temperature can be realized, which is sufficient for most room-temperature applications [3,4]. For high-temperature applications, however, a higher coercivity is needed. This is commonly achieved through the addition of the heavy-rare-earth (HRE) dysprosium into the starting alloy, which is incorporated into the Nd–Fe–B lattice in place of the Nd atoms and leads to an enhancement of the anisotropy field Ha. As a drawback, the saturation magnetization MS as well as the energy product (BH)max is decreasing [5]. Other drawbacks are the high cost of Dy and the associated supply risk due to the monopolistic market situation and geological distribution [6,7]. To overcome these problems, the so-called grain boundary diffusion process (GBDP) was proposed by Park et al. [8], where the Dy is first deposited on to the magnet´s surface
⇑ Corresponding author; e-mail: [email protected]
and then diffused into the magnet during a post-sintering annealing treatment. The idea is to concentrate the precious element in the vicinity of the grain boundaries where nucleation and propagation of magnetization reversal is initiated [1,9,10]. As a result, a significantly reduced amount of Dy is required, which in turn allows the fabrication of high-coercivity magnets with a minimized loss of saturation magnetization. Studies in sintered magnets with other HRE compounds as the diffusion medium to increase coercivity, such as Tb [11], Dy and Tb fluorides [12,13] and eutectic compositions like DyNiAl [14], found that the diffusion occurs mainly over the grain boundaries and leads to a microstructure with grains exhibiting a HRE-rich (Nd,RE)2Fe14B shell and a HRE-poor core. Also the HRE diffusion in hot deformed Nd–Fe–B magnets has received a lot of attention in the last years [15–17]. The precise determination of the HRE element distribution in Nd–Fe–B sintered magnets after a GBDP is crucial for future optimization and has been addressed in recent publications. Samardzˇija et al. used Tb as the HRE and found a linearly decreasing Tb concentration in NdFeB up to a diffusion length of 200 lm [11]. Sepehri-Amin et al. have investigated the Dy concentration in Nd–Fe–B after a GBDP using three-dimensional atom probe (3DAP) tomography. They investigated a spot 2.7 mm away from the surface
http://dx.doi.org/10.1016/j.actamat.2014.09.039 1359-6462/Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
and found a significantly enhanced Dy concentration near the grain boundaries [18]. Our study confirms the rapid decrease of the HRE content within a HRE-rich shell upon increasing the distance to the HRE coated surfaces. Despite these research efforts, some fundamental phenomena have not been addressed yet. Neither an accurate determination of the temperature-dependent HRE diffusion coefficients along the grain boundary phase, nor a direct correlation between the coercivity and HRE gradients occurring in diffusion-processed Nd–Fe–B magnets has been established. It is the aim of the present work to evaluate the influence of the annealing temperature on the magnetic properties and microstructure of Nd–Fe–B sintered magnets after a Dy diffusion process and elucidate possible diffusion mechanisms. 2. Experimental Commercial sintered Nd–Fe–B magnets with a Dy concentration of 0.3 at.%, a size of 10 mm 10 mm 3 mm and a typical remanent polarization of 1.42 T were used as the sample material. A slurry of Dy powder in alcohol was applied either to one lateral surface with the dimensions of 10 mm 3 mm or to both pole surfaces with the dimensions of 10 mm 10 mm. The coated samples were individually wrapped in molybdenum foil and annealed at elevated temperatures Ta ranging from 600 °C 6 Ta 6 1050 °C for 6 h followed by low-temperature annealing at 500 °C for 2 h. All demagnetization curves were measured at a temperature of 20 °C using a commercial hysteresis graph from MAGNET-PHYSIK Dr. Steingroever GmbH. While the coercivity HcJ was determined directly from the demagnetization curves, the remanent magnetic polarization Br was calculated according to
249
600 °C 6 Ta 6 1050 °C are shown in Fig. 1 in comparison with uncoated reference samples which have been subjected to the same annealing treatment and typical properties of samples in the initial state. It can be clearly seen that the diffusion treatment leads to an increase in coercivity with a pronounced maximum at 900 °C (Fig. 1a), while the remanent polarization of the samples is not significantly affected by the diffusion process (Fig. 1b). For Ta = 900 °C, a coercivity of HcJ = 1510 kA m–1 was determined which corresponds to an increase in coercivity DHcJ of 420 kA m–1 when HcJ is compared with samples in the initial state. A comparison with uncoated reference samples reveals that 70 kA m–1 of this coercivity increase can be attributed to an improved annealing treatment and only a DHcJ of 350 kA m–1 is induced by the Dy diffusion process. Note that the reference sample annealed at Ta = 1050 °C is not showing a reduced coercivity with respect to samples in their initial state, indicating that no significant grain growth occurs during the annealing treatment. Thus, the reduction of DHcJ at annealing temperatures higher than 900 °C is very likely an influence of accelerated bulk diffusion, which limits the formation of Dy-enriched (Nd,Dy)2Fe14B grain boundaries. The demagnetization curves of Dy-coated samples annealed at Ta = 700, 800, 900 and 1000 °C are shown in Fig. 2 in comparison with a typical demagnetization curve of a sample in the initial state prior to the diffusion treatment. The increase in coercivity induced by the grain boundary diffusion process can clearly be observed. The squareness of the demagnetization curve is not substantially altered by Dy diffusion, which indicates a rather
Br ¼ Jr0 þ Jr0 N ðl 1Þ where l is the relative permeability, N the demagnetization factor and Jr’ the open circuit remanent polarization measured in a set of Helmholtz coils [19]. The dependence of the switching field on the distance to the coated surface was measured in open circuit conditions with a custom-built scanning Hall probe setup after applying appropriate reverse fields at room temperature. For the microstructure investigation a Zeiss Leo Gemini 1530 field emission gun (FEG) equipped with a solid-state detector energy dispersive X-ray spectroscopy (EDX) detector and an electron backscatter diffraction (EBSD) system from HKL was used. In addition, wavelength dispersive X-ray spectroscopy (WDX) line scans were obtained using a Zeiss ULTRAplus 55 FEG scanning electron microscope (SEM) with an INCAWave WD spectrometer and transmission electron microscopy (TEM) measurements were done using a JEOL JEM 2100F. For the preparation of TEM lamellae a FEI Helios 600i focused ion beam setup was used. Domain observations were carried out using Kerr microscopy (evico magnetics GmbH) and polar magneto-optic contrast.
3. Results 3.1. Magnetic properties The magnetic properties of samples coated on both pole surfaces with dysprosium powder and annealed at
Fig. 1. (a) Room-temperature coercivity HcJ of bulk samples after diffusion treatment for 6 h at 600 °C 6 Ta 6 1050 °C and (b) corresponding remanent polarization Br. The coercivity and remanence of annealed reference samples without Dy coating and the average HcJ and Br in the initial state prior to the annealing treatment are indicated for comparison.
250
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
Fig. 2. Demagnetization curves of Dy-coated samples after grain boundary diffusion processes at 700 °C 6 Ta 6 1000 °C for 6 h and of a reference sample in the initial state.
homogeneous coercivity throughout the complete sample volume. Consequently, the diffusion length of Dy has to be significantly larger than 1.5 mm, which is particularly remarkable for an annealing temperature of 700 °C and suggests the presence of an adequate amount of liquid phase, even in the low-temperature regime. In the results described above, the Dy was applied to both 10 10 mm large pole faces and consequently the Dy diffused along the thickness of 3 mm of the sample. In order to investigate the diffusion depth in more detail another set of samples was prepared where the Dy was applied to one of the lateral 10 3 mm large faces of the magnets. The local magnetization reversal of these samples after a diffusion treatment at 700 °C 6 Ta 6 1000 °C was measured with a scanning Hall probe setup. The switching field HS with respect to the distance d to the coated surface could then be determined in open circuit conditions. For this purpose the magnets were exposed to a reverse field comparable to the coercivity of the magnet. The regions of the magnet which have a lower coercivity than the applied reverse field were demagnetized where the region close to the surface where the Dy was applied remains magnetized in the initial direction. The distance of the transition line between these two regions is then determined by a Hall probe. By increasing the strength of the reverse field and repeating the Hall probe scans, the switching field was measured as a function of the distance from the surface. Fig. 3 shows the dependence of DHS (d) = HS (d) – HS (6 mm) for samples after diffusion treatment at different Ta and for a sample in the initial state. For the sample in the initial state a small reduction of DHS was observed in the vicinity of the lateral surface. This reduction is not related to actual material properties, but can very likely be attributed to edge effects as the measurements have been performed in open circuit condition. After the diffusion treatment, all samples show an increase in DHS in the proximity of the coated surface. The largest increase in HS was observed for Ta = 900 °C. For this diffusion temperature, the switching field was increased by 340 kA m–1 compared with the measurements done on the sample in the initial state and a penetration depth of 3–4 mm could be determined. These values correspond well with the previous measurements after Ta = 900 °C, where an increase in coercivity of 350 kA m–1 could be attributed to the diffusion treatment. Temperatures lower or higher than 900 °C lead to a reduction of DHS. Interestingly, the penetration depth
Fig. 3. Dependence of the switching field variation DHS on the distance d to the Dy coated lateral surface of samples after a grain boundary diffusion process at 700 °C 6 Ta 6 1000 °C for 6 h and of one sample in the initial state without diffusion process. Note that for better visibility, the error bars of Dx = ± 300 lm (spatial resolution of the Hall probe) and DHs = ± 40 kA m1 are not indicated.
is also showing a maximum for Ta = 900 °C. Higher annealing temperatures do not lead to larger penetration depths, and the switching field distribution after Ta = 800 °C is very similar to the switching field distribution after Ta = 1000 °C. This peculiar behavior might be explained by an increased bulk diffusion coefficient of Dy at 1000 °C compared to 900 °C. The resulting faster diffusion into the grains might lead to a faster depletion of Dy and therefore to a reduced penetration depth. 3.2. Microstructural investigation Fig. 4 shows SEM images of the near-surface microstructure of Nd–Fe–B permanent magnets after Dy diffusion at various temperatures for 6 h using backscattered electron (BSE) contrast. Compared to the microstructure of typical Nd–Fe–B sintered magnets which consist of (Nd,Dy)2Fe14B grains surrounded by nanometer-thick rare-earth-rich grain boundaries [9], the Dy diffusion processed samples exhibit distinct differences. First, the grains in the vicinity of the coated surface show an enhanced BSE signal and appear brighter, which indicates a higher average atomic weight as compared with grains further apart from the Dy-coated surface. These grains exist up to a distance of 10 lm at 700 °C to more than 50 lm at 1000 °C. Secondly, adjacent to this zone a pronounced core–shell structure of the (Nd,Dy)2Fe14B grains can be observed, as was mentioned in Section 1. The thickness of the shells decreases from 1 lm with increasing distance and is only visible up to a depth of 500 lm for annealing temperatures of 900–1000 °C. Due to the high amount of Dy in the outermost surface grains, the remanent magnetic polarization has to be substantially reduced compared to their initial state prior to the diffusion process. For practical applications one therefore needs to consider the avoidance or removal of such surface layers to retain a high efficiency of Dy usage and a high total remanence. In our study, the surface layer has a thickness of 150 lm after applying an optimum annealing temperature of 900 °C. Consequently, the impact of the removal of such surface layers on the production costs has to be considered when weighing the economic benefits of the reduced usage of Dy through the GBDP.
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
251
Fig. 4. BSE SEM images of near surface transverse sections of Dy-coated samples after annealing for 6 h at 700–1000 °C.
EDX maps of Dy, Nd and Fe (Fig. 5) reveal that the increase in average atomic weight of the (Nd,Dy)2Fe14B grains and (Nd,Dy)2Fe14B shells is induced by the replacement of Nd by Dy. While the Fe concentration in the (Nd,Dy)2Fe14B grains remains independent of the distance to the coated surface (the slight variation of the Fe contrast is most likely a measurement artefact due to the overlap of the Fe Ka peak with the Dy La peak), a clear reduction of the Dy concentration with increasing distance to the coated surface can be observed. Besides the formation of Dy-enriched (Nd,Dy)2Fe14B grains and shells, the diffusion process has also an influence on the rare-earth-rich grain boundary phases. At least two different Dy/Nd ratios can be distinguished in the EDX maps. One grain boundary phase is both Dy- and Nd-rich (Fig. 5, squares) while the other phase is only Nd-rich (Fig. 5, circles). It seems reasonable that the Dy-poor grain boundary phases consist of Nd-oxide phases which have been present prior to the diffusion process and remained solid at 900 °C. Thereby, the diffusion coefficient of Dy within the oxide phase is much smaller as compared to the residual rare-earth-rich grain boundary phase, which becomes liquid at 670 °C [20]. The impact of the various Nd-oxide phases on the actual diffusion process could not be determined and needs to be considered in future research activities. To study the mechanisms involved in the formation of the Dy-enriched shells, WDX line scans (Fig. 6) and EBSD mapping (Fig. 7) have been performed. The WDX line scan shows that both the Dy and Nd content remain constant within a shell. Thus, the formation by a conventional bulk diffusion process seems unlikely and the results support a model in which the surface region of the (Nd,Dy)2Fe14B grains is partially melted at elevated temperatures, as was proposed recently by Sepehri-Amin et al. [18]. Upon cooling, the Dy-enriched (Nd,Dy)2Fe14B phase solidifies and crystallizes on the matrix grains to form the Dy-rich shells. EBSD has been used to study possible orientation differences between the shells and the matrix grains in a sample
annealed at 900 °C for 6 h (Fig. 7). The differences in lattice parameters are small and therefore neglected (c-axis: ˚ , Dy2Fe14B = 11.83 A ˚ [5]). As a result, Nd2Fe14B = 12.03 A all of the values for the intragranular average misorientation angle are smaller than 1° (which is about the limit of the angular accuracy of EBSD) i.e. all the grains are perfect single crystals. This means that the shells grow epitaxially on the surface of the grains. Fig. 8 shows a region with a pronounced core–shell structure in BSE contrast and the Kerr image of the very same region. The nominal easy axis of the material is outof-plane. The Kerr microscope is used with polar contrast and thereby the sensitivity direction is parallel to the easy axis of magnetization and black and white areas can be referred to domains with magnetization along the easy axis. The pattern shows so-called star-like domains and is characteristic for uniaxial materials in this direction of observation. Due to the presence of a paramagnetic grain boundary phase and consequently a decoupling of the grains, intergrain magnetostatic interactions are expected to be dominant. However, in a few cases we found the magnetic domains to extend over the grain boundaries into neighboring grains, which might be attributed to exchange interactions between the grains. In by far the most cases there was no visible difference between shells and grains in the Kerr images (Fig. 8b). A correspondence of BSE and Kerr contrast with the core–shell structure can only be detected in exceptional cases (e.g. top part of Fig. 8c). As the majority of domains run uninterrupted through the core–shell interface a relatively undisturbed transition layer seems likely which corresponds well with the EBSD analysis in Fig. 7, where no orientation difference between shells and grains could be found. 3.3. Determination of the Dy gradient To quantitatively study the Dy gradients established by the diffusion process, WDX and TEM investigations were performed on a sample annealed at 900 °C for 6 h. Fig. 9
252
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
Fig. 5. BSE SEM image of a near surface transverse section of a Dy-coated sample after annealing at 900 °C for 6 h and corresponding EDX maps of Dy, Nd and Fe. Two examples of Nd-rich grain boundary phase are indicated by circles; two examples of both a Nd- and Dy-rich grain boundary phase are indicated by squares.
Fig. 6. (a) BSE SEM image of the transverse section of a Dy-coated magnet after annealing at 900 °C for 6 h and (b) corresponding WDX line scan showing the Dy and Nd concentration. The analyzed area was at a distance of 110 lm to the coated surface.
shows the results of WDX analysis performed within the shell regions at various distances d to the Dy-coated surface. The Dy content rapidly decreases from 6 at.% to values below 2 at.% at distances d larger than 400 lm. The Nd concentration shows the opposite behavior leading to a constant sum of the Nd and Dy concentrations of 12 at.%, which corresponds to the (Nd,Dy)2Fe14B phase. The exponential decrease of the Dy concentration within the shells can be used to determine the diffusion coefficient of Dy in the liquid grain boundary phase. The Dy profile was fitted to the Grube solution to Fick’s second law for a pair of semi-infinite solids [21]: x Cðx; tÞ ¼ csurf ðcsurf cbulk Þ erf pffiffiffiffiffiffiffiffi 4Dt
Therein csurf is the surface concentration, cbulk the concentration in the bulk, x the diffusion length, t the diffusion time and D the diffusion constant. We assumed that the diffusion occurs solely over a rare-earth-rich liquid eutectic phase along the grain boundaries of the otherwise solid grains. Furthermore, the diffusion coefficients are independent of concentration and there are no convective effects within the liquid. From the fit, the diffusion coefficient of Dy within Nd–Fe–B permanent magnets at 900 °C was determined to be DDy(900 °C) = (2.5 ± 0.5) 109 cm2 s–1. Note that although the absolute values of concentration might include errors due to measurement artifacts, only the slope of the curve is governing D, which improves the reliability of the obtained results.
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
253
Fig. 7. (a) BSE SEM image of the transverse section of a Dy-coated magnet after annealing at 900 °C for 6 h and (b) EBSD orientation map of the very same region. The shells show no detectable orientation difference to the inner parts of the grains. The sample shown in this figure was annealed at 900 °C for 6 h. The map was recorded close to the sample surface.
Fig. 8. Microstructure of a Dy coated magnet after annealing at 900 °C for 6 h. (a) Correlation between magneto-optical Kerr (top) and SEM BSE (below) contrast. The respective images show the exact same region of the sample. The easy axis of the material is pointing out-of-plane and the Kerr image is showing polar contrast. (b) and (c) show magnified views of (a).
There have been previous attempts to determine D of Dy in the system Nd–Fe–B. Cook et al. [22] assembled pairs of cast and annealed single-phase compositions of R2(Fe0.714Co0.286)14B where R = Nd or Dy. After a heat treatment at 1050 °C they measured the concentration gradient of Dy with EDX. Despite the higher annealing temperature, they found a substantially smaller diffusion coefficient DDy(1050 °C) of 2.9 1012 cm2 s–1. We believe that the huge difference to our value originates from the different amounts of liquid phase. As Cook et al. have used a Dy source which remains solid during the annealing treatment, the obtained average Diffusion coefficient is strongly
influenced by bulk diffusion. In our experimental setup, the Dy concentration is clearly dominated by diffusion along the liquid phase and consequently a diffusion coefficient which is three orders of magnitude higher is not unreasonable. As the thickness of the shells rapidly decreases with increasing diffusion length and the lateral resolution of the WDX measurements is only slightly below 1 lm, the accuracy of the determined Dy/Nd concentration for d > 200 lm is limited. To overcome the limitations of WDX, scanning transmission electron microscopy (STEM) EDX has been performed at a distance d of 1.5 mm (Fig. 10).
254
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
of 1.5 mm, the increase in Dy content at a distance of 10 nm to the actual grain boundary is 0.4 at.%. From sintered magnets, it is known that this increase in Dy leads to an increase in coercivity of 200 kA m–1 at room temperature. Thus, the increase of the Dy content approximately matches the observed increase in the switching field (Fig. 3). 4. Conclusion
Fig. 9. Depth profiles of the Nd and Dy concentration after Dy diffusion treatment at 900 °C for 6 h. Each data point was obtained from a WDX measurement in a Dy-enriched shell.
The analysis shows that the peak Dy content at the grain boundary is 1 at.%. The value must be regarded as a rough estimation due to the principal errors of the EDX method when evaluating low-energy peaks such as Dy Ma. Nevertheless, it is significantly higher than the average Dy concentration prior to the diffusion process. Hence, this measurement clearly demonstrates that Dy can easily diffuse more than 1.5 mm into a magnet body, which is in good agreement with the magnetic characterization of the switching field distribution revealing a penetration depth in the order of 3–4 mm. Focusing on the Dy gradient along the line scan, a penetration depth dP of 20–30 nm into the (Nd,Dy)2Fe14B grain can be determined. Compared with the WDX measurements of the Dy content within the shells, no plateau of the Dy content can be observed, indicating a different formation mechanism possibly involving bulk diffusion. Using the diffusion formula above a bulk diffusion coefficient of DDy(900 °C) 1 1015 cm2 s1 can be estimated. It should be noted that dP is subjected to substantial errors due to the thickness of the TEM lamella and unknown orientation of the grain boundary with respect to the beam, as well as the unknown amount of liquid phase during the annealing treatment. Nevertheless, as micromagnetic model calculations have revealed that the anisotropy of a 7 nm thick region around individual Nd–Fe–B grains has significant impact on the coercivity of the bulk magnet [23], a Dy enrichment up to a penetration depth of 20 nm to 30 nm can adequately explain the coercivity increase induced by the diffusion process. Finally the observed increase of the Dy content can be compared with the switching field distribution. At a depth
Fig. 10. (a) High-resolution TEM image of a grain boundary located 1.5 mm apart from the Dy-coated surface of a magnet annealed at 900 °C for 6 h and (b) corresponding EDX line scan.
The dependence of both microstructure and magnetic properties on the annealing temperature of a Dy diffusion process has been analyzed. An optimum annealing temperature of 900 °C could be determined for an annealing time of 6 h and the coercivity increase induced by the diffusion process was 350 kA m1. The switching field distribution revealed that an impact of the Dy diffusion process can be observed up to a distance d in the range of 3–4 mm to the coated surface. However, the benefits of the diffusion process rapidly decrease for d > 1.5 mm. Coating sintered magnets on all sides can partially compensate for this size restriction, but for many practical applications the maximum thickness of the diffusion processed samples is limited to 5 mm. The microstructure analysis shows that Dy-enriched grains and several lm thick Dy-enriched shells are present in the vicinity of the coated surface, leading to a reduction of the local remanent magnetic polarization. Although the impact on the integral remanence is small due to the restricted thickness of the deteriorated surface layer, the removal of this layer (e.g. by grinding) has to be considered to restore the full remanence. The mechanism of the Dy enrichment in the vicinity of the surface was reasoned to be melting of a Dy-rich intermediate phase and subsequent solidification. For larger distances to the coated surface the thickness of the Dy-enriched shells is reduced to several nanometers. At a distance d = 1.5 mm to the coated surface, the thickness and the Dy content were analyzed by STEM EDX and both values are in good agreement with the observed magnetic properties. Acknowledgements The authors would like to thank Dr. T. Woodcock for help with the EBSD analysis. Financial support from the Federal Ministry of Education and Research (BMBF) via the PerEMot project (No. 03X4621A) is gratefully acknowledged.
References [1] O. Gutfleisch, M.A. Willard, E. Bru¨ck, C.H. Chen, S.G. Sankar, J.P. Liu, Adv. Mater. 23 (2011) 821. [2] S. Sugimoto, J. Phys. D: Appl. Phys. 44 (2011) 064001. [3] W.F. Li, T. Ohkubo, K. Hono, M. Sagawa, J. Magn. Magn. Mater. 321 (2009) 1100. [4] K. Kobayashi, K. Urushibata, Y. Une, M. Sagawa, in: Proc. 22nd Int. Workshop on Rare-Earth Permanent Magnets and Their Applications, Nagasaki, Japan, 2012, O2–1615. [5] J.F. Herbst, Rev. Mod. Phys. 63 (1991) 819. [6] V. Zepf, A. Reller, Waste Resour. Manag. 165 (2012) 3. [7] D. Bauer, D. Diamond, J. Li, M. McKittrick, D. Sandalow, P. Telleen, Critical Materials Strategy, Tech. Rep. US DOE, 2011. [8] K.T. Park, K. Hiraga, M. Sagawa, in: Proc. 16th Int. Workshop on Rare-Earth Magnets and Their Applications, Sendai, 2000, p. 257.
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
[9] T.G. Woodcock, Y. Zhang, G. Hrkac, G. Ciuta, N.M. Dempsey, T. Schrefl, et al., Scr. Mater. 67 (2012) 536. [10] N.M. Dempsey, T.G. Woodcock, H. Sepehri-Amin, Y. Zhang, H. Kennedy, D. Givord, et al., Acta Mater. 61 (2013) 4920. [11] Z. Samardzˇija, P. McGuiness, M. Soderzˇnik, S. Kobe, M. Sagawa, Mat. Charact. 67 (2012) 27. [12] M. Komuro, Y. Satsu, H. Suzuki, IEEE Trans. Magn. 46 (2010) 3831. [13] H. Nakamura, K. Hirota, T. Ohashi, T. Minowa, J. Phys. D: Appl. Phys. 44 (2011) 064003. [14] N. Oono, M. Sagawa, R. Kasada, H. Matsui, A. Kimura, J. Magn. Magn. Mater. 323 (2011) 297. [15] N. Watanabe, M. Itakura, M. Nishida, J. Alloys Compd. 557 (2013) 1. [16] H. Sepehri-Amin, J. Liu, T. Ohkubo, K. Hioki, A. Hattori, K. Hono, Scr. Mater. 69 (2013) 647.
255
[17] T.G. Woodcock, Q.M. Ramasse, G. Hrkac, T. Shoji, M. Yano, A. Kato, O. Gutfleisch, Acta Mater 77 (2014) 111. [18] H. Sepehri-Amin, T. Ohkubo, K. Hono, Acta Mater. 61 (2013) 1982. [19] S.R. Trout, IEEE Trans. Magn. 24 (1988) 2108. [20] G. Schneider, E.T. Henig, G. Petzow, H.H. Stadelmaier, Z. Metallkunde 77 (1986) 755. [21] H. Mehrer, Diffusion in Solids: Fundamentals, Methods, Materials, Diffusion-Controlled Processes, Springer, Berlin, 2007. [22] B.A. Cook, J.L. Harringa, F.C. Laabs, K.W. Dennis, A.M. Russel, J. Magn. Magn. Mater. 233 (2001) L136. [23] D. Suess, T. Schrefl, J. Fidler, IEEE Trans. Magn. 36 (2000) 3282.
ScienceDirect Acta Materialia 83 (2015) 248–255 www.elsevier.com/locate/actamat
Temperature-dependent Dy diffusion processes in Nd–Fe–B permanent magnets ⇑
K. Lo¨ewe,a, C. Brombacher,b M. Katterb and O. Gutfleischa,c a
Material Science, TU Darmstadt, Alarich-Weiss-Str. 16, 64287 Darmstadt, Germany b Vacuumschmelze GmbH & Co. KG, 63412 Hanau, Germany c Fraunhofer IWKS Project Group for Materials Cycles and Resource Strategy, 63450 Hanau, Germany Received 2 June 2014; revised 19 September 2014; accepted 21 September 2014
Abstract—Nd–Fe–B permanent magnets have been coated with 0.6 wt.% dysprosium and annealed at various temperatures to study the impact of the temperature-dependent Dy diffusion processes on both the magnetic properties and the microstructure. When optimum annealing conditions are applied the Dy processed magnets with initial coercivity of 1100 kA m1 yield coercivity increases which can exceed 400 kA m1 without a significant reduction of the remanent magnetic polarization. The improved stability against opposing magnetic fields can be observed up to a depth of 3 mm along the diffusion direction, restricting the application of the Dy diffusion process to either thin magnets or magnets with tailored coercivity gradients. While in the proximity of the Dy-coated surface, each grain has a Dy-enriched shell with a Dy content of 6 at.%; the Dy concentration decreases exponentially to 1.8 at.% after a diffusion depth of 400 lm and to 1 at.% after a diffusion depth of 1500 lm, as was found with wavelength dispersive X-ray spectroscopy and scanning transmission electron microscopy–energy dispersive X-ray spectroscopy, respectively. In the vicinity of the Dy-coated surface, the mechanism of the Dy-shell formation is attributed to the melting/solidification of a heavy-rare-earth-rich intermediate phase during high-temperature annealing. This is based on the observation that a constant Dy concentration over the width of the shells was found. Also an epitaxial relation between the Dy-poor core and the Dy-rich shell was observed by electron backscattered diffraction, which is supported by results obtained with Kerr microscopy. Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Nd–Fe–B; Permanent magnet; Coercivity enhancement; Grain boundary diffusion process; Microstructural characterization
1. Introduction Nd–Fe–B permanent magnets exhibit excellent magnetic properties and are widely used in industry as essential components of energy applications such as motors, generators, transformers and actuators [1,2]. Depending on the grain size, in commercial Dy-free sintered magnets a coercive field l0Hc of 1–1.5 MA m1 at room temperature can be realized, which is sufficient for most room-temperature applications [3,4]. For high-temperature applications, however, a higher coercivity is needed. This is commonly achieved through the addition of the heavy-rare-earth (HRE) dysprosium into the starting alloy, which is incorporated into the Nd–Fe–B lattice in place of the Nd atoms and leads to an enhancement of the anisotropy field Ha. As a drawback, the saturation magnetization MS as well as the energy product (BH)max is decreasing [5]. Other drawbacks are the high cost of Dy and the associated supply risk due to the monopolistic market situation and geological distribution [6,7]. To overcome these problems, the so-called grain boundary diffusion process (GBDP) was proposed by Park et al. [8], where the Dy is first deposited on to the magnet´s surface
⇑ Corresponding author; e-mail: [email protected]
and then diffused into the magnet during a post-sintering annealing treatment. The idea is to concentrate the precious element in the vicinity of the grain boundaries where nucleation and propagation of magnetization reversal is initiated [1,9,10]. As a result, a significantly reduced amount of Dy is required, which in turn allows the fabrication of high-coercivity magnets with a minimized loss of saturation magnetization. Studies in sintered magnets with other HRE compounds as the diffusion medium to increase coercivity, such as Tb [11], Dy and Tb fluorides [12,13] and eutectic compositions like DyNiAl [14], found that the diffusion occurs mainly over the grain boundaries and leads to a microstructure with grains exhibiting a HRE-rich (Nd,RE)2Fe14B shell and a HRE-poor core. Also the HRE diffusion in hot deformed Nd–Fe–B magnets has received a lot of attention in the last years [15–17]. The precise determination of the HRE element distribution in Nd–Fe–B sintered magnets after a GBDP is crucial for future optimization and has been addressed in recent publications. Samardzˇija et al. used Tb as the HRE and found a linearly decreasing Tb concentration in NdFeB up to a diffusion length of 200 lm [11]. Sepehri-Amin et al. have investigated the Dy concentration in Nd–Fe–B after a GBDP using three-dimensional atom probe (3DAP) tomography. They investigated a spot 2.7 mm away from the surface
http://dx.doi.org/10.1016/j.actamat.2014.09.039 1359-6462/Ó 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
and found a significantly enhanced Dy concentration near the grain boundaries [18]. Our study confirms the rapid decrease of the HRE content within a HRE-rich shell upon increasing the distance to the HRE coated surfaces. Despite these research efforts, some fundamental phenomena have not been addressed yet. Neither an accurate determination of the temperature-dependent HRE diffusion coefficients along the grain boundary phase, nor a direct correlation between the coercivity and HRE gradients occurring in diffusion-processed Nd–Fe–B magnets has been established. It is the aim of the present work to evaluate the influence of the annealing temperature on the magnetic properties and microstructure of Nd–Fe–B sintered magnets after a Dy diffusion process and elucidate possible diffusion mechanisms. 2. Experimental Commercial sintered Nd–Fe–B magnets with a Dy concentration of 0.3 at.%, a size of 10 mm 10 mm 3 mm and a typical remanent polarization of 1.42 T were used as the sample material. A slurry of Dy powder in alcohol was applied either to one lateral surface with the dimensions of 10 mm 3 mm or to both pole surfaces with the dimensions of 10 mm 10 mm. The coated samples were individually wrapped in molybdenum foil and annealed at elevated temperatures Ta ranging from 600 °C 6 Ta 6 1050 °C for 6 h followed by low-temperature annealing at 500 °C for 2 h. All demagnetization curves were measured at a temperature of 20 °C using a commercial hysteresis graph from MAGNET-PHYSIK Dr. Steingroever GmbH. While the coercivity HcJ was determined directly from the demagnetization curves, the remanent magnetic polarization Br was calculated according to
249
600 °C 6 Ta 6 1050 °C are shown in Fig. 1 in comparison with uncoated reference samples which have been subjected to the same annealing treatment and typical properties of samples in the initial state. It can be clearly seen that the diffusion treatment leads to an increase in coercivity with a pronounced maximum at 900 °C (Fig. 1a), while the remanent polarization of the samples is not significantly affected by the diffusion process (Fig. 1b). For Ta = 900 °C, a coercivity of HcJ = 1510 kA m–1 was determined which corresponds to an increase in coercivity DHcJ of 420 kA m–1 when HcJ is compared with samples in the initial state. A comparison with uncoated reference samples reveals that 70 kA m–1 of this coercivity increase can be attributed to an improved annealing treatment and only a DHcJ of 350 kA m–1 is induced by the Dy diffusion process. Note that the reference sample annealed at Ta = 1050 °C is not showing a reduced coercivity with respect to samples in their initial state, indicating that no significant grain growth occurs during the annealing treatment. Thus, the reduction of DHcJ at annealing temperatures higher than 900 °C is very likely an influence of accelerated bulk diffusion, which limits the formation of Dy-enriched (Nd,Dy)2Fe14B grain boundaries. The demagnetization curves of Dy-coated samples annealed at Ta = 700, 800, 900 and 1000 °C are shown in Fig. 2 in comparison with a typical demagnetization curve of a sample in the initial state prior to the diffusion treatment. The increase in coercivity induced by the grain boundary diffusion process can clearly be observed. The squareness of the demagnetization curve is not substantially altered by Dy diffusion, which indicates a rather
Br ¼ Jr0 þ Jr0 N ðl 1Þ where l is the relative permeability, N the demagnetization factor and Jr’ the open circuit remanent polarization measured in a set of Helmholtz coils [19]. The dependence of the switching field on the distance to the coated surface was measured in open circuit conditions with a custom-built scanning Hall probe setup after applying appropriate reverse fields at room temperature. For the microstructure investigation a Zeiss Leo Gemini 1530 field emission gun (FEG) equipped with a solid-state detector energy dispersive X-ray spectroscopy (EDX) detector and an electron backscatter diffraction (EBSD) system from HKL was used. In addition, wavelength dispersive X-ray spectroscopy (WDX) line scans were obtained using a Zeiss ULTRAplus 55 FEG scanning electron microscope (SEM) with an INCAWave WD spectrometer and transmission electron microscopy (TEM) measurements were done using a JEOL JEM 2100F. For the preparation of TEM lamellae a FEI Helios 600i focused ion beam setup was used. Domain observations were carried out using Kerr microscopy (evico magnetics GmbH) and polar magneto-optic contrast.
3. Results 3.1. Magnetic properties The magnetic properties of samples coated on both pole surfaces with dysprosium powder and annealed at
Fig. 1. (a) Room-temperature coercivity HcJ of bulk samples after diffusion treatment for 6 h at 600 °C 6 Ta 6 1050 °C and (b) corresponding remanent polarization Br. The coercivity and remanence of annealed reference samples without Dy coating and the average HcJ and Br in the initial state prior to the annealing treatment are indicated for comparison.
250
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
Fig. 2. Demagnetization curves of Dy-coated samples after grain boundary diffusion processes at 700 °C 6 Ta 6 1000 °C for 6 h and of a reference sample in the initial state.
homogeneous coercivity throughout the complete sample volume. Consequently, the diffusion length of Dy has to be significantly larger than 1.5 mm, which is particularly remarkable for an annealing temperature of 700 °C and suggests the presence of an adequate amount of liquid phase, even in the low-temperature regime. In the results described above, the Dy was applied to both 10 10 mm large pole faces and consequently the Dy diffused along the thickness of 3 mm of the sample. In order to investigate the diffusion depth in more detail another set of samples was prepared where the Dy was applied to one of the lateral 10 3 mm large faces of the magnets. The local magnetization reversal of these samples after a diffusion treatment at 700 °C 6 Ta 6 1000 °C was measured with a scanning Hall probe setup. The switching field HS with respect to the distance d to the coated surface could then be determined in open circuit conditions. For this purpose the magnets were exposed to a reverse field comparable to the coercivity of the magnet. The regions of the magnet which have a lower coercivity than the applied reverse field were demagnetized where the region close to the surface where the Dy was applied remains magnetized in the initial direction. The distance of the transition line between these two regions is then determined by a Hall probe. By increasing the strength of the reverse field and repeating the Hall probe scans, the switching field was measured as a function of the distance from the surface. Fig. 3 shows the dependence of DHS (d) = HS (d) – HS (6 mm) for samples after diffusion treatment at different Ta and for a sample in the initial state. For the sample in the initial state a small reduction of DHS was observed in the vicinity of the lateral surface. This reduction is not related to actual material properties, but can very likely be attributed to edge effects as the measurements have been performed in open circuit condition. After the diffusion treatment, all samples show an increase in DHS in the proximity of the coated surface. The largest increase in HS was observed for Ta = 900 °C. For this diffusion temperature, the switching field was increased by 340 kA m–1 compared with the measurements done on the sample in the initial state and a penetration depth of 3–4 mm could be determined. These values correspond well with the previous measurements after Ta = 900 °C, where an increase in coercivity of 350 kA m–1 could be attributed to the diffusion treatment. Temperatures lower or higher than 900 °C lead to a reduction of DHS. Interestingly, the penetration depth
Fig. 3. Dependence of the switching field variation DHS on the distance d to the Dy coated lateral surface of samples after a grain boundary diffusion process at 700 °C 6 Ta 6 1000 °C for 6 h and of one sample in the initial state without diffusion process. Note that for better visibility, the error bars of Dx = ± 300 lm (spatial resolution of the Hall probe) and DHs = ± 40 kA m1 are not indicated.
is also showing a maximum for Ta = 900 °C. Higher annealing temperatures do not lead to larger penetration depths, and the switching field distribution after Ta = 800 °C is very similar to the switching field distribution after Ta = 1000 °C. This peculiar behavior might be explained by an increased bulk diffusion coefficient of Dy at 1000 °C compared to 900 °C. The resulting faster diffusion into the grains might lead to a faster depletion of Dy and therefore to a reduced penetration depth. 3.2. Microstructural investigation Fig. 4 shows SEM images of the near-surface microstructure of Nd–Fe–B permanent magnets after Dy diffusion at various temperatures for 6 h using backscattered electron (BSE) contrast. Compared to the microstructure of typical Nd–Fe–B sintered magnets which consist of (Nd,Dy)2Fe14B grains surrounded by nanometer-thick rare-earth-rich grain boundaries [9], the Dy diffusion processed samples exhibit distinct differences. First, the grains in the vicinity of the coated surface show an enhanced BSE signal and appear brighter, which indicates a higher average atomic weight as compared with grains further apart from the Dy-coated surface. These grains exist up to a distance of 10 lm at 700 °C to more than 50 lm at 1000 °C. Secondly, adjacent to this zone a pronounced core–shell structure of the (Nd,Dy)2Fe14B grains can be observed, as was mentioned in Section 1. The thickness of the shells decreases from 1 lm with increasing distance and is only visible up to a depth of 500 lm for annealing temperatures of 900–1000 °C. Due to the high amount of Dy in the outermost surface grains, the remanent magnetic polarization has to be substantially reduced compared to their initial state prior to the diffusion process. For practical applications one therefore needs to consider the avoidance or removal of such surface layers to retain a high efficiency of Dy usage and a high total remanence. In our study, the surface layer has a thickness of 150 lm after applying an optimum annealing temperature of 900 °C. Consequently, the impact of the removal of such surface layers on the production costs has to be considered when weighing the economic benefits of the reduced usage of Dy through the GBDP.
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
251
Fig. 4. BSE SEM images of near surface transverse sections of Dy-coated samples after annealing for 6 h at 700–1000 °C.
EDX maps of Dy, Nd and Fe (Fig. 5) reveal that the increase in average atomic weight of the (Nd,Dy)2Fe14B grains and (Nd,Dy)2Fe14B shells is induced by the replacement of Nd by Dy. While the Fe concentration in the (Nd,Dy)2Fe14B grains remains independent of the distance to the coated surface (the slight variation of the Fe contrast is most likely a measurement artefact due to the overlap of the Fe Ka peak with the Dy La peak), a clear reduction of the Dy concentration with increasing distance to the coated surface can be observed. Besides the formation of Dy-enriched (Nd,Dy)2Fe14B grains and shells, the diffusion process has also an influence on the rare-earth-rich grain boundary phases. At least two different Dy/Nd ratios can be distinguished in the EDX maps. One grain boundary phase is both Dy- and Nd-rich (Fig. 5, squares) while the other phase is only Nd-rich (Fig. 5, circles). It seems reasonable that the Dy-poor grain boundary phases consist of Nd-oxide phases which have been present prior to the diffusion process and remained solid at 900 °C. Thereby, the diffusion coefficient of Dy within the oxide phase is much smaller as compared to the residual rare-earth-rich grain boundary phase, which becomes liquid at 670 °C [20]. The impact of the various Nd-oxide phases on the actual diffusion process could not be determined and needs to be considered in future research activities. To study the mechanisms involved in the formation of the Dy-enriched shells, WDX line scans (Fig. 6) and EBSD mapping (Fig. 7) have been performed. The WDX line scan shows that both the Dy and Nd content remain constant within a shell. Thus, the formation by a conventional bulk diffusion process seems unlikely and the results support a model in which the surface region of the (Nd,Dy)2Fe14B grains is partially melted at elevated temperatures, as was proposed recently by Sepehri-Amin et al. [18]. Upon cooling, the Dy-enriched (Nd,Dy)2Fe14B phase solidifies and crystallizes on the matrix grains to form the Dy-rich shells. EBSD has been used to study possible orientation differences between the shells and the matrix grains in a sample
annealed at 900 °C for 6 h (Fig. 7). The differences in lattice parameters are small and therefore neglected (c-axis: ˚ , Dy2Fe14B = 11.83 A ˚ [5]). As a result, Nd2Fe14B = 12.03 A all of the values for the intragranular average misorientation angle are smaller than 1° (which is about the limit of the angular accuracy of EBSD) i.e. all the grains are perfect single crystals. This means that the shells grow epitaxially on the surface of the grains. Fig. 8 shows a region with a pronounced core–shell structure in BSE contrast and the Kerr image of the very same region. The nominal easy axis of the material is outof-plane. The Kerr microscope is used with polar contrast and thereby the sensitivity direction is parallel to the easy axis of magnetization and black and white areas can be referred to domains with magnetization along the easy axis. The pattern shows so-called star-like domains and is characteristic for uniaxial materials in this direction of observation. Due to the presence of a paramagnetic grain boundary phase and consequently a decoupling of the grains, intergrain magnetostatic interactions are expected to be dominant. However, in a few cases we found the magnetic domains to extend over the grain boundaries into neighboring grains, which might be attributed to exchange interactions between the grains. In by far the most cases there was no visible difference between shells and grains in the Kerr images (Fig. 8b). A correspondence of BSE and Kerr contrast with the core–shell structure can only be detected in exceptional cases (e.g. top part of Fig. 8c). As the majority of domains run uninterrupted through the core–shell interface a relatively undisturbed transition layer seems likely which corresponds well with the EBSD analysis in Fig. 7, where no orientation difference between shells and grains could be found. 3.3. Determination of the Dy gradient To quantitatively study the Dy gradients established by the diffusion process, WDX and TEM investigations were performed on a sample annealed at 900 °C for 6 h. Fig. 9
252
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
Fig. 5. BSE SEM image of a near surface transverse section of a Dy-coated sample after annealing at 900 °C for 6 h and corresponding EDX maps of Dy, Nd and Fe. Two examples of Nd-rich grain boundary phase are indicated by circles; two examples of both a Nd- and Dy-rich grain boundary phase are indicated by squares.
Fig. 6. (a) BSE SEM image of the transverse section of a Dy-coated magnet after annealing at 900 °C for 6 h and (b) corresponding WDX line scan showing the Dy and Nd concentration. The analyzed area was at a distance of 110 lm to the coated surface.
shows the results of WDX analysis performed within the shell regions at various distances d to the Dy-coated surface. The Dy content rapidly decreases from 6 at.% to values below 2 at.% at distances d larger than 400 lm. The Nd concentration shows the opposite behavior leading to a constant sum of the Nd and Dy concentrations of 12 at.%, which corresponds to the (Nd,Dy)2Fe14B phase. The exponential decrease of the Dy concentration within the shells can be used to determine the diffusion coefficient of Dy in the liquid grain boundary phase. The Dy profile was fitted to the Grube solution to Fick’s second law for a pair of semi-infinite solids [21]: x Cðx; tÞ ¼ csurf ðcsurf cbulk Þ erf pffiffiffiffiffiffiffiffi 4Dt
Therein csurf is the surface concentration, cbulk the concentration in the bulk, x the diffusion length, t the diffusion time and D the diffusion constant. We assumed that the diffusion occurs solely over a rare-earth-rich liquid eutectic phase along the grain boundaries of the otherwise solid grains. Furthermore, the diffusion coefficients are independent of concentration and there are no convective effects within the liquid. From the fit, the diffusion coefficient of Dy within Nd–Fe–B permanent magnets at 900 °C was determined to be DDy(900 °C) = (2.5 ± 0.5) 109 cm2 s–1. Note that although the absolute values of concentration might include errors due to measurement artifacts, only the slope of the curve is governing D, which improves the reliability of the obtained results.
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
253
Fig. 7. (a) BSE SEM image of the transverse section of a Dy-coated magnet after annealing at 900 °C for 6 h and (b) EBSD orientation map of the very same region. The shells show no detectable orientation difference to the inner parts of the grains. The sample shown in this figure was annealed at 900 °C for 6 h. The map was recorded close to the sample surface.
Fig. 8. Microstructure of a Dy coated magnet after annealing at 900 °C for 6 h. (a) Correlation between magneto-optical Kerr (top) and SEM BSE (below) contrast. The respective images show the exact same region of the sample. The easy axis of the material is pointing out-of-plane and the Kerr image is showing polar contrast. (b) and (c) show magnified views of (a).
There have been previous attempts to determine D of Dy in the system Nd–Fe–B. Cook et al. [22] assembled pairs of cast and annealed single-phase compositions of R2(Fe0.714Co0.286)14B where R = Nd or Dy. After a heat treatment at 1050 °C they measured the concentration gradient of Dy with EDX. Despite the higher annealing temperature, they found a substantially smaller diffusion coefficient DDy(1050 °C) of 2.9 1012 cm2 s–1. We believe that the huge difference to our value originates from the different amounts of liquid phase. As Cook et al. have used a Dy source which remains solid during the annealing treatment, the obtained average Diffusion coefficient is strongly
influenced by bulk diffusion. In our experimental setup, the Dy concentration is clearly dominated by diffusion along the liquid phase and consequently a diffusion coefficient which is three orders of magnitude higher is not unreasonable. As the thickness of the shells rapidly decreases with increasing diffusion length and the lateral resolution of the WDX measurements is only slightly below 1 lm, the accuracy of the determined Dy/Nd concentration for d > 200 lm is limited. To overcome the limitations of WDX, scanning transmission electron microscopy (STEM) EDX has been performed at a distance d of 1.5 mm (Fig. 10).
254
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
of 1.5 mm, the increase in Dy content at a distance of 10 nm to the actual grain boundary is 0.4 at.%. From sintered magnets, it is known that this increase in Dy leads to an increase in coercivity of 200 kA m–1 at room temperature. Thus, the increase of the Dy content approximately matches the observed increase in the switching field (Fig. 3). 4. Conclusion
Fig. 9. Depth profiles of the Nd and Dy concentration after Dy diffusion treatment at 900 °C for 6 h. Each data point was obtained from a WDX measurement in a Dy-enriched shell.
The analysis shows that the peak Dy content at the grain boundary is 1 at.%. The value must be regarded as a rough estimation due to the principal errors of the EDX method when evaluating low-energy peaks such as Dy Ma. Nevertheless, it is significantly higher than the average Dy concentration prior to the diffusion process. Hence, this measurement clearly demonstrates that Dy can easily diffuse more than 1.5 mm into a magnet body, which is in good agreement with the magnetic characterization of the switching field distribution revealing a penetration depth in the order of 3–4 mm. Focusing on the Dy gradient along the line scan, a penetration depth dP of 20–30 nm into the (Nd,Dy)2Fe14B grain can be determined. Compared with the WDX measurements of the Dy content within the shells, no plateau of the Dy content can be observed, indicating a different formation mechanism possibly involving bulk diffusion. Using the diffusion formula above a bulk diffusion coefficient of DDy(900 °C) 1 1015 cm2 s1 can be estimated. It should be noted that dP is subjected to substantial errors due to the thickness of the TEM lamella and unknown orientation of the grain boundary with respect to the beam, as well as the unknown amount of liquid phase during the annealing treatment. Nevertheless, as micromagnetic model calculations have revealed that the anisotropy of a 7 nm thick region around individual Nd–Fe–B grains has significant impact on the coercivity of the bulk magnet [23], a Dy enrichment up to a penetration depth of 20 nm to 30 nm can adequately explain the coercivity increase induced by the diffusion process. Finally the observed increase of the Dy content can be compared with the switching field distribution. At a depth
Fig. 10. (a) High-resolution TEM image of a grain boundary located 1.5 mm apart from the Dy-coated surface of a magnet annealed at 900 °C for 6 h and (b) corresponding EDX line scan.
The dependence of both microstructure and magnetic properties on the annealing temperature of a Dy diffusion process has been analyzed. An optimum annealing temperature of 900 °C could be determined for an annealing time of 6 h and the coercivity increase induced by the diffusion process was 350 kA m1. The switching field distribution revealed that an impact of the Dy diffusion process can be observed up to a distance d in the range of 3–4 mm to the coated surface. However, the benefits of the diffusion process rapidly decrease for d > 1.5 mm. Coating sintered magnets on all sides can partially compensate for this size restriction, but for many practical applications the maximum thickness of the diffusion processed samples is limited to 5 mm. The microstructure analysis shows that Dy-enriched grains and several lm thick Dy-enriched shells are present in the vicinity of the coated surface, leading to a reduction of the local remanent magnetic polarization. Although the impact on the integral remanence is small due to the restricted thickness of the deteriorated surface layer, the removal of this layer (e.g. by grinding) has to be considered to restore the full remanence. The mechanism of the Dy enrichment in the vicinity of the surface was reasoned to be melting of a Dy-rich intermediate phase and subsequent solidification. For larger distances to the coated surface the thickness of the Dy-enriched shells is reduced to several nanometers. At a distance d = 1.5 mm to the coated surface, the thickness and the Dy content were analyzed by STEM EDX and both values are in good agreement with the observed magnetic properties. Acknowledgements The authors would like to thank Dr. T. Woodcock for help with the EBSD analysis. Financial support from the Federal Ministry of Education and Research (BMBF) via the PerEMot project (No. 03X4621A) is gratefully acknowledged.
References [1] O. Gutfleisch, M.A. Willard, E. Bru¨ck, C.H. Chen, S.G. Sankar, J.P. Liu, Adv. Mater. 23 (2011) 821. [2] S. Sugimoto, J. Phys. D: Appl. Phys. 44 (2011) 064001. [3] W.F. Li, T. Ohkubo, K. Hono, M. Sagawa, J. Magn. Magn. Mater. 321 (2009) 1100. [4] K. Kobayashi, K. Urushibata, Y. Une, M. Sagawa, in: Proc. 22nd Int. Workshop on Rare-Earth Permanent Magnets and Their Applications, Nagasaki, Japan, 2012, O2–1615. [5] J.F. Herbst, Rev. Mod. Phys. 63 (1991) 819. [6] V. Zepf, A. Reller, Waste Resour. Manag. 165 (2012) 3. [7] D. Bauer, D. Diamond, J. Li, M. McKittrick, D. Sandalow, P. Telleen, Critical Materials Strategy, Tech. Rep. US DOE, 2011. [8] K.T. Park, K. Hiraga, M. Sagawa, in: Proc. 16th Int. Workshop on Rare-Earth Magnets and Their Applications, Sendai, 2000, p. 257.
K. Lo¨ewe et al. / Acta Materialia 83 (2015) 248–255
[9] T.G. Woodcock, Y. Zhang, G. Hrkac, G. Ciuta, N.M. Dempsey, T. Schrefl, et al., Scr. Mater. 67 (2012) 536. [10] N.M. Dempsey, T.G. Woodcock, H. Sepehri-Amin, Y. Zhang, H. Kennedy, D. Givord, et al., Acta Mater. 61 (2013) 4920. [11] Z. Samardzˇija, P. McGuiness, M. Soderzˇnik, S. Kobe, M. Sagawa, Mat. Charact. 67 (2012) 27. [12] M. Komuro, Y. Satsu, H. Suzuki, IEEE Trans. Magn. 46 (2010) 3831. [13] H. Nakamura, K. Hirota, T. Ohashi, T. Minowa, J. Phys. D: Appl. Phys. 44 (2011) 064003. [14] N. Oono, M. Sagawa, R. Kasada, H. Matsui, A. Kimura, J. Magn. Magn. Mater. 323 (2011) 297. [15] N. Watanabe, M. Itakura, M. Nishida, J. Alloys Compd. 557 (2013) 1. [16] H. Sepehri-Amin, J. Liu, T. Ohkubo, K. Hioki, A. Hattori, K. Hono, Scr. Mater. 69 (2013) 647.
255
[17] T.G. Woodcock, Q.M. Ramasse, G. Hrkac, T. Shoji, M. Yano, A. Kato, O. Gutfleisch, Acta Mater 77 (2014) 111. [18] H. Sepehri-Amin, T. Ohkubo, K. Hono, Acta Mater. 61 (2013) 1982. [19] S.R. Trout, IEEE Trans. Magn. 24 (1988) 2108. [20] G. Schneider, E.T. Henig, G. Petzow, H.H. Stadelmaier, Z. Metallkunde 77 (1986) 755. [21] H. Mehrer, Diffusion in Solids: Fundamentals, Methods, Materials, Diffusion-Controlled Processes, Springer, Berlin, 2007. [22] B.A. Cook, J.L. Harringa, F.C. Laabs, K.W. Dennis, A.M. Russel, J. Magn. Magn. Mater. 233 (2001) L136. [23] D. Suess, T. Schrefl, J. Fidler, IEEE Trans. Magn. 36 (2000) 3282.