LaCo-substituted ferrite magnets, a new class of high-grade ceramic magnets; intrinsic and microstructural aspects

Journal of Magnetism and Magnetic Materials 242–245 (2002) 1270–1276

Invited paper

LaCo-substituted ferrite magnets, a new class of high-grade ceramic magnets; intrinsic and microstructural aspects c . F. Koolsa,b,*, A. Morela,b, R. Grossinger , J.M. Le Bretond, P. Tenauda,b a

Carbone LorraineFFerrites, 41, rue Pierre Brossolette, B.P. 1642, 27016 Evreux Cedex, France b Carbone LorraineFFerrites, 38830 St-Pierre d’Allevard Cedex, France c Institut fur . Experimentalphysik, Technische Universitat . Wien, Wien A-1040, Austria d GMP, UMR CNRS 6634, Universit!e de Rouen, 76821 Mont-St-Aignan Cedex, France

Abstract The science and technology of conventional ferrite magnets is reviewed, including their historical evolution and models to explain their properties. Next, a survey is given of the new LaCo-type ferrite magnets, representing a breakthrough in magnetic performance. The increased performance is explained with the classical models. Increased market share of M-ferrite magnets as well as re-intensification of background research is expected. r 2002 Elsevier Science B.V. All rights reserved. Keywords: FerriteFhexagonal; Permanent magnet; Substitution effects; Coercivity; Anisotropy field

A Permanent magnet is characterized by its demagnetization curve and this, in turn, by its remanence (Br ), polarization coercivity (HcJ ) and squareness. The magnetic performance of a magnet in a magnetic circuit is a combination of these characteristics, a classical one being the energy product (BHmax ), indicating the maximum field energy which can be transduced to the air gap. The temperature dependence of the demagnetization curve is also important, since the magnet does not always operate close to room temperature. This dependence is characterized essentially by the temperature coefficients (TC) of Br and HcJ Compared to the other main commercial magnet materials today (AlNiCo, NdFeB, SmCo), ferrite magnets have modest magnetic performance and medium TC of Br and HcJ : Nevertheless, they represent in volume by far the highest quantity and in value they share the first place with NdFeB-based magnets. The big economic success of ferrite magnets stems from the favorable performance/price ratio. Their *Corresponding author. Tel.: +33-02-32-31-70-04; fax: +3302-32-31-72-13. E-mail address: [email protected] (F. Kools).

market share would be even larger if their magnetic performance could be improved. This can hardly be expected anymore from the classical technology which has almost attained its saturation. In recent years, however, it has been recognized that LaCo or related substituents may lead to significantly improved ferrite magnets. This has initiated a new generation of highperformance ferrite magnets. In this survey, the material science and technology of the classical ceramic anisotropic ferrite magnets is highlighted first, including models to describe the magnet properties. Subsequently, the new LaCo-type substituted M-ferrites are discussed. The main effects are presented and explained in terms of underlying intrinsic and microstructural factors, referring to our own results.

1. Conventional ceramic anisotropic ferrite magnets Induced by their favorable performance/price ratio, ceramic anisotropic ferrite magnets are located in the center of the permanent magnet market, where requirements with respect to performance and/or allowable magnet volume are serious, but not extreme. The main

0304-8853/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 8 8 5 3 ( 0 1 ) 0 0 9 8 8 - X

F. Kools et al. / Journal of Magnetism and Magnetic Materials 242–245 (2002) 1270–1276

substitution, so that it is uniquely governed by the milled particle size. A variety of grades can be made, where IP is adjusted by the milled particle size and Br =HcJ by the Al/Cr substitution. For technical and economical reasons, however, there is a lower limit for the milled particle size. The corresponding upper performance level of around IP=550 mT characterizes the saturation state of the classical technology in the early 1990s. In this period, several papers appeared, reviewing the science and technology of classical ferrite magnets and its evolution [5–7].

application for high-grade ferrite magnets is as segments in various DC motors for the automotive industry. The manufacturing process of ceramic ferrite magnets is a classical ceramic process, having a few specific features. Four operations play a crucial role:

*

*

prefiring at about 12501C for compound formation, comminution of the prefired material and fine milling into alignable particles of about 1 mm, wet pressing in a magnetic field to align the particles and to shape the product, sintering at about 12501C to realize high density while keeping the grain size around 1 mm.

Most characteristic is the tendency to lateral crystal growth and the aligned pressing operation. Together, they are responsible for the specific features of ferrite magnets and their processing: anisotropic physical properties, platelet-shaped crystals and anisotropic sinter shrinkage. The evolution of the technology has proceeded in three, more or less distinct, steps: (a) basic inventions (1952–1960), (b) significant improvements (1960–1975) and (c) gradual refining of the processing (1975–1995).

The intrinsic properties stem from the M-crystal structure and, notably, from the five distinct Fesublattices. Two intrinsic properties are crucial: saturation magnetization (Js0 ) for Br and anisotropy field strength (Ha ) for HcJ : 1.1.1. Saturation magnetization Js0 and its temperature dependence The five Fe-sublattices are coupled by super-exchange, allowing only parallel (up) or anti-parallel (down) orientation. Their mutual orientation is given by the Gorter model [1,4]: 2a(up), 4f1(down), 12k(up), 4f2(down), 2b(up). Taking into account that the magnetic moment for Fe3+ amounts to 5 mB ; the total moment at 0 K/mol AFe12O19 amounts to 20 mB ; in agreement with the observed saturation magnetization for the pure compound (Js0 ). The temperature depen0 dence of Js0 is shown in Fig. 1, implying Jsð300 KÞ ¼ 0 478 mT. It is remarkable that the Js –T curve is almost linear in a broad T-region (dJs0 =dT ¼ 0:9 mT K1).

2500

700

Js

600

2000

500

K

1500 1000

Ha

300 200

500

-3

400

K (kJ.m ); Js (mT)

(a) Although the mineral ‘‘magneto-plumbite’’ and its structure (M) were known for years, the enormous potentials of the pure AFe12O19 (A=Ba, Sr,y) compound as commercial magnet material have been generally recognized only in the early 1950s. This was induced by its first commercial introduction under the trade name FERROXDURE (FXD) [1]. A few years later followed two basic technological inventions: wet pressing in a magnetic field to make anisotropic magnets [2] and the use of sinteraids to promote densification and, in particular, to suppress grain growth during firing [3]. The latter was induced by the notion that high coercivity requires small grain size (E1 mm). (b) The second period was characterized by less important, but still significant improvements in the procedure, notably an improved composition (SrM), improved sinter-aids (Si-based) and the use of Al or Cr substitutions to adjust the Br =HcJ ratio. The end of the second period coincides more or less with 25th birthday of FXD, where a first review paper was issued [4]. (c) In the third period, it became gradually clear that increased magnetic performance requires decreased milled particle size. This implies increased milling effort and more difficult wet pressing, two factors, which largely contribute to the production costs. This has led to the next definition of the magnetic performance: IP ðmTÞ ¼ Br þ 0:4m0 HcJ : This formula compensates the opposite tendencies of Br and HcJ ; specific for ferrite magnets, when changing the firing temperature or when applying Al/Cr

1.1. Intrinsic properties

-1

*

Ha (kA.m )

*

1271

100 0 0

200

400

600

0 800

T (K) Fig. 1. Temperature dependence of Js ; K and Ha for BaM.

1272

F. Kools et al. / Journal of Magnetism and Magnetic Materials 242–245 (2002) 1270–1276

. Mossbauer analysis revealed that the latter stems from the 12k-sublattice [4]. 1.1.2. Magneto-crystalline anisotropy ðfieldÞ The magnetization is strongly bound to the hexagonal c-axis. The associated energy is characterized by the anisotropy constant (K1 ). The anisotropy field derives from K1 and Js0 : Ha ¼ 2K1 =Js0 : The temperature dependence of K1 is analogous to that of Js0 ; but its increase at decreasing temperature is somewhat less, resulting in a flat Ha –T curve, having a maximum at around 300 K (Fig. 1). There is not yet a clear-cut model for the magneto-crystalline anisotropy. The contribution of dipole–dipole interaction has been calculated to be relatively small. So, the spin–orbit coupling of the Fe3+ ions must play the main role, in spite of the fact that (free) Fe3+ has no orbital momentum. Mostly, the contribution to the overall spin–orbit coupling is associated to the 2b site, but the joint 12k sites also play a significant role [4,8]. 1.2. Modelling Br and HcJ The next formulae refer to medium and high-grade ferrite magnets where crystals are supposed to be single domain in the remanent state and magnetization reversal inside the grains is supposed to be governed by domain nucleation [9]. 1.2.1. Remanence and TCðBr Þ The remanence is described by the next formula [7] Br ¼ kJs0 ¼ f ðd=d 0 Þ sJs0 :

ð1Þ

The influence of the microstructure is reflected by ‘‘k’’ consisting of three factors: the degree of alignment ‘‘f’’ (0:5of o1), the amount of M-phase ‘‘s’’ and the relative density ‘‘d=d 0 ’’. Typical values for (unsubstituted) highperformance magnets (Br ¼ 420 mT) are: f ¼ 0:92; d=d 0 ¼ 0:98; s ¼ 0:98 (k ¼ 0:88). Since k is independent of temperature, the TC(Br ) is governed by dJs0 =dT (0.9 mT K1), resulting in TCðBr Þ ¼ 0:19% for nonsubstituted grades (as measured directly). 1.2.2. Coercivity and TCðHcJ Þ The coercivity is described by the next expression [9] HcJ ¼ a Ha  NðBr þ Js0 Þ=m0 ¼ a Ha  ðN=m0 Þðk þ 1Þ Js0 :

grain shape and increases when the grain shape becomes more platelet shaped. The physical background of Eq. (2) is based on (a) nucleation of domains on grain scale and (b) pinning of the reversal process on macroscopic (product) scale. The first term reflects the field needed to nucleate a reversed domain in a typical grain, while the second term represents the corresponding demagnetization field. For an unsubstituted high-grade magnet (HcJ ¼ 270 kA m1) typical microstructural factors are a ¼ 0:52 and N ¼ 0:7: The TC(HcJ ) is easily derived from Eq. (2): dHcJ =dT ¼ a dHa =dT  ðN=m0 Þðk þ 1Þ dJs0 =dT:

ð3Þ

Substituting the intrinsic factors (unsubstituted Mferrite at 300 K: dHa =dTE0; dJs0 =dT ¼ 0:9 mT K1), and microstructure factors (N ¼ 0:7; k ¼ 0:88) results in TC(HcJ )E+1 kA m1 K1 (as measured directly). 1.3. Substitutions Various substitutions are possible in BaFe12O19 [5]: big divalent cations (Sr, Pby) on the Ba-site and small trivalent cations (Al, Cr, Mny) on the Fe-site. Also differently charged ions are possible, requiring a charge compensation on another site. Substitutions of the Fe ions have a big influence, directly affecting the magnetic infrastructure. Until recently, it appeared to be impossible to improve the magnet performance. The Js0 can be increased, indeed, but only at the cost of Ha and vice versa. The only important Fe-substitution has been Al and/or Cr, having more or less the same effect: decreased Js0 ; while K1 remains essentially constant [5–7]. The corresponding increased Ha is used to raise the magnet HcJ : 1.4. Microstructure The tendency to lateral crystal growth is related to the strongly anisotropic crystal structure, effecting anisotropic surface energy and anisotropic diffusion. This lateral grain growth should be suppressed during high-density sintering, otherwise the HcJ deteriorates. Si-based additions appeared to be effective [4]. As underlying mechanism reaction-induced grain growth inhibition (RIGGI) has been put forward [10]. It is related to the incorporation of the silica and states that, as long as the incorporation reaction is going on, grain growth is inhibited.

ð2Þ

The dependence on the microstructure is described by ‘‘k’’ and, in particular, the factors ‘‘a’’ and ‘‘N’’ (‘‘grain demagnetization factor’’). The factor ‘‘a’’ increases with decreasing grain size, thus reflecting the well-known Hc grain size dependency; the factor ‘‘N’’ is governed by the

2. LaCo-type substituted M-ferrite magnets The most striking effect of LaCo-type substitution is a significantly increased HcJ ; without a drop of Br and, in addition, a significantly decreased TC(HcJ ). Substitution

F. Kools et al. / Journal of Magnetism and Magnetic Materials 242–245 (2002) 1270–1276

classical processing; only the calcining time is longer than usual.

The various practical investigations and their promising results have provoked an intensive research in the precise effects and their backgrounds [17–20]. This will be reviewed below, based on our own studies [20–24]. 2.1. Effects on the magnet properties The main effects are listed below (Figs. 2 and 5): (1a) significant increase of HcJ up to x ¼ 0:25 (25%), (1b) significant decrease of HcJ when x > 0:25 (back to the unsubstituted value for x ¼ 0:4), (2) reduced TC(HcJ ) (50% for x ¼ 0:3), (3) slight increase of Br (3.5% for x ¼ 0:4), (4) decreased squareness of the demagnetization curve (x > 0:2), referring to SrLaCo-M, where effects 1a, 2, 3 are typical [15–19]. In addition, we found; (5) the effects are completely analogous in BaLaCo-M and SrLaCo-M [20]. 2.2. Background of effects *

*

The coercivity is explained on the basis of our coercivity model (Eq. (2)), where an increased Ha value is substituted. Early studies already indicated an increased Ha [16,17]. These have been confirmed and extended by our single point detection (SPD) Ha measurements [20] (Fig. 3). Since the values for Br and a; N (Fig. 4) appear to be quite normal, the increased HcJ is dominated by the increased Ha : The decrease of HcJ for x > 0:25 stems from the more pronounced platelet shape, as suggested earlier [17]. Based on Eq. (2), we could model it in detail [21]. The microstructure parameters a and N were measured both magnetically and by SEM observation. For x > 0:25; the N value increases sharply (Fig. 4), while the

380

-1

480 460

HcJ

340

440 420

Br

300

Br (mT)

(a) In this period, there have been scattered studies, which included occasionally LaCo-M-type compositions, corresponding to the general formula A1xRxFe12xBxO19 (ARB-M), where A=Ba,Sr..., R=La3+,Pr3+,Bi3+,y and B=Co2+,Ni2+,Fe2+, y . Only those studies are mentioned, where Hc or Ha is increased while Js0 or Br remains more or less constant. * 1959 [3]: La O is applied as milling addition to 2 3 increase the magnet Hc : * 1961 [11]: BaLaCo-M: single phase M, at least up to x ¼ 0:5; H c increases significantly with increasing x (o0.4), while Js0 remains constant. Analogous effects for BaBiCo-M are up to x ¼ 0:2: It is assumed that Co replaces Fe at the 4f2 sites. * 1974 [8]: anisotropic sintered SrLaFe2+M: normal K1 and somewhat increased Js0 at 300 K, and strongly increased K1 and slightly reduced Js0 at 0 K. The huge K1ð0 KÞ was explained by a single ion contribution of Fe2+ at 2a sites, referring to the Slonczewski model. 0 The increased Jsð300 KÞ was explained by the changed shape of the Js0 –T curve, which had become more convex. A similar effect was found for the 12k-sublattice magnetization. * 1988 [12]: (isotropic) SrBiCo-M: XRD-pure Mcompound up to xo0:3; slightly decreasing Tc ; modestly increasing Ha and a slightly decreasing Js0 when x increases (xo0:3). * 1989 [13]: BaLaCo-M thin film for magnetooptical application: Ha increases significantly with increasing xðo0:8Þ: * 1996 [14]: La-doped anisotropic SrM magnets: increased magnet performance for xE0:05: (b) The recognition that LaCo-type ferrite magnets, in principle, can represent a new generation of highperformance ferrite magnets has provoked a wave of intensified research, being initially practically oriented in order to obtain patents [15,16]. They share the next general features: * increased performance (IPE600 mT) stemming from increased HcJ ; but Br =HcJ may vary, * reduced TC(H ), up to about 50%, cJ * broad composition range ARB-M including also (Fe+B)/(A+R)a12; optimum properties for SrLaCoM (xE0:25), where Co may be partly substituted by Fe2+,

*

HcJ (kA.m )

of M-type ferrites by LaCo or related combinations (LaFe2+, BiCo,y) is known to have a beneficial effect on HcJ and/or Ha since decades, but only in recent years is it being applied for making improved magnets. So, the historical evolution is characterized by two periods: (a) early occasional studies (1959–1996) and (b) recent intensive studies on improved magnets (after 1996).

1273

400 260 0.0

380 0.1

0.2

0.3

0.4

X Fig. 2. HcJ and Br vs. x for Sr1xLaxFe12xCoxO19 magnets.

F. Kools et al. / Journal of Magnetism and Magnetic Materials 242–245 (2002) 1270–1276

1274

1900

Ha (kA.m -1 )

dHcJ / dT

1

dHa/dT; dHcJ /dT (kA.m -1 .K - )

1.00

1800 1700 1600 1500 1400 0.0

0.1

0.2

0.3

0.4

X

0.50 0.00 0.0

0.1

0.2

0.3

0.4

-0.50 -1.00 -1.50

dHa / dT

-2.00

Fig. 3. Single point detection (SPD) Ha vs. x for Sr1xLaxFe12xCoxO19 magnets.

X Fig. 5. Temperature dependence of HcJ and Ha vs. x for Sr1xLaxFe12xCoxO19 magnets.

N 0.90 0.950

f 0.70

0.930 f

a

Rjt

0.730

0.50 0.0

0.1

0.2

0.3

0.920 0.910

0.4 0.720

X

0.900 0.0

0.1

Fig. 4. Microstructural factors underlying HcJ ; measured magnetically [21].

*

*

0.940

0.740 Rjt

a, N

0.750

a value increases slightly, thus explaining the drop in HcJ : The correlation with the SEM parameters (grain thickness h; grain diameter D) is as expected: h=D and h decrease with increasing x: After transformation of h=D into a calculated demagnetization factor, a good correlation between the calculated and magnetically measured N value was found, thus supporting our HcJ model. The TC(HcJ ) is explained by the same coercivity model in the differential form (Eq. (3)), substituting apart from the experimental a; N values, the measured values for dJ 0s /dT (Econstant) and dHa =dT (Fig. 5). This formula explains that TC(HcJ ) deviates significantly from the usual one: dHa =dT is no longer close to zero as usual, but has become negative, resulting in a decreased TC(HcJ ). The explanation of Br is based on Eq. (1) and our measurements on the involved microstructural para0 meters. The factor Jsð300 KÞ is decomposed into two 0 factors: Jsð0 and a new parameter Rjt ¼ KÞ 0 0 Jsð300 =J characterizing the curvature of the KÞ sð0 KÞ 0 Js0 –T curve. Density and Jsð0 KÞ appear to be constant [20], so the increase of Br is dominated (Fig. 6) by the increase of the alignment factor (f ) and the curva-

0.2

0.3

0.4

X Fig. 6. Remanence factors vs. x for Sr1xLaxFe12xCoxO19 magnets.

*

*

ture factor (Rjt ). The latter was also found for the 12k-sublattice magnetization [23], analogous to LaFe2+M [8]. The increased f factor is due to grain growth, being consistent with the decreased h=D value and the increased N value. The relatively poor squareness had been related to the grain size distribution, which had become bimodal for x > 0:2 [20]. The analogy between Sr and Ba compounds points to Co as the active ion. Apparently, the ion size of La (SrELaoBa) and associated local elastic lattice deformation does not play a role.

2.3. Intrinsic aspects *

Cell constant ‘‘a’’ remains constant, whereas ‘‘c’’ decreases slightly and linearly up to x ¼ 0:4: The latter holds also for Tc : Neutron diffraction detects for x ¼ 0:3; no second phase and a trace for x ¼ 0:4: So, substituted M is single phase for 0oxo0:35; as found earlier [17].

F. Kools et al. / Journal of Magnetism and Magnetic Materials 242–245 (2002) 1270–1276 *

*

The Co sublattice occupation was studied by various techniques [22,23], all being at the limit of detection. Most techniques conclude Co is at the sites 2a and 0 4f2. Since Jsð0 KÞ does not vary with x for xo0:25; an even distribution over 2a(up) and 4f2(down) is likely. . However, Mossbauer and resistivity measurements also suggest the presence of some Fe2+ [23]. This may be related to the fact that the distribution of La and Co is not perfect during calcining, so that Co may be locally replaced by Fe2+. Since a special preference is found for Co at 4f2 [22] and since Co2+ and Fe2+ have practically the same atomic moment, the final conclusion is Fe2+/Co at 2a and Co at 4f2. For Co–Ti substituted M and related compounds, Co prefers other sites and acts negatively on the anisotropy [5]. Apparently, the excess charge of La and its fixed position on the Ba sites forces Co to a special site, where it can contribute to the anisotropy as in the case of LaFe2+M [8]. The increase of SPD Ha with increasing x is due to an increase of the magneto-crystalline anisotropy K1 [20]. In addition, H a increases further when T decreases similar to LaFe2+M [8]. However LaCoM shows an anomaly at To180 K, where Ha has returned to the unsubstituted value (Fig. 7) [20]. This peculiar phenomenon is not observed for LaFe2+M. In spite of this difference, many phenomena are analogous for Co and Fe, so we suppose a similar mechanism: the increased K1 stems from the single ion contribution of Fe2+/Co at the 2a site and possibly also from Co on 4f2.

2.4. Microstructural aspects The calcined grain size decreases with x [21]. This effect is attributed to a RIGGI mechanism: dissolving of LaCo in the M-lattice inhibits grain growth. The RIGGI-effect may be eliminated by high calcining, resulting in increased final Br : Low calcining or adding

3

La2O3 to the mill results in RIGGI during sintering, and, hence, to increased HcJ and decreased Br : This explains contradictory statements on Br =HcJ in the final magnet [15,16]. The bimodal grain distribution is also related to the RIGGI-effect: grains saturated with LaCo can grow again, thus constituting a second grain population. The underlying chemical inhomogeneity can be reduced significantly by applying an improved technology (patent pending). The reinforced tendency to lateral grain growth for x > 0:25 can be suppressed by applying a special milling addition [24]. 2.5. Concluding remarks *

*

LaCo-type substituted ferrite magnets can be described by the classical models for M-ferrite magnets, having the increased magneto-crystalline anisotropy at 300 K and its increased temperature dependence as a characteristic feature. The explanation of the latter is thwarted by (1) the low Co content and (2) the appearance of a quite new feature, the low-temperature anomaly of H a. It is expected that these features will cause a revival of the research on hard ferrites in the near future. The increased HcJ and reduced TC(HcJ ) of LaCotype ferrite magnets enable new applications. These are still small scale for two reasons: somewhat increased costs and relatively low Br =HcJ : The costs are somewhat higher, due to the more expensive raw materials (La,Co). However, the technology is not yet optimized. Optimization will lead to cost reduction, e.g. the increased Ha permits an increased milled particle size, so less processing costs. In addition, the increased HcJ can be translated in principle to an increased Br at the same performance, thus realizing a more balanced Br =HcJ : So, it is expected that the new LaCo-type ferrite magnets will acquire a significant market share in the near future, notably for starter motors and small motors.

References

x = 0.2 µ0Ha (T)

1275

[1] [2] [3] [4]

2

1

[5]

0

[6]

0

200

400

600

800 [7]

T (K) Fig. 7. Low-temperature Fe12xCoxO19 (x ¼ 0:2).

behavior

of

Ha

for

Sr1xLax-

[8]

J. Went, et al., Philips Tech. Rev. 13 (7) (1952) 194. A. Stuijts, et al., Philips Tech. Rev. 16 (7) (1955) 205. A. Stuijts, et al., US patent 2,900,344, 1959. C. vd Broek, A. Stuijts, Philips Tech. Rev. 37 (8) (1977) 197. H. Kojima, in: E. Wohlfarth (Ed.), Ferromagnetic Materials, Vol. 3, North-Holland, Amsterdam, 1982. H. St.ablein, in: E. Wohlfarth (Ed.), Ferromagnetic Materials, Vol. 3, North-Holland, Amsterdam, 1982. F. Kools, in: Kirk-Othmer Encyclopedia of Chemical Technology, Vol. 10, 4th Edition, 1994. F. Lotgering, J. Phys. Chem. Solids 35 (1974) 1633; F. Lotgering, J. Phys. Chem. Solids 41 (1980) 481.

1276

F. Kools et al. / Journal of Magnetism and Magnetic Materials 242–245 (2002) 1270–1276

[9] F. Kools, Proceedings of the MMA Grenoble, J. Phys. 46 (1985) C6–349. [10] F. Kools, Solid State Ionics 16 (1985) 251. [11] G. Smolenskii, A. Andreev, Bull. Acad. Sci. USSR 25 (1961) 1405. [12] G. Turilli, F. Licci, J. Magn. Magn. Mater. 75 (1988) 111. [13] A. Watada, et al., US Patent 4,797,331, 1989. [14] N. Dung, Proceedings of the International Workshop on Materials Science, Hanoi, 1999, p. 357. [15] H. Kubota, et al., Japan Patent Application Hei 10149910. [16] H. Taguchi, et al., European Patent Application EP 0 905 718 A1. [17] K. Iida, et al., J. Magn. Soc. Japan 23 (1999) 1093.

[18] H. Kubota, et al., Proceedings of the Eighth International Conference on Ferrites, ICF8, Kyoto, 2000. [19] H. Taguchi, et al., Proceedings of the Eighth International Conference on Ferrites, ICF8, Kyoto, 2000. [20] F. Kools, et al., Proceedings of the Eighth International Conference on Ferrites, ICF8, Kyoto, 2000. [21] A. Morel, et al., Proceedings of the Eighth International Conference on Ferrites, ICF8, Kyoto, 2000. [22] M. Pieper, et al., Proceedings of the Joint European Magnetic Symposia, JEMS’01, Grenoble, 2001. [23] A. Morel, et al., Proceedings of the Joint European Magnetic Symposia, JEMS’01, Grenoble, 2001. [24] F. Kools, et al., Proceedings of the European Powder Metallurgy Congress, Euro PM2001, Nice, 2001.