Journal of Magnetism and Magnetic Materials 215}216 (2000) 253}259
Invited paper
Ferromagnetic nanocomposites F. Mazaleyrat *, L.K. Varga LESIR-URA-CNRS-CNAM-ENSC, 61, av. Du Pdt Wilson 94235 Cachan, Cedex, France Research Institute for Solid State Physics and Optics, HAS, P.O.B. 49, 1525 Budapest, Hungary
Abstract A survey of magnetic nanocomposites applicable in high-frequency signal and power electronics is given. First, the preparation and properties of ribbon and powder cores from the nanocrystalline `bulka alloys (Finemet and Nanoperm) is reviewed. A technology is presented to apply continuously a large stress during the annealing and winding of the rapidly quenched ribbons in order to induce uniaxial anisotropy in it. The obtained toroidal cores with #at hysteresis curve are applicable up to 1 MHz with considerable permeability (&250). The powder cores prepared from ground Finemet with powder size of 30}400 lm are applicable up to 1 MHz and in some cases up to 10 MHz for smaller powder sizes with low permeability (&10). Finally, the most common methods used for the preparation of metallic nanoparticles are presented. Presently, the compacts prepared from nano-size (40}80 nm) iron powders do not show the expected behavior. It is anticipated that the iron-based ferromagnetic nanocomposites should replace partly the ferrite-type materials in the forthcoming years. 2000 Elsevier Science B.V. All rights reserved. Keywords: Magnetic composites; Nanoparticles; High-frequency applications
0. Introduction In the last decade, much work has been done with the aim of increasing the operating frequency in electronic devices. The tendency is to reduce the volume (or weight) of systems using inductive elements. This is observed in all types of electronics such as radio-transmission, telecommunications, micro devices, power electronics and so on. The demand for reduction of size is obvious in the micro-systems and portable devices like telephones. In contrast, the motivation for this is less simple in the case of power converters for domestic and industrial apparatus in which the weight is not so important. Actually, in these kinds of applications the price is the most important property of the products. Let us consider, for example, the plug-in AC/DC converters, which are
* Corresponding author. Tel.: #33-147-402-108; fax: #33147-402-199. E-mail address:
[email protected] (F. Mazaleyrat).
spread over the world as much as billions. This kind of converter supplies about 10}50 W for a weight of 200}400 g and a volume of 100}200 cm. The widespread procedure applies a 50}60 Hz transformer lowering the network voltage down to about 10 V, and then recti"es it. Taking into account the reduction in price of semiconductor-based units while the price of magnetic materials is stable, it is more e$cient to rectify directly the network voltage, and then to transform it after chopping at high frequency. By using this method, a 12 W supplier can be made as small as a normal plug. In the high-frequency range, ferrite is commonly used. Beside the low price, they present the advantage of high resistivity and the possibility of making cores of various forms. However, they exhibit a negative thermal coe$cient of resistivity and have relatively low Curie temperature. Moreover, the high-permeability MnZn ferrites have a high relative permittivity (up to 10) so the wave velocity in the medium is low and a dimensional resonance is observed at about 1 MHz when the core dimensions are within several mm [1]. Due to their low magneto-crystalline anisotropy they are limited in frequency by the ferromagnetic resonance (FMR), the socalled Snoek limit [2].
0304-8853/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 8 8 5 3 ( 0 0 ) 0 0 1 2 8 - 1
254
F. Mazaleyrat, L.K. Varga / Journal of Magnetism and Magnetic Materials 215}216 (2000) 253}259
nanocrystalline alloys have been used to replace the classical permalloy [7]. In the MHz frequency range, besides the nc metallic thin layers, metal oxide thin "lms, such as Fe}(Co) }M}O (M"Hf, Zr, Ti, V, Nb, Y, Cr, W, B, Al, Ce, Nd, Dy, etc.) [6,8}10] have been developed consisting of nanosized ferromagnetic particles embedded in an oxide matrix, which exhibit quality factors as high as 15 up to 100 MHz. The aim of this paper is to review the various soft magnetic materials adaptable to high-frequency power systems with improved volume/power e$ciency. Powder cores prepared from multi-domain nc powders (particle size between 20 and 1000 lm) and from mono-domain nano-powders applicable in the radio-frequency range will be presented.
1. Preparation of nanocrystalline alloy-based ribbon and powder cores
Fig. 1. Permeability and quality factor spectra of nanocrystalline and ferrite materials.
In the high-frequency range, specially, for large induction levels insulated metallic powder (permalloy, carbonyl iron, etc.) cores are also used. The air gap induced by the insulating oxide layer reduces the permeability due to localized stray "elds. As an example, the Nanocon material, which is made from 10 lm particles Sendust metal (composition close to Fe Si Al wt%) with a 18 nm Al O insulating layer exhibits a relative permeability as high as 168 up to 10 MHz [3]. The "lling factor can be as high as 99%. The "rst nanocrystalline (nc) magnetic material invented in 1988 by Yoshizawa et al. from Hitachi Metals Laboratory, has opened new possibilities [4]. Although the Finemet's decaying permeability overlaps (Fig. 1) the ones of MnZn and NiZn ferrite in the high-frequency range (which is useful for common mode chokes), the quality factor of Finemet in high-frequency region is much inferior compared to that of ferrite. Actually, these new materials are very e$cient in the audio-frequency range (up to 100 kHz) only. On the other hand, one bene"t is that due to its metallic conductivity, the permittivity is low and the dimensional resonance is never observed even at high frequency. Interesting solutions have been found in order to produce micro-magnetic devices using single-layer amorphous or nanocrystalline ribbons [5] or thin "lms [6]. The micromagnetic devices works below 1 W. Recently,
Most of the nc alloys are prepared by partial devitri"cation of an amorphous precursor. The so-called Finemet with nominal composition Fe Si B V \V Nb Cu (with usually x"13.5 or 16.5 at%) consists of Fe}Si BCC crystallites, with 10}14 nm average diameter, embedded in the residual amorphous matrix [4]. The nanocrystallization process is based on the overlapping action of nucleating (especially Cu) and grain-growth inhibiting element (Nb, with larger atomic diameter than iron). Note that, although a number of nucleating and inhibiting elements with di!erent concentrations have been tested (over 150 di!erent alloys are presented in Hitachi's patent [11]), the best results in term of permeability, core loss and thermal stability is obtained with Cu and Nb additions [12] so far. Besides the Fe}Si}B based Finemet, the second family of nc alloys are based on Fe}ETM}B elements, usually labeled as Nanoperm (ETM"Zr, Hf, Nb of 5}7 at% and B of 2}6 at%), where Cu addition is not absolutely necessary in order to produce "ne and homogeneous grain structure [13,14]. Nevertheless, the properties are improved by 1% copper addition. The best soft magnetic properties are obtained with alloys having the nominal composition of Fe Nb Zr B Cu [15]. The frequency dependence of the complex permeability of this kind of materials is fully described by eddy currents. The higher the permeability, the smaller the skin depth is. In practice, for the high-permeability nc alloys the permeability starts decreasing above 10}100 kHz, being applicable up to 300}500 kHz only. In many applications of telecommunications, a constant permeability is required in a large frequency range and in power electronics the linearity of inductance is expected as a function of current and frequency. As a consequence, one must strongly reduce the permeability without withdrawing the bene"t of the low coercivity of this kind of
F. Mazaleyrat, L.K. Varga / Journal of Magnetism and Magnetic Materials 215}216 (2000) 253}259
materials. This can be accomplished by transverse induced anisotropy in ribbons and by demagnetizing e!ect in powdered nc alloys. 1.1. Stress annealed Finemet wound cores Transverse magnetic "eld induces only several 10 J/m while stress annealing induces more than 1000 J/m [16]. According to Herzer [17], the large creep-induced anisotropy is located within the grains in magneto-elastic form due to the tensile counter-stress exerted by the amorphous phase. As shown by Hofmann and KronmuK ller [18,19], this three orders of magnitude decrease in permeability are followed by only one order of magnitude increase in coercivity. Generally, such a stress annealing is made on several 10 cm long ribbons impeding their introduction in applications. An alternative method derived from Perron's group [20] is proposed in Refs. [21] in order to produce directly wound cores by continuous passing-by stress annealing through a furnace. The speed of the winding motor controls the annealing time, while the stress is controlled from the torque of a brake at the as-cast edge of the ribbon. The magnetic properties obtained by this method are similar to that of classical stress annealing if the temperature is properly chosen (usually 1003C higher). It is interesting to note that, stress annealing of Nanoperm-type Fe Zr B Cu alloy results in longitu dinal anisotropy revealed by the Z-type loop. This result is not well understood at this time, but suggests that, since the crystalline phase is composed of pure iron, this stress-induced anisotropy is not related with atom pair ordering in the nanocrystals [22]. 1.2. Nanocrystalline powder cores Owing to the brittleness of nc Finemet alloys, powders down to 20 lm can easily be produced by grinding the ribbon samples. Since the amorphous phase is strongly stabilized by the nanocrystallization, the nanostructure is not appreciably changed. The #ake-shaped particles are molded in resin [23}25] or solder glass [26] under pressure with eventual "eld orientation of the #akes. The magnetic properties are related to the air gap distribution and can be controlled by both the #ake size and the compacting pressure with very good reproducibility. The largest permeability (6 0 0 0) is obtained after hot pressing of large #akes (1 mm) with 5% solder glass, but this results in a low cut-o! frequency (10 kHz). In contrast, the smallest permeability (ranging from 7 to 10) is obtained after cold pressing of 20 lm #akes with 50% resin having the highest cut-o! frequency (100 MHz). Finemet is probably the best alloy for this purpose because Nanoperm has a lower resistivity and has a lower stability against grinding than Finemet. Alterna-
255
tively, Hitperm (FeCo) Zr B Cu should be used in high-temperature applications [27]. Some brittle and metastable intermetallic compounds can be prepared by planar #ow casting and then commuting the ribbon into powder, preserving its nanostructure obtained from rapid quenching. For example, silicon-rich iron ribbons can be crushed into micro-sized particles with nanocrystalline structure. Special interest is devoted to Fe Si, which exhibits a high resistivity of 220 l) cm, due to its highly disordered B2 metastable nanostructure (grain size+50 nm). The magnetic polarization is relatively low, 0.6 T, but su$cient in highfrequency applications [28]. Up to now, the properties of as-cast ribbons are very promising (see next section) but much more work has to be done in order to determine the in#uence of annealing and milling on the microstructural, magnetic and transport properties.
2. Preparation of nanoparticle-based composite cores A good review of techniques used for metallic nanoparticles production is given by Kruis et al. [29]. In this paper, we are not mentioning those techniques that give dispersed particles with superparamagnetic behavior (chemical route for example) or thin "lms (sputtering). 2.1. Inert-gas condensation The inert-gas condensation is one of the most largely used preparation techniques for nanosize particles [30]. The material is evaporated under low-pressure inert gas such as pure helium. The vapor condenses into nanometer size particles that are collected on the surface of a liquid nitrogen-cooled rotating cylinder. The size of particles becomes smaller when the density of the gas is smaller and when the melting point of the metal is higher. The nearly spherical particles are collected by scraping the surface into a funnel and "nally in a compacting assembly, the latter being in the same vacuum cell. The compaction is usually conducted in two stages, the second being at about 1 GPa. The characteristic shape is a mm-thick disc with a maximum diameter of 15 mm. The "nal relative density range between 70% and 97%, depending on the metal's ductility. The particle size ranges from 2 to 100 nm. Since this technique avoids pollution, it is suitable for fundamental studies on physical properties. According to the random anisotropy model (RAM), the e!ective anisotropy is reduced if the grain size is smaller than the anisotropy exchange length, (A/K , where K is the anisotropy constant and A is the exchange constant [31]. As shown by LoK %er [32], this diminution in K and consequently in the coercive "eld is observed for particle size lower than 17 nm instead of the theoretical 30 nm. The minimal coercive "eld for the smaller particles is still 2 orders of magnitude higher
256
F. Mazaleyrat, L.K. Varga / Journal of Magnetism and Magnetic Materials 215}216 (2000) 253}259
than the expected one. For the failing of RAM, one can recall surface anisotropy, demagnetizing e!ect due to residual porosity and magneto-elastic e!ect [33]. An elaborate version of this technique is used for mass production. 50}150 nm iron and nickel powder are available at a price of 6000 $/kg (for comparison, the price of gold is 8000 $/kg), while 19}38 nm hematite and magnetite cost 240 and 1300 $/kg respectively [34]. 2.2. High-energy milling Many metastable alloys can be obtained by mechanical alloying with nanosized crystallite-structure, the powder particle size remaining above the micrometer level. In contrast, ball milling of oxides such as magnetite or hematite leads to nano-particles. From these, pure iron nano-particles are obtained by hydrogen reduction. 2.3. Cryogenic melting When a molten metal is in contact with a cryogenic liquid, a gaseous layer forms at the surface. In this, so-called calefaction layer, the metallic vapor condenses into ultra-"ne particles the size of them depending essentially on the vapor pressure. The gas produced from the cryogenic liquid (argon or nitrogen) transports the particles into a canvas "lter. In order to have su$cient vapor pressure, the metal must be heated up by several hundred degrees over its melting temperature (over 20003C for Fe, Ni, Co). Consequently, cryogenic melting of ferromagnetic metals cannot be conducted in a ceramic crucible, which reacts under these conditions. The levitation
melting method uses a RF inductor that induces eddy currents in the metal and both levitate and heat the metal [35]. A picture of the reactor is seen in Fig. 2 (left). In practice, a rod of metal is slid down into the reactor. A drop of molten metal forms at the edge and falls onto the inductors where it is levitated to complete transformation into nanoparticles. The production is continuous at a rate of about 150 g/h. The as-obtained iron nanopowders have spheroidal shape with average diameter of 32 nm (78% of particles is less than 20 nm) in optimal case, their size is perfectly spherical (see Fig. 2, right hand). After passivation by approximately 1 nm oxide layer [36], the particles are no more pyrophoric and can be manipulated in air. The 177 Am/kg measured magnetization of 30 nm pure iron particle shows a 20% reduction compared to the bulk magnetization, which is in line with the 14% oxide fraction estimated from MoK ssbauer spectrometry. As is shown by Beke [37], the magnetic moment is the same as the bulk value. Similar results are obtained for FeCo35% nanoparticles for which a 5% reduction in magnetization was observed [38].
3. High-frequency magnetic properties of nanocomposites 3.1. Theoretical frequency limits for granular materials The susceptibility of an assembly of particles with non-magnetic inter-grain spacer is usually calculated assuming a three-dimensional matrix of cube shaped
Fig. 2. Cryogenic levitation melting experimental apparatus (left hand), TEM micrograph of the as-obtained nanoparticles (Courtesy of J. Bigot and Y. Champion, CECM, CNRS)
F. Mazaleyrat, L.K. Varga / Journal of Magnetism and Magnetic Materials 215}216 (2000) 253}259
257
particles with intrinsic susceptibility s and regular gap g: G sd s G s" G " . d#s g 1#s (p\ !1) G G This simpli"cation known as non-magnetic grain boundary model (NMGB) is acceptable for spherical particles with diameter d and packing fraction p between 30 to 60%. If the particles are smaller than the exchange length and are exchange coupled, the e!ective anisotropy, 1K2, is reduced according to RAM and the susceptibility becomes J 1 s" 3k 1K2
Kd with 1K2"p , A
where J and A are the saturation polarization and exchange constant, respectively. In a composite based on ferromagnetic conductive particles, the particles must be insulated from each other and their size must be lower than the skin depth at the working frequency. Nevertheless, the frequency limitation in ferromagnetic materials is not only due to eddy currents but also to various phenomena that can appear in small-particles assembly. Below, we enumerate the di!erent sources of the frequency limitations for ferromagnetic composites. The frequency limit due to eddy currents is connected to the skin depth: 4o f " , #! pk (s#1)d where o is the resistivity. For large-resistivity materials or for small particles, another limitation can be due to domain wall resonance [39]:
f " 5
2Sd lJ 1 , 3p(1#s) 2pk
where d is the domain wall thickness (4 (A/K) ), and J is the saturation polarization. The volume fraction of the walls is given by Sd "d (n#1)/d, where n is the number of domain walls in a particle of diameter d. If the material is magnetized at low "eld, the FMR resonance appears at lK f " , % J 1 where l"3.517;10 Hz m/A is the gyromagnetic ratio. The ferromagnetic resonance frequency depends on the anisotropy constant. Eventually, shape and surface anisotropy must be taken into account, the latter being probably relevant for nanoparticles but at the time being di$cult to evaluate.
Fig. 3. Computed frequency limitations for an iron particlebased powder core. The fraction of perfectly insulated particles is 80%.
Finally, we mention the dimensional resonance, which is not important in the case of metals but for materials having a large permittivity such as ferrite. Total re#ection of the magnetic wave occurs when one dimension of the sample, dimension is half the characteristic wavelength. In Fig. 3, di!erent types of frequency limitations are compared for a composite consisting of non-magnetic, insulating matrix "lled with 80% Fe particles. Below d"3 lm, the eddy currents are no longer the "rst limitation but the FMR. The highest frequency by permeability product is reached with 10 nm exchange coupled monodomain particles. In this example, the dimensional resonance is computed assuming a 1 mm thick compacted sample. The computation shows that the dimensional resonance appears at lower frequency than FMR for particle size below 3 nm. This is due to the predicted value of the susceptibility by RAM, which reaches unrealistic high values. Presumably, the FRM is the ultimate limit in the case of nanoparticles. Similar computation for MnZn ferrite shows that the frequency limitation is due to the FMR or the wall resonance for particles smaller or larger than 10 lm, respectively. 3.2. Experimental results on nanocrystalline alloys-based ribbon and powder cores The complex permeability spectra are measured with an EG&G 5210 lock-in ampli"er below 10 kHz, HP4394A impedance analyzer between 10 kHz and 500 MHz, and HP4291B material analyzer above 500 MHz. The permeability of both stress annealed and composite Finemet are shown in Fig. 4. As expected, the band pass of the material is depending on the static initial permeability. The results are comparable with those obtained on carbonyl iron, permalloy powder core or NiZn
258
F. Mazaleyrat, L.K. Varga / Journal of Magnetism and Magnetic Materials 215}216 (2000) 253}259
Fig. 4. Frequency behavior of Finemet powder cores with 50}60% packing fraction compared with that of a ribbon core obtained by continuous stress annealing at 100 MPa.
ferrite for the same permeability and from this point of view (low "eld properties), the stress annealed Finemet has no relevant advantage. In contrast, at higher induction (some 10 mT or more), the Finemet composite exhibits one order of magnitude lower losses in a broad band due to its higher resistivity and lower coercivity than any other material. It is remarkable that both stress-annealed and powdered Finemet has the same maximum high value of 3.3 GHz in the kf product. The nanocrystalline melt-spun Fe Si has already very interesting HF properties due to its high resistivity even in the form of as-quenched ribbons. An appreciable permeability of 250 is kept constant up to 10 MHz (Fig. 5). Furthermore, work must be done in order to remove internal stresses by annealing without decomposing its metastable structure into Fe Si#Fe Si stable phases. Applications up to 100 MHz are thought to be possible in the form of powder cores. Up to now, insulated Fe nanoparticles have been obtained by partial devitri"cation of sputtered Fe}M}O glasses. The nc Fe Hf O alloy is composed of 10 nm Fe particles embedded in an Hf and O rich amorphous matrix. Interestingly, the high permeability indicates that the exchange transmission is possible through an amorphous oxide spacer (see Fig. 5 [8}10]). The results in terms of permeability and frequency limits are strikingly in conformity with the theoretical analysis presented above. The composites prepared from nanopowders are only at the beginning of investigations. The nanocomposite made from passivated iron nanoparticles by dry compaction and sintering does not show the high-frequency behavior as predicted above. The low cut-o! frequency is probably related to the eddy currents due to bad insulation between particles (Fig. 6). One explanation is related to the possibility of segregation of iron oxide during
Fig. 5. Permeability spectra of rapidly quenched nc Fe Si rib bon [28] and nc Fe-Hf-O thin layer [8}10].
Fig. 6. Complex permeability spectra (real and imaginary parts) of compacts made from oxidized nanoparticles. Fe particles are obtained by cryogenic melting and Fe}Co particles are made by plasma torch synthesis [39].
preparation, resulting in electrical percolation. Similar results are found with oxide-coated FeCo particles [39] in which the resistivity seems to be increased by only one order of magnitude.
4. Conclusions The stress-annealed Finemet is very interesting because of its possibility to store energy due to its transverse anisotropy (and not in air gaps) but has the drawback of a narrow range of permeability and a very complicated production. The Finemet powder cores proved to be suitable for high-frequency applications with properties comparable to those of NiZn ferrites. At high induction or polarization, they compete with MnZn ferrite-gapped cores but with much lower losses. Their frequency dependences are
F. Mazaleyrat, L.K. Varga / Journal of Magnetism and Magnetic Materials 215}216 (2000) 253}259
comparable to that of permalloy or carbonyl iron powder cores but shows smaller losses. Questions arise on the necessary conditions for exchange softening in nanocomposites prepared from nanosized powders, because the coercivity remains high. It should be mentioned that the ferromagnetic nanoparticles have been studied mainly in their superparamagnetic state and not in the case of high-density packed and interacting particles. In particular, the role of the intergrain matrix is not clear. The insulation of nanoparticles remains a di$cult challenge because high packing fraction can be obtained only by sintering which results in grain growth and oxide cluster formation. Furthermore, investigations in pressing technologies must be done in order to produce high-density nanocomposites with good soft magnetic properties at elevated frequencies. Many low-anisotropy nanoparticles made from iron alloys (alloyed with Co, Ni, Al, Mo, Si, etc.) have not yet been studied.
Acknowledgements This work was partially supported by NATO Science for piece no. 97130 grant. One of the authors, L.K. Varga greatly thanks the Ecole Normale Superieure de Cachan for a two-month hospitality as an invited professor.
References [1] E.C. Snellsing, in: Soft Ferrite, Properties and Applications, Butterworth, London, 1988, pp. 33}37. [2] S. Chikazumi, in: Physics of Magnetism, Krieger, Malabar, Florida, 1978, p. 330. [3] Y. Sugaya, O. Inoue, K. Kuimiya, IEEE Trans. Magn. 31 (1995) 2197. [4] Y. Yoshizawa, S. Oguma, K. Yamaauchi, J. Appl. Phys. 64 (1988) 6044. [5] O. Dezuari, S.E. Gilbert, E. Belloy, M.A.M. Gijs, Sensors Actuators A 71 (1998) 198. [6] Y. Sasaki, S. Morita, T. Hatanai, A. Makino, T. Sato, K. Yamasawa, NanoStruct. Mater. 8 (1997) 1025. [7] N. Nakota, T. Shima, H. Fujimori, J. Appl. Phys. 70 (1991) 6238. [8] Y. Hayakawa, A. Makino, H. Fujimori, A. Inoue, J. Appl. Phys. 81 (1997) 3747. [9] T. Morikawa, M. Suzuki, Y. Taga, Jpn J. Appl. Phys. 38 (1999) 718.
259
[10] H. Fujimori, S. Mitani, T. Ikeda, S. Ohnuma, IEEE Trans. Magn. 30 (1994) 4779. [11] European Patent, no. 0271657, 1988. [12] Y. Yoshizawa, K. Yamaguchi, Mater. Sci. Eng. A 133 (1991) 176. [13] K. Suzuki, N. Kataoka, A. Inoue, A. Makino, T. Masumoto, Mater. Trans. JIM 31 (1990) 743. [14] A. Makino, T. Hatanai, A. Inoue, T. Masumoto, Mater. Sci. Eng. A226}228 (1997) 594. [15] A. Makino, T. Bitoh, A. Inoue, T. Masumoto, J. Appl. Phys. 81 (1997) 2736. [16] L. Kraus et al., J. Magn. Magn. Mater. 112 (1992) 275. [17] G. Herzer, IEEE Trans. Magn. 30 (1994) 4800. [18] B. Hofman, H. KronmuK ller, J. Magn. Magn. Mater. 152 (1996) 91. [19] B. Hofmann, Thesis, Institute for Physics, Stuttgart, 1994, pp. 51}52. [20] A. Houzali, F. Alves, J.C. Perron, R. BarrueH , Rev. Sci. Instr. 66 (1995) 4671. [21] F. Alves, J.B. Desmoulins, D. HeH risson, J.F. Rialland, F. Costa, J. Magn. Magn. Mater (2000), this volume. [22] B. Hofmann, H. KronmuK ller, Nanostruct. Mater. 6 (1995) 961. [23] V. LeH ger, C. Ramiarinjaona, R. BarrueH , R. Lebourgeois, J. Magn. Magn. Mater. 191 (1999) 169. [24] V. LeH ger, Thesis, ENS de Cachan, 1999 [25] M. MuK ller, A. Novy, M. Brunner, R. Hilzinger, J. Magn. Magn. Mater. 196}197 (1999) 357. [26] D. Nuetzel, R. Reiger, J. Wecker, J. Petzold, M. Mueller, J. Magn. Magn. Mater. 196}197 (1999) 357. [27] M.A. Willard, M.Q. Huang, D.E. Laughlin, M.E. McHenry, J.O. Cross, V.G. Harris, C. Franchetti, J. Appl. Phys 85 (1999) 4421. [28] L.K. Varga, F. Mazaleyrat, J. Kovac, A. Kakay, J. Magn. Magn. Mater (2000), this volume. [29] F.E. Kruis, H. Fissan, A. Peled, J. Aerosol Sci. 29 (1998) 511. [30] R. Birringer, H. Gleiter, H.P. Klein, P. Marquardt, Phys. Lett. 102A (1984) 365. [31] R. Alben, J.J. Becker, M.C. Chi, J. Appl. Phys. 49 (1978) 1653. [32] J.F. LoK %er, W. Wagner, Phys. Rev. B 57 (1998) 2915. [33] C.L. Chien, J. Appl. Phys. 69 (1991) 5267. [34] See for example: www.nanophase.com. [35] J. Bigot, A.G. Goursat, G. Vernet, J.F. Rimbert, J. Foulard, T. Sarle, French Patent no. 8307414, 1983. [36] J.M. Grene`che, private communication. [37] D.L. Beke, Cryst. Res. Technol. 33 (1998) 1039. [38] Z. Turgut, N.T. Nuhfer, H.P. Piehler, M.E. McHenry, J. Appl. Phys. 85 (1999) 4406. [39] G. Bertotti, in: Hysteresis in magnetism, Acad. Pre, San Diego, 1998, pp. 171}179.
