GIANT MAGNETORESISTANCE FOR READ HEADS

Transcription

1 Philips J. Res , GIANT MAGNETORESISTANCE FOR READ HEADS MATERIALS by R. COEHOORN, I.C.S. KOOLS*, Th.G.S.M. RIJKS and K.-M.H. LENSSEN Philips Research Laboratories, Prof Holstlaan 4, 5656, AA Eindhoven, The Netherlands Abstract The sensitivity of magnetoresistive read heads can be increased by using layered magnetic materials showing the giant magnetoresistance effect, instead of a single magnetic film showing the anisotropic magnetoresistance effect. For this purpose, exchange-biased spin-valve layered structures are very suitable. For well-chosen compositions and nanometer-scale layer thicknesses these materials combine a fair giant magnetoresistance effect with a very small field interval in which the resistance change takes place. In this paper we give an overview of aspects which determine the functioning of materials of this class in read heads, including their preparation, magnetotransport properties and the magnetic interactions which determine the magnetization reversal process. Keywords: magnetism, giant magnetoresistance, anisotropic magnetoresistance, materials research, magnetic recording, read heads, thin films, exchange biasing, magnetic anisotropy. 1. Introduetion The discovery in 1988 of the giant magnetoresistance (GMR) effect in magnetic multilayers with nanometer-scale layer thicknesses [1,2], and subsequent research leading to materials showing the effect in very small magnetic fields, has opened options for the application of GMR materials in read heads for high density digital hard disk and tape recording. Prototype GMR based read heads for high density hard disk recording have been presented by a number of companies and institutes [3-7]. Philips has demonstrated the application of GMR materials in prototype yoke type GMR based heads for high density digital tape recording [8,9]. *) Present address: CVC Products, 3100 Laurelview Court, Fremont, CA 94538, USA. Philips Journalof Research Vol. 51 No

2 R. Coehoorn et al. In this paper we discuss the deposition, magnetoresistance and magnetic interactions for the type of layered materials which are used most frequently as the magnetoresistive element (MRE) in thin film read heads, vi0."the so-called exchange-biased spin-valve layered structures*. In Sec. 2 their structure is described, and a phenomenological description is given of the dependence of the electrical resistance on the applied magnetic field. In Sec. 3, the origin of giant magnetoresistance and its relation to layer structure, composition and temperature are discussed. In Sec. 4 we discuss growth, microstructure and thermal stability. Factors which determine the switching field interval and the exchange biasing field are discussed in Secs 5 and 6, respectively. In Sec. 7 some concluding remarks are given. The performance of GMR materials in read heads depends not only on the properties of those materials, but also on design aspects. This will be the subject of the paper by Kools et al. in this issue [10]. In that paper it is shown how patterning and the presence of softmagnetic flux guides lead to magnetostatic interactions which modify the magnetization reversal of the GMR element, and hence influence the sensitivity, micromagnetic stability, linearity and dynamic range. For discussions on this issue we also refer to Refs. [3, 4,8, 9,11-17]. Brief introductions to the GMR effectin magnetic multilayers may be found in Refs. [18, 19]. Reviews on exchange-biased spin valves and on their application in yoke-type read heads have been written by Coehoorn [20] and Kools [21]. Rijks [22,23] has presented detailed studies of specific aspects of the anisotropic and giant magnetoresistance of spin-valve layered structures, as well as of the magnetic interactions in such structures. Excellent reviews of the GMR effect in magnetic layered structures, including discussions of other classes of materials, have been written by Levy [24], Dieny [25], Fert and Bruno [26], Parkin [27], and by Gijs and Bauer [28]. 2. Structure and functioning of exchange-biased spin-valve GMR materials 2.1. Structure Exchange-biased spin-valve layered structures were first described in 1990 by Dieny et al. [29-31]. In the original structure (Fig. la), two ferromagnetic (F) layers are separated by a non-magnetic (NM) spaeer layer. A third magnetic layer of a properly chosen composition and structure, which is *) In this paper the term 'layered structure' will be used for the general case of a thin film consisting of a stack of different layers, whereas the term 'multilayer' will be used for the more specific case of a layered structure within which a sequence of layers is periodically repeated. 94 Philip. Journni of Research Vol. SI No

3 Giant magnetoresistance materials for read heads F (pinned) cap layer NM F (free) F (pinned) NM NM F (pinned) F (pinned) NM F (free) AF buffer layer buffer layer buffer layer substrate substrate substrate (a) (b) (c) Fig. 1. Schematic cross-sections of a top (a), bottom (b) and a symmetric (c) exchange-biased spin-valve structure. usually antiferromagnetic (AF), is in contact with one ofthe F layers. This will be called here a 'simple' exchange-biased spin-valve layered structure. Typical layer thicknesses are 3 to 10nm for the magnetic layers, and 2 to 3 nm for the spaeer layer. Suitable preparation conditions, e.g. growth in a magnetic field, lead to pinning of the magnetization of the F layer that is in contact with the AF layer, as a result of the exchange interaction across the interface. This is called the exchange biasing (or sometimes exchange anisotropy) effect. The unpinned ('free') F layer is usually magnetically very soft. The NM spaeer layer is sufficiently thick as to almost eliminate the magnetic coupling between the F layers. In view of the dependence of the microstructure and magnetic properties on the order in which the layers are deposited, one should distinguish 'top' spin valves (AF layer deposited on top of the second F layer, as in Fig. la) and 'bottom' spin valves (AF layer deposited just before depositing the first F layer, see Fig. 1b)* [32].The latter structure is sometimes alternatively referred to as an 'inverted' spin valvet [33]. In general, the F layers may have different compositions, and each of the F *) This terminology was introduced by Anthony et al. [32), referring to the two simple spin-valve structures which together form a symmetric spin-valve structure (Fig. le), t) This terminology was introduced by Lenssen et al. [33), in analogy to the terminology used to describe the layer ordering in Ill-IV semiconductor heterostructures. Philips Journal of Research Vol. 51 No

4 R. Coehoorn et al. layers may itself consist of layers of different compositions. The presence of only a few monolayers of a second ferromagnetic metal at the FfNM interfaces, for example, can sometimes lead to a strong enhancement of the magnetoresistance ratio (see Sec. 3). The combination of a 'top' and a 'bottom' spin valve leads to the more complicated structure of a 'dual' or 'symmetric' exchange-biased spin-valve layered structure (Fig. Ie). Unless stated otherwise, the systems discussed in this paper are simple exchange-biased spin valves Magnetoresistance The functioning of exchange-biased spin valves is shown schematically by the magnetization and resistance curves given in Fig. 2. Defining the positive field direction as the direction of the magnetization of the pinned layer, the layers have parallel magnetizations for H > O. In a small field interval close to H = 0 the magnetization of the free F layer reverses, whereas the magnetization of the pinned F layer remains fixed. Only upon the application of a large negative field (equal to the exchange biasing field Heb), the exchange biasing interaction is overcome, and the pinned layer switches, too. We note that the magnetization curve (Fig. 2a) has been given for the special case of (a) M H (b) R Fig. 2. Schematic curves ofthe magnetic moment (a) and resistance (b) of a simple exchange-biased spin-valve layered structure. Ferromagnetic layers with equal magnetic moments per unit of area have been assumed. H 96 Philips Journalof Research Vol. 51 No. I 1998

5 Giant magnetoresistance materials for read heads equal 'magnetic layer thicknesses' Msat t for the free and pinned layers, where Msat is the saturation magnetization and t is the layer thickness. In general, the magnetic layer thicknesses can be different, leading to a non-zero magnetization in the antiparallel state. The GMR effect is the increase of resistance upon the change of the angle between neighbouring magnetic layers from a parallel to an antiparallel alignment. The observed angular variation of the resistance of exchange-biased spin-valve structures is described well by the phenomenological expression R(O) = R(O = 0) +.6R GMR (l - coso)/2, (1) where 0 is the angle between the magnetization directions of the free and pinned layers [29]. The ratio.6r GMR / R(O = 0) is called the GMR ratio. For combinations of F and NM layer compositions for which the GMR effect is high the resistance curve of exchange-biased spin-valve structures shows a steep slope close to zero field, where the configuration of magnetization directions switches, and stays high until, at the exchange biasing field, also the pinned layer switches (see Fig. 2b) Crossed anisotropies Using exchange-biased spin-valve structures, it has been possible to reach very high sensitivities, combined with a very small coercivity. As shown by Rijks et al. [34], the coercivity is very small for spin valves within which the free layer has an in-plane, uniaxial anisotropy, with the easy axis perpendicular to the unidirectional anisotropy axis of the pinned layer. In such systems with crossed anisotropies, with the external field parallel to the biasing direction, the magnetization of the free layer switches by a coherent rotation of the magnetization. In contrast, in the case of parallel anisotropies magnetization reversal would be the result of domain wall movement, leading to hysteresis in the M(H) and R(H) curves. As a result, strong Barkhausen noise would be superimposed on the sensor output signal due to pinning and dep inning of domain walls. A schematic representation of the magnetization reversal processes for the cases of crossed and parallel anisotropies is given in Fig. 3. Figure 4 shows experimental resistance versus field curves for the prototype layered system. Snm Ni8oFe2o/3nm Cu/6 nm Ni8oFe2o/S nm FesoMnso, where the AF FesoMnso layer leads to the required exchange biasing of one of the permalloy Ni8oFe2o layers. For a discussion on the various methods for realizing a crossed anisotropy configuration we refer to Ref. [34].The sensitivity of this GMR material, defined as the maximum slope (or/ R)/oH of the Philips Journal of Research Vol. SI No

6 R. Coehoorn et al. parallel anisotropies biased E7 E3' E3' E3' unbiased E7 E7 1=;7 E3' crossed anisotropies biased E7 E3' E3' E3' unbiased E7 E7 &?7 E3' H < - Heb - Heb < H < - H. H = 0 H > H. Fig. 3. Schematic representation of the magnetization reversal in the F layers of a spin-valve structure with paral1el and crossed anisotropies. Ha and Heb are the induced anisotropy and exchange biasing fields, respectively. Interlayer magnetic coupling is neglected. relative resistance versus field curve, was found to be up to l8%/(ka/m) (1.4%/Oe) at room temperature. This value is much larger than that for single 30 nm permailoy films, which are presently employed in sensor applications such as read heads. The resistance change of these films, which originates from the anisotropic magnetoresistance (AMR, the dependence of the resistance on the angle between the magnetization direction and the current direction [35]) is about 1.5%. In such AMR films the resistance change takes place in a field range of typicaily 0.56 kalm (7 Oe), leading to a sensitivity of only about 3%/(kA/m). h. (d) \ \ t1 2 % H (kalm) Fig. 4. Low field magnetoresistance of Ni-Fe/Cu/Ni-Fe/Fe-Mn spin valves (layer thicknesses as given in the text), with (a) paral1el anisotropies, (b-d) crossed anisotropies. Crossed anisotropy configurations have been obtained by rotating the applied field during sputter deposition (b), heating the system to 160 C and subsequent cooling in a field, or by the combination of these two procedures (d). Taken from Ref. [34]. 98 Philip. Journal of Research Vol. SI No

7 Giant magnetoresistance materials JOl' read heads 3. Origin of the GMR effect The electrical resistance R of a 'macroscopie' electrical conductor with length e and cross-sectional area A is given by the simple expression e R = A p, (2) where p is the specific resistivity of the material from which the conductor is made. It follows from this expression that a calculation of the resistance of a conductor with 'macroscopie' dimensions does not require information about the shape of the cross-section or about the processes which cause electrical resistivity, viz. the scattering of electrons which have been accelerated by the electric field. In nonmagnetic metals such scattering processes occur due to the interaction with lattice vibrations (phonons), due to the interaction with other electrons, and due to collisions with structural defects such as impurity atoms or grain boundaries. Defect scattering occurs even at T = 0 K, whereas the two former scattering processes occur only at finite temperatures. These 'diffusive' scattering processes bring the out-of-equilibrium distribution of accelerated electrons back to the equilibrium distribution, thereby contributing to the resistance. The average distance between successive diffusive scattering processes is called the electron mean free path, À. If the thickness of a film is of the order of the mean free path, or smaller, then the resistivity of the conductor is in general no longer given by eq. (2). Such a 'mesoscopic' system, with dimensions which are still larger than the atomic dimension but smaller than a characteristic physical length scale (in this case the electron mean free path), may even show entirely new physical properties. An example of such a new property is the GMR effect in magnetic multilayers with a small layer thickness compared with the electron mean free path. Let us consider a simple exchange-biased spin-valve structure, of which the resistance is measured using a current which flows in the plane of the layers. This current-in-plane (CIP) geometry is the most common configuration for measuring the GMR of layered structures, and is the geometry used in read heads. We should distinguish the contributions to the conductivity due to spin-up (1) and spin-down (l) electrons. As we are dealing with a magnetic material, these contributions are in general different. At sufficiently low temperatures, where the spin quantum number is conserved in the vast majority of scattering processes, the conductivity can be treated as the sum of two separate contributions [36]: (3) Philips Journal of Research Vol. SI No

8 R. Coehoorn et al. For the occurrence of the GMR effect it is crucial that: (i) the angles between the magnetization directions of the ferromagnetic layers can be modified by the application of an applied magnetic field, (ii) the electron scattering probability at the interfaces or within the bulk of the layers is spin-dependent, and (iii) in the case of parallel magnetizations, the layer averaged electron meanfree path for (at least) one spin direction is larger than the thickness ofthe non-magnetic spaeer layer. The scattering process is illustrated schematically in Fig. 5, for a Ni-Fe/Cu/ Ni+Fe/Fe -Mn exchange-biased spin-valve. Scattering, within the magnetic layers or at the Ni-Fe/Cu interfaces, is strong for electrons with spin antiparallel to the magnetization direction of the permalloy and weak for electrons with spin parallel. Each layer acts as a spin-selective valve: its magnetization direction determines whether it more easily transmits spin-up or spin-down electrons. In the case of parallel-aligned magnetic layers the contribution to the conductance due to the spin-up electrons is high, leading to a high total conductance. However, antiparallel alignment of the magnetic layers results in appreciable scattering for electrons in both spin directions, and hence in a lower total conductance. The contribution to the conductance of the antiferromagnetic Fe-Mn layer is very small. lts function is only to 'pin' the magnetization of one of the F layers. The non-magnetic Cu layers only serve to decouple the F layers. They should be as thin as is required to fulfill this function. Figure 6 shows that the minimum thickness is about 2 nm. Below that thickness the coupling between the F layers prohibits the realization of a field range within which the two F layers are antiparallel [37]. The optimal thickness of the ferromagnetic NisoFe2o layers is of the order of one half of the mean free path >'T of the spin-up electrons in these layers. If the Parallel spin-up,i ;l spin-down Antiparallel spin-up F...,"11"" ~~ NM : ' F _...,_ ::: ~:_..~ r, i,/' spin-down Fig. 5. Schematic illustration of the electron scattering processes in a (NisoFe2o/CufNisoFe2o/ FesoMnso) exchange biased spin-valve structure. Scattering at the interfaces has been neglected. Scattering at the outer boundaries is considered as entirely diffusive. 100 Philips Journalof Research Vol. SI No. I 1998

9 Giant magnetoresistance materials for read heads magnetic field (knm) 8 6 ~ r 4 ii: <I 2 copper layer thickness (nm) Fig. 6. Dependence on the Cu layer thickness of the MR curves and the MR ratio at room temperature, for (NisoFe2o/Cu/NisoFe2o/FesoMnso) spin valves [37]. Layer thicknesses are given in the text. The dashed line gives a fit to the data for the Cu thicknesses for which full antiparallel alignment is obtained (see text). F layers are much thinner than À r the spin-up conductance in the parallel magnetization state is limited by strong spin-independent scattering in the buffer layer and in the AF layer (or at the interfaces with these layers). The contrast with the spin-down conductance can then be enhanced by increasing the F layer thickness. However, only in the part of the F layer which is within a distance of the order À r /2 from the interface with the space layer is the current density influenced by the magnetization direction of the other layer. The region further away from the interface may therefore be viewed as a non-active region, which reduces the GMR effect due to the fact that it shunts the active region. The factor 0.5 takes, very roughly, into account the fact that the electrons which contribute to the conduction have a distribution of angles with respect to plane of the layers. For a quantitative discussion of the GMR effect in terms of layer-dependent and spin-dependent mean free paths in the bulk of the layers and at the interfaces we refer to Ref. [20] and references therein and to Rijks et al. [22]. In Fig. 7 experimental results for the variation of the GMR ratio with the thickness of the free magnetic layer are shown for F layers consisting of NisoFe2o, Ni66CoISFeI6, and Co, measured at room temperature and at 5 K Philip. Journni of Research Vol. 51 No

10 R. Coehoorn et al. 2 free F-Iayer thickness (nm) Fig. 7. Dependence of the MR ratio on the free magnetic layer thickness, measured at 293K (a) and at 5 K (b), for (X/Cu/X/FesoMnso) spin valves, grown on 3 nm Ta buffer layers on Si(IOO) substracts, with X = Co (squares), X = Ni66Fel6Cois (triangles) and (X = Ni so Fe2o) (+), measured at 5K. Layer thicknesses: Icu=3nm (but 2.5nm for F=(X=Ni so Fe20); IF (pinned layer) = 5 nm (but 6 nm for F = Ni66FeI6CoIS)(from Ref. [22]). [22].For these systems, the optimal F layer thicknesses are 6 to 8 nm, implying that >'T is roughly 10 to 20 nm in these ferromagnetic materials. Requirements for F/NM combinations with a high GMR ratio are: (i) the F and NM materials possess, for one spin direction, very similar electronic structures, whereas for the other spin direction the electronic structures are very dissimilar, (ii) the electron transmission probability (transmission without diffusive scattering) through the F jnm interfaces is large for the type of electrons for which the electronic structures in the F and NM layers are very similar, and (iii) the crystal structures of the F and NM materials match very well. Requirements (i) and (ii) result in a large layer-averaged mean free path, and requirement (iii) results in a very low density of structural defects which limit the electron mean-free path. Cobalt and permalloy both fulfill requirements (i) and (iii) to a very large extent, if combined with copper [20].Electronically, the spin-up band structures match very well. The structural mismatch with Cu is 102 Philips Journal of Research Vol. SI No. I 1998

11 Giant magnetoresistance materials for read heads only about 2% for both ferromagnetic metals. With respect to requirement (ii), permalloy is less favourable than cobalt, because the Ni magnetic moments at the NigoFe2o/Cu interface are depressed as a result of the interaction with the non-magnetic Cu atoms. This is expected to lead to enhanced scattering at the interface for spin-up electrons [38-40]. In contrast, Co moments at a Co/Cu interface are very close to their value in the bulk metal, leading to almost perfect transmission of spin-up electrons through a Co/Cu interface. The important role of the interfaces is nicely demonstrated by the effect of deposition of Co layers with a thickness of a few monolayers at the NigoFe2o/Cu interfaces in NigoFe2o/CujNigoFe2o/Fe5oMn5ostructures [41]. The effect on the MR ratio is shown in Fig. 8. By retaining permalloy in the bulk of the F layers, the magnetic softness of the structure is to a large extent preserved, making this method of 'interface dusting' potentially attractive for sensor applications. 4. Growth and thermal stability 4.1. Deposition process and microstructure Various deposition methods have been used in order to prepare exchangebiased spin-valve layered structures, including de and rf magnetron sputtering, ion beam sputtering and evaporation deposition in an MBE (Molecular Beam Epitaxy) system. In our laboratory we have studied deposition using a URV (background pressure 10-9 Torr) IONTECR de magnetron sputtering system containing six sources. This system, with target diameters of 50 mm and a target-substrate distance of 100 mm, leads to a film thickness homogeneity better than 1% and 4% for substrates with diameters of 1 cm and 2.5 cm, respectively. The typical deposition rate is 0.2 nm/s. As partial water pressures ~ 6 ~ 4 a: ii: <l 2 ' - - _ _ -1R_M"" Ni 80 Fe20.. '.. '.~. t - o t_~ I x [1 - exp(-t/o.23)) Ni80Fe20 o Co layer thickness (nm) Fig. 8. Dependence on the Co interface layer thickness tco of the MR ratio at room temperature for spin valves with the structure (Si/[5.3 - tcol nm NisoFe2o/tco nm Co/3.2 nrn Cu/[2.2 - tcol nrn NisoFe2o/9 nm FesoMnso/1nrn Cu) [411. Philips Journal of Research Vol. SI No

12 R. Coehoorn et al. above 10-9 Torr negatively affect the GMR ratio [42]*, the use of a UHV compatible system and highly purified sputter gasses are a strict requirement. De magnetron sputtering allows for the optimization of the interface roughness and grain structure by a variation of the Ar gas pressure during sputtering, as the gas pressure influences the kinetic energy distribution of the atoms arriving at the substrate. As shown by Kools [43], a variation of the Ar pressure can be used to optimize the GMR ratio. Such optimal materials show NigoFe2o/Cu interface roughnesses of about 0.5 to 0.7 nm (root-mean-square value of interface height with respect to the average plane of the interface) [43,44]. The presence of a buffer layer between the substrate and the film may modify the crystallite orientation and microstructure. As these structural aspects strongly influence the electric and magnetic properties, the use of a proper buffer layer is an important tool for the optimization of the sensorrelevant properties of spin-valve layers. A large number of buffer layer materials have been found to positively affect the exchange biasing effect and the GMR ratio (see, e.g. Ref. [45]). A particularly useful buffer layer material is Ta, which strongly enhances the [Ill] columnar texture if present in the form of a layer of only a few nanometers thickness on substrates such as Si(100), Si02 and Si 3 N 4. This is shown in Fig. 9, which shows results of an X-ray diffraction studied by Duchateau et al. [46] for a permalloy film on Si(100). A 3 nm Ta buffer layer leads to an orientation of [111] axes to within 4 to 5 (full width at half maximum of the X-ray rocking curve) of the film normal. Similar results are obtained for complete spin-valve multilayers. Advantages of a strong [111] texture of NigoFe2o/Cu/NigoFe2o/FesoMnsoand similar spin valves are as follows. (i) The [Ill] textures promotes magnetic softness of the free magnetic layer [47], as the magnetocrystalline anisotropy energy associated with the cubic anisotropy constant K, is independent of the magnetization direction within (Ill) planes. As a result there is no in-plane anisotropy for [111] textured films, other than the uniaxial anisotropy which may be induced by growth in a magnetic field. (ii) For [111] oriented systems the strength of the exchange biasing by FesoMnso is significantly larger, and the coercivity in the exchange biasing is significantly smaller than for other orientations, or for an essentially randomly oriented film. This effect was studied in detail by Jungblut *) l.c.s. Kools (unpublished) has observed that the increased GMR ratio is the result of an increased spin-dependence of scattering (and not just due to switching between more perfectly paral1el and antiparal1el states). 104 Philip. Journalof Research Vol.51 No

13 Giant magnetoresistance materials for read heads,; ~ se 1:: 0.25 ::::J 8 3nmTa 2nmTa 1 nm Ta :i ~.<: Cl) ]0.5.. "Co -"" lea ~ : : I,! Ta layer thickness e (nm) 0.01 nota~\ o (degrees) Fig. 9. Cu - K,,(6-26) X-ray diffraction pattern for Si02/ITa nm Ta/Sû nm NisoFe2olayers, for various thicknesses ITa of the Ta buffer layer [46]. Note the square-root intensity scale. The inset displays the (Ill) peak heights as a function of ITa' et al. [48] using coherently grown systems on top of single crystals in an MBE system. (iii) We find that growth on a thin (2 to 3 nm) Ta buffer layer leads to an increased GMR ratio (typically by 10 to 15%) with respect to sufficiently strong exchange biased systems grown on Si(100) without buffer layer. As the sheet resistance is then typically 20% lower, we attribute this effect to a decreased defect scattering. The shunting effect by 2 to 3 nm Ta buffer layers is almost negligible (approximately 1%). lts resistivity is about Om. Ta mayalso be used as a cap layer. A film thickness of about 2 to 3 nm is sufficient to protect the structure against corrosion from the atmosphere. The role of Ta has been discussed extensively by Lenssen et al. [49], who studied sputter deposited and evaporated Ta buffer layers using Transmission Electron Microscopy (TEM) and diffraction. They concluded that the [111] texture is correlated with the growth of Ta in the form of a metastable nanocrystalline phase of a yet undetermined structure. Figure 10 shows a plan-view TEM micrograph of a sputter deposited 3 nm Ta/50 nm Ni8oFe2o/3nm Ta film, grown on a Si 3 N 4 membrane [22].It reveals an average lateral grain diameter of about 15nm. The grain diameter was found to increase monotonically with the permalloy layer thickness, from values less than 5 nm for layer thicknesses below 5 nm, to a value of about 20 nm for a layer thickness of 125nm [50]. An increase of the grain diameter with the layer thickness is a quite general phenomenon in thin films, and Philip. Journal of Research Vol. SI No. I

14 R. Coehoorn et al. 25nm Fig. 10. Plan-view TEM micrograph of a 3 nm Ta/50 nm Niso Fe2o/3 nm Ta film, grown on a Si 3 N 4 membrane [22]. Fig. IJ. Cross-sectional TEM micrograph of a Si(IOO)/3 nm Ta/8 nm NisoFe2o/2.8 nm Cu/6 nm NisoFe2o/IO nm FesoMnso/S nm Ta spin valve [22]. 106 Philips Journalof Research Vol. SI No. I 1998

15 Giant magnetoresistance materials for read heads was also observed for Cu layers and exchange biased spin valves [50]. The cross-sectional TEM micrograph of an exchange-biased spin valve in Fig. 11 reveals a columnar grain structure, with grains developing throughout the film. The Si substrate and Ta layers can be clearly distinguished, but the NisoFe2o, Cu and FesoMnso layers are indistinguishable due to the small atomic number difference between the elements concerned. As the grain diameter is of the order of the mean free path in Cu and in permalloy (for spin-up electrons), grain boundary scattering is expected to affect the GMR ratio in spin valves, and the AMR ratio in single permalloy films. This is confirmed by Rijks et al. [22,50].He assumed that grain boundary scattering is spin-independent, resulting in a negative effect on the GMR ratio. Growth of inverted spin valves (see Fig. 1b) ofthe type FesoMnso/NisoFe2o/ Cu/NisoFe2o requires the use of a buffer layer on top of which FesoMnso forms the required [Ill] oriented fee structure. As shown by Lenssen et al. [33] this can be accomplished by using a 3.5 nm Taj2 nm NisoFe2o buffer layer. The GMR ratios obtained are slightly (order 10%) higher than for conventional structures (AF layer grown on top). The type of asymmetry of the spin valve (conventional or inverted) that is most suitable for a certain head application, depends on the asymmetry of the position of the element with respect to soft magnetic shields or flux guides, and on the required magnetic flux through the free magnetic layer that is generated by the sense current [9] Thermal stability A number of factors limit the use or processing of exchange-biased spinvalve layered structures at elevated temperatures. First, the exchange biasing field decreases with increasing temperature. For some exchange biasing materials this effect determines the temperature range within which sensors can be applied. The temperature dependence of exchange biasing is discussed in detail in Sec. 6. Secondly, the induced magnetocrystalline anisotropy of the free layer may be affected by heating the material. This is discussed in Sec Thirdly, as multilayers are thermodynamically metastable systems the enhanced diffusion rates during annealing will irreversibly degrade-the layered structure. In the case of a negative reaction free enthalpy of the constituents of the layers involved, interface mixing will occur. This increases the diffusive scattering at the interfaces, and may decrease its spindependence. As Ni and Cu form solid solutions, such an effect may be expected for interfaces between Cu and fee Ni-Fe-Co alloys containing a fair amount of Ni. Indeed, the structure and GMR ratio of NisoFe2o/Cu/ NisoFe2o spin-valves have been observed to be stable only with respect to Philips Journal of Research Vol. SI No. I

16 R. Coehoorn et al. annealing up to Ta = 220 C during 1 hour [44] or 6.5 hours [51]. A significant (typically 30%) decrease of the GMR ratio of F/Cu/F spin-valve structures with F = (NisoFe2o)l-xCox and x = 0 to 0.75, has been reported for annealing during 1 hour at Ta = 250 C, whereas the GMR ratio is less than 40% of the original value after annealing at Ta = 300 C [51]. In contrast, as Co and Fe are almost immiscible with Cu, diffusion across Co/Cu or Co-Fe/Cu interfaces is expected to be much slower than diffusion across Ni-Co/Cu, Ni-Fe/Cu or Ni-Fe-Co/Cu interfaces. Indeed, improved thermal stability has been observed for spin-valve structures with F = C090FeIOand NM = Cu [52]:annealing during 1hour at Ta = 250 C does not affect the GMR ratio, whereas the decrease of the GMR ratio after annealing at Ta = 300 C is less than 10%. Rapid diffusion is found above Ta = 300 C, with flr/ R = 0 after annealing during 1 hour at Ta = 350 C. Finally, diffusion mayalso affect the effective magnetic coupling between the layers. As discussed in Sec. 5.3, there are three contributions to the interlayer coupling, viz. coupling via ferromagnetic bridges between the ferromagnetic layers ('pinholes'), the magnetostatic Néel-type coupling and the interlayer exchange coupling. All three contributions may be affected irreversibly by annealing. Pinhole formation may be increased due to grain boundary diffusion; the importance of this effect depends therefore on the microstructure. The other two interactions may be modified as a result of a change of the interface structure due to interdiffusion. In read heads the (usually) ferromagnetic interlayer coupling is frequently used to balance the effectively antiferromagnetic interaction that results from the stray magnetic field originating from the pinned magnetic layer. The element is then in its most sensitive point of operation (largest slope of the R(H) curve at H = 0). Thermal stability of the interlayer magnetic coupling is important in view of the requirement that the element stays at this optimum. 5. Factors determining the switching field interval The sensitivity of a GMR material (as defined in Sec. 2.3.) is high if a high GMR ratio is combined with a small field interval within which the magnetization switches from parallel to antiparallel. In this section we discuss the factors which determine the switching field interval of exchange-biased spin valves with crossed anisotropies (see Sec. 2.3.). If there is no superimposed AMR effect and no magnetic coupling across the spaeer layer, the magnetization switches in a field interval flh = 2Ha, where Ha is the anisotropy field of the free magnetic layer. However, in general the sensitivity is modified by the superimposed AMR effect and the coupling between the magnetic layers. 108 Philips Journal of Research Vol. SI No

17 Giant magnetoresistance materials for read heads 5.1. Induced magnetic anisotropy The uniaxial in-plane magnetocrystalline anisotropy of the free magnetic layer is induced by growth in a small magnetic field, strong enough to magnetically saturate the film. A field of 15 kalm is sufficient if use is made of permalloy or ternary fee type Ni-Fe-Co alloys. The induced anisotropy is believed to be the result of a very small degree of directional pair order of atoms in the otherwise random solid solution. A systematic study by Rijks et al. [23] of the anisotropy field of alloys with the composition (NisoFe20)1-xCOx,with x = 0 to 0.18, has revealed that the anisotropy field depends on the composition, x, the layer thickness, and the material in between which the layers are sandwiched (Ta or Cu in this study). In Fig. 12 experimental results for films sandwiched in between Ta layers are shown. The decrease of the anisotropy field for film thicknesses below 15 nm is relevant to the case of GMR spin valves, in which the free layer thickness is typically 5 to 8nm. The effect is most likely due to a region with a decreased degree of pair order, situated in the ferromagnetic layer close to the interface with the buffer layer. The result shows that the addition of Co to permalloy is unfavourable as far as the switching field interval is concerned. However, an advantageous effect is that the micromagnetic stability and the field range of linear operation (dynamic range) of GMR sensor elements are improved by the addition of Co [53], resulting in a decreased higher harmonics voltage output of the sensor and improved robustness of the fabrication process. The increase of the GMR ratio with the Co content (see Fig. 7), and an , Ni66Fe16Co18.,.,,---- Ni70Fe18Co ~.. I,.. (~.. Ë ~ 1.0f... Ï ~ Ni7SFe19C06 J{ I.._ ~. Ni so Fe20 0.0L-~L._~.1...,_~-'--~-'--~-'--"_' o tf (nm) Fig. 12. The anisotropy field Ha as a function of the F-layer thickness IF for four different alloys [23]. Film structures are Si(lOO)/3nm Ta/IF nm F/3 nm Ta. The lines represent a fit to the data, using a phenomenological model given in Ref. [23]. Philips Journal of Research Vol. SI No

18 R. Coehoorn et al. increase of the field efficiency of a microstructured sensor element [53] compensate in part for the effect on the sensitivity due to the increased anisotropy field. Annealing the sample can affect the anisotropy field. In practice, it is often necessary to anneal the sample after deposition in a magnetic field that is directed perpendicular to the field direction during growth, in order to rotate the exchange biasing direction over an angle of 90. This leads to a 'crossed anisotropy' configuration (Sec. 2.3.). In such a case, the anisotropy field may decrease (or, in an extreme case, the hard axis may even become the easy axis) as a result of a modification of the directional pair order due to atomic diffusion. For example, annealing of permalloy films with an anisotropy field of 0.5 kalm at 150 C during 10 minutes can lead to a reduction of the anisotropy field by a factor of almost two [23]. This may seem an interesting method for reducing the switching field interval by reducing the induced magnetic anisotropy. However, such a process is relatively illcontrolled and therefore, in view of the negative effect on the robustness of the manufacturing process, undesired. It may even be better to first anneal with the field parallel to the induced anisotropy (leading to a stabilized induced anisotropy [23]), before annealing briefly in a perpendicular field. It is important to stress that it is the magnetization direction, and not directly the applied field direction, which determines the effect of annealing on the induced anisotropy. Strain results in an additional magnetocrystalline anisotropy due to magnetorestriction. As it is difficult to control the strain in sensor elements, the magnetostriction coefficient À should be as low as possible. For ternary fee structure Ni-Co-Fe alloys minima in the absolute value of À are found close to permalloy and for the composition Ni66Co18Fe16.The latter alloy composition has therefore been considered a good candidate for the application in GMR sensor elements [54]. In actual thin films the anisotropy field is not entirely uniform but shows a (small) lateral variation. This results in a so-called ripple domain pattern ([55] and references therein) and a rounding off ofthe M(H) and R(H) curves close to the saturation field, as can be seen, e.g., in Fig. 4b Superimposed AMR effect The R(H) curve is no longer linear at H = 0 if there is a superimposed AMR effect on top of the GMR effect. From a microscopie modelling study Rijks et al. [56] have shown that, to a very good approximation, the total magnetoresistance curve is given by the sum of the curves for the two separate 110 Philip. Journalof Research Vol. SI No. I 1998

19 Giant magnetoresistance materials for read heads effects. If the current direction is parallel to the easy axis of the free F layer (the usual configuration in yoke-type read heads), the magnetoresistance (the resistance change with respect to the high-field parallel configuration) is then given by ~R(H) = ~RGMR(1 - cos 8)/2 + ~RAMR sin 2 8. (4) The AMR-term represents the resistance change due to the change of the angle (1['/2-8) between the magnetization of the free layer and the current direction. Schematic ~R(H) curves are shown in Fig. 13 for different ratios r between the anisotropic and giant magnetoresistance ratios. The presence of the additional AMR effect increases the sensitivity (8R/R)/8H if an operating point between H = 0 and H = +Ha is chosen. For these schematic curves, the largest sensitivity is found at the anisotropy field H = +Ha However, in realistic samples with a rounding off of the R(H) curves near the average anisotropy field, the point of maximal sensitivity is shifted to a lower value. Typical MR ratios obtained for (8 nm NisoFe2o/2.5nm Cu/6 nm NisoFe2o) spin valves are (~R/ R)GMR = 4.5% and (~R/ R)AMR = 1.1% [37]. It follows from Fig. 13 that as a result of the additional AMR effect the sensitivity of these spin valves, as calculated for H = 0.5H a, is almost 50% higher than expected from the GMR effect only. However, it should be remarked that in practical applications it may be preferable to operate the material around the field H = 0, where the element is micromagnetically most stable. In this case, there is no benefit from the AMR effect. -Ha o magnetic field +Ha Fig. 13. Dependence on the ratio r == b..r AMR / b..r OMR of the magnetoresistance curves for an exchange-biased spin valve with crossed anisotropies. Philip. Journalof Research Vol. 51 No

20 R. Coehoorn et al Coupling between the magnetic layers As discussed in Sec. 3 the free and pinned F layers are magnetically coupled if the NM spaeer layer is very thin. Phenomenologically, coupling leads to an offset of the field around which the free layer switches, to a field - Hcoupl' This is shown schematically in Fig. 2 for the case of ferromagnetic coupling. Some results of an experimental study by Kools et al. [57]are shown in Fig. 14. The coupling field is the sum of three contributions: ferromagnetic coupling via pin holes (direct exchange coupling), ferromagnetic Néel-type coupling resulting from positively correlated waviness of the two F/NM interfaces (magnetostatic coupling) and an oscillatory indirect exchange coupling. The strength of the former two effects, and the amplitude of the latter effect decrease with the spaeer thickness. An analysis of experimental results in terms of these contributions is given by Kools et al. [57]. We note that all three contributions are influenced by the microstructure of the material, and hence control ofthe coupling field requires very good control ofthe deposition conditions. This issue is very relevant for sensor applications, F,!:.._ I-II-l 5.0 ~ ::::. g- o 2.5 'Ï AF 2 Cu thickness (nm) 3 Fig. 14. Dependence ofthe interlayer magnetic coupling parameter J (left axis) and the coupling field Hc.upl (right axis) on the Cu layer thickness for spin valves with the structure (Si(IOO)/3.5 nm Ta/8 nm NisoFe2oltcu nm Cu/6 nm NisoFe2018 nm FesoMnso/3.5 nm Ta) [57]. The fullline through the data represents a fit assuming magnetostatic coupling as a result of corrugated interfaces (see inset) with h = 0.4 nm and L = 10 nm. Room temperature measurements. 112 Philips Journal of Research Vol. SI No. I 1998

21 Giant magnetoresistance materials for read heads because in systems of interest the coupling is in some cases about equal or larger than the anisotropy field, leading to zero sensitivity at H = O. For spin valves based on permalloy and Cu, for example, the coupling field is 0.5 to 1.0kAlm at 2.2 nm Cu thickness, whereas the anisotropy field is about 0.4 kalm. Fortunately, in a sensor element the resulting shift of the switching field interval can be balanced by: (i) the application of a bias magnetic field from a current through an integrated bias conductor, and (ii) the magnetostatic coupling between the pinned and the free layer which arises as a result of the stray magnetic field originating from the magnetization of the pinned layer, which is directed perpendicularly to the long axis of the stripe. This interaction tends to couple the two layers antiparallel. The coupling strength is inversely proportional to the width ofthe MRE stripe, and depends on the saturation magnetization and layer thicknesses of the two layers. Folkerts et al. have demonstrated prototype yoke type GMR read heads containing lû-zzm-wide GMR stripes in which this balancing has been realized without the use of a bias current [8,9]. In addition, they have given general rules for designing balanced GMR elements [15]. Interlayer coupling influences the switching field interval due to the following two effects: (i) In the case of coupling the magnetization of the pinned layer is not quite fixed upon the rotation of the free layer. This leads to a decrease of the slope of the R(H) curve and hence to an increase of the switching field interval. The effect increases upon an increase of the ratio between the coupling field and the exchange biasing field, Hcoupd Heb [23]. (ii) Lateral fluctuations in the coupling field increase the switching field interval. Direct evidence for the occurrence of laterally inhomogeneous switching has been obtained from Lorentz transmission electron microscopy. The increase of the switch field interval resulting from this effect increases, for a given composition ofthe free layer material, monotonically with the coupling field [23]. As an example, Fig. 15 shows the switching field interval for spin valves based on two different magnetic alloy compositions. For a Cu thickness of 2 nm, the switch field for both alloys is about double the value expected from the induced anisotropy only (dashed lines). Philip. Journni of Research Vol. SI No

22 R. Coehoorn et al , O~~~~~~~~~-L~ 3 i2 ~1 _.!. _!t. '! _ teu (nm) Fig. l5. Switching field interval 6.H s of the free magnetic layer in Ta/F/Cu/F/FesoMnso/Ta spin valves, as a function of the Cu layer thickness, for two alloys. The dashed lines denote the values of 2H., determined from Ta/F /Cu/Ta films (for which there is no coupling with a pinned F layer). From Ref. [23]. 6. Exchange biasing 6.1. Phenomenological description Within a simple, phenomenological description of the exchange biasing effect, the interaction energy E between the pinned F layer and the antiferromagnetic layer is given by where ()is the angle between the biasing direction and the magnetization of the pinned F layer. For a discussion of microscopie models on the origin of this interaction, we refer to Ref. [48]and references therein. The exchange biasing field varies inversely proportional with the pinned layer thickness: (5) Keb Heb = pp' ll-omsattf (6) in which Miat and t~ are the saturation magnetization and thickness of the pinned ferromagnetic layer, respectively. To obtain exchange biasing, the AF layer thickness should generally exceed a certain critical value. As an example, we show in Fig. 16a the result of a systematic study of the F and AF layer thickness dependence of the exchange biasing field for (111) oriented permalloy, exchange biased by FesoMnso [48]. The minimum AF layer thickness required is about 4 nm at room temperature. The exchange 114 Philips Journalof Research Vol. SI No. I 1998

23 Giant magnetoresistance materials for read heads E ~ l NisoFe2o thickness (nm) temperature (OC) Fig. 16. (a) Contours of constant exchange biasing field Heb (ka/m) for (Ill) oriented NisoFe2o/ Fe50Mn5o, plotted as a function of the thickness of both layers [48). Although growth was performed by MBE on a Cu(lll) single crystal, the results obtained are qualitatively very similar to those obtained by sputter deposition on a Ta buffer layer. (b) Temperature dependence of the exchange biasing field Heb for 6 nm NisoFe2o/1nm Fe50Mn5o films, sputter deposited on Si(IOO)/3nm Ta. biasing field then increases with the AF layer thickness until it is fully developed at a thickness to ~ 7 nm. At and above that AF layer thickness, and for F layer thicknesses larger than about 5 nm, eq. (6) is well obeyed. As shown in Fig. 16b, the exchange biasing field decreases approximately linearly with temperature, down to zero at a temperature which is called the blocking temperature, Tb. The blocking temperature increases slightly with the AF layer thickness, until it saturates at a thickness of about 10 to 12nm. The blocking temperature is significantly lower than the Néel temperature of bulk FesoMns single crystals, which is about 230 C [58]. The exchange biasing effect may be obtained by growth of the AF layer on top of a magnetically saturated F layer, or by cooling the system in a field from a temperature above the blocking temperature Materials The exchange biasing materials investigated most intensively are FesoMnso [48,59-68], amorphous TblOO_xCox(x around 75) [69,70], NiO [32,68,71-74], and NiMn [68,75-77]. Recently, two new materials have been proposed, viz. Pd30Pt2oMnso[78] and IrxMnlOO-x(x typically 20 to 25) [52,79]. At present, no complete overview of the properties of the latter two materials can be given, and the data given below should be considered as preliminary, as Philip. Journal of Research Vol. SI No

24 s00 I-' I-' 0\ TABLE Experimental data on exchange biasing materials. Structurally ordered and disordered materials are indicated with the labels (0) and (d), respectively. Data for Pd30Pt2oMnso and Ir2oMnso are preliminary (see text). The exchange biasing field given refers to exchange biasing of a 6 nm NisoFe2o layer. In cases where biasing of such layers was not studied, an estimate of Heb was made using eq. (6). The thickness ofthe exchange biasing layers is close to or larger than the layer thickness to, above which biasing is fully developed. The antiferromagnetic or ferrimagnetic ordering temperatures TN or Tc refer to bulk materials. Tb is the blocking temperature, or (for ferrimagnetic Tb-Co) the temperature above which biasing disappears irreversibly. The other entries in the table refer to room temperature measurements I ~ g Cl> ;::;- C C s Cl>... ~ :- Tb to Keb Heb P CC) (nm) (ml/nr') (kalm) (10-8 Om) FesoMnso (d) 140 b, 150 c 7 c 0.12 b, 0.09 c, 0.17 d 20 b, 15 c, 29 d 130 c, 74 e T N = 230 C a a - TblOO_xCox (d) >250 g, 150 h IOOg,40 h O.33 g, O.28 h 55g,46 h 236 h ;;! s: Tc = 330 C for "lil 67 < x < 83 f... 0 c :l NiO (0) 200 c,i 35 c 0.05 c, 0.06 i 9 c, 11 i!!. insulator e, T N = 252 C ~ " NiMn (0) ",450 c, 380 j 25 c O.27 c, 0.27 j 45 c,44 j 175 c ";::.,. T N = 800 C Pd30Pt2oMnsO (0) 300 k <25 k 0.11 k 19 k ~ TN '" 640 C 1 z Ir2oMn8o (d) 310 m <15 m O.14 m,n 23 m,n "- T N = 450 CP

25 ~ f 5- l s, i s. ~ ::!l ~ ~,_. -l a) Endoh and Ishikawa [58]. b) Sputter deposited spin valves, on Ta buffer layer, Rijks et al. [22,37]. C) Sputter deposited on glass and annealed at 240 C, Lin et al. [68]. d) Sputter deposited inverted spin valves, Lenssen et al. [33]. e) Rijks [22]. f) Hansen [84]. g) Results for x = 72.5, Cain et al. [69]. h) Results for x = 77.5, Freitas et al. [70]. i) Soeya et al. [72]. j) Mao et al. [80]. k) Kishi et al. [78]. I) Value estimated by linear interpolation between Néel temperatures of PdMn and PtMn [85]. m) Fuka et al. [79]. n) Ordering (by annealing) leads to a further increase of Heb [79]. P) Yamaoka et al. [86]. G) IS ~ s ~ ;::: '" ê) ~ 0;;. ~ ::li l:l... '"... IS ë:;' 'ê>... ~ l:l... ;::,. ~ ~

26 R. Coehoorn et al. only a limited amount ofinformation is available. We have included these data in view ofthe potential attractiveness ofpd30pt2omnso and IrxMnlOo_xfor the application in read heads. FesoMnso is an antiferromagnetic random substitutional fee ('y-phase) alloy. Although this phase is thermodynamically unstable at room temperature, I-phase films with thicknesses of a few tens of nanometers may be obtained by epitaxial growth on fee conforming substrates such as Cu or permalloy, to which its lattice parameter is matched within 0.4% and 2.5%, respectively. Amorphous TblOO_xCox alloys, with 67 < x < 83, are ferrimagnets with antiparallel Tb and sublattice magnetizations. The compensation temperature varies from 20 to 200 C from x = 77 to x = 70 [84]. NiO is an insulating antiferromagnetic compound with the cubic NaCI structure. NiMn and Pd30Pt2oMnso are both antiferromagnetic intermetaiiic compounds. The Ni (or Pd and Pt) and Mn atoms reside on successive (001) planes of an fee lattice which, because of the ordering, is tetragonally distorted (CuAu-I structure). In Pd30Pt2oMnsothe Pd and Pt atoms randomly occupy the sites in the same layer. Ir2oMnso, the material composition used for exchange biasing in Ref. [52], forms an antiferromagnetic random substitutional fee alloy. Experimental data on exchange biasing by these materials have been included in Table I. From the point of view ofrobustness ofthe deposition process and materials' design flexibility, FesoMnso is quite ideal. Variations in the alloy stoichiometry between Fe4sMnss and FessMn4s do not strongly affect the exchange biasing field, and deposition can be at room temperature. In addition, if the AF layer is grown first (bottom spin valve, growth of FesoMnso on a permalloy buffer layer) structural matching stili leads to the strong [111] fibre texture of the free magnetic layer that is required from the point of view of good magnetic softness and low sheet resistance (see Sec. 4) [33]. For the application in GMR tape heads the exchange biasing field is sufficiently strong at room temperature. The blocking temperature is well above the maximum temperature during operation, which is typically 80 C. However, during fabrication of the head the maximum temperature can be higher. Although after heating to a temperature close to or above the blocking temperature of about 150 C exchange biasing can be restored by cooling from above Tb in a magnetic field, it would be preferable to have an exchange biasing material available with a higher blocking temperature. As can be seen from Table I, significantly higher blocking temperatures have been found for amorphous Tb-Co, NiO, Pd30Pt2oMnso, Ir2sMn7s and, in particular, NiMn. Disadvantages of Tb-Co and NiO (with respect to FesoMnso) are the larger sensitivity of the exchange biasing field and the 118 Philips Journal of Research Vol. SI No. I 1998

27 Giant magnetoresistance materials for read heads blocking temperature to the detailed process conditions. A disadvantage of NiO is the smaller exchange biasing interaction (at room temperature) and the much larger coercivity of the exchange biased layer. Blocking temperatures of rv450 C have been obtained using antiferromagnetic NiMn layers. However, the formation of the proper ordered antiferromagnetic phase requires an anneal treatment of at least several hours at about 255 C, which in spin-valve layered structures containing permalloy/cu interfaces would lead to a drastic decrease of the MR ratio due to interface interdiffusion. A first method to circumvent this problem has recently been proposed in Ref. [77], where it was shown that a fair exchange biasing field can be realized by post-deposition vacuum annealing of (substrate/buffer layer/50 nm NiMn/ 25 nm permalloy) structure. In principle, a complete (bottom-type) could then be grown on top of this structure. Secondly, as demonstrated in Ref. [80], the thermal stability required in view of the post-deposition annealing step can be made adequate by depositing a thin (1.5 nm) layer of Co at the permalloy/cu interface. As in the case of NiMn, Pd30Pt2oMnso films are random substitutional fee-type alloys after deposition at room temperature, and transform to the ordered CuAu-I structure upon annealing. The presence of this phase is required to obtain exchange biasing. The Pd to Pt concentration ratio given above was found to be optimal with respect to the exchange biasing field obtained after annealing at 230 C during 1hour. All data given in Table I refer to this annealing procedure. Compared with NiMn, Pd30Pt2oMnsomay for some applications be advantageous in view ofthe lower annealing temperature and/or annealing times required. No post-deposition annealing step is required for obtaining exchange biasing using IrxMnlOo-x fee-type thin films with a predominant (111) texture in the approximate concentration range 10 < x < 75 [79]. Exchange biasing is not observed for films with the ordered fct-type CuAu-I structure, which in the approximate concentration interval40 < x < 60 is the thermodynamically stable crystal structure. The fee structure required for obtaining exchange biasing can, for films in this concentration interval, be obtained by growth on a (111) fee conforming substrate such as permalloy or Co [79]. Relatively high blocking temperatures and fairly high exchange biasing fields at room temperature have been reported (see Table I for Ir20MnSO),making this material potentially attractive for applications in read heads. The exchange biasing field is largest at the lower (Mn-rich) end of this composition interval, and can be increased for layers with a composition close to x = 25 by a postdeposition annealing step (a few hours at 200 C) [79]. This effect has been explained as the effect of the formation of an ordered Cu-Au-type structure. Philips Journni of Research Vol. 51 No