Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 Hydrogen induced changes of the interlayer coupling in Fe  /V V superlattices (x"11-16) D. Labergerie , C. Sutter , H. Zabel *, B. Hjo¨rvarsson Ruhr-Universita(t Bochum, Lehrstuhl fu(r Experimentalphysik/Festko(rperphysik, D-44780 Bochum, Germany Royal Institute of Technology, Materialphysics, S-100 44 Stockholm, Sweden Received 24 July 1998; received in revised form 8 October 1998 Abstract The effect of hydrogen on the magnetic exchange coupling between iron layers through vanadium spacer layers has been studied with magneto-optical Kerr effect experiments in Fe  /V V superlattices. Here x refers to the number of V monolayers varying from 11 to 16 and the Fe layer thickness is fixed at three monolayers. Without hydrogen the superlattice is antiferromagnetic (AFM) for x between 12 and 14 and ferromagnetic (FM) in all other cases. With hydrogen loading the coupling can be switched from AFM to FM and vice versa. As previously observed with neutron reflectivity measurements (Hjo¨rvarsson et al., Phys. Rev. Lett. 79 (1997) 901) the change of the interlayer coupling upon hydrogen uptake is not simply due to the expansion of the non-magnetic vanadium spacer layer but more likely to the distortion of the Fermi surface. Bilinear and biquadratic exchange couplings can be recognized by the magnetic hysteresis loops and their coupling energies have been extracted by fits to the curves. For all samples the easy axis of the magnetization is in the plane without any preferred in-plane direction. Hydrogen loading does not affect the magnetic anisotropy of these samples. 1999 Elsevier Science B.V. All rights reserved. Keywords: Interlayer exchange coupling; Anisotropy; Superlattice; MOKE 1. Introduction increasing Cr spacer layer thickness between fer- romagnetic and antiferromagnetic, exhibiting a Since the first discovery of the exchange coupling short oscillation period of about 2 monolayers of Fe layers through Cr spacer layers by Gru¨nberg (ML) [2,19,20] superposed on a longer period of et al. in 1986 [1], much experimental and theoret- about 22 ML [3]. Concomitantly, a giant mag- ical work has been devoted to this broad and netoresistance effect with a period commensurate promising field. It has been shown that the mag- with the exchange coupling period was discerned netic coupling in Fe/Cr multilayers oscillates with soon after [4,21]. Fe/Cr heterostructures and su- perlattices were the first system which has been extensively studied and for which most of the im- * Corresponding author. Tel.: #49-234-700-36-49; fax: portant properties of exchange coupled superlatti- #49-234-709-41-73; e-mail: hartmut.zabel@ruhr-uni- ces have been unraveled. At the same time, the bochum.de. fast development of deposition techniques via 0304-8853/99/$ - see front matter 1999 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 8 8 5 3 ( 9 8 ) 0 0 5 4 6 - 0 D. Labergerie et al. / Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 239 sputtering and MBE methods allowed fabricating and to determine the coupling energies between the high quality ultrathin films, trilayers, multilayers, iron layers through the vanadium spacers as a func- and superlattices of a number of other systems that tion of the spacer thickness and the hydrogen con- has triggered much experimental effort to investi- centration. gate their intriguing magnetic and transport prop- In the following, we present extensive longitudi- erties (For a most recent update of the development nal magneto-optical Kerr measurements of the in the field see for example, Refs. [5,22]). Most magnetic hysteresis of Fe recently, very exciting polarized neutron reflectivity  /V V (x"11-16) super- lattices with and without hydrogen. experiments have been performed on sputtered These are the first hysteresis measurements on Fe  /V V superlattices investigating spin structures Fe/V(0 0 1) superlattices with Fe layer thickness of and exchange couplings and how these properties only 3 ML. Extensive magnetic measurements on are altered by hydrogen uptake in the vanadium Fe/V multilayers and superlattices containing spacer layers [6]. Here the subscripts indicate the thicker Fe layers have been reported previously by number of Fe and V monolayers period. In particu- Granberg et al. [11]. lar, upon loading the multilayer with hydrogen a switching from antiferromagnetic (AFM) to fer- romagnetic (FM) exchange coupling and vice versa 2. Experimental details was observed. This is similar to the changes of the exchange coupling which have previously been A series of Fe/V superlattices all composed of observed upon hydrogen loading of Fe/Nb multi- three monolayers of iron and with varying layers [7]. In general, hydrogen loading causes vanadium thicknesses (from 11 to 16 MLs) has a volume expansion of the host lattice which is been investigated via the magneto-optical Kerr ef- linearly proportional to the hydrogen concentra- fect in the longitudinal configuration. The samples tion [8]. In case of epitaxial films with elastic were sputter grown on epi-polished MgO [1 0 0] boundary conditions prohibiting an in-plane ex- substrates, as described in more detail in Ref. [12]. pansion, the out-of-plane expansion (parallel to the They were all covered with a 3 nm Pd cap layer to film normal) can reach up to 10%, as observed for allow a fast uptake and release of hydrogen at Fe  /V V superlattices [9,10]. Hjo¨rvarsson et al. a temperature of less than 100°C and to avoid could, however, show that the vanadium lattice oxidation of the sample. The total number of expansion is not the major cause for the switching double layers in this set of samples is N"30. Prior behavior of the interlayer exchange coupling [6]. to hydrogen exposure, the samples were character- The superlattice containing 15 ML of V is at the ized with small angle X-ray reflectivity measure- border line between AFM and FM coupling. In- ments using Mo-K stead of becoming FM coupled upon hydrogen  radiation ( "0.0709 nm). A representative reflectivity curve of the Fe loading, as one would expect from an expanding  /V  sample is shown in Fig. 1. The solid line is a fit to V layer thickness, this superlattice becomes AFM the data points using a generalized Parratt algo- coupled for a limited hydrogen concentration range rithm [13,23]. The first and second order superla- indicative for a specific distortion of the vanadium ttice reflections as well as the finite size oscillations Fermi surface, thereby creating new spanning vec- are well pronounced, proving the high structural tors for the exchange coupling. quality of the superlattice with a low interface The neutron reflectivity experiments could prove roughness of typically less than 1 ML. The nominal the existence of coherent AFM and FM spin struc- ( tures in the Fe/V superlattices and determine quali-  ), as provided by the growth parameters, and the actually measured ( tatively the coupling strengths. However, for a ) bilayer periods of the samples investigated here are listed in Table 1. more quantitative analysis it is necessary to deter- The magnetic hysteresis of the Fe mine complete magnetic hysteresis loops. They  /V V superla- ttices have been investigated with an improved allow us to extract more detailed information magneto-optical Kerr effect (MOKE) setup de- about the type of coupling (bilinear, biquadratic) veloped in one of our laboratory facilities [14,24]. 240 D. Labergerie et al. / Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 Fig. 1. Small angle X-ray reflectivity of the Fe  /V  superla- ttice before loading with hydrogen. The solid line is a fit to the Fig. 2. Schematics of the sample orientation and the scattering data points using a generalized Parratt algorithm. geometry for longitudinal MOKE measurements. The MOKE measurements were performed with a magnetic field along the [1 1 0] axis, which is the hard axis in bulk samples. In our geometry the [1 0 0] axis is oriented along the film diagonal. We briefly outline the basic features of this tech- nique and the refinement used to increase the sensi- tivity of the measurements. MOKE is based on the change of the polarization state of incident light reach an angular resolution of better than 2; reflected by magnetic materials [15]. In the longitu- 10\ deg, which corresponds in the case of Fe dinal configuration, this change is directly propor- layers to a magnetization of 40 emu/cm. tional to the in-plane magnetization of the material The MOKE measurements were performed at [16]. In our case, the plane of polarization of the room temperature along the [1 1 0] axis, which is He-Ne laser beam ( "632.8 nm, P"5 mW) was the projection of the hard axis of the bulk iron in aligned perpendicular to the plane of incidence the film plane (see Fig. 2). The penetration depth (s-state). A Faraday modulation lock-in technique (&50 nm) allows us to see a large number (&20) of has been implemented in the standard setup to double layers without being disturbed by the sub- obtain a better resolution of the detected signal. strate. The sample was positioned in a high vacuum The linear polarization state of the laser beam chamber in the center of a magnetic field. Due to reflected by the sample is modulated sinoidally by the required pole distance the maximum field was passing through a Faraday rod. A second Faraday limited to 2000 Oe. The vacuum chamber was con- rod, installed downstream, turns the polarization nected to a hydrogen loading facility allowing an state in such way that the signal measured is mini- exposure of the sample up to a hydrogen gas pres- mized. For more details see Refs. [14,24]. This sure of 900 mbar. This pressure is more than suffi- refinement in the detection allows us to commonly cient to saturate the sample with hydrogen [9,10]. Table 1 Structural parameters of the different samples investigated Sample Fe  /V  Fe  /V  Fe  /V  Fe  /V  Fe  /V  Fe  /V  N 30 30 30 30 30 30  (As) 20.9 22.4 23.9 25.5 27 28.5  (As) 20.9 21.9 24.4 25.8 27.3 27.5 Rel. deviation (%) 0 !2.2 2.1 1.2 1.1 3.5 D. Labergerie et al. / Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 241 The pressure was measured via a strain gauge with a relative accuracy of 0.5%. The most important changes of the exchange coupling take place at rather low hydrogen pressures between 0 and 15 mbar. In this pressure range the measurements are completely reversible and zero hydrogen pres- sure corresponds to an essentially hydrogen free sample, while at 10 mbar the estimated hydrogen concentration in the vanadium layers is about 20%. Since the signal detected is only 10 times the resolu- tion limit of our system, we have taken several measurements (typically 5) at any given point and averaged them to improve the signal to noise ratio. Using a thin paramagnetic Nb film we have detec- ted a non-negligible Faraday rotation due to the optical windows of our hydrogen loading facility, which subsequently was subtracted from all our measurements. The time between each pressure step was at least 20-30 min in order to reach an equilibrium state of the sample and to guarantee a homogeneous hydrogen distribution in the sample. 3. Results and discussions Fig. 3. Longitudinal MOKE hysteresis loops of Fe  /V V super- lattices with x varying from 11 to 16 and at different hydrogen 3.1. MOKE measurements pressures measured at room temperature. The columns show data for a particular sample, the rows data are for zero hydrogen pressure (1st row), intermediate hydrogen pressures (2nd row), In Fig. 3 the longitudinal MOKE hysteresis and high hydrogen pressures (3rd row). The switching of the loops measured in the experimental configuration magnetic coupling from FM (characterized by a square shape as depicted in Fig. 2 are presented. All measure- with a remanent magnetization) to AFM (hysteresis curve with ments were performed at room temperature on a linear slope at small field values and negligible remanent Fe magnetization) and vice versa with changing V thickness and  /V V superlattices with x varying from 11 to hydrogen pressure can already be seen by inspection of the raw 16 ML and for different hydrogen pressures. Each data (see also text). column of Fig. 3 corresponds to one particular superlattice as indicated, and each row corresponds to different representative hydrogen pressures, starting at zero pressures in the first row, a medium and a very small remanent magnetization (e.g. hydrogen pressure (&5 mbar) in the middle row, Fe and finally a high pressures (&10 mbar) in the  /V  without hydrogen), whereas for a fer- romagnetically coupled system the hysteresis ex- bottom row. For the unloaded samples one can hibits more a square like shape with a rather large easily recognize that the magnetic coupling oscil- remanent magnetization (e.g. Fe lates between FM and AFM as a function of the  /V  without hydrogen). The Fe vanadium thickness. The period of this oscillation  /V   samples exhibit, in addition to the ferro or antiferromagnetic shape, is on the order of 3 MLs (Fe  /V  and Fe  /V  also biquadratic (90°) coupling contributions [17], are AFM). The hysteresis curve for an antifer- discernible by the slow saturation of the magneti- romagnetically coupled system is characterized by zation with increasing external field. It is interesting a linear slope as a function of the applied field to note that for the FM coupled systems the Kerr 242 D. Labergerie et al. / Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 angle in remanence to respect to the Kerr angle saturation increases with the number of vanadium MLs. This may be due to an improving structural quality of the superlattices with increasing vanadium thickness. Next we study the development of the hysteresis loops along the columns. For the first two superla- ttices (Fe  /V   ) we notice that the remanent Kerr rotation increases with increasing hydrogen pressure. Furthermore, these samples exhibit at the highest hydrogen pressure saturation fields which were outside the measurement range. It appears that hydrogen induces a stronger biquadratic coup- ling in these samples. While the overall shape of the hysteresis loops does not change for these first two superlattices, we recognize dramatic changes for all other superlattices with increasing hydrogen pres- sure (concentration). The superlattice Fe  /V  changes shape from AFM to FM, superlattice Fe  /V  from FM to AFM, and finally the super- lattice Fe Fig. 4. Remanent Kerr rotation and saturation field as a func-  /V  starts FM like, changes to AFM tion of the hydrogen pressure for different samples extracted at low pressures and returns to FM at higher pres- from the longitudinal MOKE hysteresis loops at room temper- sures. ature. In Fig. 4, the remanent Kerr rotation and the saturation field extracted from the measurements are plotted as function of the hydrogen pressure. Different interesting features in this figure are 3). In the Fe worth noting. First, hydrogen loading of the Fe  /V  sample, FM coupling occurs at  / a hydrogen pressure where the sample should still V  superlattice leads to a resonance like AFM be AFM coupled as concerns the vanadium layer coupling between 1 and 8 mbar. Indeed, if we con- thickness. In Table 2 we summarize the magnetic sider the oscillatory exchange coupling of Fe  /V V couplings and the saturation fields for the Fe as a function of the number of MLs x, for x"15-16  /V V superlattices as determined by MOKE measure- we expect FM coupling, and for x'16 again AFM ments for the unloaded and loaded samples with coupling. In contrast, only a slight hydrogen pres- hydrogen. sure (&1 mbar) in the Fe  /V  superlattice caus- It has been observed previously with FMR ing only a small lattice expansion causes, however, measurements that in Fe/V superlattices with only a dramatic change of the coupling nature. This 3 MLs of Fe the easy axis is in the plane and that clearly proves that the lattice expansion cannot there is no in-plane anisotropy [18]. We have char- explain the change of the coupling nature as al- acterized the magnetic anisotropy of the Fe ready noted previously [6]. At higher hydrogen  /V  superlattice before and after hydrogen loading of pressures the coupling on the Fe  /V  does not the sample in order to test whether these properties completely return to the FM type but shows a re- depend on the hydrogen concentration. Fig. 5 re- sidual AFM contribution, with a saturation field on produces hysteresis loop of the sample exposed to the order of 40 Oe. This result could not be ob- 800 mbar hydrogen pressure and with the external served by the neutron reflectivity measurements [6] field applied along the [1 1 0] axis (line) and the because of the constant external neutron guide field [1 0 0] axis (diamonds). The difference between which was set to 100 Oe. At this field level the these two hysteresis loops is shown by filled superlattice appears already FM coupled (see Fig. squares. Within the resolution of our system the D. Labergerie et al. / Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 243 Table 2 Summary of the exchange coupling as deduced from magneto-optical Kerr effect measurements Sample Fe  /V  Fe  /V  Fe  /V  Fe  /V  Fe  /V  Fe  /V  Coupling without H FM AFM AFM AFM FM FM H  (Oe) 360 1600 700 Coupling with H FM Becomes FM Becomes FM No result AFM Resonance AFM Resonance H  (Oe) 150 115 Fig. 5. Comparison between the hysteresis loops with external magnetic fields applied along the [1 1 0] and the [1 0 0] axis on the Fe  /V  superlattice and at a constant hydrogen pressure of 800 mbar at room temperature. sample remains isotropic even at these high hydro- be written as [17] gen pressures. This is a surprising result, since the hydrogen uptake causes a large change of the va- , E"! M nadium sublattices and eventually one on the crys- 1t$ H cos( G! &) G tal fields at the Fe/V interface. Therefore, an effect ,\ of hydrogen on the magnetic anisotropy would ! J have been conceivable. * cos( G! G>) G ,\ 3.2. Analysis and computer simulation ! J / cos( G! G>). (1) G We analyzed the magnetic hysteresis loops using Here the first term in Eq. (1) corresponds to the a simple phenomenological model containing no Zeeman energy, the second to the bilinear exchange in-plane anisotropy as justified by the present coupling characterized by a bilinear coupling con- MOKE results (Fig. 5) and by earlier FMR stant J measurements [18]. The magnetic energy per unit * between two adjacent Fe layers, and the third to the biquadratic coupling J area of a superlattice with N double layers can then /. G is the angle between the [1 0 0] direction and the 244 D. Labergerie et al. / Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 magnetic moment in the layer i, and & the angle between the [1 0 0] direction and the applied field H. M is the magnetization in saturation and t$ is the Fe layer thickness. The fit program searches the energies J * and J / that minimizes the total en- ergy E with respect to the angle G. The procedure is then repeated for each field value H. If we assume that the magnetization vectors in two adjacent Fe layers are antiparallel (pure antiferromagnetic coupling: G"! >, and J /"0), it is possible to directly evaluate the coupling constant J$, from the saturation field H according to N J$" H 4(N!1) 1M1t$ . (2) The saturation magnetization values, M, of our samples used to fit the hysteresis loops were deter- mined independently by FMR measurements at room temperature and at a frequency of 9.2 GHz. The magnetic field of 13 kOe sufficient for satura- ting the samples was applied along the [1 1 0] in- plane direction. The results are listed in Table 3. The relatively small magnetization values of the Fe Fig. 6. Bilinear and biquadratic coupling energies extracted  /V  superlattice is probably due to a smaller thickness of the Fe layer compared to the samples from the fit of the hysteresis loops for two superlattices which switch their exchange coupling nature with increasing hydrogen used in Ref. [18] or to a smaller magnetic moment pressure. Top panel: Fe of the Fe atoms at the interface between iron and  /V  , bottom panel: Fe  /V  . vanadium. Using the linear dependence of the satu- ration magnetization (per iron volume) for Fe/V multilayers to the inverse of the number of periods layers exhibit a reduced moment but also the in- n, Poulopoulos et al. have found that the Fe mag- terior Fe layer. netic moment at the interface should be reduced by Returning to the coupling nature of the superla- a factor of 0.32 as compared to bulk iron [18]. In ttices, we have plotted in Fig. 6 the variation of the other words, Fe monolayers embedded between bilinear (J nearest neighbor Fe MLs exhibit the bulk magnetic *) and biquadratic (J /) coupling ener- gies as well as their sum (J moment of 2.2 *#J /) versus the , whereas Fe monolayers at the hydrogen pressure for two respective samples (see interface to vanadium show a reduced moment of also Table 4). We also show the AFM coupling 0.32;2.2 "0.7 . Our magnetization measure- constant (J ments indicate that not only the boundary Fe $) as calculated directly via Eq. (2). We should note that for strongly AFM coupled Table 3 Saturation magnetization of the Fe layers in the Fe/V superlattices as determined by ferromagnetic resonance (FMR) at room temperature Sample Fe  /V  Fe  /V  Fe  /V  Fe  /V  Bulk 4 M (emu/cm) 469 536 417 549 1740 D. Labergerie et al. / Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 245 Table 4 Bilinear and biquadratic coupling energies determined for Fe/V superlattices with and without hydrogen Sample Fe  /V  Fe  /V  Fe  /V  Fe  /V  Fe  /V  Fe  /V  p(H)"0 J * !3.6 !2.0 !8.1 !4.1 #2.6 !1.8 J / !1.8 !1.3 #0.5 #0.2 !1.2 !2.0 J/m p(H)O0 8 mbar 4 mbar J * #1.2 no result !1.0 J / !1.2 #0.1 J/m superlattices the biquadratic coupling is very small starts from FM, switches to AFM with increasing as compared to the bilinear one and can be neglect- hydrogen concentration and returns to FM at the ed. The biquadratic coupling is more important for highest hydrogen concentration. We found that superlattices at the border line between FM and most hysteresis loops can successfully be simulated AFM coupled systems and appears to increase with with a model that takes bilinear and biquadratic hydrogen pressure as mentioned before. In this exchange coupling terms into account. In FM respect the biquadratic coupling contribution can coupled superlattices the biquadratic coupling con- be considered as a precursor to AFM coupling. In stant appears to increase with hydrogen concentra- a purely FM and AFM coupled superlattice, the tion, whereas in AFM coupled superlattices the lateral fluctuation between both exchange coup- biquadratic contribution can be neglected. The lings can be neglected. However, in superlattices experiments clearly show that hydrogen has a which are fine-tuned by the hydrogen content to dramatic effect on the coupling nature of Fe/V the border between FM and AFM coupling, min- superlattices, which cannot be explained solely on ute lateral thickness fluctuations become noticeable the basis of the lattice expansion as a function of the and contribute to the biquadratic coupling. This hydrogen concentration. Changes of the vanadium aspect will be worked out in the future in more Fermi surface upon hydrogen loading creating new detail. spanning vectors for the exchange coupling are required to explain these observations. In contrast, hydrogen has no effect on the magnetic anisotropy 4. Summary of the Fe layers. The magnetization vector of the Fe layers remains in-plane and exhibits no magneto- In summary, we have performed magneto-op- crystalline anisotropy at room temperature with- tical Kerr measurements on Fe  /V V superlattices out and with hydrogen. with x varying from 11 to 16 in order to determine quantitatively the exchange coupling constant be- tween two adjacent iron layers separated by differ- Acknowledgements ent thicknesses of vanadium spacer layers. A small and continuous variation of the vanadium thick- We would like to thank Jens Pflaum for his help ness has been achieved by loading the sample with with the FMR measurements, and Till Schmitte for hydrogen. But more important are the changes of his assistance during the MOKE measurements, the electronic structure induced by the hydrogen and P. Isberg for growing the samples. This work loading. We find that with hydrogen the coupling was supported by the Bundesministerium fu¨r Bi- can continuously be changed from AFM to FM ldung, Wissenschaft, Forschung und Technologie and from FM to AFM. In some cases the coupling under contract 03-ZA4BC1-0, by the EU-TMR 246 D. Labergerie et al. / Journal of Magnetism and Magnetic Materials 192 (1999) 238-246 network FMRX-CT98-0187. H.Z. would like to [10] G. Anderson, B. Hjo¨rvarsson, H. Zabel, Phys. Rev. B 55 thank the Volkswagen-Stiftung for support during (1997) 15905. his sabbatical and BH acknowledges financial sup- [11] P. Granberg, P. Nordblad, P. Isberg, B. Hjo¨rvarsson, port from NFR. R. Wa¨ppling, Phys. Rev. B. 54 (1996) 1199. [12] P. Isberg, B. Hjo¨rvarsson, R. Wa¨ppling, E.B. Svedberg, L. Hultman, Vacuum 48 (1997) 483. [13] A. Abromeit, Diploma Thesis, Ruhr-Universita¨t Bochum, References 1992. [14] T. Zeidler, Ph. D. Thesis, Ruhr-Universita¨t Bochum, 1997. [1] P. Gru¨nberg, R. Schreiber, Y. Pang, M.B. Brodsky, [15] M. Born, E. Wolf, Principle of Optics, Pergamon Press, C.H. Sowers, Phys. Rev. Lett. 57 (1986) 2442. Oxford, 1989. [2] J. Unguris, R.J. Celotta, D.T. Pierce, Phys. Rev. Lett. 67 [16] S.D. Bader, J. Magn. Magn. Mater. 100 (1991) 440. (1991) 140. [17] M. Ru¨hrig, R. Scha¨fer, A. Hubert, R. Mosler, J.A. Wolf, [3] S.S.P. Parkin, N. More, K.P. Roche, Phys. Rev. Lett. 64 J. Appl. Phys. 69 (1991) 4780. (1990) 2304. [18] P. Poulopolous, P. Isberg, W. Platow, W. Wisny, M. Farle, [4] M.N. Baibich, J.M. Broto, A. Fert, F. Nguyen, Van Dau, B. Hjo¨rvarsson, J. Magn. Magn. Mater. 170 (1997) 57. F. Petroff, P. Etienne, G. Creuset, A. Friederich, [19] S.T. Purcell, W. Folkerts, M.T. Johnson, N.W.E. McGee, J. Chazelas, Phys. Rev. Lett. 61 (1988) 2472. K. Jager, J. aan de Stegge, W.B. Zeper, W. Hoving, [5] J. Tobin et al., Magnetic Ultrathin Films, Multilayers and P. Gru¨ndberg, Phys. Rev. Lett. 67 (1991) 903. Surfaces, MRS Symp. Proc., vol. 475, 1997. [20] S. Demokritov, J.A. Wolf, P. Gru¨ndberg, Europhysics Let- [6] B. Hjo¨rvarsson, J.A. Dura, P. Isberg, T. Watanabe, T.J. ters 15 (1991) 881. Udovic, G. Andersson, G.F. Majkrzak, Phys. Rev. Lett. 79 [21] G. Binash, P. Gru¨nberg, F. Saurenbach, W. Zinn, Phys. (1997) 901. Rev. B. 39 (1989) 4828. [7] F. Klose, Ch. Rehm, D. Nagengast, H. Maletta, A. Weidin- [22] Proc. 15th Int. Colloquium on Magnetic Films and Surfa- ger, Phys. Rev. Lett. 78 (1997) 1150. ces, ICMFS '97, J. Physics D: Appl. Phys. 31 (1998) [8] J. Peisl, in: G. Alefeld, J. Vo¨lkl (Eds.), Hydrogen in Metals [23] P. Bo¨deker, A. Abromeit, K. Bro¨hl, P. Sonntag, N. Metoki, I. Topics in Applied Physics, vol. 28, Springer, Berlin, 1978. H. Zabel, Phys. Rev. B 47 (1993) 2353. [9] G. Andersson, B. Hjo¨rvasson, P. Isberg, Phys. Rev. B 55 [24] Th. Zeidler, F. Schreiber, H. Zabel, W. Donner, N. Metoki, (1997) 1774. Phys. Rev. B 53 (1996) 3256.