Surface Science 495 (2001) 68±76 www.elsevier.com/locate/susc Morphology of epitaxial metallic layers on MgO substrates: in¯uence of submonolayer carbon contamination M. Rickart, B.F.P. Roos, T. Mewes, J. Jorzick, S.O. Demokritov *, B. Hillebrands Fachbereich Physik and Forschungs- und Entwicklungsschwerpunkt Materialwissenschaften, Universit at Kaiserslautern, Erwin-Schr odinger-Strasse 56, D-67663 Kaiserslautern, Germany Received 30 May 2001; accepted for publication 6 August 2001 Abstract By investigating the epitaxial growth of Ag(0 0 1) and Au(0 0 1) ®lms and Fe/Ag(0 0 1) and Fe/Au(0 0 1) layered systems on MgO(0 0 1) the in¯uence of carbon contamination of the MgO surface on the morphology of the obtained ®lms is studied. A newtechnique for the preparation of carbon-free MgO(0 0 1) surfaces using ion beam oxidation is reported. This technique takes advantage of the high chemical activity of dissociated lowenergy oxygen atoms, which removes the carbon contamination from the MgO surface as con®rmed by Auger electron spectroscopy. It is shown that metallic layers grown on carbon-free MgO substrates demonstrate reduced roughness, improved crystallo- graphic quality and enhanced surface magnetic anisotropy compared to those grown on carbon contaminated sub- strates. Ó 2001 Elsevier Science B.V. All rights reserved. Keywords: Molecular beam epitaxy; Ion bombardment; Surface structure, morphology, roughness, and topography; Carbon; Magnesium oxides; Metal±metal magnetic thin ®lm structures 1. Introduction cessful epitaxial growth of 3d magnetic Co, Fe, Ni ®lms with a well-layered structure [2,3]. However, The generation of atomically ¯at and chemically for technological applications bulk single crystals clean surfaces is one primary goal of epitaxial thin are not very suitable as substrates, since they are ®lm growth. It is a well-known fact that the di cult to manufacture, expensive, and unconve- chemical structure and morphology of the sub- nient in handling. In addition, the use of metallic strate surface in¯uences dramatically the quality of single crystals impedes resistivity (and magnetore- the resulting ®lm [1]. Conventionally bulk single sistivity) measurements on grown ®lms. Therefore, crystals of cubic metals (Cu, Ag and Au) with the use of semiconducting or insulating substrates almost atomically ¯at surfaces are used for a suc- is necessary. Ag and Au bu er ®lms deposited on semiconducting GaAs(0 0 1) are widely used as templates for epitaxial growth of high quality * Fe(0 0 1) ®lms and multilayers based on Fe(0 0 1) Corresponding author. Tel.: +49-631-205-4075; fax: +49- [4,5]. The bu ers serve as barriers for As atoms, 631-205-4095. E-mail address: demokrit@physik.uni-kl.de (S.O. Demokri- which otherwise di use from GaAs in Fe and cause tov). a magnetic ``dead'' layer [6]. 0039-6028/01/$ - see front matter Ó 2001 Elsevier Science B.V. All rights reserved. PII: S0039-6028(01)01504-7 M. Rickart et al. / Surface Science 495 (2001) 68±76 69 On the other hand insulating MgO is a widely high temperature furnace [15,16] in order to re- used substrate for growth of metallic ®lms due to construct surface deteriorations arising by storage its superior high temperature stability and low in air and to diminish the carbon contamination. chemical activity [7,8]. These substrates are com- Chemical cleaning with isopropanol or etching [17] paratively inexpensive and commonly available in followed by UHV annealing at elevated tempera- a variety of orientations and sample dimensions. tures [18,19] is also a frequently used method. MgO has rock-salt structure with a lattice spacing With these methods it is possible to prepare a of 0.421 nm. The in-plane mismatch of MgO(0 0 1) smooth single crystal with a vertical surface rough- to fcc Ag(0 0 1), fcc Al(0 0 1), fcc Au(0 0 1), bcc ness of 0.2±0.4 nm as corroborated by atomic force Fe(0 0 1) and bcc Cr(0 0 1) ®lms is below4%. How- microscopy (AFM) [15]. But high temperature ever, very often an essential contamination of the treatment of the MgO surface in vacuum and/or in MgO surface by carbon is observed [9], prohibiting oxygen atmosphere does not result in a carbon- a good epitaxial growth. Depending on the carbon free surface [19], probably, due to continuous dif- concentration such a contamination leads to sev- fusion of carbon from the bulk of the crystal to the eral reconstructions of the deposited ®lms [10]. It is surface. The ion beam treatment reported here therefore important to understand the in¯uence of takes place at room temperature, thus avoiding a the surface structure and its possible contamina- possible recontamination of the surface by di u- tion on the morphology of high quality thin ®lms sion of carbon from the bulk. grown on these substrates. In this study we examine the in¯uence of carbon contamination of MgO substrates on the mor- 2. Experimental procedure and preparation of the phology of metallic bu er layers consisting of Ag substrate surface or Au and the magnetic properties of Fe thin ®lms prepared onto these bu er layers. A newtechnique MgO(0 0 1) substrates with the nominal purity for the preparation of carbon-free MgO(0 0 1) sur- of 99.9% provided by Crystal GmbH were rinsed faces using lowenergy (<100 eV) oxygen ion beam in isopropanol at ambient pressure before loading treatment to remove the carbon contamination on into the UHV. The substrate dimensions are 10 the substrate is presented. We demonstrate that 10 0:5 mm3. All samples were preheated in UHV the ion treatment not only results in a removal of at a moderate temperature of 150 °C for 30 min to carbon from the substrate, but provides a means to remove water from the surface. Three types of sam- smooth the MgO surface. ples were prepared for the experiment: type (A): The surface of commercially available MgO pre-heated MgO substrates, without any further substrates is contaminated mainly by water and cleaning before the growth of the metallic layers; carbon. Several methods have been proposed to type (B): annealed MgO substrates, heated in UHV reduce the carbon contamination and to obtain a up to 600 °C for 180 min; type (C): ion beam carbon free surface of MgO with low roughness: treated MgO substrates, treated with an atomic cleavage of the single crystals in ultra high vacuum oxygen ion beam at room temperature. The oxy- (UHV) is one of the most practised methods to gen ion beam treatment of the MgO is performed produce a defect free surface without any con- using a novel type of an excited electron cyclotron tamination [11,12]. Cleaving the substrates in air wave resonance (ECWR) controlled plasma reac- leaves water at the surface and further contami- tor (COPRA 160, CCR Technology). The reactor nation and defects even if the substrate is intro- is a high frequency, lowpressure plasma source duced into the UHV immediately. In this case the with an inductive excitation. A crucial advantage substrate has to be heated to temperatures up to of the source is its ability to produce lowenergy 600 °C for several hours to remove these adsor- (20±100 eV) ion beams. The ion beam is auto- bates [13,14]. Another possibility is to cleave the matically neutralized by a corresponding electron single crystals in air with heating subsequently in current, thus minimizing charging e ects on the oxygen atmosphere at 1000 °C for about 1 h in a substrate. The ®lament free design of the source 70 M. Rickart et al. / Surface Science 495 (2001) 68±76 allows one to perform the ion treatment without (STM). AFM with a typical lateral resolution of contamination of the surface by ®lament materi- 3 nm was used to for investigations of insulating als. The source produces an oxygen beam with a MgO. Magnetic anisotropies of Fe ®lms were de- dissociation degree up to 80%. The ion current rived from the frequencies of the spin waves de- density was chosen to be 0.1 mA/cm2 and a process termined by means of Brillouin light scattering time of 2 min, providing atomic oxygen ion doses spectroscopy [22,23]. of 6 1016 ions/cm2. Ions with the nominal energy of 35 eV and an energy distribution width below 10% of the nominal energy were used for the ion 3. Results and discussion beam treatment. The ion current density and the ion energy were monitored using a Faraday cup. 3.1. Properties of the MgO substrates The pressure in the chamber during the ion beam treatment was 6 10 4 mbar. A detailed technical For the type (A) pre-heated MgO substrate an description of the source can be found elsewhere intensive carbon (KLL: 272 eV) peak is found in [20]. the Auger electron spectrum as seen in Fig. 1a. The Metallic ®lms were prepared in an UHV multi- observed intensity of this peak corresponds to an chamber molecular beam epitaxy (MBE) system e ective carbon coverage of 0.28 ML. A similar with a base pressure of less than 5 10 11 mbar. amount of carbon impurities is also found on pre- They were deposited either by a ®ve-pocket elec- heated substrates, which has not been cleaned with tron beam evaporator or a Knudsen cell (Ag) onto isopropanol. The type (B) annealed substrates re- the MgO(0 0 1) substrates with deposition rates veal a reduced carbon peak, as it is shown in Fig. between 0.01 and 0.1 nm/s as monitored by a 1b, with an e ective carbon coverage of 0.17 ML. quartz microbalance. Type (C) substrates cleaned by oxygen ion beam Analysis of the surface morphology and the treatment do not demonstrate any carbon con- sample structure was performed in situ with low tamination above the detection limit of the AES energy electron di raction (LEED) using a rear detector, which corresponds to 0.01 ML. During viewLEED system (ErLEED, VSI GmbH) and re- the ion treatment the oxygen ions most likely clean ¯ecting high energy electron di raction (RHEED, EK-35-R Staib Instruments). For quantative ana- lysis of the di raction patterns the systems were combined with a CCD camera and a digital image processing system with a resolution of 512 512 pixels and a dynamic range of 10 bits. The same LEED system was used as a retarding Auger spectrometer to control the carbon contamination. Using Auger spectroscopy it is very di cult to determine the distribution of impurity atoms within the escape depth of the Auger electrons [21]. In this work for quantitative comparison of dif- ferent Auger spectra the observed intensities of the carbon Auger-peak (KLL line at 275 eV), with the escape depth of the electrons being 3.5 ML, have been recalculated into the e ective surface carbon coverage, expressed in percentage of one mono- Fig. 1. Auger electron spectra of MgO surfaces. Curve (a) layer [21]. The surface topography in the real space shows a pre-heated substrate, (b) annealed at 600 °C for 3 h and was studied with a commercial Park Scienti®c In- (c) a substrate treated with an oxygen ion beam. No carbon contamination was detected in spectrum (c). All spectra are struments Autoprobe VP 2 UHV device, a com- normalized to the Mg peak for comparison and vertically bined AFM and scanning tunneling microscope shifted for clarity. M. Rickart et al. / Surface Science 495 (2001) 68±76 71 and ``reconstruct'' the surface due to the high dissociation degree of the oxygen ions and a chemical reaction with carbon atoms. The ob- tained COx is then pumped out from the UHV chamber. In addition to chemical cleanness the ion beam treated MgO(0 0 1) surfaces exhibit much sharper LEED 1 1 di raction patterns with essentially no di use background compared to the pre-heated or even the annealed sample. The mean coherence length in the electron scattering process is derived from the analysis of the inverse width (FWHM) of the LEED spots multiplied with the corresponding lattice parameter [24]. Since for the case of MgO (and Fe, see below) periodical changes of the spot width as a function of the electron energy have been observed, the obtained Fig. 2. Mean electron scattering coherence length obtained by evaluation of the LEED spot width of the MgO substrates, the mean coherence length, presented in Fig. 2 can be Ag and Au layers and the Fe ®lms for the di erent preparation interpreted as a mean terrace width of the surface methods of independent substrates. The vertical broken line [24]. As it is seen from Fig. 2, the terrace width is separates regions with di erent substrate pre-treatment: an- highest for the ion beam treated substrates. AFM nealed and pre-heated substrates. topography measurements on the pre-heated sub- strates showing a rough surface in real space, [25]. For the same reason monatomic steps on corroborate the results of our LEED analysis. The MgO cannot be resolved. Therefore a direct com- measured root mean square (RMS) roughness parison between the LEED- and the AFM-data is obtained by AFM on the pre-heated MgO surfaces not possible. is 0.7 nm. As seen in Fig. 3 the ion beam treated surface is much smoother showing a measured 3.2. Growth of the metallic non-magnetic layers RMS roughness of about 0.1 nm. Note here, that due to averaging over the ®nite lateral resolution As it has been already mentioned, MgO with its of the AFM (typically 3 nm), the obtained images lattice spacing of 0.421 nm is an excellent substrate may not represent the true surface morphology for lattice matched growth of fcc Ag(0 0 1) and fcc Fig. 3. AFM images of MgO substrates, (a) pre-heated substrate (type A), RMS 0:7 nm, (b) oxygen ion beam treated substrate (type C), RMS < 0:1 nm. The insets showthe corresponding LEED pattern at an energy of 107 eV. 72 M. Rickart et al. / Surface Science 495 (2001) 68±76 Au(0 0 1) metallic ®lms. Palmberg et al. ®rst re- ported epitaxial growth of Ag(0 0 1) and Au(0 0 1) on vacuum cleaved MgO(0 0 1) surfaces with a rather high amount of disorder in the ®lms [26]. In agreement with Refs. [26,27] a three-dimensional island growth of Ag and Au on MgO has been observed in the present study, which is due to the high mobility of the Ag and Au atoms on MgO. As layer-by-layer growth is desired an intermedi- ate layer of 1 nm Fe deposited onto the MgO substrate at room temperature improves the growth of the following Ag and Au layers remarkably [28] as one can see in the RHEED images of the Ag ®lms in Fig. 4. The in-plane epitaxial relation be- tween the bcc Fe seed layer and the MgO rock-salt structure is Fe 110 kMgO[1 0 0] as observed by LEED. After deposition of the intermediate Fe layer a Fig. 5. STM image of an Ag layer demonstrating monatomic 150 nm thick Ag layer was deposited at a rate of steps. The inset shows for Ag a characteristic 1 1 LEED 0.5 ML/s at a substrate temperature of 120 °C. pattern. LEED investigations showthat the Ag layer grows with the [1 0 0]-direction parallel to the [1 0 0]- direction of the MgO substrate. STM investiga- indicating an excellent crystallographic order of tions of the Ag(0 0 1) ®lms prepared on the type the type (C) samples. (C) substrates reveal atomically smooth terraces Epitaxial Au(0 0 1) ®lms with the thickness of with the mean size of 50±70 nm, as seen in Fig. 5. 150 nm were grown on MgO/Fe(0 0 1) at a depo- LEED shows an exceptionally intense 1 1 sition rate of 0.5 ML/s and a substrate tempera- pattern with less background compared to the ture of 120°C. Similar to the growth of Ag the samples grown on type (A) and (B) substrates, [1 0 0]-direction of Au is orientated along the Fig. 4. RHEED images of a 50 nm Ag bu er layer grown onto the MgO substrate (a) without and (b) with a 1 nm Fe intermediate layer. The incident beam with the energy of E 15 keV was directed along the [1 1 0]-direction of the MgO substrate. M. Rickart et al. / Surface Science 495 (2001) 68±76 73 Fig. 6. STM images of an Au layer showing the 5 20 reconstruction and monatomic steps. The inset shows a LEED pattern at 75 eV of the 5 20 reconstruction. [1 0 0]-direction of MgO. As it is seen from the 3.3. Fe ®lms grown on Ag and Au layers STM image shown in Fig. 6, Au(0 0 1) grows in a layer-by-layer mode where the uppermost layer It is well known that Fe(0 0 1) grows in the bcc forms a 5 20 reconstruction, known from the phase in a 45° rotation of its lattice on fcc Ag(0 0 1) literature [29]. The atomic rows of the recon- and Au(0 0 1) with a corresponding in-plane lattice structed surface are orientated parallel the [1 1 0]- mismatch of less than 1% [4,6]. The high quality of directions of Au. The existence of domains with Ag and Au ®lms is an important premise for a di erent orientation ([1 1 0] or [1 1 0]) of the re- good growth of Fe ®lms on those ®lms. In this construction restores a fourfold symmetry of the section we present the results for Fe ®lms de- surface, as con®rmed by the LEED pattern, shown posited on the Ag and Au bu ers grown on in the inset of Fig. 6. For a scan area of 0:1 MgO(0 0 1) surface with di erent carbon contam- 0:1 lm2 the RMS roughness of the surface is 0.26 ination, prepared as it was described above. A 3 nm. The results of the LEED-spot width analysis, nm Fe ®lm was deposited at a temperature of presented in Fig. 2 for both Ag and Au ®lms, show 120 °C and at a rate of 0.04 ML/s. The obtained that Ag and Au ®lms grown on ion beam treated LEED patterns (Fig. 7a) con®rm an epitaxial MgO substrates (type C) exhibit a higher correla- growth of Fe(0 0 1) ®lms on both Ag and Au tion length compared to those grown on annealed bu ers. The value of mean coherence lengths de- substrates (type B) or pre-heated substrates (type rived from the width of the LEED spots for the Fe/ A). Note here, that contrary to the case of the bu er/MgO layered system increases from 2 nm MgO surfaces, the spot width of the Ag and Au for pre-heated substrates to 3 nm, obtained for LEED images do not oscillate as a function of the the systems grown on ion beam treated substrates. electron energy. It means, that the observed values No measurable di erence between the coherence of mean coherence length, presented in Fig. 2 for length for Fe/Ag and Fe/Au systems is observed. Ag and Au are determined not by atomic steps, The energy dependence of the spot width demon- but by defects within the atomic layers [24]. This is strates an oscillating behaviour, indicating that the in agreement with the STM data, which indicate atomic steps of the Fe surface limit the coherence that the mean terrace widths of the Ag and Au length. This is in agreement with the STM images surfaces are much larger than the coherence of Fe shown in Fig. 7 for Fe/Ag and Fe/Au sys- lengths, derived from LEED. tems. In both cases islands with the mean island 74 M. Rickart et al. / Surface Science 495 (2001) 68±76 Fig. 7. (a) LEED pattern E 180 eV of a 3 nm thick Fe ®lm, epitaxially grown on Au, (b) STM image of the same ®lm showing islands, (c) large scale STM image of the same Fe ®lm, (d) large scale STM image of a 3 nm thick Fe ®lm, epitaxially grown on Ag. Note that in (c) the direction [1 0 0] on Fe corresponds to [1 1 0] on Au. size of about 15 nm2 and the height of 3±4 ML is enough even to change the equilibrium orientation observed also in agreement with Ref. [30]. Note, of the ®lm magnetization [31]. The in¯uence of the that the large atomically smooth terraces of the morphology of the ®lm on its magnetic anisotro- bu ers can be clearly traced on the Fe surface, as it pies is thoroughly studied (see e.g. review[3]). is seen in Fig. 7c and d. Bruno [32] has theoretically shown that surface Surface magnetic anisotropy is a decisive char- roughness reduces the surface anisotropy contri- acteristic property of thin magnetic ®lms for butions. Thus, Fe ®lms prepared on ion beam application in e.g., magnetic sensors. Magnetic treated MgO substrates demonstrate improved anisotropy in general describes the dependence of smoothness and thus possess an enhanced surface the magnetic energy of a sample on the orientation anisotropy with respect to those prepared on the of its magnetization. Contrary to magnetic bulk annealed or pre-heated substrates. anisotropies the surface anisotropy describes the The surface magnetic anisotropy of the grown contribution of surfaces or interfaces of a ®lm to ®lms has been determined ex situ using Brillouin its magnetic energy. For thin ®lms it can be large light scattering from spin waves in the ®lm [23]. M. Rickart et al. / Surface Science 495 (2001) 68±76 75 For ex situ measurements Fe ®lms have been value of k 2 S 0:7 erg/cm2 obtained in this study capped with a 4 nm thick Au overlayer in the case for the Au/Fe/Au system is even higher than that of an Au bu er layer and 2 nm Ag and a 2 nm one known from the literature (k 2 S 0:54 erg/cm2) Cr overlayer in the case of an Ag bu er layer to [3,33]. establish chemical symmetry on both interfaces of the Fe ®lms as well as to prevent corrosion. Fol- lowing the standard approach [3] we characterize 4. Conclusion the magnetic surface anisotropy by the uniaxial out-of-plane anisotropy constant k 2 We have shown that the growth of Ag(0 0 1) and S determined by its contribution to the magnetic energy of the Au(0 0 1) ®lms on MgO(0 0 1) drastically depends ®lm as EV 2k 2 on the carbon contamination of the MgO surface S sin2 h=d, where EV is density of magnetic energy of the ®lm, h is the angle between and its smoothness. Using a newly developed the ®lm magnetization and the normal to the ®lm technique±±lowenergy ion beam treatment±±we plane, and d is the thickness of the ®lm. The sur- succeeded to prepare a carbon free MgO(0 0 1) face anisotropy energy changes the frequency of surface with a very low roughness. We have also spin waves in the ®lm, which is measured by means presented the layer-by-layer growth of Ag and Au of Brillouin light scattering. The details of the on such carbon free MgO surfaces covered by an measurement set-up and the data evaluation pro- intermediate thin Fe layer, as it is con®rmed by cedure can be found elsewhere [23]. In fact, the three independent experimental techniques (LEED, measurements reveal a higher value of k 2 RHEED, and STM). The surface magnetic an- S for the Fe ®lms prepared on the ion beam treated (type isotropies of Fe(0 0 1) ®lms, grown on Ag/Fe/MgO (C)) MgO substrates as seen in Fig. 8, which is in and Au/Fe/MgO-bu ers were measured using agreement with the better quality of the Au and Brillouin light scattering. The obtained results cor- Ag bu er layers described above. The highest va- roborate a close relationship between high values lue of k 2 of the surface anisotropy and high layer quality of S for the Ag/Fe/Ag system obtained in this study (ion-beam treated substrates) k 2 the ®lms. S 0:6 erg/ cm2 is comparable to the value of k 2 S 0:79 erg/ cm2 obtained on high quality Fe whiskers [33]. The Acknowledgements Support by the Deutsche Forschungsgemeins- chaft and the ESF programme NANOMAG is gratefully acknowledged. One of the authors (T.M.) acknowledges support by the Studienstiftung des deutschen Volkes. References [1] M.A. Herman, H. Sitter, in: M.B. Panish (Ed.), Molecular Beam Epitaxy, second ed., Springer, Berlin, 1996. [2] J.A.C. Bland, B. Heinrich (Eds.), Ultrathin Magnetic Structures, vols. I and II, Springer, Berlin, 1994. [3] U. Gradmann, Magnetism in ultrathin transition metal ®lms, in: K.H.J. Buschow(Ed.), Handbook of Magnetic Fig. 8. Evaluation of the uniaxial out-of-plane surface aniso- Materials, vol. 7, North-Holland-Elsevier, Amsterdam, tropy constants in dependence of the carbon concentration. Spin 1993. wave frequencies were measured by means of Brillouin light [4] P. Gr unberg, S. Demokritov, A. Fuss, R. Schreiber, J.A. scattering spectroscopy and the anisotropy constants calcu- Wolf, S.T. Purcell, J. Magn. Magn. Mater. 104±107 (1995) lated. 1734. 76 M. Rickart et al. / Surface Science 495 (2001) 68±76 [5] T. Leeb, M. Brockmann, F. Bensch, S. Miethaner, G. [20] M. Weiler, K. Lang, E. Li, J. Robertson, Appl. Phys. Lett. Bayreuther, J. Appl. Phys. 85 (1999) 4964. 72 (1998) 1314. [6] J.J. Krebs, B.T. Jonker, G.A. Prinz, J. Appl. Phys. 61 [21] M.P. Seah, in: D. Briggs, M.P. Seah (Eds.), Practical (1987) 2596. Surface Analysis, Wiley, NewYork, 1983, p. 181. [7] Y.C. Lee, P. Tong, P.A. Montano, Surf. Sci. 181 (1987) 559. [22] S. Demokritov, E. Tsymbal, J. Phys, Cond. Mat. 6 (1994) [8] C. Li, R. Wu, A.J. Freeman, C.L. Fu, Phys. Rev. B 48 7145. (1993) 8317. [23] B. Hillebrands, Brillouin light scattering from layered [9] F. Didier, J. Jupille, Surf. Sci. 307±309 (1994) 587. magnetic structures, Topics Appl. Phys. 75 (2000) 174. [10] G. Gewinner, J.C. Peruchetti, A. Jaegle, R. Riedinger, [24] M. Henzler, Appl. Surf. Sci. 11/12 (1982) 450. Phys. Rev. Lett. 43 (1979) 935. [25] H.N. Yang, G.C. Wang, T.M. Lu, Di raction from Rough [11] J.B. Zhou, H.C. Lu, T. Gustafsson, P. H aberle, Surf. Sci. Surfaces and Dynamic Growth Fronts, World Scienti®c, 302 (1994) 350. Singapore, 1993. [12] G. Fahsold, A. Priebe, N. Magg, A. Pucci, Thin Solid [26] P.W. Palmberg, T.N. Rhodin, C.J. Todd, Appl. Phys. Lett. Films 364 (2000) 177. 11 (1967) 33. [13] C. Duriez, C. Chapon, C.R. Henry, J.M. Rickard, Surf. [27] T. Suzuki, S. Hishita, K. Oyoshis, R. Souda, Surf. Sci. 442 Sci. 230 (1990) 123. (1999) 291. [14] T. Suzuki, R. Souda, Surf. Sci. 445 (2000) 506. [28] P. Etienne, J. Massies, S. Lequien, R. Cabanel, F. Petro , [15] F. Klinkhammer, Ch. Sauer, E.Yu. Tsymbal, S. Handschuh, J. Crystal Growth 111 (1991) 1003. Q. Leng, W. Zinn, J. Magn. Magn. Mater. 161 (1996) 49. [29] M.A. van Hove, R.J. Koestner, P.C. Stair, J.P. Biberian, [16] L.W. Guo, T. Hanada, H.J. Ko, Y.F. Chen, H. Makino, T. L.L. Kesmodel, I. Bartos, G.A. Somorjai, Surf. Sci. 103 Yao, Surf. Sci. 445 (2000) 151. (1981) 189. [17] S.S. Perry, H.I. Kim, S. Imaduddin, S.M. Lee, P.B. Merril, [30] D.E. B urgler, C.M. Schmidt, J.A. Wolf, T.M. Schaub, J. Vac. Sci. Technol A 16 (6) (1998) 3402. H.-J. G untherodt, Surf. Sci. 366 (1996) 295. [18] J. Dekoster, S. De Groote, T. Kobayashi, G. Langouche, [31] M.L. Neel, J. Phys. Rad. 15 (1954) 376. J. Magn. Magn. Mater. 148 (1995) 93. [32] P. Bruno, J. Phys. F: Met. Phys. 18 (1988) 1291. [19] S.M. Jordan, J.F. Lawler, R. Schad, H. van Kempen, [33] B. Heinrich, Z. Celinski, J.F. Cochran, A.S. Arrott, K. J. Appl. Phys. 84 (1998) 1499. Myrtle, J. Appl. Phys. 70 (1991) 5769.