Journal of Magnetism and Magnetic Materials 198}199 (1999) 396} 401 Invited paper Exchange coupling in magnetic multilayers grown on iron whiskers J. Unguris*, R.J. Celotta, D.A. Tulchinsky, D.T. Pierce Electron Physics Group, National Institute of Standards and Technology, Building 220, Room B206, Gaithersburg, MD 20899, USA Abstract Meaningful tests of theoretical predictions of magnetic multilayer properties require the fabrication of multilayers with nearly the same atomic scale precision as the theoretical models. Multilayers grown epitaxially on single-crystal Fe whisker substrates come very close to this ideal standard. We have investigated the growth, magnetic structure, and exchange coupling of Fe/(Ag, Au, Cr, Mn, V, Cu, or Al)/Fe (1 1 0) whisker structures primarily using re#ection high-energy electron di!raction (RHEED) and scanning electron microscopy with polarization analysis (SEMPA), and in some systems, confocal magneto-optic Kerr e!ect (MOKE) microscopy. In cases of nearly layer-by-layer growth, the measured oscillatory coupling periods and strengths agree well with theoretical predictions. For rougher growth, less predictable non-collinear coupling is generally observed. 1999 Elsevier Science B.V. All rights reserved. Keywords: Interlayer exchange coupling; Magnetic multilayers; SEMPA; Magnetic microstructure; Spin density waves 1. Introduction decomposition of FeCl in an H atmosphere, are among the most perfect metal crystals known [5]. More impor- In certain cases, growing magnetic multilayers on Fe tantly for thin "lm growth, nearly perfect (1 0 0) surfaces whisker substrates has achieved the nearly perfect atomic can simply be obtained by in situ ion sputtering and ordering required to make meaningful comparisons to thermal annealing [6]. Scanning tunneling microscopy theories of magnetic exchange coupling. Exchange coup- (STM) measurements of these surfaces reveal step densi- ling measurements of Fe multilayers with Ag [1], Au ties of a single atomic step per micrometer over hundreds [2,3], or Cr [4] interlayers grown on Fe whiskers have of micrometers [7]. Re#ection high-energy electron shown that many of the discrepancies between experi- di!raction (RHEED) patterns consist of very sharp dif- ment and theory can be resolved by removing or quan- fraction spots with no streaking and minimal di!use tifying the atomic scale disorder in the multilayer. We scattering. The whiskers used in these measurements have examined several other potential interlayer mate- were about 10 m long, a few tenths of a mm wide, rials and extended some of our earlier measurements on and had a rectangular cross section that stabilized the Cr. This paper summarizes our results for these various magnetic structure into two opposite domains. spacer layers. The samples were prepared and measured following procedures described previously [4,6,8]. Fig. 1 shows 2. Experimental a schematic diagram of the sample con"guration and an example of a scanning electron microscopy with polar- An essential ingredient in all of these experiments is the ization analysis (SEMPA) measurement from an Fe whisker substrate. These whiskers, grown by thermal Fe/Au/Fe sample. Wedge-shaped interlayers are grown in order to investigate the exchange coupling as a func- * Corresponding author. Tel.: #1-301-975-3712; fax: #1- tion of spacer layer thickness. Typical wedge slopes were 301-926-2746. about 1 atom layer per 10 m. SEMPA measurements E-mail address: unguris@epg.nist.gov (J. Unguris) were made before, during and after Fe deposition. 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 1 1 3 9 - 1 J. Unguris et al. / Journal of Magnetism and Magnetic Materials 198}199 (1999) 396}401 397 Fig. 1. (Bottom) Schematic diagram of the sample con"guration consisting of the Fe (0 0 1) whisker, a wedge-shaped spacer layer (Au in this example), and a thin Fe top layer. (Top) The top image shows the spatial RHEED intensity oscillation measure- Fig. 2. RHEED intensity oscillations as a function of thickness ment of the spacer which provides a precise measure of the for deposition of the various materials on to the Fe substrate at spacer thickness. The SEMPA images show two components of the temperatures indicated. The dashed lines show the abrupt the magnetization, MV and MW, where white (black) indicates change to rough or 3-d growth. magnetization in the positve (negative) x- or y-direction along and across the whisker, respectively. RHEED, both in time-resolved and spatially-resolved modes, was used to monitor the quality of the "lm growth and to measure the thickness of the "lms. Fig. 1 shows a RHEED image of the Au wedge before coating with Fe. 3. Results The growth of the various interlayer materials on the Fe whisker substrates is summarized by the thickness- Fig. 3. Polarization line scans from SEMPA images showing dependent RHEED scans shown in Fig. 2. The "lms were oscillations of MV in the top Fe layer as a function of the spacer layer thickness. grown at the temperatures indicated. Cr, Ag and Au grow nearly layer-by-layer and provide the best measurements of oscillatory exchange coupling. The exchange coupling for these spacer layers along with Mn and V is sum- This high-temperature growth can lead to alloying at the marized by the MV line scans shown in Fig. 3. The Cr/Fe interface [13,14], but growing the "rst two Cr oscillatory exchange coupling periods derived from these layers at a lower temperature of about 1003C minimizes measurements are summarized and compared to theory the interfacial alloying and increases the magnitude of [9}12] in Table 1. Mn, V, Al and Cu all grow following the Fe/Cr/Fe exchange coupling strength [15]. SEMPA di!erent variations of Stranski}Krastanov growth measurements of Fe/Cr/Fe coupling on Fe whiskers, modes; growing layer-by-layer initially and then switch- such as the one shown in Fig. 3, show nearly layer-by- ing to rough or three-dimensional growth at some critical layer switching between ferromagnetic and antiferromag- thickness. netic alignment for almost 80 layers of Cr [8,16]. Closer inspection of the SEMPA data actually reveals that two 3.1. Cr coupling periods are present; there is a strong short period nearly commensurate with the lattice and a longer Although Fe/Cr/Fe was the "rst multilayer in which period. Which of these periods is dominant depends oscillatory exchange coupling was observed, it continues sensitively on the thickness #uctuations in the Cr spacer to be an interesting system due to the rich variety of layer as has been veri"ed by STM measurements [17]. possible Cr magnetic states. The highest quality Cr "lms SEMPA measurements of only bare Cr on the Fe are grown at elevated temperatures of about 300}3503C. whisker also reveal the same short period oscillations 398 J. Unguris et al. / Journal of Magnetism and Magnetic Materials 198}199 (1999) 396}401 Table 1 The measured periods for coupling in the [0 0 1] direction for Cr, Ag, Au, and V are compared to theory. Except for V where the growth is poor, there is excellent agreement with the predictions of theory Interlayer Measured Theory Stiles [9,10] Bruno and Van Schilfgaarde Chappert [11] and Herman [12] Cr (0.144 nm/layer) 2.105$0.005 2.10 2.15 12$1 11.1 12.3 Ag (0.204 nm/layer) 2.37$0.07 2.45 2.38 5.73$0.05 6.08 5.58 Au (0.204 nm/layer) 2.48$0.05 2.50 2.51 8.6$0.3 9.36 8.60 V (0.152 nm/layer) &5 3.08 11.0 pends on the Cr thickness. These measurements clearly show the incommensurate spin density wave nature of the Cr with the expected minimum in the coupling strength at the position of the phase slip at 24 layers [18,19]. This "gure also shows how di$cult measuring coupling strengths of well-ordered Fe/Cr/Fe structures one thickness at a time can be: a change in the Cr thickness of only a tenth of a monolayer can lead to over an order of magnitude change in the exchange coupling strength. The incommensurate short-period oscillations along with the related phase slips are also very sensitive to temperature. Previous SEMPA measurements [16] of the bare Cr polarization show that the short-period oscillations exist to at least 1.8 times the bulk NeHel temperature, ¹,"311 K. Fig. 5 shows a series of Fig. 4. A series of MOKE images from an Au (10 ML)/ SEMPA measurements of the Fe/Cr/Fe coupling as Fe(15 ML)/Cr wedge/Fe whisker sample taken at various ap- a function of temperature. Fig. 5 also includes curves plied magnetic "elds, showing the "eld and Cr thickness depen- showing the thickness at which the phase slips in bare dence of the reversal of the antiferromagnetic regions (dark Cr/Fe occur. Like the Cr/Fe case, the short-period oscil- bands). The Fe/Cr/Fe exchange coupling strength is determined lations in Fe/Cr/Fe exist to nearly twice the bulk ¹ from the switching "eld. A SEMPA image of the same wedge at , and the phase slips have nearly the same temperature depen- zero applied "eld is shown at the bottom for reference. dence. Locating the phase slip is somewhat di$cult in Fe/Cr/Fe because the short-period coupling strength ap- pears to drop o! more rapidly with temperature than [16]. For both bare Cr and Fe coated Cr, the incommen- that of the long period. (Only the short-period oscilla- surate short-period oscillations result in phase slips in the tions are visible in the bare Cr/Fe data.) Note that all of coupling at about 24, 44 and 64 layers at room temper- these observations are completely reversible and are not ature owing to the slight di!erence between the lattice due to an irreversible roughening of the multilayer. The wave vector and Fermi surface spanning vector. temperature dependence of the coupling leads to rever- Further evidence for the incommensurate nature of the sals in the direction of the coupling i.e., below 420 K the short-period coupling is also seen in the magneto-optic coupling at 30 layers is antiferromagnetic, while above Kerr e!ect (MOKE) measurements shown in Fig. 4. The 500 K the coupling switches to ferromagnetic. "gure consist of a series of MOKE images of the Our measurements are qualitatively similar to neutron Fe/Cr/Fe (1 0 0) wedge taken at various applied magnetic scattering measurements which "nd a commensurate "elds. This series of images graphically shows how the spin density wave below and incommensurate spin den- strength of the antiferromagnetic exchange coupling de- sity wave above the "rst phase slip [20,21]. However, the J. Unguris et al. / Journal of Magnetism and Magnetic Materials 198}199 (1999) 396}401 399 3.3. V Vanadium grows in a Stranski}Krastanov mode with the "rst six layers growing layer by layer and then ab- ruptly switching to rougher island growth for thicker "lms. The same growth behavior with roughening start- ing after exactly six layers was observed for various #ux rates and for substrate temperatures ranging from 303 to 3003C. SEMPA measurements of an Fe/V/Fe wedge are shown in Fig. 6. For up to 20 layers the exchange coup- ling oscillates, but the growth is su$ciently rough to "lter out any potential short-period oscillations. After 20 layers no coupling oscillations are observed. From this limited data it is di$cult to derive precise oscillatory coupling periods. An average period of about 5$1 Fig. 5. The temperature dependence of the bilinear (M layers may be estimated from the measurements. This V) ex- change coupling in Fe/Cr/Fe. The phase slips measured on bare value is di!erent from calculated periods [12], but it is Cr are shown by the solid gray line; the dashed line is the about the same as the 6 layer period observed by Parkin estimated position of the next phase slip. Note that the short- [23] for sputtered V/Co multilayers. period oscillations, where visible, have opposite direction at temperatures below and above these lines. 3.4. Mn boundary dividing the two regions of spin density wave Although Mn is notable for its many di!erent struc- behavior saturates at about 300 K in the neutron tural and magnetic phases, it appears to grow on Fe measurements but extends to well over 500 K for our whiskers as a simple antiferromagnet with a bct structure samples as seen in Fig. 5. The di!erence probably can be closely matched to the substrate [24]. RHEED measure- attributed to di!erences in crystalline order, lattice con- ments show initial well-ordered layer-by-layer growth in stants, or lattice strains in samples prepared in di!erent the "rst few layers that gradually becomes rougher and ways and on di!erent substrates. "nally, abruptly switches to 3-d growth after 15}20 layers. SEMPA measurements of the bare Mn wedge 3.2. Au and Ag show an oscillating surface moment that is collinear with the Fe moment and consistent with antiferromagnetic Although FCC Ag and Au have very di!erent unit cells Mn. While the bare Mn polarization is similar to that than BCC Fe, there is little lateral mismatch between the of Cr, the Fe/Mn/Fe exchange coupling is completely BCC substrate and the FCC overlayer when rotated by di!erent. SEMPA measurements of the Fe/Mn/Fe 453. A large vertical mismatch remains, however, which wedge, Fig. 7, show that the coupling is oscillatory, but can be a problem at substrate Fe steps. Both Ag and Au not collinear. After an initial transient phase for thin Mn grow nearly layer by layer up to &50 layers, although "lms the coupling oscillates with a 2 layer period about each system has its own peculiarities: Au grows with 903 with respect to the Fe whisker magnetization. The a "ve-fold reconstruction of the top layer [2], while Ag magnitude of the oscillations varies for di!erent Fe seems to require an initial incubation period of between whisker substrates but is generally between 10 and 303, 2 and 6 layers before growing layer by layer [1]. SEMPA measurements of Ag and Au each reveal two coupling periods (see Table 1). The periods arise from the &belly' and &neck' extremal spanning vectors of the Fermi Surface of these noble metals. As can be seen in Fig. 3 the short period dominates the coupling in Au, while the longer period is strongest in Ag. The magnitude of the exchange coupling strength for Fe/Au/Fe whiskers has been measured by MOKE and found to be much larger than earlier measurements using non-whisker substrates. In fact, when the residual thickness #uctuations of the Au Fig. 6. The RHEED intensity imge shows that V grows nearly spacer layer are taken into account, the measured coup- layer-by-layer for 6 layers. The SEMPA images show exchange ling strengths [3] are very close to the theoretically coupling oscillations with a period of approximately 5 layers up predicted values [22]. to about 20 layers. 400 J. Unguris et al. / Journal of Magnetism and Magnetic Materials 198}199 (1999) 396}401 i.e. for the 103 amplitude the coupling oscillates between 80 and 1003 [25]. 3.5. Al Aluminum is as well lattice matched to Fe (1 0 0) as are Ag or Au so that one might expect similar smooth, layer-by-layer growth, but instead Al forms rough three- dimensional "lms on Fe whisker substrates, and the exchange coupling is altered dramatically. Fig. 8 shows SEMPA and RHEED measurements for Al grown on an Fe whisker. For less than two Al layers, RHEED patterns Fig. 8. RHEED images showing an arc of sharp spots for clean indicate that the growth is smooth and epitaxial, and Fe and 1 layer of Al. By three layers of Al a 3-d transmission SEMPA images show that the Fe/Al/Fe is coupled fer- pattern is observed and the top Fe layer couples at 903 to the Fe romagnetically. For Al coverages greater than 3 layers, whisker magnetization. the RHEED patterns switch from sharp spots along an arc to a lattice of spots characteristic of transmission through three-dimensional Al structures. The dramatic roughening at such low Al coverages may indicate that the roughness is not just due to Al, but the Fe substrate may also become rougher as a result of the tendency of Al and Fe to form compounds. SEMPA images show that the coupling switches dramatically from ferromagnetic to a 903 alignment when the "lm becomes rough. Unlike earlier measurements by Fuss et al. [26], no antifer- romagnetic coupling is observed, but the 903 biquadratic coupling agrees with the observations of Filipkowski et al. [27]. Neither of these other measurements used single- crystal Fe substrates however. 3.6. Cu Copper does not lattice match well to the Fe (1 0 0) Fig. 9. The SEM image of Cu deposited on a Fe whisker substrate and achieving layer-by-layer growth of Cu on showing the formation of 3-d Cu crystallites at nucleation sites. an Fe whisker is consequently very di$cult. We found that one or perhaps two layers of two-dimensional Cu could be grown epitaxially on the Fe (1 0 0) surface. could "nd nucleation sites and grow into crystallites. Subsequent layers were not stable and formed three- Fig. 9 shows SEM images from an uncleaned part of an dimensional bulk Cu crystallites at nucleation sites along Fe whisker in which several of these crystallites formed. the edge of the whisker or in regions that were not sputter No Cu crystallites were observed on the cleaned surface cleaned. A surprising feature of this rough growth was of the Fe whisker. Although the extreme mobility of the the extreme mobility of the Cu deposited on 1}2 ML of Cu on the Fe whisker is interesting, it prevented us from Cu. Even at room temperature Cu di!used over relatively making any meaningful exchange coupling measure- large distances, on the order of tenths of a mm, so that it ments on this system. 4. Conclusions Model epitaxial trilayer systems with Fe (0 0 1) whisker substrates were prepared to test theories of inter- layer exchange coupling. Although structures with V, Mn, Al, and Cu exhibited some interesting properties, the Fig. 7. The magnetization of the top Fe layer in Fe/Mn/Fe is quality of the growth was inadequate for a comparison of oriented predominantly at right angles to the whisker magneti- the observed coupling to theory. The Fe/Cr/Fe system zation. The contrast in the M continues to provide fascinating results. The strength of V image indicates a small oscilla- tion of the magnetization about 903 with a period of 2 layers. the coupling displays a minimum at the phase slip at 24 J. Unguris et al. / Journal of Magnetism and Magnetic Materials 198}199 (1999) 396}401 401 layers where a node is introduced into the spin density [7] J.A. Stroscio, D.T. Pierce, J. Vac. Sci. Technol. B 12 (1994) wave. Further, this phase slip boundary extends to tem- 1783. peratures nearly twice the bulk NeHel temperature in con- [8] D.T. Pierce, J. Unguris, R.J. Celotta, in: B. Heinrich, J.A.C. trast to neutron scattering measurements on samples Bland (Eds.), Ultrathin Magnetic Structures II, Springer, with less perfect interfaces. The excellent agreement ob- Berlin, 1994, p. 117. tained between experiment and theory for the periods of [9] M.D. Stiles, Phys. Rev. B 48 (1993) 7238. the exchange coupling oscillations for Au, Ag, and Cr, [10] M.D. Stiles, Phys. Rev. B 54 (1996) 14 679. [11] P. Bruno, C. Chappert, Phys. Rev. Lett. 67 (1991) and in the case of Au for the strength of the coupling as 1602. well, show that the theory of interlayer exchange coup- [12] M. van Schilfgaarde, F. Herman, Phys. Rev. Lett. 71 (1993) ling is now on sound footing. 1923. [13] D. Venus, B. Heinrich, Phys. Rev. B 53 (1996) R1733. [14] A. Davies, J.A. Stroscio, D.T. Pierce, R.J. Celotta, Phys. Acknowledgements Rev. Lett. 76 (1996) 4175. [15] B. Heinrich, J.F. Cochran, D. Venus, K. Totland, C. We wish to thank Bret Heinrich and coworkers for Schneider, K. Myrtle, J. Magn. Magn. Mater. 156 (1996) initially supplying us with Fe whiskers and then teaching 215. us how to grow them ourselves. This work was supported [16] J. Unguris, R.J. Celotta, D.T. Pierce, Phys. Rev. Lett. 69 in part by the O$ce of Naval Research. (1992) 1125. [17] D.T. Pierce, J.A. Stroscio, J. Unguris, R.J. Celotta, Phys. Rev. B 49 (1994) 14 564. References [18] Y. Wang, P.M. Levy, J.L. Fry, Phys. Rev. Lett. 65 (1990) 2732. [1] J. Unguris, R.J. Celotta, D.T. Pierce, J. Magn. Magn. [19] Z.P. Shi, R.S. Fishman, Phys. Rev. Lett. 78 (1997) Mater. 127 (1993) 205. 1351. [2] J. Unguris, R.J. Celotta, D.T. Pierce, J. Appl. Phys. 75 [20] E.E. Fullerton, S.D. Bader, Phys. Rev. Lett. 77 (1996) (1994) 6437. 1382. [3] J. Unguris, R.J. Celotta, D.T. Pierce, Phys. Rev. Lett. 79 [21] A. Schreyer et al., Phys. Rev. Lett. 79 (1997) 4914. (1997) 2734. [22] M.D. Stiles, J. Appl. Phys. 79 (1996) 5805. [4] J. Unguris, R.J. Celotta, D.T. Pierce, Phys. Rev. Lett. 67 [23] S.S.P. Parkin, Phys. Rev. Lett. 67 (1991) 3598. (1991) 140. [24] S.K. Kim et al., Phys. Rev. B 54 (1996) 5081. [5] S.S. Brenner, in: J.J. Gilman (Ed.), The Art and Science of [25] D.A. Tulchinsky, J. Unguris, R.J. Celotta, in preparation. Growing Crystals, Wiley, New York, 1963, p. 30. [26] A. Fuss, S. Demokritov, P. GruKnberg, W. Zinn, J. Magn. [6] A.S. Arrott, B. Heinrich, S.T. Purcell, in: M.G. Lagally Magn. Mater. 103 (1992) L221. (Ed.), Kinetics of Ordering and Growth at Surfaces, [27] M.E. Filipkowski, C.J. Gutierrez, J.J. Krebs, G.A. Prinz, Plenum, New York, 1990, p. 321. J. Appl. Phys. 73 (1993) 5963.