VOLUME 83, NUMBER 6 P H Y S I C A L R E V I E W L E T T E R S 9 AUGUST 1999 Magnetic Domain Formation in Fe Films on Cr(100) H. Hopster Department of Physics and Astronomy and Institute for Surface and Interface Science, University of California, Irvine, California 92697 (Received 29 December 1998) Magnetic domain images of thin (2 nm) Fe films on Cr(100) are presented. Upon cooling a single- domain state transforms into a state with a locally varying in-plane magnetization direction. This transformation is partially reversible and highly reproducible. It is suggested that the in-plane rotation is driven by frustration that favors 90± coupling, and that the local magnetization direction reflects mm-scale variations of the atomic scale roughness. PACS numbers: 75.70.Ak, 75.50.Ee, 75.70.Cn, 75.70.Kw There is currently a great interest in ferromagnetic using secondary electron microscopy with polarization antiferromagnetic FM AF interfaces. "Exchange bias- analysis (SEMPA) [9]. The rectangular Cr(100) crystal ing" is becoming increasingly important in novel mag- and the cleaning and growth conditions are the same netic devices. In exchange biased films the hysteresis as the ones used in Ref. [5]. However, the Cr crystal loop due to the coupling to the AF film is shifted from has been reoriented and repolished [10]. The electron zero field as if an external field was present. Although gun for SEMPA measurements consists of a Shottky discovered more than forty years ago the effect is not well field emission tip, single-lens focusing, and an octupole understood on an atomic scale [1]. In addition, short- deflection unit [11]. For sample alignment the SEM period oscillations in the exchange coupling between Fe can be operated in the usual TV mode. The low- films through Cr in very flat samples [2] have kindled energy secondary electrons are extracted into a medium- interest in the magnetic structure of Cr layers since the energy Mott detector [12]. Two components of the spin coupling is believed to be closely tied to the AF structure polarization can be measured simultaneously. Figure 1 of the Cr films. The magnetic order in thin Cr layers in shows the sample and measurement geometry. The relation to the well-known incommensurate spin density incoming electron beam is about 35± off normal and wave ordering of bulk Cr [3] is actively being investi- the extraction lens into the Mott detector about 10± off gated [4]. normal so that one measures basically the two in-plane The present study was motivated by a magneto-optical magnetization components (Px and Py), as indicated. In Kerr effect investigation of Fe films on Cr(100) [5], which addition, an intensity image can be obtained by adding showed unusual temperature dependent magnetic proper- the four channels from the Mott detector. In the intensity ties. Fe films thinner than 5 nm show anomalous tem- images the samples show no contrast except for a few perature dependencies of the coercive field and the "spots," as the one shown in the image in Fig. 1. These remanent magnetization decreases below the Cr Néel spots were used to position the sample reproducibly. All temperature (311 K). The effects were attributed to images shown are 256 3 256 pixels and about 85 mm 3 frustrated magnetization structures. Frustration is a 70 mm (due to the off-normal alignment of gun and common phenomenon when AF interactions and steps sample). The primary electron beam energy is 5 keV and are present simultaneously, and it plays a crucial role in the current is a few nA. The lateral resolution is about most theoretical models of AF FM interface structures 500 nm. Count rates in the Mott detector are typically [6]. A recent example is "biquadratic" (or 90±) coupling 5 3 104 counts sec per channel. The Fe films can be between FM layers [7]. Steps were recently shown to magnetized by a field pulse of 50 Oe in the x direction, induce the Fe magnetization on stepped Cr surfaces to which coincides with an easy axis. The Fe films show switch between parallel and perpendicular to the steps as a high secondary electron polarization of about 40%. a function of step density [8]. The sample stage can be cooled to about 2100 ±C and This paper gives a microscopic picture of the tempera- heated to about 80 ±C, which is well above the Cr Néel ture dependent magnetization of an ultrathin FM film temperature. (2 nm Fe) on a nominally flat AF substrate (Cr). A single- Figure 2 contains the principal observations of this domain state evolves into a state in which the magneti- paper. It shows the two in-plane polarization components zation direction varies on the micron scale when cooled starting in the top row with a one-domain state at RT below TN. after the film was magnetized above TN. There is a The experiments were performed in a new UHV uniform high x-polarization (left column) except for a spot system that combines molecular-beam epitaxy growth and with zero polarization in the center of the image. Upon characterization capabilities with magnetic microscopy cooling, a domain pattern as shown in the second row 0031-9007 99 83(6) 1227(4)$15.00 © 1999 The American Physical Society 1227 VOLUME 83, NUMBER 6 P H Y S I C A L R E V I E W L E T T E R S 9 AUGUST 1999 FIG. 1. Geometry of the experiment and sample orientation with respect to the polarization measurement. The two in-plane components (x and y) of the magnetization are measured. An external magnetic field can be applied in the x direction. (LT) evolves which was taken at the final temperature of 2100 ±C. This domain pattern with the emergence of a perpendicular component Py and a corresponding reduction in Px is, of course, a microscopic picture of the observed magnetization reduction reported in Ref. [5]. A pixel-by-pixel analysis shows that the absolute polarization p values ( P2x 1 P2y) stay constant at about 40%. This proves that the magnetization stays in plane (no "missing" out-of-plane component) and only rotates in plane as FIG. 2 (color). The evolution of the magnetic domain state as the temperature is changed. The images show on the left a function of temperature while the local magnetization (yellow-blue contrast) the Px and on the right the Py component value is temperature independent. It also proves that (red-green). All images are taken on the same spot, except there are no significant magnetic structures smaller than for a small thermal drift of the sample stage. The images are the beam diameter (500 nm) since averaging over those 85 mm 3 70 mm. Top row: One-domain state at RT; second structures would lead to a reduction of the polarization row: cooled down to low temperature 2100 ±C . Third row: After warming above T values. The magnetization structures are quite irregular N; fourth row: cooled down again to LT. and range from a few mm to some tens of mm. When the film warms to above the Néel temperature the domain rection. A four-pixel average was applied in order to im- pattern changes to the one shown in the third row (RT). prove statistics. The single-domain state shows a fully The images become sharper. The domains boundaries aligned magnetization. The width of the distribution is become smoother and most of the small-scale structures solely due to counting statistics. The LT state, on the other have coalesced into bigger domains. The change between hand, shows a broad asymmetric distribution of angles the LT and RT state is reversible. When cooled again the with a maximum around 25± and the average at about domain images in the lower row of Fig. 2 result in almost 45±. After warming to above TN the maxima at 0± and all detail identical to the second row. at 90± correspond to easy axes of the bcc Fe films. Thus, A survey of several spots on the sample showed that the above TN the magnetization direction is governed by crys- magnetization behavior described above is indeed typical. talline anisotropies. By comparing the RT and LT images A quantitative evaluation of the images reveals more de- in Fig. 2 it is clear that the areas that have a strong perpen- tails about the magnetization state. From the measured Px dicular component at LT are most likely to "flip over" into and Py values one can determine the local magnetization the 90± direction upon warming to RT. angle for every pixel. Figure 3 shows the distribution of Not only is the temperature cycling between LT and RT angles with respect to the macroscopic magnetization di- reversible but even the domain patterns that result after 1228 VOLUME 83, NUMBER 6 P H Y S I C A L R E V I E W L E T T E R S 9 AUGUST 1999 cently in very different systems: Fe3O4 CoO superlattices [14] and Fe films on FeF2 [15]. In order to demonstrate the roughness magnetization connection a little better, Fig. 4 shows the polariza- tion and the corresponding intensity image (digitally en- hanced) taken very close to the sample edge, where one can actually see scratches in the SEM image as well as under an optical microscope. A correlation between mag- netization and topography is obvious in a very general, qualitative way. On this "rough terrain" hardly any hori- zontal Px magnetization is left over, i.e., everything has turned almost by a full 690± with respect to the original magnetization direction. Note that it is not the direction of the scratches that determines the magnetization direction but that the magnetization is perpendicular to the original magnetization direction, i.e., dominated by the Cr inter- face moments. Also, during cooling the rough regions start switching sooner than the flat areas consistent with a roughness driven mechanism. Some slight preferential orientation of the roughness, e.g., step edges, probably is also the cause of the preferential angle in which the mag- netization turns at low temperature. In Fig. 2 almost ev- erything turns towards 190± and very little towards 290±. FIG. 3. Histogram of the angular distribution of the local This preference seems to vary locally over the sample. magnetization with respect to the axis of the macroscopic magnetization. The images from Fig. 2 were used after a four- Finally, because of the similarity of the images I want to pixel average to improve statistics. mention a study of the exchange coupling in Co Cr Co trilayers by TEM during magnetization reversal [16]. It is shown that some areas start switching consistently sooner remagnetization (above TN) and cooling again are highly than others. This is attributed to a spatially varying ex- reproducible. Reversing the magnetization direction, as change coupling due to thickness fluctuations. well as rotating the sample 90± and magnetizing in that direction, shows that it is always the same areas that turn away from the macroscopic magnetization direction, independent of the initial magnetization direction. What drives the magnetization reorientation? It seems unlikely that the images reflect some aspect of the mag- netic structure of the Cr, e.g., an antiferromagnetic domain structure, because of the reproducibility of the structures even after heating above the Cr ordering temperature. It can also be safely excluded that the domains reflect the atomic terrace structure of the Cr substrate. Atomic terrace widths on well-prepared single crystal metal surfaces are typically not more than a few hundred Å wide [13]. Thus, the magnetic structures are a hundred times larger than the terraces. The mechanism that then springs to mind is a roughness driven transition due to frustration, very simi- lar to the mechanism that leads to biquadratic coupling be- tween FM layers due to interface roughness. Since the Fe magnetization cannot follow the Cr terrace structure, because it would cost too much energy in domain walls, it tends to turn perpendicular to the Cr surface moments. The Cr surface magnetization was induced by the single- FIG. 4 (color). Domain Px and Py image in the top row domain Fe film and was frozen in when the sample was taken at RT (same parameters as in Fig. 2); below: SEM cooled through T intensity image derived from the Mott detector counts, with N. Thus the images show the biquadratic the contrast greatly enhanced. This image was obtained on a coupling already at the single AF FM interface. Evidence rough area very close to the sample edge, showing the general for perpendicular spin orientations has been reported re- correlation between roughness and magnetic structure. 1229 VOLUME 83, NUMBER 6 P H Y S I C A L R E V I E W L E T T E R S 9 AUGUST 1999 In summary, it was shown that upon cooling one- Rev. Lett. 81, 914 (1998); S. Demuynck, J. Meersschaut, domain 2-nm Fe films on nominally flat Cr(100) surfaces J. Dekoster, B. Swinnen, R. Moons, A. Vantomme, form a nonuniform magnetization state with a spatially S. Cottenier, and M. Rots, Phys. Rev. Lett. 81, 2562 varying 90± (in-plane) magnetization component. The do- (1998). main evolution is partially reversible and highly repro- [5] A. Berger and H. Hopster, Phys. Rev. Lett. 73, 193 (1994). ducible. It is suggested that atomic scale roughness is the [6] J. C. Slonczewski, Phys. Rev. Lett. 67, 3172 (1991); N. C. driving force for the magnetization reorientation, and that Koon, Phys. Rev. Lett. 78, 4865 (1997). the magnetization direction reflects the average local step [7] M. Rührig, R. Schäfer, A. Hubert, R. Mosler, J. A. Wolf, S. Demokritov, and P. Grünberg, Phys. Status Solidi A density. When warmed to above the Cr Néel temperature 125, 635 (1991); M. E. Filipkowski, J. J. Krebs, G. A. the Fe magnetization relaxes into the easy magnetization Prinz, and C. J. Gutierrez, Phys. Rev. Lett. 75, 1847 directions. Evidence for the correlation between rough- (1995); A. Azevedo, C. Chessman, S. M. Rezende, F. M. ness and magnetization direction was presented. The de Aguiar, X. Bian, and S. S. P. Parkin, Phys. Rev. Lett. origin of the 90± alignment is suggested to be frustration 76, 4837 (1996); J. Schwabenhausen, T. Durkop, and due to interface roughness similar to the mechanism in bi- H.-J. Elmers, Phys. Rev. B 15, 15 119 (1997). quadratric coupling in FM AF FM structures. [8] E. J. Escorcia-Aparicio, H. J. Choi, W. L. Ling, R. K. I thank D. L. Mills for discussions and a critical reading Kawakami, and Z. Q. Qiu, Phys. Rev. Lett. 81, 2144 of the manuscript. I would also like to thank Y. Iwasaki, (1998). J. Barthel, J. Kondis, B. P. Miller, and E. B. Maiken [9] H. Hopster, Y. Iwasaki, J. Barthel, E. B. Maiken, B. P. for their contributions in the early stages of the design Miller, J. Kondis, and Y. Yu, Mater. Res. Soc. Symp. Proc. 475, 569 (1997). and construction of the experiment. The experimental [10] The Cr(100) single crystal sample was commercially setup was made possible to a large extent by a financial oriented and polished to within 0.1± (0.25± guaranteed) donation from the Sony Research Center, Yokohama. I by Mateck, Jülich. also thank the FEI Corporation for donating the electron [11] FEI Company, Hillsboro, Oregon. column. The early stages of the experiments were also [12] D. P. Pappas, Ph.D. thesis, University of California, Irvine, supported by the NSF through Grant No. DMR 9500213. 1990. My special thanks go to our machine shop personnel at [13] At present there is no direct information available on the UCI for their excellent work. atomic terrace widths on this sample. The smoothest metal substrates available seem to be whiskers (see, e.g., Ref. [2]). Here a terrace width can reach the micron scale. [1] W. H. Meiklejohn and C. P. Bean, Phys. Rev. 102, 1413 STM experiments will be combined with SEMPA in the (1956). future. [2] J. Unguris, R. J. Celotta, and D. T. Pierce, Phys. Rev. Lett. [14] Y. Ijiri, J. A. Borchers, R. W. Erwin, S.-H. Lee, P. J. van 67, 140 (1991). der Zaag, and R. M. Wolf, Phys. Rev. Lett. 80, 608 (1998). [3] E. Fawcett, Rev. Mod. Phys. 60, 209 (1988). [15] T. J. Moran, J. Nogues, D. Lederman, and I. K. Schuller, [4] E. E. Fullerton, S. D. Bader, and J. L. Robertson, Phys. Appl. Phys. Lett. 72, 617 (1998). Rev. Lett. 77, 1382 (1996); P. Bodeker, A. Hucht, [16] A. C. Daykin, J. P. Jakubovics, and A. K. Petford-Long, A. Schreyer, J. Borchers, F. Guthoff, and H. Zabel, Phys. J. Appl. Phys. 82, 2447 (1997). 1230