PHYSICAL REVIEW B VOLUME 60, NUMBER 10 1 SEPTEMBER 1999-II Magnetization reversal and spin reorientation in Fe/Cu 100... ultrathin films E. Mentz, A. Bauer, T. Gušnther, and G. Kaindl Institut fušr Experimentalphysik, Freie Universitašt Berlin, Arnimallee 14, D-14195 Berlin, Germany Received 19 September 1997; revised manuscript received 30 September 1998 The magnetization reversal in perpendicular applied magnetic fields in low-temperature grown Fe/Cu 100 ultrathin films was studied in situ by magneto-optical Kerr effect, Kerr-microscopy, and scanning tunneling microscopy. It was found that the onset of long-range ferromagnetic order at 0.9 monolayers ML is related to the coalescence of bilayer islands of Fe. Below 3.8 ML, the magnetization-reversal process takes place in a narrow field range and is characterized by domain-wall motion. While the domain boundaries are rather smooth at thin films, domains get increasingly irregularly shaped above 3 ML, which is assigned to a decrease of the domain-wall energy. At the spin-reorientation transition from out-of-plane to in-plane magnetization at 3.8 ML, two coexisting metastable spin configurations are found. S0163-1829 99 10633-7 I. INTRODUCTION Kerr effect MOKE , is well suited for measurements in ex- ternal magnetic fields. At present, however, there are only a The reversal of magnetization in external magnetic fields few UHV Kerr microscopes in operation.10,11 in ultrathin magnetic films has attracted considerable interest In this article, we report on a combined study of low- recently.1­3 This is mainly due to the potential novel appli- temperature LT grown ultrathin films of Fe/Cu 100 by in cations in magnetic-storage and sensor technology, but also situ Kerr microscopy, MOKE, and scanning tunneling mi- due to an interest in the complex process itself. The croscopy STM . The conclusions reached for this particular magnetization-reversal process is basically governed by do- system concerning magnetization reversal in films with per- main nucleation in the vicinity of defects, steps, etc., with pendicular magnetic anisotropy and spin reorientation are ex- subsequent domain-wall motion, which is again strongly af- pected to be representative for several other systems. It is fected by film morphology and defects. Of particular interest shown that the onset of long-range ferromagnetic order in is the magnetization-reversal process at or close to a spin- Fe/Cu 100 is related to the coalescence of bilayer islands at reorientation transition: In many systems, a transition from 0.9 ML a similar effect was previously found for an out-of-plane to an in-plane easy axis of magnetization Fe/W 110 .12 For these very thin films, the shape of the mag- occurs at a specific film thickness typically a few monolay- netic hysteresis loop deviates from the squarelike shape ob- ers and/or a specific temperature. At the transition, the mag- served for slightly thicker films, which is assigned to inho- netic anisotropies are rather small, which makes micro- mogeneous film morphology and anisotropy. Domain domain states e.g., stripe domains energetically nucleation and domain-wall motion during magnetization re- favorable.4­7 However, states with uniform magnetization versal in out-of-plane magnetized thin films could be imaged out-of-plane, in-plane, or canted can be stabilized in exter- for films as thin as 1.9 ML. The observed strong influence of nal magnetic fields and might be metastable in zero field film thickness on coercive field and magnetic-domain shape giving rise to magnetic hysteresis.6­8 is explained by thickness dependent magnetic anisotropy. Of Magnetization reversal in ultrathin films is strongly af- particular interest is the coexistence of out-of-plane magne- fected by morphology on the nanometer scale,9 which makes tized domains and domains with a reduced or even vanish- it imperative to closely control growth and morphology of ing out-of-plane magnetization component at the spin- these films. Since a protective layer on top of an ultrathin reorientation thickness of 3.8 ML. The latter may either magnetic film can have a strong influence on structure, mor- have a uniform canted magnetization or an inner micro- phology, and magnetism of the film, it is desirable to per- domain structure that is not resolved by the Kerr microscope. form in situ measurements in ultrahigh vacuum UHV . Con- siderable insight into the magnetization-reversal process can II. EXPERIMENTAL DETAILS be obtained from magnetization curves magnetic hysteresis loops , even though a knowledge of average magnetizations The Fe/Cu 100 samples were prepared in a UHV system is often not sufficient to unambiguously interpret the reversal base pressure: 3 10 11 mbars by electron-beam evapo- process. In this respect, a uniform canted magnetization can- ration of Fe onto Cu 100 at 100 K. The deposition rate 1 not be distinguished from a domain state with the same av- Ć/min was measured by a quartz microbalance with an ab- erage magnetization. For an improved understanding, it is solute error of 20%, and a relative reproducibility of 5%. therefore necessary to laterally resolve the film magnetiza- Before deposition, Cu 100 was cleaned by more than 50 tion. Most of the available UHV-compatible domain-imaging sputter-anneal cycles until a very sharp low-energy electron techniques based on scanning electron microscopy or mag- diffraction LEED pattern was obtained as well as wide and netic force microscopy are restricted to measurements at contamination-free Cu terraces were seen in the STM im- zero or small external magnetic field. On the other hand, ages. Kerr microscopy, which is based on the magneto-optical The as-prepared samples were either transferred to a 0163-1829/99/60 10 /7379 6 /$15.00 PRB 60 7379 ©1999 The American Physical Society 7380 E. MENTZ, A. BAUER, T. GUšNTHER, AND G. KAINDL PRB 60 custom-built UHV-STM operating at room temperature RT or into an interconnected UHV chamber specifically de- signed for in situ MOKE and Kerr-microscopy measure- ments. For Kerr microscopy, the sample and a UHV electro- magnet were moved close to a 2 34 view port, realizing the optimal working distance 10 cm of the employed optical microscope Questar QM100 . The sample was illuminated by a Hg-discharge lamp, and a sheet polarizer and analyzer were placed in the incident and reflected beams, respectively. Images were recorded with a charge-coupled device CCD camera connected to an image-processing system that allows background subtraction and frame integration in order to eliminate topographical structures and enhance the signal-to- noise ratio, respectively. Only static domain images were recorded. To record domain images, the magnetization- reversal process was interrupted at certain points by slightly reducing the applied magnetic field. The lateral resolution of the Kerr microscope is about 3 m. The MOKE and Kerr- microscopy measurements were performed in polar geom- etry, with the sample at 130 K, magnetized along the surface normal, and illuminated at a small angle 10° with respect to the surface normal. A detailed description of the Kerr microscope and the MOKE setup will be published elsewhere.13 III. RESULTS AND DISCUSSION The system Fe/Cu 100 is known to exhibit strong rela- tions between atomic structure and magnetic properties see, e.g., Ref. 14 and references therein . For the first few mono- layers ML , which is the thickness range that is studied in the present work, Fe grows pseudomorphically on Cu 100 with a tetragonally distorted fcc structure fct .15 In this thickness range the easy axis of magnetization is perpendicu- lar to the surface out-of-plane magnetization . We have grown the films at LT in order to obtain abrupt interfaces. RT-grown Fe/Cu 100 films exhibit strong Fe-Cu interdiffu- sion at the interface, which is expected to be suppressed at LT.16 For the as-grown films, a spin-reorientation from out- FIG. 1. a and b : 400 400 Ć STM images of a 0.6-ML of-plane to in-plane magnetization is observed at 3.8 ML, and b 0.9-ML LT-grown Fe/Cu 100 exhibiting bilayer islands which shifts to higher thickness up to 6 ML for annealed before and after coalescence. In b , the nucleation of some 3-ML- films.17 high patches is observed. c Polar-MOKE hysteresis loops taken at 130 K of 0.9-ML Fe/Cu 100 annealed at A 130 K and B 300 K. The MOKE intensity is normalized to the respective lower signal; H A. Onset of ferromagnetic order is the external magnetic field. Figure 1 a shows an STM image of LT-grown 0.6-ML Fe/Cu 100 , which is characterized by isolated bilayer is- estingly, loop B deviates considerably from a squarelike lands with diameters ranging from 10 Ć to 50 Ć; no shape observed for thicker films see below , which could MOKE signal could be detected from this film. At 0.9 ML, well be caused by a locally varying degree of coalescence, the bilayer islands coalesce see Fig. 1 b , and a polar- i.e., a distribution of local coercive fields. Unfortunately, the MOKE signal is recorded, indicating long-range magnetic magnetic contrast was too small to allow the recording of order hysteresis loops A and B in Fig. 1 c . Hysteresis loop domain images. However, it is clear that the networklike film A was taken before the sample was warmed up to RT for the morphology, with structures on a nanometer scale, strongly STM measurements, loop B was recorded afterwards. Both affects the magnetization process. loops were recorded at a temperature of 130 K. The RT- It should be noted here that there is a controversy in the annealed film shows a higher MOKE signal and a larger literature concerning the onset of long-range ferromagnetic coercive field. This is explained by a supposedly higher Tc order in LT-grown Fe/Cu 100 films, ranging from about 1 resulting from a stronger coalescence of islands in the RT- ML to more than 2 ML.6,18,19 The present data clearly shows annealed film, and hence a lower reduced temperature T/Tc that at 130 K ferromagnetism occurs even below 1 ML. One leading to higher magnetization and magnetic anisotropy that reason why in some studies a delayed onset of ferromag- will cause the observed changes in the hysteresis loop. Inter- netism was found could be that in these cases the films were PRB 60 MAGNETIZATION REVERSAL AND SPIN . . . 7381 magnetized by short magnetic-field pulses, which were either not strong enough to overcome the rather large coercive field as compared to thicker films, see below or too short to allow for domain-wall motion. B. Magnetization reversal in out-of-plane magnetized films In the following, the magnetization-reversal process in 1.9-ML and 3.2-ML-thick Fe/Cu 100 films is discussed. In Figs. 2 a and 2 b , the hysteresis loop for a 1.9-ML-thick film is shown as well as a set of domain images taken during magnetization reversal in magnetic fields close to the coer- cive field HC 80 Oe. The hysteresis loop has square shape, and the magnetization-reversal process is characterized by nucleation of only few domains with reversed magnetization dark areas in the images followed by continuous domain growth. Note that the domain walls are rather straight. At 3.2 ML Fe/Cu 100 , the hysteresis loop retains its square shape but the coercive field is reduced to HC 32 Oe see Fig. 2 a . Furthermore, the magnetic domains forming during magnetization reversal see Fig. 2 c have a completely dif- ferent shape than those for 1.9 ML: There are numerous nucleation centers, and the domains have an irregular shape, since they are pinned at several points of the film. Nucleation centers and pinning sites correspond to crystal defects and polishing scratches on the Cu 100 surface as is seen by comparing the magnetic-domain images with the background-topography image not shown here . Domain nucleation and wall motion in these Fe/Cu 100 films is governed by thermal activation, and the wall velocity depends exponentially on the external magnetic field.10,20 A squarelike hysteresis shape usually indicates that the coer- cive field is determined by the magnetic field necessary for a domain nucleation that triggers an almost instantaneous mag- netization reversal. In the present case, however, the coercive field is given by the field at which domain-wall motion oc- curs on the time scale of the hysteresis-loop data recording in the order of 1 s . Therefore the squareness of the hyster- esis loops is a measure for the homogeneity of the films. At this point it should be mentioned that the investigated films are fairly smooth, with the roughness increasing towards larger thicknesses see STM images in Figs. 2 d and e . The lateral size of islands and holes is of the order of 1­3 nm. From annealing studies where the nanometer-scale mor- phology of 1.9-ML and 3.2-ML-thick Fe/Cu 100 films was altered we do not get any indication that magnetization re- FIG. 2. a Polar-MOKE hysteresis loops of LT-grown 1.9-ML versal is affected by roughness on this small length scale, and 3.2-ML Fe/Cu 100 taken at 130 K. b and c : Correspond- probably because the domain-wall widths exceed this scale. ing series of magnetic-domain images taken during magnetization A detailed discussion of the influence of film morphology on reversal around the coercive fields at b 1.9-ML and c 3.2-ML the magnetization-reversal process will be published Fe/Cu 100 . The dark areas indicate domains with reversed out-of- elsewhere.13 Barriers for domain-wall propagation are more plane magnetization. d and e : STM images of d 1.9-ML and likely be represented by Cu 100 -surface steps and defects e 3.2-ML Fe/Cu 100 . The islands protrude typically 1 ML out of on the Cu surface.10 For the 0.9-ML Fe/Cu 100 film, on the the surface while some holes are at least 2-ML deep and reach other hand, it is possible that anisotropies are large enough to down to the Fe-Cu interface. lead to sufficiently small domain-wall widths. In this case, the networklike film morphology see Fig. 1 b would have thickness.1 It is assigned to a decrease of the magnetic an- a strong impact on the magnetization-reversal process, which isotropy which is supposed to almost vanish at spin reorien- remains to be studied. tation resulting in lower energy thresholds for domain We will now address the observed thickness dependence nucleation and domain-wall motion. The observation of such of coercive field and domain shapes during magnetization strong changes in domain shape, however, is interesting. It is reversal: A reduction of HC is commonly observed in ultra- explained by a reduced domain-wall energy caused by a di- thin magnetic films by approaching the spin-reorientation minished magnetic anisotropy. To first approximation, the 7382 E. MENTZ, A. BAUER, T. GUšNTHER, AND G. KAINDL PRB 60 magnetic anisotropy Ea can be written as: Ea K2 cos2 (Kb,2 Ks,2 /d)cos2 , where is the angle between magne- tization direction and surface normal, d is the film thickness, and Kb,2 ,Ks,2 are second-order anisotropy constants for bulk and surface, respectively. Kb,2 is positive and includes the shape-anisotropy term 1/2 0M2 with M being the magnitude of the saturation magnetization. For Fe/Cu 100 , the surface- anisotropy term is negative, which explains the perpendicu- lar anisotropy of thin films. Since the anisotropy changes sign at the spin-reorientation thickness d0 3.8 ML Ref. 21 see discussion below , it follows that Ks,2 d0Kb,2 . The domain-wall energy per unit length, , depends on K2 , d, and the exchange stiffness A in the form d*(A K2 )1/2 (AKb,2 dd0 d2 )1/2. There is a maximum in (d) at d d0/2 1.9 ML, and a minimum at d0 3.8 ML. For the domain growth at 3.2-ML Fe/Cu 100 , the energy barrier for introducing longer domain walls is obviously smaller than the barrier height for domain walls to overcome pinning cen- ters, which explains the observed rugged domain shapes. For 1.9-ML Fe/Cu 100 , on the other hand, the domain walls tend to be as short as possible, in agreement with the rather straight domain boundaries. It should be noted that the ob- served, differently shaped domains are not representing ther- modynamical ground states. They are a consequence of the dynamical, thermally activated process of domain-wall mo- FIG. 3. a Polar-MOKE hysteresis loops for A 3.8-ML and tion. It should also be mentioned that in the discussion of B 4-ML Fe/Cu 100 . The approximate points at which domain domain shapes the influence of long-range magnetic dipole- images were recorded are indicated. b and c : Growth of mag- dipole interaction magnetostatic energy can be neglected: netic domains dark areas for 3.8-ML Fe/Cu 100 . d After a do- The domains are typically much wider than 10 m which is main state similar to that in c had established, the external mag- beyond the scale where-at a film thickness of less than 1 netic field was reduced to H 0 resulting in less magnetic contrast nm-significant variations in the magnetostatic energy in the dark areas. e ­ g : Domain images during magnetization occur.6 reversal at 4-ML Fe/Cu 100 . In the upper-right corner the nonmag- netic sample holder is seen. C. Spin reorientation since the average film magnetization cannot be determined In the last section, the magnetization-reversal process at accurately from the domain images. the spin-reorientation transition is discussed. The hysteresis The spin-reorientation phase transition occurs within a loop of a 3.8-ML Fe/Cu 100 film see Fig. 3 a shows that rather narrow thickness range 0.5 ML : Already at a there is still almost 100% remanence, but the hysteresis loop 4-ML-thick Fe/Cu 100 film, an almost hard-axis hysteresis no longer has square shape. Interestingly, rather similar hys- loop is measured though there is still some hysteresis effect, teresis loops were observed at the spin-reorientation transi- see Fig. 3 a . Accordingly, the magnetic contrast in Kerr- tion of Fe/Cu3Au 100 .22 Without domain imaging, however, microscopy images changes continuously for magnetic fields there could only be a speculation on the nature of the below the saturation field, and no domain nucleation pro- magnetization-reversal process. With Kerr-microscopy see cesses are observed see Figs. 3 e ­ g . Figs. 3 b and c , it is now seen that the magnetization Lower magnetic contrast of domains in Kerr-microscopy reversal starts with nucleation of magnetic domains that are images taken at 3.8-ML Fe/Cu 100 dark areas in Figs. even more irregularly shaped than that in case of 3.2-ML 3 b ­ d is directly attributed to a reduced out-of-plane Fe/Cu 100 , in consistency with the preceding discussion. magnetization component within the domains. The domains However, the magnetic contrast is lower than observed for are either in a uniform canted magnetization state including 3.2-ML Fe/Cu 100 . Upon reaching an apparent single- in-plane magnetization at or near zero external magnetic domain state,23 the magnetic contrast increases further with field or in a microdomain state with a nonresolved inner higher external magnetic fields up to the saturation field H0 domain structure. In any case, the domains exhibit a different 200 Oe. To demonstrate the magnetic contrast effects spin structure than the surrounding areas with uniform out- more clearly, the external magnetic field was reduced before of-plane magnetization stabilized in an external magnetic a single-domain state had been established inner hysteresis field prior to magnetization reversal bright areas in Figs. loop . In this case, the domain contrast diminished continu- 3 b ­ d . Both models are discussed in more detail in the ously and reversibly with decreasing and increasing re- following. versed external magnetic field compare Figs. 3 c and d , It is known that at the spin-reorientation transition micro- with the domain walls at an apparently fixed position. The domain states become energetically favorable due to a di- corresponding inner hysteresis loop is not plotted in Fig. 3 a minished magnetic anisotropy and a gain in dipole energy.6 PRB 60 MAGNETIZATION REVERSAL AND SPIN . . . 7383 Due to this fact, hysteresis loops measured at Fe/Cu ingly, one recent study by Speckmann, Oepen, and Ibach 3Au 100 with shapes like that in Fig. 3 a for 3.8-ML Fe/Cu 100 suggests the coexistence of the out-of-plane and in-plane were previously attributed to a transition from a metastable magnetization phase in remanence at the spin-reorientation out-of-plane magnetization state that persists in zero external transition of Co/Au 111 .7 Experimentally, they show that a magnetic field remanence to a microdomain state at a suf- metastable out-of-plane single-domain state exist. It remains ficiently large reversed external magnetic field.22 The net unclear, however, whether also the in-plane domain state can magnetization of a thermodynamically stable microdomain be stabilized. The coexistence of the two phases is explained state would be zero at zero external magnetic field and in- by taking higher-order contributions to the magnetic anisot- crease with field. The energy of the domain walls that have ropy into account. Indeed, the importance of higher-order to be created at the transition represents an activation barrier anisotropies at the spin-reorientation transition was pointed and explains why the out-of-plane state is metastable. So far, out in recent works.7,26 this model is consistent with the present Kerr-microscopy On the basis of the present data alone, we cannot give a data since microdomain states e.g., stripe domains with conclusive answer to the question of the spin configuration stripe widths 1 m are not resolved, and the changing within the domains with a reduced out-of-plane magnetiza- magnetic contrast in the Kerr-microscopy images would just tion component observed at 3.8-ML Fe/Cu 100 -whether it reflect the net magnetization of the microdomain state. The is uniform or exhibits a microdomain structure. A better the- Kerr-microscopy images in Figs. 3 b and c imply that after oretical understanding of domain formation in external mag- nucleation the possible microdomain phase grows by netic fields as well as domain-imaging studies with improved domain-wall displacement rather than entirely by nucleation. lateral resolution are required. For this purpose, we are cur- Therefore it could be expected that the domains have more a rently developing a UHV scanning near field optical micro- fractal-like shape2,24 rather than a stripelike configuration.6 scope SNOM for magnetic-domain imaging with submi- There is one point which is not in accordance with previ- cron resolution.27 ous theoretical and experimental studies of microdomain states, though: Small external magnetic fields 50 Oe IV. SUMMARY should already destabilize the microdomain state and cause a transition to an out-of-plane magnetized single-domain In summary, we have investigated the magnetization re- phase.6,8 In the present case, however, the transition field versal process in external magnetic fields at low-temperature amounts to about 200 Oe see hysteresis loop for 3.8-ML grown Fe/Cu 100 films for film thicknesses up to 4 ML. Fe/Cu 100 in Fig. 3 a . This fact leads to the assumption With a combined in situ MOKE, Kerr microscopy, and STM that at 3.8 ML, the effective second-order magnetic anisot- study we were able i to attribute the onset of long-range ropy constant K ferromagnetic order to the coalescence of bilayer islands at 2 has already changed sign (K2 0), favor- ing in-plane magnetization.25 Provided that the out-of-plane about 0.9 ML, ii to assign the irregular domain shapes magnetized state is still metastable, the nucleated domains above 3 ML to a reduced domain-wall energy, and iii to observed in the Kerr-microscopy images Figs. 3 b ­ d identify two coexisting spin configurations at the spin- would have uniform in-plane magnetization, in contrast to reorientation transition that occurs in a narrow thickness the preceding discussion. The inclined straight sections in the range around 3.8 ML. hysteresis loop Fig. 3 a and the changing magnetic con- ACKNOWLEDGMENT trast in the Kerr-microscopy images Figs. 3 c and d would result from a coherent rotation of the magnetization in This work was supported by the Deutsche Forschungsge- the external magnetic field canted spin orientation . Interest- meinschaft, Project No. Sfb-290/TPA6. 1 P. Bruno, G. Bayreuther, P. Beauvillain, C. Chappert, G. Lugert, Berger and R. P. Erickson, J. Magn. Magn. Mater. 165, 70 D. Renard, J. P. Renard, and J. Seiden, J. Appl. Phys. 68, 5759 1997 . 1990 . 9 D. Sander, R. Skomski, C. Schmidthals, A. Enders, and J. Kir- 2 A. Lyberatos, J. Earl, and R. W. Chantrell, Phys. Rev. B 53, 5493 schner, Phys. Rev. Lett. 77, 2566 1996 . 10 1996 . A. Kirilyuk, J. Giergiel, J. Shen, and J. Kirschner, J. Magn. Magn. 3 A. Moschel, R. A. Hyman, A. Zangwill, and M. D. Stiles, Phys. Mater. 159, L27 1996 . 11 A. Vaterlaus, U. Maier, U. Ramsperger, A. Hensch, and D. Pes- Rev. Lett. 77, 3653 1996 . 4 cia, Rev. Sci. Instrum. 68, 2800 1997 . D. P. Pappas, K.-P. Kašmper, and H. Hopster, Phys. Rev. Lett. 64, 12 H. J. Elmers, J. Hauschild, H. Hošchle, U. Gradmann, H. Bethge, 3179 1990 . D. Heuer, and U. Košhler, Phys. Rev. Lett. 73, 898 1994 . 5 R. Allenspach and A. Bischof, Phys. Rev. Lett. 69, 3385 1992 . 13 E. Mentz, A. Bauer, and G. Kaindl unpublished . 6 Y. Yafet and E. M. Gyorgy, Phys. Rev. B 38, 9145 1988 ; A. B. 14 S. Mušller, P. Bayer, C. Reischl, K. Heinz, B. Feldmann, H. Kashuba and V. L. Pokrovsky, ibid. 48, 10 335 1993 . Zillgen, and M. Wuttig, Phys. Rev. Lett. 74, 765 1995 . 7 M. Speckmann, H. P. Oepen, and H. Ibach, Phys. Rev. Lett. 75, 15 H. Magnan, D. Chandesris, B. Vilette, O. Heckmann, and J. 2035 1995 ; H. P. Oepen, M. Speckmann, Y. Millev, and J. Lecante, Phys. Rev. Lett. 67, 859 1991 . Kirschner, Phys. Rev. B 55, 2752 1997 . 16 D. A. Steigerwald, I. Jacob, and W. F. Egelhoff, Surf. Sci. 202, 8 A. Berger and H. Hopster, Phys. Rev. Lett. 76, 519 1996 ; A. 472 1988 . 7384 E. MENTZ, A. BAUER, T. GUšNTHER, AND G. KAINDL PRB 60 17 E. Mentz, D. Weiss, J. E. Ortega, A. Bauer, and G. Kaindl, J. ponent so that domains with different in-plane spin orientation Appl. Phys. 82, 482 1997 . cannot be distinguished by polar Kerr microscopy. 18 D. Li, M. Freitag, J. Pearson, Z. Q. Qiu, and S. D. Bader, Phys. 24 J. Valentin, Th. Kleinefeld, and D. Weller, J. Phys. D 29, 1111 Rev. Lett. 72, 3112 1994 . 1996 . 19 D. P. Pappas, C. R. Brundle, and H. Hopster, Phys. Rev. B 45, 25 This possibility was also discussed in Ref. 22 and dismissed since 8169 1992 . no MOKE signal was detected in longitudinal geometry, which 20 E. Fatuzzo, Phys. Rev. 127, 1999 1962 . would be expected if the easy axis is in plane. For a better 21 H. Zillgen, B. Feldmann, and M. Wuttig, Surf. Sci. 321, 32 understanding, longitudinal MOKE measurements would have 1994 . to be done here as well. 22 F. Baudelet, M. T. Lin, W. Kuch, K. Meinel, B. Choi, C. M. 26 K. Baberschke and M. Farle, J. Appl. Phys. 81, 5038 1997 . Schneider, and J. Kirschner, Phys. Rev. B 51, 12 563 1995 . 27 B. L. Petersen, A. Bauer, G. Meyer, T. Crecelius, and G. Kaindl, 23 Polar MOKE measures only the out-of-plane magnetization com- Appl. Phys. Lett. 73, 538 1998 .