PHYSICAL REVIEW B VOLUME 57, NUMBER 21 1 JUNE 1998-I Evolution of the magnetization depth profile of Fe/Cu 100... films upon thermal annealing J. Shen, Ch. V. Mohan, P. Ohresser, M. Klaua, and J. Kirschner Max-Planck-Institut fušr Mikrostrukturphysik, Weinberg 2, 06120 Halle/Saale, Germany Received 11 September 1997; revised manuscript received 5 November 1997 The annealing effect on the magnetic and structural properties of Fe/Cu 100 ultrathin films has been studied. For films below 5 ML, their magnetization and Curie temperature are reduced considerably after annealing. We explain the reduction as a result of the transition from a high-moment ferromagnetic phase to a nonferromagnetic phase in the inner layers of the Fe films after annealing. Supporting evidence comes from the low-energy electron-diffraction study, which indicates a structural relaxation from a tetragonally distorted fct structure to a true fcc structure in the inner layers of the Fe films. The surface layers of the Fe films, after annealing, are still expanded, providing the main contribution to the remaining magnetization and nonzero Curie temperature of the annealed Fe films. S0163-1829 98 00821-2 I. INTRODUCTION been realized for some years that after annealing the surface layers of the films become copper rich.15 Our recent scanning The ultrathin film system of epitaxially grown face- tunneling microscopy STM study16 has unambiguously centered-cubic fcc Fe on Cu 100 has been extensively shown that the mechanism of the copper diffusion is a sur- studied in recent years.1­8 It has been observed that this sys- face diffusion via rectangular pits which are formed during tem is metastable upon variation of thickness. The structure annealing. In Refs. 16 and 17 these pits were termed as ``pin- of the room-temperature as-grown films is face-centered te- holes'' which did not reflect the fact that the lateral extension tragonal fct with buckling below 5 ML.9 Between 5 and 11 of the pits is usually 5­10 times larger than their vertical ML,3 the films have a fct structure in the top layers but a fcc extension. Therefore in this paper we will rename them structure in the layers underneath. The structure of the films ``pits'' instead of ``pinholes.'' The diffused copper from the finally transforms to a much more stable body-centered- pits covers the surface of the films and forms a Cu/Fe/Cu cubic bcc modification above 11 ML.10 The fcc to bcc sandwich structure. This also implies that the equilibrium of structural transformation is largely driven by the energy dif- the Fe/Cu 100 films is neither two-dimensional layers nor ference between fcc and bcc Fe. The transformation starts three-dimensional clusters of Fe on the substrate, but rather a with forming dislocations at about 4 to 5 ML and proceeds Cu/Fe/Cu sandwich structure. by following the martensitic path.2,11 The appearance of the It is important to know the atomic structure and in par- dislocations is likely associated with the fct fcc structural ticular the magnetic properties of the annealed Fe/Cu 100 relaxation which occurs between 4 and 5 ML. films. This is not only because of the interest of understand- The structural changes of the Fe/Cu 100 system have ing the properties of the Fe/Cu 100 films at equilibrium, but strong influence on its magnetic behavior. In the fct thick- also because of the fact that thermal annealing is a general ness regime ( 5 ML), the films have a uniform high-spin process during the magnetic measurements such as Curie ferromagnetic phase.6 The easy magnetization axis is perpen- temperature. Apparently the physical meaning of the mea- dicular to the film surface.1 Between 5 and 11 ML, while the sured magnetic quantities at high temperatures will only be easy magnetization axis is still perpendicular, the magnetiza- well understood by studying the annealing effect. Further- tion of the films become nonuniform in depth.1,4 It has been more, a study of the annealing effect on the structure and observed that in this thickness regime only the topmost layer magnetism of the Fe/Cu 100 films will also serve as guid- has a high-spin ferromagnetic phase while the layers under- ance for some other magnetic systems such as Co/Cu 100 neath become nonmagnetic or antiferromagnetic depending Ref. 17 and Fe/Au 100 ,18 where a similar diffusion on the temperature.1,4,7,8 As a result, the total magnetization mechanism holds. of the films from 5 to 11 ML has only a value about 30­ Therefore, following our previous work on morphology,16 40 % of that of the 4 ML film. The reduction of the magne- in this paper we describe the influence of annealing on the tization has been generally believed to result from the fct to magnetization depth profile of the Fe/Cu 100 films after an- fcc structural relaxation,3 which agrees with a theoretical nealing, a subject which has hardly been touched in this prediction of the magnetic moment­atomic volume relation- ``ever green'' system. We have observed that the Curie tem- ship of fcc Fe.12 perature and the total magnetization of the Fe films become The metastability of the Fe/Cu 100 films has also shown remarkably small after annealing. We interpret these changes up when varying the temperature. Zharnikov et al.13 have as a result of the transition from a high-spin ferromagnetic shown that a 4 ML film undergoes a reversible fct to fcc state to a low-spin ferromagnetic or even a nonferromagnetic structural relaxation during temperature cycle from 160 to state in the inner layers of the Fe films, which is strongly 370 K. Further increasing the annealing temperature above supported by structural data from our low-energy electron- 400 K results in significant Fe-Cu interdiffusion.14 It has diffraction LEED studies. 0163-1829/98/57 21 /13674 7 /$15.00 57 13 674 © 1998 The American Physical Society 57 EVOLUTION OF THE MAGNETIZATION DEPTH . . . 13 675 II. EXPERIMENTAL DETAILS The experiments were performed in an ultrahigh vacuum UHV multichamber system including a molecular-beam- epitaxy preparation chamber, a STM chamber, an analysis chamber equipped with facilities for Auger electron spectros- copy AES , LEED and thin-film growth, and a magneto- optical Kerr effect MOKE chamber. The base pressure of the individual chambers is better than 5 10 11 mbar. A fully automatic video-LEED system19 has been used for re- cording LEED images as well as for measuring intensity vs energy I/V LEED curves. The sample was prepared in the analysis chamber. Prior to film deposition the copper sub- strate was cleaned by Ar sputtering followed by annealing at 870 K. After several cycles of this procedure we have achieved a clean and flat substrate. Contamination-free Au- ger spectrum, sharp LEED spots and large atomically flat terraces on the order of several hundred nanometers under STM all prove the high quality of the copper substrate. The Fe films were prepared from an iron wire (5N) heated by e-beam bombardment. During deposition the substrate was kept at room temperature (300 5 K), and the vacuum pres- FIG. 1. Polar MOKE hysteresis loops of 3 and 4 ML Fe/ sure rose from 7 10 11 to 2 10 10 mbar at a typical Cu 100 film before and after annealing. The loops were recorded at evaporation rate of 0.2 monolayer/min. The well-cleaned Fe 150 K. Annealing has caused an increase of the coercive field Hc source and the excellent vacuum have guaranteed the clean- but a decrease of the saturation magnetization Ms and remanance liness of the Fe films as examined by the AES 0.5 at. % Mr . of carbon contamination . After the film preparation, the LEED and IV-LEED mea- after annealing. This has also been confirmed by the in-plane surements were taken immediately from the as-grown films. MOKE measurements not shown here , which show a typi- Then the sample was transferred to the MOKE chamber. cal hard-axis behavior of the film. Magnetic data were recorded in situ from films both before Before annealing, the 3 ML film has a much smaller co- and after annealing. The annealing temperature was typically ercivity (Hc) than that of the 4 ML film. Generally a smaller about 490 K though other temperatures have also been tried. Hc is an indication of a more perfect structure which has less The heating rate was about 10 K/min in the temperature defects to pin the domain motion. However, according to our range between 300 and 490 K, allowing the maximum pres- previous STM results, the surface roughness of the 3 and 4 sure to be maintained below 3 10 10 mbar. The copper dif- ML films are small and comparable.16 The large difference fusion process has been observed to be dependent on both between the Hc values of the 3 and 4 ML films must come the annealing temperature and the heating rate, and a detailed from other factors causing structural imperfection. Previous discussion will be presented in a forthcoming paper.20 The LEED Ref. 3 and STM Ref. 2 studies indicate that above LEED and IV-LEED analysis of the annealed samples were 4 ML the Fe films undergo a fct fcc structural relaxation finally done in the analysis chamber. STM and AES have and a fcc bcc phase transformation simultaneously. Both also been used to examine the number of pits and the amount processes cause structural imperfection: the former results in of diffused copper in the annealed films, and similar results different interlayer spacing for different layers while the lat- to those described in Ref. 16 were obtained. ter generates dislocations. It is these imperfections that are likely responsible for the enhanced coercivity of the 4 ML film. III. REDUCED MAGNETIZATION After annealing the hysteresis loops are widened with a AND ITS STRUCTURAL ORIGIN distinctly larger Hc than that of the films before annealing. The effect is particularly strong for the 3 ML film. At 150 K, Let us first discuss the changes of the magnetic properties the coercive field drastically increases by a factor of more after annealing. For convenience, we will concentrate on than 20, from 25 Oe before annealing to 535 Oe after anneal- films with thickness of 3 and 4 ML, while we note here that ing. For the 4 ML film, the increase of Hc is much less the characteristic features are general for all films between 2 pronounced, from 230 Oe before annealing to 340 Oe after and 5 ML. Figure 1 shows the measured Kerr hysteresis annealing. It is interesting to note that after annealing, the 4 curves of the 3 and 4 ML films before and after annealing. ML film has a smaller Hc than the 3 ML film though the All the curves were recorded in the polar geometry at about situation is just reverse before annealing. 150 K. The hysteresis loops of the films before annealing Another visible effect of annealing, from Fig. 1, is the have a well-defined rectangular shape, while after annealing decrease of the magnetization of the films. At 150 K, the the corners of the hysteresis become somewhat rounded. In measured magnetization appears to have decreased by a fac- both cases the remanent magnetization (Mr) equals 100% of tor of 2 for both 3 and 4 ML films. A more strict comparison the saturation magnetization (Ms), indicating that the easy of the magnetization requires also the knowledge of the Cu- magnetization axis remains to be perpendicular to the surface rie temperature (Tc), because the chosen temperature 150 13 676 SHEN, MOHAN, OHRESSER, KLAUA, AND KIRSCHNER 57 magnetization, from Fig. 2. We estimate that the magnetiza- tion 0 K of the 3 ML film after annealing is about 40% of that of the film before annealing. By the same way we de- termined that the magnetization of the 4 ML annealed film reduces to about 45% of the original value. Two possible mechanisms could be responsible for the decrease of the magnetization after annealing. The first one is the copper diffusion towards the top of the film surface.16 It has been observed recently that capping layers on top of a 3 ML Fe film results in a reduction of the magnetization and the Curie temperature depending on the copper thickness.22 Moreover, in Ref. 16 we have demonstrated that the diffused copper mixes with the top Fe layers and form Fe-Cu surface alloy. Theoretically for an atomically ordered Fe-Cu alloy,23 the magnetic moment of Fe atoms is only slightly smaller than in pure Fe once the copper concentration is under 50%. As no such ordered alloy exists in the bulk, one could only produce Fe-Cu alloys either by stabilizing fcc Fe clusters in a Cu matrix, i.e., a cluster-type alloy, or by epitaxially stack- ing monoatomic Fe and Cu layers.24 The former is antiferro- magnetic with a NeŽel temperature of about 67 K,25 while the latter appears to be ferromagnetic its magnetic moment needs to be further examined24 . The Fe-Cu surface alloy in the annealed Fe/Cu 100 films consists of Fe-rich and Cu- FIG. 2. Temperature dependence of the saturation Ms and rem- rich patches,16 which is in between the ordered alloy and the anent magnetization Mr of the 3 ML upper panel and 4 ML lower cluster alloy. The reduced magnetization of the annealed Fe panel Fe/Cu 100 film before and after annealing. The dashed and films can also be partly caused by the alloy formation if one the full lines correspond to Mr before and after annealing, respec- assumes the Fe magnetic moment to be small in the alloy. tively, and are only for guiding the eyes. The second possible mechanism would be a magnetic phase transition. By annealing the Fe films may transform K for comparison must be well below Tc to avoid any un- from the high-spin ferromagnetic phase to a low-spin ferro- certainties caused by the rapid fall of the magnetization in magnetic phase or a nonmagnetic phase. Because the close the vicinity of Tc . To determine Tc , we have measured the connection between the magnetic moment and atomic vol- temperature dependence of Ms and Mr . The values of the ume of fcc Fe, the magnetic phase transition should result in Ms and Mr were taken from the hysteresis loops. The results or originate from a change of lattice constant. If the films for the 3 ML upper panel and 4 ML film lower panel are transform to a low-spin phase, the lattice constant of the shown in Fig. 2. It remains a matter of dispute whether one whole films should become uniformly smaller. If the films should use Ms or Mr to determine the precise value of Tc . In transform to a nonmagnetic phase, parts of the films must be principle, these two should yield the same result as far as a still ferromagnetic, otherwise there would be no magnetic film with a perfect structure is concerned. The difference signal detectable after annealing. Therefore only the non- between the temperature dependence of Ms and Mr often magnetic part of the films would have a reduced interlayer reflects the deviation of the structure of the film from the distance while the ferromagnetic part remains unchanged. perfect order. It has been suggested by Kohlhepp et al.,21 To decide which of the above two mechanisms is respon- that the Ms vs T curve would yield a more reliable Tc if the sible for the reduction of the magnetization, it is important to saturation field remains reasonably small. However, all the know 1 the exact amount of diffused copper onto the sur- Ms curves except that of the 3 ML film before annealing in face; 2 the extent of the change of the lattice constant. A Fig. 2 were measured under large field, which requires the general method to determine the copper diffusion is AES. Curie temperature to be determined by the Mr vs T curve. But in the Fe/Cu 100 system a quantitative analysis by AES The 3 ML film before annealing has a rather small saturation turns out to be difficult because of the formation of the field less than 100 Oe , but its structural perfection results in Fe-Cu alloy. We therefore decided to use STM to determine only a small difference between the Ms and Mr curves. the amount of diffused copper. As an example, Fig. 3 shows Therefore, for consistency we will use the Mr vs T curves to STM topography images of a 3 ML Fe/Cu 100 film before determine Tc in all four cases. The obtained Tc values for the a and after b annealing. Before annealing, the film shows 3 and 4 ML films are 325 and 275 K before annealing, and a good layer-by-layer morphology as the third layer has been 220 and 215 K after annealing, respectively. Apparently the more than 95% filled. After annealing, pits dark patches Curie temperature of the Fe films has been considerably re- have been formed in the film. Between the pits there exist duced after annealing. It is also interesting to note here that many small sub-monolayer-deep depressions, which are con- the 3 and 4 ML annealed films have a close Tc value around sidered as characteristic features of the Fe-Cu surface 220 K. alloy.16 The total material diffused out of the pits can be With the knowledge of the temperature dependence of determined by calculating the volume of the pits. In order to Ms , we can extrapolate Ms at 0 K, i.e., the spontaneous reduce the statistical error, we made our calculation on more 57 EVOLUTION OF THE MAGNETIZATION DEPTH . . . 13 677 FIG. 3. STM topography images of a 3 ML Fe/Cu 100 film before a and after b annealing. Before annealing the film exhib- its three exposed layers, i.e., second dark , third grey , and fourth bright layer. The third layer contributes more than 95% of the total area. After annealing, several nanometer-deep pits dark patches have been formed as a result of the copper diffusion to- wards the top of the surface. Submonolayer-deep depressions are clearly visible between the pits. These depressions are characteristic features of the Fe-Cu surface alloy, see also, Ref. 16. than 20 images (300 300 nm2) taken from various surface locations. Subtracting the Fe part 0.1 and 0.07 ML for 3 and 4 ML film, respectively , we estimated the amount of dif- fused copper to be about 0.6 and 0.4 ML for the 3 and 4 ML Fe film, respectively. Above 4 ML the copper diffusion be- comes insignificant as the films are thick enough to preclude pit formation. The structural changes of the films upon annealing have been analyzed by LEED and IV-LEED. Figure 4 shows the LEED patterns normal incidence recorded before left col- umn and after annealing right column of the 3, 4, and 5 ML Fe/Cu 100 films. Before annealing, the 3 and 4 ML FIG. 4. LEED patterns of Fe/Cu 100 films before annealing films exhibit a complicated (5 1) superstructure, while the left column and after annealing right column . All pictures were 5 ML film shows a (2 1)p2mg type of superstructure. These recorded at 108 eV. Before annealing, the 3 and 4 ML films exhibit results are consistent with previous LEED studies of the (4 1) and (5 1) superstructures, while the 5 ML film shows a room-temperature as-grown Fe/Cu 100 films.26,27 The (5 (2 1)p2mg structure. After annealing, all the films have a c(2 1) superstructures have been considered as a reflection of 2) superstructure whose intensity relative to that of the substrate the buckling of the surface atoms.27 In this respect at room- spots decreases with increasing thickness. temperature the growth of Fe on Cu 100 is by no means a real pseudomorphic growth. latter could explain the fact that the intensity of the super- Stark change of the LEED patterns has been observed for structure spots decreases with increasing thickness, since the the films after annealing, as shown in the right column of amount of diffused copper, thus the Fe-Cu surface alloy, de- Fig. 4. None of the original superstructures is visible in the creases with increasing thickness. The adsorption of the re- annealed films. Instead, another type of superstructure, c(2 sidual gases on the films' surface, however, should not de- 2), appears in the LEED patterns. The relative intensity of pend on the thickness. Therefore we tentatively attribute the the c(2 2) superstructure spots over the substrate spots de- origin of the c(2 2) superstructure to the Fe-Cu surface creases with increasing Fe thickness, reaching nearly the alloy. At this point we do not know the exact atomic arrange- level of the background at 5 ML. In this system, the c(2 ment which causes the c(2 2) pattern. 2) superstructure could be caused by two optional mecha- In any case, we can conclude that there is no buckling of nisms. First, it is known28 that the adsorption of the residual the surface atoms of the annealed films because of the dis- gas in vacuum O2, CO, etc. could cause such a superstruc- appearance of the (5 1) superstructure. This, to some ex- ture on the surface of the Fe/Cu 100 films. During annealing tent, is an indication that annealing improves the lateral order the vacuum pressure often increases slightly, which will un- of the Fe films with respect to that of the copper substrate, avoidably result in a bit larger amount of adsorption than except that the surface layer is affected by the Fe-Cu surface normal. Our AES data confirm the weight of carbon and alloy. Such an improvement, according to our IV-LEED re- oxygen peaks to slightly increase after annealing but by a sults, is not only limited in the lateral direction, but also in factor of less than 2 corresponding to less than 1% carbon the vertical direction of the Fe films. contamination . Second, the c(2 2) superstructure may Figure 5 shows the LEED 00 beam intensity as a func- simply reflect the order of the Fe-Cu surface alloy which is tion of the beam energy. The I(E) spectra of the Fe/Cu 100 formed due to the copper diffusion during annealing. The films before and after annealing are displayed in the lower 13 678 SHEN, MOHAN, OHRESSER, KLAUA, AND KIRSCHNER 57 the existence of the low-energy fct peaks. After annealing, the I(E) spectra of the 3, 4, and 5 ML Fe films upper panel in Fig. 5 are remarkably similar to the spectrum of the 5 ML film before annealing. This indicates that the inner layers of the annealed Fe films have the same interlayer distance as copper. In other words, annealing re- sults in a structural relaxation, from fct to fcc, in the inner layers of the 3 and 4 ML Fe films. The inner layers of the 5 ML film already have the fcc structure before annealing, thus no further structural change occurs upon annealing. The sur- face layers of the annealed Fe films appear to be still ex- panded, though the intensity of the two low-energy fct peaks is even smaller than for the 5 ML film before annealing. The following picture can be drawn from our STM and LEED data regarding the annealing effect on the structure of the room-temperature grown Fe/Cu 100 films. For the films below 5 ML, upon annealing, rectangular pits are formed in the Fe films, serving as channels for the diffusion of the substrate copper onto the top of the surface. The diffused copper mixes with the surface layers of the Fe films and FIG. 5. IV/LEED spectra of the 00 beam for Fe/Cu 100 films forms an Fe-Cu surface alloy. A fct fcc structural relax- before lower panel and after upper panel annealing. The dashed ation occurs in the bulk layers of the Fe films, while the lines indicate the fcc peak positions as obtained from that of the structure of the surface layers remains fct-like. At or above 5 copper substrate, while the full lines mark the fct peaks of the Fe ML, the films become thick enough to preclude the pit for- films. Before annealing, the 3 and 4 ML films have four fct peaks mation on the atomically flat terraces though a few pits may each, and the 5 ML film has only two low-energy fct peaks. After exist at some weak points of the films. A detailed discussion annealing, all the films have only two low-energy fct peaks, whose of the origin of these annealing effects will be given in intensity is even smaller than those of the 5 ML film before anneal- Sec. IV. ing. and the upper panels, respectively. In this case the primary IV. DISCUSSION beam is about 6° off the surface normal lying in the 001 plane. Before annealing, the spectra of the 3 and 4 ML films As mentioned, the amount of diffused copper for the 3 are characterized by two families of peaks which are marked ML film is about 0.6 ML. Deliberately capping such an by solid and dashed lines, respectively. The peaks on the amount of copper onto the as-grown 3 ML Fe film would dashed lines have nearly the same energy positions as those lead to a magnetization reduction by only about 15%,22 of the copper substrate. These peaks will be referred to as fcc which is not sufficient to explain the observed reduction by a peaks, since they stand for an interlayer distance which is factor of 2. The fct to fcc structural relaxation in the inner close to that of the fcc substrate. The peaks on the solid lines layers of the annealed Fe films, however, could well explain clearly have lower energy positions than the fcc peaks. Un- the reduction of the total magnetization after annealing. The der the assumptions of the kinematic theory, these peaks cor- contracted inner layers, with a smaller atomic volume, are respond to an interlayer distance which is larger than that of likely converted into a nonferromagnetic state, i.e., paramag- the substrate. These peaks will be referred to as fct peaks. netic or antiferromagnetic state, whose magnetization is sig- Within the kinematic model we have calculated the inter- nificantly smaller than the high-spin ferromagnetic state. As layer distance, which is 1.82 Ć for the fcc and 1.95 Ć for the for the topmost layers, we cannot unequivocally determine fct peaks. For the films below 5 ML, the intensity of the fcc their magnetization. On one hand, they remain expanded af- peaks decreases with increasing thickness, hinting that these ter annealing, which should still result in the high-spin fer- peaks are likely contributed from the substrate. The fct romagnetic state. On the other hand, they are present in a peaks, on the other hand, reflect the real interlayer distance form of a Fe-Cu surface alloy instead of pure Fe. The mag- of the Fe films below 5 ML. The existence of the fct peaks netic moment of the Fe atoms in the Fe-Cu alloy might de- and the superstructure suggests that the Fe films ( 5 ML) crease depending on the Cu concentration as well as on the have adopted neither the lateral nor the vertical lattice con- structure of the alloy. As mentioned earlier, so far there has stant of the copper substrate. been no experimental or theoretical work discussing the Increasing thickness up to 5 ML affects the fct peaks magnetic moment of Fe atoms in the type of Fe-Cu alloy strongly. The three high-energy fct peaks disappear, and the observed in the annealed Fe/Cu 100 films, i.e., a surface intensity of the two low-energy fct peaks become weaker and alloy formed by Fe-rich and Cu-rich patches. To match the shift slightly towards higher energies. Since the high-energy measured magnetization of the annealed Fe films, we suggest peaks are contributed by electrons with longer escape length, the following model of the magnetization depth profile of the the disappearance of these peaks indicates that the bulk lay- annealed Fe films. ers of the Fe films are no longer vertically expanded. The We first discuss the 3 ML film. The annealed 3 ML Fe surface layers, however, are still expanded as evidenced by film become virtually 3.7 ML thick owing to the additional 57 EVOLUTION OF THE MAGNETIZATION DEPTH . . . 13 679 material diffused out of the pits. This material includes 0.6 ever, the formation of the pits in the Fe/Cu 100 films has ML of Cu and 0.1 ML of Fe. If the diffused 0.6 ML Cu not only reduced the total free energy of the system due to mixes only with the topmost layer of Fe, the alloy would the copper diffusion, but also caused a stress relief in the contain 35% of Cu and 65% of Fe. The calculation in Ref. 22 film. The latter can be understood in the following way. Be- indicates that the magnetic moment of Fe atoms will only fore annealing, the Fe films have a tensile strain in the ver- reduce very little once the Fe concentration exceeds 50%. tical direction as indicated by the interlayer expansion Fig. Therefore, in this case the 1.7 ML topmost Fe-Cu alloyed 5 . The relief of the stress in the Fe films can be largely layers should have a magnetization similar to that of 1.1 ML accomplished by pit formation, as a considerable amount of Fe in the high-spin phase. The magnetization of the annealed strained volume has been removed from the pits in the films. film is about 37% of that of the 3 ML high-spin Fe before The relaxation of the vertical tensile strain directly results in annealing, which agrees well with the 40% value obtained the change of the structure from fct to fcc in the inner layers from Fig. 2. The bottom two layers in the annealed film of the Fe films, as shown by the IV-LEED data in Fig. 5. The should possess small or even zero net magnetization. The residual strain manifests itself in the still expanded top lay- detailed magnetic structure, whether it is low-spin ferromag- ers. Annealing also improves the structural order in the lat- netic or paramagnetic, remains unclear. eral direction. The buckled structure of the Fe films before The same model can be applied to the 4 ML film. After annealing is replaced by a well ordered fcc-like structure annealing, 0.4 ML of diffused Cu forms an alloy with the except the Fe-Cu alloyed topmost layers. topmost Fe layer. Here the magnetization of the 1.4 ML However, the formation of the pits might not be the only Fe-Cu alloy would equal roughly that of 1 ML high-spin reason for the fct to fcc structural relaxation. In a recent ferromagnetic Fe, which is 25% of the magnetization of the 4 temperature-dependent IV/LEED study of the Fe/Cu 100 ML film before annealing. Since Fig. 2 bottom panel shows film, Zharnikov et al.13 have shown that the fct structure of that the magnetization of the 4 ML film after annealing is the films tends to relax back to the fcc structure upon heating about 45% of that before annealing, one has to assume that to a temperature of 370 K, at or below which no pits could the three Fe layers below the alloyed layers of the annealed be formed. The main difference between the fct to fcc struc- film have a nonzero net magnetization. Such a nonzero net tural relaxation observed in Ref. 13 and in this work, is the magnetization can be resulted from two types of spin align- reversibility of this structural change. In Ref. 13, the fcc to ment in the three inner layers: 1 a perfect layered antifer- fct transition is a reversible process as the fct structure will romagnetic structure, or 2 a low-spin ferromagnetic struc- be recovered after the temperature is lowered. In the present ture. As the NeŽel temperature of the inner layers is lower work, the fct to fcc transition is a completely irreversible than the Curie temperature of the surface layer, one would process and the transformed fcc structure is stable at low expect a sudden increase of the magnetization at the NeŽel temperatures. Comparing these two studies, we conclude that temperature in the Ms vs T curve of the 4 ML annealed film the transformation from fct to fcc is a general tendency of the assuming the inner layers have an antiferromagnetic struc- Fe/Cu 100 system upon heating, while the pit formation ture. However, this phenomenon has not been observed in helps to stabilize the transformed fcc structure. Fig. 2. Thus, it is more likely that the inner layers of the 4 ML annealed film have a low-spin ferromagnetic structure. V. SUMMARY The decrease of the Curie temperature after annealing can be easily understood according to the suggested model. After In summary, we have achieved a comprehensive under- annealing, the measured TC reflects the Curie temperature of standing of the evolution of the magnetization depth profile the remaining high-spin ferromagnetic layers. The Curie of the Fe/Cu 100 films upon thermal annealing. The change temperature of the annealed film, therefore, decreases to a of the magnetic properties is closely connected with the value corresponding to 1­2 layers of Fe-Cu alloy, which is annealing-induced structural relaxation. Before annealing, determined experimentally to be about 220 K Fig. 2 . On the films are strained and have a buckled fct structure. The the other hand, the large increase of Hc after annealing has to fct structure tends to transform to the fcc structure in order to be associated with the pit formation. The pits could serve as reduce both the strain energy and the magnetic energy. Such pinning centers for the domain-wall motion. This also ex- a transformation can be accomplished by the pit formation plain why after annealing the Hc of the 4 ML film is smaller upon annealing. The annealed films have a fcc structure in than that of the 3 ML film: there are less pits in the annealed the bulk layers and a fct structure in the topmost layers. The 4 ML film than in the annealed 3 ML film.16 topmost layers are still expanded, are ferromagnetic and con- Since the magnetic phase transition after annealing is tribute to the measured magnetic signals. closely connected with the fct fcc structural relaxation in the inner layers of the Fe films it is worthwhile to discuss the ACKNOWLEDGMENTS origin of this structural relaxation. In the previous work16 we have already experimentally proved that the driving mecha- The authors would like to thank J. Barthel, F. Pabisch, nism for the pit formation is the surface free energy. How- and G. Kroder for their technical support. 13 680 SHEN, MOHAN, OHRESSER, KLAUA, AND KIRSCHNER 57 1 J. Thomassen, F. May, B. Feldmann, M. Wuttig, and H. Ibach, Kirschner, Phys. Rev. Lett. 76, 4620 1996 . Phys. Rev. Lett. 69, 3831 1992 . 14 M. Arnot, E. M. McCash, and W. Allison, Surf. Sci. 272, 154 2 J. Giergiel, J. Kirschner, J. Landgraf, J. Shen, and J. Woltersdorf, 1992 . Surf. Sci. 310, 1 1994 . 15 T. Detzel and N. Memmel, Phys. Rev. B 49, 5599 1994 . 3 S. Mušller, P. Bayer, C. Reischl, K. Heinz, B. Feldmann, H. 16 J. Shen, J. Giergiel, A. K. Schmid, and J. Kirschner, Surf. Sci. Zillgen, and M. Wuttig, Phys. Rev. Lett. 74, 765 1995 . 328, 32 1995 . 4 Dongqi Li, M. Freitag, J. Pearson, Z. Q. Qiu, and S. D. Bader, 17 A. K. Schmid, D. Atlan, H. Itoh, B. Heinrich, T. Ichinokawa, and Phys. Rev. Lett. 72, 3112 1994 . J. Kirschner, Phys. Rev. B 48, 2855 1993 . 5 J. Giergiel, J. Shen, J. Woltersdorf, A. Kirilyuk, and J. Kirschner, 18 E. R. Moog and S. D. Bader, Superlattices Microstruct. 1, 543 Phys. Rev. B 52, 8528 1995 . 1985 . 6 R. D. Ellerbrock, A. Fuest, A. Schatz, W. Keune, and R. A. 19 J. Giergiel, H. Hopster, J. M. Lawrence, J. C. Hemminger, and J. Brand, Phys. Rev. Lett. 74, 3053 1995 . Kirschner, Rev. Sci. Instrum. 66, 3474 1995 . 7 M. Straub, R. Vollmer, and J. Kirschner, Phys. Rev. Lett. 77, 743 20 Ch. V. Mohan, J. Shen, P. Ohresser, M. Klaua, and J. Kirschner 1996 . unpublished . 8 B. Gubanka, M. Donath, and F. Passek, Phys. Rev. B 54, R11 153 21 J. Kohlhepp, H. J. Elmers, S. Cordes, and U. Gradmann, Phys. 1996 . Rev. B 45, 12 287 1992 . 9 H. Magnan, D. Chandesris, B. Villette, O. Heckmann, and J. 22 M. Straub, R. Vollmer, and J. Kirschner unpublished . Lecante, Phys. Rev. Lett. 67, 859 1991 . 23 P. A. Serena and N. GarciŽa, Phys. Rev. B 50, 944 1994 . 10 M. Wuttig, B. Feldmann, J. Thomassen, F. May, H. Zillgen, A. 24 S. S. Manoharan, J. Shen, M. Klaua, H. Jenniches, and J. Kir- Brodde, H. Hannemann, and H. Neddermeyer, Surf. Sci. 291, 14 schner, J. Appl. Phys. 81, 3768 1997 . 1993 . 25 Y. Tsunoda, N. Kunitomi, and R. M. Nicklow, J. Phys. F 17, 11 K. Kalki, D. D. Chambliss, K. E. Johnson, R. J. Wilson, and S. 2447 1987 . Chiang, Phys. Rev. B 48, 18 344 1993 . 26 M. Wuttig and J. Thomassen, Surf. Sci. 282, 237 1993 ; 12 V. L. Moruzzi, P. M. Marcus, and J. Kubler, Phys. Rev. B 39, 27 S. Mušller, P. Bayer, A. Kinne, P. Schmailzl, and K. Heinz, Surf. 6957 1989 ; C. S. Wang, B. M. Klein, and H. Krakauer, Phys. Sci. 322, 21 1995 . Rev. Lett. 54, 1852 1985 . 28 A. Kirilyuk, J. Giergiel, J. Shen, M. Straub, and J. Kirschner, 13 M. Zharnikov, A. Dittschar, W. Kuch, C. M. Schneider, and J. Phys. Rev. B 54, 1050 1996 .