PHYSICAL REVIEW B
`
`VOLUME 56, NUMBER 17
`
`1 NOVEMBER 1997-I
`
`Structural and magnetic phase transitions of Fe on stepped Cu(cid:132)111(cid:133)
`
`J. Shen, M. Klaua, P. Ohresser, H. Jenniches, J. Barthel, Ch. V. Mohan, and J. Kirschner
`Max-Planck-Institut fu¨r Mikrostrukturphysik, Weinberg 2, 06120 Halle, Germany
`共Received 19 May 1997兲
`
`The magnetism and its correlation with morphology and structure of ultrathin Fe/Cu共111兲 films have been
`studied. At room temperature, the films grow in a quasi-one-dimensional form 共stripes兲 in the submonolayer
`range. Between 1.4 and 1.8 ML the stripes percolate and become two-dimensional films. The remanent
`magnetization of the percolated films was observed to be significantly more stable with respect to time than
`that of the stripes. At low thickness 共⬍2.3 ML兲 the films adopt the fcc structure from the substrate and later
`transform to bcc共110兲 structure with Kurdjumov-Sachs orientation. Experimental evidence suggests that the fcc
`films have a low-spin ferromagnetic or ferrimagnetic phase, and a perpendicular easy magnetization axis. The
`magnetization switches to an in-plane high-spin phase after the fcc to bcc structural transformation has been
`accomplished. 关S0163-1829共97兲02541-1兴
`
`I. INTRODUCTION
`
`The step decoration effect of Fe growth on Cu共111兲 has
`been noticed for some years since the first scanning tunnel-
`ing microscopy 共STM兲 study on this system.1 Recently this
`effect has been used by the authors to produce one-
`dimensional 共1D兲 Fe stripes on a stepped Cu共111兲 substrate.2
`It has been found that in the submonolayer regime the de-
`posited Fe atoms form parallel stripes along the step edges.
`The magnetism of the stripes has a superparamagnetic nature
`which is distinguished from that of a two-dimensional 共2D兲
`ferromagnetic film mainly by its time-dependent remanent
`magnetization. Further increase in the thickness leads to an
`increase of the width of the stripes and finally results in the
`two-dimensional percolation of the films. This system has
`provided an opportunity to study the transition from 1D to
`2D magnetism.
`In addition the Fe/Cu共111兲 films have fcc structure in the
`low thickness limit,3,4 which otherwise does not exist below
`1100 K in bulk. The magnetism of fcc ␥-iron appears to be
`very interesting. Theoretical calculations have shown that fcc
`␥-iron may be nonmagnetic, antiferromagnetic, low-spin, or
`high-spin
`ferromagnetic
`depending
`on
`the
`lattice
`parameters.5 Though experimental evidence from the bulk
`indicates that fcc ␥-Fe has a antiferromagnetic ground state,6
`the ultrathin films of ␥-Fe on Cu共100兲 were observed to have
`a
`high-spin
`ferromagnetic
`phase
`at
`low thickness
`共⬍5 ML兲.7 This high-spin ferromagnetic 共FM兲 phase has
`been associated with the tetragonal distortion of the Fe fcc
`structure leading to an enlarged atomic volume.8 To find out
`whether the high-spin FM phase is a general feature of 2D
`␥-Fe 共ultrathin films兲, it is important to study the magnetic
`properties of ␥-Fe on a copper substrate with another orien-
`tation, such as the 共111兲 plane.
`In contrast to the Fe/Cu共100兲 system, the structure and the
`magnetism of the Fe/Cu共111兲 system has been much less
`understood due to the very limited amount of work. So far it
`has been reported that at room temperature Fe grows pseudo-
`morphically up to 4 or 5 monolayers,4 and then transforms to
`the bcc structure with 共110兲 orientation.4,9–13 STM has
`shown that the growth of Fe on Cu共111兲 is in a three-
`
`dimensional manner at room temperature.1,14 Magnetic mea-
`surements from a copper capped Fe/Cu共111兲 film suggested
`that
`the Fe has a surprisingly small magnetic moment
`共⬃0.6␮B兲.11 The easy magnetization direction, however, ap-
`peared to depend strongly on the growth temperature.11 An
`electron-capture spectroscopy study has reported the exis-
`tence of long-ranged ferromagnetic order in a 4-ML film and
`short-ranged ferromagnetic order in one- or two-monolayer
`films.15
`The aim of this work is to systematically study the struc-
`tural and magnetic properties of the Fe/Cu共111兲 films with
`emphasis on the correlation between the structure and their
`magnetism. In situ STM,
`low-energy electron diffraction
`共LEED兲 with intensity vs energy function 共I/V LEED兲 and
`magneto-optical Kerr effect 共MOKE兲 have been used to
`study the thickness dependence of the morphology, atomic
`structure, and magnetism, respectively. The main results are
`summarized as follows.
`共1兲 In the submonolayer range the Fe films have a
`quasi-1D form which is characterized by parallel aligned
`stripes. The remanent magnetization of the stripes was ob-
`served to be time dependent.
`共2兲 Between 1.4 and 1.8 ML the Fe stripes percolate to
`form 2D films. The magnetization relaxation time of the 2D
`films is significantly longer than that of the 1D films.
`共3兲 We proved the occurrence of the fcc to bcc transition.
`The critical thickness is between 2.3 and 2.7 ML, which is
`somewhat lower than the previously reported 4 or 5 ML.4
`共4兲 The easy magnetization axis switches from perpen-
`dicular to in-plane between 2.3 and 3.0 ML, in coincidence
`with the fcc to bcc structural transition.
`共5兲 In the fcc to bcc transformation regime, the magneti-
`zation of the Fe films sharply increases. The magnetization
`of the transformed bcc Fe films is about four to five times
`larger than the extrapolated value from the low thickness fcc
`Fe films.
`The paper is organized as follows. We describe the ex-
`perimental details in Sec. II. Sections III and IV present the
`structural and the magnetic results, respectively. The corre-
`lation between the observed structural and magnetic proper-
`
`0163-1829/97/56共17兲/11134共10兲/$10.00
`
`56
`
`11 134
`
`© 1997 The American Physical Society
`
`TDK Corporation Exhibit 1011 Page 1
`
`

`
`56
`
`STRUCTURAL AND MAGNETIC PHASE TRANSITIONS . . .
`
`11 135
`
`The well-ordered atoms reflect the sixfold symmetry of the
`substrate surface, proving that the substrate surface is virtu-
`ally defect free in the terraces.
`The Fe films were prepared in the analysis chamber from
`an iron wire 共5N in purity兲 heated by e-beam bombardment.
`At a typical Fe evaporation rate of 0.2 ML/min the pressure
`increased from 5⫻10⫺11 mbar to 1⫻10⫺10 mbar. To sup-
`press interdiffusion, the copper substrate was kept at 0 °C
`during deposition.17 Afterwards the sample was further
`cooled down to 170 K in order to avoid a temperature rise
`above 0 °C during transferring from the analysis chamber to
`the MOKE chamber.
`In the MOKE chamber, the sample was placed on a ma-
`nipulator with a function of three-axis movement as well as
`azimuthal rotation. The sample can be cooled down to 40 K
`and heated up to 600 K. In the present work most of the
`measurements were carried out in the sample temperature
`range between 90 and 270 K. Since the magnetization of
`nearly all the films persists above 270 K at which the inter-
`diffusion starts to proceed, no attempts have been made to
`measure the Curie temperature in the present study. The po-
`lar and in-plane Kerr measurements are easily achieved by
`rotating the manipulator to allow the sample surface to be
`perpendicular or parallel to the external field. The maximum
`field is about 0.8 T, which appears to be large enough to
`saturate the films with thickness below 5 ML along the hard-
`axis direction.
`
`FIG. 1. STM topography image of a well-prepared Cu共111兲
`vicinal surface. The atomic steps are parallel and oriented along
`具011典 direction. The inset
`is an atomically resolved image (I t
`⫽0.25 nA, V bias⫽10 mV兲 recorded from a terrace showing the six-
`fold symmetry of the Cu共111兲 surface.
`
`ties is discussed in Sec. V. The final conclusion will be
`drawn in Sec. VI.
`
`II. EXPERIMENTAL DETAILS
`
`III. MORPHOLOGY AND STRUCTURE
`
`The experiments were performed in a multichamber sys-
`tem including an MBE preparation chamber, an STM cham-
`ber, an analysis chamber equipped with facilities for Auger
`electron spectroscopy 共AES兲, LEED and thin-film growth,
`and a MOKE chamber. The base pressure of the individual
`chambers is better than 5⫻10⫺11 mbar. The AES system has
`a cylindrical mirror analyzer 共CMA兲 allowing the detecting
`limit to about 0.1 at. % for most of the elements. A fully
`automatic video-LEED system16 has been used for recording
`LEED images as well as for measuring I/V LEED curves. In
`the present work the sample was prepared in the analysis
`chamber. The LEED and the AES measurements were taken
`immediately after film deposition. Then the sample was
`transferred to the MOKE chamber for magnetic measure-
`ments. STM images were recorded at room temperature after
`the MOKE study.
`The copper substrate has a miscut of 1.2° with respect to
`the 共111兲 orientation. The surface steps are aligned along one
`具011典 direction with a 共001兲 microfacet. The average terrace
`width is about 10 nm. Prior to film preparation the copper
`substrate was cleaned by Ar⫹ sputtering and annealing
`cycles to 700 K. The crystallographic quality and the clean-
`liness of the substrate were monitored by LEED and AES,
`respectively. Both sharp LEED p共1⫻1兲 spots and the
`contamination-free Auger spectrum indicated the high qual-
`ity of the substrate. The surface morphology of the substrate
`has also been examined by STM. As shown in Fig. 1, the
`surface consists of 具011典 oriented steps with an average ter-
`race width of about 10 nm. The inset shows a magnified
`image with atomic resolution taken from a terrace region.
`
`Figure 2 shows a series of STM topography images of
`Fe/Cu共111兲 films of different thickness. It is immediately
`visible that in the submonolayer regime 共0.3 ML, 0.8 ML兲
`the films have a quasi-1D form. The stripes are aligned par-
`allel to one another on the upper edges of the steps. Most of
`the regions of the stripes have a monolayer height at 0.3 ML
`and increase to a double-layer height at 0.8 ML. At 0.3 ML
`the amount of Fe atoms is so small that the stripes are virtu-
`ally formed by segments which are only weakly linked. At
`0.8 ML, most of the segments have coalesced and the stripes
`become much more continuous. The edges of the stripes are
`rather rough. This is because the Fe atoms have a tendency
`towards growth along all three 具011典 directions even though
`they have a preferential alignment along the direction of the
`step edges. In other words, the Fe stripes are virtually formed
`by triangular-shaped islands which are linearly arrayed and
`connected. Such kind of construction of the stripes appears
`to have some strong influence on their magnetic properties,
`in particular the magnetization relaxation process which will
`be discussed in Sec. V.
`With increasing thickness the stripes become wider and
`finally percolate between 1.4 and 1.8 ML, as seen in Figs.
`2共c兲 and 2共d兲. This percolation leads to a direct connection
`between most of the stripes 共⬎90%兲. In the present work we
`refer such a percolation as a 2D percolation while the con-
`nection between the segments within each stripe, i.e., be-
`tween 0.3 and 0.8 ML, will be named 1D coalescence. Here
`the 2D percolation, in reality, leads to a 1D to 2D morpho-
`logical transition of the films. The effect of this transition on
`the magnetism will also be discussed in Sec. V.
`
`TDK Corporation Exhibit 1011 Page 2
`
`

`
`11 136
`
`J. SHEN et al.
`
`56
`
`FIG. 3. LEED pattern of the Fe/Cu共111兲 films with different
`thickness. Below 2.3 ML the films have a p共1⫻1兲 pattern inherited
`from the substrate. From 2.3 ML on, satellite spots start to appear
`around the substrate spots. These satellite spots are caused by the
`bcc共110兲 domains with Kurdjumov-Sachs orientation.
`
`ented along the step direction. But those along the other two
`具011典 directions gradually develop to become almost equally
`important 共see the 9.0 ML image in Fig. 2兲.
`The question immediately arises: what causes the changes
`of the film morphology in the thickness regime between 2.3
`and 2.7 ML? Since the triangular shape of the islands is a
`reflection of the fcc共111兲 symmetry, losing this feature might
`imply losing the fcc共111兲 structure. Indeed, our LEED stud-
`ies have proven that the films undergo a fcc to bcc structural
`transition between 2.3 and 2.7 ML. Figure 3 shows the
`LEED patterns taken from different films at about 65 eV.
`The 1.4-ML and the 1.8-ML films have p共1⫻1兲 patterns with
`respect to the clean copper substrate, indicating the pseudo-
`morphic growth of the Fe films. Because of the increased
`roughness owing to the multilayer growth, the films show
`more diffuse LEED spots than the copper substrate. With
`increasing thickness some additional spots start to appear
`around the substrate spots. At 2.3 ML these spots are already
`visible though their intensity is very weak. At 2.7 ML or
`even higher thickness the LEED patterns are already featured
`by these additional spots. In fact, these satellite spots are
`better seen at the beam energy of about 90 eV, as shown in
`the upper-left picture of Fig. 4 for the 2.7-ML film. There
`appear to be five satellite spots around each substrate spot,
`with three in the inner circle and two in the outer circle. It
`
`FIG. 2. Series of STM images showing the morphological evo-
`lution of the Fe/Cu共111兲 films with increasing thickness. In the
`submonolayer regime 关共a兲 and 共b兲兴 the films have quasi-1D form.
`The 2D percolation occurs between 1.4 ML 共c兲 and 1.8 ML 共d兲. The
`morphology changes greatly between 2.3 ML 共e兲 and 2.7 ML 共f兲, as
`characterized by the formation of elongated domains aligning along
`the 具011典 directions.
`
`Up to 2.3 ML the film morphology is characterized by
`triangular-shaped islands reflecting the threefold symmetry
`of the fcc 共111兲 structure. As the islands are mostly bilayer or
`trilayer in height, the growth of the Fe/Cu共111兲 can be gen-
`erally referred to as multilayer growth. Above 2.3 ML, the
`morphology of the films, however, changes significantly.
`The number of the exposed layers increases. It is 2 to 3 at 2.3
`ML, 5 to 6 at 2.7 ML, and 7 to 8 at 9 ML. Moreover, the
`triangular-shaped islands are no longer visible in the images
`of the films with higher thickness 关Figs. 2共e兲–2共h兲兴. Instead
`some elongated domains appear which are aligned along the
`具011典 directions. At 2.7 ML these domains are mainly ori-
`
`TDK Corporation Exhibit 1011 Page 3
`
`

`
`56
`
`STRUCTURAL AND MAGNETIC PHASE TRANSITIONS . . .
`
`11 137
`
`FIG. 4. Demonstration of the Kurdjumov-Sachs superstructure.
`Upper left: LEED pattern of the 2.7-ML film at the beam energy of
`90 eV. Upper right: Simulated LEED pattern of Fe bcc共110兲 on
`Cu共111兲 with Kurdjumov-Sachs orientation. Note the consistency
`between the experimental and the simulated LEED patterns. Bot-
`tom: real-space schematic view of the six KS domains on the
`Cu共111兲 substrate.
`
`has already been discussed in previous work that these spots
`result from the bcc共110兲 domains in Kurdjumov-Sachs 共KS兲
`orientation.4 The KS orientation is a special case of the one-
`dimensional matching between bcc 共110兲 and fcc共111兲 共bcc
`关1¯11兴 is parallel to fcc 关101¯兴兲 with sides of the rhombic unit
`meshes of the film and the substrate to be parallel. Using the
`bulk parameters of Fe and Cu, we have simulated the LEED
`pattern of the KS orientation which is shown in the upper-
`right of Fig. 4. Except the middle two spots in the inner
`circle of the simulated pattern which are too close to be
`resolved experimentally, the experimental and the simulated
`LEED patterns agree well with each other. The six satellite
`spots reflect the six kinds of atomic relationship between the
`bcc共110兲 and fcc共111兲 structure as shown in the bottom pic-
`ture of Fig. 4. Since these KS domains have only a one-
`dimensional matching with the substrate, each domain tends
`to have a narrow width in order to reduce the stress. In prin-
`ciple these six domains should have no preference on a flat
`surface due to the sixfold symmetry of the surface. Owing to
`the mirror symmetry, there are two kinds of domains along
`each 具011典 direction. As a result, the morphology of the films
`could be viewed as ridgelike structures oriented along the
`three 具011典 directions on a similar number. However, the
`STM images have shown that there are more domains elon-
`gated along the step direction than the other two 具011典 direc-
`
`FIG. 5. IV-LEED spectrum of the 共00兲 beam for the Fe/Cu共111兲
`films. The 0’s on the y axis indicate the zero level of the intensity of
`the corresponding curves after background subtraction. Below 2.3
`ML there is no distinct peak shift. At higher thickness the peaks
`shift towards higher energies. The dashed lines indicate the peak
`position of the substrate.
`
`tions 关see the 2.7-ML image in Fig. 2共f兲兴. This can be ex-
`plained by the fact that the substrate steps are oriented only
`in one of the 具011典 directions and therefore have introduced
`a twofold symmetry in addition to the sixfold symmetry of
`the fcc共111兲 surface. The two domain configurations which
`are parallel to the steps should be more favored as compared
`to the other four domains. The dominance of the two particu-
`lar KS domains should, in principle, make the corresponding
`satellite spots have different intensity as compared to the rest
`satellite spots. However, as a result of the limited resolution
`and the rather rough surface, it is difficult to conclude this
`from our LEED data.
`The structural transition from fcc to bcc is further ana-
`lyzed by an IV-LEED study of the system. Figure 5 shows
`the 共00兲 beam intensity vs energy curves for films with dif-
`ferent thickness. These curves were obtained with the inci-
`dent angle of the primary beam to be about 6° off the surface
`normal. In this case the step direction of the substrate is
`perpendicular to the plane of incidence. The dashed lines in
`Fig. 5 mark the peak position of the copper substrate. It is
`evident that with increasing thickness up to 2.3 ML there is
`no major shift of the peak position. There seems to be only
`one family of peaks inherited from the substrate 共fcc兲, which
`is in stark contrast to the Fe/Cu共100兲 system where two fami-
`lies of peaks have been observed corresponding to fcc and
`fct, respectively.8 Above 2.7 ML, the peaks have clearly
`shifted towards higher energies. Such a shift, within the ki-
`nematic approximation, indicates that the average interlayer
`distance of the films becomes smaller than that of the sub-
`strate. Using the kinematical model, we have calculated the
`
`TDK Corporation Exhibit 1011 Page 4
`
`

`
`11 138
`
`J. SHEN et al.
`
`56
`
`mation 共about 2.7 ML兲 most of the bcc domains are oriented
`along the step direction. The number of the bcc domains
`oriented along the other two 具011典 directions only become
`significant at higher thickness 共e.g., see the 5.7-ML image in
`Fig. 2兲. Therefore, it is not surprising that the Fe films on the
`vicinal Cu共111兲 substrate transform into bcc structure at
`lower thickness than the films on a flat Cu共111兲 substrate.18
`
`IV. MAGNETIC PROPERTIES
`
`As shown in Fig. 2, in the submonolayer regime the
`Fe/Cu共111兲 films have a quasi-1D form. In our previous po-
`lar MOKE measurements we have shown that
`these Fe
`stripes exhibit hysteresis with a coercivity depending
`strongly on the temperature.2 The easy magnetization axis of
`the stripes is determined to be perpendicular to the surface,
`because no magnetic signal was detected in the longitudinal
`geometry, irrespective of the external field being parallel or
`perpendicular to the stripes. In the same work we have also
`performed time-dependent magnetization measurements by
`MOKE. The results obtained from the stripes, which has a
`submonolayer nominal coverage 共⬃0.8 ML兲, showed that at
`zero field the magnetization decays at a speed strongly de-
`pending on the temperature. For example, while the magne-
`tization of the 0.8-ML stripes decays very slowly at about
`100 K, it decays rapidly to zero 共within 20 sec兲 at 160 K. We
`associated the decay of the magnetization with the superpara-
`magnetic nature of the 1D stripes. The fact that the super-
`paramagnetic stripes exhibit hysteresis is simply because the
`measuring time of the magnetization curves is considerably
`shorter than the magnetization relaxation time of the stripes.
`Here we extend the previous work by demonstrating the
`coverage-dependent magnetization relaxation process of the
`Fe/Cu共111兲 films. Figure 7 shows the magnetization relax-
`ation curves of films with different thickness ranging from
`that of the 1D stripes to the 2D percolated film. All the
`curves were measured in the polar geometry. The 0 level of
`each curve indicates the signal of a demagnetized state of the
`films. An external field of about 0.6 T was applied at the
`time point which is marked as ‘‘field on.’’ For all the films
`the magnetization quickly reaches saturation. The external
`field was removed at the time point marked as ‘‘field off.’’ In
`all three cases the magnetization falls down, but with dis-
`tinctly different speed. At 160 K, the magnetization of the
`0.8-ML film decays down to zero within 20 sec. In contrast,
`at the same temperature after an initial quick decay the re-
`sidual magnetization of the 2-ML film is much more stable.
`The quickest decay of the magnetization is observed, how-
`ever, for the 0.3-ML film. Its magnetization decays rapidly to
`zero in a few seconds even at much lower temperature of 50
`K. The different behavior of the magnetization relaxation
`process of these films appears to be closely correlated with
`their morphology, which will be discussed in the next sec-
`tion.
`We now turn to consider the magnetic anisotropy of the
`Fe/Cu共111兲 films. As mentioned above, the easy magnetiza-
`tion axis of the submonolayer films 共the stripes兲 is perpen-
`dicular to the film surface. This in fact has been found to be
`the case for all the films below 2.3 ML of thickness. Figure
`8 shows the magnetization curves of films measured in both
`polar 共a兲 and longitudinal 共b兲 geometry. It is evident that
`
`FIG. 6. The calculated interlayer distance from Fig. 5 using the
`kinematic approximation. It is evident that the interlayer distance
`decreases above 2.3 ML. The vertical dashed line separates films
`with different LEED structure. The films have a fcc-like structure
`below 2.5 ML, and a bcc structure above.
`
`interlayer distance as a function of the film thickness. The
`results are shown in Fig. 6. Two regions can be clearly dis-
`tinguished as indicated by the dashed line: in the low thick-
`ness region where the films show a p共1⫻1兲 pattern the inter-
`layer distance of the films is close to that of the copper
`substrate, while in the high thickness region the interlayer
`distance of the films becomes apparently smaller. However,
`it has to be noted that the accuracy of the absolute value by
`our determination is influenced by the simple kinematic
`model, as the deviation of the calculated interlayer distance
`from the corresponding bulk value is about 0.001 and 0.003
`nm for the fcc共111兲 and bcc共110兲, respectively. Nevertheless,
`the fact that the interlayer distance of the films drops down
`between 2.3 and 2.7 ML by more than 0.002 nm is quite
`unambiguous. This decrease of the interlayer distance is con-
`sistent with the LEED patterns which indicate the fcc to bcc
`structural transition between 2.3 and 2.7 ML.
`The observed critical thickness of the fcc!bcc transfor-
`mation is smaller than the previously reported value4 of
`about 4 or 5 ML. This might be due to the fact that a vicinal
`copper substrate is used in the present work. Compared to
`the growth of Fe on a flat substrate, at the same nominal
`thickness the step decoration effect of the Fe growth on the
`vicinal substrate could make the films be locally thicker
`along the step edges. The fcc!bcc transformation could pro-
`ceed first in these locally thicker regions. In addition, on the
`vicinal surface the islands have an elongated shape along one
`of the 具011典 directions 共the step direction兲 even before the
`structural phase transformation. These elongated islands may
`have an easier path to convert into bcc KS domains as com-
`pared to the normal
`triangular-shaped islands on a flat
`Cu共111兲 surface. This is in fact backed by our STM obser-
`vations that the bcc domains along the step direction appears
`earlier than those along the other two 具011典 directions. In
`Fig. 2 we have shown that at the beginning of the transfor-
`
`TDK Corporation Exhibit 1011 Page 5
`
`

`
`56
`
`STRUCTURAL AND MAGNETIC PHASE TRANSITIONS . . .
`
`11 139
`
`the
`the magnetization of
`FIG. 7. Time dependence of
`Fe/Cu共111兲 films at different thickness. At the same temperature of
`160 K, after removal of the field the decay of the magnetization of
`the 2-ML film is significantly slower than that of the 0.8-ML film.
`The magnization of the 0.3-ML film decays rapidly to zero at an
`even lower temperature of 50 K.
`
`below 2.3 ML the films exhibit magnetic signals only in the
`polar geometry, indicating that the easy magnetization direc-
`tion is perpendicular to the surface. At 160 K, the polar hys-
`teresis loops of the films 共⬍2.3 ML兲 are usually character-
`ized by small remanence and coercivity. On the scale of the
`lower panel of Fig. 8共a兲, the hysteresis of the loops is virtu-
`ally invisible. The loop of the 2.0-ML film has been magni-
`fied in the upper panel of Fig. 8共a兲, where a clear hysteresis
`can be seen. With increasing thickness, the saturation field
`increases from 400 Oe for the 2.0-ML film to 850 Oe for the
`2.5-ML film. The 2.3- and 2.5-ML films, as shown in Fig.
`8共b兲, start to exhibit weak magnetic signals in the in-plane
`geometry. Further increasing thickness up to 3.0 ML, the
`in-plane magnetization curves become much more pro-
`nounced with a near rectangular shape. The perpendicular
`magnetization, on the other hand, becomes much harder to
`saturate. The saturation field for the 3.0-ML film is about
`4500 Oe, a value which is almost five times larger than that
`of the 2.5-ML film. This indicates that at a thickness of 3.0
`ML or higher the easy magnetization axis lies parallel to the
`surface. We therefore conclude that the easy magnetization
`direction switches from perpendicular to in plane in the
`thickness range between 2.3 and 3.0 ML.
`The films with in-plane
`easy magnetization axis
`共⬎3.0 ML兲 have the bcc structure with KS orientation. With-
`out the step-induced twofold symmetry, the magnetization
`curves should be identical if measured along the three 具011典
`directions. Due to the existence of the well-aligned steps, the
`step direction is no longer equivalent to the other two 具011典
`directions. We have observed 共not shown here兲 that it re-
`quires a larger field to saturate the films along the step direc-
`tion than along the other two 具011典 directions. At this stage it
`is not possible for us to determine the anisotropy direction in
`
`FIG. 8. The magnetization curves of films with increasing thick-
`ness. 共a兲 Polar curves. 共b兲 In-plane curves. Below 2.3 ML, the mag-
`netic signal can only be detected in the polar geometry. The upper
`panel in 共a兲 shows a magnified loop of the 2-ML film. When in-
`creasing thickness from 2.5 ML to 3.0 ML, the saturation field in
`the perpendicular direction greatly increases by a factor of 5. The
`in-plane signal starts to appear at 2.3 ML, and at 3.0 ML already
`develops to a well-defined rectangular loop.
`
`the plane because of the limited azimuthal angles we were
`able to cover presently. Instead, we can conclude that the
`easy magnetization axis is at least not along the step direc-
`tion.
`We also want to point out a remarkable change of the
`magnetism in the thickness range from 2.3 to 3.0 ML, that is
`the rapid increase of the saturation Kerr signal. Figure 9
`
`TDK Corporation Exhibit 1011 Page 6
`
`

`
`11 140
`
`J. SHEN et al.
`
`56
`
`cence. Its stripes are formed by short segments which are
`weakly linked or even disconnected. The 0.8-ML film, how-
`ever, has a thickness above 1D coalescence, and the stripes
`are much more continuous. In the previous work we have
`already discussed the superparamagnetic nature of these
`quasi-one-dimensional stripes.2 Each stripe is divided into
`some superparamagnetic spin blocks. Although inside each
`block the spins are ferromagnetically coupled,
`the spin
`blocks are more or less independent from each other with
`only weak coupling between them. The volume of the spin
`blocks (V b) can be obtained by fitting the measured hyster-
`esis curves. The fitted V b of the 0.3-ML film was found to be
`close to the volume of the segments. This is not surprising
`since the links between the segments are weak. The some-
`what surprising result is that the fitted V b of the 0.8-ML film
`was found only about four times larger than that of the
`0.3-ML film, though the stripes of the 0.8-ML film are rather
`continuous. We believe this is mainly due to the rough edge
`of the stripes, which results in some weak connections along
`the stripes. According to the Ne´el model,19 the superpara-
`magnetic relaxation time exponentially increases with the in-
`crease of the volume of the block unit. This explains why the
`relaxation time of the 0.8-ML film is longer than that of the
`0.3-ML film, even though the later was measured at a much
`lower temperature of 50 K. The 2-ML film, on the other
`hand,
`is already a percolated 2D film, since the two-
`dimensional percolation occurs between 1.4 and 1.8 ML. The
`2D percolation of the films may lead to a significantly more
`stable remanent magnetization by two possible mechanisms:
`共1兲 the films are still superparamagnetic but with a much
`longer relaxation time; 共2兲 a true long-range order has al-
`ready been built up in the films. It is not clear which mecha-
`nism holds in this particular case, though the fact that the
`remanence of the 2-ML film slowly decays may imply a
`superparamagnetic nature of the film. Nevertheless, the is-
`land size of the 2-ML film should be much larger than that of
`the other two films because of the 2D percolation. Therefore
`the 2-ML film has the longest relaxation time among the
`films shown in Fig. 7.
`Despite of their superparamagnetic nature, all the films
`below 2.3 ML have an easy magnetization axis perpendicular
`to the surface. At the low thickness limit the sign of the
`effective anisotropy constant is usually governed by the sur-
`face anisotropy constant (K s). This implies that the surface
`anisotropy of the Fe/Cu共111兲 system favors perpendicular
`magnetization. With increasing thickness above 2.3 ML the
`spin reorientation starts to take place. The origin of this spin
`reorientation may come from two factors. The first one is the
`interplay between the positive surface anisotropy and the
`negative shape anisotropy. Since the surface anisotropy re-
`mains constant with increasing thickness while the shape an-
`isotropy increases, it is expected that above a certain critical
`thickness the easy magnetization axis will switch from per-
`pendicular to in plane. The second one is the structural trans-
`formation from fcc共111兲 to bcc共110兲, which occurs in almost
`the same thickness regime as the spin reorientation. The two-
`fold in-plane symmetry of the bcc共110兲 plane provides an
`in-plane anisotropy in addition to the usual perpendicular
`anisotropy. The film thus may have a in-plane easy magne-
`tization axis once the bcc共110兲 structure is attained. Such fcc
`to bcc structural transformation-induced spin reorientation
`
`FIG. 9. The saturation magnetization as a function of film thick-
`ness. The measuring temperature is about 160 K. The magnetization
`abruptly increases about four to five times at the crossover from
`region I to II which is indicated by the vertical line. The vertical
`line marks the thickness at which the LEED pattern changes from
`fcc to bcc structure. The magnetization of both the fcc Fe films and
`the bcc Fe films increases linearly with thickness. Note the slope of
`the bcc films’ magnetization 共dotted line兲 to be about five times
`larger than that of the fcc films 共dashed line兲.
`
`shows the saturation Kerr signal as a function of the film
`thickness. Note all the films were saturated along the surface
`normal no matter whether their easy magnetization axis is
`perpendicular or parallel to the surface. Below 2.3 ML, the
`Kerr signals of the films have small values whose linear ex-
`trapolation 共dashed line兲 extends approximately through
`zero. With increasing thickness the Kerr signals of the films,
`in principle, should lie on the dashed line if the magnetic
`moment of the films keeps constant. However, Fig. 9 clearly
`indicates that between 2.3 and 3.0 ML the measured Kerr
`signals increase much quicker than just linearly. Structurally
`this thickness range is exactly that of the fcc to bcc transfor-
`mation. After the completion of the structural transformation
`共⬎3.0 ML兲, the Kerr signals of the bcc films increases again
`more or less linearly as indicated by the dotted line. The
`slope of the dotted line, however, is about five times larger
`than that of the dashed line. This means that the Kerr signals
`of the bcc films are about five times larger than expected
`from the linear extrapolation of the dashed line. The origin of
`this unusual increase of the Kerr signal as well as the spin
`reorientation will be discussed in the following