PHYSICAL REVIEW B VOLUME 55, NUMBER 1 1 JANUARY 1997-I Magnetic interface formation at Cr/Fe 100... and Fe/Cr/Fe 100...: Magnetic dichroism in photoemission study Giancarlo Panaccione* and Fausto Sirotti Laboratoire pour l'Utilisation du Rayonnement Electromagnetique, CNRS-CEA-MESR, F-91405 Orsay, France Elisabetta Narducci Instituto Nazionale di Fisica della Materia, Dipartimento di Fisica, Universita di Genova, I-16146 Genova, Italy Giorgio Rossi Laboratorium fušr Festkošrperphysik, Eidgenošssische Technische Hochschule­Zušrich, CH-8093 Zušrich, Switzerland and Instituto Nazionale di Fisica della Materia, Dipartimento di Fisica, Universita di Modena, I-41110 Modena, Italy Received 24 May 1996 The early stages of the growth of Cr/Fe 100 and Fe/Cr/Fe 100 interfaces have been investigated by magnetic dichroism in photoemission of Fe 3p and Cr 3p core levels as measured from chiral experiments employing linearly polarized synchrotron radiation. Evidence is obtained for a 30% larger magnetic moment of interface Cr atoms with respect to Cr atoms belonging to epitaxial ultrathin films and a 40% magnetic moment enhancement of top Fe interface atoms in the Fe/Cr/Fe 100 trilayer. The kinetic growth conditions 450 K lead to a uniform overlayer growth, without intermixing, but dominated by islanding. As a consequence the formation of a single-surface ferromagnetic domain for Fe/Cr/Fe 100 is frustrated up to two Fe monolayer ML thickness. The line shape of Fe 3p photoemission in the frustrated regime is consistent with the presence of in-plane magnetic order at 90° with respect to the substrate magnetization direction. The appearance of photoemission magnetic dichroism for Fe overlayer thicknesses exceeding 2 ML is interpreted as due to domain rotation towards the direction antiparallel to the Fe substrate magnetization. S0163-1829 97 03601-1 I. INTRODUCTION nary temperature and Cr is AF with a slightly incommensu- rate AF structure oriented along the 100 direction.4 Cr is Stacking of alternate layers of Cr and Fe makes an artifi- often defined as a layered AF solid since it can be viewed as cial solid with intrinsically anisotropic electronic properties, a stacking of ferromagnetically ordered 100 planes, antifer- among which great importance is given to the giant magne- romagnetically coupled one to the next. Antiparallel cou- toresistance effect.1 As a function of the thickness of antifer- pling across the interface and a strong enhancement up to romagnetic AF Cr layers separating next-neighboring Fe sevenfold of the Cr surface magnetic moment, with respect layers the magnetic coupling between ferromagnetic FM Fe to the bulk Cr value of 0.59 B , were predicted for an or- layers is parallel or antiparallel. The oscillations of the cou- dered Cr monolayer on Fe 100 .5,6 Total energy calculations pling follow a long period and a short period which are for the Fe/Cr/Fe system showed that the number of the Cr thought to be related to the shape of the Fermi surface.2 layers dictates the parallel or antiparallel coupling between Due to the prototypical value of the Fe/Cr/Fe structure for the Fe overlayer and substrate separated by the Cr buffer.2,7,8 the understanding of magnetically dependent electron trans- The Fe magnetic moments in the Fe/Cr/Fe multilayers are port, a large number of experiments, models, and theoretical predicted to be similar to that of bulk Fe.8 descriptions have been produced in recent years. An overall A number of experimental results were apparently incon- agreement has been reached on the double periodicity of the sistent, being strongly influenced by the actual thicknesses of magnetic coupling and on the general behavior of the giant the layers grown in different experiments, by the growth magnetoresistance, but open questions and discrepancies re- conditions which determine the degree of epitaxial order and main in the detailed description of the magnetic behavior of of atomic mixing at the interface, and also by the incertitude the atoms at the interface between Fe 100 and Cr and be- of the methods used for measuring the magnetic moments. tween Fe and Cr 100 as the multilayer grows. In particular, Extreme values for the Cr magnetic moment at the open questions are the magnitude of the magnetic moments Cr/Fe 100 interface have been quoted from in situ magne- of Fe and Cr at and close to the interface and surface region, tometer measurements giving Cr 4 B for submonolayer and the magnetic coupling of the first layers, together with its thicknesses and Cr 3 B after completion of the first dependence on growth conditions, i.e., on the morphology, monolayer;9 in the same experiment, the authors inferred perhaps metastable, assumed by the interfaces at the early from their data a delayed onset of the Cr AF stacking. Spin- stages of formation. resolved photoemission and energy loss spectroscopies of Cr Cr and Fe have the same bcc lattice at room temperature at the interface with Fe have indicated Cr 1.8 B ,10 an and very similar lattice parameters 2.87 Ć and 2.88 Ć for Fe enhancement with respect to bulk Cr Ref. 11 or a moment and Cr, respectively3 , so that epitaxial growth is possible similar to bulk Cr.12 X-ray absorption dichroism experiments with either one or the other as a substrate. Fe is FM at ordi- on multiple interfaces showed basically the bulk value for 0163-1829/97/55 1 /389 8 /$10.00 55 389 © 1997 The American Physical Society 390 PANACCIONE, SIROTTI, NARDUCCI, AND ROSSI 55 Cr.13 The Fe magnetic properties at the Fe/Cr 100 and by the chirality of the experiment.28 Knowing that photoelec- Fe/Cr/Fe stacks are strongly influenced by the effective tron diffraction effects can strongly influence the obtained structural properties of the AF Cr substrate, as shown from dichroism,32 experimental results can be compared only magneto-optical Kerr effect results.14 Both enhancement or when measured in a fixed geometry and for a fixed photon reduction of the surface Fe magnetic moment are predicted, energy as done in the present work. 3 The presence of and some theoretical and experimental results suggested the magnetic order can be observed from dichroism experiments growth of a magnetically dead layer.13,15­17 Very clear spin- even if field averaging is done naturally by 180° domains or resolved electron microscopy images show the AF coupling by antiferromagnetic ordering. In fact the line shape of the of Fe layers across Cr layers of variable thickness, with field-averaged spectra depends on the alignment of the mag- monolayer resolution.18­20 These experiments also showed netic moments, i.e., of the magnetic core hole states, inde- the incommensurability of the Cr AF order with the lattice, pendently of the degree of polarization magnetization .33 In together with a ``defect'' in the layered antiferromagnetism this case variable chirality experiments can be performed in of Cr in the 0­4 atomic layers range. Moreover, bond frus- order to probe the existence of magnetic order. 4 The trations and interface roughness, depending on growth con- LMDAD spectra are determined by the energy splitting of ditions, can either suppress the short coupling period oscil- the magnetic core hole sublevels, which is proportional to lations in the Cr overlayers or modify the orientation of Fe the magnetic moment.34 overlayers, as proposed from theoretical calculations.15,21,22 Superconducting quantum interference device SQUID measurements and polarized neutron reflectivity showed for III. EXPERIMENT Cr/Fe superlattices the importance of biquadratic interlayer LMDAD experiments were performed on the Swiss- magnetic coupling, when a magnetic frustration is French beam line SU3 and on the SU7 beam line at the present,23,24 confirming the prediction of Ref. 20. The above SuperAco storage ring in LURE Orsay . In both cases, the results indicate that the structure of the real surfaces and electron energy analyzers were placed at 45° with respect to interfaces plays a fundamental role in the magnetic order and the direction of the linearly polarized synchrotron radiation coupling. Any experimental methods that integrate the infor- from standard planar undulators impinging onto the sample. mation on magnetic order and magnetic moment lead there- Angular acceptances were, respectively, of 1° SU3 and fore inevitably to gross errors. 22° SU7 . The overall energetic resolution was 100 The aim of the present work was to study the magnetic meV for the experiments using 120 eV photons SU3 and properties in the Cr/Fe 100 and Fe/Cr/Fe 100 interfaces 250 meV for 150 eV photons SU7 . with magnetic dichroism in photoemission. The advantage of All measurements presented in this paper were obtained photoemission experiments on magnetic interfaces comes using a 100 -oriented Fe 3% Si single crystal as a substrate, from the sensitivity to the chemical species and to the sur- mounted to close the gap of a soft iron yoke.26,34 The face atoms. By exploiting magnetic dichroism in chiral ex- Fe 100 surfaces were prepared by Ar -ion sputtering and periments with linearly polarized synchrotron radiation annealing cycles. In order to avoid the surface segregation of LMDAD Refs. 25 and 26 on core levels, one can take bulk impurities Si, C, and S the final iron surfaces were advantage, with respect to the previous spin-resolved experi- obtained either by a mild sputtering-annealing cycle or by ments, of the much higher counting rates and therefore of the homoepitaxy of a thin iron overlayer onto a well-ordered but better statistics which is attainable, allowing one to deepen C-segregated Fe 100 surface. Fe and Cr were evaporated by the interpretation by separating the effects of magnetic order electron bombardment from high purity rods, with a typical from the information on magnetic moments at a qualitative, deposition rate of 0.5 and 0.2 Ć/min, respectively, and in a but fruitful level.27,28 Cr 100 layers display dichroism in pressure below 2 10 10 mbar. The thickness of the deposit photoemission since the surface contributes approximately was monitored by a quartz crystal oscillator and verified by 1/3 of the total signal which is not completely averaged by the Cr 3p and Fe 3p photoemission intensities. Analysis of the exponentially damped underlayer contributions. low-energy electron diffraction LEED patterns suggested that layer-by-layer growth is favored at high temperature II. LMDAD METHOD ( 600 K , confirming the findings of Ref. 20. However, in order to minimize interdiffusion, the growth was performed The LMDAD effect has been described in several recent at a substrate temperature of 450 K. No trace of contami- experimental and theoretical papers.25­31 We refer for the nants was detected before and after each evaporation. Va- definition of the experimental geometry and for description lence band spectra were measured to control the surface and application of the atomic model interpretation to Refs. cleanliness during the experiment. The base pressure was 27 and 29. Here it is important only to recall that 1 the sign 3 10 11 mbar. of the LMDAD dichroism, i.e., its plus-minus feature, de- The sample was magnetically saturated by current pulses fines the parallel-antiparallel magnetic alignment between through the winding of the electromagnet. All spectra were overlayer and substrate with respect to a standard ferromag- measured in remanence. Both spin polarization data, ob- netic sample. 2 The magnitude of the dichroism is propor- tained from a 100 kV Mott detector on the same Fe single tional to Msurf , i.e., to the order parameter of the ferromag- crystal and mounting, and in situ Kerr-effect measurements netic surface; it vanishes at the Curie temperature and/or for showed a squared hysteresis loop as well as 100% rema- unmagnetized samples. LMDAD is therefore sensitive to in- nence. Linear magnetic dichroism in the LMDAD mode was plane disorder and domains: A reduced LMDAD signal im- measured in the chiral geometry described in Refs. 26 and plies a reduction of the magnetic order along the axis defined 34, obtaining two mirror experiments by reversing the sign 55 MAGNETIC INTERFACE FORMATION AT Cr/Fe 100 . . . 391 FIG. 1. Left: LMDAD spectra for the two magnetization directions crosses and continuous curves of the Fe and Cr 3p core level as a function of the Cr coverage on the Fe 100 surface from up to down , as measured in the same fixed chiral geometry, for a photon energy of 120 eV at 150 K of temperature. The vertical bars indicate the energy positions of the maxima. Right: LMDAD difference curves, corresponding to the magnetization-dependent spectra, for Fe and Cr 3p. Solid circles are experimental data and solid lines are the smoothed functions. of the magnetization direction, which was parallel to the tion to the photoemission peak area. Within the error bars the Fe 100 surface and perpendicular to the scattering plane de- submonolayer and monolayer data are equal, but a sharp de- fined by the photon beam and by the photoelectron momen- crease of the width is measured at 2.5 ML, and starting from tum vector. The LMDAD magnetic asymmetry is defined as 3.5 ML a constant value is reached, up to thicker Cr films. ALMDAD (Iup Idown)/(Iup Idown), where Iup (down) are the The same analysis for the Fe 3p core level splitting is shown photoemission intensities measured for the imposed magne- in the top panel of Fig. 3, where a similar result of reduction tization in the upward downward direction. of the mJ 3/2 splitting is observed for Cr coverages larger than 1.5 ML. Figure 4 presents the Fe/Cr/Fe 100 in- IV. RESULTS terface, for 1.5 ML of Fe on top of a 12 ML Cr film grown The left panel of Fig. 1 presents the 3p core level spectra onto Fe 100 . From the comparison with the magnetization- for Fe and Cr as measured in the two mirror experiments, as dependent spectra for the Cr/Fe 100 , we observe that a the a function of the Cr coverage on the Fe 100 surface. The coupling between the Fe overlayer and the Fe 100 substrate, corresponding LMDAD difference curves, representing the across the Cr layer, is dominantly antiferromagnetic, as LMDAD dichroism, are shown in the right panel of the same shown by the reversal of the sign of Fe LMDAD; also the Cr figure. The vertical bars identify the different peak positions LMDAD signal is reversed, showing that Fe is the magnetic for the two core levels. The opposite behavior, i.e., the re- driver in the Fe/Cr interface; b the degree of magnetic order versed plus-minus feature, of the Cr dichroism with respect is small in the iron overlayer, which has a small LMDAD to the one of Fe indicates that the dominant contribution is signal. The Cr LMDAD dichroism width is within the errors from Cr antiferromagnetically coupled to the Fe 100 of the same order of the thick layer, as indicated in the right substrate.27 At 2.5 and 3.5 monolayers ML one sees a small part of Fig. 4, but the value of the Fe 3p splitting is different energy shift of both Fe and Cr 3p peaks and a marked nar- with respect to the value of the Cr/Fe 100 interface in the Cr rowing of the Cr 3p LMDAD curve. This effect is better monolayer regime. In fact, the widths of the Fe LMDAD shown in Fig. 2 where the Cr 3p LMDAD spectra, after dichroism for the Fe 100 clean surface and for the 1 ML normalization, at 1.5 ML of coverage and at 3.5 ML are Cr/Fe 100 interface have comparable values, but the compared, both aligned to the positive asymmetry peak: The LMDAD width for the 1.2 ML Fe/12 ML Cr/Fe 100 width of the Cr LMDAD spectrum of the 1.5 ML coverage is trilayer is 30% larger. The same enhancement in the value larger by 35% with respect to the 3.5 ML spectrum. Within of the Fe splitting was observed in previous experiments,26,35 the scheme of the atomic model and according to Fe 2p whose results are reported in Fig. 3 circles . Finally, Fig. 5 LMDAD data, the positive and negative peaks of the asym- shows the evolution of the Fe 3p LMDAD splitting bottom metry correspond to the energy of the mJ 3/2 panel and the Fe 3p normalized LMDAD top panel as a sublevels.27,29 The width of the Cr dichroism, i.e., the function of the Fe coverage in the Fe/Cr/Fe 100 system. mJ 3/2 energy splitting, is plotted in the bottom panel of After the first coverage with no LMDAD signal, starting Fig. 3 versus the Cr thickness on Fe 100 , after normaliza- from 1.5 ML a large Fe LMDAD splitting is found, followed 392 PANACCIONE, SIROTTI, NARDUCCI, AND ROSSI 55 FIG. 3. Top: evolution of the 3p LMDAD splitting of Fe open squares as a function of the Cr coverage in the Cr/Fe 100 and Fe/Cr/Fe 100 interfaces. Solid triangles are the results of a previ- FIG. 2. Top: comparison between the Cr 3p LMDAD curves for ous experiment on the same interface Ref. 26 . The insets show the 1.5 ML open circles and 3.5 ML solid squares coverage in the direction of coupling, parallel or antiparallel to the substrate Cr/Fe 100 interface. Solid curves are smoothed functions. The nor- Fe 100 , in the bilayer and trilayer systems. Bottom: evolution of malized dichroism curves are reversed in sign and both arbitrarily the 3p LMDAD splitting of Cr as a function of the coverage in the aligned on one side of the curve, to better show the difference of the Cr/Fe 100 and Fe/Cr/Fe 100 interfaces. splitting value. The vertical bars indicate the peak position: A dif- The exponential attenuation of the Fe signal through the Cr ference of about 35% in the width of the dichroism is recognizable. overlayer in Fig. 6 excludes the occurrence of an extended Bottom: same comparison for the 3p dichroism of Fe. Open circles, intermixing for our growth conditions, but the little magnetic 1.5 ML of Cr; solid squares, 3.5 ML of Cr; solid curve, Fe 100 order for the Fe monolayer deposited on top of the 12-layer clean surface. The LMDAD curves are aligned on the same side of Cr film Fig. 5 indicates a stepped Cr surface, which in turn dichroism. The reduction of the magnetic splitting is of 8%. suggests that the growth of Cr is in large islands rather than layer by layer. We must observe that the lack of well- by a reduction of the splitting towards the value of the behaved layer-by-layer growth in our conditions frustrates Fe 100 clean surface. the correct development of layered antiferromagnetism in the range of small coverages investigated up to 3.5 ML . We can in fact infer the presence of a Cr structural disorder from V. DISCUSSION the results of Fig. 1, where the obtained Cr dichroism shows A. Cr/Fe 100... for all the coverages the plus-minus feature that corresponds to an antiparallel Cr alignment with respect to the Fe sub- The changes of the photoemission peak shape and width strate. The breakdown of the layered antiferromagnetism be- can have several origins including bonding disorder, with havior is a signature of a stepped and frustrated Cr layer.20 unresolved chemical shifts arising from inequivalent sites, By assuming that i the spin-orbit interaction is fixed and and size-dependent core hole screening effects. In addition, ii the splitting between the m in a magnetic material the variations of the magnetic moment J 3/2 and the mJ 3/2 sublevels, which are the two pure spin-orbit states of the at the surface or at an interface are directly reflected in the multiplets,31 varies linearly with the strength of the exchange energy splitting of the core hole magnetic sublevels and, interaction, one can interpret the changes of the width of the therefore, in the energy width of the magnetic dichroism LMDAD curve reported in Fig. 3 as being proportional to spectrum. We will discuss below the relative changes in the the relative variations of the surface magnetic moment. Fig- dichroism width as defined above independently of the small ure 3 shows that in the range 0­3.5 Cr ML on Fe 100 , the chemical shifts that are observed at the interface formation. Cr mJ 3/2 splitting value decreases from 1.05 0.05 eV 55 MAGNETIC INTERFACE FORMATION AT Cr/Fe 100 . . . 393 FIG. 4. LMDAD 3p photoemission spectra as a function FIG. 6. Total photoemission intensity of Fe 3p core levels, di- of magnetization reversal crosses and solid curve for the vided by the sum of the total intensity of Cr and Fe 3p solid Fe/Cr/Fe 100 trilayer system (h 120 eV, T 150 K . The plus- circles , as a function of the Cr coverage in the Cr/Fe 100 inter- minus feature of the dichroism curve open circles is reversed for face. The solid curve is the fitted exponential function. both Fe and Cr, showing an antiparallel orientation of the Fe top layer with respect to the Fe substrate. Also the Cr LMDAD dichro- to 0.85 0.05 eV. This relative reduction of 35% in the split- ism is reversed with respect to the one measured for the same cov- ting as the coverage exceeds the first monolayer is the sig- erage as a free terminated layer. The solid curve in the LMDAD nature of an enhanced interface magnetic moment of the Cr dichroism is a smoothed function. atoms in contact with Fe. The splitting value of about 0.85 eV cannot be representative of the Cr bulk magnetic mo- ment, considering also that the thickness range over which the reduction of the magnetic splitting occurs is affected by the island growth mode: Signals from first, second, and third Cr layers are added. Nevertheless, it appears that the mag- netic moment changes gradually at least through three layers before stabilizing at the value which is measured up to 12 layers in this experiment.22 The measured enhancement of Cr at the surface of Fe 100 is large, but definitely smaller than some values reported before,9,10 or predicted by theory theoretical predictions are referred to T 0 K and for a per- fect 100 Cr monolayer . The Fe 3p LMDAD splitting is basically unaffected by the adsorption of the first monolayer of Cr, but a reduction of about the 10% is observed for higher Cr coverages. The si- multaneous reduction for both the Fe and Cr LMDAD split- tings ( Fe and Cr) suggests a change in the magnetic properties of the whole interface region at a ``critical'' thick- ness of 1.5­2 ML of Cr. This range of thicknesses is the onset of the ferromagnetic order of Cr, in qualitative agree- ment with Turtur and Bayreuther9 and Alvarado and Carbone.16 The interface between Fe 100 and a single monolayer of Cr is different from the interface between Fe 100 and an AF stacked Cr film: The latter case implies a reduction of the Fe moments near the surface while the former case does not. FIG. 5. Top: evolution of the normalized dichroism ND of the More insight into the magnetic order of Fe in the interface Fe 3p LMDAD in a Fe/Cr/Fe 100 trilayer system open diamonds region, below the Cr overlayer, can be obtained by using the with error bars , as a function of the Fe top layer coverage. The LMDAD normalized dichroism ND and by plotting it value of 1 corresponds to the normalized dichroism of the Fe 100 against the m clean surface. Bottom: evolution of the Fe 3p LMDAD splitting in J 3/2 splitting.36 As we discussed above and in Ref. 37 the width and the ND of the LMDAD are not fully the same trilayer system as a function of the Fe top layer coverage squares with error bars . The dashed line indicates the value of the independent as one can test by applying the atomic model Fe 100 clean surface splitting. The ND is obtained by dividing the and calculating the LMDAD spectra when the mJ 3/2 integral of the Fe photoemission peak at each coverage and referred splitting is artificially varied. The wider the splitting, the to the standard spectrum of the clean Fe 100 surface; this was larger is the ND since the opposite dichroic intensities over- measured at 120 eV of photon energy and 150 K of temperature. lap less and less. Conversely, if the splitting is reduced to 394 PANACCIONE, SIROTTI, NARDUCCI, AND ROSSI 55 FIG. 7. Simulation of the LMDAD behavior as a function of the mJ 3/2 splitting value, in the scheme of the atomic model of Ref. 29. Solid squares are the LMDAD of the measured Fe 100 clean surface. Inset: Comparison between the simulation open squares and the Fe 3p experimental data for the Cr/Fe 100 inter- face solid squares for the maximum negative asymmetry in the normalized LMDAD dichroism curve vs splitting value. The nor- malized value equal to 1 corresponds to the Fe 100 clean surface as measured. Data for the Fe/Cr/Fe 100 are also indicated by the arrow, showing the largest splitting value, as well as the minimum of normalized dichroism. zero, i.e., in the case of degenerate hole sublevels in the absence of magnetic moment, also the ND is reduced to zero. Such a calculation is represented in Fig. 7 by the solid squares and lines. One sees that for typical changes of the FIG. 8. Top: Fe 3p magnetization-dependent spectra for two magnetic splitting changes of the magnetic moment by different Fe coverages in the Fe/Cr/Fe 100 trilayer, h 150 eV, 30% , the related changes in the ND are less than 10%. On T 300 K; the lower coverage does not show any LMDAD right the other hand, changes of the order parameters, i.e., of the panel . The spectra for 15 ML do show LMDAD left and corre- ND, do not influence the splitting.37 In Fig. 7 we compare spond to AF i.e., antiparallel coupling with respect to the Fe 100 also the calculated values with the data for Fe 3p of the substrate. Bottom: comparison between the magnetization-averaged Fe 100 surface covered by increasing thickness of Cr. The spectra for the two Fe films. The line shapes show a marked differ- data open squares show a reduction of Fe ND for two ence between the zero LMDAD spectra and the AF coupled ferro- monolayers of Cr, followed by a reduction of the splitting as magnetic overlayer spectra. seen in Fig. 3, and by a sharp reduction of the ND. The ND reduction is large and independent on the magnetic moment 5, we observe that the maximum ND is reached for 5 ML; of the Fe substrate. This effect is further proof of the pertur- this value corresponds to both high order and higher splitting bation in the Fe near-interface layers of the substrate. The with respect to the standard reference spectrum of the layer nearest the interface, i.e., basically the only one con- Fe 100 clean surface i.e., mJ 3/2 1.06 eV and ND 1 . At tributing to the photoemission in the data point for 8 ML higher coverages the ND converges to the standard value. Cr/Fe 100 , has a severely reduced magnetic order in the Correspondingly, the bottom panel of Fig. 5 shows that the direction probed by our experiment. Intermixing at the level mJ 3/2 splitting starts very high as soon as it can be of a single interface double layer Fe-Cr cannot be excluded, measured, and then decreases, still remaining higher than the as well as a rotation of the Fe moments forming an extended Fe 100 substrate for relatively high thicknesses. These re- domain wall with the antiparallel oriented Cr interface layer. sults indicate that the magnetic moment of the top Fe film of the Fe/Cr/Fe 100 is enhanced and that the onset of a Fe B. Fe/Cr/Fe 100... magnetic order antiparallel aligned to the Fe 100 substrate Interface magnetic effects are seen from the Fe/Cr/Fe re- is for 1.5 ML thickness of the Fe top layer. In fact, Fig. 5 sults too Figs. 4 and 5 . Data from Fig. 5 show that the first and the magnetization-dependent spectra of Fig. 8 show that monolayer of Fe grown on the 12-layer Cr buffer has a large for submonolayer and monolayer Fe thicknesses no LMDAD splitting, but a small degree of alignment of the moments is measured. The absence of LMDAD has been confirmed by along the substrate magnetization parallel or antiparallel . performing experiments at 150 K, which excludes the hy- This is consistent with the results of Alvarado and Carbone pothesis of a strongly reduced Curie temperature for the Fe who measured zero spin polarization for Fe growing on a overlayer, in agreement with Ref. 16. The spectra do not Cr 100 epitaxial film up to 2 ML Fe thickness.16 From Fig. present LMDAD, but this does not correspond to a narrow- 55 MAGNETIC INTERFACE FORMATION AT Cr/Fe 100 . . . 395 ing of the photoemission peaks. All these findings suggest atoms is enhanced by 30% independently of the direction of that the magnetic moments are oriented differently. the film magnetization, a result which is in qualitative agree- As we discussed above, based on indirect evidence, the Cr ment with the theoretical analysis of Stoeffler and Gautier for surface is highly stepped, implying the existence of terraces spin-frustrated systems.22 of different height determining antiparallel Cr surface do- The experimental evidence of enhanced magnetic mo- mains to which the Fe top monolayer should couple antifer- ments near the interface and of a thickness-dependent orien- romagnetically. The Fe top layer would therefore break itself tation of the surface magnetization axis shows how delicate into domains with many in-plane NeŽel walls, which is an the energy balance is for the magnetic coupling through Cr energetically unfavorable situation. Nevertheless, if this was spacer layers, at least if these present a rough surface. From the case, the Fe 3p photoemission spectrum would look just Fig. 4 one also observes the AF coupling between the Cr and like the field average of the usual LMDAD spectra. In fact in top Fe layers; this means that as the ferromagnetic order of this hypothesis the quantization axis of the 180° domains iron sets in, the rough Cr interface becomes magnetically would still be parallel to the magnetization axis of the sub- ordered. This effect stores some extra energy in the Cr buffer strate. The Fe top layer would be unmagnetized, but its mo- layer, which influences the subsequent coupling oscillation. ments would be still aligned along the perpendicular direc- The rotation of the surface iron magnetization into the tion to the photoemission plane; so the spectrum would have 180° direction is due to the prevalence of exchange coupling the same line shape as a field-averaged spectrum of the Fe over anisotropy, i.e., to a fine energy balance which can be substrate. As a matter of fact, the spectra for submonolayer easily modified by any extra energy term like strain or sur- and monolayer coverages of the top Fe layer i.e., the one face impurities. The difficulty of reproducing fully consistent showing no LMDAD are quite different from the field av- experimental results when different growth conditions and erage of the iron substrate spectra, as easily observable in the substrates are employed is therefore easily understood. bottom panel of Fig. 8, and qualitatively resemble the line shape measured in the nonchiral geometry which can be ob- tained by rotating the quantization axis the magnetization VI. CONCLUSIONS in the scattering plane. Based on the present set of data we We have shown that the interfaces between Cr and can make the following statements: a The magnetic split- Fe 100 and between Fe and Cr/Fe 100 as grown in condi- ting of Fe 3p is present from the submonolayer regime, b tions that optimize both the degree of structural order less the absence of LMDAD cannot be explained by 180° do- than perfect and the suppression of atomic intermixing at mains aligned with the substrate quantization axis, and c the interface are dominated by antiferromagnetic coupling the spectra are compatible with the hypothesis of a nonchiral through the interface, unless magnetic frustration arises from effective geometry of the experiment, obtained by a 90° ro- interface roughness. The Cr magnetic moments at the tation of the surface quantization axis, either within the sur- Fe 100 surface are enhanced with respect to the ones in the face plane or perpendicular to it, the LMDAD being zero in thin film regime 3­12 ML . The Fe magnetic moment at the both cases. Fe/Cr/Fe 100 surface appears enhanced by 30% with re- The hypothesis of perpendicular magnetization was put spect to the Fe 100 surface value. The complexity of the forward by Alvarado and Carbone to explain the lack of spin magnetic behavior of the interface involves both the growing polarization at less than 2 ML of Fe coverage.16 Although it overlayer and the substrate near interface layers. Beyond the is a possibility, it implies a large anisotropy which for almost changes of magnetic moments, the observed changes of mag- relaxed quasiepitaxial layers is not expected. In-plane 90° netic order of the substrate can be qualitatively described as rotation may occur due to biquadratic interlayer coupling. the formation of a magnetic domain wall, between substrate The exchange energy Eex , which describes the coupling be- and overlayer, extended over several atomic planes. The en- tween layers, is proportional to both the bilinear J1 and the ergy balance governing the formation of the interface mag- biquadratic coupling J2, i.e., Eex J1cos J2cos2 , where netic wall, or extended magnetic interface, includes anisot- is the angle between the magnetization direction of two ropy, epitaxial strain, roughness, impurities, and of course layers.38,39 When J2 0, as possible in the presence of inter- exchange interlayer coupling. The balance may favor 90° face roughness and of terraces of opposite magnetization, a domains when spin frustration is large, as appears to be the 90° orientation of two magnetic adjacent layers may occur, case for the Fe monolayer on Cr/Fe 100 . instead of 0° or 180°.23,24,40 The square lattice structure of the 100 surface allows 90° domains with inequivalent an- ACKNOWLEDGMENTS isotropy energy. The behavior shown in Fig. 5 can be inter- preted then as representing the rotation of the surface iron We gratefully acknowledge M. Sacchi for fruitful discus- magnetization from 90° biquadratic interlayer coupling to sions. This work was partially supported by the EC, under 180° antiferromagnetic bilinear coupling as the Fe film the HCM program. G.R. thanks H.C. Siegmann for continu- thickness crosses 1.5­2 ML. The magnetic moment of Fe ous support. *Present address: Institut de Physique, Univ. Neucha tel, CH-2000, 2 Y. Wang, P.M. Levy, and J.L. Fry, Phys. Rev. Lett. 65, 2732 Switzerland. 1990 . 1 M.N. Baibich, J.M. Broto, A. Fert, F. Nguyen van Dau, F. 3 C. Kittel, Introduction to Solid State Physics, 5th ed. Wiley, New Petroff, P. Etienne, G. Creuzet, A. Friedrich, and J. Chazelas, York, 1976 . Phys. Rev. Lett. 61, 2472 1988 . 4 In this work we will not consider the long periodicity due to this 396 PANACCIONE, SIROTTI, NARDUCCI, AND ROSSI 55 incommensurability, the analysis being in the few monolayers edited by P. Bagus, G. Pacchioni, and F. Parmigiani Plenum, regime. New York, 1995 . 5 R.H. Victora and L.M. Falicov, Phys. Rev. B 31, 7335 1985 . 28 G. Panaccione, Ph.D. thesis, Univ. Paris VI and LURE, Orsay, 6 C.L. Fu, A.J. Freeman, and T. Oguchi, Phys. Rev. Lett. 54, 2700 France, 1995; G. Panaccione, F. Sirotti, and G. Rossi, J. Electron 1985 . Spectrosc. Relat. Phenom. 76, 189 1995 . 7 P.M. Levy, K. Ounadjela, S. Zhang, Y. Wang, C.B. Sommers, 29 G. Rossi, F. Sirotti, N.A. Cherepkov, F. Combet-Farnoux, and G. and A. Fert, J. Appl. Phys. 67, 5914 1990 . Panaccione, Solid State Commun. 90, 557 1994 . 8 J. Xu and A.J. Freeman, Phys. Rev. B 47, 165 1993 . 30 G. van der Laan and B.T. Thole, Solid State Commun. 92, 427 9 C. Turtur and G. Bayreuther, Phys. Rev. Lett. 72, 1557 1994 . 1994 ; B.T. Thole and G. van der Laan, Phys. Rev. B 50, 10 Zhongde Xu, Y. Liu, P.D. Johnson, and B.S. Itchkawitz, Phys. 11 474 1994 . Rev. B 52, 15 393 1995 . 31 G. van der Laan, Phys. Rev. B 51, 240 1995 , and references 11 T.G. Walker, A.W. Pang, and H. Hopster, Phys. Rev. Lett. 69, therein. 1121 1992 . 32 F.U. Hillebrecht, H.B. Rose, T. Kinoshita, Y.U. Idzerda, G. van 12 F.U. Hillebrecht, Ch. Roth, R. Jungblut, E. Kisker, and A. der Laan, R. Denecke, and L. Ley, Phys. Rev. Lett. 75, 2883 Bringer, Europhys. Lett. 19, 711 1992 . 1995 . 13 Y.U. Idzerda, L.H. Tjeng, H.J. Lin, C.J. Gutierrez, G. Meigs, and 33 Ch. Roth, H. Rose, F.U. Hillebrecht, and E. Kisker, Solid State C.T. Chen, Phys. Rev. B 48, 4144 1993 . Commun. 86, 647 1993 . 14 A. Berger and H. Hopster, Phys. Rev. Lett. 73, 193 1994 . 34 F. Sirotti, G. Panaccione, and G. Rossi, Phys. Rev. B 52, R17 063 15 A. Vega, C. Demangeat, H. DreysseŽ, and A. Chouairi, Phys. Rev. 1995 . B 51, 11 546 1995 . 35 G. Rossi, G. Panaccione, and F. Sirotti, in Magnetic Ultrathin 16 S.F. Alvarado and C. Carbone, Physica B 149, 43 1988 . Films, Multilayers and Surfaces, edited by E. Marinero et al., 17 C. Carbone and S.F. Alvarado, Phys. Rev. B 36, 2433 1987 . MRS Symposia Proceedings No. 384 Materials Research Soci- 18 J. Unguris, R.J. Celotta, and D.T. Pierce, Phys. Rev. Lett. 67, 140 ety, Pittsburgh, 1995 , p. 447. 1991 . 36 The ND was obtained after a background subtraction, taking into 19 J. Unguris, R.J. Celotta, and D.T. Pierce, Phys. Rev. Lett. 69, account the single-peak intensities in the total photoemission 1125 1992 . spectra, i.e., magnetically averaged. The ND value was normal- 20 J. Unguris, Joseph A. Stroscio, R.J. Celotta, and D.T. Pierce, ized to different temperatures and corrected for the two 120 eV Phys. Rev. B 49, 14 564 1994 . and 150 eV photon energies. The ND of the clean Fe 100 21 D. Stoeffler and F. Gautier, Phys. Rev. B 44, 10 389 1991 . surface measured at 120 eV photon energy was taken equal to 1 22 D. Stoeffler and F. Gautier, J. Magn. Magn. Mater. 147, 260 as a reference value. It is well known that the LMDAD dichro- 1995 . ism signal is energy dependent, as shown theoretically in Refs. 23 S. Adenwalla, G.P. Felcher, Eric. E. Fullerton, and S. Bader, 29 and 31 and also experimentally Ref. 28 . It is then necessary, Phys. Rev. B 53, 2474 1996 . when comparing two values of the ND obtained at different 24 Eric E. Fullerton, K.T. Riggs, C.H. Sowers, S.D. Bader, and A. photon energies, to take into account this energy dependence. Berger, Phys. Rev. Lett. 75, 330 1995 . We used the measured value of the LMDAD on the Fe 100 25 Ch. Roth, F.U. Hillebrecht, H. Rose, and E. Kisker, Phys. Rev. clean surface to obtain the presented normalization. Lett. 70, 3479 1993 . 37 G. Rossi, F. Sirotti, and G. Panaccione unpublished . 26 F. Sirotti and G. Rossi, Phys. Rev. B 49, 15 682 1994 . 38 M.D. Stiles, Phys. Rev. B 48, 7238 1993 . 27 G. Rossi, F. Sirotti, and G. Panaccione, in Core Level Spec- 39 J. Unguris, R.J. Celotta, and D.T. Pierce, J. Magn. Magn. Mater. troscopies for Magnetic Phenomena: Theory and Experiment, 127, 205 1993 . Vol. 345 of NATO Advanced Study Institute, Series B: Physics, 40 J.C. Sclonzewsky, Phys. Rev. Lett. 67, 3172 1991 .