PHYSICAL REVIEW B VOLUME 57, NUMBER 2 1 JANUARY 1998-II Magnetism, structure, and morphology of ultrathin Fe films on Cu3Au 100... B. Feldmann, B. Schirmer, A. Sokoll, and M. Wuttig Institut fu¨r Grenzfla¨chenforschung und Vakuumphysik, Forschungszentrum Ju¨lich D-52425 Ju¨lich, Federal Republic of Germany Received 3 July 1997 The magnetic, structural, and morphological properties of ultrathin Fe films on Cu3Au 100 have been investigated by low-energy electron diffraction including I/V measurements, Auger electron spectroscopy, medium-energy electron diffraction, and magneto-optic Kerr effect. The main aim of these studies was to establish the correlation between film structure and magnetism. For this purpose both films deposited at 300 K RT and at 150 K LT followed by subsequent annealing to 300 K were investigated. Above about 1 monolayer ML , the films exhibit a perpendicular magnetization, which switches at 3.2 0.2 ML for LT and 2.3 0.2 ML for RT films in-plane. The reduction of the switching thickness from perpendicular to in-plane with growth temperature is caused by an interdiffusion at the Fe film/substrate interface. At somewhat larger thickness a structural transition is observed. This structural transition is not related to the magnetic reorienta- tion. Contrary to other studies no evidence is found for any fcc iron modification. We rather conclude that above 5 ML, the iron film transforms from strained to unstrained bcc 100 Fe. S0163-1829 98 03802-8 I. INTRODUCTION thickness a transition to unstrained bcc iron is observed. However, up to this point, the evidence for an fcc iron phase By deposition of ultrathin films it is possible to stabilize has been mainly derived from a kinematic analysis of low- new materials with properties, which deviate from the corre- energy electron diffraction LEED I/V curves for the 0,0 sponding bulk crystals. One aim of the study of thin-film beam rather than a full-dynamical LEED analysis for several properties is to exploit them for new applications. To tailor beams. The latest study which followed the kinematical ap- films with specific properties, the knowledge about the cor- proach was performed by Lin et al.12 These authors corre- relation between the magnetic, structural, and morphological lated the observed magnetic reorientation with a change in properties is required. Ultrathin Fe films are particularly the LEED I/V curve and changes in the scanning tunneling suited for an investigation of this correlation, because Fe can microscopy STM images with growing film thickness. exist in a ferromagnetic FM bcc, FM fcc or antiferromag- They concluded that the magnetic reorientation is closely netic AF fcc modification.1 Epitaxially grown Fe films on related to a structural transition from fcc 100 to bcc 100 Cu 100 lattice parameter a 3.61 Å attracted much atten- tion because the misfit ( f ) for the antiferromagnetic fcc Fe iron. Hence, this scenario would closely resemble the behav- phase a 3.58 Å, f 0.8% Ref. 2 and the ferromag- ior of iron films deposited on Cu 100 . This is clearly an netic fcc Fe phase a 3.66 Å, f 1.3% is small. Both interesting and important finding since it extends our knowl- modifications3­6 have been stabilized by epitaxial growth on edge about the limits of metastable epitaxy. a Cu 100 substrate. FM fcc Fe can be grown up to approxi- To confirm this interesting observation and to help de- mately 4.5 monolayers ML . The quantitative structure de- velop a detailed understanding of the correlation of mag- termination of the phase revealed that FM fcc Fe is unstable netic, structural, and morphological properties we have stud- against shears in the fcc 011 direction and is only stabilized ied Fe films on a Cu3Au 100 crystal. This study finds rather by the Cu substrate.3 This result supports theoretical conclusive evidence that the correlation of structure and calculations,7 which predicted an instability of this phase. In magnetism for this system differs considerably from several order to investigate the strain-dependent properties of this conclusions of previous investigations. phase, we chose Cu3Au 100 (a 3.745 Å) as the substrate, In the next section the experimental setup will be de- which should allow to expand the in-plane and contract the scribed. Section III is divided into the presentation of the interlayer distances of a pseudomorphic fcc Fe-film. This morphological A , magnetic B , and structural C proper- substrate has the advantage, that the misfit for the FM phase ties. In Sec. IV we correlate the properties A and compare ( f 2.3%) is smaller than for the AF ( f 4.6%) and them to previous results B . Section V contains a short sum- bcc 100 phase ( f 7.7%). Hence the growth of FM mary. fcc 100 Fe might be enabled. For substrates with an even larger lattice parameter, on the contrary, the misfit for the bcc modification of iron de- II. EXPERIMENTAL creases and the misfit for the fcc iron modification increases. Hence, on those substrates which include Au 100 and The experiments were performed in an ultrahigh-vacuum Ag 100 , one expects the stabilization of bcc iron instead of chamber designed to correlate magnetism, structure, and fcc iron. This is in line with a number of previous studies.8,9 morphology of ultrathin films. Since the apparatus has al- Hence, Cu3Au 100 should be a particulary interesting sub- ready been described elsewhere,13 only a brief description of strate for the growth of iron films. Indeed a number of pre- the system will be given here. The chamber is equipped with vious studies come to the conclusion that fcc iron grows on several facilities for the preparation and analysis of thin films Cu3Au 100 at small thickness.10­12 With increasing film and surfaces. The substrate was a polished Cu3Au 100 0163-1829/98/57 2 /1014 10 /$15.00 57 1014 © 1998 The American Physical Society 57 MAGNETISM, STRUCTURE, AND MORPHOLOGY OF . . . 1015 single crystal, approximately 7 mm in diameter and 2.5 mm III. RESULTS thick. The surface is oriented to within 0.1° of the surface normal. It was cleaned by Ar-ion sputtering at room tem- For Fe films grown on a Cu 100 single crystal it was perature and at 490 K for 10 min afterwards. Subsequently shown1,15 that the properties are sensitively dependent on the the sample was annealed at 770 K for 5 min to allow a preparation conditions, such as the deposition temperature. smoothening of the surface. Then the sample was held for 20 Interface properties like roughness and interdiffusion of sub- min at 630 K to order the Cu strate and deposit atoms can be modified by the preparation 3Au 100 alloy surface, which exhibits an order-disorder transition at 660 K. After this conditions. Segregation is another effect which could influ- treatment contamination levels were below the Auger detec- ence the film properties. In principle, three factors16 deter- tion limit ( 2 at. %) and LEED showed a sharp pattern mine the segregation of substrate atoms. These are the size of the atoms, the surface free energy and the tendency to form typical for the Cu3Au surface. Due to the different atomic an alloy. Au atoms are larger in size than Fe atoms, which form factor of Cu and Au atoms, additional spots are visible supports the tendency of Au atoms to segregate to the sur- in LEED for the chemically ordered Cu3Au 100 surface, face. Due to the smaller surface free energy of Au (1.5 J/m2) which are not evolving for unreconstructed fcc 100 surfaces and of Cu (1.78 J/m2) than of bcc Fe (2.41 J/m2),16 a cov- of pure metals. In the notation of this surface, the additional ering of the film surface with substrate atoms is preferred. spots are located at (m 12 ,n 12 ), (m,n 0,1,2,3, . . . ) posi- Finally, the binary phase diagrams of Fe and Au and of Fe tions. and Cu Ref. 17 show a negligible tendency for a solid Fe was evaporated from a small disk of high purity solution of Au and Cu in Fe and no bulk alloy formation at 99.99% . The pressure rise during Fe deposition was typi- the investigated temperatures. This implies that predomi- cally below 1 10 8 Pa. After the source was turned off it nantly Au atoms may segregate to the film surface upon quickly dropped to a base pressure of 4 10 9 Pa. The re- annealing. The segregation of substrate atoms may also sulting contamination level of the Fe films was below 2%. change the Fe-substrate interface, because Fe atoms may oc- Deposition rates between 0.3 and 2 ML/min were used. The cupy the subsurface sites which are created by the segrega- film growth was monitored by measuring the medium-energy tion of Au and to a lesser extent Cu atoms. Therefore we electron-diffraction MEED intensity during deposition. For investigated for comparison the properties of films deposited Fe films grown at room temperature, the MEED curves of at 300 K room temperature RT and at 150 K low tem- the 0,0 beam show weak oscillations superimposed on a perature LT . The films deposited at 150 K were subse- descending slope up to 3 ML. However, the (1,1¯) beam quently annealed at 300 K for 10 min to improve the struc- exhibits more pronounced oscillations with constant period- tural order and to smooth the film surface. icity up to 6 ML see Fig. 2 . This allows a thickness deter- mination with a precision of 8% and enabled a calibration of the Auger intensity ratios of the Fe 703 eV and Cu 920 eV A. Morphology peaks. Our Auger calibration for Fe/Cu3Au 100 is in excel- First of all we tried to evaluate to what extent interdiffu- lent agreement with the Auger intensity ratio obtained previ- sion and/or segregation of substrate atoms influence the film ously for Fe/Cu 100 after correcting for the reduced Cu den- properties. For this purpose, the intensity of several Auger sity in Cu3Au 100 as compared with Cu 100 . This gives transitions at low energy Fe 47 eV, Cu 60 eV, and Au 69 further support for our thickness determination. Fe films eV and at higher energy Fe 703 eV and Cu 920 eV was were grown with a homogeneous thickness on the measured after deposition at 300 K and the ratios of different Cu3Au 100 sample or with a wedgelike thickness distribu- peaks were determined. Subsequently the temperature was tion. The thickness profile of the film was determined by the raised in steps of 30 K. After 20 min at constant temperature Auger intensity ratios, which were taken at different posi- the Auger intensity ratios were determined again. A change tions of the wedge. This allowed an assignment of the film of the Auger intensity ratio after such an annealing step at a properties to thickness. certain temperature is indicative of interdiffusion and/or seg- In order to investigate structural properties, the LEED regation. At the low thicknesses investigated, the Fe 703 eV pattern was observed and spot profiles were measured in dif- / Cu 920 eV ratio remained rather constant, whereas the low- ferent crystallographic directions. LEED I/V curves were energy Auger intensity ratios changed at temperatures, TDiff , taken for a quantitative, full-dynamical analysis. A compari- that led to a diffusion of substrate atoms to the film surface. son of I/V curves with analyzed spectra was used to detect These temperatures are shown in Fig. 1. As expected, the changes of structural properties. Furthermore LEED was em- tendency for substrate atoms to diffuse decreases with grow- ployed to investigate the film morphology by observing the ing thickness. Au diffuses at considerably lower tempera- energy dependence of the peak width for certain LEED tures than Cu within the investigated thickness regime. A beams. further indication for a predominant diffusion of Au atoms is Magnetic properties of the film were characterized using that the Au 69 eV peak decreases more slowly upon growing the magneto-optic Kerr effect MOKE . The light source was thickness than the Cu 60 eV signal. Whereas the Cu peak a He-Ne Laser with a wavelength of 632.8 nm. Hysteresis vanishes at 1.8 ML, a Au peak can be detected up to 4.4 ML. loops were recorded in longitudinal and polar geometry em- For this reason it cannot be excluded that a diffusion of Au ploying a polarization modulation technique. The maximum atoms to the film surface influences the film properties up to field which could be applied was 1050 Oe. The dependence 4.4 ML for RT preparation. At larger thickness no diffusion of the susceptibility on temperature was measured according of substrate atoms to the film surface could be detected even to Berger et al.14 to determine the Curie temperature. after annealing to 573 K. 1016 B. FELDMANN, B. SCHIRMER, A. SOKOLL, AND M. WUTTIG 57 FIG. 1. Thickness dependence of Cu solid squares and Au open circles diffusion to the Fe film surface. TDiff is the tempera- ture that led to diffusion. Temperatures, at which no diffusion was observed, are 30 K below these values. See text for further details. Since TDiff reaches a value of 300 K below 2 ML, a seg- regation of substrate atoms to the film surface influences the film properties already during RT preparation. To avoid this diffusion at low thickness, Fe was deposited at 150 K. Sub- sequently, the dependence of the Auger peaks on thickness before and after annealing to 300 K was evaluated. The Fe/ Cu, Fe/Au, and Cu/Au intensity ratios do not change consid- erably upon annealing. Both the Cu and Au peak vanish at FIG. 2. MEED curve of the (1,1¯) and 0,0 spot intensity during nearly equal thicknesses, namely 1.6 and 1.9 ML. Hence, a deposition at 300 and 150 K. The thickness was calibrated by an segregation of Cu and Au is negligible for low temperature evaluation of the MEED oscillations which can be observed in the preparation. (1,1¯) beam. The curves were recorded with an electron energy of 3 For a complete characterization of the growth, the MEED keV and an angle of incidence of 81.4° against the sample normal intensity of several spots was measured simultaneously dur- out of phase condition . The azimuthal angle was chosen 5.5° off ing deposition. Both the growth mode and the film structure the 001 direction. influence the MEED intensity.17 Furthermore the slope of the curves is dependent upon the scattering conditions.13 There- preparation, but the thicknesses where characteristic changes fore, one has to be cautious with a straightforward interpre- are found differ for these two preparation conditions. tation of MEED curves. The intensity of the ( 1 One might argue that Au segregation at RT is responsible 2 , 12 ) beam, which is caused by the ordered arrangement of Au and Cu for the difference in film growth as evidenced by the MEED atoms, decreases exponentially and vanishes for both growth curves. The first evidence against this assumption is the fact temperatures after deposition of 1.5 ML. The intensities of that the MEED intensity curve of films deposited at 420 K is similar to films grown at RT, although Au is able to segre- the 0,0 spot for growth at 300 and 150 K and the (1,1¯) spot gate in a larger amount to the film surface. To suppress Au at 300 K are shown in Fig. 2. During growth at 300 K, the segregation during deposition at 300 K, the following film intensity of the (1,1¯) beam, which is diffracted nearly par- preparation was chosen. Initially 2 ML were deposited at LT allel to the film surface, increases up to 1 ML due to the and annealed to 300 K to avoid segregation of Au. Subse- growing disorder of the surface. It exhibits a shoulder at 1 quently, Fe layers were deposited at 300 K. Since the MEED ML and clear oscillations up to 6 ML. The occurrence of curves for this film closely resemble the behavior of RT these oscillations indicates an oscillation of the step density films, the segregation of Au is not responsible for the differ- upon deposition. Above 6 ML, both the 0,0 and (1,1¯) ent MEED curves of RT and LT films. beam exhibit a continuously decreasing intensity up to 10 ML. The intensity of the 0,0 beam increases again above 10 ML for RT films. However, for LT films the intensity of B. Magnetic properties the 0,0 beam has its absolute minimum already at 5 ML, For both preparation conditions, a sequence of polar and before it starts to increase again. longitudinal hysteresis loops were recorded at 166 K. The The change of the MEED intensity is also accompanied Kerr ellipticity at saturation (Ms) and at remanence (Mr) as by a distinct change of the MEED pattern. With growing well as the coercive force (Hc) were determined for different thickness we observe a reduction of the spot intensity and an thicknesses. The results for (Ms) and (Hc) are shown in increase in spot size. Above a minimum in spot intensity new Figs. 3 and 4. The data for (Mr) not shown closely re- extra spots develop which increase in intensity with growing semble the behavior of (Ms). At first sight the development thickness. This behavior is observed both for RT and LT of the magnetic properties seems to be fairly similar for LT 57 MAGNETISM, STRUCTURE, AND MORPHOLOGY OF . . . 1017 FIG. 5. Curie temperature (Tc) as a function of coverage for RT down triangles and LT up triangles preparation. The tempera- tures were determined from the temperature dependence of the sus- ceptibility as described in Ref. 14. The solid and open symbols denote the values obtained from curves recorded in polar and lon- gitudinal geometry. The perpendicular lines denote the film thick- ness for LT and RT films above which Tc shows a steep increase. FIG. 3. Thickness dependence of the Kerr ellipticity at satura- tion. The linear slope is typical of a homogeneously magne- tion M tized film. At a critical thickness, the magnetization switches s and of the coercive force Hc for Fe films deposited at 300 K RT preparation . The hysteresis loops were recorded at 166 K. to in plane and increases again linearly. Hence, also at these The solid and open circles denote the values obtained from curves thicknesses the film is magnetized homogeneously in the recorded in polar and longitudinal geometry, respectively. Longitu- film. dinal curves were taken in the fcc 001 direction. For an investigation of the in plane anisotropy, hysteresis curves were measured in different crystallographic direc- and RT preparation. No hysteresis loops could be measured tions. Below 6 ML the curves are rectangular and only a below 0.9 for LT and 1.1 ML for RT films. Above these weak dependence of their shape on the in plane direction is coverages the films are magnetized perpendicular to the sur- found. Above 10 ML for RT and 5 ML for LT films, the face and exhibit a linearly increasing saturation magnetiza- loops get a more sigmoidal shape in fcc 001 and keep the rectangular shape in the fcc 011 direction. This develop- ment continues upon increasing coverage. Hence the easy direction above 6 ML is the fcc 011 direction. The slope and the values of the coercive force for both preparation conditions do not differ considerably from each other. The coercivity decreases from 40 Oe 1.1 ML to 5 Oe for RT and from 170 Oe 0.9 ML to 5 Oe for LT films. In order to investigate the temperature dependence of the switching thickness, LT wedges were grown on the substrate and the dependence of the magnetization upon temperature was determined. The thickness, at which the hysteresis curves show a deviation from a rectangular shape and the remanence becomes smaller than the saturation magnetiza- tion, was taken as an onset for the reorientation. This inves- tigation reveals that the switching thickness changes only in a narrow coverage regime from 3.1 ML at 300 K to 3.5 ML at 100 K. Below 3.1 ML, no longitudinal hysteresis loops were measured up to the Curie temperature. The perpendicu- lar magnetization is stable up to the Curie temperature in this coverage regime. The Curie temperatures for both preparation conditions are shown in Fig. 5. Below about 1 ML, no Curie tempera- ture could be measured independent of preparation. The Cu- rie temperature increases from 275 K at 1.2 ML to 430 K at FIG. 4. Thickness dependence of the saturation magnetization 4.5 ML for LT and to 450 K at 5.3 ML for RT films. The M values of the LT films are higher, but the difference between s and of the coercive force Hc for Fe films deposited at 150 K and subsequently annealed to 300 K LT preparation . The hysteresis RT and LT films is not very pronounced. Both curves show loops were recorded at 166 K. an additional, steep increase to values above 570 K within 1018 B. FELDMANN, B. SCHIRMER, A. SOKOLL, AND M. WUTTIG 57 1 ML, whereby the onset is somewhat delayed for RT films RT: 5.5 ML, LT: 4.5 ML . A comparison of Figs. 3 and 4 shows that the most pro- nounced difference is the different stability region of the magnetic anisotropy. The perpendicular magnetization direc- tion is observed up to 3.3 ML for LT films but only 2.3 ML for RT films. The different stability regime is also visible from the maximum value of the polar Kerr ellipticity at satu- ration which reaches 808 rad for LT deposition and 520 rad for RT preparation. This indicates that the deposition temperature affects the magnetic properties of the film. The average gradient of 240 10 rad/ML and 223 10 rad/ML, respectively, for both growth conditions is fairly similiar, however. As mentioned above, the most pronounced difference be- tween RT and LT films is the different stability region of the perpendicular magnetization. Possibly the difference can be explained by variations in interface and surface roughness of the iron film or an interdiffusion. In order to explore if inter- diffusion influences the magnetic interface anisotropies, dif- ferent preparation procedures were chosen. To avoid inter- diffusion at the Fe/substrate interface a homogeneous Fe film with a thickness of 2 ML was grown at 150 K. The film was annealed to 300 K for 10 min afterwards. Subsequently, a film with a wedgelike thickness distribution was deposited at 300 K. The magnetic properties of these films are shown in Fig. 6 a . Again a perpendicular magnetization at low thick- ness and a longitudinal one at higher coverage were found. FIG. 6. Ms versus coverage for different preparation conditions: The perpendicular magnetization reaches a maximum value a after deposition of 2 ML at 150 K, subsequent annealing to 300 of 740 rad at 3.2 ML, the switching thickness. This behav- K and postdeposition at 300 K. Hysteresis loops were taken at 166 ior is comparable to that of LT films but not to RT films. The K. b after deposition at 150 K. The measurements were taken at similarity implies that the interdiffusion at the Fe/substrate 150 K without previous annealing. c after deposition of 1 ML at interface is of crucial importance. To confirm this hypoth- 300 K and growth of additional layers at 150 K followed by sub- sequent annealing to 300 K. Hysteresis curves were taken at 166 K. esis, a control experiment was performed where iron films The inset describes the resulting interface and surface morphology. were deposited and analyzed at 150 K without any anneal- Solid circles denote the perpendicular magnetization, open circles ing. denote the in-plane magnetization. The ellipticity measured at 150 K can be compared to the values of RT and LT preparation as well as postdeposited films, because the temperature difference is only 16 K. the Fe-substrate interface is a reduction of the stability re- Deposition at 150 K without subsequent annealing leads to a gime of the perpendicular magnetization for LT films. film whose properties closely resemble LT films Fig. 6 b , These results show the impact of the morphological prop- indicating that the surface roughness of this iron film is not erties on magnetic properties. In particular, roughness and crucial for the switching thickness. diffusion at the film/substrate interface have a profound ef- Finally, we tested if the segregation of substrate atoms to fect on these properties. However, structural changes may the film surface could also be of importance. Therefore, 1 also influence the stability regime of the perpendicular mag- ML was deposited at 300 K to allow an interdiffusion at the netization as was shown for Fe films on Cu 100 .15,18 The Fe/substrate interface. Subsequently, additional Fe was existence of a perpendicular magnetic anisotropy in Fe/ grown at 150 K to avoid further diffusion. After annealing Cu 100 is always linked to the stabilization of the FM fcc the sample to 300 K, polar and longitudinal hysteresis loops phase, which raises the question if this phase can also be were taken at 166 K Fig. 6 c . As becomes visible in Fig. stabilized on Cu3Au 100 . Therefore we have investigated 6 c , the switching thickness is comparable to films grown at the structural properties of the films, as well. 300 K but lower than the value for LT films. This similarity implies that the segregation of substrate atoms to the film surface for RT growth does not lead to a pronounced modi- C. Structure fication of the magnetic properties of the film. These experi- ments show that the origin of the different behavior of RT The structure of the Fe films was characterized by LEED. and LT films is the interdiffusion and roughness at the Fe/ To determine the film structure in detail, LEED I/V curves substrate interface rather than the roughness of the film sur- were measured and the observed LEED patterns were quan- face and the segregation of substrate atoms to this surface. titatively analyzed. In addition, the film morphology was de- The effect of this interdiffusion of substrate and Fe atoms at rived from the energy dependence of the LEED spot width. 57 MAGNETISM, STRUCTURE, AND MORPHOLOGY OF . . . 1019 FIG. 7. LEED patterns taken at 100 K after deposition of LT films: a 2.2 ML (E 58 eV), b 4.9 ML (E 103.5 eV), RT films: c 2.7 ML (E 50 eV), d 5.0 ML (E 150 eV). Sharp integer order spots are observed for LT films up to 3.3 sharp LEED pattern and the observation of the superstructure ML. Above 1.2 ML, superstructure spots evolve Fig. 7 a . spots up to 4.8 ML. Hence the diffuse intensity observed A spot profile analysis at 1.8 ML revealed that the super- below 5 ML is mainly due to surface roughness. Below 5.4 structure is either a (n& 1)R45 or a (n& n&)R45 with ML, the LEED pattern of films deposited at 300 K exhibits n 13­ 14. Only superstructure spots around integral order also a (n& &)R45 or a (n& n&)R45 superstructure beams (k 1/n,l 1/n) including k l 0 are observed. The similar to LT films with n 13­ 14 at 1.8 ML increasing up most intense superstructure spots are always observed for the to 33 at 4.8 ML Fig. 7 c . The superstructure spots cannot negative sign, i.e., (k 1/n,l 1/n). Such a behavior is ex- be separated from the integer order spots at about 5.5 ML. pected if the characteristic interatomic distance in the scat- They are sharpest at low energies and become weaker at tering plane is slightly larger than the nearest neighbor spac- higher energy than comparable LT films. Above approxi- mately 3 ML, additional streaks evolve running in the ing in the Cu3Au 100 plane. Upon growing thickness, the fcc 001 direction, which become more intense upon increas- superstructure spots converge towards the integer order ing energy and coverage Fig. 7 d . Moreover, the back- spots. This can be explained by a growing superstructure unit ground intensity increases at these energies. This develop- cell. Above about 3 ML, a diffuse intensity arises around ment is observed up to 6 ML. At about 5.8 ML, the streaks in integral order spots, which becomes more intense upon the 001 direction vanish and a very broad distributed inten- growing thickness Fig. 7 b . Annealing to 470 K leads to a sity evolves around the main spots, which become more in- 1020 B. FELDMANN, B. SCHIRMER, A. SOKOLL, AND M. WUTTIG 57 FIG. 8. LEED patterns taken at 100 K after annealing to the temperatures as denoted below of RT films: a 6.6 ML E 93 eV, 570 K , b 14 ML E 94 eV, 570 K , c 53 ML E 57 eV, 470 K , d an LT film: 8.8 ML E 94.8 eV, 570 K . tense upon growing thickness. The intensity of integral order ing thickness the extension of the spots is decreasing at com- spots decreases and can only be observed up to 8.6 ML. parable energies, which indicates smaller angles. Finally at Above 9.5 ML, very broad spots remain and the LEED very large thicknesses ( 30 ML), round and sharper spots pattern looks similar to the pattern of LT films above 5 ML. are observed even after annealing the films to 470 K only An annealing of this phase to 570 K leads to a sharpening Fig. 8 c . No indications for faceting can be observed by and the formation of a quadratic intensity distribution. In the LEED. Since annealing to 400 K does not improve the corners of these diffuse spots an intensity enhancement is LEED pattern, the activation barrier for the interlayer mass observed Fig. 8 a . The centers of these spots are no longer transport has to be between 400 and 470 K. LT films show located at the positions of the Cu3Au 100 substrate, but are smaller spots already above 5 ML Fig. 8 d , which are located 7.6 2% inwards in the fcc 011 direction in the comparable in quality to RT films above 10 ML after anneal- LEED pattern. The 7.6% inward shift would correspond to a ing. This shows that the change of the LEED pattern takes comparable expansion of the in-plane lattice constant. This place in a smaller coverage regime than for RT-grown films. leads to a next-nearest-neighbor distance of 2.87 0.06 Å. The data presented above describe the evolution of super- The extension of these quadratic spots increases in k space structures and film morphology with growing thickness but with increasing energy. Therefore these LEED patterns can- are insufficient to determine the full structure, i.e., the pre- not be caused by a superstructure. Only facets or mosaics can cise atomic positions of the iron atoms in the film. To be responsible for this broadening. Spot profiles were mea- achieve this goal, LEED I/V curves have been recorded for sured to determine the angle of these facets.19 Upon increas- several spots for a number of different thicknesses and 57 MAGNETISM, STRUCTURE, AND MORPHOLOGY OF . . . 1021 preparation conditions. Several of these data sets have been used as input for a full-dynamical LEED analysis. The de- tailed procedure and the findings of this analysis will be presented elsewhere.20 Here we will only summarize a subset of the information obtained that is relevant to understanding the correlation between magnetic and structural properties. The full-dynamical analysis for an annealed RT film with a coverage of 53 ML reveals an interlayer distance of 1.46 0.03 Å and an in-plane lattice constant of 2.84 0.05 Å. The structure of the film is unstrained bcc 100 . A visual comparison with I/V curves measured for bcc Fe 001 single crystals21,22 shows a rather good agreement as expected from the results of our full-dynamical LEED analysis. Therefore we can use the I/V curves as a fingerprint of films with a bulklike bcc 100 structure. Any structural difference in the near-surface region of films at lower thickness should result in characteristic deviations from the curves for bulklike films. Due to the high background and the diffuse spots the curves of not annealed films are of poor quality. However, the results of a quantitative analysis of a not annealed RT film at 18.7 ML yields nearly the same structural parameters as of the annealed film with a thickness of 53 ML. Hence the structure of the Fe films is already at this coverage of nearly unstrained bcc 100 Fe. The comparison with the spectra of a LT film at 6.5 ML and RT films at 8.6 and 14.4 ML reveals no significant differences between bcc 100 spectra and be- tween each other as becomes visible in Fig. 9. Below these thicknesses, deviations from the unstrained bcc spectra are found. This can be seen in a comparison of Fig. 9 a and 9 b . The main peaks of the I/V curves are noticeably shifted FIG. 9. Comparison between LEED I/V curves of the 0,0 spot to lower energies. The shift increases upon growing energy. of RT and LT films. The angle of incidence was 8° off the surface This can be explained by an expanded interlayer spacing in normal. a high coverage RT: 14.4 ML solid line , RT: 8.5 ML comparison to bulk bcc 100 Fe. The in-plane and the inter- dashed line , LT: 6.5 ML dotted line , b different preparation layer distances were determined for a 4.7 ML RT film by a conditions: RT 5.2 ML solid line , LT 4.3 ML dashed line , post- quantitative LEED I/V analysis. In comparison to bulk deposited film 2 ML at 150 K 2.8 ML at 300 K c low coverage: bcc 100 Fe, the interlayer distance (1.53 0.05 Å) is ex- RT 3.2 ML solid line , LT 2.5 ML dashed line panded and the in-plane distance (2.65 0.06 Å) contracted. Hence the Fe grows epitaxially strained at lower coverage. in the fcc 001 direction, whereby n increases from 14 at 1.8 These structural properties do not seem to change consider- ML to 33 for RT films at 4.8 ML. The superstructure spots ably upon decreasing coverage, because the I/V curves do do not exhibit a preferential streaking in a crystallographic not show a pronounced change down to 1.5 ML compare orientation. Therefore we cannot distinguish if the atoms are Figs. 9 b and 9 c . corrugated like ``waves'' only in fcc 010 or like hillocks in In a next step, we wanted to investigate if the preparation both the fcc 010 and fcc 001 directions. From a kinematic procedure, the segregation of Au, or a different roughness simulation of the LEED pattern the corrugation amplitude have a large impact on the film structure. For these reasons can be estimated to lie between 0.2 and 0.3 Å. From these we compared LT films at 4.3 ML, RT films at 5.2 ML, and pieces of information alone it cannot be decided if only the Fe films which were postdeposited by 2.7 ML at 300 K after surface atoms or the whole film is corrugated. As mentioned deposition of 2 ML at 150 K Fig. 9 b . The comparison of above, the most intense superstructure spots are observed for I/V curves of these films shown in Fig. 9 b demonstrates slightly smaller wave vectors than the substrate spots. There- that the structural differences are not pronounced. Therefore, fore the characteristic spacing of atoms needs to be slightly different preparation procedures and the Au segregation have larger than 2.65 Å, the atomic spacing of the substrate layer. no profound impact on the structural properties. The I/V Hence, the corrugation may be caused by an incommensurate spectra of RT and LT films do not differ considerably upon growth of Fe atoms, whereby n atoms occupy n 1 sites. decreasing coverage compare Figs. 9 b and 9 c , which indicates a similar structure of the films for both preparation conditions at low coverage. In this thickness regime a de- IV. DISCUSSION tailed quantitative analysis of the superstructure remains as a task for future work. The superstructure spots are only vis- A. Correlation of magnetism, structure, and morphology ible near integral order spots including the 0,0 beam Figs. For iron films grown on Cu 100 a clear correlation be- 7 a and 7 c . This restricts the number of models to a sinu- tween magnetism, structure, and growth has been established soidal displacement of atoms, i.e., a corrugation of n atoms in recent years.3­5,23 For films grown at room temperature, a 1022 B. FELDMANN, B. SCHIRMER, A. SOKOLL, AND M. WUTTIG 57 perpendicular magnetization is observed up to approximately atomic volume. Hence, we have to conclude that the postu- 10 ML. The reorientation of the magnetization at this thick- lated stabilization of such an fcc iron phase is highly unplau- ness is caused by a structural transformation from an fcc-like sible. live layer phase to bcc iron growing with the 110 A more probable scenario is obtained if we assume the orientation.5,24 This structural transition is accompanied by a growth of a strained bcc iron modification on Cu3Au 100 . change in growth mode from layer-by-layer growth to three- For this phase an interlayer spacing of 1.65 Å is expected, in dimensional growth. For deposition at low temperature, i.e., excellent agreement with the result of our LEED analysis.20 100 to 150 K, the transition from fcc to bcc iron already For this phase, the misfit to Cu3Au 100 is with 7.7% quite takes place at around 4.5 ML and is again accompanied by a large. Presumably, the increased atomic spacing in the iron reorientation of the magnetic anisotropy. films, which leads to the formation of superstructures, offers It is tempting to try to discuss the experimental data pre- a mechanism to reduce the film stress. At a critical coverage, sented here for Fe on Cu the strained bcc iron phase can no longer be stabilized and 3Au 100 in a similar framework. Again, evidence is found for both a change in the magnetic transforms to unstrained bcc iron. The transition we observe and structural properties with growing film thickness. Yet, a takes place at slightly different thickness depending upon growth temperature and is observed at a smaller thickness for number of findings differ considerably from the behavior LT films. Such an observation is plausible as well. The mix- observed for Fe on Cu 100 . For Fe films grown on ing at the interface for RT films should lead to a gradient in Cu3Au 100 at 300 K, a magnetic reorientation from perpen- composition which could reduce the overall stress and in- dicular to in-plane is observed at 2.3 ML. For films grown at crease the critical thickness for dislocation formation. The 150 K, the transition is found at 3.2 ML. This difference is observed structural change is therefore mainly controlled by caused by a different roughness of the interface as shown in the elastic properties of the film, the misfit to the substrate, Fig. 6. and the interface properties. The magnetic reorientation is Evidence for a structural change is found as well both unrelated to the structural change. The reorientation thick- from the analysis of the observed LEED pattern and the mea- ness is caused by a competition of crystal and shape anisot- surement of LEED I/V curves. These data indicate a struc- ropy. The interface anisotropy can be modified considerably tural change for LT films around 4.5 to 5 ML and 5.5 to 6 by an intermixing at the interface which explains the depen- ML for RT films. These thicknesses are considerably larger dence of the magnetic reorientation upon growth tempera- than the critical thickness for the magnetic reorientation. Fur- ture. thermore, even though the magnetic reorientation is observed at a larger thickness for LT than RT films, the structural change is observed at a lower thickness than for RT films. B. Comparison with previous studies Confirmation for this observation comes from the measure- The longitudinal magnetization at film thicknesses above ment of Tc as a function of film thickness. Again, the steep 3 to 5 ML has already been observed by a number of increase of Tc with film thickness is observed with 4.8 ML authors.10­12,25,26 Keune et al.25 and Macedo et al.26 investi- for LT films earlier than for RT films 5.5 ML . This casts gated the magnetic properties of Fe films grown on doubt on any attempt to relate the magnetic reorientation to a Cu3Au 100 at 473 and 309 K by Mo¨ssbauer spectroscopy. structural change, in particular a structural change from fcc At 5 ML the magnetization direction is orientated in the film to bcc iron. Yet, it is clear that the film structure at larger plane. A remanent, in-plane magnetization was found at thickness is unstrained bcc iron. The question therefore room temperature by Rochow et al.10 only above 3.6 ML arises what the film sturcture is for smaller Fe coverages from spin-resolved electron energy-loss experiments. The between 1 and approximately 6 ML. Our LEED I/V data magnetic reorientation from perpendicular to in plane was show signs for a structural change, but not a very pro- observed by Baudelet et al.11 A recent paper by Lin et al.12 nounced one. For a structural transition from fcc to bcc iron, attributes this to a structural change from fcc to bcc iron. A a considerable change in the interlayer spacing from approxi- similar structural change has already been postulated by mately 1.75­1.9 Å for fcc iron to 1.43­1.56 Å for bcc iron is Rochow et al.10 Both findings were derived from a kinematic expected. This is in contrast to the results of our LEED analysis of the LEED I/V maxima of the 0,0 beam. We analysis which only shows a moderate change with film believe that for the iron films on Cu3Au 100 only a full- thickness. Furthermore, from the LEED superstructures we dynamical analysis of an extended data set with several conclude that the characteristic in-plane atomic spacing of beams can lead to reliable data. We have performed such an the Fe atoms is 3­6 % larger than the atomic spacing in the analysis and find evidence for a strained bcc modification Cu3Au 100 surface, i.e., 2.65 Å. Lin et al.12 have recently below approximately 4­5 ML. Above this thickness a struc- determined an interlayer spacing of 1.9 Å for the first few tural change to unstrained bcc iron is observed. A structural iron layers from a kinematic analysis of maxima in the I/V change in this transition regime has also been observed in the curve of the 0,0 beam. Combining this result with a 5% STM studies of Lin et al.12 Again, the observed change in expansion of the atomic spacing of the iron atoms in the film morphology, in particular, in step height as a function of plane, an atomic volume of 14.7 Å3 is obtained. This is 20% film thickness has been attributed to a transition from fcc to larger than the expected atomic volume for the ferromagnetic bcc iron. We believe that this change in step height can also fcc iron phase. Neglecting an expansion in the plane still be explained by a transition from strained to unstrained bcc produces an increase of the atomic volume by 10% com- iron in the entire film, which should change the step height pared to the ground state of ferromagnetic fcc iron. We can- between strained und unstrained areas by approximately 0.4 not envision any mechanism which could lead to such a large Å. This is in agreement with the observations of Lin et al. 57 MAGNETISM, STRUCTURE, AND MORPHOLOGY OF . . . 1023 Hence, the phase transition from strained to unstrained bcc films up to 3.1 ML at 300 K. The different switching thick- iron appears to be consistent with the experimental data pre- nesses are a result of an interdiffusion of Fe and substrate sented so far. atoms at the Fe-substrate interface, i.e., an occupation of Fe atoms at subsurface sites. This decreases the interfacial an- V. SUMMARY isotropy. Predominantly Au atoms segregate on top of the film surface during deposition at 300 K and are partly incor- A correlation between the magnetic, structural, and mor- porated in the film upon growing coverage. Above 4.8 ML phological properties of ultrathin Fe films on Cu3Au 100 for LT and 5.5 ML for RT films the film relaxes to un- was established for LT and RT preparation. Below 5 ML for strained bcc 100 Fe. The transition regime for RT films lies LT and 5.5 ML for RT films the Fe grows epitaxially between 5.5 and 10 ML and is larger than for LT films. This strained with a tetragonally distorted bcc 100 modification. transition also changes the magneto-optic properties of the The films exhibit a buckling with an amplitude of about 0.2 Fe film. Yet, the structural transition is unrelated to the mag- Å in the fcc 001 direction, which seems to be caused by an netic reorientation observed at smaller thickness. incommensurate growth. The periodicity of the buckling in- creases upon growing thickness. ACKNOWLEDGMENTS The magnetization is orientated perpendicular to the film plane above 1 ML. RT films switch to in-plane at 2.3 ML, This work was supported by the Deutsche Forschungsge- LT films at 3.3 ML. The perpendicular magnetization is meinschaft Wu 243/2 . B.F. gratefully acknowledges the stable up to the Curie temperature and can be detected for LT support of the Konrad Adenauer Stiftung. 1 U. Gradmann, in Magnetism In Ultrathin Transition Metal Films, 13 J. Thomassen, B. Feldmann, and M. Wuttig, Surf. Sci. 264, 406 edited by K. H. J. Buschow Elsevier Science, Amsterdam, 1992 . 1993 , Vol. 7, Chap. 1, pp. 74­96. 14 A. Berger, S. Knappmann, and H. P. Oepen, J. Appl. Phys. 75, 2 E. F. Wassermann, IFF-Ferienschule, Magnetismus von Festko¨r- 5598 1994 . pern und Grenzfla¨chen Forschungszentrum Ju¨lich GmbH, Ju¨- 15 M. Wuttig, H. Zillgen, and B. Feldmann, Surf. Sci. 321, 32 lich, 1993 , Vol. 24. 1994 . 3 16 S. Mu¨ller, P. Bayer, C. Reischl, K. Heinz, B. Feldmann, H. G. Tre´glia and B. Legrand, Phys. Rev. B 35, 4338 1987 . 17 Zillgen, and M. Wuttig, Phys. Rev. Lett. 74, 765 1995 . H. Okamato, T. B. Massalski, L. J. Swartzendruber, and P. A. 4 R. D. Ellerbrock, A. Fuest, A. Schatz, W. Keune, and R. A. Beck, Bull. Alloy Phase Diagrams 5, 592 1984 . 18 Brand, Phys. Rev. Lett. 74, 3053 1995 . M. Wuttig, B. Feldmann, and T. Flores, Surf. Sci. 331­333, 659 5 J. Thomassen, F. May, B. Feldmann, M. Wuttig, and H. Ibach, 1995 . 19 Phys. Rev. Lett. 69, 3831 1992 . B. Feldmann, Ph.D. thesis, Forschungszentrum Ju¨lich, 1996. 20 6 B. Schirmer, B. Feldmann, and M. Wuttig unpublished . D. Li, M. Freitag, J. Pearson, Z. Q. Qiu, and S. D. Bader, J. Appl. 21 K. O. Legg, F. Jona, D. W. Jepsen, and P. M. Marcus, J. Phys. C Phys. 76, 6425 1994 . 7 10, 937 1977 . G. L. Krasko and G. B. Olson, Phys. Rev. B 40, 11 536 1989 . 22 Z. Q. Wang, Y. S. Li, and F. Jona, Solid State Commun. 61, 623 8 S. D. Bader and E. R. Moog, J. Appl. Phys. 61, 3729 1987 . 1987 . 9 M. Stampanoni, A. Vaterlaus, M. Aeschlimann, and F. Meier, 23 D. Li, M. Freitag, J. Pearson, Z. Q. Qiu, and S. D. Bader, Phys. Phys. Rev. Lett. 59, 2483 1987 . Rev. Lett. 72, 3112 1994 . 10 R. Rochow, C. Carbone, T. Dodt, F. P. Johnen, and E. Kisker, 24 K. Kalki, D. D. Chambliss, K. E. Johnson, R. J. Wilson, and S. Phys. Rev. B 41, 3426 1990 . Chiang, Phys. Rev. B 48, 18 344 1993 . 11 F. Baudelet, M.-T. Lin, W. Kuch, K. Meinel, B. Choi, C. M. 25 W. Keune, T. Ezawa, W. A. A. Macedo, U. Glos, and U. Kirsch- Schneider, and J. Kirschner, Phys. Rev. B 51, 12 563 1995 . baum, Physica B 161, 269 1989 . 12 M.-T. Lin, J. Shen, W. Kuch, H. Jenniches, M. Klaua, C. M. 26 W. A. A. Macedo, W. Keune, and E. D. Ellerbrock, J. Magn. Schneider, and J. Kirschner, Phys. Rev. B 55, 5886 1997 . Magn. Mater. 93, 552 1991 .