PHYSICAL REVIEW B VOLUME 56, NUMBER 15 15 OCTOBER 1997-I Growth of ultrathin Fe films on Ge 100...: Structure and magnetic properties P. Ma Department of Physics, University of Western Ontario, London, Ontario, Canada N6A 3K7 P. R. Norton Department of Chemistry, University of Western Ontario, London, Ontario, Canada N6A 5B7 Received 7 April 1997; revised manuscript received 25 June 1997 The structure and magnetic properties of ultrathin Fe films grown on Ge 100 at room temperature have been studied by low-energy electron diffraction, Auger electron spectroscopy AES , angle-resolved AES, and in situ magneto-optical Kerr effect MOKE measurements. Fe initially grows on Ge 100 in a disordered fashion, with local order commencing around 4 ML. The film grows with a bcc structure for thicknesses greater than 7 ML. Our data are consistent with 6% Ge intermixed in the films. Significant intermixing starts at about 160 °C, with rapid diffusion of Fe into the bulk occurring at temperatures higher than 400 °C. A single-loop to stepped-loop to single-loop sequence of ferromagnetic loops was observed by MOKE measurements. Hyster- esis loop simulations were performed based on the coherent model, the results suggesting that strong in-plane uniaxial anisotropy exists in these films, especially for very thin films. The sequence of loops is due to the increase of the ratio between the cubic anisotropy and the uniaxial anisotropy as the film thickness increases. S0163-1829 97 03239-6 I. INTRODUCTION AES , and angle-resolved AES ARAES to study the struc- ture, composition, and growth of the films. The magneto- The study of ultrathin ferromagnetic structures has been optic Kerr effect MOKE technique is used to characterize an active research field in recent years, due to the unique the magnetic properties of the thin films. properties of such systems and the potential for new appli- cations. The ability to grow magnetic structures directly on II. EXPERIMENT semiconducting surfaces has an additional attraction in that it might be possible to integrate magnetic devices and elec- All the preparation and structural characterizations of the tronic circuits on a single chip.1 thin films were performed in an ultrahigh vacuum UHV The lattice constant of Ge is about twice of that of bcc Fe chamber main chamber with a base pressure of 1 with only a 1.3% lattice mismatch. Together with the com- 10 10 torr. This chamber is equipped with rear view mercial availability of high-quality single crystal wafers, this LEED optics and a VG CLAM electron energy analyzer for makes Ge a very attractive substrate on which to grow bcc AES and ARAES. The LEED images can be recorded on Fe films. However, there have not been many studies of ul- video tape for spot intensity and width analysis by software trathin Fe films on Ge substrates, probably because of the of our own design. The main chamber is connected with concern that intermixing between the substrate and the de- another UHV chamber MOKE chamber for MOKE mea- posits might create a thick magnetic dead layer. For example, surements. The sample can be transferred between these two it has been reported that the magnetic dead layer can be more chambers in vacuum, which allows us to perform in situ than 100 Å thick for an Fe film grown on Ge at 150 °C.2 A magnetic property measurements. The main chamber is also recent study of the growth of Fe on a S-passivated Ge 100 interfaced to a 2.5-MeV van de Graaff accelerator, for in situ surface was successful in producing ferromagnetic Fe films Rutherford backscattering spectrometry RBS , etc. with interesting magnetic properties.3 A much thinner dead N-type Ge 100 wafers supplied by Superconix were used layer less than 10 Å was observed for Fe grown on a as substrates. After rinsing in methanol, the substrates were S-passivated Ge 100 surface at 150 °C compared to the inserted into the vacuum chamber for cleaning. The surfaces 100 Å magnetic dead layer of Fe grown on clean Ge 100 were sputtered by 1.5-keV Ar ions at 0.5 A/cm2 at 15° at the same temperature. This large difference of the thick- and 15° incidence angle for 10 min at each angle, and then nesses of the magnetically dead layers of Fe on clean annealed at 800 °C for 10 min. After this cleaning process, Ge 100 and S-passivated Ge 100 indicates a need for fur- no contamination was detected with AES and a sharp (2 ther study of this system. We have therefore studied the 1) LEED pattern was observed. STM measurements re- growth of Fe on the Ge 100 surface at lower temperatures ported elsewhere4 on Ge 100 following the same cleaning with emphasis on the composition and structure related to the procedure showed that the surface was well ordered and with magnetic properties. large flat terraces. The Fe source is an e-beam evaporator In this paper, we present our study of the composition, and the flux was calibrated by RBS by measuring the amount structure, and magnetic properties of ultrathin Fe films of Fe deposited on a clean Si 100 surface under identical grown on Ge 100 at room temperature. We use low-energy conditions. The Fe films were grown at room temperature electron diffraction LEED , Auger-electron spectroscopy without any further annealing. In this paper, 1 ML of Fe is 0163-1829/97/56 15 /9881 6 /$10.00 56 9881 © 1997 The American Physical Society 9882 P. MA AND P. R. NORTON 56 FIG. 1. LEED spot intensity 1,0 beams at 120 eV as a func- FIG. 2. Ge and Fe AES intensities as functions of Fe coverage. tion of Fe coverage. The lines are least-squares fits to the exponential functions see text for details . In the Ge branch, the dotted line is the fit without an offset, while the solid line is the fit with an offset. defined as 1.22 1015 atoms/cm2. All the in situ MOKE measurements were performed in the longitudinal configura- 2. AES study tion and the applied magnetic field was along the 010 di- rection of the sample. The ex situ MOKE measurements on Figure 2 shows the AES intensity of the Ge gold capped Fe films were also made in the longitudinal LM M(1147 eV) and Fe LM M transitions at 706 eV as a configuration while the orientation of the applied field in the function of Fe coverage. The data points follow a smooth sample plane could be varied. line and no apparent breaks were detected. The lines are the least-squares fittings to an exponential function see below . The calculated attenuation lengths based on the fitting are III. RESULTS AND DISCUSSION 15.5 Å for the Ge 1147-eV Auger electrons and 13.2 Å for the Fe 706-eV Auger electrons, respectively. To extract in- A. Growth and structure of the Fe films formation about the intermixing if there is any and the growth mode from the data, we have computed the attenua- 1. LEED study tion for two simple models. One possibility is that there is a After sputter cleaning and annealing at 800 °C, a sharp uniform intermixed region throughout the film, with a small (2 1) LEED pattern was observed for the Ge 100 surface. percentage of incorporated Ge. Then, after depositing n ML Figure 1 shows the intensity of the diffraction beams of the Fe, the Ge AES intensity can be written as coincident spots from Ge 100 and the Fe overlayer, as a n function of Fe coverage. Deposition of Fe led to complete B B B extinction of the Ge diffraction beams by coverages as low t exp nd0 / 0exp i 1 d0 / , n 1 as 0.3 ML. No diffraction pattern from the substrate or over- 1 layer Fe film was observed for coverages between about 0.3 where B and 7 ML. The diffraction pattern reappeared at about 8 t and B0 are the AES intensities of bulk Ge and 1 ML Ge, respectively, is the Ge atomic density in the film, ML. LEED pictures of the substrate and the Fe films of various thickness were taken at different primary energies, d0 is the monolayer thickness, and is the attenuation length and the overlayer Fe first-order beams coincide with the of Ge Auger electrons in the film. The first term is the at- Ge 100 second-order beams. This indicates that the lattice tenuated signal from the substrate and the second term is the constant of the ordered Fe films is half of that of the sub- signal from within the film. If we ignore the difference of the strate, i.e., the lattice constant of the overlayer is the same as attenuation lengths of Ge Auger electrons in the film and in that of bulk bcc Fe, and that the lattice orientation of the the bulk Ge, we have Bt B0 / 1 exp( d0 / ) . Therefore, overlayer is registered with that of the substrate. This is con- the above equation can be simplified to sistent with the ARAES observations that will be presented B B in this section. t Bt 1 exp nd0 / . 2 The recovery of the LEED pattern around 7 ML is a very Here we see that for an intermixed film, in addition to the steep function of Fe coverage, the major part of the effect exponential dependence of the substrate signal on the film being completed within about 3 ML. The spot intensity thickness, there is a constant background that is proportional reaches its maximum value around 12 ML and then starts to to the atomic density of the substrate element in the inter- decrease as more Fe is deposited. mixed film. 56 GROWTH OF ULTRATHIN Fe FILMS ON Ge 100 : . . . 9883 In the Ge data in Fig. 2, the solid line is the least-squares fit of the data to an exponential function with offset, while the dotted line is the fit to an exponential function without offset. We see that the fit that includes the offset is satisfac- tory, and that the other fit is not as good, especially at high Fe coverages. This suggest that intermixing does take place at some point. The estimated uniform Ge concentration is 6%.A further possible model is one in which the constant background originates from a small amount of Ge segregated on top of the Fe film. In this model, after deposition of n ML Fe, the Ge AES intensity can be written as B Bt 1 x exp nd0 / Btx exp 1 n d0 / xB0 , 3 where x is the monolayer coverage of the segregated Ge assuming it is less than 1 ML and is the same for all Fe coverages , and other symbols are the same as in Eq. 1 . The first two terms are the attenuated substrate signal and the third term is the signal from the segregated Ge. Using the FIG. 3. Ratio between Ge and Fe AES intensities as a function relation between B of sputter time. The starting thickness of the Fe Film is 30 ML. The t and B0 , Eq. 3 can be rearranged as sputter rate calculated is about 1.3 ML/min. B Bt 1 x Btx exp d0 / exp nd0 / primary electrons. The scans were taken as a function of the xBt 1 exp d0 / . 4 polar angle with the azimuthal angle fixed in the 010 plane. Again, we see the Ge AES intensity exhibits an exponential To present the data clearly, different vertical offsets have dependence on the Fe coverage with a constant background. been added to the spectra except the 4 ML one. The data Mathematically, Eqs. 4 and 2 are the same. Therefore, the clearly demonstrate that the strong enhancement of the AES best fits for Eqs. 2 and 4 to the data are the same and the intensity at 0° and 45°, which is a signature of a cubic struc- calculated preexponential coefficients and the constant terms ture, only starts at 7 ML and is quickly saturated in 2­3 ML. are also the same. The calculated Ge coverage is about 0.6 This suggests that a cubic-type bonding arrangement be- ML for the best fits. tween the iron atoms is only established at about 7 ML. To distinguish these two models, we grew a thick Fe layer Interestingly, this is the same coverage as that at which our 30 ML on the Ge 100 substrate, and then monitored the AES intensities of Ge and Fe while Ar sputtering the film. For the segregation model, a significant drop 36% based on Fig. 2 of the ratio between Ge and Fe AES intensities is expected after the first one or two monolayers being re- moved. Figure 3 plots the AES ratio vs the sputter time. Based on the AES study mentioned earlier in this section, we estimated that the sputter rate is about 1.3 ML/min in this experiment. In Fig. 3 we can see that there is no noticeable decrease of the AES ratio during the initial sputter period. Therefore, the result of this experiment is in favor of the intermixing model. 3. Angle-resolved AES study Angle-resolved AES ARAES is based on the strong for- ward scattering of Auger electrons with kinetic energies of a few hundred eV, by neighboring atoms.5,6 A direct result of the forward scattering is the enhancement of the AES inten- sity along the direction between the Auger-electron emitter and the scattering atom, resulting in a characteristic angular distribution of the AES intensity along certain directions. The forward scattering effect is significant typically only for the nearest- and next-nearest-neighbor distances, and so ARAES is an excellent technique to probe the local environ- FIG. 4. Angle-resolved AES spectra of Fe films of various ment of a specific atom close to the surface. Figure 4 is a plot thicknesses. The monitored peak is the Fe Auger transition at 706 of ARAES spectra at several Fe coverages. The monitored eV excited by a 5-keV electron beam. The scan is in the 010 peak is the Fe Auger transition at 706 eV excited by 5 keV plane. 9884 P. MA AND P. R. NORTON 56 FIG. 5. Effect of annealing on a 12 ML Fe on Ge 100 . The annealing temperature was ramped linearly at 2.4 °C per min. FIG. 6. Hysteresis loops measured with in situ MOKE for Fe films up to 20 ML. The MOKE system was set up in the longitu- LEED observations Fig. 1 , indicate that long-range order is dinal configuration and the applied field was along 010 . also being established. Below 7 ML, the spectra are essen- tially featureless, although at higher sensitivity and better since it can create a thick magnetically dead layer. A well- S/N, weak peaks around 0° and 45° for can be detected in known method used to prevent intermixing involves the films between 3 and 7 ML. Polar ARAES scans were also growth of a buffer layer before deposition of the magnetic performed in the 011 azimuth to obtain more structural layer. For example, on the GaAs 100 surface, a thick Ag information about the overlayers. A similar Fe coverage de- buffer layer is usually grown before deposition of the Fe film pendence of the enhancement of the forward scattering peaks to avoid the intermixing. As we have seen so far, in the case at 0 and 55° was observed, but no peak was observed at 35°. of Fe on Ge 100 , the intermixing is not severe for tempera- This clearly indicates that the Fe layer is growing in the bcc tures between 25 and 160 °C, and this will be true for lower structure, since, for an fcc layer, a forward scattering peak temperatures. This is an important observation for the cases should be observed at 35° along the 011 azimuth. This in which a buffer layer is not desirable. result is consistent with the LEED data. C. Magnetic properties B. Thermal stability Figure 6 plots the results of MOKE measurements of sev- The thermal stability of a 12-ML Fe film was studied by eral Fe films with the applied field along the 010 direction. monitoring the AES intensities as the temperature was For Fe film thicknesses of less than 4 ML, no hysteresis loop ramped linearly from room temperature at 2.4 °C/min. Fig- was observed, and even at 4 ML there is only a trace of a ure 5 plots the AES peak amplitudes of Fe and Ge as a hysteresis loop. A narrow loop appears at about 5­6 ML and function of annealing temperature. We can see that the peak then becomes a stepped loop between 7 and 9 ML. For Fe amplitudes of Ge and Fe remain constant for temperatures coverages greater than 10 ML and up to 20 ML, only single 160 °C. Between 240 and 400 °C, there is another plateau loops were observed. This transition from single to stepped with the ratio of Fe AES amplitude to Ge AES amplitude to single loops was consistently observed in all experiments close to 1. The Fe signal quickly disappears at temperatures as a function of Fe coverage. above 400 °C. As discussed in Sec. III A, there is a small The first 3­4 ML of Fe is either magnetically dead or has amount of Ge intermixed/segregated in/on the Fe film. No an easy axis perpendicular to the film. This question cannot additional intermixing and/or segregation occurs between be resolved in situ with our MOKE system, which is set up room temperature and 160 °C but significant intermixing in the longitudinal configuration. However, about the same clearly takes place above about 160 °C, which is probably thickness of dead layer was reported in the Fe/GaAs 100 the reason for the thick magnetic dead layer reported for case.8 The hysteresis loop first appears between about 4 and 150 °C growth by Prinz.2 It is likely that a surface alloy is 6 ML and at these Fe coverages the structure of the films is formed between 160 and 400 °C. Based on the Auger sensi- still disordered LEED , but some short-range bcc order has tivity of Fe and Ge,7 the ratio of atomic concentrations of Ge been established ARAES . and Fe in the intermixed region can be estimated as about In our FMR studies on the gold-capped Fe films grown on 2:1. At 400 °C, the alloy apparently dissociated and Fe atoms Ge 100 under conditions identical to those used for the films quickly diffuse into the bulk of the Ge substrate. described in the present paper, we found that all the films Intermixing has always been a major concern in growing studied from 5 to 20 ML have in-plane uniaxial ultrathin ferromagnetic films on semiconducting substrates, anisotropy.8 As we have seen in the previous sections, the 56 GROWTH OF ULTRATHIN Fe FILMS ON Ge 100 : . . . 9885 FIG. 7. Hysteresis loops measured with ex situ MOKE system FIG. 8. Calculated hysteresis loops for a film with in-plane on a gold-capped 6.5 ML Fe film grown on Ge 100 at room tem- uniaxial anisotropy and with zero cubic anisotropy. The angles in perature. The angles indicated in the figure are the angles between the figure are the angles between the external field and the 010 the applied field and the 010 direction. direction. films have a bcc structure. Therefore, there must also be a where K1 is the fourth-order crystalline cubic anisotropy, Ku cubic anisotropy energy contribution. The stepped loops be- is the in-plane uniaxial anisotropy, is the angle between M tween 7 and 10 ML are the result of the competition between and H, and is the angle between M and 010 . In simulat- the in-plane uniaxial and cubic anisotropies. For a very thin ing the loops in Fig. 7, we let K film, the uniaxial anisotropy may have a relatively large 1 0, which corresponds to a film with uniaxial anisotropy and without cubic anisotropy. value. As the film becomes thicker, the film also becomes The results are presented in Fig. 8. In this figure, the reduced more bulklike and eventually the cubic anisotropy will domi- field is (M/K nate. Therefore, during the film growth process, the ratio of u)H, and the reduced magnetization is really the cubic and uniaxial anisotropies varies, and this ratio should generally increase as the film thickness increases. As will be demonstrated below, this ratio between the cubic anisotropy and the uniaxial anisotropy determines the shape of the hysteresis loops. Figure 7 shows hysteresis loops of a 6.5-ML Fe film capped with gold and measured with our ex situ MOKE sys- tem in the longitudinal configuration. The angles indicated in the figure are the angles between the applied field and the 010 direction in the sample plane. It is clear that the mag- netic property lacks fourfold symmetry and that 011 and 01¯1 axes are not equivalent; the former behaves like a hard axis and the latter like a soft axis. Therefore, for this film we expect that the uniaxial anisotropy is relatively large and the uniaxial axis is along 01¯1 . Similar results were also found for Fe on the GaAs 100 surface.10 In the following section, we describe the application of the coherent model11 to model and interpret the hysteresis loops observed in Figs. 6 and 7. In doing so, we assume that the uniaxial axis is along 01¯1 for all film thicknesses. Therefore the energy density of the thin film can be written as K E 1 FIG. 9. Calculated hysteresis loops for films with different ratios 4 sin22 Kusin2 4 HM cos , of cubic anisotropy and in-plane uniaxial anisotropy. The external 5 field is along 010 . 9886 P. MA AND P. R. NORTON 56 cos( ). The angles indicated in the figure are the angles properties by MOKE. The growth of Fe on Ge 100 initially between the applied field and the 010 direction. The simi- occurs in a disordered fashion, clear evidence of formation of larity between Figs. 7 and 8 shows that the uniaxial anisot- the cubic structure only appearing above 4 ML. Above 7 ML ropy is dominant for very thin Fe films on Ge 100 . In simu- an ordered overlayer of bcc structure is formed. An AES lating the results in Fig. 6, we let 0, which corresponds to study indicates that a small amount about 6% of Ge inter- an applied field along 010 . Figure 9 shows the results of the mixes with the Fe film. The magnetic properties of the films simulation. In this figure we see that with the applied field are clearly dependent on the structure of the films. A narrow along 010 , the stepped loops appear when the K1 and Ku single loop was observed between 5 and 6 ML. The hyster- have comparable values. When K1 /Ku is very small or larger esis loops become stepped over the next 3­4 ML total cov- than about 4, a normal loop should be observed. It is inter- erages between 8 and 10 ML , becoming single loops again esting to note that the stepped loops appear at about the same above 10 ML. Our calculation indicates that this single to Fe coverage as the LEED pattern reappears and approaches stepped to single loop transition is the result of the increase its maximum intensity and also at the same coverage at of the ratio between the cubic anisotropy and the in-plane which the 45° peak in the ARAES spectra appears and grows uniaxial anisotropy with the increase of the film thickness. Secs. III A 1 and III A 3 ; this is precisely the region in Significant intermixing did not occur for temperatures below which the bcc structure develops strongly. There is, there- 160 °C. Between 160 and 400 °C a Ge-Fe alloy is formed fore, a very good correlation between the structural and mag- with an atomic composition of 2 Ge to 1 Fe. With even netic properties. Also, the result of the simulation is remark- higher annealing temperatures, Fe completely diffuses into ably consistent with our FMR measurement,8 where we the bulk Ge. found that generally Ku decreases with Fe thickness while K1 increase with Fe thickness, and K1 /Ku is about 0.5 at 5 ML, 1.8 at 8 ML, and 5.8 at 10 ML. ACKNOWLEDGMENTS IV. CONCLUSIONS The authors gratefully acknowledge helpful discussions with Bret Heinrich and Keith Grifffiths, the technical assis- We have studied the room-temperature growth of Fe on tance of Leighton Coatsworth, Dan O'Dacre, and Lawrence Ge 100 by LEED, AES, and ARAES, and the magnetic Green. NSERC is acknowledged for financial support. 1 E. Schloemann, R. Tustison, J. Weissman, and J. Van Hook, J. 7 L. E. Davis, N. C. MacDonald, P. W. Palmberg, G. E. Raich, and Appl. Phys. 63, 3140 1988 . R. E. Weber, Handbook of Auger Electron Spectroscopy Physi- 2 G. A. Prinz, in Ultrathin Magneatic Structures II, edited by B. cal Electronics Industries, Eden Prarie, Minnesota, 1976 . Heinrich and J. A. C. Bland Springer-Verlag, Berlin, 1994 , p. 8 P. Ma, P. R. Norton, and B. Heinrich unpublished . 35. 9 M. Gester et al., J. Appl. Phys. 80, 347 1996 . 3 G. W. Anderson, P. Ma, and P. R. Norton, J. Appl. Phys. 79, 10 J. J. Krebs, B. T. Jonker, and G. A. Prinz, J. Appl. Phys. 61, 2596 5641 1996 . 1987 . 4 P. Ma, J. H. Horton, and P. R. Norton unpublished . 11 B. Dieny, J. P. Gavigan, and J. P. Rebouillat, J. Phys.: Condens. 5 W. F. Egelhoff, Solid State Mater. Sci. 16, 213 1990 . Matter 2, 159 1990 . 6 S. A. Chambers, Surf. Sci. Rep. 16, 261 1992 .