PHYSICAL REVIEW B VOLUME 54, NUMBER 13 1 OCTOBER 1996-I Magnetic moments in thin epitaxial Cr films on Fe 100... P. Fuchs, V. N. Petrov,* K. Totland, and M. Landolt Laboratorium fu¨r Festko¨rperphysik, Eidgeno¨ssische Technische Hochschule Zu¨rich, CH-8093 Zu¨rich, Switzerland Received 26 February 1996 The absolute magnetic moments of thin Cr overlayers on Fe 100 are directly determined by energy-resolved spin-polarized secondary-electron emission. Spin-dependent attenuation of low-energy secondary electrons is quantitatively treated, following a model by Siegmann, to extract magnetic depth profiles in the adlayer. The first monolayer of Cr couples antiferromagnetically to the Fe substrate and exhibits a maximum magnetic moment of 1.8 0.2 B per atom for a submonolayer coverage. Subsequent Cr layers show a positive magne- tization. S0163-1829 96 06538-1 The topic of induced magnetic moments in nonmagnetic steps at the Cr-Fe interface leads to frustration and hence to adlayers at ferromagnetic surfaces combines state-of-the-art multiple magnetic moment distributions with reduced aver- experimental thin film magnetometry with computational age moments. This, however, contradicts first-principles physics. Most important, however, is the quantitative experi- band-structure calculations by Coehoorn11 on Fe 100 /Cr su- mental determination of the magnetic moments since the perlattices with mixed monolayers at the interface where Fe computers provide numerical values which are to be chal- and Cr moments show almost no dependence on the nearest- lenged. Stimulating indeed is the fact that on both sides even neighbor environment. A very recent study with spin- conceptual uncertainties persist and constant progress is polarized core-level photoemission by Xu et al.12 reports made towards the understanding of the underlying physics. 1.8 B/atom for a submonolayer coverage using the same In the present study we focus on Cr epitaxially grown on polarization-to-moment conversion as in Ref. 6. Fe 100 which is the best investigated system in its family. The SPSEE surface spectrometer13 is straightforward. The We utilize spin-polarized secondary-electron emission sample, which is an Fe 100 single crystal with a Cr adlayer SPSEE as in a recent report on V/Fe 100 .1 The high sur- in the present case, is magnetized by a small horseshoe- face sensitivity of SPSEE with a quantitative treatment of the shaped electromagnet along an easy direction. It exhibits full spin dependence of the electron scattering enables us to ex- remanence at which all the measurements are performed. A tract quantitative magnetic moments of the adlayers. Sub- secondary-electron cascade is excited near the surface by an stantial progress over the earlier study is made in that we unpolarized primary-electron beam of 2000 eV. The surface- utilize the energy dependence of the scattering and reach normal emission of secondary electrons is resolved in energy consistency as a stringent test. We find that at room tempera- in a cylindrical-mirror energy analyzer and subsequently ture one monolayer ML of Cr on Fe aligns antiparallel to submitted to spin-polarization analysis in a 100 keV Mott the Fe magnetization. It has a maximum magnetic moment detector. The spin polarization is defined as P (N N )/ of 1.8 0.2 B/atom for a submonolayer coverage. Subse- (N N ). N ( ) is the number of electrons with magnetic quent layers show a positive magnetization which decreases moment parallel antiparallel to the quantization axis of the with increasing distance from the interface. detector which is chosen to lie parallel to the Fe magnetiza- For a monolayer of Cr on Fe 100 an induced magnetiza- tion direction. The Cr films are deposited on the well- tion with very large magnetic moments ranging from 3.1 to prepared Fe 100 surface at room temperature by electron- 3.9 B/atom Refs. 2­4 has been predicted to align antipar- beam evaporation. During evaporation the pressure is kept allel to the magnetization of the Fe substrate accompanied by below 10 9 Torr. The cleanliness of the substrate and of the a slight reduction of the interfacial Fe moment. These early adlayers is checked with Auger-electron analysis. The film predictions so far could hardly be confirmed by experimental thicknesses are determined by the relative changes of the Fe observations. For a submonolayer of Cr Idzerda et al. report L3M45M45, Fe M23M45M45, and Cr L3M23M45 Auger- 0.6 B/atom from soft x-ray magnetic circular dichroism.5 electron intensities upon evaporation. They exactly follow Hillebrecht et al.6 have found from spin-polarized core-level exponential attenuation laws as shown in Fig. 1, lower panel, photoemission a value of 0.5­1 B/atom. Hopster et al.7 in- with perfect consistency of the respective attenuation fer from an exchange splitting of 1.9 eV obtained by spin- lengths14 at the relevant energies. This gives evidence of a polarized electron energy-loss spectroscopy a moment of growth mode without island formation or interdiffusion. We roughly 1.9 B/atom for 1 ML Cr on Fe. Later Turtur and do not expect, however, the growth to be strictly layer by Bayreuther have determined large magnetic moments of layer. The absolute Cr thickness is based on published at- 4 B/atom and a sizable depolarization of the Fe atoms at the tenuation lengths.14 The crystalline structure of the Cr layers interface by in situ alternating gradient magnetometry.8 For is examined by low-energy electron-diffraction LEED Cr deposited on vicinal Fe surfaces the same group reports9 analysis. We find that Cr on Fe 100 displays the same strong deviations from the layered antiferromagnetic struc- LEED pattern as clean Fe for all Cr thicknesses of the ture which are ascribed to particular surface morphologies. present study. The intensity maxima occur at the same elec- Recent tight-binding calculations claim10 that the presence of tron energies as with Fe and no diffuse background is ob- 0163-1829/96/54 13 /9304 4 /$10.00 54 9304 © 1996 The American Physical Society 54 MAGNETIC MOMENTS IN THIN EPITAXIAL Cr FILMS . . . 9305 d n n /2. 4 Using Eqs. 1 ­ 4 it is possible to express the observed spin polarization P in terms of n . This allows one to determine the magnetic moment which is given by (n n ) B from P with only two adjustable parameters or 0 and d .Application of this scheme to a surface of a semi-infinite sample is not favorable since it assumes to be constant for all depths z which usually is not the case. The power of the method, however, lies in the investigation of ultrathin adlay- ers in the submonolayer-to-monolayer range where the mag- netization can be considered to be homogeneous over the adlayer thickness t. In the present study we use it to inves- tigate Cr on Fe 100 . In the case of an adlayer on a magnetic substrate the integration of Eq. 2 yields I t I sexp t i 1 exp t , 5 where I s is the emission from the substrate. It appears to be practical to compute the quantity PI I I in terms FIG. 1. Secondary-electron spin-polarization times intensity PI of the spin polarization of the valence electrons at Ekin 3 eV upper panel and Auger-electron intensities lower P0 (n n )/(n n nsp) and compare it to the ex- panel of Cr/Fe 100 versus Cr thickness. The solid lines represent perimentally determined P(t)I(t). The procedure has al- exponentials with attenuation length 4.7 Å for PI, and 10 Å, 4.2 ready been described in a recent study of V on Fe 100 .1 As Å, and 6.5 Å for the Fe L3M45M45 , Fe M23M45M45 , and a substantial progress, however, we here make use of the Cr L3M23M45 Auger lines, respectively. energy dependence of . At kinetic energies above 20 eV served. Cr is adopting the structure of bcc-Fe. All measure- the spin dependence of the scattering cross section ments are performed at room temperature with a working vanishes.16 Thus we can compare data taken at below 3 eV pressure of 2­5 10 10 Torr. and at 47 eV with the corresponding analyses using d 0.72 The measured secondary-electron spin-polarization P at nm 1 Ref. 15 and d 0, respectively. We indeed find a low energies of a bulk sample is proportional to the sample consistent picture which lends confidence to the validity of magnetization in a surface region of about 4­5 Å Ref. 15 the scheme. thickness. The determination of the magnetic moment per Using energy-resolved SPSEE we measure the spin polar- atom from P then requires quantitative knowledge of the ization P and intensity I versus Cr thickness t. The observed production and emission rates of secondary electrons. For the thickness dependence of the product P(t)I(t) of secondary production we can assume that all valence electrons are ex- electrons with kinetic energies below 3 eV is presented in cited with equal probability i. This yields for each spin Fig. 1, upper panel. For a nonmagnetic overlayer it is ex- for up or down a production rate of pected to exhibit the exponential attenuation of the substrate emission by virtue of the overlayer. The signal strongly de- i i n n viates however, from an exponential attenuation law with sp/2 , 1 attenuation length 4.7 Å Ref. 17 shown as a solid line in where n and nsp are the number of d and sp electrons, Fig. 1, upper panel. This is in pronounced contrast to the respectively, in the valence band. The emission from depth z Auger intensities depicted in the lower panel of Fig. 1. The then is governed by a spin-dependent inelastic scattering comparison with the Auger data demonstrates that the devia- cross section . This gives rise to the secondary-electron tions of the PI signal from an exponential are of magnetic current origin and do not relate to the growth properties of the ad- layer. The difference between the PI data and the exponen- dI i exp z dz. 2 tial background is shown in Fig. 2, upper panel. In principle The key to the present magnetic moment determination lies it could arise from an induced magnetization in the Cr ad- in the quantitative treatment of . Following Siegmann and layer or from a considerable demagnetization of the substrate co-workers15 the scattering cross section in transition met- surface or a combination of the two. In the particular system als can be described as being proportional to the number of d Cr on Fe 100 only a weak depolarization of the interfacial holes h 2(5 n) in the form Fe atoms compared to the bulk value is numerically predicted.2­4,12 Moreover, element-specific measurements by core-hole photoemission of Cr on Fe 100 Ref. 6 as well as 0 d 5 n 3 of Fe on Cr 100 Ref. 18 have shown that the Fe polariza- for sufficiently low kinetic energies of the electrons. The tion and exchange splitting remain unchanged compared to authors were able to determine the parameters 0 and d pure Fe. Based on these observations we neglect a reduction from a compilation of data of various transition metals. of the magnetic moment of the Fe atoms at the interface. We Straightforward application15 of this concept to a ferromag- note that with this assumption the present analysis yields an net then yields upper bound of the magnitude of the Cr moment. It further 9306 P. FUCHS, V. N. PETROV, K. TOTLAND, AND M. LANDOLT 54 FIG. 2. Upper panel Secondary-electron spin-polarization FIG. 3. Secondary-electron spin-polarization times intensity PI times intensity PI at Ekin 3 eV of Cr/Fe 100 versus Cr thickness at E of Fig. 1 after subtraction of the exponential background. Lower kin 47 eV of Cr/Fe 100 versus Cr thickness. Upper panel raw data dots and exponential background with attenuation length panel Magnetization profile obtained from the data. The corre- 4.2 Å solid line . Lower panel Same data after subtraction of sponding calculated PI see text is shown as a solid line in the the exponential background. The solid and broken lines represent PI upper panel. calculations with different parameters of the model see text . turns out that consistency between the data of various elec- sensitivity of PI to spin-dependent scattering we, for com- tron energies only exists for zero reduction of the Fe inter- parison, also have calculated PI with d 0.72 nm 1 as in face moment. the low-energy case. The result dotted line in Fig. 3 Next we attempt to determine a magnetic depth profile strongly deviates from the experimental data. Likewise as a from the PI data. As a first step starting from the bare sub- test we have computed a profile *(t) from the low-energy strate we calculate PI of a small Cr adlayer of thickness PI data setting d 0, i.e., neglected spin-dependent scatter- within which we assume the moment to be constant. We use ing altogether. *, which is not shown here, exhibits quali- Eqs. 1 ­ 5 with the parameters 1/ and d 0.72 nm 1 tatively the same shape as (t) but the absolute values are taken from Ref. 15 and obtain of this Cr adlayer by best fit more than twice as large 4.2 B for 0.3 ML Cr on to the PI data. This step then is repeated for each additional Fe 100 . The corresponding PI curve at 47 eV, however, fraction of adlayer of thickness using the precedent layers does not fit the experimental data at all dash-dotted line in of total thickness t as a new substrate. With this procedure Fig. 3 . This indicates that spin-dependent scattering at low we imply the rather strong assumption that the moment of a energies is important, indeed. The consistency of (t) of given layer does not change upon adsorption of further lay- Fig. 2 for both energies with and without scattering, respec- ers. The resulting magnetization profile (t) is shown in Fig. tively, demonstrates that the present treatment of spin- 2, lower panel, together with the corresponding PI curve as dependent attenuation is meaningful. a solid line in the upper panel. We have shown that the presence of the Fe interface in- In order to gain confidence in the quantitative treatment of duces a large spin polarization in thin Cr adlayers. For the the spin dependence of the inelastic scattering we have re- particular structure of epitaxially grown Cr on Fe 100 at peated the PI measurements at a different kinetic energy of room temperature we find that a Cr adlayer of a submono- the secondary electrons. Figure 3 shows PI at 47 eV kinetic layer coverage has an induced magnetization of 1.8 0.2 B energy as a function of the Cr thickness. This energy corre- per atom and is magnetically oriented antiparallel to the Fe sponds to the Fe M23M45M45 Auger line. It is particularly surface magnetization. The entire (t) in Fig. 2 can be re- suited because the spin dependence of the scattering vanishes garded as a magnetic depth profile of Cr on Fe 100 , but above 20 eV,16 on one hand, and the attenuation length can only with some reservation. We note that a reasonable fit to directly be determined without further parameters from the the data within the first monolayer only is obtained when decrease of the Auger intensity, on the other. The attenuation using fractional monolayers. This reflects the fact that the length turns out to be 4.2 Å. Deviations of PI from the growth at room temperature is not strictly layer by layer, as corresponding attenuation law solid line in Fig. 3, upper has been shown with tunnel microscopy.19 The analysis fur- panel again are significant; the difference is shown in the thermore implies the rather strong assumption that the mag- lower panel of Fig. 3. Using the same magnetic depth profile netic moment in a given fraction of the Cr layer does not (t) as in Fig. 2 obtained from low-electron-energy data we change upon adsorbtion of further layers. This assumption is have calculated the thickness dependence of PI without any correct for the first monolayer but, however, might be wrong adjustable parameters taking 1/ and d 0. The result is for the next few layers near the Fe interface. We emphasize, shown as a solid line in the lower panel of Fig. 3. It exhibits however, that in the range of submonolayer coverage the good agreement with the experimental data. To illustrate the determination of the magnetic moment for thin Cr films on 54 MAGNETIC MOMENTS IN THIN EPITAXIAL Cr FILMS . . . 9307 Fe 100 is a firm result. When going to thicker Cr adlayers It is a pleasure to thank H. C. Siegmann for many fruitful we observe the sign of the induced magnetization to change conversations and continuous support and to K. Brunner for in the first layer and the magnetization to remain positive for expert technical assistance. Financial support by the Sch- thicker Cr films. weizerischer Nationalfonds is gratefully acknowledged. *Permanent address: Division of Experimental Physics, Techni- 10 P. Martin, A. Vega, C. Demangeat, and H. Dreysse´, J. Magn. cal University St. Petersburg, 195251 St. Petersburg, Russia. Magn. Mater. 148, 177 1995 . 1 P. Fuchs, K. Totland, and M. Landolt, Phys. Rev. B 53, 9123 11 R. Coehoorn, J. Magn. Magn. Mater. 151, 341 1995 . 1996 . 12 Zhongde Xu, Y. Liu, P. D. Johnson, and B. S. Itchkawitz, Phys. 2 C. L. Fu, A. J. Freeman, and T. Oguchi, Phys. Rev. Lett. 54, 2700 Rev. B 52, 15 393 1995 . 1985 . 13 M. Landolt, R. Allenspach, and D. Mauri, J. Appl. Phys. 57, 3626 3 R. H. Victora and L. M. Falicov, Phys. Rev. B 31, 7335 1985 . 1985 . 14 M. D. Seah and W. A. Dench, Surf. Interface Anal. 1, 1 1979 . 4 J. Dorantes-Da´vila et al., Surf. Sci. 251/252, 51 1991 . 15 H. C. Siegmann, J. Phys. Condens. Matter 4, 8395 1992 ; G. 5 Y. U. Idzerda et al., J. Appl. Phys. 73, 6204 1993 . Scho¨nhense and H. C. Siegmann, Ann. Phys. 2, 465 1993 . 6 F. U. Hillebrecht, Ch. Roth, R. Jungblut, E. Kisker, and A. 16 H. Hopster, R. Raue, E. Kisker, G. Gu¨ntherodt, and M. Campa- Bringer, Europhys. Lett. 19, 711 1992 ; R. Jungblut, Ch. Roth, gna, Phys. Rev. Lett. 50, 70 1983 ; D. Mauri, R. Allenspach, F. U. Hillebrecht, and E. Kisker, J. Appl. Phys. 70, 5923 1991 . and M. Landolt, J. Appl. Phys. 58, 906 1985 . 7 T. G. Walker, A. W. Pang, H. Hopster, and S. F. Alvarado, Phys. 17 4.7 Å is obtained from intensity data of the present study. It Rev. Lett. 69, 1121 1992 ; H. Hopster private communication . compares well with 4.4 Å from Ref. 15. 8 C. Turtur and G. Bayreuther, Phys. Rev. Lett. 72, 1557 1994 . 18 E. Beaurepaire et al., Surf. Sci. 251/252, 36 1991 . 9 S. Miethaner and G. Bayreuther, J. Magn. Magn. Mater. 148, 42 19 D. T. Pierce, J. A. Stroscio, J. Unguris, and R. J. Celotta, Phys. 1995 . Rev. B 49, 14 564 1994 .