RAPID COMMUNICATIONS PHYSICAL REVIEW B, VOLUME 63, 140406 R Topmost layer magnetization of ultrathin Cr films on Fe 100... from proton-induced spin-polarized electron emission R. Pfandzelter,* M. Ostwald, and H. Winter Humboldt-Universita¨t zu Berlin, Institut fu¨r Physik, Invalidenstrasse 110, D-10115 Berlin, Germany Received 24 January 2001; published 16 March 2001 The magnetic ordering of the topmost surface layer of ultrathin Cr films grown on Fe 100 is studied via spin-polarized electron emission, excited by fast protons grazingly scattered from the film surface. We find that most electrons originate from the topmost layer. Based on simple assumptions we are able to deduce the layer-dependent magnetic moments from the observed spin polarization of electrons. We demonstrate that our method has a clearly smaller probing depth than conventional spin-polarized electron spectroscopies. DOI: 10.1103/PhysRevB.63.140406 PACS number s : 75.70.Ak, 79.20.Rf, 79.20.Hx Fe/Cr/Fe sandwiches and superlattices have evolved into tive spin polarization which gradually decreases with in- a prototype system to study interlayer exchange coupling and creasing film thickness. Using the same technique, although the giant magnetoresistance effect. During recent years, con- without energy resolution, Unguris et al.3,8 succeeded in ob- siderable experimental and theoretical work has been per- serving layer-by-layer oscillations in the spin polariza- formed on these artificially layered magnetic structures lead- tion starting from the fourth layer for growth at 570 K. ing to outstanding discoveries.1 The basic building block of A common feature of techniques based on electron- these multilayers are bare films of Cr epitaxially grown on induced electron emission is a signal stemming from a depth Fe 100 . In this communication we present studies on the region comprising a few layers beneath the surface. Extrac- magnetic structure of ultrathin Cr films on Fe 100 using a tion of the overlayer signal thus requires modeling and sub- different technique: spin-polarized emission of electrons in- traction of a generally spin- and energy-dependent back- duced by grazing proton impact. A striking feature of this technique is the extreme sensitivity to the topmost film layer, ground.2,14 It is evident that this poses problems in cases like in comparison to larger probing depths in conventional Cr on Fe, where interfacial alloying and largely different electron-induced electron emission spectroscopies.2 layer magnetizations are expected in Refs. 3 and 8 Cr- Considerable progress has been achieved concerning overlayer induced features in the observed spin polarization knowledge of the structural and chemical properties of ultra- amount to only a few percent of the underlying background thin Cr films on Fe 100 . Growth is epitaxial and pseudo- polarization for the first few layers grown . morphic in a wide range of temperatures.3­6 Almost perfect In this communication we demonstrate a significant re- layer-by-layer growth is achieved for substrate temperatures duction of the probing depth in spin-polarized electron emis- around 600 K, whereas lower temperatures lead to kinetic sion spectroscopy by using fast protons instead of electrons roughening of the film. Recently it has been observed that as primary particles. The protons are scattered under grazing growth of the first layer leads to the formation of a Cr-Fe angles from the surface layer without penetrating into the alloy,7­9 which is essentially confined to the surface and sub- film ``surface channeling''15 . Thus, in contrast to the al- surface layer. For converges of more than two monolayers most uniform in-depth generation of secondary electrons by ML the topmost layer consists almost exclusively of Cr primary electrons, proton-induced electrons originate mainly atoms.8,9 from the topmost surface layer. Common to both techniques Most studies have focused on the magnetic properties of is that the emission results from a transfer of kinetic energy the Cr films since the seminal discovery of a layer-by-layer of the primary particle to electrons of the film. In this respect antiferromagnetic ordering in several ML thick Cr films.3 our method differs from related techniques, where the poten- Although a number of experimental techniques, mainly tial energy of impinging ions causes emission or capture of based on spin- and, in some cases, energy-resolved scattering spin-polarized electrons.16­19 of electrons, have been applied, the magnetic behavior of the The experiments are performed at a pressure of some first few layers remained unclear. Walker et al.10 infer from 10 11 mbar. The Fe crystal is mounted to close the gap of a slightly negative asymmetries in spin-polarized electron- soft-magnetic yoke with a coil. The crystal is magnetically energy-loss spectroscopy that 1 ML Cr grown at 470 K couples antiferromagnetically to Fe, in agreement with re- saturated by current pulses along an easy axis of magnetiza- sults from magnetically sensitive core level spectro- tion 001 or 001¯ within the 100 surface plane and per- scopies.11­13 Upon further deposition, the asymmetry main- pendicular to the scattering plane. With the magneto-optic tains the same sign as for clean Fe up to coverages of about Kerr effect we find a square-shaped hysteresis loop with full 8 ML. A similar behavior is observed using electron-excited remanence near the center of the target and a reduced rema- spin-polarized electron emission secondary electron emis- nence near the edges. All data are recorded in the remanent sion spectroscopy for room temperature growth.14 After state. The Fe target is prepared by cycles of grazing sputter- background subtraction, a negative spin polarization is in- ing with 25 keV Ar ions and subsequent annealing, until ferred for the 1 ML Cr film. Subsequent layers show a posi- the surface is clean and flat mean terrace width 1000 Å , 0163-1829/2001/63 14 /140406 4 /$20.00 63 140406-1 ©2001 The American Physical Society RAPID COMMUNICATIONS R. PFANDZELTER, M. OSTWALD, AND H. WINTER PHYSICAL REVIEW B 63 140406 R ticles b . The spectra were recorded at 610 K during Cr growth. In order to discriminate secondary electrons gener- ated at surfaces in the chamber other than the target surface, we biased the target by 10 V with respect to ground.22 We estimate an increase of the effective solid angle at small elec- tron energies electron trajectory calculations show 30% at 5 eV , which at least partly compensates for the energy depen- dence of the transmission due to residual magnetic stray fields. The intensity distributions in Fig. 1 curves exhibit the well-known behavior: a pronounced peak with a maximum at 1­2 eV and a gradual decrease towards higher energies. The peak is usually ascribed to cascade multiplication owing to kinetic electron-electron collisions, whereas direct excita- tion of electrons by primary particles should dominate at higher energies. Although details of the intensity distribu- tions will be discussed elsewhere, we would like to mention two observations here: i Aside from a shift of the maxi- mum to smaller energies, the effect of the Cr coverage is weak. ii The cascade peak is less pronounced for H exci- FIG. 1. Normalized intensity distribution curves, right-hand or- tation, but we do not observe significant differences between dinate and spin polarization symbols, left-hand ordinate of elec- electron and proton excitation. This is in contrast to Ref. 19, trons excited by 25 keV H ions a and 4 keV electrons b . The where similar experiments on clean surfaces are interpreted spectra refer to the clean Fe 100 surface bottom and Cr coverages solely in terms of potential emission by ion impact. averaged over a coverage range of 0.2 ML in each case as The observed spin polarization Fig. 1, symbols is in indicated. The polarization is calculated from the measured asym- agreement with previous studies on electron-induced elec- metry with a Sherman function of 0.2. The origins are displaced tron emission:23,24 It is largest for small energies and falls vertically by constant amounts dashed lines . rapidly within the range of cascade electron energies Fig. 1 b . This is also observed for proton excitation Fig. 1 a . as checked by low-energy electron diffraction LEED , This similarity between proton and electron excitation is re- Auger spectroscopy, and angular distributions of scat- markable, as it shows that cascade effects are important even tered ions.20 for grazing angles of incidence. We note that, for the clean Cr is sublimated by electron beam heating. Growth is Fe surface, the polarization is smaller for proton impact than monitored in situ and in real time by measuring the specular for electron impact. This can be, at least partly, ascribed to intensity of grazingly scattered ions. This technique enables the larger source area reduced magnetization near the one to calibrate the coverage and quantitatively determine sample edge and the smaller probing depth thermally re- the film morphology. For the present study we choose a duced magnetization at the surface . growth temperature of about 610 K, where growth is found Although the overall shape of the spin polarization spectra to proceed in an almost perfect layer-by-layer mode, in ac- hardly changes upon growth of Cr, we observe for proton cordance with our previous studies.6 Typical growth rates are excitation a strong and nonmonotonic dependence of the val- some 10 4 ML s 1. ues of the polarization on the coverage. This is in clear con- Emitted electrons are collected within a cone of about 12° trast to the gradual decrease observed for electron excitation. full opening angle around a direction of 10° off normal and The effect becomes more evident when the measured spin enter a cylindrical sector field energy analyzer via a transfer polarization is averaged over the whole spectral range ex- lens CSA300, Focus . After energy separation and 90° cluding small energies E 2 eV) and plotted against the Cr deflection, electrons are imaged by another lens into a coverage. For proton excitation Fig. 2 a the spin polariza- LEED spin-polarization detector.21 Pass energy 80 eV , en- tion follows a series of roughly linear variations. The break- ergy resolution 3.0 eV full width at half maximum point positions coincide with integer ML coverages. This FWHM , and LEED scattering energy 104.5 eV are kept oscillatory behavior clearly differs from electron excitation constant during energy scans. Each polarization spectrum is Fig. 2 b , where, aside from the first layer, the layer-by- obtained from two identical measurements with reversed layer growth of Cr does not appear in the observed polariza- magnetization to eliminate instrumental asymmetries and tion. We note that the shape of the spin polarization curves checked by measurements on a paramagnetic Ta foil attached Fig. 2 hardly depends on the choice of energy interval for directly near the Fe sample. energies larger than a few eV, if normalized to the data for In Fig. 1 a we show intensity distribution curves and the clean Fe surface. spin polarization symbols of electrons excited by 25 keV Considering the conceptual and physical similarity of H ions incident at a grazing angle of 1.2° upon the clean both experiments, the striking difference between the ob- and Cr-covered Fe 100 surface. For comparison we also use served polarization curves may be ascribed to different prob- 4 keV electrons at oblique incidence 33° as primary par- ing depths for proton and electron excitation, respectively. 140406-2 RAPID COMMUNICATIONS TOPMOST LAYER MAGNETIZATION OF ULTRATHIN Cr . . . PHYSICAL REVIEW B 63 140406 R FIG. 3. Layer-dependent magnetic moments per atom for Cr/ Fe 100 obtained from a fit to the data from Fig. 2 according to Eq. 1 . Layers 0* and 1* alloyed layers are marked by * refer to the subsurface and surface layer, respectively, after deposition of 1 ML of Cr. The error bars cover statistical errors and systematic errors due to the choice of electron energy interval and the uncertainty of probing depth. tion, attenuation lengths have been compiled in Ref. 25. Here s 4.2 ML Fe and a 2.9 ML Cr we neglect the spin FIG. 2. Spin polarization symbols of electrons excited by im- dependence observed for the smallest energies E 10 eV). pact of 25 keV H ions a and 4 keV electrons b on Cr/Fe 100 On the other hand, for grazing proton impact, we expect versus Cr overlayer thickness. The solid lines represent model cal- 0. Yet there is a finite probability for the protons to pen- culations assuming the same layer-dependent magnetization profile etrate the surface via ledges of islands, substrate steps, or but different probing depths for proton and electron excitation see thermally displaced atoms. From trajectory simulations20 we text . The data have been normalized to the polarization for the estimate s a (0.5 0.3) ML. clean Fe surface. The layer-dependent magnetic moments obtained from a fit to the data for proton excitation Fig. 2 a , solid lines are Whereas grazingly scattered protons mainly excite electrons shown in Fig. 3. As expected, the plot closely resembles the from the top-most film layer, the probing depth for electron spin polarization curve. Note, however, that the moments excitation amounts to several layers. An antiferromagnetic deduced for the alloyed layers 0*, 1*, and 2* represent av- stacking of layers, as expected for Cr/Fe 100 , thus causes a eraged moments of Fe and Cr atoms. Our results show that gradual decrease of the measured polarization with increas- there are significant deviations from the layer-by-layer oscil- ing coverage, in agreement with our observation and previ- lations of Cr moments observed for thicker layers.3 This is ous studies on electron-induced electron emission.3,14 Never- most evident for coverages of 1 or 2 ML, where considerable theless, we do not observe a layer-by-layer change in sign for alloying occurs. From the third monolayer, the observed mo- proton impact, in agreement with the highly surface sensitive ments are similar to the rms Cr bulk moment about 0.4 experiments of Walker et al.10 B). Aside from the unexpected magnetic ordering at the begin- The layer-dependent magnetic moments can be estimated ning of growth, our data suggest another irregularity between by fitting the data for proton excitation from Fig. 2 a , as- layers 4 and 5. This would explain the anomalous phase of suming a proportionality between spin polarization and mag- the magnetic stacking observed in thicker layers.3 netization, Note that an averaged moment of 1.4 B for the inter- face layer 0* would imply huge negative moments for Cr nBa,i exp zi/ a nBs s exp d/ a atoms ( 4.7 B), if we assume Fe moments as in the bulk i P , 1 (1.92 B). Such huge Cr moments ( 4.5 B) have been also observed by Turtur and Bayreuther26 during initial na,i exp zi / a ns s exp d/ a growth at room temperature by absolute magnetometry. i With the layer-dependent magnetic moments from Fig. 3, where ns(na,i) is the number of conduction electrons per we calculated the spin polarization curve for electron excita- atom in the substrate film layer i , nBs(nBa,i) the Bohr mag- tion Fig. 2 b , solid lines , also by using Eq. 1 , but with neton number, d the film thickness, and s( a) the attenua- larger attenuation lengths. The consistency of the data sets tion length of electrons in the substrate film . na,i is slightly corroborates our assumption of different probing depths in layer dependent due to interfacial alloying.7,9 We assume proton and electron excitation. We note that the reverse pro- nBs 0.9 2.13 1.92 the prefactor takes account of thermal cedure deduction of moments from measured spin polariza- spin excitations to be independent on the depth. For a uni- tions leads to ambiguous results. form in-depth generation of electrons as in electron excita- In summary, we studied the magnetic ordering of ultrathin 140406-3 RAPID COMMUNICATIONS R. PFANDZELTER, M. OSTWALD, AND H. WINTER PHYSICAL REVIEW B 63 140406 R Cr films on Fe 100 by spin-polarized electron emission. The face sensitivity for proton excitation enables one to deduce films were grown under conditions where growth is almost magnetic moments of the topmost film layer in a straightfor- perfectly layer-by-layer and the chemical composition is well ward manner. known. In addition to conventional excitation by electrons, we used grazing impact of fast protons. The observed spin The assistance of T. Igel, A. Laws, K. Maass, and R. A. polarization of emitted electrons shows layer-by-layer oscil- Noack in the preparation of the measurements is gratefully lations for early Cr growth, in contrast to the gradual de- acknowledged. This work was supported by the Deutsche crease observed for excitation by electrons. The extreme sur- Forschungsgemeinschaft Sonderforschungsbereich 290 . *Email address: pfandz@physik.hu-berlin.de FAX: 49 30 2093 12 Z. Xu, Y. Liu, P. D. Johnson, and B. S. Itchkawitz, Phys. Rev. B 7899. 52, 15 393 1995 . 1 K. B. Hathaway, A. Fert, P. Bruno, D. T. Pierce, J. Unguris, R. J. 13 G. Panaccione, F. Sirotti, E. Narducci, and G. Rossi, Phys. Rev. B Celotta, S. S. P. Parkin, in Ultrathin Magnetic Structures II, 55, 389 1997 . edited by B. Heinrich and J. A. C. Bland Springer, Berlin, 14 P. Fuchs, V. N. Petrov, K. Totland, and M. Landolt, Phys. Rev. B 1994 , p. 45, and references therein. 54, 9304 1996 . 2 15 H. C. Siegmann, J. Phys.: Condens. Matter 4, 8395 1992 . D. S. Gemmell, Rev. Mod. Phys. 46, 129 1974 . 3 16 J. Unguris, R. J. Celotta, and D. T. Pierce, Phys. Rev. Lett. 69, C. Rau and R. Sizmann, Phys. Lett. 43A, 317 1973 . 17 1125 1992 ; D. T. Pierce, R. J. Celotta, and J. Unguris, J. Appl. F. B. Dunning, C. Rau, and G. K. Walters, Comments Solid State Phys. 73, 6201 1993 . Phys. 12, 17 1986 . 18 4 B. Heinrich and J. F. Cochran, Adv. Phys. 42, 523 1993 . H. Winter, H. Hagedorn, R. Zimny, H. Nienhaus, and J. Kir- 5 D. T. Pierce, J. A. Stroscio, J. Unguris, and R. J. Celotta, Phys. schner, Phys. Rev. Lett. 62, 296 1989 . 19 C. Rau, K. Waters, and N. Chen, Phys. Rev. Lett. 64, 1441 Rev. B 49, 14 564 1994 . 6 1990 ; C. Rau and K. Waters, Nucl. Instrum. Methods Phys. R. Pfandzelter, T. Igel, and H. Winter, Surf. Sci. 375, 13 1997 . 7 Res. B 40Õ41, 127 1989 . D. Venus and B. Heinrich, Phys. Rev. B 53, R1733 1996 . 20 R. Pfandzelter, Phys. Rev. B 57, 15 496 1998 . 8 A. Davies, J. A. Stroscio, D. T. Pierce, and R. J. Celotta, Phys. 21 J. Kirschner, Polarized Electrons at Surfaces Springer, Berlin, Rev. Lett. 76, 4175 1996 ; A. Davies, J. A. Stroscio, D. T. 1985 . Pierce, J. Unguris, and R. J. Celotta, J. Magn. Magn. Mater. 165, 22 J. Kirschner and K. Koike, Surf. Sci. 273, 147 1992 . 82 1997 . 23 J. Unguris, D. T. Pierce, A. Galejs, and R. J. Celotta, Phys. Rev. 9 R. Pfandzelter, T. Igel, and H. Winter, Phys. Rev. B 54, 4496 Lett. 49, 72 1982 . 1996 ; Surf. Sci. 377, 963 1997 . 24 H. Hopster, R. Raue, E. Kisker, G. Gu¨ntherodt, and M. Campa- 10 T. G. Walker, A. W. Pang, H. Hopster, and S. F. Alvarado, Phys. gna, Phys. Rev. Lett. 50, 70 1983 . Rev. Lett. 69, 1121 1992 . 25 G. Scho¨nhense and H. C. Siegmann, Ann. Phys. Leipzig 2, 465 11 F. U. Hillebrecht, C. Roth, R. Jungblut, E. Kisker, and A. Bringer, 1993 . Europhys. Lett. 19, 711 1992 . 26 C. Turtur and G. Bayreuther, Phys. Rev. Lett. 72, 1557 1994 . 140406-4