Electron-energy-loss spectroscopy of Fe thin films on GaAs(O01) J. Yuan, E. Gu, M. Gester, J. A. C. Bland, and L. M. Brown Cawndish Laboratoq, Cambridge CB3 OHE, United Kingdom An electron-microscopy-based technique of electron-energy-loss spectroscopy (EELS) has been used to characterize electronic and magnetic properties of ultrathin Fe films grown on GaAs(J00) surface> as a fun&on of the film thickness. Large-area electron transparent membranes for microscopic analysis are prepared by ion-beam thinning or chemical etching from the substrate side, and the toJ> surface of the ultrathin Fe film is protected by a thin Cr layer. Analysis of the Fe 2p, Cr 2p, and 0 Is absorption spectra confirms that only the Cr layer is oxidized. The locat magnetic moments of the ultrathin Fe films are deduced from the "white line" branching ratio in the Fe 2p absorption spectra. For Fe films as thin as 150 I%, the magnetic moment is not different from that found in bulk *Fe. For a 70-a Fe film, the local magnetic moment is enhanced although the average magnetization is reduced. As doping is suspected to be the cause for the departure from bulk (~-u-Fe properties. Jn the case where the 50-A film is poiycrystalline and discontinuous, spatially resolved EELS has been used to distinguish small island clusters from large crystalline particles. The large particles are aFe crystallites and the islands are probably also heavily affected by As doping. 1. INTRODUCTION (LEED) techniques can be found in Ref. 6. At the end of Fe Epitaxially grown Fe thin films on GaAs( 100) hold great deposition, the top surface is covere.d with a thin layer of Cr promise for the integration of magnetic film technology into to prevent the oxidation of Fe films. To produce electron- semiconductor and opto-electronic technologies.`-" How- beam-transparent samples for transmission electron micros- ever, many physical propertie-s of the thin films have a strong copy and transmission electron-energy-loss analysis, the thickness dependence and the origins of these are currently sandwich film is thinned from the substrate side only. This is poorly undcntood. For example, the average magnetization achieved by either ion-beam sputtering or by chemical etch- per Fe atom starts to decrease from the bulk value for films ing. The latter is partic.ularly favored since a large uniform as thick as 230 A.' To understand the microscopic origin of thin membrane can be produced using a thin layer [about 0.2 this behavior, we have to se:parate the intrinsic physical ef- /~.m thick) of the relatively chemically inert Ga,&l,,As as fects from the effect of defects, impurities, etc., associated an "ctcting stop" layer' on top of the CaAs(100) substrate. with the growth problem. Thiis can be achieved by studying EELS is conducted inside a dedicated vacuum gene.rate the Fe film with a Ioctzlized mng~etic probe. In this article, (VG)-HE5501 STEM operating at 100 keV. The transmitted we carried out such a characterization of thin and ultrathin Fe electrons are analyzed by a VG magnetic sector prism, and films by studying element-selective inner-shell ionization us- the spectrum is further magnified by a quadrapoie lens sys- ing electron-energy-loss spectroscopy (EELS)," The EELS tem before being collected on a CCD {charge-coupled measurement is carried out using a scanning transmission device&based parallel detection system.7 All the spectra have electron microscope (STEM), which not only allows us to been corrected for the dark current and flat-fidd response of study uniformly deposited films, but also permits analysis of the CCD camera. The convergence semiangle of the incident inhomogeneous films with high spatial resolution. All the Fe electron is about 7 mrad, and the collection semiangle of the fJJms are preserved from oxidation and degradation by e.n- spectrometer is about 8 mrad. Under such experimental con- capsulation between the substrate and a top protective layer. ditions, inner-shell transitions of interest (0 is at 530 eV,.Cr This distinguishes us from early studies of transition metal 2p at 576 cV, and Fe 2p at 706 eV> all obey the dipole- films that are thermally evaporated with poor control for im- selection rule. The spatial resolution of the focused electron purity and crystallinity and inevitably contains surface oxide probe is better than 5 A and the optimal energy resolution of layers. We will be able to show from the Fe 2p absorption, the EELS system is 0.3 eV. which accesses to the fmal states involving the Fe 3d con- duction hand, that the electronic structure of Fe films grown Ill. RESULTS on the GaAs(lOOj surface is bulkliie at 150 A but different A. Cr@5 &Fe(l50 &/GaAs(OOl) structure from the bulk for 70-A and 50-A ultrathin films; and that the The magnetization measurement of this 1.50-w Fe film difference is consistent with As doping. shows a pronounced fourfold anisotropy.' Thin membrane of II. EXPERIMENT the sandwich structure is produced by chemical etching. The electron diffraction pattern is consistent with an epitaxial The ultrathin Fe films are deposited on a GaAs(O01) sub- a--Fe film lattice-matched on a GaAs(001) substrate. The strate inside a UHV chamber with a base pressure better than composition analysis using the theoretical cross section? for 5X10-" mbar during the growth. A detai1e.d characterization the continum part of the Cr and Fe 2p absorption shows that of the growth process by both reflection high-energy electron the Fe and Cr concentrations are in the ratio 6.33, close to diffraction (RHEED) and ilow-energy electron diffraction the expected ratio of 6.10, which is calculated from the J, Appl. Phy8. 75 (lo), 15 May 1994 0021-6979194/75/10)/6501/3/$6.00 Q 1994 American Institute of Physics 6501 Downloaded 16 Sep 2002 to 148.6.178.13. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp Fe 2p absorption by EELS tinuous and that the grains have two characteristic sizes. The 27ooo Fe is mostly sparsely distributed in clusters of small islands of the order of 50 A across. From the analysis of the ioniza- f tion cross sections, using that of the f 19750 Cr layer as a standard, the thickness of the Fe islands is deduced to be about 20-30 e c A. The Fe 2p absorption spectrum from one of these 2 12500 nanometer-size.d islands is shown in Fig. 1. Lt is similar to a 0 that of the 70-A film, except for an additional shoulder at B 710 eV. Much less numerous are the large Fe particles with $ 5250 diameters over several thousand angstroms. The energy-loss Ei spectrum from the large particles has been taken as represen- -2ocM tative of the bulk Fe films and is also presented in Fig. 1 for 695.0 705.0 715.0 725.0 735.0 comparison purpose. Energy (A') IV. DLSCUSSION FIG. 1. The Fe 2p absorption spectra from ultrathin films of different thick- ncss. The bulk reference spectra is taken from 2000-8 thick particles (see In a one-electron picture, the inner-shell transition the text for detail). should give information about a symmetry-projected, empty density of states localized on the excited atoms, and the re- sult can be compared with the bandstructure calculation. In thickness difference of the two films as measured by quartz the case of Fe 2p absorption, it probes the empty density of microbalance, in situ. The thickness of the membrane is es- states of the conduction band with the 3d, and to a less timated from the low-loss spectrum to be about 300400 A extent 4s character. However, strong coupling between the and is very uniform. An analysis of the Cr 2p absorption partially filled d band and the remaining inner-shell p elec- spectra indicates that at least a part of Cr has been converted trons produce a significant distortion to the empty density of into oxides.' The Fe 2p absorption for the thin film is shown state. The spectra are dominated by two "white-line" struc- in Fig. 1. It is similar to that for the bulk Fe absorption tures separated by 12.5 eV, approximately the energy differ- spectrum shown on the same figure, with the energy of the ence between the Fe 2pla and 2~~~~ core levels. Information first peak at 706 eV. There is no sign of Fe oxidation since its about the electronic structure of the Fe film can still be as- white line would appear at much higher energies." sessed in these cases. For example, the density of the con- duction electrons controls the lifetime. of the core hole states, 6. Cr(25 &Fe(70 &GaAe(OOl) structure hence the width of the core leve.1 transition. In bulk metal Fe The magnetization of this 7!1-A Fe film shows a mixture (see Fig. 1), the width of the white line is quite large (3.5 of fourfold and twofold anisotr0py.s As there is no built-in eV), reflecting a large conduction electron density around Fe. etching stop on the GaAs substrate, ion-beam thinning was It is reduced to less than 2.1 eV for the 70-A and 50-A films, used to produce electron-beam transparent membrane. From indicating a large reduction in the density of the conduction electron diffraction analysis, the Fe film growth is again electron. Because of this reduction in the width, the multiplet found to be epitaxial. To avoid areas where the GaAs sub- structures in the white lines becomes visible for the spectra strate has been completely removed, we deliberately move to from these two films. These are due to resonant transition areas where the sandwich structure is over 700 A thick and from the 2y'j3dn initial states into 2p"3d"' t final states, and where signal from the GaAs substrate is unmistakable. With are sensitive to the cl-level occupancy of the Fe atom. Com- at least about 600 A of the GaAs substrate still intact, the parison of line shape of l, peak with an atomic multiplet possibility of ion-beam-induced effects on the thin Fe film is calculationi suggests that the average d-level occupancy (n) very small. The Fe 2p absorption spectrum taken from this in the 70-A film is 6, assuming the crystal-field effect is thick sample is deconvoluted to remove multiple scattering, unimportant and the result is also shown in Fig. 1. lt is clear that the 70-A Because of the spin-orbit coupling, the branching ratio Fe film is already different from the bulk and 150-A Fe film. of the white lines arising from transitions from the two sub- Thee first peak in the Fe 2p absorption spectrum from the levels of the Fe 2p state (defined as 1,(2p.&/ 70-A Fe film has a shoulder at 706 eV where the 2p absorp- [12(2p ,Izj +1;(2p&j) deviates from the statistical values tion of the bulk Fe has its maximum. i (Ref. 11) and dc.pends on the specific configuration of the partially filled 36 levels." Although a first-principle calcula- C. Cr(25 &Fe@0 &GaAs(OOi) structure tion of these effects is still not feasible for the time being, an empirical relationship has been found between the local This film is slightly different in its fabrication from the magnetic moment per Fe atom (which depends on the 3d above two lilms. The GaAs wafer containing the etching- configuration of the Fej and the branching ratio of the Fe 2p stop layer has been thinned to produce a membrane window absorption." The branching ratio of Fe 2p absorption from prior to Fe deposition. The magnetization of the film is found the bulk and 150-A Fe is found to be 0.74 (13j12=2.9j, in to be isotropicX and is associated with the polycrystalline agreement with that reported from the literature.`"*i" The cor- nature of the film, as shown by electron diffraction. In addi- responding branching ratio for the 70-A Fe film is 0.83 tion, electron microscopy shows that the Fe film is discon- (&/I,=5). This indicates an enhanced magnetic moment per 6502 J. Appt. Phys., Vol. 75, No. 10, 15 May 1994 hat7 et al, Downloaded 16 Sep 2002 to 148.6.178.13. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp atom according to the empirical relationship of Kurata and Special sandwich structure has been produced to prevent the Trinaka.1' To translate bcaf magnetic moments into volume oxidation of the ultrathin Fe fiIm at ambient atmosphere. magnetization, we need to consider the nature of IocaI mag- This allows, for the first time, observation of the electron- netic ordering. The brdnchillg ratio of the thinner film ("50 energy-loss spectrum of Fe metal films not tainted by the &`) film is not available due to poor statistics of the spec- surface oxide. For the lSO-A film, the Fe La,3 absorption trum. edge is similar in shape to that obtained from the bulk. But The ca.se for the dqmrture from the bulk Fe signal in the spectrum from the 70-w Fe film is very different. The the thin and ultrathin films is not clear. The effect of low local magnetic moment deduced from the branching ratio of dimensionality is unlikely in this particular case as the Fe 2p absorption spectra of a two monolayer and a four monolayer the Fe 2p spectrum from the 70-A Fe film is higher than that Clrn on Cu(100) substrate, obtained using in sitrt x-ray from the bulk. For the 50-w fiIm examined, the magnetiza- absorption, I3 is similar to that found for bulk Fe and the tion behavior is isotropic because the film is polycrystalline. 150-A film. A more likely cause maybe due to the atomic The film is also discontinuous and has two distinct grain diffusion across the boundary. The CrRe bilayer has been sizes. The electronic structure of the small Fe islands (about studied extensively in terms of ultrathin film, and multilayer 50 A across) is similar to that of the 70-~% Fe film. The larger fabrication and interdiffusion is generally not a problem. grains have diameters of the order of thousand angstroms Contamination of the Fe film by more diffusive As species and bulklike properties. The exact nature of the electronic. has been found in earlier deposition experiments,"q1"8'5 and structure of the deposited Fe films are still under investiga- the formation of antiferromagnetically coupled Fe,?As micro- tion. It is suspected that the changes are partially caused by crystallites has been speculatted as a possible reason for the the proximity of the interface and the related As diffusion drop of the average magnetization per atom in thin film." across the boundary, which can induce both doping as well There are some evidence for As contamination of Fe film as alloying. from irz situ Auger spectroscopy; however, there is no elec- tron diffmction evidence for FezAs microcrystalline particles in the films we have examined. On the other hand, the As- doped CX-Fe is also possible in the equilibrium phase diagram of the Fe-As binary system'" and cannot be dismissed purely from our plan-view electron microscopy and EELS results. `J. hl. Florczak and E. D. Dahlbcrg, Phys. Rev. B 14, 033X i1991). `G. A. Prinz and J. J. Krebs, Appl. Phys. Lett. 39, 397 (1981). We may speculate on the electronic structure of the As-doped 3K. T. Riggs, E. D. Dahlberg, and G. A. Prinz, J. Magn. Magn. Mat. 73, 46 Fe film. Crudely speaking, the more ele.ctronegetive As atom (1988). will be an electron acceptor and reduces the d-level occu- `.I. J. Krehs. B, T. Jonker, and G. A. Prinz, J. Appl. Phys. 61.2596 (19.873. pancy in the Fe atom. This is consistent with the reduced ' R. F. Egerton, Electron Energy Luss Spectroscopy in the Electron Miero- branch ratio and decrease in the line width of the white lines scope (Plenum, New York, 1986). in the Fe 21' absorption spectra from the Xl- and 50-A films. "E, Su, (1. Daboo, .I. A. C. Eland, M. Fester, A. J. R. Ives, L. M. Brown, According to the empirical relationship between branch ratio and J. N. Chapman. J. Mayn. Magn. Mater. 124, 180 (1993). `D. McMullan, .I. M. Rodenburg, Y. Murooka, and A. J. McGibbon, Inst. and the local magnetic moment per atom? the magnetic mo- Phys. Conf. Ser. 98, 55 (1989). ment per atom for the As-induced cr-Fe film may actually `C. Dahoo, R. J. Hicken, D. E. P. Eley, M. Gcster, S. J. Gray, A. J. R. Ives, increase (from 2.5 to 6 pR). This finding may not necessary and J. A. C. Bland {these proceed&s). be in confict with the decrease in the average magnetization "R. D. Leapman, L. A. Grunes, and P. Fejes, Phys. Rev. B 26, 614 (1982). per atom, as the relationshi,p between the two are not neces- "IG. van der Laan and I. W. Kirkman, J. Phys. Conden. Mat. 4.4189 (1992). sary in tandem parficularly for an itinerant ferromagnetic Fe " R. D. Leapman and L. A. Grunts, Phys. Rev. Lea. 15, 397 (1980). " II. &U&i and N. Tanaka, Microsc. MicKXiii?d. Microstrnct. 2, 183 (1991). sample, since the local ordering of the atomic moments is "J . G Tobin, 1 Cr. D. Waddill, and D. I'. Pappas, Phys. Rev. Lctt. 68, 3642 also important and may also be atierted by As doping. 1.1992). "I R. W&-Imp ancl R. W. Grant, Appl. Phys. Lett. 34, 630 (1979). `,, A, Chamhcrs, F. Xu, H. W. Chen, S. B. Vitomirov, S. B. Anderson, and V. SUMMARY J.`H. Weaver, Phys. Rev. B 34, 660.5 (1986). Thin Fe films grown on GaAsj100) surf&e have been IhF . A . Shunk, Consfimfion of Rinay AfZo~s, 2nd suppl. (McGraw-Hill, examined by electron-micn,scc~py-leased high-energy EELS. New York, 1949). .I. Appl. Phys., Vol. 75, No. 10, 15 May 1994 Yuan et al 6503 Downloaded 16 Sep 2002 to 148.6.178.13. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp