PHYSICAL REVIEW B VOLUME 53, NUMBER 14 1 APRIL 1996-II Soft-x-ray fluorescence study of buried silicides in antiferromagnetically coupled Fe/Si multilayers J. A. Carlisle, A. Chaiken, R. P. Michel, and L. J. Terminello Materials Science and Technology Division, Lawrence Livermore National Laboratory, Livermore, California 94551 J. J. Jia and T. A. Callcott University of Tennessee, Knoxville, Tennessee 37996 D. L. Ederer Tulane University, New Orleans, Louisiana 70118 Received 18 December 1995 Soft-x-ray fluorescence spectroscopy has been employed to obtain information about the Si-derived valence- band states of Fe/Si multilayers. The valence-band spectra are quite different for films with and without antiferromagnetic interlayer exchange coupling, demonstrating that these multilayers have different silicide phases in their spacer layers. Comparison with previously published fluorescence data on bulk iron silicides shows that the Fe concentration in the silicide spacer layers is substantial. Near-edge x-ray-absorption data on antiferromagnetically coupled multilayers in combination with the fluorescence data demonstrate unambigu- ously that the silicide spacer layer in these films is metallic. These results on the electronic structure of buried layers in a multilayer film exemplify the wide range of experiments made possible by high-brightness syn- chrotron sources. I. INTRODUCTION strongly intermixed iron silicide spacer layer, which was variously hypothesized to be a metallic compound in the Multilayer films made by alternate deposition of two ma- B2 CsCl structure4 or a Kondo insulator in the more complex terials play an important role in electronic and optical de- B20 structure.6 If the spacer layer is not metallic, then the vices such as quantum-well lasers and x-ray mirrors.1 In ad- usual theories of interlayer exchange coupling do not apply5 dition, phenomena like giant magnetoresistance and and the coupling must involve a novel mechanism. Using dimensional crossover in superconductors have emerged transmission electron microscopy TEM , the spacer layer from studies of multilayers. While sophisticated x-ray tech- has been identified as a metastable cubic iron silicide closely niques are widely used to study the morphology of lattice-matched to bulk Fe.7 However, since the exact stoichi- multilayer films, progress in studying the electronic structure ometry of the silicide was not determinable by diffraction has been slower. The short mean-free path of low-energy means, the question of whether the spacer layer is a metal or electrons severely limits the usefulness of photoemission and not has remained unanswerable. SXF and NEXAFS are ideal related electron spectroscopies for multilayer studies. Soft-x-ray fluorescence SXF is a bulk-sensitive photon- techniques to resolve exactly this type of issue. in, photon-out method to study valence-band electronic SXF and NEXAFS measurements were performed on five states.2 Near-edge x-ray-absorption fine-structure spectros- different Fe/Si multilayer films at the Advanced Light Source copy NEXAFS measured with partial photon yield can give on beamline 8.0, which is described in detail elsewhere.8 complementary bulk-sensitive information about unoccupied SXF data has previously been used to study buried layers of states.3 Both these methods are element specific since the BN Ref. 9 and Si.10 Data taken at the Fe L edge closely incident x-ray photons excite electrons from core levels. By resembles bulk Fe for all Fe/Si multilayers. Previously pub- combining NEXAFS and SXF measurements on buried lay- lished x-ray diffraction show that the Fe layers in Fe/Si mul- ers in multilayers and comparing these spectra to data on tilayers have a lattice constant close to bulk Fe.7 In conjunc- appropriate reference compounds, it is possible to obtain a tion with the Fe-edge SXF measurements, these data suggest detailed picture of the electronic structure. that silicide formation occurs through the diffusion of a small The Fe/Si multilayer system well illustrates the power of amount of Fe into Si rather than the diffusion of Si into Fe. combining the SXF and NEXAFS techniques. Fe/Si multi- The NEXAFS spectra were acquired by measuring the total layers exhibit a large antiferromagnetic AF interlayer ex- Si L emission yield with the same detector used for fluores- change coupling that is apparently similar to that previously cence. The resulting data are expected to be comparable to observed in metal/metal multilayers like Fe/Cr.4 The obser- those acquired by electron counting.3 The films used in this vation of strong antiferromagnetic coupling was initially sur- study were grown using ion-beam sputtering IBS in a prising, since this coupling is believed to be a manifestation chamber with a base pressure of 2 10 8 Torr.7 All multi- of spin-density oscillations in the nonmagnetic metallic layers were characterized using x-ray diffraction and magne- spacer layer of a multilayer.5 The interpretation of the Fe/Si tometry. The incident photon energy calibration for the coupling data was hampered by lack of knowledge about the NEXAFS data was established by comparison of the c-Si L 0163-1829/96/53 14 /8824 4 /$10.00 53 R8824 © 1996 The American Physical Society 53 SOFT-X-RAY FLUORESCENCE STUDY OF BURIED . . . R8825 FIG. 2. SXF Si L emission spectra for crystalline and amor- phous silicon films. These data were taken with an incident photon energy of 132 eV. FIG. 1. Magnetization curves for three Fe/Si multilayers. The shown in Fig. 1. The FeSi2 data has two primary features, y axis shows magnetization data normalized to the saturated value. namely s-orbital features near 90 eV, and a shoulder which The solid line indicates data for a polycrystalline Fe 30 Å/Si 20 Å extends up to 99 eV and is comprised mostly of states with 50 multilayer which has a magnetization curve much like bulk Fe. The open circles indicate data for an epitaxial Fe 40 Å/Si 14 Å d symmetry. These features have been previously identified 40 multilayer which has the high saturation field and low rema- in semiconducting bulk FeSi2 specimens.11 nent magnetization that are characteristic of antiferromagnetic in- In Fig. 3 the spectrum for the polycrystalline antiferro- terlayer exchange coupling. The polycrystalline Fe 30 Å/Si 14 Å magnetically coupled multilayer with tSi 14 Å looks simi- 50 multilayer indicated by filled circles has weaker antiferro- lar to the FeSi2 data, while the spectrum for the polycrystal- magnetic coupling than the epitaxial multilayer. absorption to other published work.13 Using these methods, the relative energy calibration error between incident and emitted photons is estimated to be less than 0.1 eV. II. RESULTS Figure 1 shows hysteresis loops for three representative Fe/Si multilayers. The polycrystalline Fe 30 Å/Si 20 Å 50 multilayer grown on glass has a magnetization curve that shows no sign of interlayer exchange coupling. This multilayer has magnetic properties like those of bulk Fe. The epitaxial Fe 40 Å/Si 14 Å 40 multilayer grown on MgO has a low remanent magnetization and a high saturation field, which are the classic signs of antiferromagnetic interlayer coupling. Data on the polycrystalline Fe 30 Å/Si 14 Å 50 multilayer fall somewhere in between these two ex- tremes. Detailed characterization of these films has been published previously.7 For purposes of comparison to the Fe/Si multilayer SXF spectra, SXF reference spectra taken at the Si L edge for the c Si and a Si samples are shown in Fig. 2. The spectra resemble previously published Si data.2,10 The peaks near 89 FIG. 3. SXF Si L emission spectra for an FeSi and 92 eV in the c Si spectrum originate from nonbonding 2 reference sample and for the two polycrystalline Fe/Si multilayers whose magnetiza- s states and sp-hybridized states, respectively.10,11 These fea- tion curves are shown in Fig. 1. The incident photon energy was tures are broadened by disorder in a Si. 132 eV. The data labeled ``uncoupled ML'' is from the Fe 30 Å/Si Figure 3 shows the Si L edge valence-band emission 20 Å 50 multilayer grown on glass. The data labeled ``AF- spectra of the FeSi2 reference sample and the same two poly- coupled ML'' is from the antiferromagnetically coupled Fe 30 Å/Si crystalline Fe/Si multilayers whose magnetization data are 14 Å 50 multilayer grown on glass. R8826 J. A. CARLISLE et al. 53 the energy gap which is expected in a semiconductor, the slope of the silicide bands near EF suggests that the Fermi level falls in the middle of an energy band. A more detailed interpretation of these spectral features will require elec- tronic structure calculations. III. DISCUSSION AND CONCLUSIONS SXF data have also been taken on an Fe/Si multilayer with tSi 14 Å but which was held at a reduced temperature of 120 K during growth data not shown . The valence-band spectra of the film grown at reduced temperature with tSi 14 Å look virtually identical to data on the film grown at 60 °C but with tSi 20 Å. The most likely explanation for this similarity is that both films have amorphous iron silicide spacer layers. The amorphous state of the spacer layer in these films must be due to the reduced Fe content compared with films which have thinner Si layers or are deposited at higher temperature. Multilayers with amorphous spacer lay- ers do not display antiferromagnetic interlayer coupling.4,7 A comparison of the data of Figs. 3 and 4 show that the peaks in the spectrum of the epitaxial AF-coupled multilayer FIG. 4. SXF Si L emission spectra solid line and Si L edge NEXAFS dashed line for the crystalline Si reference film and for are narrower than those in the spectrum of the polycrystalline the epitaxial Fe 40 Å/Si 14 Å 40 multilayer on MgO. The cross- AF-coupled multilayer. This suggests that a higher degree of ing of the valence-band data obtained from SXF and the local order occurs in epitaxial films. The nature of this order conduction-band data obtained from NEXAFS demonstrates that and the exact structure of the silicide spacer layer phase are the silicide spacer layer is metallic. not yet known. TEM studies have shown that the spacer layer in AF-coupled multilayers is a crystalline cubic iron line uncoupled multilayer with tSi 20 Å is more like c Si. silicide in the B2 CsCl phase or fcc DO3 phase.7 The TEM Peaks in the AF-coupled multilayer spectrum are noticeably diffraction patterns are not consistent with the B20 structure, narrower than those in the FeSi2 reference spectrum. Studies whose SXF data most closely resembles that of the AF- of bulk iron silicides have shown that peaks in the Si emis- coupled multilayers. Jia et al. do report SXF data on the sion spectra narrow as the iron content increases and Si-Si DO3-structure Fe3Si phase but the spectrum of this com- coordination decreases.11 Thus the data of Fig. 3 indicate that pound has a much more prominent and narrow nonbonding the Fe atomic fraction in the spacer layer of the AF-coupled s feature.11 The presence of an Fe3Si spacer can be ruled out multilayers is higher than 1/3. Overall the shape of the spec- on other grounds since this compound is ferromagnetic, in- trum from the AF-coupled multilayer is more reminiscent of consistent with the presence of antiferromagnetic interlayer SXF data on bulk B20 FeSi than of data on bulk FeSi2.11 The coupling. The possibility remains, however, that the spacer uncoupled multilayer data in Fig. 3 have a sharp peak near layer is in the DO3 structure but at a different stoichiometry. 92 eV which coincides with a feature in the c Si spectrum No SXF data on the metastable B2 silicide phase have been although the shape of the higher energy part of the valence reported although photoemission measurements show that it band more closely resembles the FeSi2 data. The narrowness is metallic.14 The magnetic properties of the B2 phase and of the 92 eV feature is evidence for a significant Fe content hypothetical off-stoichiometry phases are not known. The and low Si-Si coordination in the spacer layer of the un- observation of large biquadratic coupling in Fe/Si coupled multilayer. These observations are consistent with multilayers15,16 suggests that an antiferromagnetic or ferri- the TEM determination that the spacer layer in the uncoupled magnetic order may be present in the spacer layer. multilayers is amorphous iron silicide.7 When examined together, the SXF and NEXAFS data The presence of significant Fe in the silicide spacer layer show that Fe/Si multilayers with crystalline metallic silicide of the Fe/Si multilayers strongly suggests that the silicide is spacer layers have antiferromagnetic interlayer coupling, metallic. Unambiguous confirmation of the metallic nature of while similar multilayers with amorphous silicide spacer lay- the silicide is obtained by plotting together the SXF and ers show no interlayer coupling. Whether the amorphous sil- NEXAFS spectra as in Fig. 4. For this data set the spectrom- icide layers are metallic or semiconducting is a topic for eter energy calibration was accomplished through compari- further study. Theoretical calculations will be necessary to son with earlier work on c Si L emission12 and through get a better estimate of the stoichiometry and magnetic prop- alignment of the elastically scattered photon peak to the in- erties of the silicide spacer in the AF-coupled multilayers. cident photon energy. The energy resolution of the SXF spec- The present data should lay to rest any speculation that the tra is about 0.3 eV. The more than 1 eV of overlap between interlayer exchange coupling in Fe/Si multilayers involves a the valence-band features from the SXF and the conduction- novel mechanism. The clarity of these results on thin buried band features from NEXAFS is therefore convincing evi- silicide layers illustrates the power of photon-counting spec- dence that the silicide spacer layer of the multilayer is me- troscopies with their intrinsic bulk sensitivity for the study of tallic. While the Si bands near the Fermi level clearly show multilayer films. 53 SOFT-X-RAY FLUORESCENCE STUDY OF BURIED . . . R8827 ACKNOWLEDGMENTS 9017997, by a Science Alliance Center for Excellence Grant from the University of Tennessee, by the U.S. Department of We would like to thank P.E.A. Turchi, P.A. Sterne, and J. Energy DOE Contract No. DE-AC05-84OR21400 with van Ek for helpful discussions. This work was supported by Oak Ridge National Laboratory, and by the Louisiana Edu- the Division of Materials Science, Office of Basic Energy cational Quality Support Fund and DOE-EPSCOR Grant Sciences, and performed under the auspices of the U.S. De- LEQSF 93-95 -03 at Tulane University. This work was per- partment of Energy by Lawrence Livermore National Labo- formed at the Advanced Light Source, which is also sup- ratory under Contract No. W-7405-ENG-48, by National Sci- ported by the Office of Basic Energy Sciences, U.S. Depart- ence Foundation Grant No. DMR-9017996 and DMR- ment of Energy, under Contract No. DE-AC03-76SF00098. 1 Synthetic Modulated Structures, edited by L.L. Chang and B.C. L.J. Terminello, D.K. Shuh, and R.C.C. Perera, Rev. Sci. In- Giessen Academic, Orlando, 1985 . strum. 66, 1394 1995 . 2 9 D.L. Ederer, T.A. Callcott, and R.C.C. Perera, Synchr. Rad. News J.A. Carlisle, L.J. Terminello, E.A. Hudson, R.C.C. Perera, J.H. 7, 29 1994 . Underwood, T.A. Callcott, J.J. Jia, D.L. Ederer, F.J. Himpsel, 3 and M.G. Samant, Appl. Phys. Lett. 67, 34 1995 . J. Sto¨hr, NEXAFS Spectroscopy Springer-Verlag, New York, 10 R.C.C. Perera, C.H. Zhang, T.A. Callcott, and D.L. Ederer, J. 1992 . Appl. Phys. 66, 3676 1989 ; P.O. Nilsson, J. Kanski, J.V. 4 E.E. Fullerton, J.E. Mattson, S.R. Lee, C.H. Sowers, Y.Y. Huang, Thordson, T.G. Andersson, J. Nordgren, J. Guo, and M. Magnu- G. Felcher, S.D. Bader, and F.T. Parker, J. Appl. Phys. 73, 6335 son, Phys. Rev. B 52, R8643 1995 . 1993 . 11 J.J. Jia, T.A. Callcott, W.L. O'Brien, Q.Y. Dong, D.R. Muller, 5 K.B. Hathaway, in Ultrathin Magnetic Structures, edited by D.L. Ederer, Z. Tan, and J.I. Budnick, Phys. Rev. B 46, 9446 J.A.C. Bland and B. Heinrich Springer-Verlag, New York, 1992 . 12 1994 . See, for instance, J.-E. Rubensson, D. Mueller, R. Shuker, D.L. 6 J.E. Mattson, S. Kumar, E.E. Fullerton, S.R. Lee, C.H. Sowers, Ederer, C.H. Zhang, J. Jia, and T.A. Callcott, Phys. Rev. Lett. M. Grimsditch, S.D. Bader, and F.T. Parker, Phys. Rev. Lett. 71, 64, 1047 1990 . 13 See, for instance, A. Bianconi, R. Del Sole, A. Selloni, P. Chiara- 185 1993 . 7 dia, M. Fanfoni, and I. Davoli, Solid State Commun. 64, 1313 A. Chaiken, R.P. Michel, and M.A. Wall, Phys. Rev. B 53, 5518 1987 . 1996 . 14 H. von Ka¨nel, K.A. Ma¨der, E. Mu¨ller, N. Onda, and H. Sir- 8 IBM/TENN/TULANE/LLNL/LBL beamline, described further in ringhaus, Phys. Rev. B 45, 13 807 1992 . J.J. Jia, T.A. Callcott, J. Yurkas, A.W. Ellis, F.J. Himpsel, M.G. 15 E.E. Fullerton and S.D. Bader, Phys. Rev. B 52, 5112 1996 . Samant, J. Sto¨hr, D.L. Ederer, J.A. Carlisle, E.A. Hudson, 16 R.P. Michel and A. Chaiken unpublished .