PHYSICAL REVIEW B VOLUME 57, NUMBER 16 15 APRIL 1998-II Structural and magnetic phases of ultrathin Fe wedges and films grown on diamond 100... Dongqi Li, D. J. Keavney,* J. Pearson, and S. D. Bader Materials Science Division, Argonne National Laboratory, Argonne, Illinois 60439 J. Pege and W. Keune Laboratorium fu¨r Angewandte Physik, Gerhard-Mercator-Universita¨t, D-47048 Duisburg, Germany Received 20 October 1997 Fe wedges 0­20 Å and films have been grown epitaxially onto diamond C 100 substrates, and their metastable structural and magnetic phases were studied by means of reflection high-energy electron diffraction RHEED , the surface magneto-optic Kerr effect SMOKE , and in situ Mo¨ssbauer spectroscopy. Both the substrate and films 5 ML thick, that were grown at room temperature and subsequently annealed, show 1 1 streaks in the RHEED patterns, indicating a flat, well-ordered fcc structure. Films 5 ML thick exhibit a three-dimensional bcc RHEED pattern. No magnetic signal was observed in either longitudinal or polar SMOKE measurements that were taken between 120 and 300 K for Fe thickness 5 ML. For 5.5-ML Fe, the films are ferromagnetic with in-plane easy axis, and the saturation magnetization at 130 K increases linearly with Fe thickness. In situ Mo¨ssbauer spectra for a 4-ML-thick 57Fe film exhibit a spectral line typical of paramagnetic fcc Fe at 300 and 70 K. This line broadens considerably at 40 and 35 K due to magnetic hyperfine interaction, indicating a low-moment antiferromagnet. These results indicate that 5 ML is the fcc-bcc phase boundary that separates low-moment antiferromagnetic from ferromagnetic ground states, respectively. S0163-1829 98 05216-3 I. INTRODUCTION Mo¨ssbauer spectroscopy on a 4-ML-thick 57Fe film to pro- vide local information about magnetism and structure via the Face-centered-cubic fcc Fe -Fe is predicted to have hyperfine parameters. We observed that Fe grows onto dia- ferromagnetic and antiferromagnetic phases depending on mond 100 as two-dimensional 2D epitaxial, fcc films for atomic volume.1 In nature, the room-temperature RT phase the first 5 ML, and is paramagnetic at RT and low-spin AFM of Fe is bcc and ferromagnetic, while the fcc phase occurs at low temperature. Fe thicknesses 5 ML are rough, bcc, between 1184 and 1664 K. The fcc Fe lattice constant a and ferromagnetic, as expected for the bcc phase. These re- extrapolated down to RT is 3.59 Å. Epitaxy of Fe on differ- sults are in contrast with those of the expanded cases of ent substrates permits fcc-like phase stabilization with ex- epitaxial fcc Fe films.2­6,8­11 panded or contracted in-plane lattice constants. Fe on Cu 100 (a 3.61 Å) is a well-studied example, where the II. EXPERIMENT interplay between structural and magnetic instabilities is clearly manifested with several face-centered structural and The work at Argonne was carried out in a UHV chamber magnetic phases under different growth conditions.2­6 RT- ( 1 10 10 Torr) equipped with SMOKE, low-energy grown films 2­5 ML thick are ferromagnetic, with a face- electron diffraction LEED , RHEED, Auger spectroscopy, centered-tetragonal fct structure, while those 6­11 ML and joule-heated evaporators. Synthetic diamond substrates thick have an enhanced ferromagnetic surface and an antifer- of 4 4 mm2 were used for systematic structural and romagnetic AFM interior, composed of alternating ferro- SMOKE measurements. The diamond C 100 substrates magnetic layers.2 The complex multiplicity of phases were were prepared by etching in HCl:HNO3 solution and boiling recently described theoretically as arising from frustrated ex- in H2O:H2O2:NH4OH solution, followed by ultrasonic clean- change interactions.7 fcc Fe on fcc Co 100 ,8,9 Ni 100 ,8 ing in distilled water. The substrates were then transferred Cu-Au alloys,5,10 and Ni81Fe19 100 Ref. 11 have also been into the UHV chamber with a load-lock system, and cleaned investigated. All these cases represent expansion of the fcc by heating to 650 °C. The Fe films were grown at RT under Fe lattice, and yield similar structural and magnetic phases as a typical pressure of 1 10 9 Torr, and subsequently an- Fe/Cu 100 . Diamond 100 , on the other hand, has a lattice nealed to 400 °C after each 5-Å Fe dosage. Previous work constant of 3.57 Å, and thus provides an in-plane contraction has shown that deposition at RT followed by a 600 °C anneal for metastable fcc Fe. fcc Fe has been grown onto synthetic results in the carbon-stabilized fcc Fe phase as in stainless C 100 ,12,13 though the magnetic characterization is incon- steel.13 Our RT deposition results in no structural ordering clusive. In this work, we investigate wedge-shaped Fe films above 2.5 ML, as observed by RHEED. We anneal at a rela- of 0­20 Å on diamond 100 by means of reflection high- tively low temperature to minimize interdiffusion at the in- energy electron diffraction RHEED , low-energy electron terface, yet ensuring the formation of fcc Fe films. The nomi- diffraction LEED , and the surface magneto-optical Kerr ef- nal Fe thickness was determined with a quartz-crystal fect SMOKE in order to identify the structural and mag- thickness monitor within an accuracy of 20%, assuming a netic phases systematically. In addition, we have employed fcc structure. A typical deposition rate was 0.2 Å/min. 0163-1829/98/57 16 /10044 5 /$15.00 57 10 044 © 1998 The American Physical Society 57 STRUCTURAL AND MAGNETIC PHASES OF ULTRATHIN . . . 10 045 FIG. 1. Typical RHEED pictures taken at electron energy of 9 kV with a CCD camera. a Clean C 100 substrate. b 4.0-ML Fe/C 100 . c 7.8-ML Fe/C 100 . Wedge-shaped samples have a typical slope of 1­2 Å/mm. 100 1.795 Å . The 3D phase has a relaxed lattice spacing; The cleanliness and surface structure of the substrates and the in-plane separation of atom rows perpendicular to the the films were confirmed with Auger, RHEED, and LEED. diffraction plane is 2.04 0.02 Å. This coincides with the Several substrates were used during the experiment, each row distance of 2.03 Å for bcc Fe along the 110 direction. repeatedly etched outside the UHV chamber after Fe depo- Such an orientation is consistent with a transformation to the sition and reused. Magnetic properties were measured in situ bcc structure, with its 111 direction matching the fcc 110 via SMOKE utilizing a focused He-Ne laser beam as occurs in the fcc-to-bcc transformation of Fe/Cu 100 .14 0.2 mm in diameter scanned along an Fe wedge to obtain Assuming this orientational relationship to be the case, the hysteresis loops from the Kerr ellipticity. Both longitudinal atomic row distance should be reported relative to 110 and polar measurements were performed between 120 and rather than perpendicular to the diffraction plane. This small 300 K. correction yields a spacing of 2.01 0.02 Å along the 110 Mo¨ssbauer measurements at Duisburg utilized a 4.0 direction, again in excellent agreement with expectation. 0.8-ML-thick film of 95.5% isotopically enriched 57Fe Therefore, films thicker than 5 ML are essentially of bcc grown at RT in a UHV chamber base pressure structure. It is noted that the fcc structure is retained for a 8 10 11 Torr at a rate of 0.48 Å/min under a back- thicker film when the source and/or the substrate are dirty. ground pressure of 5 10 10 Tor. The substrate was a natu- This is, however, normally accompanied by a c(2 2) re- ral diamond C 100 of 6 6-mm2 area. Chemical etching construction. This is attributed to surfactant-mediated growth was as described above. After transferring the substrate into as occurs in Fe/Cu 100 , where an fcc layer-by-layer growth the UHV chamber, it was cleaned by annealing at 900 °C for can persist up to 40 ML or more in comparison to 10­11 ML 10 min. The nominal Fe thickness was measured by a quartz- for clean Fe/Cu 100 .15,16 crystal thickness monitor calibrated previously by RHEED- In addition to the spacing, the width of the streaks and/or intensity oscillations from fcc-Fe/Cu 100 . The cleanliness spots also changes drastically between Figs. 1 a and 1 b and surface structure of the films were determined by Auger and Fig. 1 c . The broadening of the streak width indicates a and RHEED. The 57Fe conversion-electron Mo¨ssbauer spec- decrease in the average terrace size, i.e., a rougher surface. tra CEMS were obtained in situ utilizing a channeltron de- This accompanies the transformation into the 3D bcc phase. tector. The radiation from a 100-m Ci 57Co Rh Mo¨ss- In general, the width of a streak w includes an instrumental bauer source was incident normal to the film plane. All broadening, but, ignoring this, we can estimate the lower isomer-shift values are given relative to bulk -Fe at RT. limit of the terrace sizes as 2 /w 450 Å for the 4-ML fcc phase Fig. 1 b , and 150 Å for the 7.8-ML bcc phase III. RESULTS AND DISCUSSIONS Fig. 1 c . Plotted in Fig. 2 are the row spacing in real space and the Figure 1 shows typical RHEED patterns of a synthetic lower limit of terrace size 2 /w from the RHEED patterns C 100 substrate, and Fe films of 4.0 and 7.8 ML. Both the along a wedge as a function of the nominal Fe thickness, substrate and films 5 ML thick show sharp 1 1 patterns assuming an fcc structure and 1 ML 1.8 Å. There is an with similar spacings, indicating a well-ordered fcc structure. abrupt change in both the row spacing and the terrace size at The substrate and film surfaces are reasonably flat, as indi- 5 ML. This indicates a structural phase transition from a cated by the sharp streaks in the RHEED pattern. Films fcc 100 film to a bcc 110 texturing. This structural phase 5 ML thick exhibit a 3D spotty RHEED pattern, as shown transition may possess quantitative differences on different in Fig. 1 c . Some ring-shaped features are also evident, sug- substrate crystals or on the same crystal after each etching. gesting the coexistence of a polycrystalline phase. RT depo- For example, for some wedges, the transition is more gradual sition only results in the fading of the 1 1 diamond 100 than the one indicated in Fig. 2. Nevertheless, the transition RHEED pattern, which disappears above 2.5 ML. This sug- always occurs at 5 ML. These minor differences may be gests that, at RT, Fe cannot form an ordered fcc film. The caused by different defect densities on the substrates. lattice spacings of the annealed films can be measured from Figure 3 shows typical LEED patterns for 3.2- and 7.8- the spacing on the RHEED pattern compared with that of the ML-thick Fe along the same wedge as in Figs. 1 and 2. Both C 100 reference. The Fe films first grow pseudomorphically show the same fcc 1 1 pattern as that of the diamond sub- with the in-plane lattice spacing the same as that of diamond strates, though the one for 7.8 ML has a slightly higher back- 10 046 LI, KEAVNEY, PEARSON, BADER, PEGE, AND KEUNE 57 FIG. 2. The row spacing and the lower limit of average terrace size 2 /W in real space from the RHEED patterns along an Fe wedge on diamond 100 , where W is the width of the streaks in the RHEED patterns. A structural phase transition is apparent at 5 ML. ground. No bcc phase can be identified. This apparent con- flict with the RHEED results may be due to the exposed fcc C 100 substrate or to residual fcc Fe film after the Fe balls up to form tiny 3D bcc grains. This is consistent with the fact that the bcc phase is very rough, based on our RHEED re- sults. RHEED, with a glancing incidence, is most sensitive to the top surface of a film, while LEED, with a normal- incidence electron beam, is sensitive not only to the film, but also to any exposed substrate. Our Auger measurements re- FIG. 4. Longitudinal Kerr hysteresis loops of Fe/C 100 taken at veal a C signal even for the thickest films, which may come 130 K. No magnetic signal, polar or longitudinal, was detected for from exposed substrate or C interdiffusion. films 5 ML. Figure 4 shows typical SMOKE longitudinal hysteresis loops along an Fe wedge. No magnetic signal was observed are not ferromagnetic above 120 K. Only the ordinary bcc in either longitudinal or polar measurements for Fe thickness films are ferromagnetic as expected. 5 ML. For 5.5-ML Fe, the films are ferromagnetic, with Typical CEM spectra obtained at different temperatures in-plane easy axis. Both the saturation magnetization (Ms) at from the annealed 4.0-ML-thick 57Fe film on natural C 100 130 K and the Curie temperature (TC) of the film increase are shown in Fig. 6. The RHEED pattern of this film not linearly with Fe thickness, as seen in Fig. 5. Linear fits of the shown is similar to that shown in Fig. 1 b , except that the Ms and TC indicate that they extrapolate to zero at around 5 RHEED streaks are considerably broadened. This indicates ML, instead of 0 ML, which further quantifies the onset of that the 4-ML 57Fe film is fcc, but is much rougher than the ferromagnetic ordering. TC increases from 120 to 300 K one in Fig. 1 b . within 1 ML of the onset of ferromagnetism along the Evidence for the fcc structure and paramagnetism of the wedge. This onset at 5 ML coincides with the onset of the 4-ML 57Fe film are obtained from the CEM spectrum at RT structural phase transition from fcc to bcc. These results in- Fig. 6, top . No lines of ferromagnetic -Fe are observed. dicate that the metastable, epitaxial fcc Fe films ( 5 ML) Data analysis by a least-squares computer fit17 with Lorent- FIG. 3. Typical LEED patterns of Fe/C 100 taken at an electron energy of 145 eV. a 3.2 ML. b 7.8 ML. 57 STRUCTURAL AND MAGNETIC PHASES OF ULTRATHIN . . . 10 047 zian lines yield a dominant rather narrow central line of 80% in relative spectral area and a less-intense symmetric quadrupole-split doublet satellite spectrum, of 20% in rela- tive spectral area . The strong central line exhibits an isomer shift of 0.101 0.009 mm/s at 300 K. This negative value is very close to that of paramagnetic fcc -Fe precipi- tates in a Cu matrix at 300 K ( 0.088 0.003 mm/s),18­21 and is close to isomer-shift values of paramag- netic fcc Fe/Cu 100 multilayers22,23 and films24,4 at RT ( 0.08 0.01 mm/s). Therefore, we assign the central single line in Fig. 6 to paramagnetic fcc Fe on C 100 . The measured full width at half maximum of this single line is 0.40 0.01 mm/s at RT, and remains unchanged at 70 K ( 0.41 0.02 mm/s). This indicates that fcc Fe on C 100 is still paramagnetic at 70 K. FIG. 5. The saturation magnetization Ms at T 130 K, and the The dominant fcc Fe line is found to broaden consider- Curie temperature TC along an Fe wedge on C 100 . ably at low temperatures Fig. 6 . For example, of 0.76 0.03 and 0.81 0.05 mm/s is obtained at 40 and 35 K, respectively. This observation is analogous to the case of AFM fcc -Fe precipitates in Cu below their Ne´el tempera- ture of 67 K, where a reduction in temperature leads to in- creasing line broadening due to an increasing degree of AFM ordering.18­21 The AFM state is indicated by a line broaden- ing only, since the magnetic hyperfine-field saturation value 2.4 T for large precipitates is of the order of the natural linewidth, and thus the full 57Fe six-line Zeeman pattern of fcc Fe cannot be resolved. This is a consequence of the low Fe atomic magnetic moment ( 0.7 B) of these AFM -Fe precipitates.25,26 Similar drastic line broadening at low temperatures has been reported for fcc-Fe/Cu 100 multilayers22,23 0.8 mm/s at 4.2 K, equivalent to a hyperfine field Bhf 1.6 T and fcc Fe/Cu 100 films24,4 0.5­ 0.6 mm/s at 29­35 K, equivalent to Bhf 1.1­1.3 T . Moreover, AFM ordering at 4.2 K in fcc Fe/Cu 100 multilayers has been proven by Mo¨ssbauer spectroscopy in an external magnetic field.22 The average magnetic hyperfine field estimated from the linewidth of our film is about 1.3­1.5 T at 40­35 K, thus being in agreement with typical Bhf values for the AFM fcc Fe/Cu 100 system and for AFM -Fe precipitates in Cu. Therefore, we conclude that annealed 4-ML-thick fcc Fe on C 100 is in a low-moment AFM state at low temperature. This identification clearly rules out the possibility of having superparamagnetism below 5 ML. It is interesting to note that the width of the paramagnetic fcc Fe/C 100 line at RT ( 0.40 0.01 mm/s) is slightly larger than the corresponding value of 0.30 0.01 mm/s in fcc Fe/Cu 100 ,22,23,24,4 and is remarkably larger than the linewidth observed with our spectrometer on a standard -Fe calibration foil 0.24­0.26 mm/s . The excess linewidth in the case of fcc Fe/C 100 might be caused by interstitial C impurities. It is well known that in C-containing paramag- netic bulk fcc steel austenite , the electron distribution at nearest-neighbor Fe atoms surrounding an interstitial C im- purity is perturbed, giving rise to a quadrupole splitting EQ e2qQ/2 of about 0.625­0.643 mm/s, and a of about 0.06­ 0.002 mm/s at RT.27­29 Fe atoms more distant FIG. 6. Mo¨ssbauer spectra of annealed 4.0-ML Fe/C 100 mea- from a C impurity show the characteristic fcc-Fe single line sured at RT, 70, 40, and 35 K from top to bottom, respectively . of paramagnetic austenite with a of 0.05­ 0.1 mm/s at The linewidths of the fitted central line are 0.40 0.01, 0.41 RT.27,28 The intensity area ratio of the single line to the 0.02, 0.76 0.03, and 0.81 0.05 mm/s, respectively. quadrupole doublet depends on the C concentration in the 10 048 LI, KEAVNEY, PEARSON, BADER, PEGE, AND KEUNE 57 austenite. This poses a question about the origin of the annealing,2 are highly metastable and can be transformed to quadrupole-split doublet satellite spectrum included in the the more stable AFM low-moment fcc Fe phase by annealing least-squares fitting of the data in Fig. 6. While the EQ at 570 K ( 300 °C) and cooling to RT.4 We postulate that value of 0.63 mm/s of this doublet agrees well with that of the AFM fcc Fe films on C 100 might behave in a similar Fe atoms with one nearest-neighbor C atom in austenite, the fashion after annealing. isomer shift of this doublet in Fig. 6 is 0.28 0.05 mm/s at RT, which is far from the corresponding value in austenite. IV. CONCLUSIONS Therefore, one has to look for another explanation of the doublet in Fig. 6. As a possibility, it could originate from the We have studied the structural and magnetic phases of paramagnetic fcc Fe/diamond interface. A quadrupole dou- annealed Fe films on diamond 100 . Films of 5 ML ex- blet with similar splitting ( EQ 0.60 0.08 mm/s), though hibit 2D epitaxial growth and form a fcc phase. From with a different isomer shift of 0.015 0.005 mm/s, has SMOKE, such a phase is found not to be ferromagnetic been observed at 300 K for probe-layer 57Fe atoms artifi- above 120 K, the lowest temperature explored. Films thicker cially located at the paramagnetic fcc Fe/Cu 100 interface.4 than 5 ML transform into a 3D bcc phase, which is the nor- Similar work on the fcc Fe/C 100 interface is in progress to mal ferromagnetic phase. A Mo¨ssbauer spectral line typical clarify the origin of the doublet. Due to the much larger for paramagnetic fcc 57Fe was observed for 4-ML Fe at and intensity of the fcc Fe single line in Fig. 6, our results for fcc above 70 K, while at and below 40 K the film is in a low- Fe are essentially independent of the detailed shape of the moment antiferromagnetic state. weak satellite doublet. The formation of a low-spin AFM phase for thin fcc Fe is ACKNOWLEDGMENTS in contrast to the Fe/Cu 100 case. For Fe/Cu 100 , a high- spin ferromagnetic phase exists for both RT- and low- We thank Dr. D. Pappas for valuable information and temperature-grown films, at least when the Fe thickness is discussions, and Dr. S. Jiang and U. von Ho¨rsten for techni- 5 ML. This high-spin Fe/Cu phase has a metastable fct cal assistance. One of us W.K. is grateful to the Volk- structure with reconstructions.30 There are several possible swagen Stiftung for supporting his stay at Argonne. The causes for this difference, such as lattice contraction instead work at Argonne was supported by the U.S. Department of of expansion, carbon interdiffusion, etc. It is interesting to Energy, BES-MS, under Contract No. W-31-109-ENG-38, mention that low-temperature-grown fcc Fe/Cu 100 films, and at Duisburg by the Deutsche Forschungsgemeinschaft which are in a ferromagnetic high-moment state before SFB 166 . *Present address: Department of Physics, University of Arizona, 14 K. Kalki, D. D. Chambliss, K. E. Johnson, R. J. Wilson, and S. Tucson, AZ 85721. Chiang, Phys. Rev. B 48, 18 344 1993 . 1 C. S. Wang, B. M. Klein, and H. Krakauer, Phys. Rev. Lett. 54, 15 Dongqi Li, M. Freitag, J. Pearson, Z. Q. Qiu, and S. D. Bader, J. 1852 1985 ; V. L. Moruzzi, P. M. Marcus, K. Schwarz, and P. Appl. Phys. 76, 6425 1994 . Mohn, Phys. Rev. B 34, 1784 1986 ; F. J. Pinski, J. Staunton, 16 A. Kirilyuk, J. Giegiel, J. Shen, M. Straub, and J. Kirschner, D. D. Johnson, and G. M. Stocks, Phys. Rev. Lett. 56, 2096 Phys. Rev. B 54, 1050 1996 . 1986 . 17 For least-squares fitting of the data, the computer program by R. 2 J. Thomassen, F. May, B. Feldmann, M. Wuttig, and H. Ibach, A. Brand, provided by WISSEL GmbH., D-82319 Starnberg, Phys. Rev. Lett. 69, 3831 1992 . Germany, has been used. 3 Dongqi Li, M. Freitag, J. Pearson, Z. Q. Qiu, and S. D. Bader, 18 U. Gonser, C. J. Meechan, A. H. Muir, and H. Wiedersich, J. Phys. Rev. Lett. 72, 3112 1994 , and references therein. Appl. Phys. 34, 2373 1973 . 4 W. Keune, A. Schatz, R. D. Ellerbrock, A. Fuest, Katrin Wilmers, 19 S. J. Campbell and P. E. Clark, J. Phys. F 4, 1073 1974 . and R. A. Brand, J. Appl. Phys. 79, 4265 1996 . 20 D. L. Williamson, W. Keune, and U. Gonser, Proceedings of the 5 D. J. Keavney, D. F. Storm, J. W. Freeland, I. L. Grigorov, and J. International Conference on Magnetism Nauka, Moscow, C. Walker, Phys. Rev. Lett. 74, 4531 1995 . 1974 , Vol. 1, p. 246. 6 M. Zharnikov, A. Dittschar, W. Kuch, C. M. Schneider, and J. 21 T. Ezawa, W. A. A. Macedo, U. Glos, W. Keune, K. P. Schletz, Kirschner, Phys. Rev. Lett. 76, 4620 1996 . and U. Kirschbaum, Physica B 161, 281 1989 . 7 D. Spisak and J. Hafner, Phys. Rev. B 56, 2646 1997 . 22 W. Keune, R. Halbauer, U. Gonser, J. Lauer, and D. L. William- 8 W. L. O'Brien and B. P. Tonner, Phys. Rev. B 52, 15 332 1995 . son, J. Appl. Phys. 48, 2976 1977 . 9 Ernesto J. Escorcia-Aparicio, R. K. Kawakami, and Z. Q. Qiu, 23 R. Halbauer and U. Gonser, J. Magn. Magn. Mater. 35, 55 1983 . Phys. Rev. B 54, 4155 1996 ; Ernesto J. Escorcia-Aparicio, H. 24 W. A. A. Macedo and W. Keune, Phys. Rev. Lett. 61, 475 1988 . J. Choi, R. K. Kawakami, and Z. Q. Qiu, J. Appl. Phys. 81, 4714 25 S. C. Abrahams, L. Guttmann, and J. S. Kasper, Phys. Rev. 127, 1997 . 2052 1962 . 10 D. Tillmann and E. Kisker, Solid State Commun. 100, 415 26 G. J. Johanson, M. B. McGirr, and D. A. Wheeler, Phys. Rev. B 1996 . 1, 3208 1970 . 11 W. Kuch and S. S. P. Parkin, Europhys. Lett. 37, 465 1997 . 27 P. M. Gielen and R. Kaplow, Acta Metall. 15, 49 1967 . 12 D. P. Pappas, J. W. Glesener, V. G. Harris, Y. U. Idzerda, J. J. 28 J. M. Genin and P. A. Flinn, Phys. Lett. 22, 392 1966 . Krebs, and G. A. Prinz, Appl. Phys. Lett. 64, 28 1994 . 29 S. J. Lewis and P. A. Flinn, Phys. Status Solidi 26, K51 1968 . 13 R. S. Swineford, D. P. Pappas, and V. G. Harris, Phys. Rev. B 52, 30 S. Mu¨ller, P. Bayer, C. Reischl, K. Heinz, B. Feldmann, H. 7890 1995 . Zillgen, and M. Wuttig, Phys. Rev. Lett. 74, 765 1995 .