Ultraviolet and visible Raman studies of nitrogenated tetrahedral amorphous carbon lms

Transcription

1 Thin Solid Films 366 (2000) 169±174 Ultraviolet and visible Raman studies of nitrogenated tetrahedral amorphous carbon lms J.R. Shi*, X. Shi, Z. Sun, E. Liu, B.K. Tay, S.P. Lau School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore , Singapore Received 3 July 1999; received in revised form 13 January 2000; accepted 13 January 2000 Abstract Nitrogenated, tetrahedral amorphous carbon (ta-c) lms prepared by the ltered cathodic vacuum arc (FCVA) technique have been studied using ultraviolet (UV, 244 nm) and visible (514 nm) micro-raman scattering. The nitrogen ions were produced by a RF ion-beam source with a nitrogen ow-rate varying from 0 to 10.0 sccm, which results in a nitrogen content from 0 to 10.8 at.% in the deposited lms. In the visible Raman spectra, only vibrational modes of sp 2 -bonded carbon (G and D peaks) are observed, while a new wide peak, called the T peak, located at 1090±1320 cm 21, associated with the vibrational mode of sp 3 -bonded carbon, appears in the UV-Raman spectra. In the visible Raman spectra, the G-peak width (100±113 cm 21 ) and the intensity ratio I D /I G (0.34±0.94) are both sensitive to the structural changes induced by N incorporation. In the UV-Raman spectra, the G-peak position almost linearly decreases from 1665 to 1610 cm 21, and the T- peak position increases tremendously from 1095 to 1314 cm 21 with increasing N content. The G-peak position and width, and the T-peak position, are all sensitive to the bonding structure of the lms. q 2000 Elsevier Science S.A. All rights reserved. Keywords: Raman scattering; Tetrahedral amorphous carbon 1. Introduction * Corresponding author. Tel.: ; fax: address: ejrshi@ntu.edu.sg (J.R. Shi) Tetrahedral amorphous carbon (ta-c) lms, prepared by the ltered cathodic vacuum arc (FCVA) method [1], have stimulated great interest from both scienti c and industrial perspectives in the last decade. ta-c lms have interesting and useful properties [2±6], such as extreme hardness (,70 GPa), chemical inertness, a wide Tauc optical band-gap (,2.5 ev), smooth surface and low friction, thermal stability, transparency in wide spectral range and ultra-thin achievable thickness (,3 nm). Therefore, this material is important for coating technology and electronic-device applications. It has been shown by electron energy-loss spectroscopy (EELS) that a signi cant fraction (up to 87%) of the carbon atoms in the ta-c lms form an amorphous, tetrahedral (sp 3 ) structure [5±7]. The electronic and optical properties of ta-c lms were reported to be continuously adjustable by the incorporation of nitrogen during deposition [8±10]. Nitrogen ions were ef ciently combined into the amorphous carbon network by nitrogen-ion bombardment (ion-assisted FCVA) [10,11]. Nitrogenated ta-c lms show large optical absorption coef cients, high photo-response, and are a potential solar cell material [10,12]. Raman scattering is a non-destructive technique for measuring the bonding properties of diamond-like carbon (DLC) and diamond. The conventional visible Raman spectrum of DLC, excited by 488- or 514-nm photons, is dominated by the G band at about 1560 cm 21 and a D feature around 1360 cm 21, both of which are attributed to sp 2 - bonded carbon. Recently, direct ultraviolet (UV)-Raman observations of the vibrational mode of the sp 3 -bonded carbon in the ta-c lm have been achieved using 244-nm excitation [13±15]. In the visible Raman measurement of the ta-c lm, the excitation energy (2.4 ev for 514 nm) corresponds to the p±p* transition at sp 2 sites, and this leads to a resonant enhancement of the Raman cross-section [16,17]. For UV excitation at 244 nm (5.1 ev), the energy is suf cient to excite the s states of both sp 2 and sp 3 sites. This allows the Raman spectrum to show a more equally-weighted view of the vibrational density of states for sp 2 and sp 3 sites. In this paper, UV and visible Raman scattering studies of nitrogenated ta-c lms prepared by the FCVA method are presented. The correlation between the spectral parameters and the nitrogen content is discussed /00/$ - see front matter q 2000 Elsevier Science S.A. All rights reserved. PII: S (00)00732-X

2 170 J.R. Shi et al. / Thin Solid Films 366 (2000) 169± Experimental details The nitrogenated ta-c lms were deposited by a FCVA system which is described in detail elsewhere [6]. The carbon plasma is produced from the arc spot on a cathode of 60-mm diameter, % pure graphite in high vacuum. In our system, a radial electric eld is introduced via the torus duct-wall bias, and this, coupled with the curvilinear axialmagnetic eld on a curved toroidal duct, forms the crossed electric-magnetic- eld ltering assembly. The plasma, steered by the eld through the duct to the deposition chamber, is deposited on the substrate without the unwanted macroparticles and neutral atoms. The substrate in the deposition chamber was negatively biased at 80 V, which corresponds to 100 ev of impinging carbon-ion energy in our experiment. The arc current was set to 60 A, and the toroidal magnetic eld was xed at 40 mt. The substrates used were cleaned k100l p-type silicon wafers, with an average thickness of 0.5 mm. The oxide layer on the silicon surface was removed by argon ions from a RF ion-beam source before deposition. Nitrogen gas was introduced into the deposition chamber through the RF ion-beam source, with an ion energy of 100 ev and ion-current density of 2.83 ma/cm 2 during deposition. The nitrogen partial pressure was varied between and Torr, depending on the nitrogen ow-rate monitored using a mass- ow controller; the nitrogen ow-rate is directly related to the nitrogen partial pressure. A set of nitrogenated ta-c lms with thicknesses of about 80 nm were prepared by the FCVA technique at room temperature. One pure ta-c lm was prepared for comparison. The nitrogen contents in the deposited lms were measured by Auger electron spectroscopy (AES) and Rutherford backscattering (RBS). The UV and visible Raman spectra were excited using the 244-nm line of a frequency-doubled Ar 1 laser (Coherent 90C FreD series) and the nm line of an Ar 1 laser, respectively. The UV and visible scattered light was collected in back scattering with UV-enhanced and normal CCD cameras, using Renishaw micro-raman System 2000 and 1000 spectrometers. A laser output of 20 mw was used, which resulted in an incident power at the sample of approximately 1.5 mw. Multi-layer dielectric lters working in the range of UV and visible light were used for the rejection of Rayleigh scattering light. The sample was rotated during the UV-Raman measurement to prevent the lm from structural damage caused by the high photon energy of the UV-laser. The spectral resolutions of the UV and visible spectrometers (half-width, half-maximum) were 4.0 and 2.0 cm 21, respectively. Fig. 1 shows the relationship between the nitrogen atomic fractions in the lm and nitrogen ow-rate during the deposition. The nitrogen content increases monotonically as the nitrogen ow-rate increases. It increases faster at a low ow-rate and becomes saturated at a high ow-rate. The lms with nitrogen contents of 4.2, 7.9, 9.1, 9.6 and 10.8 at.% were used for Raman measurements. The visible Raman spectra of pure and nitrogenated ta-c lms are shown in Fig. 2 in the range of 800±2000 cm 21.A wide peak around 950 cm 21 is the second-order Raman peak of the silicon substrate. The appearance of this peak is a measure of the transparency of the lms near the wavelength of nm. As the nitrogen ow-rate increases, the intensity of this peak decreases gradually. This result agrees with the previous result that the absorption coef cient of the lm increases with nitrogen content [10]. The main asymmetric peak between 1100 and 1800 cm 21 is attributed to the vibrational mode of sp 2 -bonded carbon clusters. The broad peak becomes more asymmetric with increasing nitrogen ow-rate. The asymmetric broad peak could be tted either with two Gaussian peaks [17,18], or with a Breit±Wigner±Fano (BWF)-shaped peak [6,19]. In this paper, it was tted with two Gaussian±Lorentzian mixed peaks, graphite (`G') and disorder (`D') peaks. The Gaussian ratio was automatically adjusted by Grams/32 software. The tted G and D peaks for the lm containing 10.8 at.% nitrogen are shown in Fig. 2 with dashed lines. No peak associated with nitrogen was observed in the visible Raman spectra for all lms. The tting results of peak positions and line-widths are shown as a function of the nitrogen content in Fig. 3. As the nitrogen content increases from 0 to 10.8 at.%, the G- peak position varies in a small range from 1568 to 1572 cm 21, while the D-peak position varies in a range from 1377 to 1412 cm 21. The changes of these peak positions are small, which are almost within the measuring errors. With increasing nitrogen content, the line-width (half-width at half-maxi- 3. Results and discussion The nitrogen content in the nitrogenated ta-c lms, determined by AES and RBS methods, ranges from 1.7 to 11.2 at.% as the nitrogen ow-rate increases from 0.5 to 12 sccm. Fig. 1. The relationship between the nitrogen content in the lms and the nitrogen ow-rate during lm deposition.

3 J.R. Shi et al. / Thin Solid Films 366 (2000) 169± This situation corresponds to a big G peak and a very small D peak, or a small value of I D /I G ratio. So the increase in the I D /I G ratio indicates that the fraction of sp 2 -bonded carbon atoms increases with increasing nitrogen content. Based on Robertson's `cluster model' for an amorphous carbon network [20], the C amorphous network can be treated as a few sp 2 clusters embedded in a sp 3 -bonded matrix. The sp 3 matrix controls the mechanical properties, and p states of the sp 2 clusters control the lm's electronic structure and optical band-gap. The decreases in the hardness, the stress and the optical band-gap [10] for the nitrogenated lms with a high nitrogen ow-rate are all correlated with the increase in the sp 2 -bonded carbon fraction. Since the G-line-width is partly determined by the domain or size of the amorphous carbon clusters, the I D /I G ratio is plotted versus G-line-width in Fig. 4b. The I D /I G ratio monotonically decreases from the largest value of 0.94 to the smallest value of 0.34 as the G-peak width increases from to 113 cm 21. The low I D /I G ratio and broad G-line-width for the ta-c lm indicates small sp 2 clusters in the lm [18]. Recent transmission electron microscope results by Davis et al. [21] on a-c:h superlattice structures shows there to be no evidence of clusters greater than 5 AÊ in r.f. plasma-deposited a-c:h lms. As the I D /I G ratio increases Fig. 2. Visible (514.5 nm) Raman spectra of pure and nitrogenated ta-c lms. The dashed lines represent the tting of D and G peaks for the lm with 10.8 at.% nitrogen. mum) of the D peak varies from 149 to 173 cm 21. No obvious trend can be obtained for the D-peak width. Meanwhile, the line-width of the G peak monotonically decreases from 113, for the pure ta-c lm, to 106 cm 21 for the lm with 10.8 at.% nitrogen. The lm containing 9.1 at.% nitrogen has a minimum width of cm 21. It was found that there was a correlation between the G-peak width and the lm stress. For amorphous carbon lms, the G-peak width almost linearly increases with the stress [18]. For nitrogenated ta-c lms, the stress has a small maximum for the lm deposited at the 0.5 sccm nitrogen ow-rate, and then decreases with nitrogen ow-rate [10]. Therefore, the G-line-width decreases as the nitrogen content in the lms increases. The variation of the G-line-width indicates that the G-linewidth is a good parameter for nitrogenated ta-c lms. The intensity ( tting-area of peaks) ratio of the D peak to the G peak is generally a measure for the zone edges or border phonons of the carbon clusters (which depend on cluster sizes and distributions). Fig. 4a shows the dependence of the intensity ratio I D /I G on the nitrogen content. The I D /I G ratio almost linearly increases as the nitrogen content increases. The lm containing 9.1 at.% nitrogen has the maximum I D /I G ratio of For ta-c lm, it was found that the main asymmetric peak between 1100 and 1800 cm 21 becomes quite symmetric and has a large negative value of skewness (Q-factor) as the fraction of sp 3 -bonded carbon becomes over 80% [6,19]. Fig. 3. (a) Peak position; and (b), width of the G and D peaks of the sp 2 - bonded clusters at nm as a function of nitrogen content in the lm. Solid and open circles represent G-peak position and width, and solid and open triangles represent D-peak position and width, respectively.

4 172 J.R. Shi et al. / Thin Solid Films 366 (2000) 169±174 Fig. 4. The dependence of the I D /I G ratio on: (a), the nitrogen content in the lm; and (b), the G-peak width. and the G-line-width decreases, the size of the sp 2 clusters in the nitrogenated ta-c lms becomes larger. The UV-Raman spectra for the pure and nitrogenated ta- C lms are shown in Fig. 5 in the range of 750±2000 cm 21. The UV-Raman spectrum consists of two wide peaks, referred as T and G bands [13,14], and a small, sharp atmospheric O 2 peak at 1553 cm 21. The ttings with three mixed Gaussian±Lorentzian lines are plotted for each corresponding spectrum. The decomposed T, G and O 2 peaks for the lm with 10.8 at.% nitrogen are also shown in Fig. 5. The T and G peaks are related with the vibrational modes of the sp 3 - and sp 2 -bonded carbon clusters, respectively [13± 15,22,23]. It can be seen clearly from Fig. 5 that the height of the T peak decreases as the nitrogen content in the lm increases, and the position of the T peak moves to a higher wave number. No peak associated with nitrogen atoms was observed in the UV-Raman spectra. The reason for the fact that both the visible and UV-Raman spectra are insensitive to the vibrational mode of the nitrogen atoms may be due to the resonance effect of the carbon-vibrational modes and that the Raman cross-section of the vibrational mode associated with nitrogen atoms is comparatively small. Fig. 6 shows the dependence of the peak positions and the line-widths of the G and T bands on the nitrogen content. As the nitrogen content increases, the G-peak position almost Fig. 5. UV (244 nm)-raman spectra of pure and nitrogenated ta-c lms. The solid lines represent the ttings for each corresponding spectrum. The dashed lines represent the tting of T and G peaks, and the dotted line represents the tting of the peak of oxygen in the air for the lm with 10.8 at.% nitrogen. linearly decreases from 1665, for the pure ta-c lm, to 1610 cm 21 for the lm containing 10.8 at.% nitrogen. The T-peak position exhibits the opposite behaviour, increasing from 1095, for the pure ta-c lm, to 1314 cm 21 for the lm with 10.8 at.% nitrogen. The overall position changes of the G and T peaks are 55 and 219 cm 21, respectively. The results for ta-c lm are in good agreement with the data reported by Gilkes et al. [13,14] and Merkulov et al. [15]. With increasing nitrogen content in the lm, the G-linewidth decreases from 118, for the pure ta-c lm, to 105 cm 21 for the lm with 10.8 at.% nitrogen. This behaviour is similar to that of the G-line-width determined by the visible Raman spectra. The T-peak width (275±350 cm 21 ) is much larger than the G-peak width, but no distinct trend can be observed. The large widths of the G and T peaks result from the small carbon-cluster size in the lms. The G-peak position determined by the UV-Raman spectra is much higher than that determined by the visible Raman spectra, and decreases obviously with increasing nitrogen content in the lm. It is worth discussing the reasons in more detail. Because of the resonance feature of the G peak, or p±p* transition, at sp 2 sites, small changes in the strength of the p bonds can considerably affect the G- peak position. This can occur either as a result of stress or

5 J.R. Shi et al. / Thin Solid Films 366 (2000) 169± Fig. 6. (a) Peak position; and (b), widths of the G and T peaks at 244 nm as a function of nitrogen content in the lm. Solid and open circles represent G-peak position and width, and solid and open triangles represent T-peak position and width, respectively. through changes in the clustering of sp 2 sites. Ager et al. [24] found a strong dependence of the G position on stress in ta-c lms deposited by FCVA method. They showed that by delaminating the highly-stressed ta-c lms (allowing the lms to relax), the G position drops by about 20 cm 21.An average stress shift of 1.9 cm 21 /GPa was obtained for biaxial-plane stress in hard ta-c lms. For the nitrogenated ta-c lms, the difference in the compressive stress between the pure ta-c lm and the lm containing 10.8 at.% nitrogen is about 5.6 GPa. The stress-induced shift (,10.6 cm 21 ) could represent one part of the total shift of the G position. As the G-peak position, determined by the visible Raman spectra, stays almost unchanged as the nitrogen content in the lm increases (the internal stress decreases), the stress may not be a reasonable mechanism for the large change in the G- peak position measured by UV-Raman. Therefore, other mechanisms are expected. One possible mechanism is that there is a sp 2 -cluster size distribution in the lms, and the average sp 2 -cluster size increases with increasing nitrogen content in the lm. Raman scattering from sp 2 -carbon clusters with various sizes could be selectively resonanceenhanced, and the varying average cluster size of sp 2 carbon is responsible for the shift of the G-peak position. Resonant Raman spectra of hydrogenated amorphous carbon (a-c:h) lms have been reported by Wagner et al. [16] and Yoshikawa et al. [25,26]. The G-peak position of sp 2 -carbon clusters decreases with increasing wavelength of the excitation laser. This behaviour was well interpreted in terms of p±p* resonant Raman scattering from sp 2 -carbon clusters with different sizes [25]. Size-dependent resonance Raman scattering was observed recently in single-wall carbon nanotubes [27,28]. Another possible mechanism is that the sp 2 bonds may not be linear, but are bent because of the high internal compressive stress. The morphologies of nongraphitizing carbon materials reported by Harris showed sp 2 -carbon clusters made of bent sp 2 bonds [29]. The stretching mode of the sp 2 bonds becomes higher because of the bending of the bonds. The T-peak position decreases tremendously as the nitrogen content in the lm increases. The variation of the T-peak position is strongly correlated with the nitrogen content, hence the content of sp 3 bonding in the lm. Merkulov et al. [15] reported that the Raman peak associated with the sp 3 bonding varies from 1150, for the lm with 75 at.% sp 3 -C atom, to 1400 cm 21 for the sputtered a-c with 6 at.% sp 3 -C atom. The data are in good agreement with ours. The nearest neighbors of a given sp 3 -carbon atom may be either sp 3 -, or sp 2 -bonded carbon. For the ta-c lms with an sp 3 fraction of more than 80%, the nearest neighbors of the sp 3 site contain either zero or one sp 2 -bonded carbon atom, therefore the bonds between sp 3 -carbon atoms (sp 3 Zsp 3 bonds) are predominant. As the sp 3 fraction decreases, the content of the sp 2 -bonded carbon increases. The bonds between sp 3 - and sp 2 -carbon atoms (sp 3 Zsp 2 bonds) gradually become the predominant bonds. This results in the upward movement of the T peak because of the high vibrational frequency of the sp 2 -bonded atoms. When the sp 2 -bonded carbon atoms are dominant, the sp 3 Zsp 2 bonds are embedded in the sp 2 Zsp 2 bonds, and the latter are the predominant bonds. The T-peak position is coming close to the D-peak position measured by visible Raman. As the clusters in amorphous carbon lms are quite small (around 5 AÊ ) [21], the vibrational modes of the zone edges have a signi cant contribution in the Raman spectra. It is reasonable to assume that the zone edge contribution of sp 2 -carbon clusters is mixed in the T peak. Based on this assumption, the very large T-peak width and the shift of the T-peak position can be interpreted. The big Fig. 7. Height ratio of I T /I 1390 as a function of the nitrogen content in the lm.

6 174 J.R. Shi et al. / Thin Solid Films 366 (2000) 169±174 change in the T-peak position demonstrates that the T peak is sensitive to the bonding structure of the amorphous carbon lm. The intensity ratio of T peak to G peak varies with the nitrogen content in the lm. No obvious dependence of the I T /I G ratio on the nitrogen content was observed. This fact further enhances the discussion above that the T peak is not only the measure of the vibrational mode of the sp 3 ±sp 3 bonds, but also a measure of the vibrational modes of the sp 3 ±sp 2 and sp 2 ±sp 2 bonds, as the sp 2 fraction of the carbon atom increases. The count at 1390 cm 21 was used as a reference to compare the T-peak height. Fig. 7 shows the relationship between the height ratio, I T /I 1390, and the nitrogen content in the lm. The I T /I 1390 ratio decreases monotonically with increasing nitrogen content. 4. Conclusion The visible Raman spectra are only sensitive to vibrational modes of sp 2 -bonded carbon atoms, while a new wide band, called the T-band, in the range of 1090±1320 cm 21, associated with sp 3 -bonded carbon atoms, appears in the UV-Raman spectra. For the visible Raman, the G- and D-peak positions of sp 2 -bonded carbon vary in a small range, while the intensity ratio I D /I G increases from 0.34 to 0.81 as the nitrogen content in the lms increases. The G- line-width and the intensity ratio, I D /I G, are sensitive to the structural changes induced by nitrogen incorporation. For the UV-Raman, the G-peak position monotonically decreases from 1665 to 1610 cm 21, and the T-peak position increases tremendously from 1095 to 1314 cm 21.TheGpeak position and width, and the T-peak position, are all sensitive to the bonding structure of the nitrogenated ta-c lms. The T peak is a measure of the vibrational modes of the sp 3 Zsp 3 and sp 3 Zsp 2 bonds, and then the sp 2 Zsp 2 bond, as the sp 3 -fraction of carbon atom decreases. References [1] D.R. McKenzie, D. Muller, B.A. Pailthorpe, Phys. Rev. Lett. 67 (1991) 773. [2] J. Robertson, Prog. Solid State Chem. 21 (1991) 199. [3] P.J. Martin, S.W. Filipczuk, R.P. Netter eld, J.S. Field, D.F. Whitnall, D.R. McKenzie, J. Mater. Sci. Lett. 7 (1988) 410. [4] S. Falabella, D.B. Boercker, D.M. Sanders, Thin Solid Films 236 (1993) 82. [5] P.J. Fallon, V.S. Veerasamy, C.A. Davis, et al., Phys. Rev. B 48 (1993) [6] S. Xu, D. Flynn, B.K. Tay, S. Prawer, K.W. Nugent, S.R.P. Silva, Y. Lifshitz, W.I. Milne, Philos. Mag. B 76 (1997) 351. [7] C.A. Davis, G.A.J. Amaratunga, K.M. Knowles, Phys. Rev. Lett. 80 (1998) [8] V.S. Veerasamy, J. Yuan, G. Amaratunga, et al., Phys. Rev. B 48 (1993) [9] X. Shi, H. Fu, J.R. Shi, L.K. Cheah, B.K. Tay, P. Hui, J. Phys.: Condens. Matter 10 (1998) [10] L.K. Cheah, X. Shi, J.R. Shi, E.J. Liu, S.R.P. Silva, J. Non-Cryst. Solids 242 (1998) 40. [11] L.K. Cheah, X. Shi, H. Fu, B.K. Tay, Electron. Lett. 33 (1997) [12] L.K. Cheah, X. Shi, E. Liu, J.R. Shi, Appl. Phys. Lett. 73 (1998) [13] K.W.R. Gilkes, H.S. Sands, D.N. Batchelder, J. Robertson, W.I. Milne, Appl. Phys. Lett. 70 (1997) [14] K.W.R. Gilkes, H.S. Sands, D.N. Batchelder, J. Robertson, W.I. Milne, J. Non-Cryst. Solids 227±230 (1998) 612. [15] V.I. Merkulov, J.S. Lannin, C.H. Munro, S.A. Asher, V.S. Veerasamy, W.I. Milne, Phys. Rev. Lett. 78 (1997) [16] J. Wagner, M. Ramseiner, C. Wild, P. Koidl, Phys. Rev. B 40 (1989) [17] M.A. Tamor, W.C. Vassell, J. Appl. Phys. 76 (1994) [18] J. Schwan, S. Ulrich, V. Batori, H. Ehrhardt, S.R.P. Silva, J. Appl. Phys. 80 (1996) 440. [19] S. Prawer, K.W. Nugent, Y. Lifshitz, et al., Diam. Rel. Mater. 5 (1996) 433. [20] J. Robertson, Diam. Rel. Mater. 4 (1995) 297. [21] C.A. Davis, S.R.P. Silva, K.M. Knowles, G.A.J. Amaratunga, M.W. Stobbs, Phys. Rev. Lett. 75 (1995) [22] D. Beeman, J. Silverman, R. Lynds, M.R. Anderson, Phys. Rev. B 30 (1984) 870. [23] C.Z. Wang, K.M. Ho, Phys. Rev. Lett. 71 (1993) [24] J.W. Ager, S. Anders, A. Anders, I.G. Brown, Appl. Phys. Lett. 66 (1995) [25] M. Yoshikawa, G. Katagiri, H. Ishida, A. Ishitani, T. Akamatsu, J. Appl. Phys. 64 (1988) [26] M. Yoshikawa, N. Nagai, M. Matsuki, H. Fukuda, H. Ishida, A. Ishitani, I. Nagai, Phys. Rev. B 46 (1992) [27] M. Sugano, A. Kasuya, K. Tohji, Y. Satio, Y. Nishina, Chem. Phys. Lett. 292 (1998) 575. [28] A. Kasuya, Y. Sasaki, Y. Satio, K. Tohji, Nishina, Phys. Rev. Lett. 78 (1997) [29] P.J.F. Harris, Int. Mater. Rev. 42 (1997) 206.