Transmission electron microscopy and electron energy loss spectroscopy of C60 fullerite

u/

Ultramicroscopy 41 (1992) 1-9 North-Holland

y

Transmission electron microscopy and electron energy loss spectroscopy of C60 fullerite Y a h a c h i Saito a, N o r i t o m o Suzuki and Masato Tomita d

b,

H i s a n o r i S h i n o h a r a c, T a k a y o s h i H a y a s h i d

a Department of Electrical and Electronic Engineering, Faculty of Engineering, Mie University, Tsu 514, Japan Toyota Central Research and Development Laboratories, Nagakute, Aichi 480-11, Japan c Department of Chemistry for Materials, Faculty of Engineering, Mie University, Tsu 514, Japan a NTT Interdisciplinary Research Laboratories, Musashino, Tokyo 180, Japan Received 13 November 1991

The crystal structure and the electronic structure of crystalline C60 grown from a benzene solution have been studied by transmission electron microscopy and energy loss spectroscopy. Electron diffraction patterns and high-resolution images reveal an hcp structure (a = 1.01 _+0.01 nm, c = 1.65_+0.02 nm) with stacking disorders. Electron energy loss spectra in a low-loss region revealed not only two collective excitations at 6.5 and 26 eV but also several single-electron excitations. From carbon K-shell excitation, spectral information on unoccupied "rr* and or* bands has been obtained.

1. Introduction

The fullerenes, all-carbon molecules with spherical or pseudospherical closed-cage structures such as C60 and C70, are new interesting modifications of carbon. Their extreme high stability and low reactivity were first discovered by Smalley and his coworkers [1]. The recent synthesis of macroscopic amounts of fullerenes by Kr~itschmer and coworkers [2,3] has stimulated a variety of experimental and theoretical studies. The fullerenes have attracted even more attention since the discovery that alkali-metal doping of the solid phase of crystallized fullerenes, called fullerites, leads to superconductivity [4,5]. The structure of the soccerball-like C6o cluster is now well established by X-ray diffraction [6], ~3C NMR [7-10] and other spectroscopic studies [2,3,10-13]: a truncated icosahedron with 60 vertices, 12 pentagonal faces, and 20 hexagonal faces. All carbon sites are equivalent, but there are two different C - C bond lengths (0.139 nm for edges shared by two hexagons, 0.144 nm for edges shared by a hexagon and a pentagon). As in

graphite, the four valence electrons are hybridized in a s p 2 configuration leading to three strong ~ bonds and a weaker w bond. The electrons are considered to be localized in a hollow spherical shell with a diameter of 0.71 nm (defined by the position of the carbon nuclei) and a thickness of ca. 0.3 nm. Since the C60 molecules interact only weakly (Van der Waals interaction) with one another the spectroscopic data are expected to predominantly reflect the electronic and geometric structure of the individual molecules. The understanding of the crystal structure and the electronic structure of the C6o in a solid phase is an interesting subject in its own right and also a prerequisite for the understanding of the superconducting properties of doped systems. The crystal structures of solid C60 and C7o have been studied by X-ray diffraction [14-16], electron diffraction [3,17] and STM [18]. Vacuum-deposited C60 films have an fcc structure (a = 1.4198 nm) [14,19]. However, solid C60 grown from solutions have a variety of crystal structures depending on solvents used. For example, C6o fullerites grown from a benzene solution have an

0304-3991/92/$05.00 © 1992 - Elsevier Science Publishers B.V. All rights reserved

2

Y Saito et al. / TEM and EELS of C6o fullerite

hcp structure (a = 1.0018 nm, c = 1.6374 nm) with stacking disorders, but grown from a toluene solution they have an fcc structure (a = 1.4171 nm) [16]. On the other hand, both the solid Cv0 grown from a toluene solution and that formed by vacuum deposition have the hcp structure (a = 1.0554 nm, c = 1.729 nm) [16]. The electronic structure of the molecular C~,0 has been studied by optical spectroscopy in hexane solutions [10,11] and photoemission spectroscopy in the gas phase of C ~ ions [20]. In the crystalline form there have been studies with nearly every spectroscopic technique available, e.g., photoelectron [21], inverse photoelectron [22], X-ray absorption [23,24], X-ray emission [25], and electron energy loss spectroscopy [26-29]. In this paper we report on TEM (transmission electron microscopy) and EELS (electron energy loss spectroscopy) studies of C60 fullerite grown from a benzene solution. The crystal structure of the fullerites is revealed by electron diffraction and high-resolution TEM. Information on the electronic structure of w and cr electrons was obtained through valence band excitation and carbon ls core excitation spectra.

Fig. 1. Carbon smoke generated by resistive heating of graphite rods in 100 Torr He. The graphite rods are behind clamps supporting the rods. C7( I dissolved in benzene. The purity was checked by ~3C N M R spectroscopy and high-pressure liquid chromatography.

2. Experiment Fullerene-rich carbon soot was produced by resistive heating or by arc heating of graphite rods in an atmosphere of 20-100 Torr helium. This technique for preparing fine particles is known as gas-evaporation [30]. Fig. 1 shows a picture of carbon smoke emanating from the heated graphite rods. The carbon soot deposited on the inner surfaces of a vacuum chamber was collected and then extracted via a Soxhlet extractor with benzene. After extraction a dark red-brown liquid containing C60, C70, and other fullerenes was obtained [3,9,10]. A further separation and purification of C60 is performed by column chromatography on neutral alumina with n-hexane, which gives a purple-magenta solution of C60 (99% purity) [9,10,23,26]. Fig. 2 shows a picture of crude soot, benzene extract, chromatographed C6o and

Fig. 2. Fullerene-rich soot (A), benzene extract (B), chromatographed C60 (C) and C70 (D) in benzene.

Y. Saito et al. / TEM and EELS of Coo fullerite

Specimens for electron microscopy were obtained by dropping benzene solutions of the chromatographed C60 on copper grids covered with a perforated (holey) carbon film. After evaporation of the benzene solvent under ambient temperature and atmosphere, crystallites of C60 were left. We have examined crystal habits and structures of these crystallites by T E M with a tilting stage. The crystallized C60 have the crystal habits of hexagonal plates and rods, which have been reported previously [17]. In the present study, we used plate crystallites for T E M and EELS because this morphology provides a film thickness of 50 to 500 nm suitable to T E M and EELS. High-resolution TEM images were taken with a JEM-2000EX (C S= 0.7 mm) microscope operated at 200 kV. Some high-resolution images were processed to filter out noise in the images with an L U Z E X III image processor. EELS was carried out in an H-800 electron microscope equipped with a serial analyzer or in an HF-2000 electron microscope equipped with a parallel analyzer (Gatan Model 666). The latter microscope is a specially designed U H V - T E M with a field emission gun, about which a detailed report will appear elsewhere. The energy of a primary electron beam for EELS was 200 keV.

3

An STEM mode with a spot size of about 20 nm was employed in the H-800 microscope, and a TEM mode with an analyzing area of about 300 nm in diameter in the HF-2000 microscope.

3. Results and discussion 3.1. T E M observation

Fig. 3 shows a T E M image and the corresponding diffraction pattern of a hexagonal thin plate of pure C60 with the incident beam parallel to the c axis. Diffraction patterns taken from various directions by tilting the samples can be interpreted as an hcp lattice with stacking disorder [17], which is consistent with the powder X-ray diffraction data for C60 fullerite prepared from a benzene solution [16]. The lattice parameters are a = 1.01 +_ 0.01 nm, c = 1.65 + 0.02 nm. Bent extinction contours and a speckled contrast characteristic of defects are observed. The defects may be introduced by electron irradiation, which brings about desorption of benzene occluded within the crystallites. Prolonged observation under electron irradiation deteriorates the diffraction pattern; higher-order reflections (ini-

Fig. 3. (a) TEM image of a hexagonal platelet of C60, and (b) the correspondingdiffraction pattern taken along the c axis.

4

Y Saito et al. / TEM and EELS of C'~o fidlerite

tially observed up to around 5th-order reflections of 10.0) disappear and sharp spots change to diffusive ones. The solid C60 is apt to suffer from radiation damage. Therefore, the electron dosage must be kept as low as possible to obtain highresolution images. Fig. 4 shows a typical high-resolution image of an edge region of a C(, 0 platelet. The corresponding diffraction pattern is also shown as an inset, indicating that the incident beam is parallel to the c axis. An enlarged and noise-filtered image is shown in fig. 5a. The corresponding Fouriertransformed pattern is also shown in fig. 5b. We notice the characteristic features of hexagonal distribution of black spots in the image. A rhombus depicted in the picture is a unit cell of the hcp lattice projected along the c axis. For a guide to the scale, circles corresponding to the size of a C60 ball (0.71 nm in diameter) are also shown in fig. 5a. It is found that the bright regions correspond to the shell of the C ~ , and dark spots to the cavities inside the molecules. As is well known, a regular hcp lattice has two equivalent sites within the projected unit cell; one is a corner site and the other is one of the two

sites dividing internally a long diagonal of the rhombus in a ratio of 1 : 1 : 1 . Contrary to this expectation, in the observed image, apparently similar dark spots are found at three positions within the projected unit mesh: one position is a corner, and the other two are both the points dividing the long diagonal. This feature is characteristic of an fcc structure, corresponding to the three kinds of stackings, "A", " B " and " C " stackings, perpendicular to the { 1 11 ) directions. However, the diffraction patterns from the C60 crystals are inconsistent with those expected from an ideal fcc structure. Since the introduction of stacking faults to the hcp structure brings about three kinds of stacking, " A " , " B " and "C", it is concluded that the lattice image with seemingly characteristic features of the fcc structure reflects the disorder in stacking sequence of C~,0 molecular layers. 13C N M R studies [7,8] of solid C6~ have revealed that the C~,0 molecules in the solid state must be rotating at rate higher than 10 ~ s-1 at ambient temperature. Since the electron micrographs presented here are recorded at ambient temperature, images with atomic resolution

Fig. 4. High-resolution image of an edge region of a C¢,~ platelet. The corresponding diffraction pattern is inserted.

E Saito et al. / TEM and EELS of Cm~ fullerite

should not be expected even if the microscope does have enough resolving power. It must be necessary to cool the sample to a temperature much lower than the liquid-nitrogen temperature to explore the atomic structure of the C60 cage.

5

3.2. E E L S 3.2.1. Low-loss region Fig. 6a shows an energy-loss spectrum of solid C~,0 in a low-loss region, which is recorded by the

b Fig. 5. (a) Enlarged and noise-filtered image, and (b) the corresponding Fourier-transformed pattern. A unit cell of the hcp lattice projected along the c axis and circles mimicking hollow C~0 balls which are located on a close-packed layer are shown.

6

Y. Saito et al. / TEM and EELS of Coo fullerite

zero-loss

~ a p h i t ~ 4m C J H

/

' Energy

2'0 Loss

'

4'0

'

(eV)

Fig. 6. EELS in low-loss region; (a) solid C60, (b) highly oriented pyrolytic graphite (HOPG), (c) vacuum-deposited amorphous carbon, and (d) diamond.

serial EELS with low resolution ( A E - - 2 eV, F W H M of a zero-loss peak). The electron probe was placed within a region of a plate crystal overhanging a hole of the microgrid so that the electron b e a m did not pass through the supporting carbon film. We see that the spectrum is dominated by two bulk plasmons at 6.5 and 26 eV, which are due to collective excitations of a group of v electrons and all the valence ('rr and or) electrons, respectively. The assignment is verified by the recent high-resolution EELS study [29] showing that the imaginary part E2 of the dielectric constant does not show any peaks at those energy values. Moreover, the plasmon energies revealed in the present study agree well with the energy values predicted theoretically [31]. For comparison with other forms of carbons, spectra f r o m graphite, a m o r p h o u s carbon (vacuum-deposited carbon film) and diamond are also shown in fig. 6. The spectrum of graphite is quite similar to that of C60 fullerite when the resolution of the spectrometer is as low as ca. 2

eV. The two peaks in the spectrum at 7 and 27 eV for graphite are identified to be due to the so-called -rr and cr plasmons, respectively. According to the free-electron model of plasma oscillation, graphite would show the rr plasmons at 13 eV and the ~r plasmons at 25 eV, whereas the observed peaks shift to 7 and 27 eV, respectively, owing to interband transitions involving the cr electrons [32]. For solid C60, the free-electron resonance energies are 11 eV for the "rr plasmon and 22 eV for the so-called e plasmon, where the value of 1.678 g / c m 3 is used for the density of solid C60 [3]. Interactions of collective oscillations and single-electron excitations would cause the shifts of the plasmon energies relative to the free-electron resonance energies, like the plasmons in graphite. On the other hand, amorphous carbon shows a discernibly different spectrum (fig. 6c) from that of C60, and diamond shows a totally different spectrum (fig. 6d). Carbon atoms in diamond are tetrahedrally bonded by the sp 3 orbitals with long-range order, whereas amorphous carbon is considered to have a complicated network structure possessing both trigonally and tetrahedrally coordinated carbons, with more of the former than the latter. In other words, the structure of the so-called amorphous carbon is a complicated network consisting of minute segments of graphite layers. In amorphous carbon, a totally collective effect on electrons does not occur, since rr and cr bands may not be definitely distinguished due to the lack of long-range order, and its structure is extremely complicated. Therefore, the resonance peak originating from rr electrons is not observed. Though it was a rare case, however, a faint shoulder at 7 eV was observed in our amorphous films especially for thick films ( > 100 rim). This may be due to the fact that the electron beam encountered a region of the sample where relatively large segments of graphite were contained. While the C60 fullerite resembles graphite in view of collective oscillations of valence electrons, the former form of carbon shows more rich fine features originating from single-electron excitations than the latter when examined with much higher resolution. Fig. 7 shows such a spectrum in

Y.. Saito et al. / TEM and EELS of C6o fullerite

a region below 28 eV with an energy resolution of 1.0 eV, which was recorded with the parallel EELS. Several broad humps or shoulders other than the bulk plasmon peaks (shown by arrows) are observed around 5, 10, 14 and 17-19 eV, as indicated by vertical lines. These features are more clearly revealed in the recent high-resolution EELS study [29]. The first optically allowed transition (hu-t]g) for C60 should be observed at 3.7 eV, which is actually observed in the ultraviolet(UV)-visible absorption spectrum of C60 both in a hexane solution [10,11] and in a solid form [3] as well as in a photoelectron spectroscopy of solid C60 [21]. The corresponding structure is hardly observed in fig. 7, though there seems to be a very faint hump at 3.7 eV (indicated by a vertical dotted line). In a high-resolution EELS study of our sample, performed very recently [29], the hu-tlg transition is clearly observed. The broad hump around 5 eV, which correlates with the strong UV absorption band at 260 nm [10,11] and with the strong XPS (X-ray photoelectron spectroscopy) peak at 4.8 eV [21], is due to single-electron transitions from bonding rr to antibonding "rr* orbitals. The other features above ca. 10 eV originate from "rr-cr*, or-w* and ~-cr* transitions. 3.2.2 Carbon K-shell excitation Since the core level is sharp in the case of carbon, ELNES (energy loss near-edge structure)

U~

1500

C

o c3 1000

-

©

o

J

7:3

o

500

o

c-

EL 0

5

Energy

10

15

20

25

L o s s (eV)

Fig. 7. EELS of C60 in an energy region below 28 eV. This spectrum is recorded with the parallel EELS with an energy resolution of 1.0 eV.

7

oo (a)

O*

-,-I Ul

c

rr

4J H

' ' %o ....

Energy

Loss

4;o'

(eV)

Fig. 8. EELS in the region of carbon K-shell excitation; (a) solid C60, (b) H O P G , (c) vacuum-deposited amorphous carbon, and (d) diamond.

reflects a density of states above the Fermi level except for the threshold where relaxation effects arising from core-exciton formation bring about a large peak or a sharp rise [33]. An energy-loss spectrum in the carbon K-shell excitation region of the C60 fullerite is shown in fig. 8, where spectra of the other forms of carbon are also shown. The spectra were taken with the serial spectrometer with low resolution (2.3 eV). A sharply rising peak at 284 eV and a band ranging from 290 to 305 eV, near the K-edge, are assigned to a transition to antibonding rr* orbitals and to a transition to antibonding or* orbitals, respectively, following the assignment for graphite and amorphous carbon. The XANES (X-ray absorption near-edge structure) of the C60 fullerite has also exhibited the 7 " and ~r* resonance

8

K Saito et al. / TEM and E E L S of Ct, o fidlerite

bands [23,24]. The presence of a "rr* resonance at 284 eV is characteristic of unsaturated (sp or sp 2) carbon bonds, supporting the proposed sp 2 hybridization in the C60. A broad peak centered at around 320 eV is caused, at least in part, by electrons which undergo both the inner-shell and plasmon excitation. Multiple excitation of plasmons (plural scattering) may be responsible for humps observed in a higher energy-loss region ( > 345 eV). We pay our attention to E L N E S (energy loss near-edge structure) observed in an energy range up to about 50 eV above the onset of the core excitation. ELNES features of the C60 distinguishable from those of graphite and amorphous carbon are as follows: (1) a shoulder is observed at 290 _+ 2 eV as indicated by an arrow (A) in fig. 8, and (2) the width of the ~* band is about 15 eV, which is between the width for graphite ( ~ 18 eV) and that for amorphous carbon ( ~ 13 eV). The shoulder at 290 eV has been observed neither in graphite nor in amorphous carbon, much less in diamond whose ELNES is quite different from that of C~,o. The K-edge excitation spectrum of C60 taken by the parallel EELS with a higher resolution of 1.0 eV is shown in fig. 9, in which the density of unoccupied states calculated by Saito and Oshiyama [34] is superimposed. They calculated the SCC

~f-

T

x:~

molecular orbitals (MO's) of the C60 by the local-density approximation under the densityfunctional theory. The density of states (DOS) curve is obtained by broadening the degenerated MO levels with Gaussians of 0.62 eV F W H M for the lowest four MO's (tlu, tlg, t2~, hg) and 0.82 cV for the other higher MO's (h,, gg, gu, tg, and so on). The experimental spectrum exhibits many fine features, which correlate well with the calculated unoccupied DOS. Peaks at 284, 286 and 288 eV correlate with transitions into "rr * bands derived from tlu, (tlg , tzu , hg), and (h~, gg, gu, tg) molecular orbitals, respectively. However, the experimentally observed peak at 284 eV assigned to the t ~ transition is too strong for its low DOS, from the comparison between the intensity and the DOS for the other two peaks. This high intensity of the onset peak is probably due to the contribution of core-excitonic enhancement, as has been mentioned above. Shoulders and humps in ~* bands are also ascribed to the peaks in the unoccupied DOS. The shoulder observed in the low-resolution spectrum (fig. 8a) corresponds to the broad hump ranging from 290 to 293 eV in fig. 9. The present study shows that distinct features of the C~0 fullerite can be observed conventionally by using an EELS mode in an electron microscope even not devoted to EELS.

4. Conclusion

TT---]

6?

,0 £2

68,8 -

©

•~O

, ~2o% p-'

&00'

~

zu'J73

280 !N

Fig.

9. E E L S

in the

'

'

.,

i '

" i~'

29C @ NC ,,.

near-K-edge

. " '.

-

~'

300 0,89 excitation

J" ?

~ ?,",

C 4 of C60,

which

is

recorded with the parallel EELS with an energy resolution of 1.0 eV. Under the spectrum is shown the unoccupied DOS (broken curve) calculated by Saito and Oshiyama [34].

Electron diffraction and high-resolution T E M revealed that the C60 fullerites grown from a benzene solution have the hcp structure with stacking disorders. EELS of C60 are quite distinguishable from those of graphite and diamond. In a low-loss region two collective excitations and several single-electron excitations are observed. The carbon K-edge excitation spectrum of C60 exhibits rich features which correlate excellently with the unoccupied DOS calculated. Many fine structures in the spectra reminiscent of small molecules reflect the high degeneracy of energy levels in the C60 molecule due to its extremely high symmetry (point group Ih).

Y Saito et al. / TEM and EELS of Coo fullerite

Acknowledgements T h e a u t h o r s w o u l d like t o t h a n k S. S a i t o a n d A. O s h i y a m a of N E C for c o m m u n i c a t i o n of their results prior to publication.

The

work

is s u p -

p o r t e d in p a r t by t h e Y a m a d a S c i e n c e F o u n d a tion and the Okasan-Kato Cultural Foundation.

[15]

[16]

[17] [18]

References [1] H.W. Kroto, J.R. Heath, S.C. O'Brien, R.F. Curl and R.E. Smalley, Nature 318 (1985) 162. [2] W. Kr~itschmer, K. Fostiropoutos and D.R. Huffman, Chem. Phys. Lett. 170 (1990) 167. [3] W. Kr~itschmer, L.D. Lamb, K. Fostiropoulos and D.R. Huffman, Nature 347 (1990) 354. [4] A.F. Hebard, M.J. Rosseinsky, R.C. Haddon, D.W. Murphy, S.H. Glarum, T.T.M. Palstra, A.P. Ramirez and A.R. Kortan, Nature 350 (1991) 600. [5] K. Tanigaki, T.W. Ebbesen, S. Saito, J. Mizukk J.S. Tsai, Y. Kubo and S. Kuroshima, Nature 352 (1991) 222. [6] J.M. Hawkins, A. Mayer, T.A. Lewis, S. Loren and F.J. Hollander, Science 252 (1991) 312. [7] C.S. Yannoni, R.D. Johnson, G. Meijer, D.S. Bethune and J.R. Salem, J. Phys. Chem. 95 (1991) 9. [8] R. Tycko, R.C. Haddon, G. Dabbagh, S.H. Glarum, D.C. Douglass and A.M. Mujsce, J. Phys. Chem. 95 (1991)518. [9] R. Taylor, J.P. Hare, A.K. Abdul-Sada and H.W. Kroto, J. Chem. Soc. Chem. Commun. (1990) 1423. [10] H. Ajie, M.M. Alvarez, S.J. Anz, R.D. Beck, F. Diederich, K. Fostiropoulos, D.R. Huffman, W. Kr/itschmer, Y. Rubin, K.E. Schriver, D. Sensharma and R.L. Whetten, J. Phys. Chem. 94 (1990) 8630. [11] R.L. Whetten, M.M. Alvarez, S.J. Anz, K.E. Schriver, R.D. Beck, F. Diedrich, Y. Rubin, R. Ettl, C.S. Foote, A.P. Darmanyan and J.W. Arbogast, Mater. Res. Soc. Symp. Proc. 206 (1991) 639. [12] D.S. Bethune, G. Meijer, W.C. Tang and H.J. Rosen, Chem. Phys. Lett. 174 (1990) 219. [13] C.I. Frum, R. Engleman Jr., H.G. Hedderich, P.F. Bernath, L.D. Lamb and D.R. Huffman, Chem. Phys. Lett. 176 (1991) 504. [14] R.M. Fleming, T. Siegrist, P.M. March, B. Hessen, A.R. Kortan, D.W. Murphy, R.C. Haddon, R. Tycko, G. Dab-

[19]

[20] [21]

[22]

[23] [24]

[25]

[26] [27] [28] [29] [30] [31] [32] [33] [34]

9

bagh, A.M. Mujsce, M.L. Kaplan and S.M. Zahurak, Mater. Res. Soc. Symp. Proc. 206 (1901) 691. P.A. Heiney, J.E. Fisher, A.R. McGhie, W.J. Romanow, A.M. Denenstein, J.P. McCauley Jr., A.B. Smith and D.E. Cox, Phys. Rev. Lett. 66 (1991) 2911. M. Takata, Y. Kubota, M. Sakata, J. Harada, Y. Saito. H. Shinohara, H. Nagashima and Y. Ando, presented at Meeting Physical Society of Japan, 29 September 1991. Y. Saito, N. Suzuki, H. Shinohara and Y. Ando, Jpn. J. Appl. Phys. 30 (1991) 2857. Y.Z. Li, J.C. Partin, M. Chander, J.H. Weaver, L.P.F. Chibante and R.E. Smalley, Science 252 (1991) 547. Y. Saito, N. Suzuki, M. Terauchi, R. Kuzuo, M. Tanaka, H. Shinohara, A. Ohshita, M. Ohokohchi and Y. Ando, Proc. Int. Symp. Phys. and Chem. of Finite Systems, Richmond, 1991, to be published. S.H. Yang, C.L. Pettiete, J. Conceicao, O. Cheshnovsky and R.E. Smalley, Chem. Phys. Len. 139 (1987) 233. J.H. Weaver, J.L. Martins, T. Komeda, Y. Chen, T.R. Ohno, G.H. Kroll, N. Troullier, R.E. Haufler and R.E. Smaltey, Phys. Rev. Lett. 66 (1991) 1741. M.B. Jost, N. Troullier, D.M. Poirier, J.L. Martin, J.H. Weaver, L.P.F. Chibante and R.E. Smalley, Phys. Rev. B 44 (1991) 1966. H. Shinohara, H. Sato, Y. Saito, K. Tohji, Y. Udagawa, Jpn. J. Appl. Phys. 30 (1991) L848. L.J. Terminello, D.K. Shuh, F.J. Himpsel, D.A. Lapiano-Smith, J. St6hr, D.S. Bethune and G. Meiier, Chem. Phys. Lett. 182 (1991) 491. Y. Saito, K. Kurosawa, H. Shinohara, S. Saito, A. Oshiyama, Y. Ando and T. Noda, J. Phys. Soc. Jpn. 60 (1991) 2518. Y. Saito, H. Shinohara and A. Ohshita, Jpn. J. Appl. Phys. 30 (1991) L1068. Y. Saito, H. Shinohara and A. Ohshita, Jpn. J. Appl. Phys. 30 (1991) Ll145. P.L. Hansen, P.J. Fallon and W. Kr~itschmer, Chem. Phys. Lett. 181 (1991) 367. R. Kuzuo. M. Terauchi, M. Tanaka, Y. Saito and H. Shinohara, Jpn. J. Appl. Phys. 30 (1991) L1817. R. Uyeda, Prog. Mater. Sci. 35 (1991) 1. G. Barton and C. Eberlein, J. Chem. Phys. 95 (1991) 1512. R.F. Egerton, Electron Energy Loss Spectroscopy in the Electron Microscope (Plenum, New York, 1986) p. 163. E.J. Mele and J.J. Ritsko, Phys. Rev. Lett. 43 (1979) 68. S. Saito and A. Oshiyama, Phys. Rev. Lett. 66 (1991) 2637.