A NEXAFS characterization of ion-beam-assisted carbon-sputtered thin films

ELSEVIER

Diamond

A NEXAFS

and Related

characterization

de M&allurgie

Physique,

URA

b, J. Delafond 131 CNRS,

F-86022 b Laboratoire

pour I’Utilisation

du Ruyonnement Campus

’ Luborutoire

d’Etudes Received

d’Orsuy,

Poitiers 2090,

de Poitiers,

a, G. Hug ’

40 Avenue

de Recteur

Pineau,

Frcmw CNRS

F-91405

CEA-MEN,

Orsay

ONERA-CNRS,

1994; accepted

carbon-sputtered

a, N. Junqua

UniversitC Cedex,

Electromugn~tique, B6timent

des Microstructures, 27 July

4 (1995) 200~206

of ion-beam-assisted thin films

M. Jaouen a, G. Tourillon a Luhorutoire

Materials

Centre

Uniwrsituire

de Pks

Sud

Cede-x, France

BP 72, F-92322

in final form 17 November

ChBtillon,

France

1994

Abstract Thin carbon films were deposited by sputtering with various species by means of the technique of dynamic ion mixing (DIM). Transmission electron microscopy experiments reveal that the films are amorphous and contain contaminants, mainly oxygen. Density values are determined by glancing X-ray reflectometry. Friction and wear resistance measurements have been accomplished. Near-edge X-ray absorption fine structure (NEXAFS) experiments are performed at both CK and OK edges. The CK edge NEXAFS data for a film obtained with DIM and an argon-methane mixture filling the sputtering source show a strong peak that is attributed to C-H and O=COH bonds, while the number of sp2-hybridized carbon sites is quite small. More damped features are observed for a film deposited without DIM when an argon-hydrogen mixture is used. However, the number of sp’-hybridized carbon sites is then greater. When DIM is used, we observe in the NEXAFS spectrum that the high energy ion irradiation induces

hydrogen release. A correlation of the NEXAFS with the wear behaviour is suggested. Keywor&

Amorphous

carbon;

Ion-assisted

deposition;

X-ray absorption;

1. Introduction For some years the economical and technological properties of amorphous (a-C) and hydrogenated amorphous (a-C:H) carbon films have attracted the interest of many researchers [ 1,2]. These films exhibit properties such as hardness, electrical insulation, chemical inertness, optical transparency, smoothness, resistance to wear and sputtering. They can be produced by various deposition techniques (see Ref. [3] and references cited therein): carbon arc deposition, arc plasma CVD, ion beam deposition, laser evaporation, ion beam sputtering, evaporation with concurrent ion bombardment, as well as r.f. and d.c. sputtering. However, the physical properties of amorphous carbon films can vary dramatically depending on the deposition process, the temperature [4] and the hydrogen content. In particular, it has been

shown [S] that it may not be possible, at least for a-C:H films, to optimize all the growth parameters for obtaining simultaneously all the desirable properties. In this paper we focus our attention on the improvement in friction and wear resistance of stainless steel 0925-9635,!95;$09.50 0 1995 Elsevier Science S.A. All rights reserved SSDI 0925-9635(94)00252-5

Wear

(304L) plates coated with carbon films. To achieve good adhesion, the film deposition was assisted during growth with an independently controlled high energy ion beam, a process known as dynamic ion mixing (DIM). For a deeper understanding of these carbon coatings, besides knowledge of the wear properties, comprehensive structural studies are needed. To reach this goal, we have used many techniques. In particular, we concentrate our attention on high energy resolution near-edge X-ray absorption fine structure (NEXAFS) experiments to obtain a more reliable correlation of the real film structure with the observed wear behaviour.

2. Deposition method A schematic diagram of the apparatus [6] used in this study is shown in Fig. 1. The vapour flux is obtained by ion sputtering of a pure carbon target (greater than 99.99%) using a broad Kaufman [7] ion source. This source was fed with Ar or a mixture Ar + H? or

M. Jaouen et al. / Diumond and Relured Materials Thickness monitor

Rotating substrate

Q -f=l-Vacuum system

system

for

producing

3. Wear behaviour and TEM characterizations coatings

by

dynamic

3.1. Friction and wear resistance results

Ar + CH,. The acceleration voltage was 1.2 kV and the typical current intensity 120 mA. The ions hit the target at an incident angle of 4.5”. The deposition chamber has been designed to operate in line with a 200 kV ion implanter [6]_ The rotating sample holder is positioned at an angle of 45” to the ion beam and the vertical axis, allowing continuous deposition and implantation. The vacuum system has a base pressure less than 10e4 Pa and the pressure during the deposition process is held below lop2 Pa. All depositions were performed at room temperature. The thickness was monitored by a calibrated quartz oscillator. The mean deposition rate is of the order of 0.6 nm min- ‘. For the DIM experiments the ratio of implanted ions to deposited atoms was held constant at 1% by monitoring the implanter current. The substrates were freshly cleaved rock salt for transmission electron microscopy (TEM) observations, fused quartz plates for NEXAFS experiments, stainless steel discs (diameter 32 mm, thickness 8 mm) for wear resistance tests and Si wafers for X-ray thickness measurements. The stainless steel plates were mechanically polished to a mirror finish. The substrates were ultrasonically cleaned in a bath (distilled water plus ethanol) for 10 min, rinsed and then dried with Ne gas before the deposition. Four coatings (labelled Cl-C4) were examined: the first three are expected to be a-C:H films, while the fourth is an a-C coating. The deposition conditions are listed in Table 1. We also give the film densities measured by grazing X-ray reflectometry in the angleresolved dispersive mode with a position-sensitive proportional counter [S]. This experiment was not possible for the unassisted coating Cl, because it shows cracks Table 1 Parameters Sample

Cl c2 c3 c4

201

and stripping. No such behaviour was observed for the ion-beam-assisted samples. Again we verify here the now well-established fact [9] that the bombardment of a growing film by energetic particles produces beneficial modifications such as improved adhesion (mixing of the interface) and change in the internal stresses. Obviously a coating such as sample Cl will not be adequate for mechanical design.

, holder Kaufman source

Fig. 1. Experimental ion mixing.

4 (1995) 200-206

used to deposit

carbon

Sputtering (Kaufman

films ClLC4. species source)

Ar (20%) + H, (80%) Ar (37%) + CH, (63%) Ar (20%) + H, (80%) Ar

Thicknesses,

The apparatus used to measure the kinetic coefficient of friction was a pin-on-disc tribometer [lo] consisting of a spherical ruby ball (5 mm in diameter) sliding on the coated disc. Discs, as well as the ruby ball, were cleaned for 45 min using ethanol in an ultrasonic cell before friction measurements. The friction coefficient was measured in air at room temperature (RT) rotational under an speed of 5 rev min- ’ without lubrification applied load varying between 0.5 and 6 N. The best result was obtained for sample C2 as shown in Fig. 2, where the evolution of the kinetic friction coefficient p is plotted vs. the number of turns under a load of 6 N. After a certain number of passes, ,B reaches a steady state value close to 0.1, with a very low amplitude of the “stick-slip” process. At this stage no

P=6N

P = OSN

I-

1

It 0.5

0

r

I

I

0

I

I

1

2000

1000

3000

turn number Fig. 2. Friction coefficient p as a function of the turn number for sample C2 (load P = 6 N). Inset: uncoated stainless steel (P =0.5 N).

as well as density Mixing ions

None Ar+ (150 keV) Ar+ (160 keV) Ar+ (160 keV)

values determined

by glancing

X-ray reflectometry,

Thickness

Density

(nm)

(g cm-‘)

270 120 120 120

Impossible 2.1 & 0.2 2.4 + 0.2 2.3 k 0.2

are also given

to measure

wear cracks were visible on the disc. Looking at the behaviour of an uncoated stainless steel disc shown in the inset of Fig. 2, one can see clearly the improvement in the wear resistance of the material supplied by the carbon overlayer. It must be outlined that this very good behaviour is obtained with an unusually thin coating. the more commonly used thicknesses being in the range of several microns. The same kinds of results have been observed for sample C4, but with a higher steady state value ~1z 0.2. The situation was found not to be so good for sample C3. For a load of 6 N there occurs a first steady state with a small friction coefficient (ccz 0.2) and a small stick-slip amplitude. After 65 turns there appears a second steady state with a higher friction coefficient (~~0.3) and a stick-slip amplitude (Ap ~0.1) which looks like that of the uncoated material.

3’06 -l[i 2

E

: B a 9 2 a

2 106 --

Ep = 24.5

C-K edge 1106-

0 100 0

100

200

300

400

500

Energy Loss (eV)

5 105

r

3.2. TEM observations The films deposited on rock salt were removed by dissolving the latter in doubly distilled water and then transferred to copper grids 3 mm diameter for TEM experiments. The films were sufficiently thin to be electron transparent and thus there was no need for ion beam milling. To establish the composition of film C2, a Gatan parallel electron energy loss spectrometer (PEELS 666) in a Jeol 4000 FX microscope working at 400 kV was used. The position (24.5 eV) and width (21.2 eV) of the plasmon loss peak (Fig. 3(a)) are consistent with the existence of an amorphous carbon phase [l 11, although it should be noted that PEELS will not detect hydrogen in this amorphous film. Moreover, along with the CK edge, the OK edge at 532 eV. the CrL2.3 edges at 572 eV and the FeL,.3 edges at 708 eV can be observed in the energy loss spectrum (Fig. 3 (b)). Oxygen is a possible contaminant arising from the relatively poor vacuum in the reaction chamber. The presence of iron and chromium indicates a sputtering of the walls of the chamber by the ions coming from the Kaufman source during growth. A crude estimate of the impurity content made by removing the background from the EEL spectra and estimating the areas under the edges [ 121 indicates that Fe, Cr and 0 concentrations were less than 3%, 2% and 8% respectively. NEXAFS experiments will show that all films contain oxygen as contaminant. Analytical TEM examinations were performed at 200 kV in a Jeol 200 CX microscope. All films were found to be amorphous, with no marked differences between them. A typical selected area diffraction (SAD) pattern obtained from sample C2 is shown in the inset Fig. 4. Two diffuse rings characteristic of amorphous carbon [ 131 are observed. The bright field electron micrograph in Fig. 4 exhibits an “orange-skin”-like contrast that may be interpreted as a tendency to the formation of clusters on a subnanometre scale.

olo”l

Jl

I

I

200 300 400 500 600 700 800 900 1000 (bl

Energy Loss (eV)

Fig. 3. (a) Plasmon loss peak from aample C2; (b) detail of the 200 1000 cV loss region showing peaks from carbon at 18X eV. oxygen at 523 cV. chromium al 575 eV and iron at 70X cV.

Fig. 4. Bright field electron micrograph of lilm c‘2 (scale bar=O.l /im). Inset: electron diffraction pattern showing two ditfilse rings chnracterw tic of amorphous carbon.

Unfortunately, the TEM results do not provide enough microstructural information to allow us to understand the differences observed in the friction and wear behaviours of the tested coatings.

203

M. Jaouen et ul. / Diamond and Related Materials 4 (1995) 200-206

4. NEXAFS

results and analysis

4.1. Experiment The experiments were conducted at the Laboratoire pour 1’Utilisatiot-r du Rayonnement Electromagnetique (LURE, Orsay, France) on the VUV Super-AC0 storage ring. They were carried out on the SACEMOR beam line [ 141 using a high energy TGM monochromator better than 0.1 eV (1200 lines mm-’ grating, resolution at the CK edge). For NEXAFS the incident beam I, was monitored by collecting the total electron yield from an 85% transmission copper grid freshly coated with gold. The total electron yield from the sample, I, was then normalized with respect to I,. The electrons were collected over 1 s for each data point, the energy step being 0.1 eV. For comparison, diamond and highly oriented pyrolytic graphite (HOPG) samples have also been probed. The spectra for a-C, a-C:H and diamond were recorded at normal X-ray incidence (E parallel to the surface). Normal incidence was chosen for convenience and because it is expected that NEXAFS spectra from amorphous samples Cl-C4 will reveal no angular dependence according to their disordered nature as is observed in TEM experiments. The tetrahedral symmetry of diamond eliminates polarization-dependent effects [ 151. For HOPG an X-ray incidence close to the “magic angle” (55” from the c axis) was used to suppress polarization-dependent effects [ 161. We do not attempt to remove any possible surface oxygen contamination by argon sputtering for two reasons: first, it is well known that sputtering treatment provokes structural changes in thin carbon films [ 171; second, the oxygen atoms are distributed within the whole volume as was revealed by EELS. This last point was confirmed by NEXAFS experiments performed at the OK edge in the fluorescence mode, which probe nearly the entire sample thickness. 4.2. Results The NEXAFS spectra for diamond, amorphous samples and HOPG are shown in Fig. 5. The data for graphite and diamond are in good agreement with those previously reported in the literature [17,18]. From the examination of the diamond spectrum it can be pointed out that the overall energy resolution was quite good owing to the distinguishable presence of the sharp exciton peak [ 191 localized 0.19 eV below the C 1s ionization energy. On the other hand, the data collected over disordered carbon films Cl-C3 look quite different from those already obtained either with X-ray absorption spectroscopy (XAS) [ 16,201 or with EELS [3,21-231. This is more striking for films Cl and C2, especially at the absorption edges, which show very strong peaks close to 290 eV.

/I 280

290

/ 300 Enwzy

Fig. 5. CK edge NEXAFS spectra at the ‘magic’ angle and diamond. observed as indicated.

1

310

320

@V)

of samples ClLC4, HOPG recorded The p* resonance and s* bands are

The assignment of the various features in carbon K shell absorption spectra has been discussed in previous papers [ 15,16,20] and the labelling in Fig. 5 reflects that assignment. They are characterized by the presence of a rt* resonance at 285 eV characteristic of unsaturated (sp or sp2) carbon bonds and a broad o* feature that arises from a superposition of the signatures of many different bonding configurations in the matrix. From the examination of the o* region the presence of a significant sp-bonded component must be precluded, since it would cause a rather sharp cr* intensity around 310 eV. The film C2 generated by sputtering the carbon target with a mixture of Ar + CH, is of most interest. The rc* resonance centred close to 285 eV is observed to be the smallest among the four analysed samples Cl-C4. Generally one finds that a carbon K edge NEXAFS spectrum consists of a peak close to 285 eV due to excitations from the 1s level to empty rr* states of sp2 sites, followed by a step at around 289 eV due to transitions from the Is level to empty o* states of both sp2 and sp3 sites. Therefore it is theoretically possible to extract from the data the fraction of sp2 sites in any sample by normalizing the area of the n* resonance with the area of a large section of the spectrum and comparing this ratio with the value for graphitized carbon that

contains only sp* sites [4,23]. This method has been shown to be independent of the hydrogen concentration of the material [24]. Unfortunately, we had no amorphous material with exactly determined sp’ content that was suitable as a reference spectrum to yield a calibration of the ratio of the 7-c*and D* integrals. Furthermore, this approach must be considered with caution, because the x* transition intensity variation depends strongly on the orbital delocalization. Thus the results presented here only allow the comparison of the investigated films. Towards this aim, the spectra presented herein were normalized at 320 eV, since the non-resonant continuum at this energy is atomic like [ 1.51. The n* resonances are then fitted by gaussian line shapes and their areas compared. They are found to be nearly equal for samples Cl, C3 and C4, while that for C2 is only 35% of the other three. From this rather crude estimation one may thus think that sample C2 has a rather high sp3 content. It will reach 65% if we assume film C4 to be 0% sp3 hybridized as is the most probable case. However, one must keep in mind that there exist sharp resonances at the CK edge of film C2 that require a more detailed analysis. Fig. 6 shows an enlarged view of the C2 CK edge spectrum as well as that related to its second derivative. It is clearly seen that there exist at least fout resonances (labelled a-d), the first one obviously being related to the ls+7c* (C=C) transition. To obtain a better insight into the C2 sample excitation spectrum, we have used a curve-fitting technique [ 151. Owing to the rather good energy resolution of the TGM monochromator, it is assumed here that the line shape of the continuum step is determined by the lifetime of the core hole. Following Paratt 1251. this implies that the continuum step is then described by a tan ’ function: I step

of the step. H is the step height, r,_ (0.4 eV) is the width of the step and E is the energy. The decline of the step function above the edge is modelled as usual [ 151 by an exponential decay. Since the monochromator resolution dominates, the TC*resonance peaks will have a gaussian line shape [ 151 given by I, = H exp

-

1 E-E, - ~2 r,:c

i

where H is the maximum of the function, r, is the full width at half-maximum (FWHM) of the peak, E, is its position and c is a constant defined by c = 2( In 4)‘!‘= 2.355. The best-fit results are shown in Fig. 7 and the related parameters are listed in Table 2. Using the results collected in the database [ 151, the peaks can be assigned to ls+7~* (C=C), ls+7z* (C=O), ls+x* (C-H) and ls-+n* (O-C-OH) transitions respectively. Overall the fit is satisfactory, except right near the step where the limitations of the tan -l function in representing the continuum step become apparent. Furthermore, the use of only one function to model the step is too crude an approximation in the present case, because there must exist at least two (or maybe three) ionization potentials (IPs) owing to the chemical shift of the C Is level. A more adequate description of the edge jump would use two (or three) tan- ’ functions. but unfortunately we did not perform any X-ray photoelectron spectroscopy (XPS) measurements to obtain 2.5

=H

where E, (289.1 eV) is the position

of the inflection

point

0.0

280 a

b

c

2x5

290

295

300

305

310

315

Energy (eV)

d

fitted with an arctan Fig. 7. Sample C? CK edge spectrum function (~~~ ) and four gaussian functions (--- ) to model the resonances a-~d (see text). Table 2 Resonances a d (see text) positions and widths fitting procedure for sample C2. The assignments are also given Peak 282

284

286 Energy

288

290

E,; (eV)

Position

Width FWHM

284.8 2X6.4 287.5 2X8.5

1.0 1.7 1.0 1.0

deduced from the of the resonances

Assignment (eV)

292

(eV)

Fig. 6. Enlarged view of coating C2 CK edge (0 ) as well as its second derivative (&) showing the four resonances labelled a-d (see text).

: :

ls+rr*(C=C) ls+rr*(C=o) Is-rr*(C II) Is-x*(0 C OH)

M. Jaouen et al. 1 Diamond and Reiated Materials

the C 1s TPs. Therefore the fit presented here must be used with care; in particular, the resonance heights (except for the n* (C=C) transition) have no significant physical meaning. In the same way we do not try to fit the broad o* band that involves solid state effects. One can note, however, some humps that can be attributed to ls+o* (C-C) (about 293.5 eV), ls-+cr* (C=C) (about 300eV) and Is+@ (O=C-OH) (oT about 296.6 eV; 0; about 303.5 eV) resonances. To summarize, the main point to keep in mind during the discussion that follows is that these results allow us to know the nature of the various bonds involved in this sample. 4.3. Discussion The results quoted above may help us to obtain a better understanding about the C2 sample network. The first striking result is the weakness of the rr* (C=C) resonance: the number of sp* (graphite-like) bonds is quite small. Furthermore, the rc* (C=C) resonance is rather broad compared with the HOPG one the presence of several (rci = 1.1 eV). This indicates cluster sizes corresponding to a certain cluster size distribution [26]. These clusters are oriented randomly because no polarization effects are observed. The appearance of a rc* (C=O) resonance at the CK edge clearly shows that most oxygen atoms are bounded to carbon, most probably at the cluster boundaries. This is also evidenced by the examination of the OK edge (see Fig. 8) which exhibits a well-defined rc* (O=C) resonance. The existence of the Is-C-H peak at 287.5 eV as in cyclohexane (C6Hi2) [27] proves unambiguously that we have a hydrogenated material (a-C:H). The hydrogen atoms saturate some carbon dangling bonds and lead to the formation of sp3 sites. Finally, there also exist acidic-like bonds (peak labelled d in Table 2) so that some carbon atoms at the sp2 cluster boundaries involve bonds with COOH and CO functions. To summarize, we think that the C2 sample network can be described by a two-phase model: subnanometre “graphite-like” sp2

2.5 1s-xx* (o=c)

0.0

I--,530

’

535

I

540

I

545

I

550

spectra

205

clusters terminated either with hydrogen or a COOH acidic function. However, the rather structureless o* region indicates the absence of long-range order in the film. The clusters, which contain some sp3 sites, may be viewed as being embedded in a “random matrix” that includes a polymer-like component at the boundaries. This last component is responsible for the good sliding properties we observed during wear tests: it acts as shock absorber when the material is strained. To some extent the same description holds for film Cl. However, the rc* (C=C) resonance is enhanced compared with that obtained for sample C2. Although the C-H peak still exists, it should be noted that the edge is broadened and that the O=C-OH contribution is strongly damped. Here the oxygen atoms are not as well bounded to carbon or hydrogen atoms. This behaviour is confirmed by the OK edge examination (see Fig. S), which shows a weak and very broad n* (O=C) resonance; in this case the OK edge shape looks more like one related to the presence of an alcohol function

t-151. The situation becomes quite different for sample C3, whose CK edge shows no evidence of much hydrogen in the film. There remains only a small O=C-OH contribution, the C=O bonds being strongly distorted (see Fig. 8). The comparison between the CK edges related to films C2 and C3 respectively confirms that a high energy deposition method not only improves the adhesion properties but also leads to hydrogen release during growth [ 28,291. These changes may be explained by bond breaking caused by ionization or displacement processes and relaxation of the microscopic structure. On the other hand, the film C2 growth was also ion beam assisted. Therefore all these experimental results evidence that hydrogen trapping during growth is enhanced when an Ar + CH, mixture fills the Kaufman source. We do not have a clear explanation of this behaviour, but one might assume that CH, molecules hitting the carbon target induce the formation of some CH, radicals (x ~4) in the flux of particles impinging on the growing film. This is less probable when H, molecules and Ar ions are used; in that case there exist fewer C-H bonds in the film, most of them being broken inside the material by the high energy irradiation. Finally, sample C4 looks like the a-C samples reported in Refs. [20,23] and requires no further comments, except for the very weak hump around 289 eV. This may come either from surface contamination or more probably from the residual water vapour that condenses on the surface during growth and therefore from the formation of some COOH-like radicals.

I

555

540

Energy (eV)

Fig. 8. OK edge NEXAFS

4 (1995) 200-206

of samples

Cl-C3.

5. Conclusions The wear behaviour of amorphous carbon deposited by DIM has been studied as a function of the sputtering

species. The best result was obtained by filling the Kaufman source with Ar + CH, for a deposition assisted with 150 keV Ar+ ions. This high energy ion beam mixing induces both stress relief and adhesion of the coating to the substrate. TEM observations reveal that such films are amorphous. NEXAFS experiments show unambiguously, for the first time with this technique to our knowledge, that some coatings (sample C2. Ar + CH, with DIM; sample Cl, Ar + H, without DIM) are hydrogenated. This conclusion can be correlated with density measurements, since the lowest value is that for film C2 (see Table 1). From the NEXAFS results we suggest that this sample may be viewed as a network composed of two phases: nanocrystalline sp2-hybridized clusters involving sp3 sites embedded in a “polymerlike” component that includes oxygen atoms. To test this hypothesis, one needs further experiments such as EELS imaging (EELSI) which may yield a precise localization of sp2 or sp3-hybridized carbon [30] and oxygen atoms by selecting adequate integration energy windows. One can also think of performing EXAFS experiments, but unfortunately the too high amount of oxygen precludes any EXAFS analysis such as that reported in Ref. [ 161 which would allow us to obtain a detailed understanding of the coating structure on the atomic scale.

Acknowledgements We would like to thank C. Fayoux for performing the depositions and A. Naudon for the glancing X-ray reflectometry measurements.

References Prog. Solid State Ckern., 21 (1991) 199. F. Rossi, in W. Gissler and H.A. Jehn (eds.), Advanced Techniques Stw@ce Engineering, 1992, p. 37 I. J.J. Cuomo, J.P. Doyle, J. Bruley and J.C. Liu, J. Vu. Sci. Technol. A, 9 (1991) 2210.

[l] J. Robertson, [2] [3]

141 J.J. Cuomo, J. Bruley, J.P. Doyle, D.L. Pappas, K.L. Saenger. J.C. Liu and P.E. Batson, MRS Smp. Proc., 20_7 ( 1991) 247. J.D. Warner, D.C. Liu and J.J. Pouch, I51 S.A. Alterovitz, /. Electrochem. Sot., 133 (1986) 2339. G. Laplanche and J. Delafond, 161 M. Jaulin. S. Pimbert-Michaux. Surf: COClf. Techno/., 37 (1989) 225. 171 H.R. Kaufman, Adv. Elrctron. Elrctron Phys., 36 ( 1974) 265. J. Chihab, P. Goudeau and J. Mimault. .I. Appl. C81 A. Naudon, Cr~stullogr., 2-7 ( 1989) 460. [91 J.I% Riviere and J. Delafond, Surf: EUR., Y (1993) 59. [IO1 P. Moine. 0. Popoola and J.P. Villain, Ser. Metull.. -70( 1986) 305. 1111 R.F. Egerton and M.J. Whelan, J. Elrcrron. Specrrox. Relat. Phe~~ow~.,3 ( 1974) 232. 1121 R.F. Egerton. m EELS in r/w Elec/ron bJkroscopc, Plenum, New York. 1986. A. Oberlin and E. Bouillon. Cornpos. Sci. 1131 M. Monthioux, 7‘e(~/l/lo/.. 37 (1990) 21. J. Phw. c 141 P. Parent. C. LalTon. A. Cassuto and G. Tourdion, Chwl., in press. 1151 J. Stijhr in NEXAFS Spectroscopy. Springer, New York, 1992. Cl61 G. Cornelli, J. StBhr. G.J. Robinson and W. Jark. Phys. Rw. B, 3x (1993) 1004. P.J. Love and V. Rehn, P/zy.s. Rut,. B, 33 1171 R.A. Rosenberg. ( 1986) 4034. [181 J.F. Morar, F.J. Himpsel, G. Hollinger, J.L. Jordan, G. Hughes and F.R. McFeely, Phys. Rec. B, 33 ( 1986) 1346. G. Hughes and J.L. 1191 J.F Morar, F.J. Himpsel. G. Hollinger, Jordan. Ph_w. Rec. Lett., 54 ( 1985) 1960. R. Carr. T.K. Sham. M. Strongin, 1201 D. Wesner, S. Krummacher, W. Eberhardt, S.L. Weng, G. Williams. M. Howells. F. Kampas, S. Heald and F.W. Smith, Phys. Rec. B. 28 (1983) 2152. D.K. Veers, M.D. Rubin, C.B. c211 N.H. Cho. K.M. Krishnan, Hooper, B. Bhushan and D.B. Bogy. J. il/ltrtrr. Res.. 5 (1990) 2543. 1221 B. Bhushan, A.J. Kelloc, N.H. Cho and J.W. Ager III. .I. Muter. Ra.\.. 7 ( 1992) 404. V.S. Veerasamy, C.A. Davis. J. Robertson, v31 P.J. Fallon, G.A.J. Amaratunga, W.I. Milne and J. Koskinen, P~JX Rev. B. 48 (1993) 4777. and L.M. Brown. Diwwntl Rrlur. Melter., 2 [241 P.J. Fallon ( 1993) 1004. 1251 LG. Paratt, Rec. &Jod. Phys.. 31 (1959) 616. [261 R. Kleber, K. Jung, H. Ehrhardt, I. Miilhing, K. Breuer. H. Metz and F. Engelke, Thin Solid Films. 205 (1991) 274. 1271 G. Comelli and J. Stiihr, Surj: Sci.. -700 (198X) 35. C281 K. Tanaka, M. Okada, T. Kohno, M. Yanokura and M. Aratam, Nut,/. Jtustrurn. M~rlzods Phy~ Rex B, 5X ( I99 1 ) 34. 1291 A. Fuchs, K. Jung, H. Ehrhardt. I. Mtilhing and K. Breuer. 7’hin Solid Fib. ,717 ( 1992) 48. [301 L.M. Brown. Nature, 366 (1993) 721. D.A. Muller. Y. Tzou, R. Raj and J. Sicox. Nuture, 366 (1993) 725 P.E. Batson. h’~~tur~.366 ( 1993) 727.