Interactions of deuterium and hydrocarbon species with the diamond C(111) surface

Surface Science Letters 257 (1991) L633-L641 North-Holland

Surface Science Letters

Interactions of deuterium and hydrocarbon C( 111) surface Yoshitaka

Mitsuda

I, Taro Yamada,

T.J. Chuang,

species with the diamond

H. Seki

IBM Research Division, Almaden Research Center, 650 Harry Road, San Jose, CA 95120-6099,

R.P. Chin, J.Y. Huang

USA

and Y .R. Shen

Department of Physics, University of California and Material Sciences Division, Lawrence Berkeley Laboratory, Berkeley, CA 94720, USA

Received 13 June 1991; accepted for publication 29 July 1991

AES, XPS, TDS and LEED have been used to study the adsorption, the desorption and the reaction behavior of D atoms and CH, (1 IX I 3) species on the diamond C(111) surface. It is found that a small amount of chemisorbed D atoms can induce the (2 x 1) to (1 X 1) phase transition and hold the sp3 structure up to nearly 1400 K. The structural transformation can also be monitored by the optical second-harmonic generation technique. Thermal desorption of deuterium occurs mainly around 1300 K. Adsorption of CH, can also induce the phase change but not as readily as the D atoms. Annealing of the adlayer at a high temperature results in the partial conversion to graphitic carbon. D atoms can react with both the adsorbed CH, and the surface graphite. The surface bonding characteristics and the relevance of present results to the diamond CVD process are briefly discussed.

Chemical vapor deposition (CVD) at relatively low pressures with a hydrogen-hydrocarbon gas mixture to produce diamond thin films by a plasma or a hot-filament technique has attracted great attention in recent years [l]. In the CVD process, H atoms and certain hydrocarbon species (C,H,), in particular, the methyl radical, have been considered very important reactants [2-51. Obviously knowledge of the chemisorption, the reaction and the desorption behavior of H and C,H, on a diamond surface is essential for a better understanding of the mechanisms and the basic steps involved in the growth of diamond films. The adsorption of H atoms on the diamond C(111) and C(100) surfaces has been investigated to a considerable extent [6-111. Whereas, studies of hydrocarbon species adsorbed on such crystal ’ Present address: Institute of Industrial Science, University of Tokyo, 7-22-l Roppongi, Minato-ku, Tokyo 106, Japan. 0039-6028/91/$03.50

surfaces remain very scarce [12]. Recently, attempts to identify the chemical species at the surface during the diamond nucleation and the growth process using mass spectrometry and X-ray photoemission spectroscopy (XPS) have been reported [13,14]. We present here the initial results of the study on the interactions of D atoms and CH, species with a C(111) surface using thermal desorption spectroscopy (TDS), XPS, Auger electron spectroscopy @ES) and low energy electron diffraction (LEED) methods. In the present study, deuterium gas instead of hydrogen is employed in TDS to avoid the ubiquitous hydrogen background of the UHV chamber. The chemical behavior of D atoms, in the scope of present investigation, is basically the same as the H atoms. In fact, we have’ also investigated the interaction of H atoms with a C(111) surface with the highly surface-sensitive optical second-harmonic genera-

0 1991 - Elsevier Science Publishers B.V. All rights reserved

Y. Mitsuda et al. / Interactions of D and CH, species with the C(lI1)

tion (SHG) technique 1151,and found good correlation of its rest& with those of D -adsorption obtained by AES and LEED. It is observed that the bare C(111) surface reconstructs to a (2 X 1) and/or (2 x 2) structure. The two patterns are practically indistinguishable by LEED and will be designated simply as (2 X 1) hereafter. Exposure of the clean surface to a small amount of D atoms causes the surface to transform into the (1 x 1) structure. The critical surface coverage needed for the phase transition is determined to be about 5% of a monolayer (ML) or less. The chemisorbed D atoms desorb mainly around 1300 K. AES spectra exhibit fine features which are sensitive to surface chemical and structural changes. The reversible phase transformation between the (2 x 1) and the (1 X 1) structures are evident from the AES spectra, the SHG signals and the LEED patterns. Exposure of CH, gas through the hot tungsten filaments results in the adsorption of CH, (with 1 IX 5 3) on the surface. D atoms can react with the adsorbed CH, species and cause the surface to deuterate. The results relevant to the diamond CVD process are also briefly discussed. A UHV chamber (base pressure 5 X lo-‘* Torr) equipped with TDS, XPS and AES spectrometers (PHI Mode1 5481, as described previously 1161,is used with a turbo molecular pump as a supplement to the ion pump for the main chamber. A second turbo pump is employed for differential pumping of the small chamber which houses the quadruple-mass-spectrometer (QMS) used in TDS. LEED patterns are obtained by the reverse-view LEED optics (Princeton Research Inst~ments). A polished type 2A diamond C(lll> crystal (8 mm diameter, 0.5 mm thickness, Dubbeldee Corp.) is cleaned by a H$O,-HNO, acid at 90°C followed by H,O rinsing and ultrasonic cleaning in acetone and methanol solvents to remove metallic and graphitic contaminants on the surface. The crystal is mounted on a MO cup with a W(5% Re)-W(26% Re) thermocouple attached near the front surface. Sample heating in UHV for surface cleaning, annealing and TDS is carried out with a RF-induction heater. The sample temperature is monitored by both the thermocouple and an optical pyrometer (Wahl, In-

surface

frared Thermometer, Model DHS-52). The temperature readings by the two methods agree within f50”C which is considered to be the experimental uncertainty. Dosing and adsorption of D atoms and CH, species are accomplished by passing D, and CH, gases through the W-filaments maintained at 2000°C. During the gaseous exposure through the filaments (typically 10 min), the sample is heated to about 500 o C due to the radiation from the filaments. Without the hot filaments, the exposure of C(111) to D, and CH, gases does not result in any adsorption on the surface. Typical experimental procedure involves the cieaning of the surface by heating to lZO”C, and the characterization at room temperature by XPS, AES and LEED before and after the gaseous exposures as well as after the annealing of the sample to various temperatures. For the SHG measurements, a small UHV chamber (base pressure 5 x lo- lo Torr) equipped with a turbo molecular pump is used. The 532 nm light pluses (20 ps pulse width, 10 Hz) generated from a mode-locked laser are incident at 46 o onto a C(113) crystal (3 x 3 x 0.25 mm”). The reflected second-harmonic signal at 266 nm is detected after proper filtering by a photomultiplier tube and processed by photon counting electronics and a computer. After the sample is installed in the UHV, a sharp LEED pattern can be obtained right away even before the bake-out of the system. The sample “as is” exhibits a (1 x 1) structure and the AES shows the typical fully hydrogenated spectrum along with a small oxygen signal, but no other contaminants. When it is heated to 1250°C LEED exhibits a (2 X 1) (and/or (2 X 2) as noted above) pattern, as displayed in fig. la. AES reveals a characteristic spectrum (fig. 2a) which is quite stable under the e-beam used in AES. In AES and LEED, the e-beam is directed at about 1 mm from the edge of the crystal held by a MO cup. No significant charging or severe surface damage (further discussed below) by the e-beam is observed under the experimental condition. The XPS spectrum shows a single symmetrical C(ls) peak at 285 eV. The chemical shifts of this core level peak due to D and CH, adsorption are apparently too small to be resolved by the spec-

Y. ~its~a

et al. f

I~teracti5~

of D

and CH,

species withthe COll) surfae

Fig. 1. LEED patterns (E, = 172 eV) from the C(111) surface exposed to D atoms: (a) bare surface, (2 x 1); (b) 0 s 0.05 ML, (1 x 1); (c) B = 1 ML, (1 x l);(d) sample (c) annealed to 1380 K, (1 X 1); (e) annealed to 1420 K, (2 x 1).

trometer. Also, XPS (with an Al anode) is not particularly surface sensitive because of the long escape depth of the photoelectrons through the diamond crystal. In the XPS study by Belton et al. [14] of the CVD diamond on top of a Si substrate, the surface species, apart from the Sic and the diamond layers, could not be detected either. Employment of grazing angle detection and soft X-ray photons (e.g., synchrotron radiation) in the 350-450 eV range in the future should enhance the surface sensitivity. When a bare (2 X 1) surface is exposed to D atoms at a dosage sufficient to yield a coverage of 0.05 ML or less, the structure starts to transform rapidly to 0 x 1). The phase transition is evident from both LEED (fig. lb) and AES (fig. 2b). The surface coverage (0) is determined from TDS measurement (presented below). The changes in the AES (the first derivative) spectra are very striking: (1) the shift of the characteristic peak

minimum from 272.5 to 269.0 eV, (2) the appearance of new features at 240.0, 252.5 and 277.9 eV, and (3) the increase in the peak height ratio between the 256.0 eV and the 248.0 eV peaks. At a higher coverage, the peak minimum shifts further down in energy, e.g., to 268.0 eV at 8 = 0.5 ML. At the saturation coverage (1 ML), the peak minimum settles at 267.5 eV (fig. 2~1, i.e., 5 eV below that of the bare (2 X 1) surface. The corresponding LEED pattern is depicted in fig. Ic; The most prominent features in the AES speci trum of the fully deuterated (1 X 1) surface are the positive peaks at 240.0, 246.5, 251.5 and 254.5 eV, a shoulder at 259.5 eV and the negative peaks at 267.5 and 277.5 eV. The KLL Auger transitions involve 1 K-core level and 2 valence’ electrons. The observed fine features are directly related to the 2s-2p chemical bonding configuration [17]. The escape depth of the Auger elecJ trons near 270 eV is less than 10 A in diamond

I’. Mitsuda et al. / Interactions of D and CH, species with the Ccl1 1) surfuce

267.5 ,,,L.i, 230

270

250

290

Kinetic Energy (eV) Fig. 2. AES spectra of the C(111) surface exposed to D atoms: (a) bare surface; (b) 0 I 0.05 ML; (c) 8 = 1 ML; (d) sample (cl annealed to 1330 K;(e) annealed to 1380 K, (f) annealed to 1-42rlc

[9]. Thus, AES is very sensitive to surface chernical and structural changes. It should be noted that prior work with AES on diamond [9,18-201 did not detect such fine features possibly because of surface contaminants, inadequate annealing of the crystal, severe e-beam damage and/or low spectral resolution. We have examined the e-beam irradiation effect on both the bare reconstructed and the fully deteurated surfaces. Under the normal condition to obtain the AES spectrum (2 kV, - 2 PA for 4 min), e-beam does not have any effect on the bare surface. This is confirmed by translating the sample during the spectral scan.

The deuterated surface is slightly affected by the e-beam with the 251.5 eV peak being the most sensitive to the irradiation. With the e-beam on the same spot for 10 min, the peak height of this feature decreases by about 20%. Even with an irradiation time longer than 30 min, however, the peak structure can still be well recognized. To date, this and other features at 240.0, 259.5 and 277.5 eV have never been observed so clearly. These features resemble the X-AES spectrum obtained by Mizokawa et al. [21] with an X-ray source on an air-exposed diamond surface which undoubtedly was fully hydrogenated. X-AES in principle is less damaging to the crystal. The signal is, however, much weaker and it takes a longer time to acquire a good spectrum. The D-exposed surface at 0 = 1 ML is annealed to various temperatures. Below 1200 K, neither the AES nor the LEED shows any change. At about 1300 K, the AES peak minimum shifts upwards to 268.5 eV (fig. 2d) and at 1380 K, the features at 253.0 and 278.5 eV decrease substantially in intensity (fig. 2e). The LEED pattern is still essentially (1 X 1) (fig. Id). At around 1420 K, the surface reconstructs completely to (2 X 11 (figs. le and 2f). The phase transition between the two structures is completely reversible and the LEED pattern does not show any degradation due to the repetitive D exposures and the heating cycles. The thermal desorption behavior of the Dcovered surface is investigated by TDS. At a heating rate of about 40 K/s, the desorption signal (detected mainly as D2) is observed between 1150 and 1420 K with the maximum desorption rate near 1300 K. The TDS spectra are displayed in fig. 3. The D, desorption signal is proportional to 8 and increases with the exposure of D atoms up to a saturation level beyond which the signal is independent of the exposure (top two curves of fig. 31. The surface coverage at the saturation is designated as 8 = 1 ML. This is consistent with the attenuation of the C(ls) XPS signal due to the chemisorption of D atoms. The behavior is in great contrast to the C(1001 surface exposed to H atoms. According to Hamza et al. [6], the H, desorption flux was never observed to saturate in the H/C(lOO) system and the diffu-

Y. Mitsuda et al. / Interactions of D and CH, species with the C(Il1)

I

I

I

I

C(11 1 l/D

1

I

I

I

I

1200

1300

1400

15oc

Temperature

(K)

Fig. 3. TDS spectra (Da mass signals) of C(111) exposed to D atoms: (bottom curves) at 0 I 0.05 ML, (top curves) at saturation coverage (1 Ml) represented by two different exposures with the dashed curve 10 times higher in dosage than the solid curve.

sion of H atoms into the subsurface region was inferred. This apparently does not occur on the C(111) surface. The signal at saturation coverage can be used to determine the value of 0 under the various exposure conditions. At an exposure equivalent to 0 = 0.05 ML or less (bottom curves of fig. 3), a definite change from (2 X 1) to (1 X 1) is observed by AES and LEED. At an initial coverage of 0 = 0.5 or 1 ML, the (1 x 1) structure persists when heated even up to 1380 K. At this temperature only about 10% of the initially adsorbed D atoms remain on the surface. The surface can reconstruct only by heating to 1420 K or higher when all D atoms are desorbed. Clearly a few percent of a ML of D atoms on the surface is sufficient to hold the (1 X 1) pattern. In the prior study by Hamza et al. 171, it was reported that annealing after hydrogen desorption was required for the complete (2 x 1) reconstruction in the H-C(111) system. The structural transformation from (2 X 1) to (1 X 1) due to H adsorption is also detected by SHG. Fig. 4 shows the typical SHG signal as a function of dosing time when a bare, well-annealed C(111) surface is exposed to H atoms. A sudden increase followed by a rapid drop in the

surface

SHG intensity is clearly observed in the initial phase of the exposure (less than 10% of the saturation dosage). It can be correlated with the initial H adsorption on the (2 x 1) surface followed by the induced phase transition to (1 x 1). The SHG signal then increases with dosing and reaches a saturation level. This is associated with further H chemisorption on the (1 X 1) surface. We have also employed a sum-frequency-generation (SFG) technique for surface vibrational spectroscopy [15] to probe the surface along with the SHG measurements. The observed vibrational mode can be identified as the C-H stretch mode for H adsorbed on the top sites of the C(lll)(1 X 1). The spectral intensity can also be used for the calibration of H coverage. Details of the SFG study will be reported elsewhere [22]. It is remarkable that less than 0.05 ML of D atoms can induce the (2 X 1) to (1 X 1) phase transition and maintain the (1 x l)-sp3 structure with 95% (or higher) of the surface bonds dangling even at 1380 K. It is well known that alkali metals can induce surface reconstruction of metal substrates at a few percent of a monolayer coverage [231. Adsorption of hydrogen and oxygen on metal substrates can also cause phase transformation [24], but the surface coverages required for the reconstruction are usually higher than the alkali overlayers and the D-C(111) system reported here. Unlike the adsorbate-metal systems where charge transfer and long range interactions may operate, the D-C(l11) bonding is covalent and completely localized. Thus the ability of the

I

L

I

I

I

I

0 Dosing

I

I

I

I

20 Time (min)

I

40

Fig. 4. SHG signal (polarization, p-in and p-out) as a function of H dosing time with H, gas maintained at 6 x lo6 Torr through a hot W-filament.

Y Mitsuda et al. / Interactions

of D and CH, species with the C(llli

sparsely adsorbed D atoms to induce the drastic phase change is rather surprising. Of the various structures considered for the C(lll)-(2 x 1) surface, the r-bonded-chain model proposed by Pandey [25,26] has been found to be the most energetically favorable by ab initio LCAO (linear combination of atomic orbitalsl computations [27]. In this picture, the C atoms in the top rows have sp* configuration and form a chain of P-bonds on the surface. Since a D (or H) atom tends to bond only with one C atom regardless of coverage, it is conceivable that a small amount of such atoms adsorbed on a chain can cause some kind of “unzipping” of the entire chain to a (1 X 1) structure once the r-bonded linkage is ruptured in certain places. Furthermore, at the substrate temperature near 500°C during dosing, or higher in the annealing steps,

surface

the adsorbed D atoms may be quite mobile on the surface. The mobility may help the small amount of D to uphold the (1 x 1) structure. The (2 X 1) reconstruction occurs only when all D atoms are desorbed. This is one way to interpret the present result. The exposure of a bare C(lll)-(2 X 1) surface to the CH, gas through the hot W filaments produces a very different result from that of D atoms. At a surface coverage (e.g., 0 5 0.1 ML) where D atoms can readily transform (2 x 1) into (1 X l), the (2 X 1) structure persists in the case of CH, exposure (fig. 5b). The AES spectrum exhibits a small shift of the peak minimum to a lower energy along with small changes in the spectral features. Even at a coverage estimated at 13= 0.3-0.5 ML, the (2 x 1) pattern is still clearly visible although it is weakened substantially (fig.

Fig. 5. LEED patterns (E, = 172 eV) from C(111) exposed to CH, through hot W filements; (a) bare surface, (2 x 1); (b) 0 = 0.1 ML, (2 X 1); (c) fJ z 0.3-0.5 ML, weak (2 X 1); (d) 0 = 1 ML, (1 x 1); (e) sample (d) annealed to 1275 K, (1 x 1) and weak (2 x 1); (f) annealed to 1475 K, (2 x 1).

Y. Mitsuda et al. / Interactions ofD and CHz species with the C(f 1I) surface I

I

I

I

I

I

C(1 11 KHx

I 230

I 250

I

272.5

II

I

270

I 290

Kinetic Energy (eVf Fig. 6. AES spectra of C(111) exposed to CH, through hot W filaments: (a) bare surface; (b) 0 = 1 ML; (c) sample (b) annealed to 1175 K; (d) annealed to 1275 IQ (e) annealed to 1475 K; (0 after 4 cycles of dosing and heating treatments.

5~). In this case the approximate value of 8 is estimated from the amount of gaseous exposure and the shift in the AES peak minimum. The results here again correlate very well with the SHG signal variation with dosing. At an exposure of 8 X 10e5 Torr for 10 min of CH, through the filaments, the (1 X 1) pattern is observed by LEED (fig. 5d). The AEZS spectrum shows the peak minimum at 270.0 eV (fig. 6b). Further exposures (up to 3 x 10e4 Torr, 10 min) do not cause much change in either the LEED pattern or the AES spectrum. The coverage at this exposure is about 1 ML and the hydrocarbon species

on the surface is designated as CH, with 1 I n 1s 3. The surface species is certainly not graphitic or amorphous carbon and surely not CH, because the molecule simply does not stick on diamond at these temperatures. In comparing fig. 6b (CH,, B = 1 ML) with fig. 2e (D, 6 = 1 ML), we find certain similarities but also major differences in the AES fine structures. The most notable distinctions are the magnitudes of the peak shifts, the relative peak heights and the absence of a resolvable feature at 277.5 eV on the CH,-covered surface. The surface contains very littIe, if any, graphitic (or amo~hous) carbon. The amount of H atoms coadsorbed on the surface is smaller than CHX, even though under the exposure condition, H atoms are also produced from the cracking of CI-I, by the W-filaments at 2000°C. When the CH,-covered surface is subsequently exposed to D atoms, a change in the AES spectrum from fig. 6b to fig. 2b and then to fig. 2c is readiry observed. Apparently D atoms can react with the adsorbed CH, species and cause the surface to be more highly deuterated. In the exposure of CH,-hot filament alone, CHX species evidently has a higher sticking probability to the (2 x 1) surface than the H atoms which are also present. When the CH,-covered (1 X 1) surface at 6 = 1 ML is annealed to 1075 K or lower, the B spectrum shows little change. Upon annealing to 1175 K, subtle changes in AES are detected (fig 6~) while the LEED pattern is still (1 x 1). At 1275 K, a weak (2 x 1) structure begins to emerge (fig. Se) and the peak minimum of the AES spectrum shifts to 271.5 eV (fig. 6d). Heating to 1475 K causes the surface to reconstruct completely to (2 x 1) (fig. 5f) and the adsorbed CHX to partially convert into graphitic (or amorphous) carbon. This is evident from the AES spectrum (fig. 6e) in which the peak height at 250.0 eV increases with respect to that at 260.0 eV. Upon repeated cycles of dosing and heating to 1475 K, the amount of graphitic carbon on the surface increases. Fig. 6f shows the AES spectrum after 4 cycles of such treatments. The characteristic graphitic feature at 250.0 eV is clearly evident. The surface graphite so produced can be reduced or eliminated (depending on the coverage) by

Y Mitsuda et al. / Interactions of D and CH, species with the C(lI1) surface

repetitive exposure to H (or D) atoms and heating to 1475 K. It is observed that even with the amount of graphite slightly higher than 1 ML (e.g., fig. 6f), the (2 x 1) surface reconstruction by heating to 1475 K and the phase transition to (1 x 1) by exposure to D atoms can proceed without much effect by the presence of the graphite. It is also clear that when the CHX-covered surface is heated to 1275 K or higher, some surface species desorb. The chemical identities of the desorbed species have not yet been determined. These may be CH, and H, produced by recombination reactions on the surface in the thermal desorption process. The results show that both the adsorption and the desorption of CH, on the C(111) surface behave very differently from those of the D atoms. At an equivalent coverage in the dosing stage, CH, species are incapable of converting (2 x 1) to (1 x 1) as readily as the D atoms; Likewise, CH, cannot hold the (1 X 1) structure as well as the D in the annealing and desorption phase. The bonding characteristics of CH, to the C(111) surface must be fundamentally different from those of D (or H). In the Pandey model of the (2 X 1) surface, it is conceivable that at a low to medium coverage (e.g., f3< 0.3 ML), CH, may bond onto the r-bonded-chain in a manner that allows the chain in the bare regions to maintain its r-bonded character. A possibility that comes to mind is where a CH, attaches itself at a double bond to form a cycle-propane-like structure with 2 adjacent carbons in the chain. According to the calculations by Vanderbilt and Louie [271, the C-C bond of the relaxed chain is contracted by 4.4% from the bulk C-C distance. In contrast to the hydrogen termination, the cyclopropane-like structure may help maintain this tighter bond and stabilize the chain. The CH, bonded in this manner is likely to be relatively immobile and may hinder the unzipping of the entire chain. At a higher coverage (e.g., 0 2 0.5 ML), lateral interactions may force the adsorbed CH, to bond with only one surface carbon and eventually cause the surface to become (1 x 1). The ability of H (or D) atoms to uphold the (1 X 1) structure even at small surface coverages may be strongly related to the role H atoms play

in the diamond CVD. In most CVD reactions, a large quantity of H, with respect to the hydrocarbon gas such as CH, is used. Under the exposure to a plasma or hot-filament, a relatively high concentration of H atoms is expected. Even in the diamond CVD with a hydrocarbon feed gas without H,, such as a pure CH,OH or CH,OH plus Ar [28], a substantial concentration of H atoms can be generated. These H atoms can maintain the diamond surface in the (1 x 1) structure and thus promote the growth in the sp” configuration. In addition, H atoms can react with graphitic carbon to reduce sp2 contaminants during the CVD film growth. The current study leaves some important questions unanswered such as the identity of the adsorbed CH, species and the desorption products in the thermal annealing process. The origin of the CH, species whether they are generated in the gas phase by the hot filaments or from the excited CH, molecules colliding with the diamond substrate is also unclear. We are using SFG and other spectroscopic methods to resolve these and related questions. The authors wish to thank J. Goitia for technical assistance, G.D. Kubiak for suggestions concerning the hot filament, H.F. Winters for helpful discussions, and M.R. Philpott and J.W. Coburn for encouragement and support. R. Chin acknowledges the support of a 1987 Science and Engineering Scholarship from the Natural Sciences & Engineering Research Council of Canada. J.Y. Huang acknowledeges the support of a postdoctoral fellowship from IBM. The work of the Berkeley group was supported by the Director, Office of Energy Research, Office of Basic Energy Sciences, Materials Sciences Division of the US Department of Energy under contract No. DE-AC03-765F00098.

References

111See recent reviews, e.g., J.C. Angus and C.C. Hayman, Science 241 (1988) 913; A.R. Badzian and R.C. DeVries, Mater. Res. Bull. 23 (1988) 385; D.V. Fedoseev, V.P. Varnin and B.V. Deryagin, Russian Chem. Rev. 53 (1984) 435.

@I C.J. Chu, M.P. D’Evelyn, R.H. Hauge and J.L. Margrave, J. Mater. Res. 5 (1990) 2405. [31 S.J. Harris and L.R. Martin, J. Mater. Res. 5 (1990) 2313; S.J. Harris and A.M. Weiner, J. Appl. Phys. 67 (1990) 6520; S.J. Harris, Appl. Phys. Lett. 56 (1990) 2298. [41 L.R. Martin and M.W. Hill, Appl. Phys. L&t, 55 (1989) 2248; J. Mater. Sci. Lett. 9 (19QO)621. 151 F.G. Celii, P.E. Pehrsson, H.-t. Wang and J.E. Butler, Appl. Phys. Lett. 52 (1988) 2043; F.G. Celii and J.E. Butler, Appl. Phys. L&t. 54 (1989) 1031; F.G. Celii, P.E. Pehrsson, H.-t. Wang, H.H. Nelson and J.E. Butler, in: Advances in Laser Science IV, Eds. W.C. S&valley and J. Gale, AIP Conf. Proc. No, 191(1989) 747. 161A.V. Hamza, G.D. Kubiak and R.H. Stulen, Surf. Sci. 237 (1990) 35. [71 A.V. Hamza, G.D. Kubiak and R.H. Stulen, Surf. Sci. 206 (1988) LS33. [81 H. Namba, M. Masuda and H. Kuroda, Appl. Surf. Sci. 33/34 (1988) 187. [91 B.B. Pate, Surf. Sci. 165 (1986) 83; B.B. Pate, M. Oshima, J.A. Silberman, G. Rossi, I. Lindau and W.E. Spicer, J. Vat. Sci. Technol. A 2 (1984) 957. 1101J.E. Lowther, Solid State ~mmun. 56 (1985) 243. ml B.J. Wadawski, D.T. Pierce, N. Swanson and R.J. Celotta, J. Vat. Sci. Technol. 21 (1982) 368. WI S.P. Mehandru and A.B. Anderson, J. Mater. Res. 5 (1990) 2286. R31 S.J. Harris, A.M. Weiner and T.A. Perry, Appl. Phys. Lett. 53 (1988) 1605. D41 D.N. Belton, S.J. Harris, S.J. Schmieg, A.M. Weiner and T.A. Perry, Appl. Phys. Lett. 54 (1989) 416.

1151 Y.R. Shen, Nature 337 (1989) 519, and references therein. [16] T.J. Chuang, Surf. Sci. Rep. 3 (1983) 1; T.J. Cbuang, H. Seki and I. Hussla, Surf. Sci. 158 (1985) 525. [17] T.W. Haas, J.T. Grant and G.J. Dooley III, J. Appl. Phys. 43 (1972) 1853. [18] S.V. Pepper, Appl. Phys. Lett. 38 (1981) 344; M. Dayan and S.V. Pepper, Surf. Sci. 138 (1984) 549. 1191T.J. Moravec and T.W. Orent, J. Vat. Sci. Technol. 18 (1981) 226. [20] P.G. Lurie and J.M. Wilson, Surf. Sci. 65 (1977) 453; 65 (1977) 476; 65 (1977) 499. [21] Y. Mizokawa, T. Miyasato, S. Nakamura, K.M. Geib and C.W. Wilmsen, Surf. Sci. 182 (1987) 431. [22] R. Chin, J.Y. Huang, Y.R. Shen, T.J. Chuang, H. Seki and M. Buck (to be published). [23] See, e.g., C.J. Barnes, M. Lindroos, D.J. Holmes and D.A. King, Surf Sci. 219 (1989) 143; P. H[iiberle, P. Fenter and T. Gustafsson, Phys. Rev. B 39 (1989) 5810; WC. Fan and A. Ignatiev, Phys. Rev. B 38 (1988) 366. [24] See, e.g., D.A. King, Phys. Ser. T 4 (1983) 34; A.H. Smith, R.A. Barber and P.J. Estrup, Surf. Sci. 136 (1984) 327. [25f G.D. Kubiak and K.W. Kolasinski, Phys. Rev. B 39 (1989) 1381. 1261 K.C. Pandey, Phys. Rev. B 25 (1982) 4338. [27] D. Vanderbilt and S.G. Louie, Phys. Rev. B 30 (1984) 6118. [28] M. Buck, T.J. Chuang, J.H. Kaufman and H. Seki, Mater. Res. Sot. Symp. Proc. 162 (1990) 97.