Valence-band spectra of hydrogenated diamond (111) surface

Valence-band spectra of hydrogenated diamond (111) surface

fB!AMOND RELATED MATERIALS Diamond Valence-band Kazuo Yamamoto, National and Related Materials 4 (1995) 520-523 spectra of hydrogenated Shigeru ...

426KB Sizes 0 Downloads 2 Views

fB!AMOND RELATED MATERIALS Diamond

Valence-band Kazuo Yamamoto, National

and Related

Materials

4 (1995) 520-523

spectra of hydrogenated

Shigeru Suehara, Institute,for

Research

diamond

Toshihiro Ando, Shunichi Yoichiro Sato

in Inorganic

Materials,

1-l Namiki,

(

111) surface

Hishita, Mutsukazu

Tsukuba,

Ibaraki

Kamo,

305, Japan

Abstract Boron-doped semiconducting diamond thin films were epitaxially grown on a natural diamond ( 111) single crystal by microwaveplasma-assisted chemical vapor deposition. We investigated chemisorption of hydrogen on the diamond surface by X-ray photoemission spectroscopy measurements and first-principles molecular orbital calculations. By comparison between the experimental and calculated spectra, we conclude that hydrogen termination of the epitaxially grown diamond (Ill) surface is in the form of the CH, group. Keywords:

Spectroscopy;

Surface characterization;

Homoepitaxy;

1. Introduction Despite a widespread interest in chemisorption of hydrogen on diamond thin films grown by chemical vapor deposition (CVD), including the surface reaction of a hydrogen-hydrocarbon mixture, there appears to be a shortage of information on it which would lead to elucidation of the growth mechanism of diamond. Hydrogen is thought to play important roles in the growth process, including maintenance of sp3 hybridization at the growth surface. Although it has been reported that the hydrogen-terminated diamond surfaces have a ( 1 x 1) structure for a ( 111) surface [l-lo] and a (2 x 1) structure for a (100) surface [ 1 l-131, atomic arrangements at the surface and/or chemisorbed species on the surface remain uncertain. Recently, high-resolution electron-energy-loss spectroscopy (HREELS) experiments clarified that the epitaxially grown diamond( 11 l)-(1 x 1) and (lOO)-(2 x 1) surfaces were terminated by well-ordered methyl (CH,) and monohydride (CH) groups respectively [ 141. As for the (111) surface, although the CH, group termination of the as-polished (1 x 1) surface had also been reported by HREELS [3], scanning tunneling microscopy (STM) observations of CVD diamond (111) surfaces [ 151 and He scattering and diffraction data [8,9] and electronstimulated desorption ion angular distributions (ESDIAD) [ 51 from as-polished diamond ( 111) surfaces favored a simple hydrogen termination of dangling bonds along the surface normal. 0925-9635/95/$09.50 0 1995 Elsevier Science S.A. All rights reserved SSDl 0925-9635(94)05326-X

Microwave

plasma CVD

There is a similar controversy in some theoretical studies on the chemisorption of hydrocarbons onto the ( 111) surface. Gradient-corrected local spin-density functional cluster calculation [ 161, ab initio molecularorbital calculation including electron correlation treatments [ 171 and semi-empirical atom superposition and electron delocalization molecular orbital (ASED-MO) calculations [IS] showed that the CH, termination is less stable than the H termination because of large steric repulsions between adjacent hydrogen atoms. In contrast semi-empirical slab modified intermediate neglect of differential overlap (slab-MINDO) calculation revealed that the CH, termination is more favorable than the H termination when the large relaxation is taken into account [ 191. Valence-band photoemission spectroscopies provide localized probes for chemical bonding states in the region of solid surfaces. In this study we investigate chemisorbed states of hydrogen on CVD diamond (111) surfaces by comparison between experimental and theoretical valence-band spectra.

2. Experiment

and calculations

Natural Ia-type diamond single crystals, which were polished to within 4” of the (111) direction, were used for substrates. Before being loaded into the deposition chamber, the diamond substrates were rinsed in a heated H,SO,-HNO, solution in order to remove metallic and

521

K. Yamamoto et al.lDiamond and Related Materials 4 (1995) 526523

graphitic contaminants. Boron-doped semiconducting diamond epitaxial layers were grown on the substrates by the microwave-assisted CVD method [20]. The addition of boron-doped layers contributes to suppressing the charging effect during the photoemission processes. Gaseous mixtures of methane (l.O%), diborane (0.8 ppm) and hydrogen were fed at the rate of 50 ml mini and the pressure was kept at 40 Torr. The supplying microwave power was about 300 W. The substrate temperature was kept at 770-800 “C. The above growth conditions are almost the same as those in a previous paper [ 141. XPS measurements were performed using a Vacuum Generators ESCALAB 200 spectrometer with a base pressure of the order of lo-’ Pa. Spectra were recorded using a Mg Kcc source with an instrumental resolution of 1.0 eV. After deposition of diamond by the CVD method, the samples were transferred to the ultrahigh vacuum chamber through the atmosphere. Ordinary cleaning methods such as Ar ion sputtering and hightemperature annealing were not used in this study. Mild annealing (200-300 “C) was carried out before the XPS measurements in order to remove contaminants such as physisorbed hydrocarbons, oxygen and water. In fact, it was confirmed that the XPS oxygen 1s peak is negligibly small. Theoretical calculations were performed with the use of the self-consistent-charge discrete variational Xcc (SCC-DV-Xa) method, details of which have been reported elsewhere [21]. In this method, the HartreeFock-Slater (HFS) equation for a cluster model is selfconsistently solved with the use of a localized exchange potential (Xa potential). The adjustable (exchangecorrelation) parameter a was taken to be 0.7 as usual. Numerical carbon lss2p and hydrogen 1s atomic orbitals, which were obtained as solutions of the atomic HFS equations, were used as basis sets.

3. Results and discussion Fig. l(a) shows an XPS valence-band spectrum of a diamond (111) surface epitaxially grown by the CVD method. This spectrum is given by subtracting a nonlinear background from a measured raw spectrum [22]. The energy scale is referenced from the valence-band maximum (=O). The spectrum shows the following features: (A) a very broad and weak structure from 0 to - 10 eV; (B) a narrower, intense peak located at - 11 eV; (C) a fairly broad, less intense peak located at - 16 eV; and (D) a weak peak between the above intense peaks, which is located at - 13 eV. Next, we have calculated the XPS valence-band spectra of six cluster models shown in Fig. 2, and compared them with the experimental spectrum. These cluster models are composed of a substrate part and an

(g) I

I

N

/ -25

I

-20

I -15

Energy

I -10

I -5

I

0

5

( eV )

Fig. 1. Experimental XPS valence-band spectrum of hydrogenated diamond (111) surfaces after removal of a nonlinear background (a) compared with calculated spectra of the six models (b)-(g): (b) the optimized fully CH,-covered surface model (model 2); (c) the fully CH,-covered surface model (model 3); (d) the fully H-covered surface model (model 4); (e) the intermixing CH,- and H-covered (H-rich) model (model 5); (f) the intermixing CH,- and H-covered (CH,-rich) model (model 6); (g) the bare surface model (model 1). In (a) the experimental raw data (dots) are smoothed to yield a solid line [23]. Solid and broken lines shown in (b)-(g) are convoluted spectra with 2.0 eV and 1.0 eV FWHM gaussian functions respectively.

adsorbate

part. The substrate is represented by a threecluster model (Fig. 2(a)). The bonds of layer Ci,H,i carbon atoms at the edges of the substrate part are saturated with hydrogen atoms so as not to leave any unsaturated dangling bonds. The CC and the C-H bond lengths are 1.545 A and 1.107 A respectively, which are extracted from data for bulk diamond and methane. The surface layer is assumed to be unrelaxed and unreconstructed. Models 2 and 3 (Fig. 2(b)) represent a fully CH,-covered diamond ( 111) surface, in which each CH, group is rotated 30” about the surface normal axis passing through the methyl C atom. In model 2 we used the optimized coordinates for the CH, adsorbates by Zheng and Smith [ 191. Model 4 represents a fully H-covered diamond ( 111) surface. Models 5 and 6 simulate an intermixture of CH,- and H- terminations. For comparison, we calculated a bare diamond (lll)(1 x 1) surface model (Model 1 in Fig. 2(a)). Figs. l( b)-1 (g) show the calculated spectra. The calculated spectra were obtained as follows. We have first calculated local densities of states (LDOS) of the atoms located at

K. Yamumoto

522

et al.!Diumond

und Related Matrriuls

(a>

(d)

H

(6

Fig. 2. Schematic illustrations of the hydrogenated cluster models: (a) model 1; (b) models 2 and (d) model 5; (e) model 6.

diamond ( I1 1) 3; (c) model 4;

the center in the cluster models (marked with C(a) and H(a) for adsorbate atoms and C(s1) and C(s2) for substrate atoms in Fig. 2) instead of the total DOS, in order to remove effects of the edges of the cluster models. Line spectra are calculated as the product of the occupied C 2s, C 2p, and H 1s LDOS and the corresponding subshell photoionization cross-sections from [ 241. The line spectra were then convoluted with a gaussian function with 1.0 or 2.0 eV full width at half-maximum height

4 i lW5j

520-523

(FWHM) to facilitate the comparison with the experimental spectrum. The above method using LDOS has successfully reproduced experimental photoemission spectra [ 25,261. As can be seen in Fig. 1, the overall features of the calculated spectrum from model 2 (the optimized fully CH,-covered surface model) agree well with the experimental one except that the width of the calculated band marked B, C and D is narrower than that of the experimental one. The experimental spectrum is not reproduced by the calculations for the other models. Although the spectrum of model 3 (the non-optimized fully CH,-covered surface model) is similar to that of model 2, the height of the peak corresponding to C is lower than that corresponding to D. As shown in Fig. l(d), a peak at -6 eV of the fully H-covered model (model 4) is stronger as compared with the experimental one. Since the photoionization cross-section of C2s is one order larger than that of C2p, this peak, which is mainly composed of C(s2) 2s character, is remarkable. In addition, a surface level associated with dangling bonds emerges at 3 eV for the case of the bare diamond ( 11 l)-( 1 x 1) model (Fig. l(g)). Surface states on the diamond (ill)-(2 x 1) surface have been characterized with the use of angle-resolved photoemission [27] and theoretical band calculations [28,29]. The fact that the surface states were not observed in our experiments shows that the CVD-grown diamond surface is terminated with some chemisorbed species. Therefore, it is concluded that the homoepitaxially grown diamond ( 111) surface is covered with the CH, group. This result is consistent with the HREELS experiments [ 141. The inconsistency with the other experiments may be due to a difference in the growth conditions of those specimens. Fig. 3 shows LDOS for model 2. The LDOS analysis reveals that a broad and weak structure from 0 to - 10 eV is mainly composed of C2p states with slight admixtures of C2s and Hls states. The main intense peaks between - 10 eV and - 15 eV are made up of hybridized orbitals having predominent C2s character. Finally, in XPS measurements using Mg Kr irradiation, it is difficult to detect hydrogen because the photoionization cross-section of Hls is extremely small. Comparison between experimental and calculated chemivalence-band spectra is useful for investigating sorbed states of hydrogen on diamond surfaces.

4. Conclusion We have measured X-ray photoemission spectra of homoepitaxially grown diamond ( 11 I) by a plasma CVD method and calculated spectra for various hydrogenated diamond ( 111) models by the DV-Xa method. From the overall agreement between the experimental and calculated spectra, we have concluded that the ( 111) surface

K. Yamamoto

et al./Diamond

and Related Materials

Fig. 3. Calculated valence-band spectrum and line spectra of LDOS of model 2. The former spectrum, which is the same as that shown in Fig. l(b), is obtained by convolution of the line spectra with a 1.0 eV FWHM gaussian function.

is terminated by CH, groups, which are relaxed in order to reduce large steric repulsions between adjacent hydrogen atoms.

Acknowledgements We wish to thank Professor H. Adachi (Kyoto University) for use of the DV-Xc( calculation program. We are indebted to S.C. Lawson (National Institute for Research in Inorganic Materials, NIRIM) for a critical reading of this manuscript and for valuable comments. Special thanks go to T. Aizawa and T. Hatano (NIRIM) for fruitful conversations, support and encouragement through this work.

References [l] [2]

J.J. Lander and J. Morrison, Surf. Sci., 4 (1966) 241 B.B. Pate, Surf. Sci., 165 (1986) 83.

4 (1995) 520-523

523

D.T. Pierce, N. Swanson and R.J. Celotta, c31 B.J. Waclawski, J. Vat. Sci. Technol., 21 (1982) 368. c41 B.B. Pate, M.H. Hecht, C. Binns, I. Lindau and W.E. Spicer, J. Vat. Sci. Technol., 21 (1982) 364. and R.H. Stulen, Surf. Sci., 206 c51 A.V. Hamza, G.D. Kubiak (1988) L833. C61 Y. Mitsuda, T. Yamada, T.J. Chuang, H. Seki, R.P. Chin, J.Y. Huang and Y.R. Shen, Surf. Sci. Lett., 257 (1991) L633. c71 R.P. Chin, J.Y. Huang, Y.R. Shen, T.J. Chuang, H. Seki and M. Buck, Phys. Rev. B, 45 (1992) 1522. C81 G. Vidali and R.F. Frankl, Phys. Reo. B, 27 (1983) 2480. c91 G. Vidali, M.W. Cole, W.H. Weinberg and W.A. Steele, Phys. Rev. Lett., 51 (1983) 118. Cl01 S.-Tong Lee and G. Apai, Phys. Rev. B, 48 (1993) 2684. K. Hamada and N. Fujimori, Cl11 T. Tsuno, T. Imai, Y. Nishibayashi, Jpn. J. Appl. Phys., 30 (1991) 1063. and R.H. Stulen, Surf. Sci., 237 Cl21 A.V. Hamza, G.D. Kubiak (1990) 35. R.A. Rudder, R.J. Markunas, D. Huang and Cl31 R.E. Thomas, M. Frenklach, J. Chem. Vapor Deposition, 1 (1992) 6. Cl41 T. Aizawa, T. Ando, M. Kamo and Y. Sato, Phys. Rev. B, 48 (1994) 18348. Jpn. J. Appl. Phys., 32 (1993) Cl51 H. Sasaki and H. Kawarada, L1771. Cl61 M.R. Pederson, K.A. Jackson and W.E. Pickett, Phys. Rev. B, 44 (1991) 3891. S. Lunell and J.-O. Carlsson, Phys. Rev. B, 48 Cl71 K. Larsson, (1993) 2666. and A.B. Anderson, .I. Mater. Rex, 5 (1990) Cl81 S.P. Mehandru 2286. Cl91 X.M. Zheng and P.V. Smith, Surf. Sci., 261 (1992) 394. and N. Setaka, J. Cryst. c201 M. Kamo, Y. Sato, S. Matsumoto Growth, 62 (1983) 642. and C. Satoko, J. Phys. Sot. Jpn., 45 c211 H. Adachi, M. Tsukada (1978) 875. c221 P.M.A. Sherwood, in D. Briggs and M.P. Seah (eds.), Practical Surface Analysis by Auger and X-ray Photoelectron Spectroscopy, Wiley, New York, 1983, Appendix 3. and L. Marmet, Rev. Sci. Instrum., 54 (1983) 1034. ~231 P. Marchand We used the binomial smoothing, which is a gaussian filter. The algorithm is derived from this article. c241 J.H. Scofield, J. Electron Spectrosc. Relat. Phenom., 8 (1976) 129. and E. Miyazaki, J. Chem. Phys., 87 c251 M. Orita, I. Kojima (1987) 4162. S. Hishita and A. Nukui, Phys. Rev. C261 S. Suehara, K. Yamamoto, B, 50 (1994) 7981. ~271 F.J. Himpsel, D.E. Eastman, P. Heimann and J.F. Van der Veen, Phys. Rev. B, 24 (1981) 7270. C281 K.C. Pandy, Phys. Reo. B, 25 (1982) 4338. ~291 D. Vanderbilt and S.G. Louie, Phys. Rev. B, 29 (1984) 7099; 30 (1984) 6118.