Site seectivity in the initial oxidation of the Si 111-7=7 surface

Transcription

1 Applied Surface Science ž / Site seectivity in the initial oxidation of the Si 111-7=7 surface Jeong Sook Ha a,), Kang-Ho Park a, El-Hang Lee a, Seong-Ju Park b a Research Department, Electronics and Telecommunications Research Institute, Taejon, , South Korea b Department of Materials Science and Engineering, Kwangju Institute of Science and Technology, Kwangju, , South Korea Received 8 July 1997; accepted 19 November 1997 Abstract The site selectivity of oxygen in the initial oxidation of the SiŽ =7 surface was investigated using a scanning tunnelling microscope. At room temperature, two different oxygen-induced features, bright and dark sites, were observed indicating two different oxygen chemisorption configurations. In particular, the major dark features were observed preferentially at the center adatom sites of the faulted half of the dimer adatom stacking-fault Ž DAS. structure over corner ones, which is different from previous results. We explain this in terms of the difference in the potential energy curves of Ž two sites combined with the effect of the slow oxygen exposure 1=10 Torr. that we used. At low oxygen pressure, oxygen atoms have sufficient diffusion time to be adsorbed on the most stable sites. This proposal is supported by the pressure dependence of the site selectivity that we measured. Furthermore, when the sample was exposed to oxygen at high temperatures, the dark features still preserved the strong site preference on the center adatom sites supporting both our hypothesis and the previously reported idea that center adatom sites are thermally more stable than the corner ones. q 1998 Elsevier Science B.V. PACS: Ch; Mq; My Keywords: Site selectivity; Oxidation; Si 111-7=7; Scanning tunnelling; Microscope; Thermal stability 1. Introduction The delicate control of the initial oxidation stage of Si becomes more important as modern, very large scale integrated device technology demands that individual device components be made smaller. Up to now, many efforts to understand the mechanisms of initial oxidation have been made using various surw1 14 x. Among these, the face analysis techniques scanning tunnelling microscope Ž STM. has become ) Corresponding author. Tel.: q ; fax: q ; jsha@idea.etri.re.kr. the most powerful tool due to its unprecedented spatial resolution. At room temperature, initial oxidation of SiŽ = 7 surface results in two different surface features. Bright and dark sites were observed as a result of two different chemisorption configurations. These two features show a strong site selectivity to the adatom sites of the faulted half to unfaulted half of the dimer adatom stacking-fault Ž DAS. structure w5 7 x. Especially, the dark features which are the major oxidation product show a site preference of corner adatoms to center adatoms w5,6 x, or no signifiw7 9 x. cant site selectivity r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved. PII S

2 318 ( ) J.S. Ha et al.rapplied Surface Science At high temperatures, it has been known that the surface etching reaction and the oxide film growth compete with each other depending on the oxygen pressure and the surface temperature w10,11 x. However, the site selectivity of oxidation has not been well investigated. Site preference may be changed with temperature since the thermal stability and the chemical reactivity of each reaction site might be different at different temperatures. Therefore, we investigated the site selectivity of oxidation with variation of surface temperature in the initial oxidation of SiŽ = 7 surface by taking the STM images of samples dosed with oxygen at room temperature and 5508C. It was found that both the bright and dark features preferred the faulted half to the unfaulted half at room temperature and 5508C. The bright features, which were only observed at room temperature, showed a strong site selectivity for the corner adatom sites with respect to center ones. This is consistent with previous results w5 7 x. However, the dark sites, which are the major oxidation product, showed a strong preference for the center adatom sites, which has not been reported yet. Furthermore, the strong site selectivity of dark features was preserved on the SiŽ =7 surface exposed to oxygen at 5508C. In this paper, we propose a potential energy curve model combined with the effect of slow oxygen exposure to explain the temperature dependent site selectivity of the oxygen-induced dark features in the initial oxida- tion of the Si 111-7=7 surface. 2. Experimental The experiments were performed in an ultrahigh vacuum Ž UHV. system with a base pressure of 1.5 = 10 y10 Torr, which is equipped with a sample loadlock for fast sample transfer into UHV. The STM used in this experiment was home-built. The samples were cut from B-doped, p-type SiŽ 111. wafer with a resistivity of 1.0 V cm Ž Virginia Semiconductor.. After being installed in vacuum, the samples were thoroughly outgassed overnight at 6508C. A clean and atomically well ordered SiŽ = 7 surface was obtained by repeated heating to 12508C and slow cooling down to room temperature. O2 dosing was done by back filling the chamber with researchgrade O2 at the pressures between 1.0=10 and 5.0=10 y8 Torr. All hot filaments were shot down during the oxygen doses to minimize O2 ionization. When the SiŽ =7 surface was exposed to O 2 at high temperature, the sample was kept at that temperature for 5 min after turning off O2 gas and then cooled to room temperature. All STM measurements were done at room temperature with electrochemically etched tungsten tips. The sample bias voltage was below "3 V and the tunnelling current was 1 na. 3. Results and discussion Fig. 1 shows the STM images taken from the SiŽ =7 surface exposed to O2 at room temperature with variation of O2 exposure time. All STM images were taken with a positive sample bias voltage of 3.0 V. Fig. 1a d shows the surface exposed to Ž y6 0.12, 0.24, 0.48, and 0.96 L 1 Ls1=10 Torr s. of O 2, respectively. However, these images do not show the same area. The faulted and unfaulted halves of the DAS structure were determined by imaging the occupied states of SiŽ =7 surface with a negative sample bias voltage. In the STM images, two different features due to O2 exposure are shown: a few bright and many dark sites. As the oxygen exposure increased, the number of major dark sites increased while that of the bright sites did not seem to change noticeably. From the previous STS, UPS results wx 5 and the tight-binding slab calculations wx 9, it has been found that the oxygen-induced bright adatom sites involve Si adatoms that maintain their dangling bond but are perturbed by a neighboring oxygen. It was suggested that the most plausible configuration involves an oxygen atom inserted in one of the adatom back bonds and causes a bright appearance since the dangling bond is empty due to charge transfer to the oxygen and it is more localized on the adatom. The dark features had been interpreted as an adatom with two oxygen atoms, one on the on-top site and the other in one of the back-bonds. Since oxygen is more electronegative than Si and its adsorption causes a reduction in the state density at energies significantly above the Fermi level, it appears dark in the STM image. However, a recent STM observation showed

3 ( ) J.S. Ha et al.rapplied Surface Science Fig. 1. STM images taken from the SiŽ =7 surface exposed to O at room temperature with variation of O exposure: Ž. a 0.12 L, Ž. 2 2 b Ž. Ž L, c 0.48 L, d 0.96 L. Oxygen partial pressure was 1.0=10 Torr, sample bias voltage was q3.0 V, and the tunneling current was 1.0 na. that the dark features represent a combination of atomic and surface molecular oxygen w12 x. The bright features appeared preferentially at the corner adatom sites of the faulted half to center ones, which is consistent with the previous results w5 7 x. The site preference of oxygen adsorption is expected to be strongly related to the electronic states of the adatoms w13,14 x. It has been known that the faulted half is preferred to the unfaulted half, and the corner adatom site is preferred to the center one since the adsorption starts in the adatom site with higher electronic states and the occupation of corner adatom dangling bond states is higher than that of center adatoms due to weaker charge transfer from the former sites to rest atoms. The dark features also appeared to prefer the adatom sites of the faulted halves to those of the unfaulted halves. Such site selectivity becomes much clearer as the surface oxygen coverage increases, i.e., the unfaulted halves stand out in the image since most dark sites appear in the faulted halves: in Fig. 1d, dark triangles Ž faulted half. and bright reciprocal triangles Ž unfaulted half. are clearly seen. This is consistent with the previous work w5,6,8 x. However, these dark sites strongly preferred the center adatom sites to corner adatom sites by about 2 : 1 in contrast to the previous results w5,6,8,9 x: strong preference to corner adatom sites w5,6x and no significant site selectivity w7 9 x. We have performed several successive scans over the same surface area and found that there was no change in the site selectivity as time evolves. This different site selectivity of dark features from the previous results cannot be explained simply by the electronic states of the adatoms as in the case of bright sites. We consider that the preference of dark features for the center adatom sites is attributed to the slow Ž oxygen exposure 1 = 10 Torr. used in this experiment, which could have given oxygen atoms sufficient diffusion time to be adsorbed on the most stable sites, the center adatom sites wx 7, on the surface. To obtain more experimental evidence for this suggestion, we have measured the site selectivity of the dark features with variation of oxygen partial

4 320 ( ) J.S. Ha et al.rapplied Surface Science Fig. 2. Site selectivity of the dark feature R CErCO, ratio of center adatom sites to corner ones, as a function of oxygen partial pressure at room temperature. Total oxygen dose was fixed at 0.24 L at each oxygen partial pressure. areas for the formation of new dark sites, indicating that the dark sites perturbed the local electronic structure and facilitated the formation of more dark sites in their vicinity. Fig. 3a shows the STM image taken from the SiŽ =7 surface which was exposed to 0.96 L of O2 at 5508C at oxygen partial pressure of 1.0= 10 Torr. In striking contrast, the number of major dark sites due to oxygen exposure is much smaller than that at room temperature. The lower density of dark sites on the surface exposed to oxygen at 5508C with respect to that at room temperature suggests pressure at a fixed oxygen dose as shown in Fig. 2. At a very low oxygen partial pressure of 1=10 Torr, the site selectivity R CErCO, ratio of center adatom sites to corner ones, was more than 2.0. When the oxygen pressure was increased, R CErCO decreased substantially but remained larger than 1.0. Even though it is quite surprising that change of oxygen partial pressure by only five times could result in a noticeable decrease of R CErCO, this result clearly shows that the oxygen pressure has a strong effect on the site selectivity of oxygen in the initial oxidation of SiŽ =7 surface. Based upon these experimental results, we can intuitively consider that the lateral interaction between adsorbed oxygen species may become significant at higher O2 flux, resulting in the decrease of surface diffusion compared to the case at low O 2 pressure. When oxygen does not diffuse sufficiently, adsorption is expected to only take place in the metastable sites and it is not going to be so selective. In order to confirm that these contradictory trends to the previous results do not originate from contaminants such as water, we analyzed the residual gas composition with variation of oxygen partial pressure using quadrupole mass spectrometer. The mass spectra did not show any significant hint of reactive chemicals accompanying the oxygen gas dose. Therefore, we can conclude that this site selectivity is mostly due to oxygen. It was also observed that dark areas seemed to appear to act as nucleation Fig. 3. STM images taken from the SiŽ =7 surface exposed to 0.96 L of O at 5508C. Oxygen partial pressure was Ž. 2 a Ž. y7 1.0=10 Torr and b 1.0=10 Torr, respectively. Sample bias voltage was q3.0 V, and the tunneling current was 1.0 na.

5 ( ) J.S. Ha et al.rapplied Surface Science that very little oxygen is left on the surface exposed to O2 at 5508C under oxygen partial pressure of 1.0=10 Torr. There are two possibilities to explain this: Ž. 1 the sticking coefficient is too low at this temperature, Ž. 2 oxygen leaves the surface as a form of SiO, i.e., surface etching occurs. On the matter of sticking coefficient, it has been experimentally known that the sticking coefficients are 1.4 = 10 y2 and 9=10 y3 at room temperature and 5508C, respectively w10 x. From the XPS measurements, it was also observed that about three times higher exposure of O2 at 6008C than at room temperature corresponded to the same oxygen coverage w15 x. Therefore, the difference in the sticking coefficient cannot explain the large difference in the number of dark sites. The second possibility seems to be more plausible. When scanned in the other areas, there appeared some monolayer-deep depressions which were not observed at room temperature. When the O2 partial pressure was increased to 1=10 y7 Torr, such depressions, so called etch marks, were more frequently observed and the Si atoms were resolved inside the etch marks as shown in Fig. 3b. Surface etching has been well known in the high temperature oxidation using a low pressure of O w10,11 x 2. The etching reaction has been known to compete with the oxide film growth depending on the oxygen pressure and the surface temperature w16 x. It is expected that at the low oxygen pressures used here, the maximum desorption rate of SiO is fast enough to desorb oxygen as SiO, leaving an etched silicon surface as observed in the previous work w10 x. When we count the number of dark features at different adatom sites in Fig. 3a, we can find that the dark sites still prefer the adsorption at faulted half to unfaulted half by about 2 : 1. The preference for the center adatom sites to the corner ones Ž 2 : 1. by dark sites observed in room temperature oxidation is still preserved. Such a strong preference for the center adatom sites was also observed in the oxidation at 7308C under the oxygen exposure of 30 L using oxygen partial pressure of 1.0=10 y7 Torr. The preference for center sites at high temperatures can be explained in terms of the difference in thermal stability between two adsorption sites. This also supports the conclusion drawn by Pelz and Koch wx 7, where short time annealing of the oxygen covered SiŽ 111. surface at 625 K caused a net shift in the Fig. 4. Schematic potential energy curves of corner adatom site and center adatom site for dark features as a function of a reaction coordinate. preferred location for both bright and dark sites from corner to center sites. We propose a potential energy curve model combined with the effect of low oxygen pressure we used to understand the strong site selectivity of oxygen-induced dark features observed in this work. As shown in Fig. 4 for O2 adsorption at room temperature and low O2 pressure, O2 adsorbed at the corner adatom site may diffuse to the center adatom site and reaches an equilibrium between the two different sites with 2 : 1 preference even at room temperature since O2 is exposed at low O2 pressure and the barrier height for surface diffusion to center adatom site is lower than that for desorption at corner adatom site. This proposal can be supported by the experimental result that the site selectivity was not changed in the several successive scans over the same surface area after oxygen exposure was terminated. At the temperature of 5508C, the oxygen adsorbed at corner adatoms can easily move to center adatom sites, which have much deeper potential well than that of corner site as experimentally evidenced by the site preference to center sites at high temperatures. At this temperature, oxygen can also be adsorbed directly to center adatoms, since it has sufficient kinetic energy to overcome the energy barrier for adsorption. For adsorption at room temperature and the same O2 dose under a high O2 pressure, it is expected that the lateral interaction becomes significant as the O2 dose is increased resulting in a decrease of surface diffusion of chemisorbed oxygen. This can subsequently reduce the site preference as experimentally observed in Fig. 2.

6 322 ( ) J.S. Ha et al.rapplied Surface Science Conclusions Initial oxidation of the Si 111-7=7 surface was investigated as a function of surface temperature using STM. At room temperature, two different oxygen-induced features, bright and dark sites, were observed indicating two different oxygen chemisorption configurations. As the oxygen dose increased, the number of dark sites increased whereas that of bright sites did not seem to change noticeably. Both the bright and dark features preferred the adatom sites of the faulted half to those of unfaulted half of the DAS structure and the bright sites preferred the corner adatom sites to center ones, which is consistent with the previous results. However, the major dark features preferred the center adatom sites, which is different from the previous results. We explain this particular site selectivity in terms of the potential energy curves of two different sites combined with the effect of low oxygen pressure we used. At low oxygen pressure, oxygen atoms would have sufficient diffusion time to be adsorbed on the most stable sites, center adatom sites. This hypothesis can be supported by no change in the site preference for center adatom sites observed in several successive scans over the same surface area as well as the pressure dependence of site selectivity. Furthermore, the persistent site preference of the dark features for the center adatom sites at high temperatures suggests that the center adatom sites are thermally more stable than corner ones, supporting both the previous result and the potential energy model we proposed. Acknowledgements This work has been financially supported by the Korean Ministry of Information and Communications. References wx 1 H. Ibach, H.D. Bruchman, H. Wagner, Appl. Phys. A 29 Ž wx 2 G. Hollinger, F.J. Himpsel, Phys. Rev. B 28 Ž wx 3 P. Morgen, W. Wurth, E. Umbach, Surf. Sci. 152r153 Ž wx 4 M. Nishijima, K. Edamoto, Y. Kubota, H. Kobayashi, M. Onchi, Surf. Sci. 158 Ž wx 5 Ph. Avouris, I.-W. Lyo, F. Bozso, J. Vac. Sci. Technol. B 9 Ž wx 6 T. Hasegawa, M. Kohno, S. Hosoki, Jpn. J. Appl. Phys. 33 Ž wx 7 J.P. Pelz, R.H. Koch, J. Vac. Sci. Technol. B 9 Ž wx 8 F.M. Leibsle, A. Samsavar, T.-C. Chiang, Phys. Rev. B 38 Ž wx 9 I.-W. Lyo, Ph. Avouris, B. Schubert, R. Hoffman, J. Phys. Chem. 94 Ž w10x J. Seiple, J. Pacquet, Z. Meng, J.P. Pelz, J. Vac. Sci. Technol. A 11 Ž w11x F. Donig, A. Felz, M. Kulakov, H.E. Hessel, U. Memmert, R.J. Behm, J. Vac. Sci. Technol. B 11 Ž w12x R. Martel, Ph. Avouris, I.-W. Lyo, Science 272 Ž w13x R.J. Hammers, R.M. Tromp, J.E. Demuth, Phys. Rev. Lett. 56 Ž w14x S. Ihara, T. Uda, M. Hirao, Appl. Surf. Sci. 60r61 Ž w15x Y. Ono, M. Tabe, H. Kageshima, Phys. Rev. B 48 Ž w16x F.W. Smith, G. Ghidini, J. Electrochem. Soc. 129 Ž