Vacancy creation on the Si(111)−7 × 7 surface due to sulfur desorption studied by scanning tunneling microscopy

:~:.:.:..~.:.x.::.:.:.~ ,../.... :.:__.:i,.,,,..:, ~:.:.~.~ :,,:,

:::<::;::::.: ...... p:......

Surface Science Letters 297 (1993) L113-L118 North-Holland

” ”..a: .,:,:,,: .:

surface science letters

Surface Science Letters

Vacancy creation on the Si( ill)-7 x 7 surface due to sulfur desorption studied by scanning tunneling microscopy L. Koenders

a, P. Moriarty b, G. Hughes b and 0. Jusko a ’ Physikalisch-Technische Bundesanstalt, Bundesallee 100, D-38116 Braunschweig, Germany b Physics Department, Dublin City University, Glasnevin, Dublin 9, Ireland

Received 12 July 1993; accepted for publication 31 August 1993

Ultra high vacuum scanning tunneling microscopy @TM) has been used to study the adsorption and subsequent thermal desorption of a sulfur overlayer deposited in situ, at room temperature, on the Si(lll)-7 X 7 surface. Little ordering is visible in the STM images of this overtayer 1-2 monolayers thick, with the underlying silicon surface retaining the (7 X 7) reconstruction with weakened low energy electron diffraction (LEED) spot intensity. STM images of this surface following a thermal anneal at 375°C revealed the presence of a number of monolayer deep “holes” or voids in the (7 x 7) surface. The appearance of these voids is consistent with a coalescence of vacancy defects induced by the s&fur desorption process as also observed for oxygen induced etching of the silicon surface. In addition, we have observed small regions of the (6 x fiIR30” and ~$4x 2) reconstructions within areas exhibiting a high degree of surface disorder, with the metastable (5 X 5) reconstruction also being found. Both reconstructions and disorder are attributed to vacancy diffusion to step edges causing a redistribution of surface silicon atoms.

The passiva~ion of the Si(lll)-(7 X 7) surface with group V elements, such as As and Sb, is a well-studied and well-understood phenomenon [1,2]. A monolayer group V species termination of the Si(ll1) surface results in an ideal (1 X 1) bulk termination with three surface silicon dangling bonds quenched by each group V atom. As an alternative to such passivation, Kaxiras has investigated the possibility of silicon surface restoration using a group VI element such as sulfur [3]. He notes that as sulfur tends to form two-fold coordinated bulk structures it is unlikely to restore the three-fold coordinated Si(lll> surface. Although the deposition of sulfur on the Ge(100) surface has been reported to provide such an ideal bulk (1 X 1) te~ination 143,further calculations by Kaxiras 131 have indicated that a sulfur termination of the Si(100) surface is also both chemically and energetically unfavourable. A photoemission surface core-level study of sulfur adsorption on Si(100) by Weser et al. [51 reported that for room temperature adsorption, the sulfur bonded to the surface atoms with no 0039-6028/93/$06.00

change in the (2 X 1) Si(l~) surface LEED pattern. At higher adsorption temperatures there was evidence of sulfur penetrating into the bulk. In a recent STM study of the sulfur-Si(001) system we have confirmed that there is no ideal termination following room temperature adsorption Id]. Deso~tion of the sulfur overlayer at 325°C resulted in the appearance of a c(4 X 4) reconstruction. From the STM images we proposed a missing dimer defect mode1 with the creation of the defects most likely caused during the sulfur desorption process. Feltz et al. [7] have used in situ STM imaging to investigate high temperature oxygen induced etching of the Si(lll)-7 X 7 surface. They demonstrated that during the reaction at temperatures around 65O”C, the initial etching process was via a step flow mechanism in the top silicon Iayer. At later stages the appearance of monolayer deep holes nucleating at structural imperfections in the surface layer was noted. These observations were thus explained by a vacancy coalescence mechanism where vacancy creation was due to desorp-

0 1993 - Elsevier Science Publishers B.V. All rights reserved

L. Koenders et al. / STM study of vacancy creation on Si(lllj-7

tion of a silicon oxide species. A comparable study of the initial oxidation of the Si(lll)-7 x 7 at elevated temperatures (4Oo”C-600°C) by Seiple et al. [8] also demonstrated an oxygen induced etching process with the creation of surface holes and step edge retraction due to the coalescence of vacancies. Bedrossian and Klitsner [9] have observed similar vacancy coalescence and step retraction for sputter induced vacancies due to Xe ion bombardment. Furthermore they have determined that vacancy mediated sputtering can also lead to the appearance of surface regions containing a variety of local metastable reconstructions but no long range order. A number of metastable reconstructions other than the equilibrium 7 x 7 have been observed on the laser annealed Si(ll1) surface by Becker et al. [lo]. A recent STM study of the thermal desorption of ultrathin oxide layers from the Si(100) surface by

x 7 due to sulfur desorption

Johnson et al. [ll] has shown that this process produces a high degree of pitting and a more ragged appearance to step edges. The authors concluded that the desorption of oxygen causes a redistribution of surface silicon atoms. Mobile Si atoms, not directly involved in SiO desorption, are trapped at terrace edges or under remaining oxide patches. In this Letter we firstly confirm the non-ideal termination of the silicon surface with sulfur deposited at room temperature and thus discuss the appearance of monolayer deep holes in the Si(lll)-7 x 7 surface caused by the desorption of the sulfur overlayer at 375°C. We propose that these holes are the result of the coalescence of vacancies created during the thermal desorption of the sulfur overlayer. This colaescence leads to exposure of underlying material, both within the hole regions and at step edges. As this new sur-

Fig. 1. Empty state image of the Si(lll)-7 X 7 surface after exposure to molecular sulfur (approximately 1.5 monolayer coverage). Some corner holes of the 7 X 7 unit cell remain visible.

L. Koenders et al. / STM study of vacancy creation on Si(lll)-7

Fig.

Ia) Monolayer deep enhanced

holes in the silicon surface image highlighting presence

x 7 due to sulfur desorption

after thermal desorption of the sulfur overlayer at 372i”C. (b) Contrast of 7 X 7 reconstruction at the bottom of a surface hole.

L. Koenders et al. / STM study of vacancy creation on Si(Ill)-7

face material is exposed, regions exhibiting a high degree of surface disorder are observed within which there are small reconstructed areas. The UHV tunneling microscope used in this study was from Omicron Vakuumphysik GmBH (base pressure 6 X lop9 Pa) and the vacuum vessel also contained a four grid LEED/Auger electron spectroscopy @ES) system and electron beam heater. All STM imaging was at room temperature using etched tungsten tips. The Si(ll1) samples were cut from commercially available wafers (Wacker). After rinsing in acetone and ethanol, the samples were clamped into a tantalum holder, inserted in the UHV chamber and heated slowly to degas. After subsequently flash annealing a number of times to 900°C a sharp (7 X 7) pattern was observed with LEED. Sample temperature was monitored by a thermocouple mounted on the sample holder. AES did not detect the presence of oxygen or carbon. STM images revealed the characteristic adatom pattern with few vacancies, present on terraces separated by bi-layer steps. This clean surface was thus exposed to a beam of molecular sulfur from an electrochemical cell in a connecting chamber with a base pressure of 2 x 10v8 Pa, The cell, as described by Heegeman et al. [12], was operated under conditions required to produce a beam of almost exclusively S, molecules. From the STM image (I/bias = + 2 V; tunneling current = 0.5 nA) of the sulfur covered surface illustrated in fig. 1 it is clear that no ordered termination of the 7 x 7 surface is present. A saturation coverage was not reached with any exposure - in this particular image the thickness of the sulfur overlayer, as determined from AES studies, is approximately 1.5 monolayers. LEED studies indicated that the surface retained a (7 X 7) diffraction pattern, with visibly weakened spot intensity and an increase in diffuse background. This suggests that the interaction between the Si(ll1) surface and sulfur at room temperature is not strong enough to induce a structural rearrangement of the silicon surface. Imaging at negative bias polarity (filled states) was unstable with no reproducible images formed. It is possible to see in a number of places the (7 X 7) unit cell corner holes. From their position we can deduce

x 7 due to sulfur desorption

that there is no preferential adsorption in either the faulted or unfaulted half of the unit cell at this coverage. After annealing of the sulfur covered surface at 375°C the STM images reveal the presence of a number of monolayer deep holes in the otherwise well ordered 7 X 7 surface. It should be noted that the presence of sulfur could not be detected on the annealed surfaces above the detection limit of the AES system used (approximately 0.1 monolayer). In fig. 2a the monolayer deep holes are clearly visible with fig. 2b illustrating the presence of the (7 x 7) reconstruction at the bottom of a surface hole. Seiple et al. [S] have observed the occurrence of similar behaviour after oxidation of the Si(lll)-7 x 7 surface at 500°C as have Bedrossian and Klitsner [9] following the bombardment of the Si(lll)-7 x 7 surface with 225 eV Xenon ions at elevated temperatures. Both groups conclude that the depressions in the surface are due to coalescence of mobile vacancies created due to SiO desorption at low oxygen pressure [S], and ion sputtering [9], respectively. Desorption of a SiO species exposing a new silicon layer, which is, in turn, (7 x 7) reconstructed has also been described in some detail by Feltz et al. [7]. Their data demonstrates the important role that surface diffusion plays in the reaction process. However they cite two possible diffusing ad-particles - mobile vacancies in the topmost silicon layer or mobile oxygen adatoms - with the latter suggestion described as being less likely. The progress of the oxygen etching reaction was found in that work to dependent on the mobility of the vacancies (or oxygen adatoms), the density of defects in the (7 X 7) structure and the terrace size. If a low defect density existed and/or the vacancy or oxygen adatom mobility was high then the diffusing species could reach the nearest step edge - i.e., the etching would occur via a step flow mechanism. With a high defect density or large terrace size Feltz et al. propose that mobile vacancies or oxygen adatoms could cluster forming monolayer deep pits. We suggest that the desorption of a sulfur overlayer is a further process that leads to the creation of random vacancies which, at the thermal desorption temperature, are sufficiently mo-

L. Koenders et al. / STM study of vacancy creation on Si(lll)-7

bile to coalesce and create holes in the silicon surface layer. Interestingly, the dendritic forms of step edges seen to occur at lower sample temperatures (and corresponding lower vacancy mobility) during oxygen induced Si(lll)-7 x 7 etching by Feltz et al. [7] were not observed in the course of this work. This would suggest that the terrace width on the clean surface was sufficiently large over the majority of the surface to cause mobile vacancies to preferentially coalesce and form holes rather than reach step edges. The theoretical work of Kaxiras [3] also supports the mechanism of vacany defect creation due to sulfur desorption. He suggests that as the S-Si bond energy is larger than the Si-Si bond energy this is likely to result in the formation of volatile Si,S, molecules. Annealing of the sulfur covered sample may thus lead to an etching of the silicon surface. An etching of silicon has also been observed by Holm [13] during silicon-sulfide based vapor phase growth where he stresses that the most important factor for vapor-phase transport efficiency is the difference in partial pressures of

X 7 due to sulfur desorption

SiS, and S, depending on temperature and pressure [14]. However, the etching of monolayer deep holes was not observed after sulfur desorption from the Si(OO1) surface [61. This may relate to a difference in activation energy for Si diffusion on Si(ll1) and Si(OO1)surfaces. Values of 0.2 to 1.6 eV have been determined for activation energy in the case of silicon diffusion on Si(ll1) surfaces [151, whereas Brocks et al. [16] have found a value of 0.6 eV for diffusion parallel to the dimer rows on the Si(OO1)surface and a value of 1.0 eV for diffusion perpendicular to the rows. Following the creation of a large number of surface defects, the remaining surface atoms may not exhibit the (7 X 7) reconstruction, but instead remain largely disordered or reconstruct into other local metastable structures. This appears to be particularly noticeable at step or hole edges as can be clearly seen in fig. 3. A small area of c(4 x 2) reconstruction appears within a surface region exhibiting no long range order close to a step edge. More distant from the edge the undisturbed (7 X 7) periodicity is clearly observed. In

Fig. 3. High degree of surface disorder near to a step edge. Note the appearance of a small region of c(4 X 2) reconstruction within the disordered region.

L. Koenders et al. / STMstudy

of vacancy creation on Si(Ill)-7

this image we believe that the local randomness of the surface structure has been caused by the coalescence of mobile vacancies. Furthermore, very small areas of the (6 X fi)R30” reconstruction have also been observed in the course of this work, most often at the lower surface of a step or hole edge. Both the c(4 x 2) and (fi X &)R30” reconstructions have previously been observed on the laser annealed Si(ll1) surface [lo], and attributed to close packed adatom surfaces with each adatom populating a closed triangular double site, following the work of Northrup [17]. Although the (a X 6)R30” reconstruction has also been attributed to boron dopant segregation at the Si(lll) surface [18], we have not observed this structure on clean Si(ll1) samples from the same wafer without prior sulfur desorption and therefore believe that the (a X fi)R30 reconstruction observed is a silicon adatom structure. Finally, the appearance of a (5 x 5) reconstruction was also observed following sulfur desorption. Feenstra and Lutz [193 have performed an STM and scanning tunneling spectroscopy (STS) study of the (5 X 5) reconstruction prepared via annealing of the cleaved Si(ll1) surface at 320°C. During STM investigations of low temperature epitaxial growth of silicon, Kohler et al. [20] imaged the (5 x 5) after growth at 520°C noting that previous electron diffraction work [21] had reported a disappearance of this structure when the substrate temperature during deposition exceeded 520°C. The (5 x 5) phase observed in this study may again be attributed to reconstruction (at the sulfur desorbing temperature of 375°C) of surface material following creation and coalescence of a large number of vacancy defects. Seiple et al. [S] have similarly explained the appearance of the (5 x 5) structure during studies of the elevated temperature oxidation of Si(ll1) in terms of vacancy coalescence exposing new surface material which may thus undergo reconstruction. In conclusion, sulfur desorption from the Si(lll)-7 x 7 surface has been found to lead to creation of monolayer deep surface holes via a low temperature etching mechanism. Surface material remaining after etching may reconstruct

x 7 due to sulfur desorption

into metastable structures other than the (7 x 7) or may exhibit local disorder. P.M. would like to acknowledge funding from the Irish-German Scientific Collaboration Agreement. We would also like to thank both U. Braasch and D. Schwohnke for organising the visit of P.M. to PTB and S. Schmidt for preparing the STM tungsten tips. We are very grateful to H. Wolff for all his technical assistance. References [l] M. Copel, R.M. Tromp and U.K. Kiihler, Phys. Rev. B 37

(1988) 10756. [2] D.H. Rich, T. Miller, G.E. Franklin and T.-C. Chinag, Phys. Rev. B 39 (1989) 1438. [3] E. Kaxiros, Phys. Rev. B 43 (1991) 6824. [4] T. Weser, A. Bogen, B. Konrad, R.D. Schnell, C.A. Schug and W. Steinmann, Phys. Rev. B. 35 (1987) 8184. [s] T. Weser, A. Bogen, B. Konrad, R.D. Schnell, C.A. Schug and W. Steinmann, in: Proceedings of the Eighteenth International Conference on the Physics of Semiconductors, Ed. 0. Engstorm (World Scientific, Singapore, 1987). [6] P. Moriarty, L. Koenders and G. Hughes, Phys. Rev. B 47 (1993) 15950. [7] A. Feltz, U. Memmert and J. Behm, Chem. Phys. Lett. 192 (1992) 271. [8] J. Seiple, J. Pecquet, 2. Meng and J.P. Pelz, in press. [9] P. Bedrossian and T. Klitsner, Phys. Rev. B 44 (1991) 13783. [lo] R.S. Becker, A. Golovchenko, G.S. Higashi and B.S. Swartzentruber, Phys. Rev. Lett. 57 (1986) 1020. [ll] K.E. Johnson, P.K. Wu, M. Sander and T. Engel, Surf. Sci. 290 (1993) 213. [12] W. Heegeman, K.H. Meister, E. Bechtold and K. Hayek, Surf. Sci. 49 (1975) 161. [13] C. Holm, private communication. [14] C. Holm and E. Sirtl, J. Cryst. Growth 54 (1981) 253. [15] F. Allen and E. Kasper, in: Silicon Molecular Beam Epitaxy, Eds. E. Kasper and J.C. Bean, Vol. 1 (CRC Press, Boca Raton, FL, 1988) p.65, and references therein. [16] G. Brocks, P.J. Kelly and R. Car, Phys. Rev. Lett. 66 (1991) 1729. [17] J. Northrop, Bull. Am. Phys. Sot. 31 (1986) 584. [18] P. Bedrossian, R.D. Meade, K. Mortensen, D.M. Chen, J.A. Golovchenko and D. Vanderbilt, Phys. Rev. Lett. 63 (1989) 1257. [19] R.M. Feenstra and M.A. Lutz, J. Vat. Sci. Technol. B9 (19911 716. [20] U. Kohler, J.E. Demuth and R.J. Hamers, J. Vat. Sci. Technol. A 7 (1989) 2860. [21] M. Horn, PhD Thesis, Universitit Hannover (1988).