Theory of optical reflectance anisotropy of the natural Si110surface Bernardo S. Mendoza* Centro de Investigaciones en Optica, A.C., Leo ´n Guanajuato, Mexico Rodolfo Del Sole Istituto Nazionale per al Fisicadella Materia, Dipartimento di Fisica, II Universita ` di Roma Tor Vergata, Rome, Italy Anatoli I. Shkrebtii Department of Physics, University of Toronto, Toronto, Canada Received 23 January 1998 We investigate theoretically the optical reflectance anisotropy of the natural Si110surface, including the surface intrinsic anisotropy and the surface local-field effect. We model a rough Si110surface, determining a number of defect structures by total-energy minimization. We show that a structure with missing Si atoms and partial H coverage gives a reflectance anisotropy in agreement with experiment, in contrast to previous calculations for the ideally H-terminated Si110surface. We show that both the intrinsic anisotropy and the surface local-field effect are essential for such an agreement. S0163-18299851720-1 Reflectance anisotropy spectroscopy RASis one of the few optical techniques that can probe the surface and inter- face structure of cubic materials. 1–7 It measures the differ- ence between the normal-incidence optical reflectance of light polarized along the two principal axes in the surface plane as a function of the photon energy. Since the bulk optical properties of cubic crystals are isotropic, any ob- served anisotropy must be related to the lower symmetry of the surface. RAS data are typically obtained in the visible- ultraviolet spectral range, thus providing information about electronic structure modifications due to the creation of the surface, reconstructions, adsorbates, etc. Theoretical calcula- tions of the reflectance anisotropy RAfor a few important clean and adsorbate covered semiconductor surfaces are in reasonable agreement with experiments. 8–11 For a clean sur- face of a cubic semiconductor, an important contribution to the RA comes from surface states and surface recon- struction. 6,12 However, RA has also been found at the natural oxygen or hydrogen passivatedunreconstructed Si110 surface, 1,7 where no optically active surface states are ex- pected in the frequency range of interest. In this case, the RA may arise due to the intrinsic anisotropy IAcoming from the effect of wave-function termination at the surface, and to the surface local-field effect LFE. The LFE alone yields a RA in good agreement with experiment for the natural Ge110Ref. 13and Si110Ref. 14surfaces; while the calculated IA line shape is substantially different from the experiment. 8 On the other hand, when the two effects are combined for the ideal Si110:H, 14 the surface LFE modifies the IA line shape, making it closer to the experimental RA of the natural surfaces. However, relevant discrepancies with experiment still remain, especially above 4 eV. Hence, de- spite the definite progress of the theory for the clean and adsorbate covered well ordered surfaces, 8–11,15 the RA of the natural Si110surface is not yet understood. Nevertheless, because of its stability and reproducibility, it has been pro- posed as a RAS standard. 16 Therefore, there is still a need to understand the physical mechanism underlying the RAS of natural Si110, and the purpose of this paper is to investigate systematically this surface. Actually, the key point seems to be in the prepara- tion of the samples. As Ye et al. have recently shown, 17 etch- ing is indeed crucial to preparing an ideal that is, well or- dered and ideally H-terminatedSi110surface. Depending on the etching time and conditions, one can go from an atomically rough surface to a practically ideal termination. Moreover, not only etching conditions, but also the hydroge- nation process, determines the surface structure. For in- stance, as Watanabe has recently shown, 18 hydrogenating the surface in hydrofluoric acid or in hot deoxygenated water gives a surface with missing Si atoms or ideally H covered, respectively. It seems that all available RAS measurements at natural Si110have been performed on surfaces of the former type, that is with many defects. 1,7 In view of this evidence, we model the natural Si110 surface as a surface with atomic defects. We determine vari- ous defect structures by total-energy minimization and use them to calculate RA spectra according to the LFE-IA model. We show that a rough Si110surface, with missing Si atoms and partial H coverage, produces a RA in good agreement with experiment. Following Ref. 14, we briefly sketch the procedure taken to calculate the RA. To model the semi-infinite crystal we use a slab consisting of N atomic 110layers, extending from z =0 to z =Nd , d being the interlayer distance. We define the RA signal as S RA R / R 01 ¯ 0 -R / R 001 , 1 where we have taken the normally incident light to be polar- ized along 001( x ) or 11 ¯ 0( y ). Neglecting the LFE, the surface contribution to normal-incidence reflectance is given by 19 R / R i =4 / c Im4 D ii hs / B -1 , 2 RAPID COMMUNICATIONS PHYSICAL REVIEW B 15 MAY 1998-II VOLUME 57, NUMBER 20 57 0163-1829/98/5720/127094/$15.00 R12 709 © 1998 The American Physical Society