Intermixing and buried interfacial structure in strained Ge/Si(105) facets

The strained {105} facet, fundamental in the heteroepitaxial growth of Ge/Si(100), is investigated through a combination of scanning tunneling microscopy, reflectance anisotropy spectroscopy, and density functional theory simulations. Besides providing a strong independent confirmation of the proposed structural model, optical measurements give insight into Si/Ge intermixing, reveal hidden signatures of the buried interface, and give access to a complementary viewpoint of the epitaxial growth with respect to standard top-layer probing. Strained subsurface atoms are found to strongly determine the electronic and optical properties of the whole reconstruction. Moreover, we demonstrate how their unique spectral fingerprint is a sensitive probe of the local chemical bonding environment and allows the stoichiometry of atomic bonds to be monitored within and beneath the surface layer.

Figure 1

STM images of (a) the RS-reconstructed Ge/Si(105) surface at a coverage of $\sim$4 ML of Ge, and (b) a Ge/Si(001) pyramid showing RS-reconstructed {105} facets.

Figure 2

(a) Schematic structure (top and side views) of the Ge/Si(105) rebonded-step surface reconstruction for a 4-ML Ge coverage. Pink circles are Ge; subsurface Si layers are shown below (in green). Dashed lines show the surface unit cell. (b) Experimental and simulated STM images ($28\times 33)$ Å${}^2$ for filled ($V=-1.5$ V; $I=1.5$ nA) and empty ($V=+1.0$ V; $I=0.7$ nA) states at constant current.

Figure 3

(a) Measured RAS spectra for the clean Si(105) surface (lower curve, blue), and after absorption of 4 ML of Ge and annealing at 870 K (upper curve, red). The residual optical anisotropy (measured at the fully oxidized surface) has been subtracted in each case. (b) Computed RAS spectra at 4 ML Ge of the RS reconstruction model with (dashed line) and without (solid line) subtraction of the optical background (hydrogenated surface), compared with the signal for the PD model (dotted line). A Lorentzian broadening of 0.15 eV is used. (c) DFT-LDA band structure of the 4-ML coverage slab, superimposed on the projected bulk band structure of silicon, plotted within a quarter of the surface Brillouin zone (see inset: $\overline{\Gamma }$–$\overline{K}$ corresponds to the $\left[0\overline{1}0\right]$ direction). Band splitting derives from the presence of two horseshoe units per cell. Isosurfaces of ${\left|\psi \right|}^2$ near the $\overline{J}$ point for selected surface states are shown with respect to the atoms of the horseshoe structure [Fig. 2a].

Figure 4

Left: Computed RAS signals at a 4-ML coverage for different interface compositions, compared with experiment (dots) and a pure silicon RS-reconstructed slab (below). Right: Corresponding depth profile of Ge, and side views of the structures used in the simulations.

Figure 5

(a) STM images ($50\times 50$ Å${}^2$) of the (105) surface for increasing coverage $\theta$ of Ge. (b) Measured RAS spectra near the main peak position for different $\theta$. The signal from the unreconstructed surface has been subtracted in each case. Lines are a guide to the eye. Baselines are indicated on the left. (c) Calculated spectra. (d) Intensities of the P and S features as a function of $\theta$ (experiment: dashed lines; theory: solid lines). The corresponding number of nearest-neighbor Ge-Ge bonds in the subsurface tetrahedron and changes in stoichiometry are indicated.