It is found that the symmetric dimer reconstruction of the BC8 silicon (001) surface to be stable with respect to both of the possible unreconstructed terminations of the bulk. A summary of the energies found for the (001) surfaces of silicon in the BC8 structure considered here is given in Table 6.1. Unlike various silicon and germanium surfaces, the antisymmetric dimer is not favoured. In all cases the surface dimers reduce in length below that of the bulk bond length. In doing so the bonds become slightly stronger than that of a Si-Si single bond, with the electronic charge within the bond increasing to more than 2.0 electrons. Where two unbonded electrons per surface atom remain, a bond to an atom in the bulk is formed. This reconstruction forms five fold coordinated atoms with all electrons involved in bonding and is quite close to being stable. The possibility of under coordinated atoms forming donated bonds with neighbouring atoms at little energy cost suggests new topological possibilities for amorphous silicon. The low energy of this surface may also provide new insight into the nature of the bonding on amorphous silicon surfaces, and into silicon clusters which are stabilised because of their high surface/volume ratio.
Table 6.1: Surface energies of the relaxed (001) surfaces of BC8 silicon.
In spite of the 17.5% additional broken bonds per unit area over the diamond Si (111) cleavage plane the calculated surface energy (0.0796eV/Å ) is lower than that of any previously reported silicon surface. Although ab initio studies cannot consider all possible atomic configurations and reconstructions, the central result that BC8 Si has surfaces of lower energy than diamond Si remains. There is a strong implication that amorphous silicon will behave likewise.
The stability of this surface with respect to other diamond Si surfaces suggests that the metastability of BC8 will be enhanced because of impeded nucleation and growth of the diamond structure at the BC8 silicon surface. This result is of particular relevance in view of the recent report of pressure-induced phase transitions in Si by nano-indentation. A semimetallic region formed by such a process might be expected to have a very long lifetime, determined by the stability of the semiconductor-metal boundary.
A qualitative explanation of the lower BC8 surface energy can be ventured by considering the strain energy required to distort bonding away from the prefect tetrahedral angle. Any reconstruction of a Si-I surface will cause such distortions, but since the bonds in BC8 silicon are already distorted away from the tetrahedral configuration, in some cases the surface reconstruction actually reduces this distortion. This argument also applies to amorphous silicon and suggests that amorphous silicon surfaces may be lower in energy than those in crystalline diamond Si.