In this thesis we apply a variety of computational methods based on density-functional
theory (DFT) to the study of defect centres in bulk silicon and
silicon nanostructures.
Firstly, we discuss the system-size convergence of point defect properties
in the supercell method for deep-level defects in bulk silicon; we consider both
the vacancy and gold impurity.
For the vacancy, we investigate systematically the main contributions to
the finite size error that lead to the well-known slow convergence with respect
to system size of defect properties, and demonstrate that different properties of
interest can benefit from the use of different k-point sampling schemes. We also
present a simple and accurate method for calculating the potential alignment
correction to the valence band maximum of charged defect supercells by using
maximally-localised Wannier functions, and show that the localised view of the
electronic structure provided by them gives a clear description of the nature
of the electronic bonding at the defect centre.
For the gold impurity, we show that the system becomes a non-spin-polarised
negative-U centre due to the effect of Jahn-Teller distortion, thus providing a
simple explanation for the absent electron paramagnetic resonance signal for
gold in silicon. The calculated transition levels are found to be in excellent
agreement with experimental measurements.
We then investigate the segregation of arsenic impurities in silicon close to
an interface with amorphous silica. We employ a multiscale approach, generating
a realistic disordered interface structure from Monte Carlo simulation,
with a continuous random network model of the system parametrised from
DFT. We calculate the segregation energy using DFT for a large number of
substitutional sites encompassing all the oxidation states of silicon, and show
that the results can be understood with a minimal model based only on the
local strain and volume of the defect site.