Ab Initio Modelling of Defect Complexes in Semiconductors

The work presented in this thesis started with the implementation of numerically tabulated pseudopotentials into the modelling package, AIMPRO. The algorithm scaled linearly with system size and could be incorporated efficiently into a Gaussian orbital program. Methods for modelling diffusion were then developed and were tested using a set of non-physical functions and also by looking at self diffusion in silicon by the well understood process of concerted exchange. Both of these tests were found to be successful. The dimer-hybrid method was found to be most robust and was used to study the diffusion of nitrogen in diamond. First, the diffusion of P1 centres by concerted exchange was found to have a very large migration barrier of about 8 eV. This was not in agreement with experimental values of 5.5 eV. This throws into doubt the accepted model of diffusion. We have found that a vacancy mediated process has a much lower barrier, provided that a source of vacancies can be identified. In addition, a lower energy pathway exists in the negative charge state. A new method of population analysis was developed and incorporated into AIMPRO, based on the calculation of wave function contributions for each of the atomic Voronoi polyhedra. The method also employed ideas from statistical mechanics in order to divide the contributions among the atoms. AIMBOX was tested in numerous cases and was found to overcome many of the limitations of the Mulliken population analysis.