Dislocations in semiconductors
A set of codes with 3D periodic boundary conditions has been developed to model dislocations in semiconductors. Several schemes have been used to investigate the atomic structure of dislocations; classical potentials incorporated in a Molecular Dynamics framework, a tightbinding k-space scheme and ab initio pseudopotential codes developed at Cambridge and Edinburgh. An error has been detected in previous work that modelled dislocations using periodic boundary conditions. It is demonstrated, for the 90° and 30° Shockley partials, that a mismatch at the periodic boundaries leads to erroneous atomic and particularly electronic structures. A new approach is proposed which through its geometry obviates this problem. The Stillinger Weber potential has been found to predict a completely new type of reconstruction for the 90° partial. Recent work by other authors confirms this and predicts significantly different results to earlier work. A thorough investigation has been made into the bonding processes involved in the core of the 90° partial. This study has involved reproducing much of the earlier work to understand why there has been such poor agreement between various authors. The reconstruction of the 90° partial is found to involve a symmetry lowering displacement intimately connected to its electronic structure. The band-gap is predicted to be clear of states, except for the possibility of shallow states at both band edges, which contradicts the findings of the most recent work on this partial by other authors. The interaction of phosphorous with the 90° partial has been studied using the tightbinding model. The Hamiltonian has been parameterised by comparing the predictions to an earlier ab initio cluster method study. Good qualitative agreement with the ab initio work is obtained, including the prediction of a strong dislocation locking effect by phosphorous. Preliminary studies on the unreconstructed 30° partial show that phosphorous is strongly bound to the three-fold coordinated sites resulting in no states in the indirect band-gap. The modelling of interstitial copper at the core of the 90° partial has been initiated. The ab initio codes have been used and new silicon and copper pseudopotentials tested. The first attempt to model copper located interstitially in the core was not successful and the reasons for this have been identified. However, it is evident from this investigation that the neutral copper strongly repels and does not form bonds with the surrounding silicon atoms. A review is given of two techniques that have been developed to obtain the thermally averaged structure and concentration of vacancies at dislocations, together with a preliminary investigation on the Frank partial.