Other theoretical approaches

  The problem of oxygen complexes in silicon has been addressed using many other theoretical approaches. Although ab initio methods such as ours provide a very complete description of the systems examined, we are limited in the system sizes that we can study. In addition we are unable to tackle molecular dynamics type problems which would require too much computing power. There are therefore other computational methods used, the most notable of these being semi-empirical methods and parameterised potential methods. Deák and Snyder have had much success using CNDO and MINDO/3 semi-empirical methods for pioneering work on large oxygen complexes and y-lid based structures [28], and potentials such as the Stillinger-Weber (SW) and Jiang-Brown (JB) have been used to dynamically model self-interstitial clustering [29] and oxygen diffusion processes [30,31]. However parameterised methods often suffer from transferability problems and do not give many oxygen systems correctly (e.g. Si-$\rm\hat{O}$-Si of 180$^\circ$ in O_i and O_s under CNDO [28] and JB [31] respectively). With ab initio calculations used to obtain many of the parameters anyway, first-principles studies should be preferred over empirical or semi-empirical calculations where the size of the problem permits.