While thermodynamic concepts allow the calculation of thermal equilibrium concentrations of point-defects, they give no information on how equilibrium is reached and which are the physical processes that generate or annihilate point defects in the silicon crystal. At higher temperatures a silicon atom can occasionally acquire sufficient energy from lattice vibrations to leave the lattice site and, hence, an interstitial and a vacancy are generated. This process is reversible and known commonly as Frenkel pair process if point-defects are generated and as bulk process if point defects recombine.
There is no need for the number of interstitials and vacancies to be equal in thermal equilibrium. The local dynamic equilibrium can be expressed by (3.1-12) which allows a supersaturation of interstitials along with a local undersaturation of vacancies at the same lattice site.
Recently various estimates for the recombination velocity of point-defects have been proposed. Dunham suggested from his phosphorus in-diffusion experiments that I-V recombination has to be fairly quick to ensure a high supersaturation of interstitials near the phosphorus diffusion front [Dun91]. Otherwise enhanced diffusion would be overestimated in the tail region. Cowern presented short term boron annealing experiments which lead to the conclusion that the diffusion of interstitials is limited due to bulk recombination [Cow91].
Another possibility for the generation of point-defects are reactions occurring at the surface during oxidation, nitridation, and silicidation. When the silicon crystal is heated, point defects flow from the surface into the bulk and the opposite type is transported to the surface. This process is known as Schottky process. As the detailed kinetics of the surface reactions are not well understood, an exact determination of reaction rates is not possible at present.
The major source of point-defect generation during processing is the radiation damage of the silicon lattice by ion implantation. The amount of generated point-defects depends on the implantation dose. Modern Monte Carlo simulations are able to predict the number of generated point-defects, but suffer from a lack of realistic recombination models during implantation [Hob88] [Hob95].
Finally, precipitation of oxide impurities in the silicon bulk can have significant impact on the point-defect population in the bulk. The increase of the volume during formation of the oxide precipitates causes strain in the lattice matrix. A strain relief can be given by vacancy absorption or interstitial emission. The nucleation mechanism of the precipitates is still unknown and currently under investigation.
In addition, it is also possible for point-defects to coalesce into interstitial or vacancy aggregates. Thus, the number of excess point-defects available for enhanced diffusion is limited. The point-defects are assumed to recombine at sinks within the bulk region and to be emitted from the point-defect clusters. Further growth or dissolution of the point-defect clusters depends on the local point-defect concentration related to a solubility limit for the point-defects [Cow93].
Finally, there are recombination processes in the bulk with other species than point-defects involved, which can act as a sink for point-defects. The most established one is the Frank-Turnbull recombination, where impurity-defect pairs are recombining at lattice defects (see Section 3.1-18 and 3.1-19).
