Multiple pathways for permanent deactivation of boron-oxygen defects in p-type silicon

- Boron-oxygen (BO) defects are permanently deactivated via two independent pathways.
- Thermal deactivation of BO occurs during rapid thermal ramp down.
- 'Regeneration' of BO occurs during illuminated annealing.
- Thermal deactivation and regeneration have different underlying mechanisms.
- Thermal deactivation likely involves defect dissociation into precursors.
- Regeneration is likely a hydrogen passivation effect.
- Presence of hydrogen strongly influences regeneration but not thermal deactivation.
- A four-state defect model is proposed to account for the thermal deactivation pathway.

In this work, it is shown that there are at least two separate pathways for the permanent deactivation of boron-oxygen defects - a purely thermal pathway involving the dissociation of defect precursors, and another mechanism based on passivation of the fully-formed defect during an illuminated annealing process (regeneration). Based on investigations on fired and non-fired p-type Czochralski silicon wafers, a thermal reduction in the net concentration of BO defects is confirmed and found to be caused by rapid cooling following a belt-furnace firing process. This thermal deactivation occurs independently of any subsequent permanent deactivation induced by illuminated annealing and is likely related to the dissociation of the defect and subsequent loss of defect precursors to other species. Further, by varying the surface dielectric present on the wafers during a rapid thermal process, it is demonstrated that in the absence of hydrogen in the wafer bulk, applying an illuminated annealing process does not result in any significant permanent deactivation of defects. In contrast, thermal deactivation occurs independently of the presence of hydrogen in the wafer bulk. This demonstrates that permanent deactivation via the thermal pathway and that via the illuminated annealing (regeneration) pathway occur independently of each other and have different underlying mechanisms. The implications of multiple pathways for permanent deactivation are discussed and the well-known three-state model is revisited in light of this information. A fourth state to represent the end point of thermal deactivation (State D) is proposed for more accurate modelling of BO defect kinetics.