Discussion

The results give strong support to the idea [141,90] that VO_n complexes preferentially form during annealing of irradiated O-rich Si. The VO₂ defect gives three IR active modes but only one has been detected. VO₃ gives three high frequency LVMs which have all been detected. V_nO complexes on the other hand give modes close to VO, i.e. around 830 cm^-1. VO₂ appears to have no connection with TDs [90] but may be a nucleus for oxygen precipitation.

One of the remaining questions is why [O_i] does not decrease when VO anneals out at 350$^\circ$ and VO₂ is formed. We originally proposed this was due to oxygen dimers [99] complexing with vacancies, and recent work by Londos et al [79] seems to confirm this. The third, slow, formation process for VO₂ occurs while VO levels remain relatively static, suggesting that this is not a VO transformation process. It is known that vacancy aggregates exist in the material and frequently release and re-trap isolated vacancies. It therefore seems likely that these isolated vacancies in the structure are soaking up oxygen dimers. Alternatively there could be a breakdown of VO centres as they trap dimers, and the release of O_i through this mechanism roughly balances the loss of O_i through trapping of O_i by VO centres.

Different treatments will place different emphasis on each reaction step, but it now seems that the following processes contribute to the breakdown of VO and the formation of VO₂ (the numbering corresponding to the three regimes discussed in Section 5.1.2):

Processes such as

are unlikely due to the low concentration of VO centres as compared to other defects such as O_i. Note that unlike normal Cz-Si there will be a large number of Si_i present in this material due to the irradiation. This complex behaviour of VO destruction and VO₂ formation provides strong evidence for the presence of interstitial oxygen dimers in Cz-Si, since without them it is difficult to explain the observed reaction processes.

There is also some uncertainty about the formation kinetics of the VO₃ defects. Corbett et al [141] pointed out that the growth of the VO₃ bands was anomalously fast and required either a very non-uniform distribution of O_i or a diffusivity of VO₂ higher than that of O_i. However more recent work by Lindström et al [90] states that the activation energy for the growth of VO₃ is close to that of the diffusion of O_i, suggesting that O_i diffuses to immobile VO₂. If Corbett is correct there could naturally be O_i present quite close to many of the VO₂ centres. This could occur by

$$\rm VO +O_{2i} \rightleftharpoons V + O_i + O_{2i} \rightarrow VO_2 + O_i$$

which could leave the O_i quite close to the VO₂ centre. An alternative is that VO₂ is breaking down to VO and then trapping a dimer before the VO can lose its second oxygen. The activation energy observed by Lindström is then actually the activation energy for the formation of the dimers in the first place, since once formed they can rapidly migrate to these VO centres. It seems a new study of the growth of VO₃ is called for which should also include the effects of mobile dimers.

Finally we note that when O_i is complexed with a defect imposing tensile strain on the lattice, such as VO, N₂ [100] or C_s [101], the O_i-related LVM is reduced from 1136 cm^-1 to around 1000 cm^-1. TDs also have modes in this region [102] and this suggests that they might consist of O atoms surrounding a core which gives rise to tensile strain. This need not be a vacancy, as many defects such as interstitial N₂ pairs [103], C_i [104] and Si_i [105] can also give dilated Si-Si bonds. This will be discussed further in Chapter 9.