Trimer formation

It is next important to consider the 1006 cm^-1 mode. It was suggested earlier that this could be assigned to the trimer, O_3i, since it too is normally present with the 1012 cm^-1 mode, and since O_i is the most common impurity in Cz-Si it seems likely that O_2i will complex with this. Of the modes discussed here, only the 1012 and 1006 modes are present in as-grown Cz-Si [239]. This also suggests that it must be one of the earlier complexes to form, with a trimer being the next simplest structure after the dimer. This mode has not been correlated with any of the early TDs (TD1-4).

The 1012 cm^-1 dimer mode anneals out under first order reaction kinetics [149]. This is similar to the

$$\rm VO + O_i \rightarrow VO_2$$

reaction [79], which also obeys first order reaction kinetics since [O_i]>>[VO] and so effectively remains constant throughout the reaction. Therefore this observation is consistent with a reaction scheme $\rm O_{2i} + O_i \rightarrow O_{3i}$.

Both the 1012 and 1006 behave similarly when exposed to oxygen dispersion treatments (see Figure 6.12). This behaviour is consistent with a dimer / trimer model, since in a 2 hour dispersion treatment at 1100$^\circ$C, the integrated intensity level of the 1012 cm^-1 level halves and the 1006 cm^-1 practically disappears. In the following anneal at 390$^\circ$C the 1012 level then drops further as the dimers are not being replaced by O_i diffusion. However the 1006 rapidly starts to reappear. This seems inconsistent if it is assigned to a large defect centre but follows naturally if it is assigned to trimers that are forming from the remaining supply of dimers in the material. The graph is also useful in later discussions of TD1, since the dispersion treatment effectively removes the as-grown `pool' of trimers which has a knock-on effect on the TD1 formation rate. We therefore conclude that the 1006 cm^-1 can be assigned to the oxygen trimer, O_3i.

  

Figure: The initial changes of the 1006 and 1012 bands during the adjustments to their respective equilibrium levels and the growth of TD+1 (1s-3p$\pm$) for an as-grown sample and a dispersed sample (2h at 1100$^\circ$C). Note that the TD data applies to the r.h.axis, and the 1006/1012 data to the left, hence the initial dispersed 1006 level is almost zero. Taken from  Reference [239].