The role of silicon self-interstitials

  In later oxygen precipitation, quartz precipitates are formed with an associated ejection of Si_i. This has led to a general assumption that there are many Si_i present in Cz-Si, and from this to the idea that they may lie in the core of the thermal donor. This has been supported by the Deák model which is based around Si_i. This is further supported by the fact that carbon suppresses thermal donors, which has led to another common assumption that this is occurring through Si_i absorption by the reaction

$$\rm C_s + Si_i \rightarrow C_i$$

This has led to many unusual postulates to explain experimental data, for example that there are two types of dimer complex, `normal' and `O₂^*' which can eject Si_i atoms[266]. However these assumptions are flawed. It is known that C_iO_n defects (the `P-line' complexes) form in the first 15 minutes of annealing[197] at 450$^\circ$C, showing the presence of either C_i or Si_i to produce C_i via the mechanism given above. However these are at extremely low concentrations, and indeed, act as evidence that Si_i concentrations are low as a consequence.

In addition, as described above it is not possible as far as we are aware to construct a single Si_i structure that does not possess gap states, and so thermal donor models based around a core Si_i are unable to explain the electrically inactive forms of the TDs. Carbon suppression of thermal donors can occur through other mechanisms, primarily C_s absorption of O_i and O_2i[267], and possibly also a self-catalysed C_i suppression mechanism (described briefly in Section 9.3.1).

Long time annealing FTIR studies[239] show absorption at 999 cm^-1 associated with the thermal donors. However after long times this disappears and is replaced with broad absorptions at 1012, 1006, and finally 1015 cm^-1, believed to be associated with the shallow thermal donors. This would be consistent with a picture whereby over long times the thermal donors transform into quartz, releasing large quantities of Si_i. These are then able to initially transform all available C_s into C_i, leading to a glut of C_i-based STD formation (CHO₄ complexes). Thereafter the remaining Si_i combine to form $\langle$311$\rangle$ rod-like precipitates.

This therefore suggests that Si_i are not present in large quantities in the early stages of Cz-Si annealing but are formed later when quartz precipitation occurs. In addition they are not involved in the structure of thermal donors[268].