Experimental Evidence

  Lindström and Hallberg annealed Cz-Si samples at 370$^\circ$C and 450$^\circ$ for 25 hours, and the variation in absorption of the various observed LVMs provide strong support for assignment of the 1012 cm^-1 mode to the oxygen dimer. Their results are shown in Figure 6.11 and are reproduced from Reference [239]. Note that the bands at 975, 988 and 999 cm^-1 are associated with the early thermal donors, TD1, TD2 and TD3 respectively.

At 370$^\circ$C where O_i is not mobile, there is a gradual decrease in the 1012 cm^-1 dimer band from its residual levels, accompanied with an associated increase in the 975-1006 cm^-1 bands. These changes have been directly correlated [160], and are consistent with a thermal donor model constructed from dimers (see Chapter 9). There is no notable change in the O_i levels, suggesting the available dimers are becoming depleted as they form higher order complexes, but the lack of O_i migration means that new dimers are not being formed and hence the dimer concentration is dropping.

At 450$^\circ$C when oxygen is mobile (Figure 6.11b), the story is very different. There is a rapid initial drop in [O_i], accompanied with a rapid rise in dimer concentration. Meanwhile there is also an increase in the other modes. The dimer concentration then tails off, before reaching a steady state growth along with a similar growth in the 975-1006 modes, and an associated drop in [O_i]. This can be understood in terms of an initial glut of dimer formation due to O_i atoms which are close to the capture radius required to form a dimer. However once these are exhausted, there is then a drop to steady state O_i diffusion from a more homogenous background level; the behaviour of the 975-1006 modes is then simply a knock-on from the fluctuation in dimer levels (note that this is different from the explanation given in Reference [239]).

  

Figure: The initial changes of the bands 975-1012 cm^-1 and [O_i] for annealing at (a) 370$^\circ$C (top) and (b) 450$^\circ$C (bottom). Taken from Reference [239]. In the 450$^\circ$C run, the 975 cm^-1 signal was too weak for inclusion, and has a maximum that appears and disappears too quickly to appear on this scale.