The formation of VO₂

  The OV defects anneal out at around 300$^\circ$C and a new line appears at 894 cm^-1 (4 K) which then grows in intensity [141]. This band shifts to 889 cm^-1 at room temperature. The original proposal [141] was that the absorption is due a VO₂ defect, i.e. two O atoms sharing a vacancy, each bonded between two of the Si vacancy neighbours (Figure 5.2). The symmetry is then D_2d. There are three processes which contribute to the demise of VO.

1.

Initially a fast process in which the [O_i] rapidly increases. This may be related to the break up of aggregates of Si_i introduced during the irradiation. This then leads to the reaction: VO + Si_i $\rightarrow$ O_i [81,82,79].

2.

A second slower process. This is assumed to be: VO + O_i $\rightarrow$ VO₂ and caused by VO diffusion to the immobile (at 350$^\circ$C) O_i. This assignment is based on the 1.86 eV activation energy [80], the migration barrier for VO diffusion.

3.

Not all VO is lost by this second process and a third involves VO being captured by some (unknown) defect and producing O_i. Roughly half of VO can meet their fate in this third process.

Support for the VO₂ assignment of the 889 cm^-1 defect comes from its IR-absorption intensity which is approximately proportional to [O_i]² when VO has disappeared. However, there is no loss of O_i, as measured by IR-absorption, during the slow stage when the 889 cm^-1 defect is being formed [81]. This initially provided the necessity for the third formation mechanism given above, however recent studies by Londos et al [79] show two distinct activation energies in the slow process, with a switch in dominance at about 360$^\circ$C. In addition they showed that above 380$^\circ$C VO₂ absorption continues to increase, even though VO levels have almost stabilised.

Further, early uniaxial stress studies on the 889 cm^-1 defect [77] concluded that its symmetry was lower than D_2d; unless the symmetry was D_2d, two O-related LVMs would be expected. However, prompted by our theoretical investigations, Bech Neilsen et al [83] re-examined the defect with uniaxial stress and showed it to possess either C_2v or D_2d symmetry. Although it is difficult to distinguish between the two, they have recently performed a range of stress-induced dichroism experiments that were able to unambiguously show D_2d symmetry [83] (see below). Finally, for mixed ¹⁶O-¹⁸O samples, only two LVMs were detected and not three [84,85], with isotopic shifts indicative of a single oxygen atom. This shows that the two O-modes in this model for the 889 cm^-1 centre are decoupled even though their separation can only be a few Å. This objection is not fatal as earlier modelling studies [86] found that the O atoms were decoupled in the LVMs.

The difficulties [87] described here have led to an alternative assignment of the 889 cm^-1 LVM to V₃O [78]. This defect would possess only one LVM and possess C_1h symmetry. Presumably, in this model vacancies are thermally released from VO or V_n complexes and form stable divacancies which subsequently trap VO. EPR studies have suggested [88] that complexes such as V₂O and V₃O, with S= 1, form and anneal out around 300$^\circ$C although Davies et al [89] suggest that V₂O may be responsible for an LVM at 1005 cm^-1 (10 K). This defect anneals out around 450$^\circ$C.