We also modelled the di-square structure shown schematically in Figure 9.1. Earlier CNDO/S calculations showed this to be more stable than the di-y-lid structure [105]. However there are several things wrong with this model, the primary fault being the associated defect states. The defect actually has four electrons in the gap near to the conduction band, since each square acts as a double donor. Hence running it in the +2 charge state still leaves two electrons close to Cb (see Figure 9.10d). This is sufficient to exclude it as a thermal donor model. In addition it was less energetically stable than the di-y-lid model.
For these reasons we exclude the 4Oi di-square model as a potential thermal donor model.
Calculating the modes for one of the two squares gave strong modes at 725, 544 and 533 cm-1, and hence this suggests that O-square structures on their own cannot account for the observed experimental TD vibrational modes. However the mode at 725 cm-1 is close to the lower modes observed, and hence it is possible that a square structure in combination with other Oi could be responsible for TD1. This mode corresponded to an in-plane stretch of one of the trivalent oxygen atoms.