The 1012 cm^-1 LVM

Until recently no LVMs were assigned to the oxygen dimer. However it has been suggested that a vibrational mode at 1012 cm^-1 (first observed in 1972 [146]) could be due to the oxygen dimer [147].

There are several reasons for this. Firstly it is present after heat treatments at a wide range of temperatures: 650$^\circ$C [102,148], 1100$^\circ$C [149] and 1350$^\circ$C [148], as well as in as-grown material at concentrations of about $1.5\times10^{15}$cm^-3 [270]. The line is definitely oxygen related given its 43 cm^-1 isotopic shift when going from ¹⁶O to ¹⁸O [239]. Its annealing properties (discussed in Section 6.5.1) are also consistent with that of a rapidly diffusing dimer. Given that it is present over such a range of conditions it seems sensible that it should be the simplest oxygen complex possible, namely a dimer.

Assuming that the 1012 cm^-1 mode is responsible for the dimer, simple rate equation analysis led to a binding energy in the range 0.1 $\leq$ E_b $\leq$ 0.2 eV [147]. This is consistent with the idea that the dimer should not be important in the kinetics of TD formation at higher temperatures [272]. Fits to plots of Log[1012] vs. Log(1/T) gave a more accurate dimer binding energy of 0.16 eV [147,239]. Recent detailed analysis has suggested a figure of $\rm E_b = 0.3 \pm 0.05$ eV [150].

In addition to the 1012 cm^-1 mode, there is a secondary weaker mode at 1006 cm^-1 which is normally observed under the same conditions. It is also present in as-grown material, but at concentrations of about $1.5\times10^{14}$ cm^-3 [270], and shows similar annealing properties. If the 1012 cm^-1 mode is produced by the dimer, given that the most common defect in silicon is O_i, it is likely that the 1006 cm^-1 mode is caused by the oxygen trimer, O_3i. Little is known about this defect, but crucially, from fitting their models to experimental data, Markevich and Murin showed the fits were greatly improved for the 1006 cm^-1 mode if they assumed that this defect was highly mobile [147].

There is a strong correlation between the 1012 cm^-1 band and the 1107 band due to O_i [102]. The band shows very little temperature dependance (unlike O_i, suggesting any bonding in the defect is very different to O_i); it shifts from 1013 to 1012.2 cm^-1 on going from 300K to 10K. With ¹⁸O it drops to 969 cm^-1. It is always present in as-grown material.

After we suggested a dimer should give rise to more modes, further examination has determined modes at 1105 and 1062 cm^-1 (1060 cm^-1 at 10K) which are correlated with the 1012 cm^-1 [151]. The 1062 cm^-1 mode is nearly twice as intense as the other two, and is very broad (FWHM $\sim$7 cm^-1 at 10K). More recently the 1105 cm^-1 was shown to be unconnected. There are also two lower modes at 685 and 552 cm^-1, the 552 cm^-1 exhibiting only a small shift with oxygen isotope.

Further evidence has come from irradiation experiments by Lindström and Hallberg [150]. Electron irradiated material was annealed, giving a high concentration of VO₂ centres. This was then proton irradiated (which leads to Si_i production), and using FTIR a decrease in the VO₂ 889 cm^-1 signal was observed concurrently with the appearance of lines at 1012 and 1062 cm^-1. This is consistent with the reaction $\rm VO_2 + Si_i \rightarrow O_{2i}$.

The 1012 cm^-1 band is not seen in C-rich material [239], where presumably most dimers are being trapped by C_sO_n complexes.

Note that there is another, electrically active, defect that forms after long time annealing, which also possesses a LVM at 1012 cm^-1. The FWHP of this new 1012 cm^-1 is much broader than the early 1012 cm^-1 line and rapidly swamps it. This is also correlated with a lower mode at 744 cm^-1. These modes come from a fundamentally different defect and should not be confused with the dimer modes. This mode overlap originally led to incorrect assignment of the 1012 cm^-1 line to some sort of shallow thermal donor centre[239].