Formation processes for N-O defects

A complication with the unravelling of the processes by which oxygen atoms aggregate in the presence of these impurities is the existence of more than one type of oxygen complex. In N-doped material the dominant nitrogen defect is the pair which itself forms electrically inactive NNO defects [103,177]. These have to be distinguished from the shallow donor defects originating from N_i, and the question then arises as to which defect is actually formed.

High temperature annealing is normally used to break up any defect complexes present in the material, and this will include N_2i pairs. On cooling, the diffusing N_i defects are trapped either by O_i or N_i. A high concentration of N_i implies that N_2i pairs will be preferably formed, whereas a high oxygen concentration may result in N_iO_i defects. Both defects subsequently complex with oxygen to create either electrically inactive NNO or STD defects in the form of NO_2i. Sun et al [169] showed that for N to have an effect on O precipitation it is essential to pre-anneal at 1100$^\circ$C; this step will be required to break up the N_2i pairs and any larger oxygen precipitates, and homogenously distribute the N_i and O_i.

It is likely that the N-pair is more stable than N_iO_i and hence slow cooling probably results in higher concentrations of NNO. Hara et. al. [190] examined the effect of cooling rate on N-rich Cz-Si, and found that in quenched Si they obtained high concentrations of STDs, whereas in slow cooled Cz-Si the STD concentration was much lower. This is consistent with the above explanation. Slow cooling will also lead to out-gassing of nitrogen, leaving less N_2i pairs to mop up O_i forming NNO. Similarly, evidence that high N concentrations do not necessarily result in high concentrations of STDs has been shown by Griffin et. al. [193] using PTIS on N doped samples annealed at 450$^\circ$C. They found that N suppressed TD growth -- behaviour similar to that of carbon doped Si where C-O complexes are formed. In addition, for low doping levels of N ($1.4 \times 10^{15}$cm^-3), they observed a large number of STDs, whereas for a higher N doping level ($9 \times 10^{15}$ cm^-3), the STD concentration was strongly reduced. This result is consistent with N_2i pairs forming preferentially at higher concentrations, since N_i atoms are able to combine before encountering any O_i.

The formation kinetics of the NO_2i defects will be dependent on the quantities of each reaction component, and the rate constants for many different reactions:

Here N_2i is the nitrogen pair defect, N_2iO is the electrically inactive NN-O defect, and N_iO_2i is the defect proposed here as a STD. In theory there could be other reaction components such as N_3i, as well as reaction paths involving other point defects such as C.