An investigation of the possibility that oxygen diffusion in Czochralski silicon is catalyzed during clustering

The rates of loss of isolated oxygen from solution at the lowest temperatures (T < 400°C) are close to those expected for O_i-O_i interaction in a dilute solution, yet the dependence of rates on [O_i] at slightly higher temperatures (400 < T < 500°C) indicates that the predominant clusters formed contain many more then two atoms. The implication that the kinetics of O-clustering at low temperatures are anomalous coincides with the assumption that thermal donors consist of different numbers of clustered O-atoms. Attribution of the implied fast evolution of a series of clusters to rapid dimer diffusion does not appear to be consistent with observations relating to the defects subsequently detected. Even if aspects of this evolution suggest the involvement of self-interstitials ejected during clustering, no reasonable way has been found that these defects might account for the serial evolution of large O-clusters. A natural explanation of growth of a series of relatively large clusters is that a barrier to dimerization exists and that the rate of O_i-O_i interaction is much greater than that of [O_i]-loss. The assumption of a dilute solution seems to be questionable.