Effect of spatial defect distribution on the electrical behavior of prominent vacancy point defects in swift-ion implanted Si

Samples of epitaxially grown n-type silicon have been implanted at room temperature with low doses (10(6)-10(9) cm(-2)) of He, C, Si, and I ions using energies from 2.75 to 46 MeV. Deep level transient spectroscopy studies reveal that the generation of divacancy (V(2)) and vacancy-oxygen (VO) pairs has a distinct ion mass dependence. Especially, the doubly negative charge state of the divacancy, V(2)(=/-), decreases in intensity with increasing ion mass compared to that of the singly negative charge state of the divacancy, V(2)(-/0). In addition, the measurements show also a decrease in the intensity of the level assigned to VO compared to that of V(2)(-/0) with increasing ion mass. Carrier capture cross-section measurements demonstrate a reduction in the electron capture rate with increasing ion mass for all the three levels V(2)(-/0), V(2)(=/-), and VO; but a gradual recovery occurs with annealing. Concurrently, the strength of the V(2)(-/0) level decreases in a wide temperature range starting from below 200 degrees C, accompanied by an increase in the amplitudes of both the VO and V(2)(=/-) peaks. In order to account for these results a model is introduced where local carrier compensation is a key feature and where two modes of V(2) are considered: (1) V(2) centers located in regions with a high defect density around the ion track (V(2)(dense)) and (2) V(2) centers located in regions with a low defect density (V(2)(dilute)). The V(2)(dense) fraction does not give any contribution to the V(2)(=/-) signal due to local carrier compensation, and the amplitude of the V(2)(=/-) level is determined by the V(2)(dilute) fraction only. The spatial distributions of defects generated by single-ion impacts were simulated by Monte Carlo calculations in the binary collision approximation, and to distinguish between the regions with V(2)(dense) and V(2)(dilute) a threshold for the defect generation rate was introduced. The model is shown to give good quantitative agreement with the experimentally observed ion mass dependence for the ratio between the amplitudes of the V(2)(=/-) and V(2)(-/0) peaks. In particular, the threshold value for the defect generation rate remains constant (similar to 1.2 vacancies/ion/A) irrespective of the type of ion used, which provides strong evidence for the validity of the model. Annealing at temperatures above similar to 300 degrees C is found to reduce the spatial localization of the defects and migration of V(2) occurs with subsequent trapping by interstitial oxygen atoms and formation of divacancy-oxygen pairs.