Oxygen '96

Early Stages of Oxygen Precipitation in Silicon

---

MAGNETIC RESONANCE ON HEAT TREATMENT CENTRES IN SILICON

J. -M. Spaeth

Fachbereich Physik, Universität Paderborn, 33095 Paderborn, Germany

Keywords: oxygen, silicon, EPR, ENDOR, impurities Heat treatment centres, believed to be oxygen aggregate centres, having "shallow" levels in the gap, are generated by annealing oxygen containing silicon at 450 degrees C. They can be obtained in a paramagnetic charge state such that electron paramagnetic resonance methods can be used to investigate their microscopic structures. In particular, when the magnetic isotope ¹⁷O is diffused into float-zone (Fz) silicon, one can study the participation of oxygen in the heat treatment centres. There are two prominent EPR spectra associated with heat treatment centres; the so-called NL8 spectrum and the NL10 spectrum. Both spectra differ in their g-values and have rather narrow lines and show no hyperfine structure. With electron nuclear double resonance (ENDOR) we investigated NL8 and NL10 centres in various Fz or Czochralski (Cz) grown silicon single crystals, which had different acceptor dopings and impurities. Into Fz-Si ¹⁷O was diffused. It turned out that in all samples NL8 is a unique oxygen aggregate defect which does not incorporate any impurity and which is a double donor (thermal double donor TDD). From recent ENDOR measurements on ¹⁷O nuclei in various stages of the heat treatment development by annealing up to 200 hours, it was concluded that the core of the TDDs contained 2 oxygens in a Si vacancy and a Si interstitial compatible with the model proposed by Deák et al. The other oxygens are thought to be in interstitial positions between two Si atoms along <110> directions. Various different ENDOR spectra were obtained for the NL10 centres depending on the nature of the compensating acceptors or impurities present in the crystals. Therefore, it seems that several NL10 defects have different microscopic structures depending on the impurities present. Also the energy levels seem different. For example, NL10 in P-doped silicon is not very shallow. Our results support the suggestion that the EPR spectrum of NL10 centres is not characteristic for one specific microscopic structure of an oxygen aggregate centre. The proposal made in the literature that NL10 defects are correlated with the TDDs (NL8) must be discarded.

Last modified: Sat Mar 9 14:50:47 GMT 1996 JG