Modeling the effect of boron on microdefect formation

Abstract

As microelectronic devices demand larger diameter wafers and reduced microdefect concentration and size, one of the most crucial issues in the semiconductor industry becomes the understanding of defect formation in silicon crystal growth by the Czochralski crystal growth method. Microdefect patterns in the crystal are known to be associated with the native point defects, namely self interstitials and vacancies. The focus of this thesis is the analysis of the Oxidation Induced Stacking Fault Ring (OSF-Ring) which is known to correspond to a neutral region where the concentrations of the self-interstitials and vacancies are essentially in balance. The dynamics of the OSF-Ring with changes in crystal growth operating conditions, crystal growth rate and temperature field, has been explained quantitatively for crystals containing only self-interstitials and vacancies (Sinno, 1998). The neutral ring separates a core of vacancy rich crystal from a self-interstitial rich external ring. Microvoids form in the core by vacancy aggregation. Dislocation loops form in the self-interstitial rich region. The presence of dopants in the crystal also affects microdefect distributions by interacting directly with self-interstitials and vacancies. At low dopant concentrations, the OSF-Ring position depends exclusively on the operating conditions. Crystal growth experiments by Dornberger et al. (1997) have shown the influence of high boron doping levels (1x10 15 - 2x10 19 cm-3 ) on the OSFRing position; high boron concentrations shift the OSF-Ring toward the center of the crystal without changes in growth conditions. The OSF-Ring diameter shrinks as a result of a shift of the point defect balance in favor of self-interstitials because of either vacancy depletion or selfinterstitial addition. The point defect dynamics model described by Sinno (1998) for silicon consists of point defect transport by Fickian diffusion and bulk crystal motion, and the consumption or generation of point defects by recombination. The crystal temperature field and crystal shape data was obtained from the heat transfer simulations of Dornberger et al. (1997b) for Czochralski crystal growth systems. The intrinsic point defect model of Sinno (1998) was extended in this thesis to include boron-point defect interactions in the crystal. The following reactions are considered: ... The objective of this thesis is to expand the model of Sinno (1998) by accounting explicitly for the formation of boron complexes by reactions with interstitials and vacancies. Conservation equations for each boron defect species were developed using thermophysical data obtained from electronic structure calculations (Rasband and Clancy, 1996; Luo, Rasband and Clancy, 1998). The model is parametrized using the experimental OSF-Ring data of Dornberger et al. (1997) which relates the position of the OSF-Ring to the doping concentration of boron in the melt. The results are used to postulate a mechanism for boron-mediated OSF-Ring dynamics. OSF-Ring dynamics is set by point defect and impurity dynamics within a narrow region near the melt-crystal interface. The effects of boron on IV dynamics can be understood qualitatively from reactions (1)-(6). Near the melt-crystal interface the complexes BI and B21 store interstitials temporarily and dampen depletion of interstitials caused by IV recombination. Then, interstitials are released by the dissolution of BI and B2I yielding higher interstitial concentrations. All of these processes occur in the region where the self-interstitial and vacancy balance is still evolving. These observations lead to the conclusion that the kick-out reaction is fast enough to consume interstitials quickly and release them back within a short axial distance to achieve lower effective interstitial diffusion and to shift the OSF-Ring inward. This phenomenon is known as the chemical pumping mechanism (Hu, S.M., 1994) and is observed in both interstitial and vacancy release mechanisms of BI, B2I and BV. These boron complexes can therefore act as traps which slow intrinsic point defect diffusion. This also means that the kick-out mechanism is assumed to be dominant and is the cause of on the OSF-Ring shift at high boron concentrations. Our extension of the microdefect model includes boron point defect chemistry described in Equations (2) to (6). The extended model is able to predict the experimental observations with only adjustment of binding entropies.