Modeling Carbon Diffusion and its Impact on Boron Diffusion in Silicon and Silicon Germanium

Abstract

<p> The integration of silicon germanium (SiGe) in the base of heterojunction bipolar transistors (HBTs) has recently put the alloy into prominence to produce fast-switching transistors. However, the thin highly doped SiGe base makes the transistor susceptible to base dopant outdiffusion during device processing, which results in device performance degradation. Adding carbon to the base was shown to significantly suppress boron outdiffusion and help retain the narrow as-grown profile. Dopant behavior in the presence of various species needs to be well understood and modeled for two reasons: (1) to have accurate and predictive process simulators; and (2) to obtain insight into process development. </p> <p> Modeling carbon diffusion and its role in suppressing boron diffusion in silicon and SiGe has been studied by several groups. While boron diffusion is well-established, different modeling regimes have been developed for carbon diffusion. Each of the existing studies has focused on subsets of the available experimental data. We present a consistent and complete model that accounts for carbon and boron diffusion in silicon and SiGe, under equilibrium and non-equilibrium conditions. In our regime, carbon diffusion is modeled according to the kick-out and Frank-Tumbull mechanisms for diffusion; in addition, we incorporate the carbon clustering phenomenon. To completely model boron diffusion, we account for the boron-interstitial clustering (BICs) effect and the { 311} defects that are associated with boron transient enhanced diffusion (TED). In the developed model we make use of the well-established literature data for carbon diffusion, as well as boron diffusion and Si self-diffusion. The model was verified by simulating experiments that involve boron and/or carbon diffusion in silicon and SiGe and cover the complete temperature range of 750 - 1070 °C. The test structures include published experiments in addition to recent experimental results obtained through collaboration, and feature diffusion in inert and oxidizing ambients, under rapid thermal annealing (RTA) conditions, as well as in the presence of implant damage. We also investigated the validation of the model without the inclusion of either the clustering or the Frank-Turnbull reactions. </p>