Boron segregation in silicon germanium carbide and silicon carbide alloys and application to p-channel MOSFETs

Si_1−x−yGe_xC_y alloys are of great interest for scaled Si-based devices. Adding germanium to silicon allows for the adjustment of both bandgap and strain, making possible devices such as the heterojunction bipolar transistor (HBT) and strained-Si MOSFET. Introducing small amounts of substitutional carbon can also adjust strain and bandgap, and has been shown to dramatically reduce the diffusion of boron and phosphorus atoms, allowing for increased control over dopant profiles. This has been important for the scaling of HBTs. Several factors, however, limit the usefulness of carbon, and many other issues remain unexplored. Carbon has a very low solubility in silicon and Si_1−xGe_x and can form undesirable silicon carbide (SiC) precipitates, degrading the properties of the films. Possible atomic interactions between carbon and other dopants during thermal processing are largely unknown, and may have important technological implications for diffusion and electrical properties. There has also been very little investigation of carbon in polycrystalline silicon films.; This work addresses several of these issues. First, a new application of Si_1−x−yGe_xC_y alloys for controlling dopant diffusion is presented, that of as a gate material for scaled p-channel MOSFETs. Devices with thin layers of polycrystalline Si_1−x−y Ge_xC_y in the gate are shown to have increased threshold voltage stability vs. conventional devices and good electrical characteristics. Boron segregates to the polycrystalline Si_1−x−yGe _xC_y layers during post-implant anneals, reducing diffusion out of the gate and into the substrate. This new boron segregation effect is then examined in detail. Boron is found to segregate to both polycrystalline and single-crystal Si_1−x−yGe_xC_y and Si_1−yC_y (no germanium) during thermal anneals. Electrical measurements show no boron deactivation for low carbon concentrations, even with extensive annealing, indicating that inactive boron tied up with carbon-related defects is not driving the segregation. In addition, diffusion experiments indicate that boron is not becoming immobilized at carbon-related traps. Gradients of interstitial silicon, caused by substitutional carbon, are proposed as a driving force for the boron segregation. This hypothesis is supported by simulation results using coupled point defect-dopant diffusion models.