2100
Dynamical Imaging of Nickel Disilicide Nucleation and Step Flow Propagation in Defect-Engineered Si Nanowire

Tuesday, 7 October 2014: 11:50
Expo Center, 1st Floor, Universal 18 (Moon Palace Resort)
W. Tang (University of California Los Angeles), S. T. Picraux (Los Alamos National Laboratory), A. M. Gusak (Department of Theoretical Physics, Cherkasy National University), K. N. Tu (University of California, Los Angeles), and S. A. Dayeh (University of California San Diego)
The Ni silicides and Si metal-semiconductor contacts are a vital element in the state-of-the-art commercial transistor devices. The contact formation process needs to be precisely engineered in order to achieve reproducible electrical properties over billions of transistors, especially for ultra-scaled transistors when contact resistance contribution is significant. The latest source/drain (S/D) engineering technology designs the S/D crystal structure (e.g. intentional incorporation of stacking faults) that strains the channel to enhance device performance. Understanding the role of structural alternation, or defects, in Si-Ni reaction is therefore important in achieving precise control of the contact formation process at an atomic scale. Here, we present a study of Si-Ni reaction by lattice-resolved in-situ transmission electron microscopy (TEM) and found that presence of defects in Si can fundamentally change the silicide nucleation mechanism and growth behavior.

We use vapor-liquid-solid (VLS) grown Si nanowire (NW) as the Si source and incorporate two types of defects (twin boundaries (TB) running down NW axial direction and grain boundaries (GB) on NW surface). Si NWs are dispersed on silicon nitride (50nm) TEM membrane and react with photolithography defined Ni electrodes, at 300C on an in-situ TEM heating stage. Dynamic high resolution TEM imaging resolves the axial growth of NiSi2 into Si NW as layer-wise repeating nucleation and propagation of NiSi2 (111) plane. Previously reported studies show 2D Ni silicides nucleate homogeneously in the center of defect-free NWs (Figure a-b), and the growth proceeds in a layer-by-layer manner with long incubation time between sequential layer growth (Figure c). We captured the nucleation of each NiSi2 layer at the initial stage and demonstrated that TB and GBs are preferred heterogeneous nucleation sites. An example of nucleation and step propagation from the TB is shown in Figure d. Experimentally measured growth rate (Figure e) qualitatively fit with our analytical model based on classic nucleation theory, which consider the energy of different defects. Defective sites with the heterogeneous nucleation barrier in descending order are TB, "corner", and GB, (Figure f) suggesting that higher degree of defectivity suppresses the nucleation barrier more effectively. We further extend our model to depict NiSi2 nucleation as a biased random walk stochastic process and show that twin defects in NiSi2have high interfacial energies (in contrast to other typical  FCC materials like copper or silicon). We calculate this defect energy, which together with our experimental observations, should alert and guide contact silicide formation processes and strain engineering in future Si-based technology nodes.

Figure captions: (a-b) homogeneous nucleation of Ni silicide in a single crystal Si NW (c) Layer-by-layer repeating homogeneous nucleation in silicide growth (d) NiSi2 heterogeneously nucleates from the “corner” site. Scale bar is 3 nm. (e) The silicide grows at a steady rate when guided by the TB. Inset shows layer-by-layer growth behaviour in the heterogeneous nucleation NiSi2 growth. (f) Reduction in the nucleation barrier at different heterogeneous sites noted in the inset schematic which is a 3D view of three different heterogeneous nucleation sites in a defective Si nanowire.