Modeling of {311} facets using a lattice kinetic Monte Carlo three-dimensional model for selective epitaxial growth of silicon

Abstract

Using lattice kinetic Monte Carlo, we provide a quantitative physically based atomistic model for the selective grown of Si-based materials, and explain {311} facet formation while remaining computationally efficient. Descriptions of the local atomistic configurations critical for the developing of {100}, {110}, {111}, and {311} facets are given. The model also introduces (a) three different microscopic growth rates for local {100} planes, needed to properly simulate the formation of perfect {100} terraces in miscut substrates and (b) a break-up energy to account for bonding during {311} facet formation. The model has been verified against experimental results.

Atomistic modeling of epitaxial growth of semiconductor materials

“…It differs from the model presented in the previous section by that four, instead of three, local configurations of Si atoms incorporation, namely {100 l }, {100 h }, {110}, and {111} are considered. The definition of two configurations for {100} helps to better control the amorphous/crystalline interface, and takes into account the number of second neighbors to distinguish between low {100 l } and high {100 h } coordinations [11]. In addition, unlike setting the normal or defective tags to the crystallized Si atoms, the model takes into account the actual geometry of twin formation during filling new lattice atoms.…”

“…Small size structures favor the application of the modeling methods that enable to obtain the evolution at atomic level, such as ab initio [7], molecular dynamics (MD) simulations [8,9], and lattice kinetic Monte Carlo (LKMC) modeling [10,11]. As compared to the ab initio and MD simulations, LKMC methods have a higher scaling potential with preservation of reasonable computation efficiency so that the systems of different sizes can be modeled at low cost.…”

Lattice kinetic Monte Carlo modeling of germanium solid phase epitaxial growth

“…As it has been observed in previous works [4,9], it is important to consider a difference in those prefactors depending on the neighbor coordination number encountered at the time of bonding into the lattice. This is modeled by introducing a lower rate in the {100} microscopic configurations when less first neighbors are found (6 and 7 in this simulation) making the bond less probable.…”

Improved physical models for advanced silicon device processing

“…To overcome the limitations of non-lattice KMC models in a computationally efficient code, Martin-Bragado et al developed a lattice KMC model that reproduces different planar SPER velocities and the formation of defects during regrowth [94,[111][112][113]. Stress information can be extracted from finite-element methods [111] and fed back and forth to the lattice KMC code.…”