A Hybrid Molecular Dynamics Kinetic Monte Carlo Simulation Methodology
This thesis describes a novel hybrid computational methodology in which the Molecular Dynamics and Kinetic Monte Carlo methods are concurrently combined. This hybrid methodology has been developed to simulate phenomena which are unfeasible to treat with either Molecular Dynamics or Kinetic Monte Carlo alone, due to the wide range of time scales involved and the need for highly detailed atom dynamics. Is is shown that the hybrid methodology can reproduce the results of a larger (more atoms) all Molecular Dynamics simulation at a significant reduction in computational cost (run time) - due to the replacement of Molecular Dynamics atoms with Kinetic Monte Carlo atoms. The hybrid methodology has been successfully used to study the dynamics of epitaxial stacking fault grain boundaries. This work identified that grain boundary motion was hindered by atoms lodging in off-lattice sites, and also by overlayer islands built up by adatom deposition. It was verified that the ‘kink flip” move is a key element in the motion of grain boundaries. Methods for enhancing the hybrid methodology were researched. It was shown that by an optimal choice of damping parameter γ, wave reflections back into the Molecular Dynamics domain could be minimised. This is expected to enable the hybrid methodology to operate successfully with smaller Molecular Dynamics domains, making larger and/or longer simulation runs feasible. This research included the derivation of the dispersion relation for the discrete case with damping and net reflectivity formulas. These are believed to be new results. The hybrid model can be applied to a wide variety of MD and KMC methods. Other MD potentials such as Embedded Atom or Modified Embedded Atom could be employed. The KMC component can be developed to use a more refined lattice or an ”on the fly” KMC method could be employed. Both the MD and KMC components can be extended to handle more than one species of atom. Parallelised versions of the MD and KMC components could also be developed. Any situation where the problem can be decomposed into distinct domains of fine scale and coarse scale modelling respectively, is potentially suitable for treatment with a hybrid model of this design.