Self-Learning Kinetic Monte Carlo: From 2D on-lattice to 3D off-lattice and beyond

Abdelkader Kara
University of Central Florida

I will present the developments in the Self-Learning Kinetic Monte Carlo (SLKMC) with pattern-recognitions going from 2D fcc-on-lattice (SLKMC1) to 3D-off-lattice (SLKMC4) via SLKMC2: 2D fcc/bcc-on-lattice and 2D-off-lattice (SLKMC3). Applications of the SLKMC series include 2D island diffusion and coalescence. On-lattice as well as off-lattice versions of SLKMC were used to study homo and hetero systems. I will present results on early stages of Ag (Cu) 2D-island diffusion and coarsening on Ag(111) (Cu(111)) at different temperatures and island size distributions. The activation barriers were calculated using semi-empirical interaction potentials based on the embedded-atom method. For large island sizes, we find that the diffusion coefficient follows a power law. Small islands however do not follow this law, due to their tendency to collectively diffuse. A crossover between diffusion due to collective motion to that due to periphery motion will be discussed. During early stages of island diffusion, we found that coarsening proceeds as a sequence of islands of selected sizes, resulting in peaks and valleys in the island-size distribution dictated by the relative energetics of edge-atom diffusion and detachment/attachment processes.
Finally, I will present recent development of a 3D off-lattice SLKMC using a robust and novel pattern recognition scheme. A scheme to extend the 3D-off-lattice SLKMC to systems with diverse atomic sizes will be briefly introduced

Back to Workshop II: Atomistic and Mesoscale Modeling of Materials Defects