Developments of the kinetic Monte Carlo method giving improved accuracy and increased versatility for the description of atomic diffusivity on metal surfaces were reported. The on-lattice constraint built into a proposed self-learning kinetic Monte Carlo was relaxed, leaving atoms free to occupy 'off-lattice' positions to accommodate several processes responsible for small-cluster diffusion, peripheral atom motion and hetero-epitaxial growth. This technique combines the ideas embedded in the self-learning kinetic Monte Carlo method with a new pattern-recognition scheme fitted to an off-lattice model in which relative atomic positions were used to characterize and store configurations. Application of a combination of the 'drag' and the repulsive bias potential methods for saddle point searches allows the treatment of concerted cluster, and multiple- and single-atom, motions on an equal footing. This tandem approach had helped reveal several new atomic mechanisms which contribute to cluster migration. This off-lattice self-learning kinetic Monte Carlo method was applied to the diffusion of 2D islands of Cu (containing 2-30 atoms) on Cu and Ag(111), using the interatomic potential from the embedded-atom method. For the hetero-system Cu/Ag(111), this technique uncovered mechanisms involving concerted motions such as shear, breathing and commensurate-incommensurate occupancies. Although the technique introduces complexities in storage and retrieval, it did not introduce noticeable extra computational cost.

Off-Lattice Self-Learning Kinetic Monte Carlo: Application to 2D Cluster Diffusion on the FCC(111) Surface. A.Kara, O.Trushin, H.Yildirim, T.S.Rahman: Journal of Physics - Condensed Matter, 2009, 21[8], 084213