Surface-mediated
processes, such as epitaxial growth, heterogeneous
catalysis, and etching, are typically modeled by Kinetic Monte Carlo
(KMC) methods. Traditionally, the KMC simulations are based on a top-down
approach, where the simulation parametersthe rates for the
corresponding atomistic processesare obtained by manually
fitting the simulation output to the experiment. More recently, following
the development of Density Functional Theory (DFT), an alternative
bottom-up approach has been developed, obtaining the atomistic rates
from activation energies and attempt frequencies procured by DFT.
Nevertheless, the procedure still requires a labor-intensive fine-tuning
of the rates to improve the match between simulation and experiment.
Accordingly, we propose to modify the traditional top-down and bottom-up
approaches by automating the search of the atomistic rates with the
help of an evolutionary algorithm. On the basis of a power spectral
density analysis of both the experimental and simulated images, the
procedure is applied to characterize wet etching of silicon and epitaxial
growth of silver as examples of typical surface-mediated processes.