A genetic algorithm based Kinetic Monte Carlo simulation for the evolution of complex surface in anisotropic wet etching

Abstract

A new genetic algorithm (GA) based Kinetic Monte Carlo (KMC) method is developed for atomistic simulations of the evolution of complex multivalued surfaces appearing during anisotropic etching of crystalline silicon. In previous KMC studies the atom-specific rates are calibrated by matching the surface morphologies but the orientation-dependence of the etch rate is described correctly in few etching conditions [1]. By combining a genetic algorithm with the KMC method, the simulation converts the experimental macroscopic etch rates into atomistic Monte Carlo removal probabilities. The optimized etch rates of a group of etching conditions, i.e. KOH and KOH/IPA at different concentrations and temperatures, show good agreement with the experiments. In addition, since the atomistic reactivity function used by the KMC model uses 5 parameters to control all the atomistic removal rates, a small set of silicon orientations is sufficient to carry out the GA optimization process while effectively fitting the etch rates of a wide range of {hkl} planes. Moreover the underlying octree based model has the ability to generate hexahedral element meshes for the integration between process simulator and FEA performance analysis tool.