Atomistic simulator of polycrystalline thin film deposition in three dimensions

Abstract

Thin film deposition involves processes from 10−13 to 102 s in time scale. It is desirable, and also challenging, to include important physics of all the time scales in modeling the deposition. The hierarchical bridging of ab initio, molecular dynamics, and lattice kinetic Monte Carlo has proven feasible in incorporating important time scales of thin film deposition. However, one bottleneck is the representation of polycrystals within the lattice kinetic Monte Carlo method. A brutal force representation of N grains each with linear dimension L requires additional computer memory of order NL3, beyond that of a single crystal. Instead, we here propose and implement a memory efficient algorithm of three consecutive two-dimensional mappings to represent the N grains. As a result, the additional memory requirement is on the order of NL2. Using energetics representative of Al, we demonstrate the new implementation by simulating texture competition with and without sufficient diffusion. Our demonstrations show that the implementation (1) allows the physical representation of texture evolution, and (2) enables atomistic simulations of polycrystalline thin films of 0.15--0.20 m or larger in linear dimension, on a standalone PC in the year of 2004; without the memory efficient algorithm, this would have not been possible until a decade later.