Random Walk on Lattices: Graph Theoretic Approach to Modeling Epitaxially Grown Thin Film

TL;DR

This paper introduces a graph theoretic model for simulating epitaxial thin film growth with long atomic diffusion paths, enabling efficient surface morphology predictions with linear computational complexity.

Contribution

It presents a novel Markovian approach to model adatom diffusion on lattices during thin film growth, reducing computational cost and capturing complex surface features.

Findings

Efficient linear-time algorithm for surface morphology calculation.

Successful simulation of thin film growth on flat and dislocated substrates.

Observation of a rectangular spiral ridge with smooth front features.

Abstract

Immense interests in thin-film fabrication for industrial applications have driven both theoretical and computational aspects of modeling its growth with an aim to design and control film's surface morphology. Oftentimes, smooth surface is desirable and is experimentally achievable via molecular-beam epitaxy (MBE) growth technique with exceptionally low deposition flux. Adatoms on the film grown with such a method tend to have large diffusion length which can be computationally very costly when certain statistical aspects are demanded. We present a graph theoretic approach to modeling MBE grown thin film with long atomic mean free path. Using Markovian assumption and given a local diffusion bias, we derive the transition probabilities for a random walker to traverse from one lattice site to the others after a large, possibly infinite, number of hopping steps. Only computation with linear-time complexity is required for the surface morphology calculation without other probabilistic measures. The formalism is applied to simulate thin film growth on a two-dimensional flat substrate and around a screw dislocation under the modified Wolf--Villain diffusion rule. A rectangular spiral ridge is observed in the latter case with a smooth front feature similar to that obtained from simulations using the well-known multiple registration technique. An algorithm to compute the inverse of a class of sub-stochastic matrices is derived as a corollary.