Level-set methods for the simulation of epitaxial phenomena

Mark F. Gyure, Christian Rätsch, Barry Merriman, Russel E. Caflisch, Stanley Osher, Jennifer J. Zinck, Dimitri D. VvedenskyPhysical 1998 · DOI

TL;DR

This paper introduces a novel "island dynamics" model for simulating epitaxial growth, leveraging the level-set method to efficiently track evolving island boundaries. This approach bridges the gap between atomistic simulations and continuum models, enabling the study of large-scale phenomena while retaining crucial details of island formation and coalescence.

What Problem Does This Paper Address?

Epitaxial growth, the process of depositing thin films layer by layer onto a substrate, is fundamental to modern electronics and materials science. However, accurately simulating this process is incredibly challenging due to the vast range of length and time scales involved. Atomistic processes like individual atom hopping and nucleation occur at the nanoscale, while the resulting film morphology can span microns or more.

Current simulation methods fall into two main camps:

  • Analytic/Continuum Models: These, like rate equations or partial differential equations for surface height, are computationally efficient and good for understanding large-scale behavior (e.g., scaling, stability). However, they struggle to capture the fine details of surface morphology and are often unsuitable for describing atomic-scale roughness or the initial stages of island formation.
  • Kinetic Monte Carlo (KMC) Simulations: These models can incorporate a wide range of atomistic processes and their rates can, in principle, be derived from first-principles calculations. However, KMC simulations are computationally expensive and typically limited to small system sizes and short timescales, making it difficult to study phenomena relevant to practical device fabrication.

A significant gap exists in methods that can seamlessly bridge these scales, particularly for complex scenarios like heteroepitaxial growth (where lattice mismatch introduces strain) or vapor-phase epitaxy (which involves fluid dynamics).

Methods

The core innovation presented is the application of the level-set method to model epitaxial growth. Here's how it works:

  • Level-Set Representation: Instead of explicitly tracking the jagged boundaries of islands, the level-set method represents each boundary as the zero contour of a smooth, higher-dimensional function (denoted by $\phi$). This function implicitly defines the interface.
  • Evolution Equation: The motion of the boundary is governed by a partial differential equation that describes how the level-set function $\phi$ evolves over time. This equation is essentially a conservation law, where the velocity of the interface is incorporated.
  • Island Dynamics Model: The authors propose a specific model for epitaxial growth based on this framework:
  • Continuum in Lateral Directions: The film's lateral dimensions (e.g., x-y plane) are treated as continuous. This allows for the simulation of large-scale island structures.
  • Discrete in Growth Direction: The model retains a discrete aspect in the growth direction, effectively treating layers. Overhangs and undercuts are prevented by using a single-valued function $\phi$ for all layers.
  • Adatom Density: A simplified model assumes a spatially uniform adatom (surface atom) density, which influences island growth.
  • Island Velocity: The velocity of island boundaries is made proportional to the local adatom density (Eq. 2).
  • Nucleation: New islands nucleate when the adatom density reaches a certain threshold, with their location chosen randomly.
  • Numerical Technique: The level-set equation is solved numerically using high-order accuracy, essentially non-oscillatory (ENO) methods, which are crucial for handling the complex topological changes (coalescence, splitting) that occur during growth without introducing spurious oscillations.

Datasets

Not discussed in the paper. The paper focuses on developing and demonstrating a simulation methodology rather than training on or evaluating against specific experimental datasets.

Key Insights and Ideas

  • Bridging the Scale Gap: The level-set method offers a powerful way to unify atomistic and continuum descriptions of growth. It allows for the efficient tracking of complex topological changes (like island coalescence) that are computationally prohibitive for pure atomistic methods, while still capturing the essential physics of island formation.
  • Implicit Interface Tracking: Representing interfaces as level sets inherently handles topological changes (merging, splitting) without requiring explicit re-meshing or complex algorithms, which is a significant advantage over other boundary tracking methods.
  • Flexibility for Complex Physics: The framework is designed to be extensible. By coupling the level-set evolution to other field equations (e.g., for strain or fluid dynamics), it can accommodate more realistic scenarios like heteroepitaxial growth or vapor-phase epitaxy.
  • Quantitative Agreement with Rate Equations: The authors demonstrate that their level-set "island dynamics" model, despite its different underlying mathematical approach, quantitatively reproduces results from established "real space" rate equation models for simple growth scenarios. This validates the accuracy of their numerical implementation.

Key Results

  • The paper successfully implements a level-set based "island dynamics" model for epitaxial growth.
  • The model can simulate the evolution of island morphologies, including nucleation, growth, and coalescence, across multiple layers.
  • A comparison with "real space" rate equations shows excellent agreement in scaled island size distributions, validating the level-set approach for this basic model.
  • The study shows that island density scales as $(D/F)^{-1/2}$ at low coverages, consistent with theoretical predictions and KMC simulations for the pre-aggregation regime.

Limitations and Future Work

  • Basic Model Assumptions: The presented model uses simplified assumptions, such as a spatially uniform adatom density and irreversible attachment.
  • Future Work: The authors explicitly state that their immediate future work involves solving a diffusion equation for the adatom density and using the density gradient at island edges to determine boundary velocity. This will allow for more realistic descriptions of adatom diffusion. They also envision extending the model to incorporate local strain fields for heteroepitaxial growth and hydrodynamic fields for VPE. The potential for incorporating crystal symmetries into the boundary velocity is also highlighted.

Summary updated

Abstract

We introduce a model for epitaxial phenomena based on the motion of island boundaries, which is described by the level-set method. Our model treats the growing film as a continuum in the lateral direction, but retains atomistic discreteness in the growth direction. An example of such an ``island dynamics'' model using the level-set method is presented and compared with the corresponding rate equation description. Extensions of our methodology to more general settings are then discussed.
BibTeX
@inproceedings{gyure1998level,
  author = {Mark F. Gyure and Christian Rätsch and Barry Merriman and Russel E. Caflisch and Stanley Osher and Jennifer J. Zinck and Dimitri D. Vvedensky},
  title = {{Level-set methods for the simulation of epitaxial phenomena}},
  booktitle = {Physical},
  year = {1998},
  doi = {10.1103/physreve.58.r6927},
}
Figure