A multiple level set method for modeling grain boundary evolution of polycrystalline materials

Xinwei Zhang, Jiun‐Shyan Chen, Stanley OsherInteraction 2008 · DOI

TL;DR

This paper introduces a novel "multiple level set method" to accurately simulate how grain boundaries evolve in polycrystalline materials. By employing a predictor-corrector approach and a ghost cell method for periodic boundaries, it effectively handles complex topological changes and avoids issues found in traditional front-tracking methods, offering a more robust and efficient simulation framework.

What Problem Does This Paper Address?

Polycrystalline materials, fundamental to many engineering applications, are composed of numerous small grains separated by grain boundaries. The way these grains grow and change (microstructure evolution) significantly impacts material properties like strength and conductivity. Grain boundary migration, driven by factors like curvature, is a key process in this evolution. Simulating this accurately is challenging because grain structures can undergo complex topological changes (grains merging or splitting). Traditional methods like front-tracking struggle with these changes, often requiring complex remeshing or ad-hoc rules. This paper aims to provide a more robust and computationally efficient method for modeling this evolution, particularly addressing the complexities of multi-grain interactions and topological transformations.

Methods

The core of this work is a multiple level set method. Unlike single level set methods used for two-phase problems, this approach assigns an independent level set function to each grain, treating the polycrystalline material as an N-phase system.

Key methodological contributions include:

  • Predictor-Multicorrectors Approach: To handle the interactions at triple junctions (where three grain boundaries meet), a predictor-corrector iterative process is used. First, each level set function is evolved independently (predictor phase). Then, these functions are corrected to ensure consistency and eliminate gaps at the triple junctions (corrector phase). This process is iterated to achieve grid-resolution accuracy.
  • Ghost Cell Approach for Periodic Boundary Conditions: To simulate representative macroscopic behavior from a unit cell, periodic boundary conditions are crucial. Instead of solving complex constrained problems, the authors introduce a ghost cell approach. This involves creating virtual cells outside the domain that mirror the geometry and signed distance functions of the grains on opposite sides of the unit cell, allowing for accurate local calculations without explicit global constraints.
  • Local Level Set Method: For computational efficiency, the level set function is solved only in a narrow band around the interface, reducing computational cost from O(n²) to O(n).

Datasets

Not discussed in the paper. The paper focuses on simulating material behavior based on physical principles and material parameters, rather than training on empirical datasets.

Key Insights and Ideas

  • Level Sets Naturally Handle Topology: The level set method inherently handles topological changes (like merging or splitting of interfaces) without requiring explicit geometric rules or remeshing, which is a significant advantage over front-tracking methods.
  • Triple Junctions are Key: The accurate modeling of triple junctions is critical for simulating grain boundary evolution. The predictor-multicorrectors approach effectively resolves the gaps that can arise at these junctions, leading to more accurate simulations.
  • Ghost Cells Simplify Periodicity: The ghost cell approach provides an elegant and computationally efficient way to implement periodic boundary conditions, avoiding the complexity of Lagrange multipliers or penalty methods.
  • Curvature-Driven Migration is Well-Modeled: The method accurately captures grain boundary migration driven by curvature, a primary force in grain growth.

Key Results

The paper demonstrates the effectiveness of the proposed multiple level set method through several numerical examples:

  • Topological Changes: The method successfully models both Type 1 (T1) and Type 2 (T2) topological changes in grain structures, which are crucial for accurate simulation of grain growth and recrystallization. Figures show these changes occurring naturally through the interaction of level set functions.
  • Uniform and Random Grain Structures: Simulations of both uniform (with imperfections) and random grain structures show realistic grain boundary migration and evolution. Grains with fewer than six sides tend to shrink, while those with more tend to grow, consistent with physical expectations.
  • Comparison with Front Tracking: The results are compared with a front-tracking method. While both methods show similar outcomes for uniform grain growth, the multiple level set method demonstrates superior handling of topological changes in random grain structures, avoiding errors associated with ad-hoc rules and piecewise linear approximations in front tracking.
  • Computational Efficiency: The local level set method, combined with the predictor-multicorrectors approach, offers a competitive computational time. For a random grain network case, the multiple level set method took 216 minutes, comparable to 181 minutes for the front-tracking method, but with improved accuracy in handling topological changes.

Limitations and Future Work

  • Limitations:
    • The paper focuses on curvature-induced driving forces. Other driving forces, such as those related to strain energy differences between anisotropic grains, are not considered in this work.
  • While the predictor-multicorrectors approach reduces gaps to grid resolution, it doesn't eliminate them entirely without sufficient iterations.
  • Future Work:
    • Extend the framework to three-dimensional (3D) grain growth simulations.
  • Incorporate additional driving forces, particularly those arising from anisotropic grain properties (e.g., strain energy differences).
  • Further explore the computational efficiency improvements.

Summary updated

Abstract

In this paper, we model grain boundary evolution based on a multiple level set method. Grain boundary migration under a curvature-induced driving force is considered and the level set method is employed to deal with the resulting topological changes of grain structures. The complexity of using a level set method for modeling grain structure evolution is due to its N-phase nature and the associated geometry compatibility constraint. We employ a multiple level set method with a predictor-multicorrectors approach to reduce the gaps in the triple junctions down to the grid resolution level. A ghost cell approach for imposing periodic boundary conditions is introduced without solving a constrained problem with a Lagrange multiplier method or a penalty method. Numerical results for both uniform and random grain structures evolution are presented and the results are compared with the solutions based on a front tracking approach (Chen and Kotta et al. 2004b).
BibTeX
@inproceedings{zhang2008multiple,
  author = {Xinwei Zhang and Jiun‐Shyan Chen and Stanley Osher},
  title = {{A multiple level set method for modeling grain boundary evolution of polycrystalline materials}},
  booktitle = {Interaction},
  year = {2008},
  doi = {10.12989/imm.2008.1.2.191},
}
Figure