Multicontinuum homogenization. General theory and applications

Eric T. Chung, Yalchin Efendiev, Juan Galvis, Wing Tat LeungJournal of Computational Physics 2024 · DOI

TL;DR

This paper introduces a general framework for multicontinuum homogenization, a technique used to model complex materials with multiple scales and high contrasts. The method develops a systematic way to derive macroscopic equations by considering multiple "continua" (distinct material behaviors) at each point in the material, offering a more robust approach than traditional homogenization for challenging problems.

What Problem Does This Paper Address?

Many real-world problems, such as fluid flow in porous media, involve materials with intricate structures at multiple scales and significant variations in properties (high contrast). Simulating these phenomena directly on a fine grid is computationally prohibitive. Traditional homogenization techniques, which average properties over a representative volume, often struggle when material properties vary drastically or depend on the scale itself. This paper addresses the challenge of developing a general homogenization framework that can accurately capture the behavior of such complex, high-contrast media by explicitly accounting for multiple distinct material behaviors (continua) at each macroscopic point.

Methods

The core of the proposed method is a novel expansion of the solution that represents it as a combination of multiple macroscopic variables and associated "cell problems." These cell problems are local, microscopic problems solved within representative volume elements (RVEs). The key innovation lies in:

  • Multicontinuum Expansion: The solution is expressed as a product of multiple macroscopic variables and their corresponding microscopic functions derived from cell problems. This allows for capturing different material behaviors simultaneously.
  • Cell Problem Formulation: Cell problems are formulated as saddle-point problems with integral constraints. These constraints are crucial for ensuring that the macroscopic variables accurately represent the average behavior of each continuum. The paper details the formulation of cell problems for constants, linear, and quadratic behaviors.
  • Integral Localization: A technique to localize integrals over coarse computational domains to RVEs, leveraging the RVE concept.
  • Piecewise Smooth Approximation: Macroscopic variables are assumed to be smooth functions, allowing them to be factored out of integrals over RVEs.

The derivation then involves substituting this expansion into the original fine-scale equations and applying these localization and approximation steps to derive a system of macroscopic equations. The paper also discusses the connection to mixture theories and presents applications to elliptic equations and first-order mixed systems.

Datasets

Not discussed in the paper. The paper focuses on theoretical development and numerical examples derived from defined mathematical problems, not on empirical datasets.

Key Insights and Ideas

  • Beyond Single Continuum: The fundamental insight is that for highly heterogeneous and contrasting materials, a single macroscopic variable might not be sufficient. Instead, multiple "continua" (distinct material behaviors) need to be considered at each macroscopic point.
  • Cell Problems as Building Blocks: The cell problems are not just for averaging; they are designed to capture specific behaviors (constant, linear, quadratic) of each continuum, providing the microscopic basis for the macroscopic description.
  • Handling High Contrast: The framework is explicitly designed to handle high contrast, a significant limitation of many existing homogenization methods. This is achieved by carefully formulating the cell problems and their constraints.
  • General Framework: The paper aims to provide a general, systematic approach that can be applied to various types of linear systems, including scalar elliptic equations and mixed systems, offering a unified methodology.

Key Results

The paper presents a formal derivation of the multicontinuum homogenization framework and demonstrates its applicability through numerical examples.

  • Generalized Macroscopic Equations: The method yields macroscopic equations that include additional terms compared to standard homogenization, reflecting the influence of multiple continua and high contrast.
  • Numerical Validation: Numerical examples for scalar elliptic equations and mixed systems show good convergence as the mesh size decreases. The results indicate that the proposed approach provides accurate approximations of averaged solutions, even in challenging high-contrast scenarios.
  • Error Reduction: The numerical results demonstrate that the error remains small when an appropriate number of "layers" (related to the oversampling in cell problems) are used, suggesting the robustness of the method.

Limitations and Future Work

  • Cell Problem Formulation Complexity: The paper acknowledges that formulating cell problems correctly, especially to guarantee the decay of local solutions, is a challenging task. This is crucial for accurate approximations, particularly in non-symmetric problems.
  • Definition of Continua: The current framework assumes that continua are associated with pre-defined subregions. The authors suggest that future work could explore using spatial functions or local spectral problems to define these continua, similar to nonlocal multicontinuum approaches.
  • High Contrast Implications: While the method is designed for high contrast, the paper notes that in certain high-contrast regimes, the multicontinuum equations might simplify to single-continuum homogenization. Further investigation into these transitions is implied.
  • Extension to Non-linear Problems: The current work focuses on linear systems. Extending the framework to nonlinear problems would be a natural and important next step.

Summary updated

BibTeX
@article{chung2024multicontinuum,
  author = {Eric T. Chung and Yalchin Efendiev and Juan Galvis and Wing Tat Leung},
  title = {{Multicontinuum homogenization. General theory and applications}},
  journal = {Journal of Computational Physics},
  year = {2024},
  doi = {10.1016/j.jcp.2024.112980},
}
Figure