A structure-preserving discontinuous Galerkin scheme for the Cahn-Hilliard equation including time adaptivity

Golo A. Wimmer, Ben S. Southworth, Qi TangJournal of Computational Physics 2024 · DOI · OpenReview

TL;DR

This paper introduces a novel structure-preserving numerical method for simulating the Cahn-Hilliard equation, a fundamental model for phase separation. By combining a discontinuous Galerkin (DG) spatial discretization with adaptive time-stepping and a robust preconditioning strategy, the authors achieve accurate, stable, and efficient simulations, particularly in challenging advection-dominated regimes.

What Problem Does This Paper Address?

The Cahn-Hilliard equation is crucial for modeling phenomena like spinodal decomposition and microstructure evolution. However, simulating it presents significant challenges:

  • Stiffness: The equation involves a biharmonic operator, leading to very stiff behavior that requires implicit time-stepping.
  • Multiple Time Scales: The dynamics can exhibit both very fast and very slow processes, necessitating adaptive time-stepping for efficiency.
  • Structure Preservation: The equation has fundamental physical properties like mass conservation and energy dissipation, which numerical methods must preserve to yield physically meaningful results.
  • Advection Dominance: When coupled with fluid flow, advection terms can become dominant, leading to numerical instabilities and accuracy degradation if not handled properly.

Existing methods often struggle to balance these requirements. For instance, standard DG methods might introduce penalty terms that break structure preservation, while continuous Galerkin (CG) methods can suffer from accuracy issues in advection-dominated scenarios. This paper aims to bridge this gap by developing a DG scheme that inherently preserves the equation's structure while being robust and efficient.

Methods

The authors propose a novel structure-preserving discontinuous Galerkin (DG) spatial discretization for the Cahn-Hilliard equation. Key aspects include:

  • Mixed Formulation: The equation is reformulated using a mixed approach, introducing an auxiliary variable for the chemical potential.
  • Compatible Finite Element Spaces: The discretization employs compatible finite element spaces that form a discrete de-Rham complex. This allows for a stable, mixed finite element discretization of the Laplacian operators without requiring facet integrals or penalty terms, thus preserving structure.
  • Auxiliary Flux Variables: Two auxiliary variables are introduced for the (negative) discrete phase field flux and the discrete phase field's gradient. These are discretized in divergence-conforming spaces.
  • Upwinding for Advection: For the advection term (when coupled with flow), classical DG upwinding is used, which provides necessary transport stabilization.
  • Adaptive Time-Stepping: The scheme incorporates the TR-BDF2 method, a second-order, L-stable, implicit Runge-Kutta method. This method provides an embedded solution, enabling local error estimation and adaptive time-step control.
  • Robust Preconditioning: A preconditioning strategy is developed for the resulting implicit nonlinear systems. This strategy involves a row-swapped reformulation of the Jacobian and a block diagonal preconditioner, designed to be robust and scalable.

Datasets

Not discussed in the paper. The paper focuses on numerical methods and uses manufactured solutions and standard test cases for evaluation.

Key Insights and Ideas

  • DG for Structure Preservation: The core insight is that by carefully choosing compatible finite element spaces and a mixed formulation, a DG scheme can be constructed that naturally preserves the Cahn-Hilliard equation's mass conservation and energy dissipation properties, avoiding the pitfalls of traditional penalty-based DG methods.
  • Duality with CG: The proposed DG scheme is shown to be the dual of a standard conforming mixed CG-based discretization with respect to the de-Rham complex. This theoretical connection provides a strong foundation for understanding its properties.
  • Adaptive Time-Stepping for Efficiency: The adaptive time-stepping is crucial for efficiently handling the vastly different time scales present in phase separation dynamics, allowing the solver to take large steps during slow evolution and small steps during rapid events like mergers.
  • Robust Preconditioning for Implicit Solves: Developing a robust preconditioner is essential for efficiently solving the large, implicit systems arising from the DG discretization and implicit time-stepping, especially in challenging regimes.

Key Results

  • Accuracy: The numerical tests demonstrate the expected convergence rates for the DG discretization, consistent with theoretical predictions for mixed finite element methods.
  • Structure Preservation: The scheme successfully preserves mass conservation (with errors on the order of 10⁻¹¹) and energy dissipation properties across various test cases.
  • Stability: The DG upwinding provides good stability in advection-dominated regimes, leading to improved accuracy compared to CG methods in such scenarios.
  • Adaptive Time-Stepping Performance: The adaptive time-stepping successfully handles rapid events (mergers/absorptions) by reducing step sizes and efficiently steps over slow dynamics, showing time step variations over several orders of magnitude.
  • Solver Robustness: The preconditioning strategy proves robust, with iteration counts remaining manageable even with increasing spatial refinement or smaller diffusion coefficients.
  • Advection-Dominated Regime: The method accurately captures the formation of filament structures in the presence of strong shear flow, a behavior expected for advection-dominated phase separation.

Limitations and Future Work

  • Constant Diffusion Coefficient: While the paper notes that the structural properties hold for non-constant diffusion, the derivation of the preconditioning strategy relies on the assumption of a constant diffusion coefficient. Extending the preconditioning to handle variable diffusion coefficients robustly is a potential area for future work.
  • Coupling to Compressible/Incompressible Flows: The paper briefly discusses coupling to fluid flow but focuses on a predefined velocity field. Future work could involve fully coupling the Cahn-Hilliard scheme with Navier-Stokes equations, potentially using DG formulations for the flow field.
  • Non-dimensionalization: The paper mentions that the Cahn-Hilliard equations can be cast in a non-dimensional form and that the preconditioning strategy works well for the chosen parameters. However, for very strongly advection-dominated cases, the current preconditioning might need adjustments, potentially involving methods like AIR for transport-dominated problems.

Summary updated

Abstract

We present a novel spatial discretization for the Cahn-Hilliard equation including transport. The method is given by a mixed discretization for the two elliptic operators, with the phase field and chemical potential discretized in discontinuous Galerkin spaces, and two auxiliary flux variables discretized in a divergence-conforming space. This allows for the use of an upwind-stabilized discretization for the transport term, while still ensuring a consistent treatment of structural properties including mass conservation and energy dissipation. Further, we couple the novel spatial discretization to an adaptive time stepping method in view of the Cahn-Hilliard equation's distinct slow and fast time scale dynamics. The resulting implicit stages are solved with a robust preconditioning strategy, which is derived for our novel spatial discretization based on an existing one for continuous Galerkin based discretizations. Our overall scheme's accuracy, robustness, efficient time adaptivity as well as structure preservation and stability with respect to advection dominated scenarios are demonstrated in a series of numerical tests.
BibTeX
@article{wimmer2024structure,
  author = {Golo A. Wimmer and Ben S. Southworth and Qi Tang},
  title = {{A structure-preserving discontinuous Galerkin scheme for the Cahn-Hilliard equation including time adaptivity}},
  journal = {Journal of Computational Physics},
  year = {2024},
  doi = {10.1016/j.jcp.2025.114097},
}
Figure