TL;DR
This paper introduces novel high-order discontinuous cut finite element methods (CutFEM) for solving linear hyperbolic conservation laws with stationary and moving interfaces. These methods are crucial for accurately simulating phenomena like wave propagation in heterogeneous materials, offering robust and efficient solutions without requiring mesh alignment to interfaces.
What Problem Does This Paper Address?
Simulating physical phenomena governed by hyperbolic conservation laws often involves interfaces where material properties change abruptly. Traditional finite element methods (FEM) typically require meshes to conform to these interfaces for accuracy and stability. This becomes computationally challenging and complex when interfaces are moving or when dealing with complex geometries. Standard discontinuous Galerkin (DG) methods, while powerful, also face these meshing challenges. This paper addresses the need for robust, high-order numerical methods that can handle these interface complexities efficiently, particularly when the mesh is unfitted to the interface.
Methods
The authors extend the discontinuous Galerkin (DG) framework to develop CutFEM for hyperbolic conservation laws. The core idea is to use a fixed, quasi-uniform background mesh that is independent of the interface's position.
- Cut Finite Element Method (CutFEM): This approach allows elements to be cut by the interface. Interface conditions are imposed weakly, ensuring conservation and stability.
- Discontinuous Galerkin (DG) Framework: The methods are built upon the DG framework, which uses discontinuous piecewise polynomial spaces for trial and test functions. This inherently handles discontinuities and allows for high-order accuracy.
- Stationary Interfaces: For stationary interfaces, a space-time CutFEM is proposed. Interface conditions are weakly enforced through penalty terms in the weak form.
- Moving Interfaces: A space-time CutFEM is developed, employing discontinuous elements in both space and time. This formulation is computationally more demanding but offers a unified approach. To mitigate this, a hybrid approach is suggested where the space-time CutFEM is localized to the interface region, coupled with a standard explicit DG method elsewhere.
- Stabilization: Ghost penalty stabilization terms are incorporated to ensure stability, especially for elements cut by the interface, and to control the condition number of the resulting linear systems.
- Weak Enforcement of Interface Conditions: Interface conditions are imposed weakly, which is a key feature of CutFEM, allowing for flexibility in mesh design.
Datasets
Not discussed in the paper. The paper focuses on numerical methods and their theoretical analysis, validated through numerical examples with known exact solutions or established benchmark problems.
Key Insights and Ideas
- Unfitted Meshes are Viable for High-Order Methods: The paper demonstrates that high-order accuracy and stability can be achieved even when the computational mesh is not aligned with material interfaces, a significant advantage for complex simulations.
- Weak Enforcement of Interface Conditions is Key: Imposing interface conditions weakly through penalty terms is crucial for maintaining conservation and stability, especially in the presence of cut elements.
- Space-Time CutFEM for Moving Interfaces: The development of a space-time CutFEM provides a unified framework for handling moving interfaces, although it comes with increased computational cost. The proposed localization strategy offers a practical compromise.
- Conservation is Achieved Through Careful Penalty Selection: The choice of penalty parameters at the interface is critical for ensuring both conservation and stability. The paper provides specific conditions for these parameters.
- Stabilization Controls Mass Matrix Conditioning: The ghost penalty stabilization plays a vital role in keeping the mass matrix well-conditioned, even for cut elements, preventing severe time-step restrictions.
Key Results
The paper presents theoretical analysis and numerical results for both stationary and moving interfaces.
- Accuracy: Numerical experiments show that the proposed CutFEM methods achieve optimal order of accuracy for both scalar and acoustic problems, comparable to standard DG methods on fitted meshes. For instance, in the scalar problem with a stationary interface, the L² and L∞ errors exhibit convergence rates consistent with the polynomial degree used.
- Conservation: The methods are shown to be conservative, especially when the penalty parameters satisfy specific conditions derived from the conservation laws. Numerical results for conservation error demonstrate this, often showing errors on the order of machine epsilon. For example, Figure 4 shows that when the conservation condition is met, the error is very small, while violating it leads to significant errors.
- Stability: Theoretical analysis (e.g., Theorem 3.1 and 3.4) establishes the stability of the proposed schemes, showing that the energy does not grow with time under certain conditions on the penalty parameters.
- Efficiency: While the space-time CutFEM for moving interfaces is implicit and computationally demanding, the paper suggests a locally implicit approach that balances accuracy and efficiency by localizing the complex space-time treatment to the interface region.
Limitations and Future Work
- Temporal Accuracy Degradation: The authors note that temporal accuracy can sometimes be degraded in coupled cases (e.g., locally implicit methods), and further work is needed to understand and mitigate this.
- Generalization to Systems and Multi-Dimensions: Extending the coupling between space-time elements and method-of-lines DG methods to systems of equations and to multiple spatial dimensions, while maintaining stability, accuracy, and conservation, is identified as a challenging but important future direction.
- Explicit Time-Stepping: The implicit nature of the space-time CutFEM is a drawback. Future work could explore ways to achieve explicit time-stepping in large parts of the domain.
- Suboptimal Error Estimate: While numerical results show optimal accuracy, the derived theoretical error estimate is noted as suboptimal. Further refinement of the analysis could potentially yield tighter bounds.
Summary updated