TL;DR
This paper introduces a new "invariant-region-preserving" (IRP) limiter for high-order Discontinuous Galerkin (DG) methods applied to multi-dimensional hyperbolic conservation laws. This limiter ensures that numerical solutions stay within physically meaningful bounds (invariant regions), crucial for systems like the compressible Euler equations, without sacrificing accuracy for smooth solutions.
What Problem Does This Paper Address?
Hyperbolic conservation laws, which describe phenomena like fluid flow and wave propagation, often have solutions that must remain within specific physical constraints. For example, density and pressure must be positive. While numerical methods aim to approximate these solutions, they can sometimes produce unphysical results (e.g., negative density) that violate these constraints. This is particularly challenging for high-order DG methods, which are known for their accuracy but can be prone to oscillations near discontinuities. Existing methods to enforce these "invariant regions" can be complex, implicit, or reduce accuracy. This paper tackles the problem of developing an explicit, high-order accurate, and efficient limiter that preserves these invariant regions for multi-dimensional systems, specifically the compressible Euler equations.
Methods
The core contribution is an explicit invariant-region-preserving (IRP) limiter. This limiter works by taking a polynomial approximation within a computational cell and, if it deviates from the invariant region, it modifies it using a linear convex combination with the cell average. The key is that this combination is designed to bring the solution back into the invariant region while ensuring that the order of accuracy is maintained, provided the cell average is not too close to the boundary of the invariant region.
The paper also develops a generic algorithm for incorporating this IRP limiter into high-order finite volume and DG schemes. For DG schemes, they establish sufficient conditions under which arbitrarily high-order schemes maintain the IRP property. These conditions involve ensuring that the projected one-dimensional system shares the same invariant region as the full multi-dimensional system.
For the compressible Euler equations, they specifically derive the invariant region and apply the general results to develop high-order IRP DG schemes for both one and two dimensions. They also analyze the CFL (Courant-Friedrichs-Lewy) conditions required for these schemes to be IRP.
Datasets
Not discussed in the paper. The paper focuses on numerical methods and their theoretical properties, validated through numerical experiments on standard test cases.
Key Insights and Ideas
- Explicit IRP Limiter: The development of an explicit limiter is a significant advantage, simplifying implementation and avoiding iterative solvers often required by implicit methods.
- Preserving Accuracy: A crucial insight is that the IRP limiter can be designed to maintain high-order accuracy for smooth solutions, a common pitfall for many limiting techniques.
- Generalizability: The framework is developed for general hyperbolic conservation laws with a convex invariant region, making it applicable to a wide range of physical systems.
- Projected 1D System Condition: The requirement that the projected 1D system shares the same invariant region is a key theoretical insight for extending IRP properties to multi-dimensional DG schemes.
- Mild Limiter: The IRP limiter is described as "mild," meaning it intervenes only when necessary to preserve the invariant region, thus minimizing its impact on smooth solutions.
Key Results
- Theoretical Guarantees: The paper provides theoretical proofs that the proposed IRP limiter maintains high-order accuracy for smooth solutions and that the DG schemes with the IRP limiter preserve the invariant region under certain CFL conditions.
- Numerical Validation: Extensive numerical experiments are presented for the compressible Euler equations in 1D and 2D.
- Accuracy: Tables show that the IRP DG schemes achieve the expected high-order accuracy for smooth solutions, comparable to schemes without the limiter.
- Robustness: Figures demonstrate that the IRP limiter effectively prevents unphysical oscillations and overshoots/undershoots near discontinuities and vacuum regions, which are present in schemes using only positivity-preserving limiters. For instance, in the Sod shock tube problem, the IRP limiter reduced velocity oscillations more effectively than a standard positivity-preserving limiter.
- Efficiency: The explicit nature of the limiter contributes to computational efficiency.
Limitations and Future Work
- Boundary Proximity: The paper notes that accuracy might deteriorate when cell averages are very close to the boundary of the invariant region, as indicated by a large factor in the accuracy estimate.
- Oscillation Damping: While the IRP limiter is effective, the paper acknowledges that for "stronger oscillations," more stringent limiters might be needed, especially in 2D cases.
- Future Work: The paper suggests that the developed framework can be extended to other systems and that further research into stricter limiters for more challenging scenarios is warranted. The work also focuses on forward Euler time discretization, and the authors mention that high-order SSP time discretizations would preserve the IRP property due to their convex combination nature.
Summary updated