Well-balanced finite volume schemes for nearly steady adiabatic flows

Luc Grosheintz-Laval, Roger KäppeliJournal of Computational Physics 2020 · DOI

TL;DR

This paper introduces novel well-balanced finite volume schemes for simulating nearly steady adiabatic flows in astrophysics. These schemes precisely preserve steady states, significantly improving accuracy and reducing computational costs compared to standard methods, especially when dealing with small perturbations around equilibrium.

What Problem Does This Paper Address?

Simulating astrophysical phenomena like stellar accretion and wind often involves the Euler equations with gravity. A key challenge is that these systems possess "steady states" where the fluid is in equilibrium (e.g., hydrostatic or steady adiabatic flow). Standard numerical methods struggle to accurately represent these steady states, introducing errors that can obscure small but important deviations. This necessitates extremely fine grids to resolve these deviations, leading to prohibitive computational costs. The paper addresses this by developing "well-balanced" schemes that inherently preserve these steady states, allowing for accurate simulations of perturbations on top of equilibria.

Methods

The core innovation lies in a "local equilibrium reconstruction" coupled with a "well-balanced source term discretization." Instead of reconstructing the flow based on piecewise constant or linear approximations, the schemes reconstruct it based on a local equilibrium profile that satisfies the steady adiabatic flow conditions (Bernoulli's equation). This equilibrium profile is then combined with a perturbation term, which is reconstructed using standard methods. The source terms (due to gravity) are also carefully discretized to ensure they perfectly balance the flux terms for the equilibrium state. The authors extend these methods to Cartesian, cylindrical, and spherical coordinates, and also propose a dimension-by-dimension extension for multi-dimensional problems.

Datasets

Not discussed in the paper. The paper focuses on numerical methods and their validation through simulated experiments.

Key Insights and Ideas

  • Preserving Steady States is Crucial: The paper strongly demonstrates that standard numerical schemes fail to maintain steady states accurately, leading to significant errors that grow over time. Well-balanced schemes, by design, eliminate these errors for equilibrium states.
  • Local Equilibrium Reconstruction is Key: The novel approach of reconstructing the flow based on a local steady-state solution, rather than just local data, is the foundation for achieving well-balancing.
  • Impact on Perturbations: The most significant benefit of well-balanced schemes is their ability to accurately capture small perturbations on top of steady states. This is crucial for studying phenomena like waves or instabilities that are subtle deviations from equilibrium.
  • Flexibility in Numerical Fluxes: The proposed schemes are designed to work with any consistent numerical flux for the Euler equations, making them easily integrable into existing finite volume frameworks.
  • Coordinate System Adaptability: The extension to cylindrical and spherical coordinates is vital for astrophysical applications where these geometries are common.

Key Results

  • Exact Preservation of Steady States: The well-balanced schemes preserve steady adiabatic flows to machine precision, as evidenced by extremely small equilibrium errors (on the order of 10⁻¹³ to 10⁻¹⁵) in numerical experiments.
  • Superior Accuracy for Perturbations: For small amplitude perturbations, the well-balanced schemes achieve errors that are 2-3 orders of magnitude smaller than unbalanced schemes. This improvement is dramatic, with errors at low resolutions comparable to unbalanced schemes at high resolutions.
  • Second-Order Accuracy Maintained: The schemes achieve their formal second-order accuracy for smooth solutions, and importantly, the well-balancing does not negatively impact the robustness of shock-capturing capabilities.
  • Performance in Complex Scenarios: The schemes demonstrate good performance in more complex scenarios, including discontinuous flows and a realistic stellar accretion problem with a multi-physics equation of state.

Limitations and Future Work

  • Streamline Alignment: The multi-dimensional extension is only truly well-balanced if the streamlines of the steady flow are aligned with the computational grid axes.
  • Non-Adiabatic Flows: The current schemes are designed for adiabatic flows. Future work should focus on developing well-balanced schemes for non-adiabatic steady states, which are common in astrophysics due to radiation losses.
  • Coordinate Singularities: The use of curvilinear coordinates (cylindrical, spherical) can introduce coordinate singularities, which can affect grid regularity and time step sizes. Further research is needed to address this.
  • Higher-Order Accuracy: Extending the well-balanced schemes beyond second-order accuracy is a desirable direction for future research.

Summary updated

BibTeX
@article{grosheintzlaval2020well,
  author = {Luc Grosheintz-Laval and Roger Käppeli},
  title = {{Well-balanced finite volume schemes for nearly steady adiabatic flows}},
  journal = {Journal of Computational Physics},
  year = {2020},
  doi = {10.1016/j.jcp.2020.109805},
}
Figure