Local linear stability of dual-pairing summation-by-parts methods for nonlinear conservation laws

This paper demonstrates that recently developed high-order dual-pairing summation-by-parts methods, which utilize an entropy-stable volume upwind filter, achieve both provable entropy stability and crucial local energy stability for nonlinear conservation laws, thereby preventing high-frequency mode dominance and enabling accurate simulations of turbulent flows.

The Gist

Imagine you are trying to simulate the weather, the flow of blood in an artery, or the movement of ocean currents on a computer. These are all examples of nonlinear conservation laws—complex mathematical rules that describe how things like energy, mass, and momentum move and change.

The challenge is that these systems are chaotic. They can form shockwaves (like a sonic boom) or turbulence (like swirling eddies in a river). If your computer simulation isn't careful, it can go haywire, producing nonsense results or crashing entirely.

This paper introduces a new, smarter way to build these computer simulations so they are both robust (they don't crash) and accurate (they don't get polluted by digital noise).

Here is the breakdown using simple analogies:

1. The Problem: The "Two-Headed" Monster

To simulate these complex flows, scientists use high-order methods (very detailed, high-resolution math). However, these methods have a tricky problem: they often struggle to satisfy two conflicting requirements at the same time.

• Requirement A: Entropy Stability (The "Safety Net")
  Think of this as a safety net. It ensures the simulation never explodes into infinity. It guarantees that the total energy doesn't magically appear out of nowhere. If a simulation is "entropy stable," it won't crash.

  • The Flaw: A simulation can be safe (not crashing) but still be wrong. It might be "stable" while filling up with invisible, high-frequency digital noise (like static on an old TV) that ruins the picture.

• Requirement B: Local Linear Stability (The "Noise Filter")
  This is about keeping the small details clean. Imagine you are listening to a symphony. If the orchestra is playing perfectly, but there is a tiny, high-pitched squeak in the back that gets louder and louder, eventually it drowns out the music.

  • The Flaw: Many current "safe" methods allow these tiny, unresolved digital squeaks (high-frequency waves) to grow. They don't crash, but the result is a muddy, inaccurate mess.

The Big Question: Can we build a method that is both safe (won't crash) and clean (won't let digital noise take over)?

2. The Solution: The "Dual-Pairing" Filter

The authors propose a new method called Dual-Pairing Summation-by-Parts (DP SBP).

Think of a standard simulation method as a single pair of hands trying to juggle balls. Sometimes it drops them (instability).
The DP SBP method is like having two pairs of hands working in perfect sync.

• One pair handles the "forward" motion.
• The other pair handles the "backward" motion.
• Together, they balance each other out perfectly, ensuring no energy is lost or gained incorrectly.

3. The Secret Ingredient: The "Volume Upwind Filter"

The paper's biggest discovery is about a specific ingredient they add to this two-handed juggling act: Volume Upwinding.

• The Analogy: Imagine you are walking down a hallway with a strong wind blowing.
  • If you just walk normally (standard methods), the wind might push you off balance, or you might trip over invisible obstacles (numerical noise).
  • Upwinding is like leaning into the wind. It's a deliberate, calculated push against the flow to keep things steady.
  • Usually, you only do this at the edges of your simulation (the boundaries).
  • The Innovation: This paper proves that if you apply this "leaning into the wind" technique throughout the entire volume of the simulation (not just the edges), it acts as a filter.

What does this filter do?
It acts like a noise-canceling headphone for your math. It specifically targets those annoying, high-frequency digital squeaks (the unresolved wave modes) and damps them down before they can grow large enough to ruin the simulation.

4. The Proof: From Burgers' Equation to Turbulent Oceans

The authors tested this idea on two levels:

1. The Simple Test (Burgers' Equation):
   They started with a simple wave equation. They showed that without their "volume filter," the simulation developed wild, growing errors (like a feedback loop). With the filter, the errors died out, and the simulation remained perfectly stable.

2. The Hard Test (2D Turbulence):
   They simulated a massive, swirling ocean current (barotropic shear instability) that creates full-blown turbulence.

   • Without the filter: The simulation quickly filled with digital static. The beautiful swirling patterns were destroyed by noise, and the energy spectrum (the "sound" of the turbulence) looked like garbage.
   • With the filter: The simulation captured the swirling vortices beautifully. It resolved the tiny details of the turbulence without getting overwhelmed by noise. It produced a "power law" (a specific mathematical pattern found in real-world turbulence) that matched real physics.

The Takeaway

This paper solves a long-standing puzzle in computational physics. It proves that you can have your cake and eat it too:

1. Robustness: The simulation won't crash (Entropy Stability).
2. Accuracy: The simulation won't be ruined by digital noise (Local Linear Stability).

By using this new Dual-Pairing method with a built-in "volume upwind filter," scientists can now run high-resolution simulations of complex phenomena like weather, explosions, and turbulence with a level of reliability and clarity that was previously impossible. It's like upgrading from a shaky, static-filled TV to a crystal-clear 4K screen.

Technical Summary

1. Problem Statement

The development of high-order accurate numerical methods for nonlinear conservation laws faces a fundamental challenge: achieving simultaneous nonlinear (entropy) stability and local linear (energy) stability.

• Nonlinear/Entropy Stability: Ensures the boundedness of numerical solutions and prevents blow-up, which is crucial for handling shocks and turbulence.
• Local Linear Stability: Ensures that the asymptotic numerical growth rate of perturbations does not exceed the continuous growth rate of the underlying physical system.
• The Conflict: Many existing high-order entropy-stable schemes (e.g., skew-symmetric split-form SBP or DGSEM methods) are robust against nonlinear blow-up but suffer from local linear instability. These methods often allow unresolved high-frequency wave modes to grow exponentially, contaminating the solution with spurious oscillations and destroying accuracy, even if the solution remains bounded. The paper addresses the open question of whether a single method can satisfy both stability criteria.

2. Methodology

The authors analyze the Dual-Pairing (DP) Summation-by-Parts (SBP) framework recently introduced by Duru et al. [5]. The methodology involves:

• Theoretical Framework:

  • Dual-Pairing Operators: Utilizing a pair of forward ($D_+$) and backward ($D_-$) finite difference or discontinuous Galerkin operators that satisfy the SBP property. This allows for flexible treatment of boundary conditions and interface coupling.
  • Skew-Symmetric Split-Form: Reformulating the nonlinear conservation laws into a skew-symmetric form (convex combination of conservative and advective forms) to facilitate entropy conservation without invoking the chain rule.
  • Volume Upwinding: Introducing an intrinsic volume upwind filter (parameterized by $\gamma$) into the semi-discrete scheme. This filter acts as a dissipation mechanism specifically targeting unresolved high-frequency modes.

• Stability Analysis:

  • Linearization: The authors linearize the nonlinear semi-discrete equations around a smooth base-flow ($U$) to study the evolution of small perturbations ($\delta u$).
  • Eigenvalue Analysis: They examine the spectrum of the linearized discrete operator $Q$. A method is deemed locally energy stable (strictly stable) if the maximum real part of the eigenvalues ($\lambda_{max}$) satisfies $\lambda_{max} \leq \eta_c + O(\Delta x)$, where $\eta_c$ is the continuous growth rate.
  • Optimal Parameter Derivation: For the inviscid Burgers' equation, the authors derive optimal upwind parameters ($\gamma_{opt}$) as a function of the order of accuracy and the gradient of the base-flow. They show that $\gamma_{opt}$ scales with $(1-\alpha)\max|\Delta x \partial_x U|$, where $\alpha$ is the skew-symmetric splitting parameter.

• Numerical Experiments:

  • Models: Inviscid Burgers' equation (1D) and Nonlinear Shallow Water Equations (SWEs) in 1D and 2D.
  • Tests:
    1. Eigenvalue Spectra: Comparing schemes with and without volume upwinding ($\gamma=0$ vs. $\gamma>0$).
    2. Perturbation Growth: Evolving initial conditions perturbed by the fastest-growing eigenmodes to observe error amplification.
    3. Turbulence Simulation: Simulating 2D barotropic shear instability (Kelvin-Helmholtz instability) to assess the resolution of turbulent scales and energy cascades.

3. Key Contributions

1. Proof of Local Linear Stability: The paper demonstrates that the entropy-stable volume upwind filter inherent in the DP SBP method is sufficient to ensure local linear stability. This resolves the trade-off where skew-symmetric forms usually destabilize linearized systems.
2. Optimal Upwind Parameters: The authors provide explicit formulas for the optimal volume upwind parameter ($\gamma_{opt}$) for various orders of accuracy (both odd and even) for the Burgers' equation. They show that $\gamma_{opt} \to 0$ as the grid is refined ($\Delta x \to 0$) for smooth flows, ensuring consistency.
3. Unified Stability: The work establishes a framework for designing high-order methods that are provably entropy-stable (nonlinearly stable) and locally energy-stable simultaneously.
4. Turbulence Resolution: The study validates the method's ability to resolve fully developed turbulence without the pollution of grid-scale noise, a common failure mode in standard entropy-stable split-form schemes.

4. Results

• Eigenvalue Spectra:
  • Without Upwinding ($\gamma=0$): The eigenvalue spectra of entropy-stable skew-symmetric schemes are split between the left and right half-planes. Positive real parts indicate exponential growth of grid-scale errors, leading to instability.
  • With Upwinding ($\gamma=\gamma_{opt}$): The spectra shift entirely to the left half-plane (or have non-positive real parts), satisfying the strict stability condition.
• Perturbation Evolution:
  • In the absence of upwinding, perturbations grow exponentially until nonlinear saturation occurs, leaving the solution contaminated with high-frequency noise.
  • With optimal upwinding, perturbations decay or remain bounded at the level of machine precision, preserving the accuracy of the base flow.
• 2D Turbulence (Kelvin-Helmholtz Instability):
  • DGSEM ($\gamma=0$): The simulation becomes overwhelmed by numerical noise within 20 days, destroying the vortex structures. The energy and enstrophy spectra show no physical power-law behavior.
  • DP DG ($\gamma>0$): The method accurately captures the vortex structures and the development of turbulence up to 80 days.
  • Spectral Analysis: The DP DG method exhibits a Kraichnan inertial range with a $k^{-3.5}$ power law for kinetic energy and a $k^{-1.5}$ law for enstrophy, consistent with 2D hydrodynamic turbulence theory. In contrast, the standard DGSEM fails to show these trends due to noise.

5. Significance

This work is significant because it bridges a critical gap in high-order numerical methods for nonlinear PDEs.

• Reliability: It provides a theoretical and practical pathway to designing methods that do not just "survive" (nonlinear stability) but also "perform accurately" (local linear stability) in complex regimes like turbulence.
• Efficiency: By using an intrinsic volume upwind filter rather than external artificial viscosity or filtering, the method maintains high-order accuracy while suppressing spurious modes.
• Application: The successful simulation of 2D barotropic shear instability with fully developed turbulence demonstrates the method's potential for geophysical fluid dynamics and other fields requiring high-fidelity turbulence modeling.
• Theoretical Insight: It clarifies that the "destabilizing" terms in skew-symmetric forms (residuals of the chain/product rules) can be effectively neutralized by the upwind dissipation inherent in the DP SBP framework, allowing for robust, high-order simulations of nonlinear conservation laws.