A Flux-Correction Form of the Third-Order Edge-Based Scheme for a General Numerical Flux Function

This paper introduces a flux-correction form of the third-order edge-based scheme that enables the direct application of general numerical flux functions while preserving third-order accuracy through the use of the U-MUSCL scheme with $\kappa = 1/2$, as verified by numerical results on irregular tetrahedral grids.

The Gist

Imagine you are trying to predict how a river flows around a complex city of buildings. To do this, computer scientists break the river into millions of tiny, invisible Lego blocks (a grid) and calculate how the water moves from one block to the next.

The goal of this paper is to make those calculations sharper, faster, and more flexible, specifically for a method called the "Third-Order Edge-Based Scheme."

Here is the breakdown of the problem and the solution, using simple analogies:

The Problem: The "Rigid Recipe"

For years, scientists had a very precise recipe (a mathematical formula) to calculate how water flows between two Lego blocks. This recipe was great because it was third-order accurate, meaning it was incredibly precise, even on messy, irregular grids where the blocks weren't perfect cubes.

However, this recipe had a strict rule: it only worked if you used a specific, pre-approved type of "flavoring" (a specific mathematical flux function) to mix the water.

If a scientist wanted to use a new, fancy, or highly specialized flavoring (like a new formula for hypersonic jets or chemical reactions), they had to spend months trying to rewrite that new flavoring to fit the old recipe. It was like trying to force a square peg into a round hole. If the new flavoring had complex ingredients or tuning knobs, the whole process became a nightmare.

The Solution: The "Flux-Correction" Adapter

Hiroaki Nishikawa, the author of this paper, invented a clever adapter.

Instead of forcing the new flavoring to change its shape to fit the old recipe, he added a small "correction term" (a little extra ingredient) to the mix.

Think of it like this:

• The Old Way: You have a specific sandwich maker. You can only make sandwiches if you cut the bread into a perfect square first. If you have round bread, you have to spend hours squaring it off.
• The New Way: You keep your round bread (the general flux function). You put it in the sandwich maker, but you add a special "squaring sauce" (the flux correction) that magically makes the round bread behave exactly like a square one inside the machine.

How It Works (The Magic Trick)

1. The General Flux: The scientist can now plug in any flavoring they want (like HLLC or LDFSS, which are popular formulas for high-speed airflows). They don't need to rewrite the code to fit the old recipe.
2. The Correction Term: The computer adds a tiny mathematical "nudge" to the calculation. This nudge compensates for the fact that the new flavoring wasn't designed for the old recipe.
3. The Result: The final calculation remains third-order accurate. The "nudge" fixes any errors that would have ruined the precision, ensuring the river flow prediction is still razor-sharp.

Why This Matters

• Flexibility: Scientists can now use the best, most modern tools for specific problems (like chemical reactions in hypersonic flight) without having to rebuild their entire simulation engine.
• Simplicity: It makes the computer code easier to write and maintain. You don't need to be a math wizard to adapt new formulas; you just plug them in and add the correction.
• Proof: The author tested this with some of the most complex math formulas used in aerodynamics. The results showed that the "adapter" worked perfectly, keeping the high accuracy even on messy, irregular grids.

The Bottom Line

This paper is like inventing a universal power adapter. Before, you needed a specific plug for every specific country (or specific math formula). Now, you can plug in any device (any flux function) into the same outlet, and a little built-in converter ensures it works perfectly without losing any power (accuracy). This allows engineers to solve complex fluid dynamics problems much faster and with more freedom.

Technical Summary

1. Problem Statement

The third-order edge-based scheme is a highly efficient discretization method for solving the Euler equations on arbitrary tetrahedral grids. It achieves third-order accuracy without requiring curved high-order grids or the computation of second derivatives. However, a significant limitation exists in its current formulation:

• Restrictive Flux Form: The scheme traditionally relies on a specific upwind flux structure (the sum of an arithmetic average of linearly extrapolated fluxes and a dissipation term).
• Implementation Difficulty: While it is theoretically possible to rewrite other numerical flux functions (like HLLC or LDFSS) into this specific form, doing so requires significant effort. This is particularly problematic for complex, mature flux functions that include specific modifications or tuning parameters (e.g., for chemically reacting hypersonic flows).
• Goal: The paper aims to develop a formulation that allows the direct use of any general numerical flux function within the third-order edge-based scheme while preserving third-order accuracy.

2. Methodology

The author proposes a Flux-Correction Form that modifies the standard discretization to accommodate general flux functions.

Core Concept

Instead of forcing a general flux function $\Phi$ to match the specific "average + dissipation" structure, the method replaces the arithmetic average of extrapolated fluxes with the general numerical flux evaluated at the edge midpoint, augmented by a specific correction term.

Mathematical Derivation

1. Standard Expansion: The paper analyzes the truncation error of the standard third-order scheme. It shows that third-order accuracy is achieved because the leading second-order error term is "factored" (it vanishes on regular grids) and the source term quadrature is designed to cancel specific error terms.
2. The Correction Term: To maintain this accuracy when using a general flux $\Phi_{jk}(u_L, u_R, \hat{n}_{jk})$, the discretization is modified to:
   $$\sum [\Phi_{jk} + \delta f_{jk}] |n_{jk}| = \text{Source Term}$$
   Where $\delta f_{jk}$ is the flux correction defined as:
   $$\delta f_{jk} = \frac{1}{8} \left[ \left(\frac{\partial f}{\partial w}\right)_j \nabla w_j - \left(\frac{\partial f}{\partial w}\right)_k \nabla w_k \right] \cdot \Delta x_{jk}$$
3. Accuracy Preservation: Through Taylor series expansion, the author proves that setting the correction coefficient to $C = 1/4$ (resulting in the $1/8$ factor in the final formula) ensures that the combined term $(\Phi_{jk} + \delta f_{jk})$ matches the Taylor expansion of the exact flux up to the third order, provided the left and right states are computed correctly.

Implementation Details

• State Extrapolation: The left ($u_L$) and right ($u_R$) states at the edge midpoint must be computed using the U-MUSCL scheme with the parameter $\kappa = 1/2$.
  • This specific value of $\kappa$ ensures that the extrapolation is exact for quadratic functions on arbitrary grids (assuming gradients are computed exactly for quadratic functions).
  • This eliminates the need for second derivatives.
• Gradient Computation: Gradients are computed using an implicit edge-based method to ensure stability and accuracy.
• Source Term: An accuracy-preserving source quadrature formula (compact formula) is used to maintain the third-order order of accuracy.

3. Key Contributions

• General Flux Compatibility: The primary contribution is the derivation of a flux-correction term that allows any numerical flux function (e.g., HLLC, LDFSS, Roe) to be used directly in the third-order edge-based scheme without rewriting the flux function itself.
• Theoretical Proof: The paper provides a rigorous mathematical proof that the proposed form preserves third-order accuracy on arbitrary tetrahedral grids, provided the U-MUSCL scheme with $\kappa=1/2$ is used for state extrapolation.
• Simplified Implementation: The method simplifies the integration of third-order schemes into existing CFD solvers. Developers only need to implement the U-MUSCL extrapolation and add the correction term, rather than deriving complex algebraic forms for every new flux function.

4. Results

The proposed method was validated using the Method of Manufactured Solutions (MMS) on a high-aspect-ratio domain with irregular tetrahedral grids.

• Test Setup:
  • Equations: Steady Euler equations.
  • Fluxes Tested: HLLC, LDFSS, and Roe fluxes.
  • Comparison:
    • HLLC(3rd-FC) and LDFSS(3rd-FC): General fluxes using the new correction form.
    • Roe(3rd-FR): Original Roe flux using the traditional extrapolation form.
    • Roe(3rd-FC): Roe flux using the new correction form.
    • Roe(2nd): Standard second-order scheme.
• Findings:
  • Accuracy: All third-order variants (HLLC, LDFSS, and Roe) successfully achieved third-order convergence rates ($O(h^3)$) on the irregular grids.
  • Comparison: The third-order schemes were significantly more accurate than the second-order scheme.
  • Equivalence: The Roe flux performed identically whether implemented in the original form or the new flux-correction form, validating the theoretical derivation.

5. Significance

• Algorithmic Flexibility: This work removes a major barrier to adopting high-order methods in complex CFD applications. Researchers can now utilize sophisticated, tuned flux functions developed for specific physics (e.g., hypersonics, combustion) without the burden of re-deriving them for third-order edge-based schemes.
• Efficiency: By avoiding the need for curved high-order grids or second derivatives, the method maintains the computational efficiency of the edge-based approach while boosting accuracy.
• Future Impact: The author notes that this technique will facilitate the development of robust, high-order solvers for anisotropic viscous grid adaptation and complex flow simulations, as the same flux function can be seamlessly used for both second- and third-order solvers.