Numerical method for strongly variable-density flows at low Mach number: flame-sheet regularisation and a mass-flux immersed boundary method

This paper presents a robust numerical method for simulating strongly variable-density, low-Mach-number flows in combustion systems by integrating a fractional time-step model with flame-sheet regularisation and an extended mass-flux immersed boundary method to handle thermal gradients and complex burner geometries on Cartesian grids.

The Gist

Imagine you are trying to film a slow-motion video of a candle flame. The air around the flame is moving very slowly, but sound waves (which are just pressure waves) are zipping through that air at the speed of a bullet.

If you try to use a standard camera (a standard computer simulation) designed for fast-moving objects, you run into a problem: to catch the slow motion of the flame, you have to take pictures so fast that you capture millions of sound waves in between. This makes the computer calculation take forever, like trying to count every single grain of sand on a beach just to measure the size of a single pebble.

This paper presents a clever new "camera" and a new way of "developing the film" specifically for these slow-moving, hot, variable-density flows (like fires and engines).

Here is a breakdown of their solution using simple analogies:

1. The Problem: The "Speed Trap"

In normal air, sound travels much faster than the wind. In low-speed flows (like a candle), the wind is a snail, but sound is a jet.

• The Old Way: Standard simulations try to track both the snail and the jet simultaneously. This is computationally expensive and often inaccurate because the "jet" (sound) overwhelms the "snail" (flow).
• The New Way: The authors say, "Let's ignore the jet." They mathematically strip away the sound waves, assuming they happen instantly. This leaves them with a simplified system that only tracks the slow, heavy movement of the air and the heat.

2. The "Fractional Step" Method: The Chef's Recipe

To solve the equations for this simplified flow, they use a technique called the Fractional Step Method.

• The Analogy: Imagine you are a chef trying to make a complex soup. Instead of trying to chop vegetables, boil water, and season the broth all at once (which might result in a mess), you break it down:
  1. Step 1 (Predictor): Chop the veggies and guess how the soup will taste.
  2. Step 2 (Correction): Taste the soup, realize it's too salty, and adjust the water level to fix the balance.
  3. Step 3 (Pressure): The "pressure" in the fluid acts like the water level in the pot. If the pot is too full (too much mass), you drain some; if it's too empty, you add some. This step ensures the soup (the fluid) doesn't magically appear or disappear.

3. The "Flame Sheet" and the "Blur"

In their model, they assume the chemical reaction (the fire) happens instantly. This creates a "Flame Sheet"—a razor-thin line where fuel turns into fire.

• The Problem: In math, a razor-thin line is a "discontinuity." It's like a cliff edge. Computers hate cliff edges because they cause "shaky" results (numerical oscillations).
• The Solution (Regularisation): Instead of a razor-thin line, they introduce a "soft focus" or a "blur." They smooth out the transition between fuel and fire over a tiny, invisible distance.
  • The Metaphor: Imagine a black-and-white photo with a sharp, jagged line between black and white. It looks pixelated and ugly. They apply a "feather" tool to blend the black and white slightly. The photo looks smoother, and the computer can handle it without crashing, while still looking exactly like a flame to the human eye.

4. The "Ghost" Burner (Immersed Boundary Method)

Real burners (like the nozzle on a rocket or a candle wick) are round and complex. But the computer grid they use is made of perfect squares (like graph paper).

• The Problem: How do you fit a round circle into a square grid without cutting the squares into tiny, messy triangles?
• The Solution (IBM): They use the Immersed Boundary Method.
  • The Analogy: Imagine pouring water into a bucket full of square blocks. You don't need to cut the blocks to fit the bucket. Instead, you tell the water, "If you hit a block, stop."
  • The Twist: Usually, this method just stops the water. But this paper adds a special feature: Mass Flux. They tell the "ghost" blocks, "Not only stop the water, but also push new fuel out of the block." This allows them to simulate fuel being ejected from a round burner using a simple square grid, saving massive amounts of computer time.

5. The "Odd-Even" Dance

When you put numbers on a grid, sometimes the computer gets confused. It might think the pressure is high on the left, low in the middle, and high on the right, creating a "checkerboard" pattern that doesn't exist in reality.

• The Fix: The authors use a special "flux interpolation" technique.
  • The Metaphor: Imagine a dance floor where partners are holding hands. If they only talk to their immediate neighbors, they might get out of sync. This method forces the dancers to check in with the person two steps away, ensuring the whole floor moves in a smooth, coordinated wave rather than a chaotic jitter.

Summary of Results

The authors tested their new "camera" and "recipe" on three scenarios:

1. Spinning Vortices: To check if the math is accurate (it is).
2. Flow between Cylinders: To test their "Ghost Burner" technique (it works perfectly).
3. The Double Tsuji Flame: A complex fire scenario where fuel shoots out of a cylinder into a stream of air. They compared their results with a famous commercial software (OpenFOAM) and found their method matched it perfectly but was built with a simpler, more robust foundation.

In a nutshell: This paper gives scientists a faster, more stable, and easier-to-use tool for simulating fires and low-speed hot air flows, handling complex shapes and sharp temperature changes without breaking a sweat.

Technical Summary

1. Problem Statement

Low-Mach-number flows (where flow velocity is significantly lower than the speed of sound) present significant mathematical and computational challenges due to the disparity between acoustic and convective time scales.

• The Challenge: Standard compressible flow solvers become inefficient or inaccurate because they must resolve fast acoustic waves to satisfy the Courant–Friedrichs–Lewy (CFL) condition, even though these waves have negligible impact on the fluid dynamics in low-Mach regimes.
• Specific Context: The paper focuses on reacting flows (combustion) with strong variable density (due to large temperature gradients) and complex geometries.
• Key Difficulties:
  1. Flame Sheet Approximation: Assuming infinitely fast chemistry (Burke–Schumann limit) creates discontinuities in properties (density, temperature, derivatives) at the flame front, leading to numerical oscillations in finite-difference schemes.
  2. Complex Boundaries: Simulating fuel ejection from arbitrary geometries (e.g., cylindrical burners) on Cartesian grids requires robust handling of boundary conditions and mass flux.
  3. Pressure-Velocity Coupling: Ensuring mass conservation without spurious pressure oscillations ("odd-even decoupling") on collocated grids.

2. Methodology

The authors propose a simplified fluid dynamics model based on the fractional time-step (projection) method, integrated with specific techniques for combustion and immersed boundaries.

A. Mathematical Model

• Low-Mach Asymptotics: The governing equations are derived by taking the limit as the Mach number ($M \to 0$). This eliminates acoustic waves while retaining convective and diffusive dynamics.
• Governing Equations: The system solves for density ($\rho$), velocity ($\mathbf{v}$), mixture fraction ($Z$), and excess enthalpy ($H$).
  • Thermodynamics: The ideal gas law relates pressure and temperature ($p = \rho T$). Thermodynamic pressure ($p_0$) is spatially uniform but evolves in time based on global mass/energy constraints.
  • Combustion: Modeled using the Burke–Schumann limit (infinitely fast chemistry). Reactants are separated by an infinitely thin flame sheet.
  • Coupling Functions: Instead of solving individual species equations, the model uses Mixture Fraction ($Z$) and Excess Enthalpy ($H$) as transport variables. These are conserved scalars that allow the reconstruction of temperature and species concentrations.
  • Transport Coefficients: Temperature-dependent viscosity and thermal conductivity are modeled using a power law ($\mu, \kappa \propto T^\sigma$).

B. Numerical Scheme

• Temporal Discretization: A predictor-corrector scheme is used:
  • Predictor: Second-order Adams–Bashforth (explicit).
  • Corrector: Second-order Adams–Moulton (implicit).
  • This combination ensures second-order global temporal accuracy.
• Spatial Discretization:
  • Grid: Collocated grid (all variables stored at cell centers).
  • Stability: To prevent "odd-even decoupling" (checkerboard pressure oscillations), auxiliary fluxes are defined at cell interfaces. Interpolation operators transfer data between cell centers and interfaces, creating a strong coupling between the pressure (mass conservation) and velocity (momentum conservation) equations.
  • Pressure Solution: A Poisson equation for pressure is solved at each step to enforce the continuity equation (mass conservation). The system uses an augmented matrix method to handle the pure Neumann boundary conditions inherent to low-Mach flows.

C. Key Technical Innovations

1. Flame-Sheet Regularization:
   • The discontinuity at the flame front (where $Z=1$) causes numerical instability. The authors replace the Heaviside step function with a regularized hyperbolic tangent function ($\tanh$).
   • Novelty: Instead of regularizing temperature directly, they regularize the derivative of the function relating temperature to mixture fraction. This ensures smooth transitions in both the property and its gradient, preventing oscillations in the finite-difference scheme.
2. Mass-Flux Immersed Boundary Method (IBM):
   • Extends the standard penalization method (treating solids as porous media with low permeability) to handle mass injection.
   • A forcing term is added to the momentum equation to enforce velocity boundary conditions.
   • Crucially, a mass source term ($S_m$) is added to the continuity equation to account for fuel ejection from the burner surface, allowing simulations of fuel injection from arbitrary geometries (like a cylinder) on a Cartesian grid.

3. Key Contributions

• Robust Low-Mach Solver: Development of a stable, second-order accurate solver for strongly variable-density flows that handles large temperature gradients without resolving acoustic waves.
• Advanced Regularization: A specific regularization technique for flame-sheet derivatives that maintains numerical stability while preserving the physics of diffusion flames.
• Extended IBM: A novel formulation of the Immersed Boundary Method that explicitly handles mass flux (fuel ejection), enabling the simulation of complex burner geometries (e.g., cylindrical burners) without body-fitted meshes.
• Collocated Grid Implementation: A rigorous implementation of flux interpolation on collocated grids to ensure mass conservation and eliminate pressure oscillations in variable-density reacting flows.

4. Results and Verification

The method was validated through four distinct test cases:

1. Taylor–Green Vortex (Incompressible):

   • Verified spatial and temporal accuracy.
   • Result: Achieved theoretical second-order spatial convergence and second-order global temporal accuracy (third-order local truncation error). Confirmed the pressure behaves as a time-averaged constraint.

2. Taylor–Couette Flow (IBM Validation):

   • Tested the penalization method against an exact analytical solution for flow between rotating cylinders.
   • Result: Demonstrated that using a shifted characteristic function (extending the solid boundary slightly into the fluid) significantly improves accuracy compared to the standard mask function, especially for larger Darcy numbers.

3. Thermally Driven Cavity (Variable Density):

   • Simulated natural convection in a differentially heated square cavity at high Rayleigh numbers ($Ra = 10^2$ and $10^7$).
   • Result: Excellent agreement with benchmark data for streamlines, isotherms, and Nusselt numbers. Validated the solver's ability to handle strong density gradients and buoyancy effects.

4. Double Tsuji Flame (Full Combustion + IBM):

   • Simulated a cylindrical burner ejecting fuel into a counterflow of oxidizer. This combined all features: variable density, combustion, flame regularization, and mass-flux IBM.
   • Result: The flame structure, temperature fields, and velocity profiles matched well with OpenFOAM simulations and analytical expectations. The method successfully captured the complex interaction between the cylindrical burner geometry and the diffusion flame.

5. Significance

This work provides a comprehensive and efficient numerical framework for simulating low-Mach combustion problems where:

• Geometry is complex: The IBM allows for arbitrary burner shapes without expensive mesh generation.
• Density varies strongly: The low-Mach formulation avoids the stiffness of compressible solvers while capturing buoyancy and thermal expansion.
• Chemistry is fast: The flame-sheet approximation with regularization allows for the study of diffusion flames without resolving the thin reaction zone, significantly reducing computational cost.

The proposed method is stable, accurate, and computationally efficient, making it a valuable tool for analyzing complex reacting flows in engineering applications such as gas turbines, furnaces, and safety analysis, where both geometric complexity and strong thermal gradients are present.