Investigation of Differential Diffusion and Strain Coupling in Large Eddy Simulations of Hydrogen-Air Flames

This study validates a flamelet-based Large Eddy Simulation approach for hydrogen-air flames, demonstrating that it effectively captures the coupling between differential diffusion, strain, and curvature to accurately predict flame structure and reaction rates without requiring complex strained flamelet databases.

The Gist

The Great Hydrogen Race: How a Computer Model Learned to Predict the Unpredictable

Imagine you are trying to predict the path of a very fast, very energetic runner (Hydrogen) trying to sprint through a crowded, windy stadium (Air). This runner is special: they are incredibly light, move faster than anyone else, and react instantly to the wind.

This paper is about a team of scientists who built a super-advanced video game simulation to watch this runner. Their goal? To figure out how to predict exactly where the runner will go and how fast they will burn, even when the wind is blowing hard and twisting the track.

Here is the breakdown of their adventure, explained simply:

1. The Problem: The "Lightweight" Runner

Hydrogen is a fantastic fuel because it's clean (no carbon). But it's a nightmare to control.

• The Issue: Unlike gasoline, hydrogen is so light that it diffuses (spreads out) much faster than the heat it creates. In physics terms, this is called Differential Diffusion.
• The Analogy: Imagine a race where the runner (hydrogen) is so light that a gentle breeze blows them ahead of their own shadow (heat). In a normal fire, the fuel and heat stay together. In a hydrogen fire, the fuel runs ahead, creating pockets of "super-fuel" that can make the fire explode or change shape unexpectedly.
• The Challenge: Scientists have used computer models for years that assume fuel and heat move at the same speed (like a heavy runner). These models work great for gas or wood, but they fail miserably with hydrogen because they can't account for the fuel "running away" from the heat.

2. The Experiment: The Windy Obstacle Course

To test their ideas, the scientists looked at a real experiment done at a university in Norway.

• The Setup: They shot a stream of hydrogen and air at a blunt, cone-shaped object (a "bluff body"). This created a swirling vortex, like water going down a drain, which held the flame in place.
• The Goal: They wanted to see if their computer model could predict the flame's shape and speed without needing to know every single tiny detail of the wind (which would take a supercomputer 100 years to calculate).

3. The Solution: The "Unstretched" Map

Usually, to predict how a flame behaves in strong winds, you need a massive library of pre-calculated "flame maps" that show how the flame looks when it is being stretched, squished, and twisted by the wind. This is like having a map for every possible wind speed and direction. It's huge, expensive, and hard to use.

The scientists asked a bold question: "What if we just use a simple, flat map (an 'unstretched' flame) and let the computer figure out the wind effects on the fly?"

They built a new model that:

1. Uses a simple map: It starts with a basic, calm hydrogen flame.
2. Adds a "Smart Correction": It has a special rule that says, "Hey, the wind is blowing the light hydrogen fuel ahead of the heat! Let's adjust the fuel mix right now."
3. Let's the Wind Do the Work: Instead of pre-calculating every twist, the simulation resolves (calculates) the big swirls of wind directly.

4. The Results: The "Chasing" Flame

When they ran the simulation, something amazing happened.

• The "Chasing" Effect: The model showed that because hydrogen is so light, the flame actually "chases" the areas of high wind strain. It pulls itself closer to the blunt object, making the flame shorter and more intense than the old models predicted.
• The "Super-Fuel" Pockets: The model correctly identified that the wind was pushing extra hydrogen into certain spots, creating "super-fuel" pockets. This made the fire burn hotter and faster in those specific areas.
• The Victory: The new model's predictions matched the real-world experiments almost perfectly. It got the shape, the speed, and the temperature right.

5. Why This Matters (The Big Picture)

Think of the old way of modeling hydrogen as trying to navigate a city using a paper map from 1950. It's okay for a slow walk, but useless for a Formula 1 race.

This new approach is like giving the driver a GPS that updates in real-time.

• Simplicity: You don't need a massive library of pre-calculated wind scenarios. You just need the basic rules of how hydrogen behaves.
• Efficiency: It saves huge amounts of computing power.
• Future Design: This means engineers can now design better, safer hydrogen engines and power plants. They can predict exactly how a hydrogen flame will behave in a jet engine or a power turbine without needing a supercomputer the size of a building.

In a nutshell: The scientists proved that you can predict the chaotic, fast-moving behavior of a hydrogen fire using a simpler, smarter model that accounts for the fact that hydrogen is a "lightweight runner" that likes to get ahead of the heat. This is a huge step toward making hydrogen a safe and practical fuel for our future.

Technical Summary

1. Problem Statement

Hydrogen combustion offers a pathway for decarbonization but presents unique challenges due to its physical and chemical properties, specifically high molecular diffusivity, low activation energy, and high burning velocity. A critical phenomenon in hydrogen flames is differential diffusion (preferential diffusion), where light species (like $H_2$) diffuse faster than heavier species. This effect couples strongly with flame stretch (the sum of tangential strain and curvature), significantly altering flame structure, reaction rates, and stability, particularly in ultra-lean conditions.

Current Large Eddy Simulation (LES) models for premixed flames often rely on flamelet-based approaches. However, traditional flamelet models assume species equidiffusivity (unity Lewis number) and are typically built on unstretched, one-dimensional (1D) flamelet databases. It remains unclear whether these simplified models, which do not explicitly include strain in the database, can accurately capture the complex interplay between differential diffusion, strain, and curvature in turbulent hydrogen flames where strain levels are high.

2. Methodology

The authors conducted Large Eddy Simulations (LES) of a premixed hydrogen-air flame stabilized by a bluff-body (an unconfined configuration developed at NTNU).

• Numerical Framework:

  • Solver: OpenFOAM-v9 using the finite volume method with low-Mach approximation.
  • Turbulence Modeling: A one-equation subgrid scale (SGS) model for turbulent kinetic energy ($k_{sgs}$).
  • Combustion Model: A flamelet-based approach coupled with a Presumed Filtered Density Function (FDF) closure.
  • Thermochemistry: The database consists of unstretched, 1D premixed hydrogen-air flamelets parametrized by the progress variable ($c$, based on $Y_{H_2O}$) and Bilger's mixture fraction ($z$).
  • Differential Diffusion Implementation: The model incorporates differential diffusion effects via a correction method (extended from Mukundakumar et al. and Ferrante et al.) that modifies the transport equations for mixture fraction, progress variable, and enthalpy. This involves adding source terms and correcting diffusion fluxes based on local Lewis number variations.

• Simulation Cases:
  Two cases were compared:

  1. $Le = 1$: Unity Lewis number assumption (no differential diffusion).
  2. $Le \neq 1$: Differential diffusion effects included.

• Validation:
  Results were validated against experimental data (Particle Image Velocimetry for velocity fields and OH*-chemiluminescence for flame structure) provided by the TNF Workshop archive.

3. Key Contributions

• Validation of Unstretched Flamelets for Strained Flames: The study demonstrates that an LES model using a database of unstretched flamelets can successfully predict the macroscopic effects of strain and differential diffusion in hydrogen flames, provided that the majority of the strain and curvature are resolved by the LES grid.
• Mechanism of Strain-Diffusion Coupling: The paper elucidates how resolved tangential strain interacts with differential diffusion to redistribute the mixture fraction, leading to local enrichment and increased reaction rates, even without a strained flamelet database.
• Quantification of Strain vs. Curvature: The work highlights that in bluff-body stabilized flames, tangential strain is the dominant driver of mixture fraction redistribution, overshadowing curvature effects (unlike in slot-burner configurations where thermodiffusive instabilities dominate).

4. Key Results

• Flow Field Validation: Both LES cases ($Le=1$ and $Le \neq 1$) accurately predicted the mean velocity fields and RMS fluctuations, showing good agreement with experimental PIV data.
• Flame Structure and Length:
  • The $Le \neq 1$ case predicted a shorter flame length and a reaction rate peak shifted closer to the bluff-body edge.
  • This result showed significantly better agreement with experimental OH*-chemiluminescence data compared to the $Le=1$ case.
• Mixture Fraction Redistribution:
  • The model captured the expected decrease in mixture fraction in the preheat zone (due to 1D differential diffusion) and a significant increase (overshoot) in the post-flame region.
  • This overshoot was driven by positive tangential strain coupled with differential diffusion, leading to local mixture enrichment.
• Reaction Rate and Temperature:
  • The local enrichment caused by strain led to superadiabatic temperatures (peaking around 2000 K) and increased reaction rates.
  • The model successfully mimicked the effect of a negative Markstein length (where positive strain increases reaction rate) using only unstretched flamelets.
• Strain vs. Curvature Dominance:
  • Analysis of the Probability Density Functions (PDFs) revealed that the flame stabilization is dominated by tangential strain rather than curvature.
  • The "unstretched" conditional average of mixture fraction aligned with the global average, confirming that strain (resolved by the grid) drives the deviation from 1D flamelet behavior, not curvature.

5. Significance and Implications

• Modeling Simplification: The findings suggest that for turbulent hydrogen-premixed flames with high resolved strain, it is not necessary to generate computationally expensive databases of strained flamelets. An unstretched flamelet database, when coupled with a differential diffusion correction and sufficient grid resolution to resolve the strain, is sufficient to capture the correct thermochemical state.
• Design of Combustors: This offers a practical and computationally efficient pathway for designing novel hydrogen combustors (e.g., for gas turbines) where accurate prediction of flame stability and NOx formation (linked to temperature peaks) is critical.
• Physical Insight: The study confirms that the "chasing" behavior of hydrogen flames toward high-strain regions is a result of the coupling between resolved strain and preferential diffusion, a mechanism that can be captured by modern LES frameworks without resorting to complex, pre-tabulated strained flamelets.

In conclusion, the paper establishes that resolved strain in LES acts as a proxy for the missing strain effects in unstretched flamelet databases, provided that differential diffusion is correctly modeled. This validates a simplified yet robust approach for simulating complex hydrogen combustion dynamics.