Effects of preferential concentration on the combustion of iron particles -- A numerical study with homogeneous isotropic turbulence

This numerical study demonstrates that preferential concentration in turbulent iron particle combustion significantly extends total combustion time and reduces peak temperatures compared to uniform distributions, with the effect intensifying at higher equivalence ratios and Reynolds numbers due to macroscopic oxygen depletion zones and particle clustering.

The Gist

The Big Picture: Iron as a "Green" Fuel

Imagine you want to power a city without burning fossil fuels. Scientists are looking at iron powder as a super-clean fuel. When iron burns, it turns into rust (iron oxide) and releases heat, but it doesn't release carbon dioxide. The catch? Iron doesn't melt or evaporate like wood or coal; it stays as solid particles while it burns.

The problem is that when you blow these tiny iron particles through a furnace with air, the air isn't smooth. It's turbulent, like a chaotic river with swirling eddies. This turbulence does something weird to the iron particles: it doesn't spread them out evenly. Instead, it acts like a cosmic vacuum cleaner that sucks them into tight groups, leaving huge empty spaces in between.

This phenomenon is called "preferential concentration." Think of it like a crowded dance floor where the music (turbulence) suddenly forces everyone to huddle in tight circles, leaving the rest of the floor completely empty.

The Experiment: A Digital Sandbox

The researchers used a supercomputer to run a "virtual experiment." They created a digital box filled with swirling air and iron particles. They wanted to see: Does huddling together make the iron burn faster or slower?

They tested three main variables:

1. How "sticky" the particles are to the air (Stokes number).
2. How wild the turbulence is (Reynolds number).
3. How much fuel vs. air they had (Equivalence ratio).

The Surprising Findings

1. The "Crowded Room" Effect (Clustering)

When the iron particles are spread out randomly (like people walking in a park), they burn efficiently. But when they get forced into clusters (like people packed into a tiny elevator), things change.

• The Analogy: Imagine a group of people trying to eat a single pizza. If they are spread out, everyone gets a slice easily. If they are all crammed into a tiny closet, the people in the middle can't reach the pizza. They starve while the people on the edge eat everything.
• The Result: In the iron clusters, the particles in the center run out of oxygen because the outer particles are "eating" all the air first. This causes the inner particles to burn much slower. In some cases, the total time to burn everything took eight times longer than if the particles were spread out randomly.

2. The "Empty Room" Effect (Voids)

While the clusters are starving, the empty spaces (voids) between them are actually burning faster than average. Because there are so few particles there, there is plenty of oxygen for everyone. It's like having a whole pizza to yourself in an empty room.

3. The "Group Hug" Problem

The researchers found that it's not just about how tight a single group is, but how close the groups are to each other.

• The Analogy: If you have two separate crowded rooms, the people in the middle of each room struggle to get food. But if you push those two rooms right next to each other, they merge into one giant, super-crowded room. Now, the "oxygen starvation" gets even worse because the groups are helping each other block the air supply.
• The Result: The study showed that you can't just look at one cluster to predict how long it will take to burn. You have to look at the whole map of where the clusters are. If clusters are neighbors, the burn time gets even longer.

Why This Matters

The researchers tried to create a simple math formula to predict how long iron would take to burn based on how clumped it was.

• The Good News: They found a pattern! In the "crowded" zones, the burn time shoots up exponentially as the crowd gets tighter. In the "empty" zones, it stays steady.
• The Bad News: The formula isn't perfect. Because the air is so chaotic, the particles move around while they are burning. Sometimes a particle starts in a crowd but drifts out, or a crowd splits up. This makes it hard to predict the exact burn time for a single particle just by looking at where it started.

The Takeaway

This study is a crucial step toward building real iron-fuel power plants.

• The Lesson: If you want iron to burn efficiently, you need to prevent it from clumping up. If it clumps, the fire sputters and takes forever to finish.
• The Future: Engineers now know that simply blowing air isn't enough; they need to design their furnaces to break up these "crowds" of iron particles so every single one gets a fair share of oxygen.

In short: Turbulence turns iron fuel into a game of "musical chairs" where the music stops, and the particles huddle together. The ones in the middle of the huddle starve, making the whole process inefficient. To fix this, we need to keep the particles dancing apart, not huddled in a corner.

Technical Summary

1. Problem Statement

Iron powder is emerging as a promising carbon-free renewable energy carrier. However, in industrial combustors, iron particles burn in a turbulent, particle-laden flow. A critical phenomenon in such flows is preferential concentration, where turbulence causes particles to cluster into dense regions (clusters) and sparse regions (voids).

The core problem addressed is understanding how this clustering affects the combustion process. Specifically, the authors investigate:

• How much does clustering extend the total combustion time compared to a random (Poisson) distribution?
• How do the turbulent Reynolds number ($Re_\lambda$) and global equivalence ratio ($\phi$) influence this extension?
• Can the combustion time of individual particles be deterministically predicted based on their initial spatial distribution (clustering)?

2. Methodology

The study employs Direct Numerical Simulations (DNS) of gas-phase point-particle interactions within a forced Homogeneous Isotropic Turbulence (HIT) field.

• Numerical Setup:

  • Domain: A periodic cubical box with a uniform Cartesian grid.
  • Gas Phase: Compressible Navier-Stokes equations solved using the NTMIX-CHEMKIN solver (8th-order spatial, 3rd-order temporal). Turbulence is enforced via a forcing scheme to maintain constant $Re_\lambda$.
  • Particle Phase: Iron particles are modeled as Lagrangian point particles.
    • Combustion Model: A "switch" model determines the oxygen consumption rate, limited by either solid-state diffusion of Fe ions through the FeO layer or gas-phase diffusion of $O_2$ to the surface.
    • Thermodynamics: Includes convection, radiation, and evaporation of Fe and FeO.
  • Coupling: Two-way coupling is applied for mass, momentum, and energy.

• Simulation Parameters:

  • Three sets of simulations were conducted varying:
    1. Stokes Number ($St$): 1, 10, 50 (based on Kolmogorov timescale).
    2. Turbulent Reynolds Number ($Re_\lambda$): 5, 10, 20.
    3. Global Equivalence Ratio ($\phi$): 0.25, 0.5, 0.75 (assuming FeO as the oxidation product).
  • Initial Conditions: Particles initialized in a Poisson (random) distribution at $T=1200$ K. They evolve into clusters under turbulence before ignition ($t=100$ ms).
  • Analysis Metrics:
    • Combustion Time ($\tau_B$): Normalized against a 0D constant-volume suspension model ($\tau^*_B$).
    • Clustering Metric: Voronoï decomposition was used to calculate the normalized Voronoï volume ($V_{norm}$) for each particle. Small $V_{norm}$ indicates dense clusters; large $V_{norm}$ indicates voids.

3. Key Contributions

1. Quantification of Clustering Effects: The study provides a rigorous statistical framework linking particle spatial distribution (via Voronoï volumes) to combustion kinetics in turbulent iron flames.
2. Identification of Regimes: It distinguishes between "cluster" regions (particle-dense) and "void" regions (particle-sparse), revealing distinct combustion behaviors in each.
3. Mechanism of Time Extension: The paper identifies that while microstructure (local clustering) dictates local oxygen depletion, the macrostructure (proximity of multiple clusters) is a critical, often overlooked factor in extending combustion times.
4. Predictive Framework: It proposes a semi-empirical correlation to estimate combustion time extension based on initial Voronoï volumes, while highlighting the limitations of deterministic prediction due to complex inter-cluster interactions.

4. Key Results

A. Impact of Clustering on Combustion Dynamics

• Temperature Evolution: Clustered distributions exhibit a slower and smoother temperature rise with a lower peak mean temperature compared to Poisson distributions. This is due to localized $O_2$ depletion in dense clusters.
• Combustion Time Extension:
  • The total combustion time for a clustered distribution is significantly extended compared to a random distribution.
  • At $Re_\lambda = 20$ and $\phi = 0.75$, the total combustion time can be extended by up to 8 times compared to the 0D suspension model.
  • However, 50% of the particle population in clustered distributions burns at a time comparable to the random distribution; the extension is driven by a fraction of particles in the densest regions.

B. Influence of Parameters

• Stokes Number ($St$): Clustering magnitude is strongly sensitive to $St$. Significant clustering (deviation from Poisson statistics) was observed primarily at $St=1$ under the simulated low $Re_\lambda$ conditions.
• Reynolds Number ($Re_\lambda$): Increasing $Re_\lambda$ enhances the magnitude of clustering (higher standard deviation of Voronoï volumes) but does not significantly alter the timescale of cluster formation. Higher $Re_\lambda$ leads to longer combustion times due to stronger clustering.
• Equivalence Ratio ($\phi$): Increasing $\phi$ significantly extends the completion time of combustion. Higher $\phi$ leads to faster $O_2$ depletion in particle-rich regions, causing a "starvation" effect that prolongs burning.

C. Correlation between Spatial Distribution and Combustion Time

The study analyzed the relationship between the initial normalized Voronoï volume ($V_{norm}$) and the normalized combustion time ($\tau^*_B$):

• Cluster Region ($\log_{10}(V_{norm}) < -0.5$): Shows a strong logarithmic dependence. As $V_{norm}$ decreases (denser clusters), $\tau^*_B$ increases exponentially. Particles in the center of clusters burn significantly longer due to severe $O_2$ depletion.
  • Fit: $\tau^*_B \approx -7.908 \cdot \log_{10}(V_{norm}) - 2.557$
• Void Region ($\log_{10}(V_{norm}) > -0.5$): Shows an asymptotic behavior. Particles in voids burn faster, approaching the isobaric combustion time (uncoupled mode).
  • Fit: $\tau^*_B \approx 0.69 + 0.235 / V_{norm}$

D. Limitations of Spatial Prediction

• Time-Averaging: Time-averaging $V_{norm}$ over the combustion duration did not significantly improve the correlation with $\tau^*_B$.
• The "Macrostructure" Gap: The study found that the initial intra-cluster structure ($V_{norm}$) alone cannot deterministically predict combustion times.
  • Reason: The proximity of adjacent clusters (inter-cluster structure) creates larger, merged $O_2$ depletion zones. Particles in isolated clusters burn faster than those in adjacent clusters, even if their local $V_{norm}$ is identical.
  • Consequently, models based solely on initial particle distribution tend to underpredict combustion times for particles in multi-cluster regions and overpredict for isolated clusters.

5. Significance and Conclusion

This research provides critical insights for the development of iron powder combustion technology for carbon-free energy storage.

• Industrial Relevance: It highlights that simply increasing turbulence or equivalence ratio can drastically alter combustion efficiency and residence time requirements due to preferential concentration.
• Design Implications: Reactor designs must account for the potential for "dead zones" where oxygen is depleted in dense particle clusters, leading to incomplete combustion or significantly extended burn times.
• Future Directions: The study suggests that future models must move beyond single-particle or intra-cluster statistics to include inter-cluster macrostructures (cluster spacing and adjacency) to accurately predict combustion duration in turbulent flows.

In summary, while preferential concentration significantly extends iron particle combustion times through local oxygen depletion, the complex, dynamic interaction between multiple clusters makes deterministic prediction based solely on initial particle distribution challenging.