Recursive regularised lattice Boltzmann method for magnetohydrodynamics

This paper presents and validates a robust recursive regularised lattice Boltzmann method for incompressible magnetohydrodynamics that utilizes a double-distribution formulation and fourth-order Hermite expansion to enhance numerical stability, reduce lattice-dependent artifacts, and eliminate the need for explicit velocity gradient calculations.

The Gist

Imagine you are trying to simulate a chaotic dance between two invisible partners: fluid (like water or molten metal) and magnetism. In the real world, these two are locked in a tight embrace; the fluid moves the magnetic field, and the magnetic field pushes back on the fluid. This is called Magnetohydrodynamics (MHD).

Scientists want to predict how this dance plays out on a computer, but it's incredibly difficult. The dance involves tiny, fast-moving swirls and sudden, violent snaps (called "reconnection") that are hard to capture without the simulation either crashing or becoming a blurry mess.

This paper introduces a new, smarter way to run these simulations. Here is the breakdown in simple terms:

1. The Old Way: The "Rough Sketch" Artist

The most common way to simulate fluids on a computer is called the Lattice Boltzmann Method (LBM). Think of this as a grid of tiny tiles. On each tile, we track little "particles" moving around.

• The Problem: The standard method (called BGK) is like a sketch artist who draws very fast but makes mistakes when the scene gets too chaotic. When the magnetic field gets too strong or the fluid gets too thin (low viscosity), the sketch becomes wobbly, the numbers explode, and the simulation crashes.
• The Fix: Scientists have tried other methods (like MRT and CMs) that are like hiring a team of specialists to check the drawing. They work well but are computationally heavy and complex.

2. The New Solution: The "Recursive Regularised" Method

The author, Alessandro De Rosis, proposes a new technique called Recursive Regularisation (RR).

The Analogy: The "Smart Editor"
Imagine the standard simulation is a student writing a story.

• The BGK method: The student writes the story quickly but includes a lot of "noise"—random typos, gibberish words, and sentences that don't make sense physically. When the story gets intense (a storm or a fight), the noise drowns out the plot, and the story falls apart.
• The RR method: This is like a smart editor who reads the student's draft before it's published.
  1. The editor looks at the draft.
  2. Instead of just fixing typos, the editor uses a special rulebook (math called Hermite expansion) to understand the intent of the story.
  3. The editor strips away all the "gibberish" (spurious noise) that doesn't belong in a real physical fluid.
  4. Crucially, the editor does this recursively. They don't just fix the first sentence; they use the logic of the first sentence to fix the second, then the third, and so on.

Why is this special?

• No Calculus Required: Usually, to fix these errors, you need to calculate how fast the fluid is changing speed at every single point (gradients). This is slow and prone to errors. The RR method is clever: it reconstructs the "clean" version of the data just by looking at the particles themselves, skipping the messy math.
• The Hybrid Approach: The author kept the magnetic field simulation simple (using the old, fast method) but applied this "Smart Editor" only to the fluid part. This keeps the code fast but makes the fluid part incredibly stable.

3. The Test: The "Orszag–Tang Vortex"

To prove this works, the author tested it on a famous, difficult scenario called the Orszag–Tang Vortex.

• The Scenario: Imagine two giant whirlpools spinning in opposite directions, with a magnetic field tangled inside them. As they spin, they stretch the magnetic field until it snaps (reconnection), creating a storm of turbulence.
• The Results:
  • Low Speed: When the dance is slow, all methods (Old, New, and Specialist) look the same. They all get the answer right.
  • High Speed (Turbulence): When the dance gets violent, the "Old Method" (BGK) crashes on coarse grids (low-resolution screens). The "Specialist Methods" survive but are heavy.
  • The Winner: The Recursive Regularised method survived the chaos on all grids. It didn't crash, and it captured the sharp, thin lines of the magnetic field (current sheets) without blurring them out.

4. The Trade-off

Is it perfect? Almost.

• The Cost: Because the "Smart Editor" has to do a bit more math to clean up the data, the simulation runs slightly slower than the simplest method (about 10-20% slower).
• The Benefit: However, because it is so stable, you can often use a lower resolution (fewer grid tiles) and still get a good result. This means you might save time overall because you don't need a supercomputer to run it.

Summary

This paper presents a new "Smart Editor" for fluid simulations. It takes a fast but fragile method and adds a layer of mathematical cleanup that removes the "noise" causing crashes. It allows scientists to simulate violent magnetic storms and fluid turbulence on computers that would otherwise fail, making it a powerful tool for understanding everything from solar flares to liquid metal cooling in nuclear reactors.

Technical Summary

1. Problem Statement

Magnetohydrodynamics (MHD) involves the coupled dynamics of electrically conducting fluids and magnetic fields. Numerically simulating incompressible resistive MHD flows is challenging due to:

• Strong Non-linear Coupling: The interaction between velocity and magnetic fields creates complex dynamics.
• Multi-scale Transport: The coexistence of viscous and magnetic diffusion mechanisms.
• Solenoidal Constraints: Both the velocity ($\nabla \cdot \mathbf{u} = 0$) and magnetic field ($\nabla \cdot \mathbf{b} = 0$) must remain divergence-free.
• Numerical Instability: Standard methods often suffer from spurious oscillations, excessive dissipation, or instability at low viscosities (high Reynolds numbers) and strong Lorentz forces.

While the Lattice Boltzmann Method (LBM) offers a mesoscopic alternative to traditional discretizations, standard single-relaxation-time (BGK) models struggle with stability in high-Reynolds-number MHD regimes. Advanced models like Multiple-Relaxation-Time (MRT) or Central Moments (CM) improve stability but often increase computational complexity or break Galilean invariance. There was a lack of systematic application of Recursive Regularisation (RR)—a technique known for improving stability in pure hydrodynamics—to coupled MHD systems.

2. Methodology

The author proposes a Recursive Regularised Lattice Boltzmann Method (RR-LBM) for 2D incompressible MHD flows using a double-distribution function approach:

• Double-Distribution Framework:
  • Fluid Dynamics: Evolved using a scalar distribution function ($f_i$) on a D2Q9 lattice.
  • Magnetic Field: Evolved using vector-valued distribution functions ($g_j$) on a D2Q5 lattice, following the BGK scheme proposed by Dellar.
• Recursive Regularisation (RR) for Fluids:
  • Instead of the standard second-order equilibrium, the method employs a fourth-order Hermite expansion of the equilibrium distribution.
  • Non-equilibrium Reconstruction: The non-equilibrium distribution is reconstructed recursively from lower-order Hermite coefficients.
    • The second-order non-equilibrium coefficient ($a^{(2)}_1$) is computed directly from the distribution functions ($f_i - f^{eq}_i$), avoiding explicit evaluation of velocity gradients ($\nabla \mathbf{u}$).
    • Higher-order coefficients ($a^{(3)}_1, a^{(4)}_1$) are derived recursively from lower-order ones using the macroscopic velocity.
  • Filtering: This process selectively filters spurious non-hydrodynamic modes (lattice artifacts) while retaining physical consistency.
• Hybrid Strategy: The magnetic field solver remains a standard BGK scheme, while the fluid solver is enhanced with RR. This maintains the locality of the algorithm and avoids the high computational cost of full MRT or CM schemes for the magnetic component.

3. Key Contributions

• First Systematic Application: This work presents the first comprehensive investigation of Recursive Regularisation specifically for MHD flows, bridging the gap between hydrodynamic stabilization techniques and coupled multiphysics systems.
• Hybrid Formulation: It successfully integrates RR into Dellar's double-distribution MHD framework, enhancing fluid stability without altering the magnetic evolution scheme.
• Gradient-Free Regularisation: The method reconstructs non-equilibrium moments without requiring explicit finite-difference evaluation of velocity gradients, preserving the local nature of LBM.
• Comprehensive Benchmarking: The study rigorously compares the RR-LBM against three other collision models:
  1. Standard BGK.
  2. Raw Moments (RMs) based MRT.
  3. Central Moments (CMs) based LBM.

4. Results

The method was validated using the Orszag–Tang vortex, a canonical benchmark featuring current sheet formation and transition to turbulence, across various Reynolds numbers ($Re$).

• Low Reynolds Number ($Re = 200\pi$):

  • All models (BGK, RMs, CMs, RR) showed excellent agreement with spectral reference solutions.
  • On coarse grids, RR exhibited slightly higher dissipation (marginally lower peak current/vorticity) compared to BGK, but this discrepancy vanished with grid refinement.
  • All models maintained the solenoidal constraint ($\nabla \cdot \mathbf{b} \approx 0$) effectively, with divergence errors controlled primarily by the magnetic lattice (D2Q5) rather than the fluid collision operator.

• High Reynolds Number / Turbulent Regimes ($Re = 2500, 5000$):

  • Stability: The standard BGK scheme failed (became unstable) on coarse grids as current sheets formed and magnetic reconnection intensified.
  • Robustness: Both RR and moment-based (RMs/CMs) schemes remained stable and converged across all grid resolutions.
  • Accuracy: RR accurately captured the peak current density, vorticity growth, and the transition to MHD turbulence, matching the performance of the more complex moment-based schemes.
  • Physics: The method correctly reproduced the energy transfer from magnetic to kinetic energy during reconnection events and the formation of intermittent current sheets.

• Computational Cost:

  • BGK: Surprisingly, BGK was not the fastest due to the cost of direct population-based equilibrium reconstruction.
  • RMs/CMs: These provided the most efficient performance, offering a balance between stability and speed.
  • RR: Had the highest runtime due to the additional algebraic operations required for Hermite projections and recursive reconstruction. However, it maintained linear scaling and strict locality.

5. Significance

• Robust Stabilization: The study demonstrates that recursive regularisation is a viable and robust strategy for stabilizing MHD simulations, particularly in regimes where standard BGK fails.
• Versatility: It offers a systematic route to extend regularised LBM methods to coupled multiphysics systems beyond pure hydrodynamics.
• Trade-off Analysis: The paper clarifies the trade-off between computational cost and stability. While RR is slightly more expensive than moment-based methods, it provides a simpler implementation than full MRT/CM schemes while delivering comparable stability for high-Re MHD flows.
• Future Directions: The work lays the foundation for extending these methods to 3D lattices and developing consistent regularisation strategies for magnetic populations to further reduce lattice artifacts in extreme transport regimes.

In conclusion, the proposed RR-LBM provides a robust, accurate, and versatile framework for simulating complex MHD phenomena, successfully balancing numerical stability with the preservation of physical incompressible limits.