Recursive regularisation in lattice Boltzmann magnetohydrodynamics

arXiv:2602.14667 · physics.flu-dyn, physics.comp-ph · Submitted 2026-02-16 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Recursive regularisation in lattice Boltzmann magnetohydrodynamics".

Mira: Here's the gist of the paper. Recursive regularisation in lattice Boltzmann magnetohydrodynamics presents a new method for enhancing numerical stability and accuracy in incompressible magnetohydrodynamic (MHD) flows by recursively reconstructing non-equilibrium moments from lower-order information The proposed formulation provides a robust and versatile framework for MHD simulations and offers a systematic route for extending regularised lattice Boltzmann methods to coupled multiphysics systems.

Kai: The approach is based on a hybrid double-distribution strategy where the magnetic field is evolved using the standard BGK lattice Boltzmann scheme, while a dedicated regularisation procedure is applied to the fluid solver.

Mira: This enhancement involves a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice and avoids explicit evaluation of velocity gradients.

Kai: The resulting non-equilibrium regularised populations are reconstructed from physically consistent Hermite coefficients, which is expressed as f neq reg i = wi 1/2c 4s a(2)1,αβH(2)i,αβ + 1/6c6s a(3)1,αβγH(3)i,αβγ + 1/24c8s a(4)1,αβγδH(4)i,αβγδ.

Title and authors: Kai: The title itself, "Recursive regularisation in lattice Boltzmann magnetohydrodynamics," points right at what they are doing here, focusing on that recursive method for MHD flows. It suggests an iterative way of refining the solution.

Mira: And it immediately tells us that this isn't just a small tweak; they are building a whole new framework for handling the complexities of magnetohydrodynamics using lattice Boltzmann methods. They aren't just looking at one part, like the fluid or just the magnetic field separately.

Lev: From my side, when I hear "recursive regularisation," I immediately think about how that translates to real hardware. It implies a process where you use some initial information to build a better version of your solution, and then use that better version to refine the next step.

Kai: Right, so they are using this recursive reconstruction of non-equilibrium populations from those lower-order pieces to enhance the fluid solver part of the model while keeping the magnetic field evolution standard.

The paper's summary: Mira: The summary explains that they use a hybrid double-distribution strategy where the magnetic field follows a standard BGK lattice Boltzmann scheme, and then they apply this specific recursive regularisation procedure to the fluid solver itself. It’s about using a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice to filter out those non-hydrodynamic contributions that cause trouble.

Kai: So, they are taking a standard approach, like BGK for the magnetic field, and then injecting this Hermite-based reconstruction into how they calculate the fluid's non-equilibrium parts. It sounds like a way to keep things stable without having to explicitly calculate all those tricky velocity gradients.

Lev: That avoiding explicit evaluation of velocity gradients is a big thing for me when thinking about running this on real hardware. Explicit derivatives are computationally expensive and can introduce numerical noise that just blows up the simulation quickly, especially in those low viscosity regimes where you need stability most.

Mira: And the way they reconstruct the non-equilibrium distribution from these physically consistent Hermite coefficients is what makes it work, because they are ensuring that whatever they get back is physically sensible within the context of their chosen expansion.

The paper's improvements: Kai: The main improvement they highlight is that this method preserves the correct incompressible MHD limit, and it actually improves numerical stability at low viscosities while reducing lattice-dependent artefacts in the solution. That means even when you are simulating things that are very thin or have very low magnetic diffusivity, the results should be cleaner than before.

Mira: It also addresses a problem mentioned earlier: the BGK model is known to suffer from limited numerical stability at low viscosities and sensitivity to lattice non-hydrodynamic modes, which is definitely amplified in MHD flows because of the Lorentz force and how energy transfers across different spatial scales.

Lev: If we were trying to run this on a real system, that improved stability at low viscosity would be huge for us. It means we could push the limits of accuracy in those difficult physical regimes without immediately hitting numerical breakdown, which is where most traditional schemes fail on hardware.

Kai: And they show this works well when tested against the Orszag–Tang vortex, which is a standard benchmark for non-linear MHD dynamics with strong velocity and magnetic field coupling. The results match the spectral reference data for peak current density and vorticity in the early stages of evolution.

Conclusion: Mira: So, to wrap up, this paper shows that recursive regularisation is robust enough to be transferred to coupled MHD systems in a simple and systematic way while still preserving the correct incompressible MHD limit. It offers a practical route for regularised lattice Boltzmann solvers in broader multiphysics settings.

Kai: It confirms that the method captures key physical mechanisms, like peak current intensification and vorticity growth, which is essential for understanding how MHD turbulence works, even though it doesn's be a complete fix for everything.

Lev: From an error-correction perspective, I see that while this recursive regularisation is robust compared to moment-based approaches, it still comes with a moderate increase in per time step cost because of all those Hermite projections and recursive reconstructions they have to do.

Mira: That’s true, the runtime is higher than the BGK model, but the paper shows that this extra cost is manageable because it preserves the strictly local character of the lattice Boltzmann method and keeps linear scaling with system size.

Kai: So, we're left with a method that offers improved numerical robustness at coarser resolutions or under more demanding physical conditions by offsetting that additional cost. It’s a versatile framework for simulations, and it gives us a systematic way to extend regularised lattice Boltzmann methods into coupled multiphysics systems.

Alessandro De Rosis

Department of Mechanical and Aerospace Engineering, The University of Manchester

physics.flu-dyn, physics.comp-ph

Submitted: 2026-02-16

Updated: 2026-11-15

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: Recursive regularisation in lattice Boltzmann magnetohydrodynamics presents a new method for enhancing numerical stability and accuracy in incompressible magnetohydrodynamic (MHD) flows by…

Key concepts

Recursive Regularisation
This technique enhances the fluid solver by recursively reconstructing non-equilibrium populations from lower-order information using a fourth-order Hermite expansion. It filters out spurious non-hydrodynamic contributions without needing explicit velocity gradients, improving isotropy and robustness in flows.
Hermite Expansion
The equilibrium distribution function is expanded using a fourth-order Hermite polynomial on the D2Q9 lattice. This mathematical tool allows the method to capture higher-order physical details of the distribution while avoiding the need to directly calculate velocity gradients, which are computationally intensive.
Incompressible MHD Limit
The formulation is designed to correctly reproduce the physics of incompressible magnetohydrodynamics, ensuring that simulations adhere to the solenoidal constraint (∇ · b = 0). This preservation of the physical limit is crucial for maintaining fidelity in magnetic field evolution and physical accuracy.

Terminology

Summary

Recursive regularisation in lattice Boltzmann magnetohydrodynamics presents a new method for enhancing numerical stability and accuracy in incompressible magnetohydrodynamic (MHD) flows by recursively reconstructing non-equilibrium moments from lower-order information The proposed formulation provides a robust and versatile framework for MHD simulations and offers a systematic route for extending regularised lattice Boltzmann methods to coupled multiphysics systems. p. 0

Where the paper says it

The resulting scheme preserves the correct incompressible MHD limit, improves numerical stability at low viscosities, and reduces lattice-dependent artefacts. The proposed formulation provides a robust and versatile framework for MHD simulations and offers a systematic route for extending regularised lattice Boltzmann methods to coupled multiphysics systems.

Methodology of the Recursive Regularisation

The approach is based on a hybrid double-distribution strategy where the magnetic field is evolved using the standard BGK lattice Boltzmann scheme, while a dedicated regularisation procedure is applied to the fluid solver. This enhancement involves a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice and avoids explicit evaluation of velocity gradients. p. 1

Where the paper says it

We adopt a hybrid double-distribution strategy in which the magnetic field is evolved using the BGK scheme proposed by Dellar10 , while the fluid solver is enhanced through recursive regularisation of the non-equilibrium moments. The approach is based on a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice and avoids the explicit evaluation of velocity gradients.

The resulting non-equilibrium regularised populations are reconstructed from physically consistent Hermite coefficients, which is expressed as f neq reg i = wi 1/2c 4s a(2)1,αβH(2)i,αβ + 1/6c6s a(3)1,αβγH(3)i,αβγ + 1/24c8s a(4)1,αβγδH(4)i,αβγδ. p. 3
Where the paper says it

Compute the non-equilibrium Hermite coefficients: a (2) 1,αβ = X i H (2) i,αβ (fi − f eq i ), a (3) 1,αβγ = uαa (2) 1,βγ + uβa (2) 1,αγ + uγa (2) 1,αβ, a (4) 1,αβγδ = uαa (3) 1,βγδ + uβa (3) 1,αγδ + uγa (3) 1,αβδ + uδa (3) 1,αβγ− h uαuβa (2) 1,γδ + uαuγa (2) 1,βδ + uαuδa (2) 1,βγ+ uβuγa (2) 1,αδ + uβuδa (2) 1,αγ + uγuδa (2) 1,αβi . Reconstruct the non-equilibrium regularised populations: f neq,reg i = wi 1 2c 4 s a (2) 1,αβH (2) i,αβ + 1 6c 6 s a (3) 1,αβγH (3) i,αβγ+ 1 24c 8 s a (4) 1,αβγδH (4) i,αβγδ .

Validation and Performance Across Regimes

The method is validated using the Orszag–Tang vortex as a canonical benchmark for non-linear MHD dynamics featuring strong velocity–magnetic field coupling. p. 1

Where the paper says it

The approach is based on a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice and avoids the explicit evaluation of velocity gradients. The method is validated using the Orszag–Tang vortex, a canonical benchmark for non-linear MHD dynamics featuring strong velocity–magnetic field coupling, current-sheet formation, and a transition toward MHD turbulence26.

Numerical results show that in the well-resolved early evolution, all schemes closely match spectral reference data for peak current density and vorticity. p. 4
Where the paper says it

5), the flow remains smooth and well resolved. In this regime, all lattice Boltzmann formulations exhibit excellent agreement with the reference data.

As the flow develops sharp gradients and current sheets, the recursive regularisation introduces a modest additional dissipation on coarse grids, leading to slightly lower extrema compared to BGK. p. 5
Where the paper says it

For instance, at Re = 2500, variations in jmax between N = 500 and N = 2000 remain below 1%, indicating that modest resolutions are sufficient prior to current-sheet formation. As reconnection develops (t ≳ 1), sharp gradients emerge and the standard BGK formulation becomes unstable on coarse grids, systematically failing as current concentrations intensify.

Importantly, this discrepancy diminishes under grid refinement: on the finest meshes, the RR results converge to the same level of agreement achieved by the competing collision operators. p. 4
Where the paper says it

Importantly, these differences vanish rapidly under grid refinement. At N = 512, all collision models converge toward the spectral reference with relative discrepancies of approximately 1.

Stability and Constraint Preservation

The preservation of the solenoidal constraint, ∇ · b = 0, is essential for physical fidelity in MHD simulations. p. 5

Where the paper says it

Results obtained with different collision models are compared against a spectral reference solution. The preservation of the solenoidal constraint, ∇ · b = 0, is essential for physical fidelity in MHD simulations.

All tested lattice Boltzmann formulations display clear rapid decay of the maximum divergence error with increasing resolution, and the differences between collision models remain negligible. This indicates that the divergence behaviour is primarily controlled by the discrete magnetic-field representation (D2Q5) rather than by the regularisation of non-equilibrium moments. p. 5
Where the paper says it

Notably, the LBM results remain slightly below the spectral reference values. The near-identical behaviour across collision models demonstrates that divergence control is primarily governed by the discrete magnetic-field representation rather than by the specific regularisation of non-equilibrium moments.

Computational Trade-offs

The recursive regularisation entails a moderate increase in per-time-step cost due to the additional Hermite projections and recursive reconstructions performed during the collision step. Nevertheless, the algorithm preserves the strictly local character of the lattice Boltzmann method and maintains linear scaling with system size. p. 2

Where the paper says it

(21) It is worth noting that the LBM allows us to compute the electric current jz = ∇ × b locally from the populations as28 jz = − ωm c 2 s εαβ X j ej,αgj,β − ej,αg eq j,β , (22) where X j ej,αg eq j,β = uβbα − uαbβ, (23) and εαβ is the Levi-Civita permutation tensor, without the need to evaluate non-local finite-difference operators. Recursive regularised lattice Boltzmann method We now derive a recursive regularisation strategy for magnetohydrodynamic flows.

The RR model displays the highest runtime. p. 4
Where the paper says it

94% for jmax and 10. 09% for ωmax, while the recursive regularised model displays marginally higher dissipation, reaching 4.

p. 4
Where the paper says it

94% for jmax and 10. 09% for ωmax, while the recursive regularised model displays marginally higher dissipation, reaching 4.

However, the near-perfect overlap of the RMs and CMs curves in Fig. p. 5
Where the paper says it

The near-identical behaviour across collision models demonstrates that divergence control is primarily governed by the discrete magnetic-field representation rather than by the specific regularisation of non-equilibrium moments. 10 BGK RMs CMs RR 0.

5 indicates that the additional velocity-dependent shift required by the CMs approach incurs a negligible cost relative to the dominant streaming operations.

Conclusion on Robustness

Recursive regularisation demonstrates robustness comparable to moment-based approaches while retaining the simplicity and efficiency of a single-relaxation-time framework. The accurate capture of peak current intensification, vorticity growth, and magnetic topology evolution confirms that the method preserves the essential physical mechanisms governing MHD turbulence. p. 1

Where the paper says it

The method is validated using the Orszag–Tang vortex, a canonical benchmark for non-linear MHD dynamics featuring strong velocity–magnetic field coupling, current-sheet formation, and a transition toward MHD turbulence26. The impact of regularisation on numerical stability, dissipation, and the accurate capture of key MHD features are assessed.

The improved numerical robustness offered by the RR scheme can offset this additional cost by enabling stable simulations at coarser resolutions or under more demanding physical conditions.

Pruning redundant reasoning tokens

The paper presents a recursive regularised lattice Boltzmann method for incompressible magnetohydrodynamic (MHD) flows. The approach is based on a double-distribution formulation, in which the magnetic field is evolved using a standard BGK lattice Boltzmann scheme, while the fluid solver is enhanced through a Hermite-based recursive regularisation of the non-equilibrium moments. p. 0

Where the paper says it

Recursive regularised lattice Boltzmann method for magnetohydrodynamics Alessandro De Rosis1 Department of Mechanical and Aerospace Engineering, The University of Manchester, Manchester M13 9PL, United Kingdom (Dated: 18 February 2026) We present and test a recursive regularised lattice Boltzmann method for incompressible magnetohydrodynamic (MHD) flows. The approach is based on a double-distribution formulation, in which the magnetic field is evolved using a standard BGK lattice Boltzmann scheme, while the fluid solver is enhanced through a Hermite-based recursive regularisation of the non-equilibrium moments.

The method exploits a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice, allowing higher-order isotropy to be retained while selectively filtering spurious non-hydrodynamic contributions. p. 0
Where the paper says it

The approach is based on a double-distribution formulation, in which the magnetic field is evolved using a standard BGK lattice Boltzmann scheme, while the fluid solver is enhanced through a Hermite-based recursive regularisation of the non-equilibrium moments. The method exploits a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice, allowing higher-order isotropy to be retained while selectively filtering spurious non-hydrodynamic contributions.

How the training signal is built

The non-equilibrium distribution is reconstructed from physically consistent Hermite coefficients, avoiding explicit evaluation of velocity gradients. The regularisation procedure reconstructs the non-equilibrium distribution from physically consistent Hermite coefficients, avoiding explicit evaluation of velocity gradients. p. 0

Where the paper says it

The method exploits a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice, allowing higher-order isotropy to be retained while selectively filtering spurious non-hydrodynamic contributions. The regularisation procedure reconstructs the non-equilibrium distribution from physically consistent Hermite coefficients, avoiding explicit evaluation of velocity gradients.

The resulting scheme preserves the correct incompressible MHD limit, improves numerical stability at low viscosities, and reduces lattice-dependent artefacts. p. 0
Where the paper says it

The regularisation procedure reconstructs the non-equilibrium distribution from physically consistent Hermite coefficients, avoiding explicit evaluation of velocity gradients. The resulting scheme preserves the correct incompressible MHD limit, improves numerical stability at low viscosities, and reduces lattice-dependent artefacts.

Where the gains come from

The recursive regularisation procedure systematically reconstructs higher-order non-equilibrium moments from lower-order information, leading to improved isotropy and robustness in hydrodynamic flows. This approach is based on a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice and avoids the explicit evaluation of velocity gradients. p. 1

Where the paper says it

We adopt a hybrid double-distribution strategy in which the magnetic field is evolved using the BGK scheme proposed by Dellar10 , while the fluid solver is enhanced through recursive regularisation of the non-equilibrium moments. The approach is based on a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice and avoids the explicit evaluation of velocity gradients.

The method is validated using the Orszag–Tang vortex, a canonical benchmark for non-linear MHD dynamics featuring strong velocity–magnetic field coupling. p. 1
Where the paper says it

The approach is based on a fourth-order Hermite expansion of the equilibrium distribution on the D2Q9 lattice and avoids the explicit evaluation of velocity gradients. The method is validated using the Orszag–Tang vortex, a canonical benchmark for non-linear MHD dynamics featuring strong velocity–magnetic field coupling, current-sheet formation, and a transition toward MHD turbulence26.

Synthesis across flow regimes

Taken together, the low-Reynolds-number and turbulent results provide a coherent picture of the role of non-equilibrium control in MHD lattice Boltzmann models. In mildly non-linear regimes dominated by smooth structures, all collision operators perform equivalently, confirming that regularisation does not degrade accuracy in resolved flows. p. 8

Where the paper says it

Synthesis across flow regimes Taken together, the low-Reynolds-number and turbulent results provide a coherent picture of the role of non-equilibrium control in MHD lattice Boltzmann models. In mildly non-linear regimes dominated by smooth structures, all collision operators perform equivalently, confirming that regularisation does not degrade accuracy in resolved flows.

As the dynamics transition toward strongly non-linear behaviour characterised by thin current sheets, magnetic reconnection, and intense shear layers, the numerical challenge shifts from accuracy to stability and robustness. p. 8
Where the paper says it

In mildly non-linear regimes dominated by smooth structures, all collision operators perform equivalently, confirming that regularisation does not degrade accuracy in resolved flows. As the dynamics transition toward strongly non-linear behaviour characterised by thin current sheets, magnetic reconnection, and intense shear layers, the numerical challenge shifts from accuracy to stability and robustness.

Final Summary Statement

In summary, our results demonstrate that recursive regularisation can be transferred to coupled MHD systems in a simple and systematic manner, delivering a robust stabilisation strategy while preserving the correct incompressible MHD limit. Beyond the benchmark considered here, the proposed formulation provides a practical route to regularised lattice Boltzmann solvers for broader multiphysics settings. p. 0

Where the paper says it

The resulting scheme preserves the correct incompressible MHD limit, improves numerical stability at low viscosities, and reduces lattice-dependent artefacts. The proposed formulation provides a robust and versatile framework for MHD simulations and offers a systematic route for extending regularised lattice Boltzmann methods to coupled multiphysics systems.

Computational cost

The computational cost is discussed in Fig. 5, which summarizes the normalised runtime of the four lattice Boltzmann models as a function of the total number of grid points, N2. p. 9

Where the paper says it

Computational cost We conclude the performance assessment of the lattice Boltzmann formulations for MHD with an evaluation of their computational cost. 5 summarises the normalised runtime of the four lattice Boltzmann models as a function of the total number of grid points, N2 .

The RR model displays the highest runtime. This behaviour is a direct consequence of the more elaborate collision step, which involves the projection of non-equilibrium populations onto Hermite polynomials followed by recursive reconstruction procedures. p. 7
Where the paper says it

Across both turbulent regimes, the recursive regularisation consistently reproduces the converged extrema with accuracy comparable to moment-based formulations while preserving stability even on the coarsest grids. This demonstrates that recursive filtering of non-equilibrium moments is sufficient to control under-resolved dynamics and accurately capture reconnection-driven intermittency, without requiring explicit moment reconstruction.

The RR approach trades increased runtime for enhanced numerical robustness. p. 0
Where the paper says it

Recursive regularised lattice Boltzmann method for magnetohydrodynamics Alessandro De Rosis1 Department of Mechanical and Aerospace Engineering, The University of Manchester, Manchester M13 9PL, United Kingdom (Dated: 18 February 2026) We present and test a recursive regularised lattice Boltzmann method for incompressible magnetohydrodynamic (MHD) flows. The approach is based on a double-distribution formulation, in which the magnetic field is evolved using a standard BGK lattice Boltzmann scheme, while the fluid solver is enhanced through a Hermite-based recursive regularisation of the non-equilibrium moments.

The single most informative sentence about the paper's subject and finding

The proposed formulation provides a robust and versatile framework for MHD simulations and offers a systematic route for extending regularised lattice Boltzmann methods to coupled multiphysics systems.

The magnetic field is evolved using auxiliary vector-valued distribution functions defined on a D2Q5 lattice. The relaxation frequencies ω and ωm are related to the kinematic viscosity ν and magnetic diffusivity η through ν = 1/(ω - 1/2) and η = 1/(ωm - 1/2) respectively. p. 2

Where the paper says it

3 g ⋆ j (x, t) = g eq j (x, t) + (1 − ωm) g neq j (x, t). (11) The relaxation frequencies ω and ωm are related to the kinematic viscosity ν and magnetic diffusivity η through ν = 1 ω − 1 2 , η = 1 ωm − 1 2 , (12) respectively.

The equilibrium distribution f eq i is fixed throughout this study to the fourth-order Hermite formulation given in Eq. p. 2
Where the paper says it

(19) The Hermite tensors are defined as H (0) i = 1, H (1) i,α = ci,α, H (2) i,αβ = ci,αci,β − c 2 s δαβ, H (3) i,αβγ = ci,αci,βci,γ − c 2 s (ci,αδβγ + ci,βδαγ + ci,γδαβ), H (4) i,αβγδ = ci,αci,βci,γci,δ − c 2 s X perm ci,αci,βδγδ+ c 4 s X perm δαβδγδ, (20) where δαβ is the Kronecker delta and (α, β, γ, δ) ∈ {x, y} 4 . The equilibrium distribution f eq i is fixed throughout this study to the fourth-order Hermite formulation given in Eq.

(19). The macroscopic fields are obtained as ρ = X i fi, u = P i fici ρ, b = X j gj. p. 2
Where the paper says it

The equilibrium distribution f eq i is fixed throughout this study to the fourth-order Hermite formulation given in Eq. Finally, the macroscopic fields are obtained as ρ = X i fi , u = P i fici ρ , b = X j gj .

The LBM allows us to compute the electric current jz = ∇ × b locally from the populations as 28. p. 2
Where the paper says it

Finally, the macroscopic fields are obtained as ρ = X i fi , u = P i fici ρ , b = X j gj . (21) It is worth noting that the LBM allows us to compute the electric current jz = ∇ × b locally from the populations as28 jz = − ωm c 2 s εαβ X j ej,αgj,β − ej,αg eq j,β , (22) where X j ej,αg eq j,β = uβbα − uαbβ, (23) and εαβ is the Levi-Civita permutation tensor, without the need to evaluate non-local finite-difference operators.

The initial velocity and magnetic fields are prescribed as u(x, t = 0) = u0 [− sin(y), sin(x)], b(x, t = 0) = b0 [− sin(y), sin(2x)]. The flow is characterised by the Reynolds and magnetic Prandtl numbers, with the former being defined as Re = u0L ν (35) where L = 2π denotes the characteristic length scale, ν is the kinematic viscosity, and u0 = b0 = 2. p. 4

Where the paper says it

All simulations are performed in a doubly periodic square domain Ω = [0, 2π]×[0, 2π], discretised using a uniform Cartesian grid of N × N lattice nodes. The flow is characterised by the Reynolds and magnetic Prandtl numbers, with the former being defined as Re = u0L ν , (35) where L = 2π denotes the characteristic length scale, ν is the kinematic viscosity, and u0 = b0 = 2 being the root-mean-square of the fields in Eqs.

The Mach number of our simulations is set to Ma = ˜u0/cs ≃ 0. p. 4
Where the paper says it

The flow is characterised by the Reynolds and magnetic Prandtl numbers, with the former being defined as Re = u0L ν , (35) where L = 2π denotes the characteristic length scale, ν is the kinematic viscosity, and u0 = b0 = 2 being the root-mean-square of the fields in Eqs. The Mach number of our simulations is set to Ma = ˜u0/cs ≃ 0.

009, where ˜u0 = 0. p. 4
Where the paper says it

The Mach number of our simulations is set to Ma = ˜u0/cs ≃ 0. 009, where ˜u0 = 0.

2 π−1 is the reference velocity expressed in lattice units. p. 4
Where the paper says it

009, where ˜u0 = 0. 2 π −1 is the reference velocity expressed in lattice units.

Spatial derivatives are consistently evaluated using the isotropic lattice gradient ∂αuβ(x) = c−2s X i [wiuβ(x + ci)ci,α] and ∂αbβ(x) = c−2s X i [wibβ(x + ci)ci,α]. p. 4
Where the paper says it

The choice of ˜u0 as a multiple of π −1 further facilitates sampling of flow quantities at salient discrete time steps. Spatial derivatives are consistently evaluated using the isotropic lattice gradient ∂αuβ(x) = c −2 s X i [wiuβ(x + ci)ci,α] , ∂αbβ(x) = c −2 s X i [wibβ(x + ci)ci,α] , (36) which is fully consistent with the underlying lattice symmetry31–33 .

The preservation of the solenoidal constraint, ∇ · b = 0, is essential for physical fidelity in MHD simulations.

The proposed formulation provides a robust and versatile framework for MHD simulations and offers a systematic route for extending regularised lattice Boltzmann methods to coupled multiphysics systems. The method is validated using the Orszag–Tang vortex as a canonical benchmark for non-linear MHD dynamics featuring strong velocity–magnetic field coupling. The impact of regularisation on numerical stability, dissipation, and the accurate capture of key MHD features are assessed. p. 1

Where the paper says it

The method is validated using the Orszag–Tang vortex, a canonical benchmark for non-linear MHD dynamics featuring strong velocity–magnetic field coupling, current-sheet formation, and a transition toward MHD turbulence26. The impact of regularisation on numerical stability, dissipation, and the accurate capture of key MHD features are assessed.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Recursive regularisation can be transferred to coupled MHD systems in a simple and systematic manner, delivering a robust stabilisation strategy while preserving the correct incompressible MHD limit. This allows AI systems to reliably simulate complex multiphysics flows, such as those involving plasma physics or geophysical flows, without introducing spurious lattice artifacts in the fluid solver.

  2. The method enables AI systems to accurately capture reconnection-driven intermittency and magnetic energy transfer across multiple scales in turbulent regimes. This capability is crucial for simulating astrophysical phenomena where magnetic field topology undergoes rapid changes, ensuring that the simulation reflects the true physical mechanisms of MHD turbulence rather than numerical dissipation artifacts.

  3. The formulation preserves the correct incompressible MHD limit and improves numerical stability at low viscosities while reducing lattice-dependent artefacts. AI systems can therefore be deployed in regimes characterized by low magnetic diffusivity or strong Lorentz forces where traditional methods fail, leading to more accurate and stable solutions for liquid-metal technologies or fusion applications.

  4. The scheme demonstrates robustness comparable to moment-based approaches in turbulent regimes while retaining the simplicity and efficiency of a single-relaxation-time framework. This means AI systems can achieve high accuracy in complex turbulent flows without requiring the computational overhead associated with explicit moment reconstruction, balancing numerical robustness and computational cost effectively.

Abstract

Recursive regularisation (RR) improves the stability of lattice Boltzmann schemes, but its structure and limitations in magnetohydrodynamics (MHD) remain insufficiently understood. This work connects RR to central-Hermite-moment (CHM) collisions on D2Q9 and D3Q27 through a closed-form reconstruction valid for arbitrary second-order relaxation rates. The central moments of the Maxwell stress coincide, up to sign, with the recursion coefficients, providing a systematic basis for MHD collision operators. Common two-dimensional raw- and central-moment schemes thus correspond to regularisation with a unit bulk relaxation rate. In three dimensions, a CHM implementation reproduces RR to round-off, while a monomial central-moment collision differs at fourth order and produces a four- to twelve-fold larger kinetic-energy error in the tested flow. Chapman--Enskog analysis shows that, with a magnetic field, the hydrodynamic recursion reconstructs the third-order non-equilibrium with a relative error of 78 -- 96% (4% without a field). A local correction largely removes it, yet changes the solutions by less than their discretisation error, consistent with its Burnett-order character. Three-dimensional vortex simulations show approximately second-order convergence at Re=100 and agree with pseudo-spectral solutions through the peak current at Re=1000 and, with a unit bulk rate, at Re=3000. There, single-rate RR develops a compressive instability whose onset depends on the grid, lattice Mach number and initial magnetic configuration. For the standard initial field, a unit bulk rate suppresses it at Ma=0.034 for N at least128. This improvement does not extend to BGK collisions or prevent failure at Re=10,000. These results show that robust three-dimensional MHD simulations require regularisation and bulk dissipation to be considered together.

Sources

Related papers