Add-ons for Lattice Boltzmann Methods: Regularization, Filtering and Limiters

We describe how regularization of lattice Boltzmann methods can be achieved by modifying dissipation. Classes of techniques used to try to improve regularization of LBMs include flux limiters, enforcing the exact correct production of entropy and manipulating non-hydrodynamic modes of the system in relaxation. Each of these techniques corresponds to an additional modification of dissipation compared with the standard LBGK model. Using some standard 1D and 2D benchmarks including the shock tube and lid driven cavity, we explore the effectiveness of these classes of methods.

💡 Research Summary

The paper investigates how to improve the stability and accuracy of lattice Boltzmann methods (LBM) by adding three distinct “add‑ons” that modify the dissipation inherent in the standard LBGK scheme. The authors focus on (i) Entropic Lattice Boltzmann (ELBM), (ii) Entropic‑based flux limiters, and (iii) Multiple‑Relaxation‑Time (MRT) collision operators. All three techniques can be interpreted as introducing additional, targeted dissipation to suppress non‑physical oscillations that arise in regions of sharp gradients such as shock waves or high‑Reynolds‑number flows.

Entropic Lattice Boltzmann (ELBM).
ELBM enforces a discrete H‑theorem by adjusting a scalar factor α in the collision step so that the entropy S remains constant: S(f + α(f⁎ − f)) = S(f). This leads to a nonlinear equation for α that must be solved at every lattice site and time step. The authors describe a robust root‑finding algorithm based on successive quadratic (parabolic) approximations with cubic convergence, supplemented by a Newton‑type correction. When the non‑trivial root does not exist, two fallback strategies are considered: the Ehrenfest rule (force α = 1, i.e., jump directly to equilibrium) and the positivity rule (choose the largest α that keeps all populations non‑negative). Both fallback rules increase local dissipation, especially in highly non‑equilibrium cells, thereby damping post‑shock dispersive ripples.

Entropic flux limiters.
The limiter approach rewrites the distribution as f = f_eq + k · (f − f_eq) and rescales the non‑equilibrium amplitude k by a factor φ(r), where r is a ratio of neighboring macroscopic variables. φ(r)≈1 retains a high‑order (second‑order) scheme in smooth regions, while φ→0 reduces the scheme to first order near steep gradients. This construction preserves the macroscopic moments (density and momentum) because f and f_eq share the same moments; only the direction of the non‑equilibrium component is altered. The paper implements several classic limiter functions (Van Leer, Minmod, Superbee) and demonstrates that the limiter acts as a pointwise “filter” that adds just enough dissipation to prevent spurious oscillations without sacrificing overall accuracy.

Multiple‑Relaxation‑Time (MRT) models.
MRT generalizes the single‑parameter BGK collision by diagonalizing the collision operator in moment space and assigning a distinct relaxation rate λ_i to each moment. Hydrodynamic moments (density, momentum) retain the usual relaxation linked to viscosity, while non‑hydrodynamic (ghost) modes can be relaxed much faster. By tuning λ_i for the non‑hydrodynamic modes, the method selectively damps high‑frequency, non‑physical components while leaving the physical dynamics essentially unchanged. The number of free parameters depends on the lattice (e.g., D2Q9, D3Q19) and the set of conserved quantities.

Numerical experiments.
Two benchmark problems are used: (a) a one‑dimensional shock tube (Sod problem) and (b) a two‑dimensional lid‑driven cavity at Reynolds numbers up to 10⁴. In the shock tube, the standard LBGK exhibits pronounced post‑shock oscillations. ELBM eliminates almost all of these oscillations because α grows large in the shock region, providing strong extra dissipation. MRT reduces the oscillations to a moderate level, while the limiter only damps them locally, preserving higher accuracy elsewhere. In the cavity flow, LBGK becomes unstable for Re > 10³. MRT and the limiter‑augmented schemes remain stable and capture the vortex structure correctly. ELBM, being the most dissipative, yields a smoother flow field but at the cost of attenuating the strength of the primary vortex.

Performance and cost.
ELBM incurs the highest computational overhead due to the per‑site root‑finding; MRT adds a modest matrix‑vector cost (≈1.2–1.5× LBGK); limiters are essentially cost‑free, requiring only a scalar function evaluation. The authors therefore suggest a hybrid strategy: use ELBM or MRT where global stability is critical (e.g., strong shocks), and employ limiters in regions where only local gradient control is needed.

Conclusions.
The study demonstrates that the key to stabilizing LBM lies in how, where, and how much additional dissipation is introduced. Global entropy‑preserving dissipation (ELBM) is most effective at suppressing non‑physical oscillations but can over‑smooth the solution. Local, gradient‑dependent dissipation (flux limiters) preserves overall accuracy while still controlling spurious modes. MRT offers a flexible, mode‑specific dissipation that can be tuned to balance stability and fidelity. The authors provide practical guidelines for selecting and combining these add‑ons based on problem characteristics, thereby extending the applicability of LBM to higher Mach and Reynolds number regimes.