Towards nonlinear thermohydrodynamic simulations via the Onsager-Regularized Lattice Boltzmann Method

This paper presents a theoretical analysis and numerical validation of the Onsager-Regularized Lattice Boltzmann Method, demonstrating that it achieves higher-order accuracy and mitigates lattice isotropy errors in nonlinear thermohydrodynamic simulations on standard lattices without requiring external correction terms.

The Gist

Imagine you are trying to simulate how a fluid (like water or air) moves and carries heat using a computer. In the world of physics simulations, there is a popular tool called the Lattice Boltzmann Method (LBM).

Think of LBM as a giant, digital game of "connect-the-dots" played on a grid. Instead of tracking every single molecule, the computer looks at tiny packets of fluid sitting on the intersections of a grid (like squares on a chessboard). At each step, these packets move to their neighbors and bump into each other, transferring energy and momentum. Over time, these tiny bumps and moves recreate the complex flow of a river or the turbulence of wind.

However, there's a catch. The standard grid used in these games (called the D2Q9 lattice) is like a square chessboard. It has a problem: it treats "straight" directions (up, down, left, right) differently from "diagonal" directions.

The Problem: The "Square Grid" Bias

Imagine you are walking through a city with a perfect grid of streets. If you walk North or East, it's easy. But if you try to walk Northeast (diagonally), you have to zigzag. In a computer simulation, this "zigzagging" creates fake errors.

When the fluid moves diagonally, the standard grid gets confused. It creates spurious errors—ghostly ripples and wrong temperatures that don't exist in real life. To fix this, scientists usually have to add "patches" or "correction terms" to the code. But these patches are like duct tape: they are messy, hard to apply to different situations, and they slow down the computer.

The Solution: The "Onsager-Regularized" (OReg) Method

This paper introduces a new, smarter way to play the game called the Onsager-Regularized (OReg) method.

Instead of patching the grid with duct tape, the authors redesigned the rules of the game itself. They used a principle from thermodynamics (called Onsager's Principle) which basically says: "Nature is efficient; it minimizes waste and follows a specific path when things go wrong."

Here is how the OReg method works, using a simple analogy:

1. The "Predictor-Corrector" Chef

Imagine a chef (the computer) trying to bake a cake (simulate fluid flow).

• The Old Way (Standard LBM): The chef guesses the ingredients, mixes them, and then realizes, "Oh no, I added too much sugar because the scale was slightly off." They have to manually subtract the extra sugar (the correction term) to fix it.
• The OReg Way: The chef has a special, self-correcting recipe. Before they even mix the batter, they look at the ingredients and automatically adjust the amounts based on how the cake should behave. If the grid is "square" and biased, the recipe automatically adds a little extra "diagonal" ingredient to balance it out.

2. The "Smart Viscosity"

In fluid physics, viscosity is how "thick" or "sticky" a fluid is (like honey vs. water).
The paper shows that the OReg method acts like a smart, shape-shifting viscosity.

• If the fluid is moving straight, the viscosity stays normal.
• If the fluid tries to move diagonally across the square grid, the OReg method automatically changes the viscosity just enough to cancel out the grid's bias.
• It's like driving a car with active suspension: if the road is bumpy on the left, the car automatically stiffens the left shock absorber to keep the ride smooth. The OReg method stiffens the "fluid rules" to keep the simulation smooth, without needing any external help.

Why This Matters

The authors tested this new method on two difficult scenarios:

1. Rotating Waves: Simulating a wave spinning in a square grid. The old methods made the wave look jagged and wrong. The OReg method made it look perfectly smooth, just like in real life.
2. Shockwaves: Simulating a sudden explosion or pressure change (like a supersonic jet). The old methods created "noise" (fake ripples) that made the picture messy. The OReg method cleaned up the noise completely.

The Big Takeaway

This paper is a breakthrough because it proves you don't need to use complex, heavy, and slow "patches" to fix the flaws of the standard grid.

• Old Way: "The grid is flawed, so let's add a complicated correction term to fix it."
• New Way (OReg): "The grid is flawed, so let's change the way the fluid particles interact so they naturally ignore the flaw."

This allows scientists to run faster, more accurate, and more stable simulations of complex fluids (like blood flow, weather patterns, or engine combustion) on standard computers. It turns a "square peg in a round hole" problem into a seamless, natural flow.

In short: The authors found a way to make a square grid behave like a perfect circle, simply by teaching the fluid particles how to dance better, rather than forcing the grid to change.

Technical Summary

1. Problem Statement

The Lattice Boltzmann Method (LB) is a powerful tool for simulating fluid dynamics, but standard implementations on common lattices (e.g., D2Q9, D3Q27) suffer from insufficient lattice isotropy. This deficiency introduces spurious numerical errors (anisotropy artifacts) in the macroscopic equations, particularly when simulating nonlinear thermohydrodynamic phenomena or flows at arbitrary temperatures.

• Current Limitations: To recover accurate Navier-Stokes-Fourier (NSF) dynamics on standard lattices, existing state-of-the-art methods often require non-local correction terms or complex recursive regularization schemes. These approaches compromise the computational efficiency and parallel scalability of LB, making them difficult to generalize across different collision kernels and equilibrium distributions.
• The Gap: There is a need for a fully local, correction-free, and assumption-free framework that can handle nonlinear thermohydrodynamics on standard lattices without sacrificing the inherent advantages of the LB method.

2. Methodology

The authors propose and analyze the Onsager-Regularized (OReg) Lattice Boltzmann method. The core methodology involves:

• Theoretical Framework: The study employs a generalized, stencil-independent theoretical analysis using the Chapman-Enskog (CE) multiscale expansion. Unlike traditional analyses that assume compatibility conditions hold for all orders of approximation, this work explicitly evaluates the non-equilibrium distribution function ($f^{(neq)}$) based on linear irreversible thermodynamics and Onsager's Symmetry Principle.
• The OReg Scheme:
  • The non-equilibrium contribution is defined via viscous and thermal irreversible processes (Eq. 4).
  • It is projected onto lattice stencils to create a fully local formulation (Eq. 6) that depends on the non-equilibrium populations ($f^{(neq)}$) and local gradients.
  • The update rule (Eq. 7) acts as a one-step predictor-corrector method, where the prediction is corrected using the OReg formulation.
• Equilibrium Distributions: The authors derive the hydrodynamic limits for two specific equilibrium distributions on the standard D2Q9 lattice:
  1. Thermal Guided Equilibrium: A distribution designed to satisfy higher-order moment constraints (specifically $O(u^4)$ accuracy).
  2. Second-Order Polynomial Equilibrium: The standard polynomial form commonly used at the reference temperature ($O(u^2)$ accuracy).

3. Key Contributions

The paper makes several significant theoretical and practical contributions:

• Assumption-Free Analysis: The authors derive the macroscopic limit without assuming the validity of compatibility conditions for higher-order terms, providing a rigorous proof of the OReg scheme's accuracy.
• Automatic Viscosity Compensation:
  • When coupled with the Guided Equilibrium on D2Q9, the OReg scheme automatically adjusts the lattice viscosity to compensate for insufficient lattice isotropy.
  • This results in $O(u^4)$ accuracy at the reference temperature ($\theta = 1/3$) and $O(u^2)$ accuracy at arbitrary temperatures.
  • This is a significant improvement over the bare BGK model, which typically yields only $O(u^3)$ and $O(u)$ accuracy, respectively.
• Polynomial Equilibrium Improvement: Even with the standard second-order polynomial equilibrium, the OReg scheme improves the kinetic model accuracy to $O(u^3)$, an order of magnitude better than the bare BGK model.
• Correction-Free Implementation: The method achieves these accuracies without external non-local correction terms, preserving the local nature and parallel efficiency of LB.

4. Numerical Results

The theoretical claims were validated through numerical benchmarks on quasi-one-dimensional problems:

• Rotated Decaying Shear Wave:
  • Axis-aligned waves: All schemes (BGK, Projected Regularized, OReg) performed correctly.
  • $\pi/4$-rotated waves: Standard BGK and Projected Regularized (PR) schemes failed to recover the correct viscous dissipation rate due to lattice anisotropy. The OReg scheme successfully recovered the exact dissipation rate without corrections.
• Isothermal Shocktube:
  • Simulations were conducted at elevated temperatures ($\theta \neq 1/3$) and low viscosities.
  • Oscillations: The standard BGK scheme produced large spurious oscillations. While the Essentially Entropic (EE) and PR schemes reduced these, they did not eliminate them. The OReg scheme completely eliminated spurious oscillations.
  • Accuracy: The OReg scheme captured density profiles with ~98.8% accuracy (at $\theta=0.35$) and ~98.2% accuracy (at $\theta=0.4$), demonstrating robustness across different grid resolutions.

5. Significance

This work establishes a generic theoretical foundation for performing fully local, correction-free, nonlinear thermohydrodynamic simulations on standard lattices.

• Scalability: By removing non-local correction terms, the OReg scheme facilitates highly scalable simulations on complex geometries and high-performance computing architectures (e.g., GPUs).
• Physical Fidelity: It enables the simulation of physically challenging flows (high Reynolds and Mach numbers, thermal effects) that were previously prone to anisotropy-induced errors.
• Future Applications: The framework is easily integrable into existing high-performance codes (e.g., waLBerla) and offers a promising alternative for defining boundary conditions (e.g., Grad boundary conditions) and coupling with thermal boundary models.

In summary, the OReg scheme represents a paradigm shift from modeling accuracy via complex corrections to achieving stability and accuracy through a thermodynamically consistent, local regularization of the non-equilibrium distribution.