A Hybrid Gas-Kinetic Scheme and Discrete Velocity Method for Continuum and Rarefied Flows

This paper proposes a hybrid gas-kinetic scheme and discrete velocity method that integrates the strengths of both approaches through a numerical collision time and adaptive strategies, enabling accurate and efficient simulations of flows across both continuum and rarefied regimes.

The Gist

Imagine you are trying to predict how air moves around a space vehicle. The challenge is that the air behaves very differently depending on how high you are.

• Down low (Continuum Flow): The air is thick and crowded, like a busy highway where cars are bumper-to-bumper. They bump into each other constantly, moving as a smooth, continuous fluid.
• Way up high (Rarefied Flow): The air is thin and sparse, like a few cars driving across a vast, empty desert. They rarely bump into each other and fly freely.

For decades, scientists have had two different "tools" to simulate these scenarios, but neither tool works perfectly for both:

1. The "Smooth Flow" Tool (GKS): This is like a super-efficient traffic cop. It's amazing at predicting the busy highway (continuum flow) because it assumes cars are always interacting. But if you try to use it for the empty desert, it fails because it assumes cars are bumping into each other when they aren't.
2. The "Free Flight" Tool (DVM): This is like a tracker for individual particles. It's perfect for the empty desert because it follows every single car. But if you try to use it for the busy highway, it becomes incredibly slow and messy. It tries to track every tiny collision, which takes forever, and it often gets "fuzzy" or inaccurate when the traffic is dense.

The New Solution: A "Smart Hybrid" Tool

The authors of this paper created a Hybrid Gas-Kinetic Scheme. Think of this as a chameleon tool that can instantly switch its personality depending on the environment.

Instead of forcing the computer to use one method for the whole journey, this new method acts like a smart traffic manager who knows when to use the "Traffic Cop" and when to use the "Particle Tracker."

How does it decide?
It uses a special "timer" called a Numerical Collision Time.

• In the thick air (Continuum): The timer tells the system, "We are in a crowd; use the efficient Traffic Cop method." It ignores the slow, particle-by-particle tracking to save time.
• In the thin air (Rarefied): The timer says, "We are in the desert; switch to the Particle Tracker." It stops assuming constant collisions and lets the particles fly free.
• In the middle (Shocks): Sometimes, even in thick air, there are sudden, violent changes (like a shockwave or a sonic boom). Here, the tool adds a little bit of the "Particle Tracker" logic back in. This acts like a safety cushion, adding just enough "friction" to stop the simulation from getting unstable and crashing, ensuring the shockwave is captured sharply.

The "Adaptive" Feature: Saving Energy

The paper also introduces a "smart switch" based on how fast the air is moving and how thin it is.

• If the air is thick and moving slowly, the tool only uses the fast Traffic Cop method.
• If the air is thin, it only uses the Particle Tracker.
• It only uses the complex "mix" of both when absolutely necessary.

This is like a hybrid car that runs on electricity in the city (efficient) and switches to gas only when you need to go fast or climb a steep hill. This strategy makes the computer run 10 times faster for smooth flows and 2 times faster for rarefied flows compared to previous methods, without losing accuracy.

The Proof: Three Test Drives

The authors tested this new tool on three specific scenarios to prove it works:

1. The Flat Plate (Smooth Highway): They simulated air flowing over a flat surface. The new tool matched the perfect theoretical answer almost exactly, but did it much faster than the old methods.
2. The Cavity (The Wind Tunnel): They simulated air swirling inside a box with a moving lid. They tested this in three different "air densities" (thick, medium, and thin). In all cases, the new tool matched the results of the most accurate (but very slow) reference methods, but finished the job in about half the time.
3. The Shock Wave (The Sonic Boom): They simulated a sudden, violent compression of air. This is the hardest part because the air changes instantly. The old methods either got "wobbly" (oscillated) or were too slow. The new hybrid tool, thanks to its "safety cushion" (the numerical collision time), captured the sharp shock perfectly without wobbling, while still being faster than the competition.

The Bottom Line

This paper presents a new way to simulate gas that is fast, accurate, and robust. It doesn't just work for one type of flow; it seamlessly handles everything from the thick air near the ground to the thin air in space, and even the violent shockwaves in between. By intelligently switching between two existing methods, it solves the "speed vs. accuracy" problem that has plagued scientists for years.

Technical Summary

1. Problem Statement

Simulating gas flows across the entire spectrum from continuum to rarefied regimes presents a significant challenge due to the conflicting requirements of different numerical methods:

• Continuum Regime: The Gas-Kinetic Scheme (GKS) is highly efficient and accurate for continuum flows (Navier-Stokes limit) but fails to reliably capture rarefaction effects.
• Rarefied Regime: The Discrete Velocity Method (DVM) is well-suited for rarefied flows (Boltzmann equation) but suffers from low computational efficiency, slow convergence, and excessive numerical dissipation when applied to continuum regimes due to the stiffness of the collision term and upwind reconstruction issues.
• Existing Unified Methods: While Unified Gas-Kinetic Schemes (UGKS) and Discrete Unified Gas-Kinetic Schemes (DUGKS) address multiscale flows, they remain computationally intensive because they require evaluating time-dependent distribution functions at cell interfaces for every time step.

The core problem is the lack of a method that is simultaneously computationally efficient, accurate, and robust across all Knudsen ($Kn$) and Mach ($Ma$) numbers without relying on complex macroscopic equation discretizations.

2. Methodology

The authors propose a Hybrid GKS–DVM method that integrates the strengths of both approaches through a novel balancing strategy.

Core Mechanism: Numerical Collision Time

The hybrid scheme constructs the time-evolution distribution function at cell interfaces by linearly combining the equilibrium component from GKS and the non-equilibrium component from DVM. This combination is weighted by a numerical collision time ($\tau_n$):

$$f^{hybrid} = (1 - e^{-\Delta t / \tau_n}) f^{GKS} + e^{-\Delta t / \tau_n} f^{DVM}$$

• $f^{GKS}$: Represents the equilibrium distribution function derived from the Chapman-Enskog expansion (suitable for continuum).
• $f^{DVM}$: Represents the non-equilibrium distribution reconstructed via an upwind scheme (suitable for rarefied flows).
• $\tau_n$ (Numerical Collision Time): Defined as $\tau_n = \frac{\mu}{p} + C \left| \frac{p_L - p_R}{p_L + p_R} \right| \Delta t$.
  • Unlike the physical collision time ($\tau$), $\tau_n$ includes a pressure-gradient-dependent term. This is crucial for shock capturing in continuum flows, as it ensures the DVM component (which provides numerical dissipation) remains active in shock regions even when the physical collision time is small.

Asymptotic Behavior

• Continuum Limit ($\Delta t \gg \tau_n$): The weighting factor $e^{-\Delta t / \tau_n} \to 0$. The scheme recovers the GKS equilibrium solution, satisfying the Navier-Stokes equations.
• Rarefied Limit ($\Delta t \ll \tau_n$): The weighting factor approaches 1. The scheme recovers the DVM solution, capturing free-molecular flow behavior.
• Shock Regions: In continuum flows with shocks, the pressure gradient term in $\tau_n$ increases the effective collision time, preventing the DVM contribution from vanishing. This introduces necessary numerical dissipation to stabilize shock capturing without smearing smooth flow regions.

Adaptive Strategies

To further reduce computational cost, the authors propose adaptive switching based on local $Kn$ and $Ma$:

1. Continuum & Subsonic ($Ma < 1, Kn \le 0.001$): Only the GKS flux is evaluated.
2. Continuum & Transonic/Supersonic: Both GKS and DVM fluxes are evaluated (using the hybrid formula) to ensure shock stability.
3. Rarefied/Transition ($Kn > 0.001$): Only the DVM flux is evaluated.

3. Key Contributions

1. Novel Hybrid Formulation: A direct coupling of GKS and DVM distribution functions without requiring explicit discretization of macroscopic governing equations (unlike some previous hybrid methods).
2. Numerical Collision Time ($\tau_n$): The introduction of $\tau_n$ solves the stability issue of shock capturing in the continuum limit while maintaining asymptotic preserving properties.
3. Adaptive Efficiency: The development of flow-regime-dependent switching strategies that dynamically select the most efficient flux evaluation method, significantly lowering computational overhead.
4. Robustness: The method avoids the excessive numerical dissipation of pure DVM in continuum flows and the instability of pure GKS in rarefied flows.

4. Results and Validation

The method was validated against three benchmark cases:

• Case 1: Flat-Plate Boundary Layer (Continuum, $Re=10^5$):

  • The adaptive hybrid method matched Blasius solutions with high accuracy.
  • Efficiency: The adaptive strategy reduced computational cost by approximately 10x compared to the non-adaptive hybrid approach.

• Case 2: Lid-Driven Cavity Flow (Rarefied to Continuum):

  • Tested at $Kn = 10.0$ (Free-molecular), $0.075$ (Slip), and $2.05 \times 10^{-4}$ (Continuum).
  • Rarefied ($Kn=10$): The hybrid method matched DSMC and UGKS results but required only ~50% of the wall-clock time of UGKS.
  • Slip ($Kn=0.075$): Achieved similar accuracy to UGKS with ~2.3x speedup.
  • Continuum ($Kn \approx 10^{-4}$): The hybrid method outperformed pure DVM (which showed vortex deviations due to dissipation) and matched reference solutions (Ghia et al.) efficiently.

• Case 3: Shock Structures (1D, $Ma=2.0$):

  • Tested at $Kn = 1.0, 0.1, 0.001$.
  • Shock Capturing: Pure GKS and hybrid methods using physical $\tau$ failed to resolve sharp shocks at $Kn=0.001$ due to insufficient dissipation (oscillations). The hybrid method using numerical $\tau_n$ successfully captured the shock structure without oscillations.
  • Efficiency: In the continuum shock case, the hybrid method reduced cost by ~10% compared to UGKS. In rarefied shock cases, the speedup was more significant (~1.6x).

5. Significance

This work presents a significant advancement in computational fluid dynamics for multiscale flows:

• Bridging the Gap: It successfully unifies the high efficiency of GKS for continuum flows with the physical fidelity of DVM for rarefied flows in a single framework.
• Computational Savings: By introducing adaptive strategies, the method achieves order-of-magnitude speedups in continuum regimes and 2x speedups in rarefied regimes compared to state-of-the-art unified schemes (UGKS).
• Shock Robustness: The specific design of the numerical collision time ($\tau_n$) resolves a long-standing difficulty in kinetic schemes: capturing shocks in continuum flows without introducing excessive numerical dissipation in smooth regions.
• Practical Application: The method is particularly valuable for aerospace applications involving vehicles traversing from ground level to near-space, where flow regimes transition rapidly from continuum to rarefied.