Simulating fluid interfaces where the densities of the two phases differ by orders of magnitude—such as water and air (a 1,000:1 ratio) or liquid metal and vapor—is notoriously difficult. As noted in a comprehensive survey by Wang et al., maintaining numerical stability and convergence at these high-density ratio boundaries remains a major open challenge in computational fluid dynamics.
When standard fluid solvers attempt to model these interfaces, they frequently suffer from severe numerical instabilities. As Fei et al. point out, these instabilities manifest as unphysical "spurious velocities" at the interface. These artificial velocity fluctuations can deform the interface unrealistically or cause the entire simulation to crash, forcing researchers to rely on artificially low, non-physical density ratios.
To understand why these simulations blow up, and how a recent wave of research fixes them, we have to look at how mass and momentum are coupled at the discrete level.
The Root Cause: Mismatched Fluxes and Broken Invariance
In a typical diffuse-interface or phase-field model, the distribution of different fluid phases is tracked using an order parameter or volume fraction, which evolves according to an advection-diffusion equation. To connect this interface-tracking step to the fluid dynamics, standard methods calculate the local mixture density $\rho$ algebraically based on the local volume fractions (e.g., $\rho = \alpha_1 \rho_1 + \alpha_2 \rho_2$) and plug this directly into the standard Navier-Stokes momentum equations.
However, Huang et al. demonstrated that this naive algebraic coupling introduces a fundamental mathematical inconsistency. When the density changes sharply across an interface, updating the density algebraically can directly contradict the mass conservation equation.
Standard Approach (Inconsistent):
[Phase-Field Transport] ──> Calculate Density Algebraically ──> [Standard Navier-Stokes]
│
Result: Broken Galilean Invariance
& Spurious Velocities
Consistent Approach:
[Phase-Field Transport] ──> Derive Consistent Mass Flux ──────> [Modified Momentum Equation]
│
Result: Strict Energy Conservation
& High-Density Stability
This mismatch breaks Galilean invariance—the physical principle that the laws of motion should remain identical regardless of the observer's constant velocity. If a simulation lacks Galilean invariance, simply moving the entire computational domain at a constant velocity will introduce thermodynamic errors and artificial forces.
Intuitively, if a grid cell contains a tiny fraction of a heavy fluid next to a light fluid, a small numerical error in tracking the interface can cause a massive, unphysical jump in the calculated local density. If the momentum equation updates the velocity field using this mismatched density, it creates an artificial momentum source. The resulting unphysical velocity fluctuations quickly cascade into simulation failure.
The Solution: Consistent Flux Coupling
To achieve numerical stability at extreme density ratios, a series of papers by Huang et al. developed a framework based on consistent mass and momentum transport.
The core insight is that mass and momentum must be transported across the computational mesh using the exact same numerical fluxes. Instead of calculating the density algebraically and inserting it into a standard momentum equation, the authors redefine the momentum flux as the tensor product of the velocity and a consistent mass flux derived directly from the phase-field equation.
As detailed in Huang et al. and Huang et al., ensuring that the mass flux in the momentum equation is identical to the flux in the mass conservation equation guarantees both Galilean invariance and strict kinetic energy conservation. Because the momentum equation is always perfectly synchronized with how mass is actually moving across cell boundaries, unphysical velocity fluctuations are eliminated. This mathematical alignment allows frameworks to handle extreme density ratios—up to $10^9:1$—without crashing.
Extending Consistency to Multiple Phases and Components
Real-world scenarios often involve more than just two simple fluids; they feature multiple immiscible phases (like oil, water, and gas) and dissolved components (like surfactants or salts).
When scaling up to $N$ phases, ensuring that the sum of all volume fractions equals exactly one at every point in space becomes an inverse problem. Huang and Ardekani solved this by introducing a coupled linear system that distributes volume changes across all phases simultaneously, ensuring local and global mass conservation even at density ratios of one million to one.
Furthermore, if components are dissolved within these phases, their transport can cause localized volume changes. Huang and Ardekani and Huang et al. showed that to maintain Galilean invariance in these complex systems, the mass flux within the momentum equation must explicitly incorporate the mass contributions and diffusion fluxes of both the background fluid phases and the dissolved chemical components. This prevents numerical desynchronization and concentration oscillations near interfaces.
This consistent formulation has proven highly versatile, enabling stable simulations of moving contact lines on curved solid surfaces (Huang et al.) and complex phase changes, such as gas bubbles becoming trapped or released by an advancing solidification front (Huang et al.).
Alternative Perspectives on Decoupling and Stabilization
While tightly coupling mass and momentum fluxes is a prominent solution, other researchers have approached high-density-ratio stability through different forms of decoupling or mathematical stabilization:
- Velocity-Based Momentum Decoupling: In the context of the Lattice Boltzmann Method (LBM), Wu et al. developed a kinetic solver that completely decouples fluid density from the momentum equations through a velocity-based formulation. By separating the sharp density gradients from the velocity updates, they achieved stable multi-fluid simulations at density ratios up to 2,000:1 while reducing memory usage by 80%.
- Implicit Phase-Momentum Mapping: For computer graphics applications, Xu et al. introduced an implicitly stable mixture model. Instead of relying on explicit approximations of interphase forces, they used an implicit phase-momentum mapping to relate individual phase velocities to the overall mixture momentum sources, combined with a mixture viscosity model to smooth out erratic drift velocities.
- Pressure Diffusion and Bulk Viscosity: Working with weakly compressible models, Lu demonstrated that explicitly introducing pressure diffusion and bulk viscosity through a predictor-corrector step can stabilize interfaces at 1,000:1 density ratios, even in completely inviscid fluids where standard models fail.
- High-Frequency Coupling: From a pure mathematical analysis standpoint, Deng and Yin proved that density fluctuations are often the primary bottleneck for system stability because they decay slower than velocity. By mathematically coupling density and velocity estimates specifically within the high-frequency range, they showed it is possible to guarantee that density variations do not impede the system's convergence to a stable state.
Ultimately, whether achieved through identical numerical fluxes, velocity-based decoupling, or implicit mapping, the consensus across recent literature is clear: stabilizing high-density-ratio simulations requires a mathematically rigorous treatment of how mass variations affect the transport of momentum.
Key Foundational Work
Consistent, essentially conservative and balanced-force Phase-Field method to model incompressible two-phase flows (2020) — contributed 5 concept elements to this direction.
Modeling and simulation of N-phase-M-component incompressible flows (2020) — contributed 3 concept elements to this direction.
A consistent and conservative model and its scheme for N-phase-M-component incompressible flows (2021) — contributed 3 concept elements to this direction.