Mixture-aware closure of the N-phase Navier--Stokes--Cahn--Hilliard mixture model

This paper establishes a unique, thermodynamically admissible closure for N-phase Navier--Stokes--Cahn--Hilliard models by enforcing PDE-level reduction consistency under the merging of identical phases, which uniquely determines the free-energy structure and mobility matrix to include Maxwell--Stefan-type mobilities as a special case.

The Gist

Imagine you are a chef trying to simulate how different ingredients mix in a giant, invisible pot. Some ingredients are oil, some are water, and some are air bubbles. In the world of computer simulations, these ingredients are called "phases."

For a long time, scientists had a recipe (a mathematical model) to simulate how two ingredients mix, like oil and water. But when they tried to add a third, fourth, or even a hundredth ingredient, the recipe got messy. The math would break if you tried to pretend two ingredients were actually the same thing, or if you tried to remove an ingredient that wasn't there.

This paper introduces a new, smarter recipe for simulating mixtures with any number of ingredients (called an $N$-phase model). The authors, Marco ten Eikelder and Aaron Brunk, created a set of rules that ensures the simulation behaves logically no matter how you label or combine your ingredients.

Here is the breakdown of their discovery using simple analogies:

1. The Problem: The "Labeling" Confusion

Imagine you have a bucket of red paint and a bucket of blue paint.

• Scenario A: You have a bucket of "Red" and a bucket of "Blue."
• Scenario B: You have a bucket of "Red," a bucket of "Blue," and a third bucket also labeled "Red."

In the old math models, if you tried to merge the two "Red" buckets in Scenario B to act like one big "Red" bucket, the computer simulation would get confused. It might calculate the physics differently just because you used two labels instead of one. It's like if a recipe for a cake changed its taste just because you wrote "sugar" twice on the ingredient list instead of once.

The authors wanted a model that understands that two labels for the same thing are physically the same thing. If you merge two identical phases, the simulation should act exactly as if you had only one phase to begin with.

2. The Solution: The "Mixture-Aware" Rules

The authors developed a set of "axioms" (unbreakable rules) for their mathematical model. Think of these as the laws of physics for their simulation pot.

• The "Merge" Rule: If you have two phases that are physically identical (same density, same stickiness, same chemical nature), merging them into one label must not change the outcome of the simulation. The math must automatically "collapse" into a simpler version that works perfectly for the remaining ingredients.
• The "Ghost" Rule: If an ingredient is missing (zero amount), it must stay missing. The simulation shouldn't suddenly create a ghost bubble of that ingredient out of thin air.

3. The New Recipe: What Does the Math Look Like?

To make these rules work, the authors figured out exactly what the "ingredients" of the math must look like. They found that there is only one specific way to write the equations that satisfies all these rules.

• The Energy Part (The "Taste"):
  The model uses a specific type of energy formula. It has two main parts:

  1. The "Mixing" Part: This is like the natural tendency of things to spread out (entropy). It's mathematically similar to how people mix at a party; it prefers a balanced distribution.
  2. The "Interaction" Part: This accounts for how much ingredients like or dislike each other. If they dislike each other (like oil and water), they separate. If they are identical, they mix perfectly.
  3. The "Surface" Part: This handles the boundary between ingredients. It acts like a rubber band trying to keep the interface between oil and water smooth.

• The Movement Part (The "Traffic"):
  The model also dictates how ingredients move (diffuse) past each other. The authors found that the "traffic rules" for this movement must follow a specific pattern called Maxwell-Stefan.

  • Analogy: Imagine a crowded dance floor. If you want to move, you have to swap places with someone else. The math says the ease of swapping depends on how many people are on the floor. If a specific dance partner (phase) isn't there, you can't swap with them. This ensures that if a phase is absent, it stays absent.

4. Testing the Recipe

The authors didn't just write the math; they ran computer simulations to prove it works.

• The "Ghost" Test: They simulated a bubble rising in a liquid but told the computer there was a third ingredient that wasn't actually there. The simulation correctly ignored the ghost ingredient, and the bubble behaved exactly as it would in a two-ingredient world.
• The "Merge" Test: They simulated a scenario where two ingredients were actually the same (e.g., two types of water). They told the computer to treat them as one big pool. The simulation smoothly merged them without glitching, behaving exactly like a standard two-ingredient simulation.
• Complex Scenarios: They successfully simulated a bubble rising through two different layers of liquid (three ingredients) and even a complex scene with a bubble, a droplet, and two liquid layers (four ingredients).

Why This Matters (According to the Paper)

The paper claims this is the first practical way to simulate complex mixtures with many ingredients while ensuring the math remains consistent. Before this, scientists had to choose between models that were easy to calculate but broke the laws of physics when ingredients were merged, or models that were physically correct but impossible to use for complex, multi-ingredient scenarios.

This new "Mixture-Aware" closure provides a single, unified framework that works for 2, 3, 4, or even $N$ phases, ensuring that the computer simulation respects the physical reality that identical things should behave identically, regardless of how you name them.

Technical Summary

1. Problem Statement

The paper addresses a fundamental gap in the modeling of multiphase flows using the Navier–Stokes–Cahn–Hilliard (NSCH) framework. While $N$-phase phase-field models are widely used to describe complex mixtures with diffuse interfaces, existing constitutive closures (definitions of free energy and mobility) often lack mixture awareness.

Specifically, current models typically enforce "reduction consistency" only when a phase is absent (i.e., the volume fraction $\phi_\alpha = 0$). However, they fail to satisfy the physical requirement that merging physically identical phases (e.g., two labels representing the same fluid) should not alter the governing equations.

• The Challenge: If two phases are identical, merging them into a single phase should result in a system mathematically equivalent to an $(N-1)$-phase model. Existing closures (such as those by Boyer-Lapuerta or Steinbach) often violate this, leading to inconsistencies where the physics changes simply due to arbitrary labeling or splitting of a single phase into sub-labels.
• The Goal: To derive a thermodynamically admissible, unique constitutive closure for the $N$-phase NSCH model that is invariant under the merging of identical phases at the Partial Differential Equation (PDE) level.

2. Methodology

The authors employ an axiomatic approach rooted in continuum mixture theory to derive the closure.

A. Governing Equations

The study utilizes the $N$-phase NSCH system with non-matching densities, formulated in terms of:

• Mass-averaged velocity ($\mathbf{v}$).
• Volume fractions ($\phi_\alpha$) satisfying the saturation constraint $\sum \phi_\alpha = 1$.
• Chemical potentials ($\mu_\alpha$) derived from a Helmholtz free energy density $\Psi(\phi, \nabla \phi)$.
• Diffusive fluxes ($J_\alpha$) driven by gradients of effective chemical potentials, governed by a mobility tensor $\mathbf{M}$.

B. Structural Axioms

The authors impose a set of structural axioms to constrain the constitutive choices:

1. Local First-Gradient Free Energy: $\Psi$ depends on $\phi$ and $\nabla \phi$.
2. Constant Capillarity: The second derivatives of $\Psi$ with respect to gradients are independent of local composition.
3. Mobility Properties: Symmetry, positive semi-definiteness, constraint compatibility ($\mathbf{M}\mathbf{1}=0$), and degeneracy (flux vanishes if a phase is absent).
4. Reduction Consistency (The Core Axiom): When two phases ($p$ and $q$) are physically identical, the $N$-phase system must reduce exactly to an $(N-1)$-phase system upon merging $\phi_p$ and $\phi_q$. This applies to both the capillary force (momentum balance) and the diffusive flux (mass balance).
5. Label Invariance: The model must be invariant under permutation of phase labels.

C. Derivation Process

1. Free Energy Derivation: By enforcing reduction consistency on the capillary forces, the authors prove that the bulk free energy must consist of:
   • An ideal mixing term (entropic): $\sum W_\alpha \phi_\alpha \log \phi_\alpha$.
   • A symmetric pairwise interaction term (enthalpic): $-\sum \chi_{\alpha\beta} \phi_\alpha \phi_\beta$.
   • A constant-coefficient quadratic gradient penalty: $\sum \kappa_{\alpha\beta} \nabla \phi_\alpha \cdot \nabla \phi_\beta$.
   • Result: Linear gradient terms are ruled out by objectivity and isotropy.
2. Mobility Derivation: By enforcing reduction consistency on diffusive fluxes, the mobility tensor is constrained to a pairwise-exchange form with bilinear degeneracy:
   • $M_{\alpha\beta} \propto -\phi_\alpha \phi_\beta$ for $\alpha \neq \beta$.
   • This structure naturally includes Maxwell–Stefan diffusion as a special case.

3. Key Contributions

1. Unique Thermodynamic Closure: The paper demonstrates that under the axioms of mixture-awareness, the admissible free energy and mobility structures are uniquely fixed. There is no ambiguity in the functional form; it must be the ideal-mixing plus quadratic interaction plus quadratic gradient form, and the mobility must be pairwise-exchange.
2. PDE-Level Reduction: Unlike previous works that focused on energy value reduction, this work enforces reduction at the governing equation level. This ensures that the dynamics (velocities, pressures, fluxes) remain consistent when identical phases are merged.
3. Rejection of Existing Models: The authors mathematically prove that popular existing closures (e.g., Boyer–Lapuerta and Steinbach free energies) are not mixture-aware. They fail to reduce correctly when identical phases are merged, leading to unphysical behavior in simulations involving label splitting.
4. Practical Parameterization: The authors provide a concrete parameterization for numerical implementation, including:
   • A polynomial approximation of the logarithmic singularity in the free energy to ensure smoothness at $\phi=0$.
   • Calibration of interaction parameters to match standard Ginzburg–Landau barrier heights.
   • A specific mobility form compatible with Maxwell–Stefan diffusion.

4. Results

The authors validate the theoretical findings through numerical experiments using a symmetric structure-preserving finite element method.

• Verification of Mixture-Awareness:
  • Absent Phase Test: Simulations confirm that if a phase is initialized as zero, it remains zero (Corollary 3.7).
  • Merging Identical Phases: Simulations of a rising bubble in a ternary system where two phases are identical show that the system behaves exactly like a binary system. The merged phases evolve identically, and the dynamics match the reduced $(N-1)$-phase model perfectly.
  • Spatial Separation: Even when identical phases are spatially separated (e.g., one bubble at the bottom, another at the top), the model correctly treats them as a single phase, maintaining consistency.
• Complex Multiphase Simulations:
  • Ternary Case ($N=3$): A gas bubble rising through two stratified liquid layers. The model correctly handles three distinct surface tensions, large density/viscosity ratios, and topological changes (interface penetration vs. trapping).
  • Quaternary Case ($N=4$): A simulation involving a falling droplet and a rising bubble in a stratified liquid. The framework successfully handles four distinct phases and their complex interactions without numerical instability.

5. Significance

• Theoretical Rigor: This work resolves a long-standing ambiguity in $N$-phase phase-field modeling by establishing a "gold standard" closure based on physical invariance principles rather than ad-hoc construction.
• Computational Robustness: The derived closure enables the first practical numerical realization of the recently proposed $N$-phase NSCH model. It ensures that simulations are robust against arbitrary choices in phase labeling, which is crucial for complex applications like additive manufacturing or emulsion dynamics where phases may be split for numerical convenience.
• Broader Applicability: While derived for incompressible NSCH, the reduction-consistency arguments apply broadly to multiphase mixture models, including gradient-free Maxwell–Stefan diffusion systems.
• Future Directions: The paper opens the door for rigorous mathematical analysis (existence, well-posedness) of the system and extends the framework to compressible mixtures and models with multiple momentum equations.

In summary, the paper provides a mathematically rigorous and numerically validated framework for simulating $N$-phase flows that respects the physical reality that "identical phases are identical," regardless of how they are labeled in the computational model.