Enhanced numerical approaches for modeling insoluble surfactants in two-phase flows with the diffuse-interface method

This paper proposes two simple and practical numerical enhancements for modeling insoluble surfactants in two-phase flows using the diffuse-interface method, which improve accuracy without increasing computational cost and include a new challenging benchmark case for future method evaluation.

The Gist

Imagine you are watching a drop of oil floating in a glass of water. Now, imagine sprinkling some soap (a surfactant) onto the surface of that oil drop.

In the real world, this soap doesn't just sit there; it moves around, clumps together, and changes how the oil drop behaves. It makes the drop stretch, shrink, or wiggle in strange ways because the soap changes the "stickiness" (surface tension) of the oil's skin.

Scientists want to use computers to predict exactly how this happens. But simulating this on a computer is like trying to paint a moving, shape-shifting picture with a brush that is slightly too thick. The "brush" is the computer's grid, and the "paint" is the soap.

This paper is about fixing the brush so the computer can paint a much more accurate picture of how soap moves on a liquid surface.

Here is the breakdown of their solution, using simple analogies:

The Problem: The "Thick Brush" Issue

In computer simulations, the boundary between oil and water isn't a razor-thin line; it's a blurry zone (like a fuzzy edge). The soap lives in this fuzzy zone.

The researchers found that the old ways of calculating where the soap goes had two main problems:

1. The "Rough Edge" Problem: The old math tried to calculate how the soap moved by looking at the steepness of the fuzzy edge. But because the edge is blurry and jagged on a computer grid, calculating the "steepness" creates a lot of noise and errors. It's like trying to measure the slope of a mountain using a ruler that is made of rubber—it just doesn't work well.
2. The "One-Size-Fits-All" Problem: The old method forced the "soap zone" to be exactly the same width as the "oil-water fuzzy zone." But sometimes, the soap needs a wider or narrower zone to be calculated accurately. Forcing them to be the same size was like trying to wear shoes that are too big just because your socks are a certain size. It made the simulation either unstable or inaccurate.

The Solution: Two Simple Tweaks

The authors proposed two "hacks" to make the simulation much better without making the computer work harder.

1. Change the "Map" (The F-Type Model)

The Analogy: Imagine you are tracking a crowd of people (the soap) moving through a foggy hallway.

• The Old Way (fd-type): You tried to count how many people were in the fog itself. Since the fog is thick and shifting, your count was messy and full of errors.
• The New Way (f-type): Instead of counting people in the fog, you count the people on the clear path underneath the fog. The path is smooth and flat. Even though the fog is still there, your count is now much more accurate because you aren't fighting the fuzziness of the fog.

In plain English: They changed the math to calculate the soap's movement based on a smooth, flat variable rather than a jagged, steep one. This reduced the "noise" and errors significantly.

2. Decouple the "Ruler" from the "Fog" (The Delta Function)

The Analogy: Imagine you are painting a picture of a thin wire (the interface).

• The Old Way: The width of your paintbrush (the "delta function") was glued to the thickness of the wire. If you wanted a finer wire, you had to use a tiny brush, which made the paint splatter and miss the wire. If you wanted a thick brush for better coverage, you had to make the wire thicker, which ruined the shape of the wire.
• The New Way: They unglued the brush from the wire. Now, they can use a thick, soft brush to paint the soap (which gives a smooth, accurate result) while keeping the wire itself thin and sharp (which keeps the shape of the oil drop accurate).

In plain English: They allowed the mathematical "zone" where the soap is calculated to be wider than the actual interface. This lets the computer handle the soap more smoothly without blurring the shape of the liquid drop.

Why Does This Matter?

The researchers tested these ideas on some very difficult scenarios:

• Spinning Vortices: They spun the liquid around until the drop stretched into long, thin tails (like taffy).
• Shear Flow: They pushed the liquid so the drop squashed and stretched.

The Results:

• The old methods often crashed or gave wrong answers when the drop got too stretched.
• The new methods stayed stable and accurate, even when the drop was being twisted into a pretzel shape.
• They even created a "Super Hard Test" (a benchmark) that is so difficult that even their new method struggles with it. This is great news because it gives other scientists a new, tougher challenge to beat, pushing the field forward.

The Bottom Line

This paper is like a mechanic telling you: "You don't need a brand new engine to make your car faster. You just need to change the oil filter (the math formulation) and adjust the tire pressure (the width settings)."

By making these two small, smart adjustments, they made computer simulations of soap and liquids much more reliable, cheaper to run, and capable of handling the most chaotic, messy fluid flows imaginable.

Technical Summary

1. Problem Statement

The accurate numerical simulation of two-phase flows containing insoluble surfactants is critical for applications ranging from detergents to microfluidics. Surfactants adsorb at the interface, reducing surface tension and inducing Marangoni effects, which significantly alter flow dynamics.

While the diffuse-interface method (specifically the phase-field method) is widely used for its ability to handle topological changes and conserve fluid mass, modeling surfactant transport on these interfaces remains challenging. Existing models often suffer from:

• Discretization errors: Arising from the numerical evaluation of spatial derivatives of variables with sharp gradients (specifically the volumetric surfactant concentration).
• Coupling constraints: The width of the delta function (used to localize surfactant transport to the interface) is traditionally tied to the interface width. This creates a trade-off: a wider delta function improves surfactant transport accuracy but degrades interface capturing and reduces the allowable time step; a narrower interface improves capturing but may fail to resolve the surfactant profile accurately.
• Lack of rigorous benchmarks: Most existing studies rely on simple test cases that do not stress the limits of surfactant transport models under severe interface deformation.

2. Methodology

The authors utilize the Accurate Conservative Diffuse-Interface (ACDI) model for fluid mass conservation and propose two specific enhancements to improve surfactant transport accuracy without increasing computational cost or complexity.

A. Approach 1: Reformulation of the Transport Equation

The study compares two generalized surfactant transport models:

1. $f_d$-type model: Solves for the volumetric surfactant concentration ($f_d = f \delta_\Gamma$). The right-hand side involves the gradient of $f_d$ ($\nabla f_d$). Since $f_d$ has a sharp gradient normal to the interface, this term introduces significant discretization errors.
2. $f$-type model: Solves for the interfacial surfactant concentration ($f$). The right-hand side involves the gradient of $f$ ($\nabla f$). Since $f$ is approximately flat in the normal direction, its gradient is smoother.

Proposal: The authors propose adopting the $f$-type model. Although mathematically equivalent to the $f_d$-type model at equilibrium, the $f$-type model significantly reduces discretization errors in practical simulations by avoiding the spatial derivatives of sharp-gradient variables.

B. Approach 2: Decoupling Delta Function Width from Interface Width

Conventionally, the delta function width ($\delta_\Gamma$) is derived directly from the phase-field variable $\phi$, forcing it to match the interface width ($W$).
Proposal: The authors introduce a hybrid approach combining the ACDI method with the Level-Set method.

• The phase-field variable $\phi$ is used for mass conservation and interface capturing.
• A level-set function $\psi$ is advected and reinitialized to maintain a signed distance property.
• A new parameter $\hat{\epsilon}$ is introduced to define a separate delta function width ($\hat{W} = 4\hat{\epsilon}$) based on $\psi$.
• This allows $\hat{W}$ to be set independently (typically larger, e.g., $3\Delta x$ to $6\Delta x$) to optimize surfactant transport resolution, while keeping the interface width $W$ small (e.g., $2.04\Delta x$) to maintain interface capturing accuracy and time-step stability.

3. Key Contributions

1. Identification of Numerical Discrepancy: The paper demonstrates that while $f_d$-type and $f$-type models are theoretically equivalent, the $f$-type model is superior in practice due to reduced discretization error. This distinction has been overlooked in previous literature.
2. Independent Width Control: The study provides a practical method to decouple the delta function width from the interface width using a level-set auxiliary variable. This resolves the conflict between interface resolution and surfactant transport accuracy.
3. Novel Benchmark Case: The authors propose a challenging benchmark test involving a bubble in a 2D vortical flow with intensified deformation (period $T=2$). This case generates sharp interface features (tails) where normal vector calculations become inaccurate, causing convergence failure in surfactant transport. This serves as a rigorous standard for future method evaluation.
4. Implementation Efficiency: The proposed methods add negligible computational cost (the level-set reinitialization is minor compared to solving the pressure Poisson equation) and preserve the discrete conservation of both fluid and surfactant mass.

4. Results

The proposed approaches were validated through a series of numerical tests:

• 2D Uniform Flow (Diffusion):
  • The $f$-type model consistently outperformed the $f_d$-type model in accuracy and stability, particularly at high resolutions and low diffusion coefficients ($D=10^{-9}$), where the $f_d$-type model became unstable.
  • Decoupling widths: Using a wider delta function ($\hat{W} = 6\Delta x$) with the $f$-type model achieved nearly second-order accuracy, whereas fixing $\hat{W}$ to the interface width resulted in higher errors.
• 2D and 3D Vortical Flow (Advection/Deformation):
  • The $f$-type model remained stable and accurate even under severe interface deformation, whereas the $f_d$-type model became unstable with narrow delta functions.
  • Width Sensitivity: In vortical flows, increasing the delta function width too much ($\hat{W} > 5\Delta x$) increased errors due to velocity gradients distorting the surfactant profile. An optimal range of $3\Delta x$ to $6\Delta x$ was identified.
• Shear Flow (Marangoni Effect):
  • A simulation of a surfactant-laden drop in shear flow successfully reproduced the drop deformation and Marangoni forces, showing excellent agreement with reference data (Teigen et al., 2011).
• Challenging Benchmark (High Deformation):
  • In the intensified vortical flow test, the surfactant error did not converge with grid refinement due to inaccuracies in the normal vector at sharp interface features.
  • The authors demonstrated that reducing the diffusion coefficient ($D$) or modifying the normal vector calculation (setting $n=0$ where $|\nabla \psi|$ deviates from unity) could mitigate errors, though convergence rates remained below first order. This highlights a remaining challenge for the field.

5. Significance

This paper provides a practical, low-cost framework for significantly improving the accuracy of surfactant-laden two-phase flow simulations. By shifting to the $f$-type formulation and decoupling the delta function width, researchers can achieve higher fidelity without the penalty of increased computational cost or degraded interface capturing.

Furthermore, the introduction of a rigorous, non-converging benchmark case addresses a gap in the literature, offering a necessary tool for evaluating the robustness of future surfactant transport models, particularly those dealing with complex, highly deforming interfaces. The work establishes a new standard for the numerical modeling of insoluble surfactants in the diffuse-interface community.