Grid-agnostic volume of fluid approach with interface sharpening and surface tension for compressible multiphase flows

This study presents a grid-agnostic volume of fluid method for compressible multiphase flows that employs an ant-diffusive volumetric body force for interface sharpening and surface tension modeling, demonstrating accurate validation against analytical solutions and experimental data across various flow regimes.

The Gist

Imagine you are trying to simulate a complex dance between two liquids, like oil and water, or air and water, inside a computer. In the real world, the line where these two liquids meet is razor-sharp and distinct. But in a computer simulation, things get messy.

The Problem: The "Blurry" Boundary

Think of a computer grid as a giant chessboard where each square holds a piece of data about the fluid. When the computer tries to calculate how the fluids move, it has to guess what happens between the squares. This guessing game creates a kind of digital fog. Over time, the sharp line between the oil and water starts to blur, like a watercolor painting left out in the rain.

This "blurring" is bad news. If you are simulating a rocket engine or a medical injector, you need to know exactly where the fuel is and where the air is. If the boundary is fuzzy, your simulation might think the fuel is leaking or mixing too early, leading to wrong answers.

The Solution: A "Digital Eraser" and "Magnet"

The researchers in this paper developed a clever two-part fix that works on any shape of grid, not just perfect squares.

1. The "Digital Eraser" (Interface Sharpening):
   Imagine you have a blurry photo of a circle. You want to make the edge crisp again. The researchers created a special mathematical "force" that acts like an eraser. It doesn't just sit there; it actively pushes the blurry pixels back into a sharp line.

   • The Magic Trick: Usually, these sharpening tools only work on perfect square grids (like a standard chessboard). But real-world objects (like airplane wings or rock pores) are weird shapes. This new method is grid-agnostic. It's like having a magic eraser that works equally well on a square tile, a triangle, or a weirdly shaped hexagon. It looks at the neighborhood of every point and knows exactly how to sharpen the line, no matter how the grid is built.

2. The "Magnet" (Surface Tension):
   In the real world, surface tension is like a tight rubber skin that tries to pull a drop of water into a perfect sphere. In the computer, this force needs to be calculated precisely. The researchers added a "magnetic" force that pulls the interface into the shape nature wants it to be.

How They Tested It

To prove their new "magic eraser" and "magnet" work, they ran three types of tests:

• The "Star to Circle" Test: They started with a digital star shape (which has pointy, jagged edges). They turned on the surface tension. Just like a real drop of water would round itself out, the digital star smoothed into a perfect circle. They checked the pressure inside and found it matched the laws of physics perfectly.
• The "Spinning Slot" Test: They took a disk with a slot cut out of it (like a Pac-Man shape) and spun it around. Usually, the slot would get blurry and disappear. With their new method, the slot stayed sharp and distinct even after many spins, proving the "digital eraser" is working hard.
• The "Droplet Pinch-Off" Test: This is the real-world application. Imagine a stream of water being hit by a strong wind. The wind stretches the water until it snaps, creating tiny droplets. The researchers simulated this. They found that depending on how strong the wind (speed) and the water's "stickiness" (viscosity) were, the droplets broke apart in different ways. Their simulation predicted the size and shape of these droplets so accurately that it matched data from other scientists.

Why This Matters

The biggest breakthrough here is flexibility. Before this, if you wanted to simulate fluid flow around a complex object (like a car or a human heart), you had to force the grid into a specific shape, which was hard to do and often inaccurate.

Now, with this new method, engineers can build a grid that fits the object perfectly—whether it's a jagged rock formation or a smooth airplane wing—and still get a razor-sharp, accurate simulation of how fluids interact. It's like upgrading from a pixelated, blocky video game to a high-definition movie, but for scientific engineering.

In short: They invented a universal tool that keeps fluid boundaries sharp and accurate, no matter how weird the shape of the container is, allowing for much better predictions in everything from rocket fuel to oil extraction.

Technical Summary

Here is a detailed technical summary of the paper "Grid-agnostic volume of fluid approach with interface sharpening and surface tension for compressible multiphase flows" by Marziale et al.

1. Problem Statement

In computational fluid dynamics (CFD), the Finite Volume Method (FVM) discretization of multiphase flow equations inherently introduces artificial numerical diffusion. Over time, this diffusion causes the interface between immiscible fluids (e.g., liquid and gas) to smear, losing sharpness. This smearing degrades the accuracy of critical physical phenomena such as:

• Capillary effects and surface tension forces.
• Heat transfer rates across interfaces.
• Phase transition dynamics.

While Adaptive Mesh Refinement (AMR) can mitigate this by refining grids near interfaces, it introduces interpolation errors and algorithmic complexity. Level Set methods are easier to implement but often fail to strictly conserve mass without additional corrections. Existing Interface Sharpening (IS) schemes are effective but are typically restricted to structured quadrilateral grids. Many engineering applications (e.g., aircraft aerodynamics, petroleum reservoirs, coastal wave breaking) involve complex geometries requiring unstructured or arbitrary grids. There is a lack of robust, conservative interface sharpening methods that are grid-agnostic (applicable to arbitrary cell distributions) and capable of handling compressible flows.

2. Methodology

The authors propose a novel framework that casts interface sharpening as an antidiffusive volumetric body force within a compressible multiphase flow solver. The methodology consists of four main components:

A. Mathematical Formulation

• Governing Equations: The system solves the compressible Euler equations for a mixture, tracking partial density ($\rho^{(I)}\phi$), total density ($\rho$), momentum ($\rho\vec{u}$), and energy ($\rho e$).
• Interface Sharpening Term: Instead of modifying the advection equation directly, the sharpening is introduced as a source term $\dot{d}$ in the partial density equation. It is derived from a negative diffusion term ($f_{sharp}$) based on the work of Chiu and Lin:
  $$\dot{d} = \rho^{(I)} \Gamma \tilde{f}_{sharp}$$
  where $\tilde{f}_{sharp}$ involves the divergence of a flux designed to compress the interface.
• Surface Tension: Modeled using the Brackbill Continuum Surface Force (CSF) model, added as a body force to the momentum equation and a work rate term to the energy equation.
• Smoothing: To ensure stability in derivative calculations for sharp discontinuities, the volume fraction field ($\phi$) is convolved with a Gaussian filter ($\tilde{\phi} = \phi * G$) before computing gradients.

B. Grid-Agnostic Discretization

The core innovation is the generalization of gradient and divergence calculations to arbitrarily constructed cell centers (unstructured grids).

• Topological Approach: The mesh is represented as a Directed Acyclic Graph (DAG) or Hasse diagram to define neighbor relationships without relying on uniform spacing or orthogonal edges.
• Gradient Computation: The authors utilize a control volume approach where the gradient of a scalar field at a point (center or vertex) is calculated by summing fluxes over the facets of a local control volume formed by the point's neighbors. This allows for accurate gradient estimation on simplex, hexahedral, or mixed-element meshes.
• Implementation: The method leverages PETSc DMPlex for mesh generation and topological data structures.

C. System Closure and Flux Evaluation

• Equation of State (EOS): Each phase follows a Stiffened Gas EOS. A Newton-Raphson solver is used to close the system by solving for auxiliary variables (phase densities, temperature, pressure) ensuring thermodynamic equilibrium ($p^{(I)} = p^{(II)}$).
• Pressure Relaxation: A specific modification is applied to the pressure EOS to prevent unphysical negative pressures when density is slightly underestimated in high-density-ratio scenarios.
• Flux Evaluation: The AUSM+up (Advection Upstream Splitting Method) scheme is used for flux evaluation at cell faces, adapted to calculate mixture properties (speed of sound, Mach number) based on local volume fractions.

3. Key Contributions

1. Grid-Agnostic Sharpening: The first derivation of an interface sharpening term as a body force that is valid for arbitrary cell center distributions, removing the dependency on structured quadrilateral grids.
2. Compressible Multiphase Capability: The method is fully integrated into a compressible flow solver, handling high density ratios and stiffened gas equations of state, which is critical for applications like underwater explosions and rocket propulsion.
3. Robust Numerical Stability: The use of Gaussian convolution for gradient calculation and a specific pressure relaxation function ensures stability even with high density ratios and sharp curvature gradients.
4. Conservative Formulation: Unlike some Level Set methods, this Volume of Fluid (VOF) approach with the proposed sharpening term maintains strict mass conservation for both phases.

4. Results and Validation

The model was validated through a series of rigorous test cases:

• Curvature Convergence: A unit sphere test demonstrated first-order convergence ($\approx 0.97$) in curvature calculation, confirming the accuracy of the gradient discretization on unstructured grids.
• Interface Sharpening (Static Circle): A diffuse circular interface was successfully sharpened into a binary jump on both quadrilateral and simplex (triangular) meshes. The interface thickness converged to a single cell width ($\epsilon^* \to 1$) with minimal shape distortion.
• Zalesak Disk: A classic advection test involving a slotted disk rotating 360 degrees. The method showed higher convergence orders (1.19 for $L_1$ norm) compared to previous literature, preserving the sharp slot geometry with minimal numerical diffusion.
• Young-Laplace Consistency (Shape Recovery): A four-pointed star interface was allowed to relax under surface tension. The simulation correctly predicted the final equilibrium radius and the pressure jump ($\Delta p = \sigma / R$) according to the Young-Laplace condition, with errors converging to zero.
• Mass Conservation: In a 3D shape recovery test with a high density ratio (air/water), the mass loss due to the sharpening term was found to be negligible ($O(10^{-7})$), even for the lighter phase.
• Droplet Pinchoff: The model simulated droplet breakup driven by shear flow across various Weber (We) and Ohnesorge (Oh) numbers. The resulting droplet shapes (deformation, bag breakup, multimodal, shear breakup) and child droplet size ratios matched established literature correlations and physical expectations.

5. Significance

This study provides a versatile and precise tool for simulating complex multiphase flows in real-world engineering scenarios where geometry dictates the mesh structure. By decoupling interface sharpening from grid topology, the method enables:

• Accurate simulation of complex geometries: Such as flow around aircraft fuselages, through fractured rock, or over coastal reefs, without the need for excessive grid refinement or complex AMR logic.
• Reliable prediction of interfacial dynamics: Crucial for designing fuel injectors, understanding boiling/condensation, and modeling underwater explosions.
• Scalability: The method demonstrated sublinear strong scaling efficiency, making it suitable for high-performance computing (HPC) environments.

In summary, Marziale et al. have successfully bridged the gap between high-fidelity interface sharpening techniques and the geometric flexibility required for modern engineering CFD applications.