ALE-57 Two-Phase Flow Benchmark

• ALE-57 Benchmark is a standardized computational framework that assesses finite element methods for simulating two-phase incompressible Navier–Stokes flows with moving interfaces, exemplified by rising bubble tests.
• It compares Eulerian and ALE FEM discretizations by measuring metrics such as bubble sphericity, rise velocity, and interface topology to evaluate accuracy and stability.
• The benchmark underpins convergence studies and robustness analysis, guiding the calibration and development of advanced interface-tracking solvers in complex two-phase flow scenarios.

The ALE-57 benchmark refers to a class of computational experiments for evaluating and comparing numerical methods for two-phase incompressible Navier–Stokes flows with moving interfaces. These benchmarks, most notably the “rising bubble” problems as introduced and popularized in [Hysing et al., 2009], are widely used in the finite element methods (FEM) community to rigorously quantify the accuracy, stability, and efficiency of interface-tracking algorithms, including both Eulerian and Arbitrary Lagrangian–Eulerian (ALE) discretizations. The benchmark scenarios are specifically designed to probe the robustness of numerical schemes under significant interface deformation while maintaining high-fidelity resolution of dynamic phenomena such as bubble rise velocity, sphericity, and interface topology.

1. Definition and Mathematical Formulation

The ALE-57 benchmark centers on the numerical solution of the two-phase incompressible Navier–Stokes equations with explicit tracking of the interface $\Gamma(t)$ between two immiscible fluids, typically denoted as phases $+$ and $-$. The governing equations in the bulk domains $\Omega_{\pm}(t)$ are:

$$\begin{aligned} \rho\left(\frac{\partial \vec{u}}{\partial t} + (\vec{u} \cdot \nabla)\vec{u}\right) - 2\mu \nabla \cdot D(\vec{u}) + \nabla p &= \vec{f} \quad \text{in } \Omega_{\pm}(t), \ \nabla \cdot \vec{u} &= 0 \quad \text{in } \Omega_{\pm}(t), \end{aligned}$$

with appropriate boundary and interface conditions related to surface tension $\gamma$, normal $\vec{\nu}$, and curvature $\kappa$. In the ALE framework, these equations are reformulated to accommodate a time-dependent domain via a prescribed domain velocity $\mathcal{W}$, resulting in a reinterpreted time derivative:

$$h_t = \left.\frac{\partial h}{\partial t}\right|_{ALE} - \mathcal{W}\cdot \nabla h,$$

and correspondingly, the momentum equation takes the form:

$$\rho\left(\left.\frac{\partial \vec{u}}{\partial t}\right|_{ALE} + ((\vec{u} - \mathcal{W})\cdot\nabla)\vec{u}\right) - 2\mu \nabla\cdot D(\vec{u}) + \nabla p = \vec{f}.$$

At the weak level, this leads to discrete variational formulations in evolving spaces, as detailed in the ALE-57 benchmark studies.

2. Numerical Implementation: Eulerian vs. ALE FEM Discretization

The benchmark supports detailed comparison of two prominent approaches: the “fitted” Eulerian FEM and the ALE FEM.

• Eulerian FEM: At each time step, the bulk mesh is conformed to the moving interface, and if the interface moves significantly, mesh adjustment or remeshing is required. This process necessitates interpolation of previously computed fields onto the new mesh, introducing an extra computational step.
• ALE FEM: The mesh itself evolves in time according to a domain velocity $\mathcal{W}$, minimizing the need for ad-hoc velocity interpolation and retaining coherence between time steps unless the mesh becomes highly distorted.

Both methods utilize piecewise linear finite element spaces for the bulk and interface. ALE schemes additionally involve constructing an ALE mapping $x_h$ relating the reference mesh to the moving mesh.

In the fully discrete setting, for time step $m+1$ with time increment $\tau$, the ALE FEM weak formulation is represented as:

$$\begin{aligned} \left( \frac{\rho^m}{\tau}\vec{U}^{m+1}, \vec{\xi} \right)_{\Omega^{m+1}} + \left( \rho^m ((\vec{U}^{m+1} - \mathcal{W}^{m+1})\cdot \nabla)\vec{U}^{m+1}, \vec{\xi} \right)_{\Omega^m} \ + 2 (\mu^m D(\vec{U}^{m+1}), D(\vec{\xi}))_{\Omega^m} - (P^{m+1}, \nabla \cdot \vec{\xi})_{\Omega^m} \ - \gamma \langle \kappa^{m+1}\vec{\nu}^m, \vec{\xi} \rangle_{\Gamma^m} = \left( \frac{\rho^m}{\tau}\vec{U}^m, \vec{\xi} \right)_{\Omega^{m+1}} + (\rho^m \vec{f}_1^{(m+1)} + \vec{f}_2^{(m+1)}, \vec{\xi})_{\Omega^m}, \end{aligned}$$

where all notation follows the conventions established in the source studies.

3. Benchmark Problem Specifications and Measurement Metrics

The principal test problems—benchmark problem I and II—specify physical parameters (densities $\rho_{\pm}$, viscosities $\mu_{\pm}$, surface tension $\gamma$) and initial interface geometry. For instance, in benchmark I:

• $\rho_+ = 10^3$, $\rho_- = 10^2$
• $\mu_+ = 10$, $\mu_- = 1$
• $\gamma = 24.5$
• Computational domain $\Omega = (0,1) \times (0,2)$
• Initial bubble: circle of radius $1/4$

Critical quantitative measures for algorithm assessment include:

Metric	Description	Typical Value Range in Benchmark I
Sphericity $s_{min}$	Minimal ratio of bubble surface area to sphere	≈ 0.901 at minimum
Maximum rise velocity $V_c$	Peak vertical velocity of barycenter	≈ 0.241
Final vertical position	Bubble barycenter position at end	≈ 1.082

These values, computed at specific simulation milestones, are compared to reference solutions to evaluate method performance.

4. Convergence Experiments and Numerical Results

Convergence experiments in ALE-57 are conducted using exact solutions for expanding sphere (bubble) scenarios. The expanding sphere solution is characterized by a radius evolving according to $r(t) = e^{\alpha t} r(0)$ and a velocity field $\vec{u}(z, t) = \alpha z$, with the pressure profile ensuring force balance. Discrete errors are quantified via measurements such as the maximum distance between numerical and exact interface locations, and deviations in computed velocity and pressure.

The experiments demonstrate that both ALE and Eulerian schemes achieve high order accuracy:

• The interface is captured with minimal geometric discrepancy.
• Calculated velocity and pressure fields closely match analytic solutions.
• No significant deterioration in mesh quality occurs, obviating the need for frequent remeshing.

These findings confirm the algorithms' ability to handle evolving two-phase interfaces with substantial topological change.

5. Comparative Performance and Efficiency

A key consideration in the ALE-57 benchmark is the relative efficiency and robustness of ALE versus Eulerian discretizations. It is widely believed that ALE methods are advantageous due to the elimination of velocity interpolation steps inherent in Eulerian remeshing. However, numerical evidence indicates:

• For regular domains, CPU times for both approaches are comparable.
• The cost of velocity interpolation in Eulerian schemes becomes pronounced primarily when domains are nonconvex.
• All tested methods (Eulerian and ALE) maintain mesh quality via inherent tangential interface redistribution, requiring few or no full remeshing operations.

This suggests that the choice between ALE and Eulerian approaches should be informed by domain geometry and practical remeshing concerns rather than an inherent efficiency differential for the tested parameter ranges.

6. Algorithmic Robustness and Mesh Quality

Algorithmic stability and the preservation of mesh quality are essential outcomes reported in ALE-57 evaluations. All formulations incorporate auxiliary equations—such as a discrete curvature condition—to achieve controlled tangential motion of interface nodes. This built-in mechanism maintains high-quality interface meshes over long simulations, enhancing robustness and minimizing numerical artifacts associated with mesh degeneration.

A plausible implication is that the reliability of such schemes extends to a wide range of two-phase flow scenarios, provided that extreme domain deformation is avoided or handled with occasional global remeshing.

7. Significance and Application in Two-Phase Flow Simulations

ALE-57 benchmarks substantiate the accuracy, robustness, and practical equivalence of state-of-the-art fitted front tracking FEM for two-phase incompressible flows. Their use as a standard for verifying and calibrating interface-tracking solvers facilitates direct comparison across research efforts and ensures consistent evaluation using physically meaningful and reproducible metrics.

Such benchmarks are instrumental in guiding the development of computational fluid dynamics packages for scientific and engineering applications where evolving interfaces, such as bubbles, droplets, or vesicles, are central to the problem dynamics.