💻 computer science

A hybrid sharp-diffuse interface approach to accurately model melt pool dynamics with rapid evaporation in laser-based processing of metals

This paper presents a hybrid sharp-diffuse interface approach that combines a sharp-interface CutFEM thermal model with a diffuse-interface level-set flow solver to accurately simulate melt pool dynamics and rapid evaporation in laser-based metal processing, significantly outperforming purely diffuse-interface models in accuracy and computational efficiency.

Original authors: Nils Much, Andreas Koch, Christoph Meier, Magdalena Schreter-Fleischhacker

Published 2026-06-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Nils Much, Andreas Koch, Christoph Meier, Magdalena Schreter-Fleischhacker

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to predict how a drop of water behaves when a super-hot laser beam hits a piece of metal. This isn't just a simple splash; it's a chaotic dance of melting, boiling, and vaporizing that happens in a fraction of a second. This process is crucial for things like 3D printing metal parts or welding, but it's incredibly hard to simulate on a computer because the physics are so extreme.

This paper introduces a new, smarter way to run these simulations. The authors call it a Hybrid Sharp–Diffuse Interface (HSDI) approach. Here is the breakdown of what they did and why it matters, using simple analogies.

The Problem: The "Blurry Photo" vs. The "Sharp Photo"

To understand the new method, you first need to understand the old problem.

The Old Way (Diffuse Interface): Imagine trying to take a photo of a sharp line between black and white, but your camera is slightly out of focus. The line becomes a blurry gray gradient. In computer simulations, this "blur" happens at the boundary where the metal turns into gas. Because the laser heats the metal so intensely, the temperature changes drastically right at that boundary. When the simulation blurs this line, it gets the temperature wrong.
Why this is bad: The forces that drive the metal to move (like the pressure from evaporating gas) depend exponentially on that temperature. Think of it like a light switch: a tiny change in temperature doesn't just turn the light up a little; it flips the switch from "off" to "blindingly bright." If your simulation gets the temperature slightly wrong because of the "blur," it predicts the wrong forces, leading to a completely wrong prediction of how the metal pool will behave.

The Solution: A "Hybrid" Camera

The authors realized they didn't need to blur the whole picture. They needed a way to keep the boundary sharp where it matters most, while still handling the messy, moving parts of the fluid.

They created a Hybrid System that uses two different tools for two different jobs:

The "Sharp" Tool (For Heat): For calculating temperature, they use a method called CutFEM. Imagine this as a high-resolution camera that can cut perfectly through the grid of the simulation. It doesn't blur the line between metal and gas; it sees the sharp edge. This ensures the temperature is calculated with extreme precision right where the laser hits.
The "Diffuse" Tool (For Flow): For calculating how the liquid metal moves, splashes, breaks apart, and merges back together, they use a Diffuse Interface method. This is like the blurry camera mentioned earlier. It's not perfect for temperature, but it is incredibly robust and good at handling complex shapes (like a bubble popping or a stream breaking into droplets) without the computer crashing.

The Magic Trick:
The real innovation is how they connect these two. The "Sharp" tool calculates the precise temperature at the boundary. Then, the computer takes that sharp temperature and "smears" it just a tiny bit into the "Diffuse" flow area. This allows the flow simulator to use the correct, high-precision temperature data to calculate the forces, even though the flow itself is being simulated with the "blurry" method.

The Results: Getting the Same Answer with Less Work

The authors tested this new hybrid method against the old "all-blurry" method using several benchmarks:

Accuracy: In tests mimicking laser heating, the new method was 100 times more accurate (two orders of magnitude) than the old method when using the same computer grid size.
Efficiency: Because the new method is so accurate, they didn't need a super-fine grid. They could use a grid that was 10 to 100 times coarser (larger blocks) and still get the same level of accuracy as the old method. This means simulations that used to take days or require supercomputers could potentially run much faster.
Real-World Test: They ran a 3D simulation of a laser melting a metal plate. The new method successfully handled the chaotic formation of a "keyhole" (a deep vapor depression) and the violent splashing of the melt pool without failing, proving it is robust enough for real engineering problems.

The Bottom Line

Think of the old simulation method as trying to navigate a stormy sea with a foggy map. You might get there, but you'll likely miss the details and make mistakes.

This new Hybrid Sharp–Diffuse method is like having a high-definition map for the dangerous cliffs (the temperature boundary) while using a reliable, flexible compass for the open ocean (the fluid flow). It gives engineers a much clearer, more accurate picture of what happens when lasers melt metal, allowing them to design better 3D printers and welding processes without needing to run simulations that are impossibly expensive or slow.

Technical Summary: A Hybrid Sharp–Diffuse Interface Approach for Melt Pool Dynamics

Problem Statement
Predictive simulation of melt pool dynamics in laser-based metal processing (e.g., laser beam welding, powder bed fusion) requires resolving complex thermo-hydrodynamic interactions at the melt–gas interface. Key driving forces, such as evaporation-induced recoil pressure and temperature-dependent surface tension, depend sensitively—often exponentially—on the interface temperature. Standard diffuse-interface (DI) formulations, while robust in handling complex topology changes like breakup and coalescence, smear extreme thermal gradients across a finite transition region. This smearing introduces critical errors in the predicted interface temperature, which subsequently degrade the accuracy of interface force predictions and the resulting melt pool dynamics. Conversely, sharp-interface (SI) methods offer high thermal fidelity but often struggle with the numerical complexity and robustness required for highly dynamic multi-phase flows involving large density ratios and topological changes. Existing models typically force a trade-off between physical fidelity and computational feasibility, often requiring prohibitively fine mesh resolutions or extensive parameter calibration to compensate for thermal inaccuracies.

Methodology
The authors propose a Hybrid Sharp–Diffuse Interface (HSDI) approach that couples a sharp-interface thermal model with a diffuse-interface flow model to achieve high-fidelity modeling without incurring the computational costs of a fully sharp-interface multi-phase simulation.

Sharp-Interface Thermal Model (CutFEM): The heat transfer problem is solved using a sharp-interface unfitted finite element method (CutFEM). This formulation accurately resolves the discontinuities in temperature and heat flux at the metal–gas interface. The authors present two variants:
- SI-2P (Coupled Two-Phase): Solves the heat equation in both the metal and ambient gas phases, weakly enforcing temperature continuity across the interface using Nitsche's method.
- SI-1P (Single-Phase): Solves the heat equation only in the metal domain, treating the gas phase as adiabatic except for imposed heat fluxes. This variant is motivated by the negligible heat capacity of the ambient gas in unresolved evaporation models, offering higher computational efficiency.
  Both variants employ face-based ghost-penalty stabilization to handle small cut elements.
Diffuse-Interface Flow Model: The multi-phase flow problem, characterized by large density ratios and complex interface dynamics, is modeled using a robust level-set-based one-fluid DI finite element formulation. This approach utilizes a Continuum Surface Force (CSF) model for surface tension and recoil pressure, allowing for the natural handling of interface breakup and coalescence.
Consistent Coupling Strategy: To bridge the two frameworks, the sharp-interface temperature field is extended into a narrow band surrounding the metal–gas interface using a closest-point projection algorithm. This extended temperature field ( $\tilde{T}_\Gamma$ ) is then used to evaluate temperature-dependent interface forces (recoil pressure and surface tension) within the diffuse-interface flow framework. This ensures that the highly accurate thermal data from the SI model drives the DI flow solver consistently.
Numerical Framework: The implementation utilizes the deal.II library with matrix-free operator evaluation, adaptive mesh refinement, and MPI-based parallelization. A partitioned operator-splitting scheme is employed for time integration, treating coupling terms explicitly while solving individual fields implicitly.

Key Contributions

Novel Hybrid Formulation: This work presents the first HSDI approach for two-phase thermo-hydrodynamics that couples SI CutFEM temperature modeling with DI FEM flow modeling based on an unresolved evaporation model.
Thermal Accuracy: The development of CutFEM-based SI formulations (SI-2P and SI-1P) that enable highly accurate predictions of the temperature field, particularly at the melt pool surface, where standard DI models fail.
Coupling Algorithm: A consistent strategy for extending the sharp-interface temperature into the diffuse transition region, enabling the accurate evaluation of temperature-dependent source terms in the DI flow solver.
Benchmarking: The proposal of a novel 2D benchmark example representing a laser-induced vapor depression to comprehensively assess the accuracy of coupled thermo-hydrodynamic models.

Results

Thermal Convergence: In purely thermal benchmarks mimicking laser–metal interactions, the SI models exhibited second-order spatial convergence ( $O(h^2)$ ), whereas standard DI models showed only first-order convergence ( $O(h^1)$ ).
Accuracy vs. Mesh Resolution: For a target accuracy of 1% in interface temperature and recoil pressure, the SI models allowed for finite element sizes two orders of magnitude larger than those required by DI models. Specifically, the SI approach achieved the benchmark tolerance with a mesh roughly 50 times coarser than the DI approach.
Coupled Thermo-Hydrodynamics: In the novel 2D vapor depression benchmark, the HSDI approach was found to be one order of magnitude more accurate than a purely DI model on the same mesh. The HSDI approach successfully predicted the stable and unstable (keyhole) dynamics of the vapor depression with significantly lower error in interface geometry and kinetic energy compared to the DI model.
3D Robustness: A 3D proof-of-principle simulation of stationary laser-induced melting of a bare metal plate demonstrated the robustness of the HSDI approach in handling complex, highly dynamic two-phase flows, including keyhole formation, breakup, and coalescence, without numerical instability.
Efficiency of SI-1P: The single-phase thermal variant (SI-1P) was shown to preserve accuracy comparable to the fully coupled SI-2P variant while offering computational efficiency, validating the assumption that gas-phase heat transfer is negligible in this context.

Significance
The paper claims that the HSDI approach provides an unprecedented balance of predictive accuracy and computational efficiency for melt pool simulations. By decoupling the thermal and flow problems into their respective optimal formulations (sharp for thermal, diffuse for flow), the method overcomes the limitations of purely DI models, which require impractically fine meshes to resolve thermal gradients, and purely SI models, which struggle with complex topological changes. The methodology enables high-fidelity predictions of melt pool dynamics on mesh resolutions feasible for industrially relevant length and time scales. The authors emphasize that this approach is applicable to a broad range of temperature-driven flow problems involving strongly localized heat sources, such as laser beam welding and laser powder bed fusion, and serves as a foundation for future work incorporating more detailed physics like ray-tracing and temperature-dependent absorptivity.

The Problem: The "Blurry Photo" vs. The "Sharp Photo"

The Solution: A "Hybrid" Camera

The Results: Getting the Same Answer with Less Work

The Bottom Line

More like this