A unified sharp-diffusive phase-field model for bulk and interfacial cohesive fracture

This paper proposes a unified sharp-diffusive phase-field model that incorporates a localized interfacial source term to independently control interface toughness and accurately simulate both bulk and interfacial cohesive fracture without requiring complex corrections or fine mesh refinement.

The Gist

Imagine you are trying to predict how a complex machine, like a jet engine or a high-tech bridge, will break when it gets stressed. These machines aren't made of one single block of metal; they are multiphase materials, meaning they are like a mosaic made of different tiles (different metals, fibers, or ceramics) glued together.

The biggest problem engineers face is the glue (the interface) between these tiles. In the real world, if the glue is weak, the tiles will peel apart before the tiles themselves crack. But in computer simulations, it's incredibly hard to model this "peeling" accurately without the computer getting confused or needing to be impossibly powerful.

Here is a simple breakdown of what this paper does to solve that problem, using some everyday analogies.

The Old Problem: The "Blurry" Glue

Imagine you are drawing a picture of a crack forming between two different colored blocks of clay.

• The Traditional Method: Old computer models treated the "glue" as a blurry transition zone. They couldn't draw a sharp line between the clay and the glue. Instead, they smeared the properties out.
• The Result: To get the math right, the computer had to zoom in so incredibly close (using tiny, tiny pixels) that the simulation would take days or weeks to run. Even then, the "strength" of the glue often came out wrong because it was accidentally mixed with the strength of the clay next to it. It was like trying to measure the exact thickness of a piece of tape by looking at a blurry photo of it.

The New Solution: The "Sharp" Glue with a Magic Switch

The authors of this paper have built a new model that acts like a high-definition camera combined with a magic switch.

1. The "Sharp" vs. "Diffusive" Mix

Think of the material as a road.

• Diffusive (Old way): The road slowly fades from asphalt to gravel. It's hard to tell exactly where the asphalt ends and the gravel begins.
• Sharp (New way): The authors use a special technique (called the $\Omega^2$-model) that allows the road to have a perfectly sharp edge. The asphalt stops, and the gravel starts instantly.
• The Analogy: Imagine a knife cutting through butter. Old models tried to describe the cut as a "softening" of the butter over a wide area. This new model describes the cut as a clean, sharp slice. This allows the computer to see the exact moment the "glue" fails without needing millions of tiny pixels.

2. The "Magic Switch" (The Source Term $q_\phi$)

This is the paper's biggest innovation.

• The Problem: Even with a sharp knife, if you just tell the computer "the glue is weak," the math still gets confused because the "blur" of the surrounding material tries to pull the numbers up.
• The Fix: The authors introduce a localized "source term" (think of it as a specialized override button or a spotlight).
• How it works: When the simulation reaches the interface (the glue), this "spotlight" turns on. It tells the computer: "Ignore the surrounding clay for a second. Right here, the glue has exactly THIS strength and THIS toughness."
• The Result: The computer can now set the glue's strength independently, just like a human engineer would specify it on a blueprint, without needing complex math corrections.

Why This Matters (The "Aha!" Moment)

In the real world, cracks do two tricky things:

1. They peel along the glue (debonding).
2. They smash through the material (matrix cracking).
3. They sometimes switch between the two.

Imagine a crack traveling down a road. It hits a patch of weak glue. Does it peel along the glue, or does it smash through the concrete?

• Old models often got this wrong or required so much computing power to figure it out that they were impractical for real-world designs.
• This new model acts like a smart GPS. It instantly calculates the path of least resistance. It can show the crack peeling along the glue for a few inches, then suddenly deciding, "This glue is too strong now," and smashing through the concrete.

The Bottom Line

This paper presents a unified, sharp-diffusive phase-field model.

• Unified: It uses one set of rules for both the "glue" and the "material."
• Sharp-Diffusive: It keeps the math smooth enough to be stable (diffusive) but sharp enough to show exact breaks (sharp).
• Efficient: Because it doesn't need to zoom in infinitely to see the glue, it runs much faster on computers.

In short: The authors have invented a new way to simulate how complex materials break. It's like upgrading from a low-resolution, blurry map to a high-definition, real-time GPS that can instantly tell you exactly where a crack will go, whether it will peel or smash, and how strong the glue really is—all without crashing your computer. This helps engineers design safer, lighter, and stronger planes, cars, and buildings.

Technical Summary

1. Problem Statement

Multiphase materials (e.g., composites, alloys) exhibit complex failure mechanisms where interfacial debonding competes with matrix cracking. Traditional phase-field modeling faces a fundamental challenge in these heterogeneous systems:

• Non-locality of Fracture Energy: In standard phase-field models, the regularization of fracture energy is non-local. This causes the effective interfacial toughness to be inherently coupled with the properties of the surrounding bulk phases.
• Inaccurate Interface Properties: Achieving prescribed interfacial toughness usually requires complex constitutive mapping, ad-hoc corrections, or extremely fine mesh refinement near interfaces to resolve the diffusive zone.
• Limitations of Existing Solutions: Previous methods (e.g., multi-phase-field or cohesive zone elements) often rely on the assumption of a finite diffusive interface width, leading to high computational costs due to strict mesh requirements and failing to capture "sharp" displacement discontinuities naturally within a continuum framework.

2. Methodology

The authors propose a unified sharp-diffusive phase-field model based on the recently developed $\Omega_2$-model. The methodology integrates three core components:

A. The $\Omega_2$-Model Framework

The model decouples the description of crack geometry and stiffness degradation by introducing two fields:

1. Phase-field ($\phi$): Remains responsible for energetic regularization (smoothing the crack surface to avoid mesh sensitivity).
2. Damage field ($\omega$): An independent variable that exhibits a Dirac-like concentration. This allows for the emergence of strong displacement discontinuities (sharp cracks) within a continuum framework, distinct from discrete cohesive elements.

B. Analytical Interfacial Source Term ($q_\phi$)

To resolve the non-locality issue where interface toughness is averaged with bulk properties, the authors introduce a strongly localized analytical source term, $q_\phi$, into the phase-field governing equations:
$$q_\phi = -\frac{2\Delta G_c^I}{\ell}(1-\phi)\omega \mathbb{1}_{B_I}(x)$$
Where:

• $\Delta G_c^I = \bar{G}_c - G_c^I$ is the difference between the average bulk fracture energy and the target interfacial fracture energy.
• $\mathbb{1}_{B_I}(x)$ is the indicator function for the interface region.
• Function: This term acts as a compensatory driving force that precisely reduces the effective fracture energy at the interface to match the prescribed local material property ($G_c^I$), independent of the surrounding bulk.

C. Unified Cohesive Law

The model unifies the description of bulk and interfacial failure using the parametric cohesive laws proposed by Feng and Li. The traction-separation behavior is governed by:
$$\hat{\sigma}_{eff} = 1 - \phi^* \quad \text{and} \quad \hat{\Delta}_{eff} = \Omega(\phi^*)$$
This formulation applies to both bulk and interface regions, with parameters ($G_c$, $\sigma_c$) determined directly by local physical properties without correction factors. The model supports various softening laws (linear, exponential, and the single-parameter $p$-model).

3. Key Contributions

1. Sharp-Diffusive Hybridization: The model successfully bridges the gap between diffusive phase-field regularization and sharp cohesive zone behavior. It captures emergent strong discontinuities (sharp cracks) naturally within a continuum setting, avoiding the need for discrete cohesive elements.
2. Independent Interface Control: By introducing the source term $q_\phi$, the model allows for the independent and precise prescription of interfacial fracture toughness and strength. It eliminates the need for complex corrections or the assumption that interface properties are averages of adjacent bulk materials.
3. Computational Efficiency: Due to the strong localization of the damage field $\omega$, the model can accurately resolve interfacial failure using only a single layer of elements at the interface, regardless of the phase-field length scale $\ell$. This significantly reduces the mesh refinement requirements compared to traditional diffusive models.
4. Unified Framework: It provides a single set of parametric equations to describe mixed-mode (tension-shear) cohesive failure in both bulk and interfacial regions, seamlessly handling the competition between matrix cracking and interface debonding.

4. Results and Validation

The model was validated through four numerical benchmarks:

• Example 1 (Uniaxial Tension of Three-Phase Bar):

  • Demonstrated that the model accurately reproduces various cohesive laws (linear, exponential, $p$-model) at the interface.
  • Confirmed that the damage field $\omega$ remains strictly localized to the interface, decoupled from the phase-field bandwidth $\ell$.
  • Showed convergence to theoretical curves even with coarse meshes ($h/\ell = 1/5$).

• Example 2 (Double-Cantilever Beam - DCB):

  • Simulated pure Mode-I interfacial fracture.
  • Results showed excellent agreement with analytical traction-separation laws.
  • The model spontaneously selected the failure path along the interface without kinematic constraints, capturing the sharp displacement jump before the full phase-field band developed.

• Example 3 (Single-Fiber Reinforced Composite):

  • Captured the complex competition between interfacial debonding and matrix cracking.
  • Successfully simulated the kinking instability, where a crack initiates at the weak interface, propagates along it, and then deflects into the stronger matrix upon reaching a critical state.

• Example 4 (Crack Impinging on Inclined Interface):

  • Evaluated the competition between crack deflection (along the interface) and penetration (into the matrix) under mixed-mode loading.
  • The model accurately predicted the transition between deflection and penetration based on the ratio of interface-to-matrix fracture energy ($G_c^I/G_c^M$) and the inclination angle, matching Linear Elastic Fracture Mechanics (LEFM) theoretical predictions.

5. Significance

This work offers a robust and computationally efficient tool for predicting complex fracture trajectories in sophisticated engineering materials.

• Theoretical Advancement: It resolves the long-standing issue of non-locality in phase-field modeling of multiphase materials, allowing for the direct use of local material properties without ad-hoc corrections.
• Practical Application: The ability to model sharp interfaces with a single element layer makes it highly suitable for simulating large-scale heterogeneous structures (e.g., fiber-reinforced composites, multi-phase alloys) where traditional methods are computationally prohibitive.
• Versatility: The unified framework simplifies the simulation of mixed-mode failures, making it applicable to a wide range of engineering problems involving interfacial debonding and matrix cracking.