💻 computer science

Phase-Field Peridynamics

This paper introduces a novel phase-field peridynamics approach that replaces irreversible bond deletion with continuous energetic and kinematic degradation to maintain the accuracy of the bond-associated correspondence formulation and ensure thermodynamic consistency, thereby overcoming numerical instabilities in fracture modeling.

Original authors: Kai Partmann, Christian Wieners, Michael Ortiz, Kerstin Weinberg

Published 2026-04-13

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Kai Partmann, Christian Wieners, Michael Ortiz, Kerstin Weinberg

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 simulate how a piece of glass shatters when dropped, or how a metal beam cracks under pressure. For decades, computer scientists have struggled with this because cracks are messy: they are sudden breaks, jagged edges, and discontinuities that don't play nice with the smooth math usually used to describe how materials bend and stretch.

This paper introduces a new, smarter way to simulate these breaks using a method called Phase-Field Peridynamics. Here is the story of how it works, explained without the heavy math.

The Problem: The "Snap" vs. The "Fade"

To understand the new method, let's look at the old way of doing things.

The Old Way (Bond Deletion):
Imagine a giant net made of rubber bands connecting thousands of tiny dots (representing atoms or particles). In traditional computer models, when a rubber band stretches too far, the computer simply snaps it. It deletes the bond instantly.

The Flaw: In complex models, if you delete too many rubber bands in one area, the "net" loses its shape. It becomes wobbly and unstable. The computer gets confused about how the remaining dots should move, leading to a crash or a fake result. It's like trying to hold up a tent by cutting half the poles; the structure collapses before the fabric even tears.

The New Way (Phase-Field Peridynamics):
The authors of this paper say, "Don't snap the rubber band. Instead, let it fade."
Instead of deleting a bond instantly, the model introduces a "dimmer switch" for every single rubber band. As the band stretches, it doesn't disappear; it just gets weaker and weaker, losing its ability to hold weight, until it is essentially useless.

The Benefit: The net never loses its shape because the connections are still there, just "ghostly." This keeps the computer's math stable, even when the material is falling apart.

The Two-Step Dance: Energy vs. Geometry

The paper introduces a clever trick to make this "fading" work perfectly. They use two different "dimmer switches" for two different jobs:

The Energy Switch (The "Strength"):
This switch controls how much force the bond can carry. As damage happens, this switch turns down the strength of the bond smoothly. This ensures the energy calculation is correct (matching the physics of how much energy it takes to break a crack).
The Geometry Switch (The "Shape"):
This is the secret sauce. Even if a bond is totally broken (carrying zero strength), the model keeps it in the "geometry" calculation for a little while. It's like keeping a broken pole in the tent frame just to help the tent keep its shape, even though the pole can't hold any weight anymore.
- Why? This prevents the computer from getting confused about the shape of the material near the crack. It stops the "wobbly net" problem that caused the old methods to crash.

The "Normalization" Secret Sauce

The authors also solved a tricky math problem: How do we know exactly how much to "fade" the bonds so the energy matches real life?

In physics, there's a rule (Griffith's theory) that says breaking a crack requires a specific amount of energy. If you fade the bonds too slowly, the crack takes too much energy to grow. Too fast, and it's too easy.

The authors derived a precise "recipe" (a normalization constant) to get this right.

The Analogy: Imagine you are pouring water into a bucket with a hole in the bottom. You need to know exactly how fast to pour so the water level rises at the perfect rate. The authors calculated this "pouring rate" for any shape of the "hole" (the mathematical kernel) using a simple ratio of two integrals. It's a universal key that ensures the simulation obeys the laws of physics, no matter how you set up the grid.

The Proof: Putting it to the Test

The team tested their new method on three famous scenarios:

The Straight and Curved Cracks (Mode I & II):
They simulated a plate being pulled apart (straight crack) and a plate being sheared (curved crack). The old method crashed during the shearing test because the "net" got too wobbly. The new method handled it perfectly, showing the correct curved path.
The Branching Crack (Boundary Tension Test):
They simulated a crack that splits into two branches (like a lightning bolt). They tested this with different mathematical "shapes" for the bonds. The results were identical every time, proving their "recipe" for the energy switch works universally.
The Kalthoff-Winkler Experiment (The Bullet Test):
This is a famous real-world experiment where a steel plate is hit by a projectile, causing cracks to shoot out at a specific 70-degree angle.
- The Result: The new model predicted this 70-degree angle perfectly and showed a smooth, realistic speed for the crack tip. The old models were a bit jittery and predicted the wrong angle.

The Big Picture

Think of this paper as upgrading the "crash test" for computer simulations.

Old Simulations: Like a puppet show where the strings are cut when the puppet breaks. The puppet falls apart chaotically, and the math breaks.
New Simulation: Like a puppet show where the strings turn into invisible, ghostly threads when the puppet breaks. The puppet falls apart realistically, but the invisible threads keep the math stable and the physics accurate.

By combining the stability of "fading" bonds with a precise mathematical recipe for energy, this new Phase-Field Peridynamics approach allows engineers to simulate complex, dynamic fractures (like car crashes or earthquake damage) with a level of accuracy and stability that was previously impossible.

1. Problem Statement

The paper addresses a critical stability issue in Peridynamics (PD), a non-local continuum theory well-suited for fracture modeling because it avoids spatial derivatives that become undefined at discontinuities.

The Specific Challenge: While the Bond-Associated Correspondence (BAQP) formulation is the most accurate PD framework (eliminating zero-energy modes and surface effects), it suffers from numerical instability when modeling fracture using the traditional irreversible bond deletion method.
The Mechanism of Failure: In correspondence formulations, the deformation gradient is approximated using a "shape tensor" derived from the neighborhood of bonds. As a crack propagates, bonds are deleted. This progressive removal degrades the shape tensor, potentially making it ill-conditioned or singular. This leads to a breakdown of the kinematic approximation and numerical failure, particularly in complex loading scenarios like Mode II shear or dynamic branching.
Limitations of Existing Solutions: Traditional phase-field methods applied to PD often require fine mesh resolution to resolve diffuse damage bands and may not fully leverage the accuracy of the bond-associated correspondence framework without introducing similar stability issues.

2. Methodology

The authors propose a novel Phase-Field Peridynamics (PFPD) approach that integrates a phase-field fracture model into the bond-associated correspondence framework. The core innovation is replacing abrupt bond deletion with continuous degradation.

Key Methodological Components:

Bond-Associated Phase-Field Parameter ( $s$ ): Instead of a binary state (intact/deleted), each bond is assigned a continuous phase-field parameter $s \in [0, 1]$ , where $s=0$ is undamaged and $s=1$ is fully damaged.
Dual Degradation Strategy:
1. Energetic Degradation ( $g(s)$ ): The stress contribution of a bond is continuously reduced using a quadratic degradation function $g(s) = (1-s)^2$ . This ensures smooth energy dissipation.
2. Kinematic Degradation ( $h(s)$ ): Crucially, the shape tensor (used for the deformation gradient) is not degraded immediately. A separate kinematic degradation function $h(s)$ is introduced, which remains unity ( $h=1$ ) until the damage exceeds a threshold $s_c$ (e.g., 0.95). Only then does it degrade. This preserves the invertibility and conditioning of the shape tensor even in highly damaged regions, preventing the numerical instabilities associated with bond deletion.
Variational Formulation: The damage evolution is derived from a bond-wise energy minimization principle, ensuring thermodynamic consistency with Griffith's fracture theory. The evolution law uses a history variable to enforce irreversibility.
Analytical Normalization Constant ( $c_0$ ): A major theoretical contribution is the derivation of an exact analytical expression for the normalization constant $c_0$ $c_{0}$ . This constant ensures that the total energy dissipated by the diffuse crack in the non-local model matches the Griffith energy release rate ( $G_c$ $G_{c}$ ) per unit crack area.
- The derivation considers a planar crack in an infinite domain with a spherical kernel.
- The result is expressed as a ratio of two one-dimensional integrals over the unit interval, dependent only on the chosen kernel function $\omega_s(r)$ :
  $c_0 = \frac{\int_0^1 \omega_s(\rho\delta) \rho^3 d\rho}{2 \int_0^1 \omega_s(\rho\delta) \rho^2 d\rho}$
- This allows for exact calculation for any spherical kernel (e.g., cubic spline, linear) without numerical approximation errors.

3. Key Contributions

Stabilized Phase-Field Peridynamics: The introduction of the dual degradation scheme (separating energetic and kinematic degradation) allows the use of the highly accurate bond-associated correspondence formulation for fracture problems without the numerical instabilities caused by bond deletion.
Exact Normalization Constant: The paper provides a rigorous, closed-form analytical derivation for the normalization constant $c_0$ for general spherical kernels. This resolves a common source of error in non-local phase-field models where the energy dissipation often deviates from the theoretical $G_c$ due to kernel truncation or approximation.
Algorithmic Implementation: The authors present a detailed algorithm (Algorithm 2) comparing the new PFPD approach with the standard BAQP model, highlighting the shift from binary bond flags to continuous phase-field variables and history tracking.
Thermodynamic Consistency: The framework is rigorously derived from an energy minimization principle, ensuring consistency with Griffith's theory and the second law of thermodynamics.

4. Numerical Results and Verification

The proposed model was validated through several numerical examples using the Julia package Peridynamics.jl:

Mode I and Mode II Fracture:
- Mode I: The model correctly predicted straight crack propagation.
- Mode II (Shear): The standard BAQP model with bond deletion became unstable and failed due to shape tensor degradation. The PFPD model remained stable and correctly predicted the curved crack path, demonstrating superior robustness.
Boundary Tension Test (BTT) - Normalization Verification:
- Simulations were run with different horizon ratios ( $m_\delta$ ) and kernel functions (cubic B-spline and linear).
- Result: The crack branching patterns, branching angles, and branching points were consistent across all variations. This confirmed that the derived normalization constant $c_0$ correctly scales the energy dissipation regardless of the kernel or horizon size.
- Crack Tip Velocity: The velocity remained below the theoretical limit of $0.5 c_R$ (half the Rayleigh wave speed), and the velocity profile was smoother than in bond-based models.
Kalthoff-Winkler Experiment:
- A dynamic impact test on a pre-notched steel plate.
- Result: The PFPD model predicted a crack initiation angle of approximately 70°, matching experimental observations. The standard Ordinary State-Based (OSB) model showed deviations.
- The PFPD model produced smoother crack tip velocity profiles and avoided the numerical oscillations seen in other formulations.

5. Significance

This work represents a significant advancement in computational fracture mechanics by bridging the gap between the accuracy of correspondence-based peridynamics and the stability required for complex dynamic fracture.

Robustness: It solves the long-standing issue of shape tensor singularity in correspondence PD during fracture, making the most accurate PD formulation viable for practical engineering applications involving crack branching and complex loading.
Theoretical Rigor: The analytical derivation of the normalization constant provides a solid theoretical foundation for non-local phase-field models, removing the need for empirical tuning of energy parameters.
Versatility: The approach is applicable to various kernel functions and horizon ratios, making it a versatile tool for simulating dynamic fracture in 3D structures without the mesh sensitivity issues of classical FEM or the instability of traditional bond deletion.

In summary, the paper introduces a robust, thermodynamically consistent framework that enables the simulation of complex dynamic fracture phenomena using the most accurate peridynamic formulations available, overcoming the numerical limitations that previously hindered their application.