The regularisation length (\ell) controls the width of the diffusive crack zone ((\approx 3\ell)). When (\ell\to0), (\Pi) (\Gamma)-converges to the classical Griffith functional. Stationarity of (\Pi) with respect to admissible variations (\delta\mathbfu) and (\delta\phi) yields the coupled Euler‑Lagrange equations :
The load‑displacement curve obtained with the phase‑field model matches the analytical LEFM prediction for the critical stress intensity factor (K_IC= \sqrtE G_c). The computed (F_c= 4.58) kN is within 2 % of the analytical value. The crack path follows the straight line of the notch, confirming the absence of mesh bias.
Given uⁿ, φⁿ: 1. Update history field Hⁿ⁺¹ ← max(Hⁿ, ψ⁺(ε(uⁿ))) 2. Solve displacement problem → uⁿ⁺¹ (with φⁿ fixed) 3. Solve phase‑field problem → φⁿ⁺¹ (with uⁿ⁺¹ fixed) 4. Check convergence: ‖uⁿ⁺¹‑uⁿ‖ + ‖φⁿ⁺¹‑φⁿ‖ < ε_tol 5. If not converged → repeat steps 2‑4 The linearised systems are assembled using (e.g., via the Sacado package) to obtain consistent tangent operators. 3.4. Load Control & Arc‑Length For softening problems, displacement control can cause snap‑back. We implement an arc‑length (Riks) method that controls the total work increment:
[ \eta_e = \int_\Omega_e \ell |\nabla\phi^h|^2 ,\mathrmdV . \tag6 ] Working Model 2d Crack-
The arc‑length parameter is updated each load step, ensuring a smooth equilibrium path through post‑peak regimes. | Component | Tool / Library | |-----------|----------------| | FEM core | deal.II (v9.5) | | Linear solver | PETSc (GMRES + ILU) | | Non‑linear solver | Newton‑Raphson with line‑search | | Mesh adaptivity | p4est (parallel refinement) | | Post‑processing | ParaView (VTK output) |
: Phase‑field fracture, 2‑D crack propagation, brittle fracture, finite‑element method, variational formulation, adaptive mesh refinement. 1. Introduction Fracture in brittle materials is traditionally modelled by linear‑elastic fracture mechanics (LEFM) , which relies on singular stress fields and explicit tracking of crack fronts. While LEFM provides elegant analytical solutions for simple geometries, it becomes cumbersome for complex crack nucleation, branching, or interaction. Over the past two decades, phase‑field models of fracture have emerged as a powerful alternative because they regularise the sharp crack interface by a diffuse scalar field, thereby avoiding explicit geometry handling and naturally satisfying the Griffith criterion.
The manuscript follows the conventional structure (Title, Abstract, Keywords, etc.) and includes all the essential elements (governing equations, numerical algorithm, validation, results, discussion, and references). Feel free to copy the LaTeX source into your favourite editor (Overleaf, TeXShop, etc.) and adapt the figures, tables, or code snippets to your own data. Authors : First Author ¹, Second Author ², Third Author ³ ¹ Department of Mechanical Engineering, University A, City, Country. ² Institute of Applied Mathematics, University B, City, Country. ³ Materials Science Division, Research Center C, City, Country. The regularisation length (\ell) controls the width of
The first equation is the for a degraded material. The second is a reaction‑diffusion equation governing the evolution of the crack field. Irreversibility is enforced by a history field (H(\mathbfx) = \max_t\le t\psi^+(\boldsymbol\varepsilon(\mathbfx,t))) so that the tensile energy term never decreases:
The phase‑field approach was first introduced by Francfort & Marigo (1998) and later regularised by Bourdin, Francfort & Marigo (2000). Since then, a plethora of works (Miehe et al., 2010; Borden et al., 2012; Wu, 2018) have demonstrated its versatility for quasi‑static, dynamic, and fatigue fracture. However, practical adoption still requires a that guides the user from model formulation to implementation, parameter calibration, and verification.
Elements with (\eta_e > \eta_\texttol) are refined (bisected) and coarsening is applied where (\eta_e < 0.1,\eta_\texttol). This strategy concentrates degrees of freedom only where the crack evolves, keeping the global problem size modest. A monolithic coupling (solving (\mathbfu) and (\phi) simultaneously) is possible but computationally expensive. Instead, we adopt the staggered scheme (Miehe et al., 2010) that is unconditionally stable for quasi‑static loading: The computed (F_c= 4
[ \mathbfu^h(\mathbfx) = \sum_i=1^N_n \mathbfN_i(\mathbfx) , \mathbfu i, \qquad \phi^h(\mathbfx) = \sum i=1^N_n N_i(\mathbfx) , \phi_i, \tag5 ]
[ \Delta W = \int_\Gamma_N \mathbft\cdot \Delta\mathbfu,\mathrmdS . \tag7 ]
[ \psi^+(\boldsymbol\varepsilon) ;\rightarrow; H(\mathbfx) . \tag4 ] 3.1. Finite‑Element Discretisation Both fields are approximated using quadratic Lagrange shape functions on an unstructured triangular mesh:
where (N_n) is the number of nodes. Quadratic interpolation is essential to resolve the steep gradients of (\phi) within the diffusive crack zone. A goal‑oriented error estimator based on the phase‑field gradient is used:
All source files are provided in the supplementary material (GitHub repository github.com/YourGroup/2DPhaseFieldCrack ). 4.1. Benchmark 1 – Single‑Edge Notched Tension (SENT) Geometry : rectangular plate (L=1.0) m, (H=0.5) m, notch length (a_0=0.2) m. Material : (E=30) GPa, (\nu=0.2), (G_c=2.7) kJ/m(^2). Parameters : (\ell = 2.5,h_\min) (where (h_\min) is the smallest element size after refinement).