Abstract
This study explores fatigue crack propagation in face-centered cubic metals using an adaptive concurrent multiscale framework that couples coarse-grained molecular dynamics with the finite element method. The approach dynamically refines the mesh and activates atomistic regions as the crack advances, enabling high-resolution modeling near the crack tip while preserving computational efficiency in less critical regions. A crack-free surface identification method is used to track the evolving crack tip. Coarse-graining of dislocation plasticity mitigates large deformations, ensuring stable crack growth during fatigue cycles. A refinement scheme activates virtual atoms when the crack tip approaches the FEM domain, with the continuum model imposing displacements. The framework is applied to simulate fatigue failure in single-crystal aluminum. The results demonstrate agreement with fully atomistic and non-adaptive multiscale models in terms of crack trajectory, stress intensity versus crack growth rate, and the Paris law exponent, while achieving up to a 46% reduction in computational cost compared to the non-adaptive approach.