Grain-boundary-mediated dislocation nucleation controls the onset of plasticity in crystalline materials, yet its atomistic kinetics are difficult to resolve because conventional molecular dynamics (MD) rarely reaches the timescales relevant to these processes. Accelerated MD methods based on collective variables (CVs), such as adaptive boost MD (ABMD), can overcome this limitation, but they generally require the rare event of interest to be specified a priori. Here, we present an exploration-driven workflow for quantitative accelerated MD and apply it to dislocation nucleation from a Cu Σ9 symmetric tilt grain boundary under uniaxial tension. In the exploratory stage, we modify shuffling accelerated MD (SAMD) by replacing the nearest-neighbor off-centering absolute displacement with the smoother nearest-neighbor off-centering displacement, thereby reducing artifacts while efficiently sampling rare events. The sampled initial and final states are then used to identify the transition pathway, and a CV constructed nonempirically from the saddle-point structure is employed in ABMD for quantitative free-energy evaluation. This workflow captures the stress-dependent transition in grain-boundary dislocation nucleation from a single-dislocation mode to a multiple-dislocation mode without prescribing the mechanism in advance. The switching stress between the two modes is consistent with previous studies, and the activation free energies obtained by ABMD vary smoothly with stress and agree with those obtained from conventional MD where direct comparison is possible. These results show that the present methodology combines mechanistic discovery with quantitative accuracy, providing a general route for identifying and evaluating competing defect-activation processes in solids.

Close

PrevNext

View figure in article

• Disclinations & dislocations
• Grain boundaries
• Plasticity
• Molecular dynamics