Fig. 2: Stress-velocity predictions.
From: Learning dislocation dynamics mobility laws from large-scale MD simulations
a Velocities of isolated edge and screw dislocations predicted using a GNN mobility function trained on DDD trajectories. The learned GNN model accurately reproduces linear mobilities employed in DDD simulation. b Velocities of isolated edge and screw dislocations predicted using a GNN mobility function trained on large-scale MD trajectories of BCC tungsten deformed under [001] tension and compression. The trained GNN mobility function predicts a non-linear force-velocity relation for the edge dislocation and a strong asymmetry in the velocities of screw dislocations sheared in the twinning (T) and anti-twinning (AT) direction.
