Dislocation–grain boundary interactions in Ta: numerical, molecular dynamics, and machine learning approaches

The motivation of this work was to find the appropriate molecular dynamics (MD) and slip transmission parameters of dislocation–grain boundary (GB) interaction in tantalum that correlate with the stress required for the grain boundary to deform. GBs were modeled using [ 1 1 2], [ 1 10], and [111] as rotation axes and rotation angle between 0° and 90°. Dislocation on either {110} or {112} slip planes was simulated to interact with various GB configurations. Drop in shear stress, drop in potential energy, critical distance between dislocation and GB, and critical shear stress for dislocation absorption by the GB were the parameters calculated from MD simulations of dislocation–GB interactions. Machine learning models eXtreme Gradient Boosting and SHapley Additive exPlanations (SHAP) were used to find the correlation between the various parameters and yield stress of the GB configurations. Machine learning results showed that the MD parameters—critical distance between the dislocation and GB, drop in shear stress; and slip transmission parameter— m^' have a stronger correlation with yield stress. The SHAP results sorted the prominent slip plane and rotation axis affecting the yield stress. The configurations with dislocation on {112} slip plane, and configurations with [111] rotation axis were difficult to deform (higher yield stress of GB) than {110} slip plane and [ 1 1 2] and [ 1 10] rotation axes configurations. Graphical abstract