Large-scale atomistic simulations of cleavage in BCC Fe using machine-learning potential

Suzudo, Tomoaki   ; Ebihara, Kenichi   ; Tsuru, Tomohito   ; Mori, Hideki*

Body-centered-cubic transition metals, such as Fe and W, cleave along the {100} plane. To find out the mechanism of this response, atomistic simulations of curved crack-fronts of bcc Fe were conducted at 0 K using an interatomic potential created by an artificial neural network (ANN) technique. We discovered that dislocations can be emitted from the curved crack fronts along the {110} crack plane, and this phenomenon explains why the cleavage is observed only along the {100} plane. In addition, the cleavage simulations along {100} at the elevated temperature were found to be accompanied by plasticity; namely, they represented more realistic fracture.