Fracture simulations using large-scale molecular dynamics

We report on recent molecular-dynamics (MD) fracture simulations of mode-I tensile loading at high strain rates. Because cracks emit sound waves, previous simulations became unreliable beyond one sound traversal time. Using massively parallel MD, we show how to eliminate unwanted boundary effects and study unimpeded crack propagation mechanisms. In order to represent tensile stress conditions near the crack tip, we employ uniaxial, homogeneously expanding periodic boundary conditions, examining the effects of strain rate, temperature, and interaction potential. Because our samples are sufficiently large, we see dislocations being emitted from the crack tip at nearly the shear-wave sound speed ${\mathit{c}}_{\mathit{s}}$. As they move many lattice spacings away from the crack, they slow down, finally moving at about 2/3${\mathit{c}}_{\mathit{s}}$. Each time dislocations are emitted, the crack tip ``fishtails,'' and at sufficiently high strain, the crack can fork; dislocations can climb and become nucleation sites for additional microcracks. We find that we can suppress ductile behavior by including viscous damping in the equations of motion, thereby demonstrating a transition to brittle crack propagation as static, zero-strain-rate conditions are approached. Finally, we show that, by altering only the attractive tail of the pair potential, we can change a ductile material into a brittle one. Under dynamic crack propagation, the distinction between ductile and brittle behavior is blurred: in brittle materials, dislocations are asymptotically bound to the crack tip, while in ductile materials, they can escape.