High performance computing simulations of spall phenomenon in a submicron thick nanocrystalline aluminum

Abstract

The spall failure phenomenon is investigated at nano-scale in a plate-on-plate impact configuration using large-scale molecular dynamics (MD) high performance computing (HPC) simulationson three multi-billion 0.5 μm thick nanocrystalline aluminum (nc-Al) systems, each with an average grain size in the range of 18–100 nm and each under impact velocities in the range of 0.7–1.5 km s−1. Using a material conserving atom section-averaging process, distributions of several mechanical responses were obtained and the extremely transient spall failure phenomena were located in the interaction zones of the stress release waves released from the impacted and free ends of the atom systems. The spall zones’ thicknesses were estimated along with the spall strengths. For the nc-Al atom systems, the spall strength was observed to increase as the impact velocity increased from 0.7 to 1.5 km s−1, but only showed a slight decrease as the grain size increased from 18 to 100 nm. The spall strengths thus estimated from the stress release waves’ interaction zones were found to be conservative when compared to the traditional estimates obtained from the pullback velocity signatures in the free-end velocity evolutions. The MD results were further analyzed using a crystal analysis algorithm and a twin dislocation identification method to obtain the densities of the atomistic microstructures evolving under the interaction of the stress release waves. High-fidelity large-scale HPC simulation results showed that certain dislocation partials strongly influenced the spall response. The Stair-rod type dislocation partials increased by more than a factor of 10 during the interaction of the stress release waves as the spall failure commenced.