Abstract: In this talk I will explain how atomic-scale modelling can be used to make accurate predictions of materials failure processes. Prominent examples include brittle fracture of ceramics used in photovoltaic cells and plastic deformation in Ni-based superalloys used in jet engine turbine blades. These challenging multiscale problems can only be tackled by combining accurate quantum mechanical simulations of breaking bonds with larger scale approximate models including millions of atoms.
I will show examples from my group’s work where predictions have been tested quantitatively in experiments carried out by our collaborators, ranging in speed from very fast catastrophic brittle fracture where cracks travel at speeds of km/s, close to the speed of sound in the material, down to very slow “stress corrosion” cracking where cracks travel at mm/s or even slower under the concerted action of stress and chemistry. I will also give an overview of current work bringing tools from machine learning and data science to bear on materials science problems.