Fracture and plasticity are the dominant failure processes underlying many materials reliability issues. They are also some of the most challenging multiscale modelling problems, requiring both an accurate description of chemical processes occurring near crack tips or dislocation cores and the inclusion of much larger model systems. These requirements can be met simultaneously by combining quantum mechanical descriptions of chemically active regions with classical atomistic models that capture the long-range elastic behaviour of the surrounding crystal matrix, using approaches such as the `Learn on the Fly’ (LOTF) scheme.
I will review recent methodological advances that: (i) improve the efficiency of the scheme using machine learning and identify remaining limitations with uncertainty quantification approaches; (ii) extend to rare events by computing minimum energy paths in multiscale sytems; (iii) extend to metallic systems. Together, these enable processes such as dislocation motion in the presence of impurities to be described.
Warwick Centre for Predictive Modelling – School of Engineering – University of Warwick