Collective phenomena in dislocation plasticity: jamming, avalanches and yielding

When crystalline materials are externally loaded, the irreversible changes in the shape, i.e. plastic deformation, results from the motion of dislocations that are line-like defects in the crystal lattice. Thus, the mechanical properties of crystalline materials, including most metals, are dependent on both the interactions between dislocations with other dislocations and dislocations with other types of crystal defects. Observing crystal plasticity on micron-scale reveals that it proceeds in discrete intermittent events that are avalanches of collective dislocation motion. The size and duration of these avalanches follow power-law distributions which is typical for critical phenomena. As the avalanches are dependent on the features of the unique, initial dislocation structure, micron-scale samples exhibit fluctuations in the stress-strain response and mechanical properties. This dissertation consists of two parts where we study crystal plasticity with 2D and 3D discrete dislocation dynamics simulations. In the first part, we focus on the seemingly stochastic, avalanche-dominated stress-strain curves of dislocation systems. In Publication I, we use supervised machine learning methods to predict 2D single system stress-response from the initial state of the system. And in Publication II, we find correlations between the subsequent avalanches of the stress-strain curve in both 2D and 3D systems. The second part considers 3D dislocation simulations with disorder in the form of precipitates. As precipitates block dislocation motion, they are commonly added to increase yield strength of the crystal. In Publication III, we study the effect of precipitate density and strength on the avalanche distributions and yield stress. We find a phase transition between low precipitate density systems, where dislocation-dislocation interaction dominates and the systems exhibit extended criticality, and high precipitate density systems, where dislocations pin to the defects and the systems possess a distinct critical point. Finally in Publication IV, we use unsupervised machine learning to locate the phase transition by using solely the dislocation structures extracted from systems with varying precipitate density.