Modeling and simulation of plasticity from dislocation density and deformation in electrochemical actuation

Abstract

In materials science, the study of microstructures demands efforts from modeling and simulations even though it often begins from experimental observation. In this thesis, two kinds of microstructures are investigated by the computational approach, one is dislocation microstructure, and another is the free surface in nano-porous metals. For dislocation microstructures, a dislocation-density dynamics framework for modeling dislocations at an “intensive” resolution scale finer than the dislocation core is established. The previous continuum dislocation models were incapable of describing the Peierls stress, due to the invariance of the misfit energy and a lack of means to trigger configurational changes in the dislocation core as the dislocation moves. Here, the inter-dislocation elastic interaction is accounted for via Mura’s formula, and the interaction within the dislocation core is modeled by introducing a phenomenological formalism, leading to the expected Peierls stress and a stable width of the dislocation as it travels. This framework is implemented numerically with a divergence-preserving finite-volume method for curved dislocations gliding on 2D slip planes. Simulation examples of various dislocation mechanisms exhibit excellent preservation of continuity of dislocation densities during their evolution, while the detailed core structures and Peierls stress are elucidated. Then, the model based on convolutional neural networks (ConvNets) is built to understand the dislocation microstructures from the density of geometrically necessary dislocations (GND). The flow strength of a crystal is determined by the total dislocation density, which is difficult to measure accurately in the case of high dislocation contents. On the other hand, the density of GND is related to crystal rotations which can be conveniently measured by electron diffraction experiments or calculated via simulations. In the model, the ConvNets are applied to extract the hidden information in the GND distribution maps to estimate the total dislocation density. The pre-trained ConvNets demonstrates their ability to predict the distribution of total dislocation density from a GND density map. Compared with previous methods involving extra efforts to track individual dislocations or other quantities, the present post-processing method is quick and convenient to apply. Finally, a multi-scale, multi-field simulation approach is used to model the electrochemical actuation behavior of nano-porous Ni in water environment upon potential loading. Specifically, molecular dynamics simulations with reactive force-field potentials and a modified charge-equilibrium (QEq) method are used to calculate the surface stress built up in Ni(100) surface with water electrolyte due to a voltage loading across the interface. The calculated surface stress is then used in a mesoscale finite-element (FE) model to compute the actuating stress set up in a single hexagonal unit cell of the Ni nanohoneycomb structure. The single-unit actuating stress is eventually used in a continuum FE model at a larger scale, to calculate the bending of an entire bilayered cantilever which replicates experimental conditions. The actuation deflection of the bilayered nanohoneycomb nickel is predicted to be 41.4 um at 0.43 V vs. the point of zero charge (PZC), which is in excellent agreement with the experimental value of 45-62 um and proves the ability of present approach.