From Rate Effects and Avalanches in Dislocation-Mediated Plasticity to Computer-Assisted High-Entropy Alloy Design

Abstract

Plastic deformation of crystalline solids on the micron-scale is dominated by fluctuations that are directly observable as strain bursts with a broad size distribution. Characterizing the sample-to-sample variation of the plastic response thus poses a challenge and creates exciting prospects of optimizing the mechanical properties of materials by controlling the dislocation dynamics. The research outcomes of this thesis advance the fundamental understanding of crystal plasticity, and present machine-learning based methods to design materials with optimal mechanical properties.

Discrete dislocation simulations present an optimal balance between the accessible length and time scales and the realism of the description, compared to techniques such as molecular dynamics best suited for short length and time scales as well as continuum models of crystal plasticity that are unable to properly describe deformation avalanches. To this end, strain-controlled loading is applied to an ensemble of statistically equivalent samples and the resulting stress-strain curves as well as the statistical properties of strain bursts and the related stress drops as a function of the imposed strain rate are characterized. Based on the different fluctuation statistics and the avalanche length scales observed resulting from the inclusion of spherical precipitates a direction towards bridging to continuum models is discussed.

The need for materials with superior properties has led to the discovery of high-entropy alloys. However, exploring their vast compositional space to find the alloys with optimal properties poses a challenge best handled by computational methods. Molecular dynamics simulations are able to reliably reproduce the interatomic interactions in crystalline solids and thus describe the plastic deformation of microscopic samples. This work combines molecular dynamics with Bayesian optimization, and creates a framework to efficiently map the compositional space of high-entropy alloys by suggesting only a few candidate alloys needing costly validation. Most importantly, the devised algorithm is transferable to any high-entropy alloy system with the possibility to extend to crystals with various microstructural features.