Hard work, plastic flow: a data-centric approach to dislocation-based plasticity

Metals as structural materials are at the core of our society.

Almost everything we physically interact with includes some form of metal manufactured to specific properties and formed into a desired shape.

Consequently, the understanding and design of the balance between ductility and strength of metals are one of the primary disciplines of materials science. On a fundamental level, this is the description of crystalline line defects called dislocations.

At the atomic scale, the current understanding of dislocations is often on the level of individual dislocation properties.

At the component scale, collective behavior is commonly formulated in continuum variables with the drawback of limited applicability over a wide range of possible scenarios.

Our current understanding still shows a gap in how individual dislocation properties translate into their collective behavior. To address this long-standing question, I propose a data-centric approach.

First, a comprehensive dataset of dislocation ensemble trajectories for various loading and initial conditions is created using discrete dislocation dynamics as well as molecular dynamics simulations and iteratively extended.

The trajectories are subsequently analyzed with tools borrowed from graph theory and time-series analysis to capture the network character of dislocation structures.

Subsequently, a novel class of plasticity models is developed: instead of human-derived state variables, I will `let the data speak for itself’ to bridge the gap between individual and collective dislocation behavior.

The project solves two timely challenges in materials science: One is the described gap, and the other is a demonstration of effective research data management of complex materials data providing solutions to data generation, storage, accessibility, data fusion, reuse, and analysis using the example of dislocation trajectories.