Public defence in Engineering Physics, M.Sc. (Tech) Henri Salmenjoki

Most metals have crystalline structure where yielding and irreversible (plastic) deformation are caused by the motion of line-like crystal defects, dislocations. Because yielding is in many cases harmful, understanding of dislocation phenomena is highly pursued to design and improve crystalline material properties. For example to increase the yield stress of a crystal, one common procedure is alloying where precipitates are added to the crystal to block dislocation motion.

However, when observing plastic deformation on micronscale, for instance in micropillars, the deformation occurs in bursts of various sizes. The bursts are dislocation avalanches arising from the collective motion of dislocations. These avalanches are similar to e.g. earthquakes as their size and duration follow power-law statistics. In macroscopic crystal, the amount of dislocations is high which leads to seemingly smooth plastic deformation as the different dislocation structures average out the bursts. But in micronscale crystals the avalanches dominate the mechanical properties. For instance in compression tests of micropillars, the measured stress-strain curve consists of discrete strain jumps which are unique for every sample. Thus on micronscale, mechanical properties can vary with the initial dislocation structure of the sample which leads to stochastic aspects in crystal plasticity.

In this dissertation, we study dislocation phenomena related to yielding with discrete dislocation dynamics simulations. We take two approaches: Firstly, we focus on the seemingly stochastic nature of crystal plasticity. We use supervised machine learning to predict the stress-strain response of simulated dislocation structures. The used method, i.e. neural network, succeeds in learning the overall shape of the stress-strain curve well but fails to reproduce single avalanches. We continue to examine the role of the avalanches on the stress response predictability by finding correlations between dislocation avalanches that reveal that the avalanches complicate the predicting by nature. Secondly, we add precipitates to our simulated systems. This way we are able to impact the yield stress and also the ensuing dynamics inside the system. By varying the density and strength of the precipitates, we extract two distinct phases of dislocation dynamics where either the dislocation-dislocation or dislocation-precipitate interactions dominate.

Opponent is Professor Michael Zaiser, Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany

Custos is Professor Mikko Alava, Aalto University School of Science, Department of Applied Physics

Contact details of the doctoral student: [email protected], 0505143552

The public defence will be organised in Otaniemi (Otakaari 1, lecture hall M1).

The dissertation is publicly displayed 10 days before the defence in the publication archive Aaltodoc of Aalto University

Electronic thesis