Browsing by study line "Beräkningsmaterialfysik"
Now showing items 21-22 of 22
-
(2021)The plasma-facing materials of future fusion reactors are exposed to high doses of radiation. The characterization of the radiation damage is an essential part in the study of fusion relevant materi- als. Electron microscopy is one of the most important tools used for characterization of radiation damage, as it provides direct observations of the microstructure of materials. However, the char- acterization of defects from electron microscope images remains difficult. Simulated images can be used to bridge the gap between experimental results and models. In this thesis, scanning transmission electron microscope (STEM) images of radiation damage were simulated. Molecular dynamics simulations were employed in order to create defects in tungsten. STEM images were simulated based on the created systems using the multislice method. A data- base of images of h001i dislocation loops and defects produced from collision cascade simulations was generated. The simulated images provide insight into the observed contrast of the defect structures. Differences in the image contrast between vacancy and interstitial h001i dislocation loops were reported. In addition to this, the results were compared against experimental images and used in identification of a dislocation loop. The simulated images demonstrate that it is feasible to simulate STEM images of radiation damage produced from collision cascade simulations.
-
(2023)High-entropy alloys (HEAs), esteemed for their exceptional resistance to radiation damage, carry considerable potential for deployment within fusion reactors. Nonetheless, due to their compositional complexity, comprehending the diffusion behaviour in these multifaceted alloys continues to be a daunting task. This thesis proposes a novel approach to modelling vacancy diffusion in body-centred cubic (BCC) HEAs, particularly Mo-Nb-Ta-V-W. The methodology involves the tactical application of collective variable-driven hyperdynamics (CVHD) to procure data for training a Gaussian process regression (GPR) and feed-forward neural network (FNN) model. The trained FNN model is subsequently employed within kinetic Monte Carlo (KMC) simulations for accurately predicting jump rates, whereas the GPR model is used to elucidate experimental findings related to the behaviour of vacancies in \mbox{Mo-Nb-Ta-V-W}. The robustness of the FNN model is manifested by its capacity to generalise to unseen data, whilst the efficacy of the overall method is corroborated by Monte Carlo molecular dynamics (MCMD) simulations. The CVHD methodology, uniquely capable of functioning at finite temperatures, can capture the entropic contribution to the free energy and model kinematics explicitly. This singular ability facilitates a more comprehensive understanding of the system's behaviour under authentic conditions. The findings presented in this thesis signify a considerable stride forward in the study of HEAs, providing a robust framework for the design of advanced materials. These results underscore the potential of the CVHD-trained FNN-KMC methodology in exploring complex environments, thereby establishing a firm foundation for future investigations and emphasising the need for its continued refinement and augmentation.
Now showing items 21-22 of 22