Predictive Crystal Plasticity Modeling of Face-Centered Cubic Metals using Multiscale Computational Techniques
Restricted (Penn State Only)
- Author:
- Shimanek, John
- Graduate Program:
- Materials Science and Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 14, 2024
- Committee Members:
- John Mauro, Program Head/Chair
Zi-Kui Liu, Co-Chair & Dissertation Advisor
Reuben Kraft, Outside Unit & Field Member
Allison Beese, Co-Chair & Dissertation Advisor
Wesley Reinhart, Major Field Member - Keywords:
- crystal plasticity
density functional theory
dislocation dynamics
finite element method
graph theory - Abstract:
- Metals represent a useful material class due to their typical high strength and toughness, which originate in both their capacity for and resistance to deformation. Over a wide range of temperatures, plasticity within crystalline metals is mediated by the motion of dislocations. The atomic origin of these two-dimensional lattice defects combined with their critical role in the manifestation of bulk deformation behavior motivates a multiscale perspective on their modeling. With this in mind, the present work takes several viewpoints on the phenomenon of dislocation-mediated plasticity. Starting from first-principles calculations based on density functional theory, the ideal shear strength is probed for pure Ni and its dilute alloys. Trends in the calculated properties of each alloy with respect to atomic properties of the alloying element indicated the importance of electronic properties at the shear plane during slip. The connection to higher length scales was made through application of a Peierls-Nabarro framework to estimate parameters for a crystal plasticity model. Comparison to literature tensile deformation data suggested the importance of screw-type dislocation segments in the strain hardening response. Experimental data, including polycrystal and single crystal macroscopic tensile stress-strain curves, is often used to parameterize crystal plasticity models that aim to bridge length scales from the bulk to the scale of individual grains. However, considerations of slight misorientations away from nominal loading orientations can become important in crystal plasticity models, especially for direct loading along high-symmetry orientations and when using models with high latent hardening effects. Since the assumed orientation of experimental data can so affect the resulting plasticity behavior, a flexible accounting of orientations was included in a Bayesian optimization framework for parameterizing hardening models against experimental data from the literature. The common problem of non-unique parameter sets, which leads to uninterpretable model parameters, was reduced by the simultaneous comparison to multiple datasets. Using the simultaneous optimization scheme, it was shown that increasing the scope of comparison data can turn an acceptable parameter set for the chosen hardening models from being non-unique to non-existent, at least according to the framework's ability to search high dimensional parameter space. Finally, to connect macroscale hardening behavior to microscopic mechanisms, discrete dislocation dynamics was used to follow the effects of elastic anisotropy, as manifested in the lattice friction stress of dislocations across their character angles, on the initial hardening rate of dislocation networks. Increasing the friction stress for screw components, with a smooth interpolation to edge components, resulted in a higher initial hardening rate that was not explained by any decrease in the segment lengths of the harder dislocation network. Furthermore, the mean free path over the full range of dislocation characters showed only a local suppression of plasticity for nearly screw-type segments, suggesting a slight difference in the motion of segments resulted in the significant hardening rate increase. The harder networks were shown to better retain dislocation density upon initial straining, which otherwise unlocked more weakly bound configurations. Using structured describing the connections between segments, the applications of several graph analysis methods showed that the harder networks tended to better maintain connectivity during deformation. The analysis demonstrates that a localized effect of atomic properties manifests as a stark difference in bulk hardening due to its impact on the dislocation network evolution over the course of deformation. The structured representation of dislocation network simulation data in the present work enables future investigations into the quantitative characterization of dislocation networks and their evolution, which remains a key step toward the construction of predictive models and a full multiscale understanding of dislocation-mediated crystal plasticity.