Metallic glass matrix composites have enormous potential stemming from the interplay between crystalline and amorphous phases. This work models such a composite using shear transformation zone dynamics (a modified kinetic Monte Carlo method) for the amorphous phase, and a local Taylor dislocation model for the crystalline phase. An N-factorial experiment using the model is presented examining the effects of crystalline volume fraction, microstructure length scale, and yield stress of the crystalline phase. Each replicate is analyzed for maximum stress, maximum strain, strain energy dissipation, strain localization, and strain partitioning between phases. Regression analysis is used to identify statistically-significant trends in the data. The experiment shows that strain delocalization and the consequent ductility are facilitated by a crystalline phase with a substantially lower yield stress than that of the amorphous matrix. It also shows that increasing crystalline volume fraction alone is insufficient to promote strain delocalization in the case of a crystalline phase with high relative yield stress, and that a lower yield stress for the crystalline phase implies lower maximum stresses supported by the composite. Therefore designers must balance the need for ductility and delocalization against the composite yield stress by finding an optimal combination of volume fraction and crystalline mechanical properties. This work provides continuous functional forms for the relationships between these properties to aid in that optimization process.

Metallic glass matrix composites (MGMCs) have been developed to improve upon the ductility of monolithic metallic glass. These composites utilize a secondary crystalline phase that is grown into an amorphous matrix as isolated dendritic trees. This work seeks to understand the mechanisms underlying strain delocalization in MGMCs in order to better direct efforts for continual progress in this class of material. A mesoscale modelling technique based on shear transformation zone (STZ) dynamics is used to do so. STZ dynamics is a coarse grained technique that can provide insight into the microscopic processes that control macroscopic behavior, but which can be difficult to resolve experimentally. A combined simulated-experimental approach to extract the individual material properties of the amorphous and crystalline phases is presented. Numerically, STZ dynamics is used to simulate nanoindentation of the crystalline and amorphous phases respectively. The indented phases are modelled as discs with varying thickness embedded in the other phase. Indentation depths are held constant. Experimentally, nanoindentation is carried out on DH2 and DH3 MGMC composites under varying loads at Stony Brook University (SBU). Specimens are cross-sectioned and using scanning electron microscopy, indentation sites are chosen so that the indenter targets individual phases. For both experimental and simulated nanoindentation, hardness and modulus values are calculated from the load-displacement data. The experimental and simulated values are normalized and compared. Good agreement between results suggests accurate characterization of the individual phases at low loads on both DH2 and DH3 composites. Length scales at which indentations begin sampling outside the intended phase are presented. Work is then presented on simulated uniaxial tensile loading of MGMCs. Dendritic microstructural sizes are varied and shear banding characteristics are measured. A competition of shear band nucleation and propagation rates that previously had only been seen in monolithic metallic glasses under certain loading conditions is found to exist in MGMCs as well. The stages of shear banding in MGMCs are presented and the influence of dendrites on shear band nucleation and propagation are discussed. It is proposed that the introduction of dendrites into the amorphous matrix work to inhibit shear band propagation and encourage shear band nucleation to delocalize strain in MGMCs. In particular, it was found that smaller dendrite sizes and spacings are better at doing so.