Structure and dynamics of metallic glass-forming liquids upon third component additions

Abstract

Even though there exist empirical criteria to predict what constitutes a good glassformer, the process of glass-formation for Bulk metallic glasses (BMGs) is not well understood. This can mainly be attributed to a shortage of composition-dependent data of the fundamental (thermo-)physical melt properties such as viscosity, diffusion, melt structure and phase selection. A good glass-forming ability (GFA) is usually found only in systems with a high number of constituents, which massively complicates the interpretation of systematic, composition-dependent measurements since the interactions of all elements amongst each other need to be considered. Accordingly, the more the composition of a system needs to be changed to achieve a measurable improvement in the GFA, the more difficult it becomes to isolate the origin of the improvement. Considering these challenges, it is most suitable that there exist two systems given by zirconium-copper (Zr-Cu) and nickel-niobium (Ni-Nb) which not only are excellent glass formers with only two components each, but also hugely profit from a minor (only a few at.%) addition of aluminum (Al) for Zr-Cu and sulfur (S) and phosphorus (P), respectively, for Ni-Nb. This thesis aims at providing composition-dependent data of structure and dynamics of Zr-Cu-Al and Ni-Nb-S/P, processed mainly by electrostatic levitation (ESL). Hence, melt viscosity and density measurements are carried out using ESL, the former in combination with the oscillation drop technique. In order to acquire static structure factors, both High-energy X-ray diffraction (HEXRD) at the German electron synchrotron (DESY) institute and neutron diffraction (ND) at the Institut Laue-Langevin (ILL) institute are employed. Due to the existence of a Ni-isotope with negative scattering length, the Bhatia-Thornton partial structure factor SCC(Q) can be determined via ND, which represents the chemical order in the melt. Ni diffusion coefficients are acquired via Quasi-elastic neutron scattering (QENS) at the PSI institute. The melt viscosity measurements of Zr-Cu-Al reveals that the GFA of this particular system can be predicted by the melt viscosity at the liquidus temperature. However, the densities showed that the decrease in atomic mobility cannot be explained by denser atomic packing. This may be attributed to the existence of directional bonds, rendering the usage of a hard sphere model inapt. The static structure factors further demonstrate a correlation between the intensity of the first peak and the dynamics and the GFA. This change in the peak intensities may be linked to the occurrence of chain-like structures as predicted by Zr-Co-Al simulations. Contrary to this, the self-diffusion and viscosity measurements for Ni-Nb-S and NiNb-P exhibit similar values compared to binary Ni-Nb within the boundaries of uncertainty. However, the change in peak intensity of the static structure factors − which cannot be explained by the varied Ni-Nb ratio or a change in the scattering contrast − points to the introduction of more disorder in the liquid structure with the addition of both S and P. Since S- and P-rich phases, which trigger premature crystallization, have been reported to form only beyond critical third component concentrations of 3 at.% S and 2.5 at.% P, the enhanced GFA of Ni-Nb-S and Ni-Nb-P may be attributed to the difference in the liquid and solid structure. In a narrow concentration interval below the above-mentioned critical third component concentration, long-ranged diffusion of the third component atoms may be promoted, leading to a delay in crystallization. The main finding of this thesis can thus be described as follows: In Zr-Cu-Al, the GFA improvement upon Al addition is linked to a slowdown in dynamics. In Ni-Nb-S and Ni-Nb-P, no correlation between the improvement of the GFA upon S and P addition, respectively, and the dynamics is found. Instead, the GFA improvement is linked to the phase selection. This implies that defining a universal rule for identifying a good glass former is more complicated than suggested by the empirical criteria, and that the primary factors influencing the GFA significantly vary depending on the system.