Metallic glasses are a class of material that have been developed to have many unique and useful properties, but have yet to see major application. The amorphous structure of these metals makes them great candidates for applications that need a material that has high strength and can be formed as easily as plastics. However, as a result of their metastable structure, alloys only form in a limited number of systems. This difficulty in locating feasible alloys frustrates designing alloys for new applications. In this thesis, we propose a series of computational tools based on classical molecular dynamics to accelerate the alloy design process.
The first tool presented is a method to rapidly produce accurate alloy interatomic potentials, which are computationally-efficient approximations of bonding in metals. Previously, in order to run an atomistic simulation of a metal alloy, one either had need to resort to slow ab initio calculations, rely on unreliable mixing rules to generate the interatomic potential, or spend a large amount of effort creating a alloy potential. To overcome this, we developed a method to create many-component alloy potentials from a library of binaries created by combining previously-developed elemental potentials and fitting the alloy terms. As we will show, potentials created using this method are superior to those generated using mixing rules and just as accurate in simulating metallic glasses as those made for specific alloys using more time-intensive methods.
Additionally, we have developed a novel tool for predicting how easily an alloy will form a metallic glass. Our method uses a combination of parameters that describe the kinetics and energetic stability of the supercooled liquid. To assess the kinetics, we calculate the fragility of the liquid, which describes how viscosity changes as a function of temperature. To evaluate the relative energetic stability, we determine the fraction of efficiently-packed clusters, which have been argued to be the source of stability in supercooled liquids. First, we assessed the correlation of each of these properties against glass-forming ability in a well-studied system and found that fragility is a better predictor for glass formation. Using that knowledge, we then develop a figure of merit, which was successfully used to identify the best glass forming compositions in the Cu-Zr binary and Cu-Zr-Ti ternary systems.
In the third part of this work, we demonstrate a tool for locating alloys with an optimal set of physical properties. To accomplish this, we have developed a tool that uses optimization algorithms to virtually design an alloy without experimental input, which required our method for rapidly generating interatomic potentials. Using this tool, we identified Ni66.4Nb21.4Al12.2 as the ternary metallic glass with the highest elastic modulus in the Ni-Nb-Al-Zr-Ti-Ta system. Also, we predict Ti70Ni30 to be the metallic glass with the a high stiffness-to-density in the Ni-Cu-Zr-Ti quaternary with the locally best glass-forming ability, as found using our computational tool. Pending experimental evaluation, this method is a possible way to enable rapid and inexpensive design of novel metallic glasses.