Date of Award

Spring 5-15-2020

Author's School

McKelvey School of Engineering

Author's Department

Materials Science & Engineering

Degree Name

Doctor of Philosophy (PhD)

Degree Type



Metallic glasses have drawn significant attention due to their unique properties, such as high strength, excellent elastic energy storage capacity, and versatile processability. However, why some liquids can easily form metallic glasses while others donմ is still unclear. Since metallic glasses are formed when liquids are cooled fast enough to bypass crystallization, we hope to better understand glass formation by investigating the structural evolution and thermophysical properties of the liquids as they are cooled toward the glass transition. Multiple molecular dynamics simulations suggest a crossover temperature for the dynamics near the liquidus temperature, which corresponds to the onset of cooperative structural rearrangements and may be the beginning of the glass transition. In this dissertation, a possible structural signature of this onset of cooperativity is first identified using high-energy synchrotron X-ray scattering studies and viscosity measurements on electrostatically levitated liquids. We also address the practical question of how to predict glass formation from properties of the high temperature liquids. A method to accurately predict the glass transition temperature in metallic glasses from properties of the equilibrium liquids is proposed. It uses the viscosity and the thermal expansion coefficient for the equilibrium liquid. Using the predicted glass transition temperature and a fragility parameter developed from the liquid properties, a new prediction formula is generated, which only uses the liquid properties. While the prediction formula works for most cases, in some cases, it fails. The analysis of these anomalous cases demonstrates that the structural similarity between the liquid and crystal phases plays an important role in the glass formability. This is the first demonstration of this important controlling factor for glass formability. We also used machine learning (Lasso regression and Random Forest) to predict the glass formability and identify important predictors. The identified important predictors are in good agreement with those from the empirical rules. Finally, the evolution of the Cu46Zr54 liquid structure is investigated by elastic neutron scattering (with isotopic substitution) and synchrotron X-ray scattering studies. The experimental results show that the number of Cu-Cu and Zr-Zr atom pairs increases as the temperature decreases, while the number of Cu-Zr atom pairs decreases on cooling. This result disagrees with predictions from previous molecular dynamics studies, suggesting that the potentials used in the molecular dynamics simulations should be reassessed.


English (en)


Kenneth F. Kelton

Committee Members

Katharine Flores, Erik Henriksen, Li Yang, Zohar Nussinov,


Permanent URL: