Abstract
The nature of glass transition, whether it is thermodynamic or dynamic, is a major puzzle in science. A similar challenge exists in glass-to-liquid transition, i.e., glass melting. In this thesis, we study surface premelting, surface melting, and bulk melting of colloidal glasses using video microscopy with the single-particle resolution.
The background and experimental systems are reported in Chapters 1 and 2. In Chapter 3, using a thermally attraction-tunable colloidal system, we epitaxially grow stable colloidal glasses in a process similar to physical vapor deposition. We find that in the surface region, the structural and dynamic parameters saturate at different depths, which define a surface liquid layer and an intermediate glassy layer. The power-law growth of both layers and the melting front behaviors at different heating rates and for mono- and multi-layer systems are similar to crystal premelting and melting, suggesting that crystal premelting and melting can be generalized to amorphous solids.
It is commonly believed that ultrastable glass melts through liquid nuclei growth, i.e. nucleation mechanism, while normal glass with low stability melts catastrophically. However, such picture lacks of study and most simulations and experiments about glass melting focused on ultrastable glasses. In Chapter 4, we investigate normal glass melting through a thermally size-tunable colloid system, and directly visualize two melting modes. Using machine learning, we show that the glasses melt through preexisting structural “defects”, following a heterogeneous nucleation
mechanism. As the density of structural “defects” increases, glass melts catastrophically. High structural ”disorder” corresponds to a lower local stability and a higher chance of melting, which connects structure and melting dynamics and explains the different melting modes in normal and ultrastable glasses.