Glass transition-the ubiquitous phenomenon in nature-is generally regarded as the process in which a viscous liquid circumvents crystallization but evolves continuously into a disordered solid state directly during fast cooling. It is regarded as one of the most key scientific problems in condensed matter physics and materials sciences all the time. In glass transition, an evident change of structures is still not observed in the experiment although dynamics in the system has changed a lot. As a new alloy, glass transition in a metallic glass also has similar features. Furthermore, it is always used for a mode system to study glass transition because of its structural feature, i.e., disorderly close packing. And understanding the mechanism of glass transition is also very significant to gain in-depth understanding of properties and structural design of materials. From thermodynamic and dynamic perspectives, the thesis focused on two key scientific issues on the structural change and dynamic evolution in glass transition by using molecular dynamics simulations, i.e., the origin of excess entropy and the change of microscopic structures in thermal glass transition and the correlation between dynamic heterogeneity and glass transition, and studied glass transition in metallic glasses systematically, then proposed a statistical model of entropic scenario on glass transition based on short-range structure and dynamic heterogeneity in metallic glass.

Initially, we reveal the origin of excess entropy and the change of microscopic structures in thermal glass transition. Through the computation of thermodynamic properties and evolution of phononic features in glass transition via molecular dynamics simulations, total entropy, vibrational entropy and a trend of configurational entropy at the entire temperature range are obtained. Except the temperature range near 0 K, the change of entropy in glass transition is confirmed to be originated from mostly configurational entropy while the vibrational entropy is trivial. The findings are in agreement with recent inelastic neutron scattering experiments and provide a compelling evidence for the Adam–Gibbs entropic scenario on glass transition. Besides, by performing statistics on diversity of local Voronoi polyhedra in glass transition in virtue of Shannon information entropy, it is found that Shannon entropy experiences striking variation during glass transition, which is in accordance with the sudden increase in configurational entropy in the system and lays a solid atomic structure foundation and provide a specific explanation for the Adam–Gibbs entropic scenario.

Afterwards, the correlation between dynamic heterogeneity and glass transition is studied by us. Based on the activation-relaxation technique (ART) , spectrum of activation barrier in glass-forming system with different thermodynamic states can be accquired. It is noticed that the probability distribution function can be decomposed into two distinct modes: cascade relaxation and localized relaxation. As temperature gradually increases, the mode dominating dynamics in the system transforms from localized relaxation to cascade relaxation. Further study on the spatial distribution of activation barriers shows that low-barrier regions are highly localized. After glass-liquid transition, low-barrier regions are extended. The high mobile regimes tend to percolate through the whole sample and lead to the reduction of dynamic heterogeneity in the end. What's more, Shannon entropy and statictical diversity of barriers are defined and decrease a lot after glass-liquid transition, which corresponds to the change of stretching exponent in Kohlrausch-Williams-Watts (KWW) equation as temperature varies. They can both effectively characterize the level of dynamic heterogeneity and are regarded as quantitative signatures for glass transition.

The above research innovatively introduces the concept of Shannon information entropy and statistically characterizes the short-range structure and dynamics of glass. It provides a confirmation of atomic credibility for Adam-Gibbs entropic scenario on glass transition since 1958 and clarifies its structural origin. Furthermore, based on Shannon information entropy, we quantitatively characterize the degree of dynamic heterogeneity by using statistical diversity of activation energy barriers and correlate it with glass transition. From thermodynamic and dynamic perspectives, our study focuses on the structure of glass and then provides a novel physical view for glass transition.