Sommaire:

Understanding the mechanisms that lead to glass formation is one of the open problems for the condensed matter research. Numerous questions remain unanswered, because the tremendous increase of relaxation times during the cooling process prevents the exploration of equilibrium properties of supercooled liquids at very low temperature. Computer simulations of glass-forming liquids are nowadays able to reach equilibrium at temperatures comparable to the Mode-Coupling crossover temperature, which is well above the experimental glass transition temperature. As a consequence, simulations lag eight orders of magnitude behind experiments in terms of equilibration times. Progress to close this gap has been slow, and stems mostly from hardware improvements.

In this thesis we make an important step to close this gap. We combine the use of a Monte Carlo algorithm, known as the swap algorithm, with the design of novel glass-forming models. We systematically test numerous models using both discrete mixtures and polydisperse systems. We discuss the role that polydispersity and particle softness play in avoiding crystallization and in efficiently reaching previously unexplored regimes. We study the dynamical processes taking place during swap Monte Carlo simulations. We demonstrate that in some cases our technique is able to produce thermalized configurations at temperatures inaccessible even by experiments.

In this newly accessible regime, we investigate some open questions concerning the glass transition. We show that a hard sphere fluid can be equilibrated at, and even beyond, the jamming packing fraction. We measure the configurational entropy in extremely supercooled liquid, finding a strong dimensional dependence that supports, on the one hand, the existence of an ideal glass transition at a finite temperature in three dimensions and, on the other hand, its absence in two dimensions. We detect the increase of amorphous order quantified through a static point-to-set length throughout the glass formation. We measure the critical exponents introduced in the mean-field theory of glasses much closer to the supposed ideal glass transition. Finally, we reveal the absence of a sharp geometric transition in the potential energy landscape across the Mode-Coupling crossover.

The models and the algorithms developed in this thesis shift the computational studies of glass-forming liquids to an entirely new territory, which should help to close the gap between theory and experiments, and get us closer to solve the long-standing problem of the glass transition.

Pour plus d'informations, merci de contacter Ninarello A. S.