In this thesis we study super-cooled liquids through simulations of hard interacting particles: hard disks in two dimensions, and hard spheres in three dimensions. These systems provide simple models of glass-forming liquids in general and are good models for colloidal glass-forming systems specifically. Whether static structure exists, and the nature of any collective motion are two key questions about super-cooled liquids. Both involve correlation, and this motivates us to use information theory: we use Shannon entropy and mutual information to provide general, unbiased measurements of disorder and correlation. We use mutual information to define an order agnostic order parameter in ther hard disk system and show that static correlations are short-ranged in these simulations. Many of the simulations described in this thesis are performed in the isoconfigurational ensemble. This technique gives access to probability distributions of particle trajectories conditioned on the initial system configuration. By using information theory to investigate these distributions we are able to answer questions about predictability in quantitative terms. We measure the predictability of the dynamics of our hard sphere system; then, by classifying the structure of the initial configuration, we search for the salient features that have predictive power over the dynamics. We find that classifying the initial configuration in terms of geometrical motifs and local volume fraction gives some predictive power. By using information theory to quantify dynamic correlations between particles we show that this predictivity exists because the initial configuration specifies the positions of early-relaxing particles and slow-relaxing relatively stable regions. Most particles belong to neither of these regions and are presumably responsible for the unpredictable behaviour of the system. The final results chapter in this thesis measures changes in shape of the dynamical regions in our model systems using fractal dimension.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:683695 |
| Date | January 2014 |
| Creators | Dunleavy, Andrew J. |
| Publisher | University of Bristol |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0022 seconds