EL-Glass · The energy Landscapes of metastable states of supercooled liquids

The energy Landscapes of metastable states of supercooled liquids

As a cooling liquid approaches the glass transition temperature its viscosity increases very rapidly, eventually becoming indistinguishable from a solid. During this transformation the underlying molecular structure changes only very little, remaining in a disordered configuration. Thus the usual description of a freezing process as a first order transition driven by a competition between energy and entropy is inapplicable, at least in the simplest form. To this day there is no universally agreed explanation for why glasses act like solids. Potential energy landscapes play an important role in guiding theories and experiments. Motion in phase space on a potential energy surface characterised by deep minima and large barriers is a natural explanation for slow dynamics. Such rough landscapes are generally thought to be a precondition for glassy dynamics. They are sufficient to explain large viscosities, but it is not yet agreed upon how this observation translates into the rapid changes of viscosity upon cooling seen in experiments. We intended to focus our study within the framework of metastable states, using state-of-the-art computational techniques to look at both local equilibrium and dynamical properties of liquids trapped in such states. Description of main results: The main results are related to understanding the dynamics and thermodynamics of escape from metastable states. Through my research on the liquid cavity in an environment of frozen particles we have found clear evidence that there exists a thermodynamic length scale associated with the glass transition. Escape from a metastable state only becomes possible if the reconfiguring region is larger than this length scale. This is a principal claim of the random first order transition theory (RFOT), which distinguishes it from most other theoretical descriptions of the glass transition. This work will be published soon, and we expect it to be influential in the community. Additionally, in a published report, we have compared numerous function minimizers and illustrated their sensitivity to certain parameters. We have demonstrated that the L-BFGS algorithm is the fastest algorithm, but that its speed comes at the cost of some lack of precision in delineating basin boundaries. We've shown that for some applications the FIRE algorithm can be a better choice. We have developed a new algorithm for thermodynamic sampling called Superposition Enhanced Nested Sampling, which is significantly more efficient than existing algorithms in certain circumstances. These results will be published soon. The software package PELE, written in collaboration with other members of the group, has been very helpful to colleagues in Cambridge and elsewhere. As an open source, collaborative project, we expect that it will continue to be useful to the energy landscape community. Expected final results and impact: We expect the main results of this project to be the demonstration that there is a thermodynamic (static) length scale related to the glass transition. This length scale has been a key component of some theories (specifically, RFOT theory), and one which distinguishes it from competing theories. As such it has been the subject of intense debate in recent years. This project will likely not end the debate conclusively, but We believe it is the strongest piece of evidence to date that the thermodynamic length scale exists. Theories that discount this effect will either have to refute the evidence, or somehow incorporate it into their description of the glass transition. Either way, it represents a significant step forward in our understanding of the physics behind the glass transition.

Data: CORDIS, © European Union