From cage-jump motion to macroscopic diffusion in supercooled liquids

The evaluation of the long term stability of a material requires the estimation of its long-time dynamics. For amorphous materials such as structural glasses, it has proven difficult to predict the long-time dynamics starting from static measurements. Here we consider how long one needs to monitor the dynamics of a structural glass to predict its long--time features. We present a detailed characterization of the statistical features of the single-particle intermittent motion of structural glasses, and show that single--particle jumps are the irreversible events leading to the relaxation of the system. This allows to evaluate the diffusion constant on the time--scale of the jump duration, which is small and temperature independent, well before the system enters the diffusive regime. The prediction is obtained by analyzing the particle trajectories via a parameter-free algorithm.