The Jamming Transition: Everything you (possibly never) wanted to know about packing spheres

• Rafael Diaz Hernandez Rojas

Packing spheres, i.e. generating configurations where spheres do not overlap with each other, is a problem of interest in mathematics, information theory and, naturally, physics. In this talk I will focus on a special class of them: amorphous packings (configurations where spheres are placed "randomly") and their connection with the so called "jamming transition". This transition has proven to be very relevant because it informs our understanding of glasses, rigidity of amorphous solids, granular matter and even some types of neural networks. Finally, I will present a new numerical algorithm developed to generate jammed packings of (infinitely) hard spheres.

Dicussion

Accessing large-scale molecular simulations through machine learning

• Flavio Giuliani

Molecular dynamics (MD) simulations are among the most used numerical methods for studying a system of many interacting atoms. However, simulations with quantum accuracy (ab-initio) are currently limited to a small number of atoms or short time scales, due to the bottleneck of computing the electronic energy and forces at each step. Recent machine learning methods are able to infer the forces with high precision but at a significantly lower computational cost, allowing for large-scale MD at ab-initio level.

Discussion