Quantum-Mechanical Molecular Dynamics for Distributed Computing and AI-hardware

Anders Nikklasson
Los Alamos National Laboratory

We present recent developments of graph-based electronic structure theory for distributed quantum-mechanical Born-Oppenheimer molecular dynamics simulations including tens-of-thousands of atoms, and how low-precision AI-hardware such as Tensor cores can be used to calculate the electronic structure with mixed-precision algorithms, while still providing accurate forces for a robust long-term energy stability.

Presentation (PDF File)

Back to Workshop I: Increasing the Length, Time, and Accuracy of Materials Modeling Using Exascale Computing