Tensor cores represent a new form of hardware acceleration specifically designed for deep neural network calculations in artificial intelligence applications. Tensor cores provide extraordinary speed and energy efficiency, but with the caveat that they were designed for matrix-matrix multiplications using only low-precision floating point operations. We demonstrate how tensor cores can be applied, with high efficiency, to the challenging and numerically sensitive problem of quantum-based Born-Oppenheimer molecular dynamics. The interatomic forces are calculated on-the-fly from an electronic structure that is obtained from a generalized deep neural network and allows for high performance in excess of 100 Tflops on a single Nvidia A100 GPU. Stable molecular dynamics trajectories are generated using the framework of extended Lagrangian Born-Oppenheimer molecular dynamics, which combines computational efficiency with long-term stability, even when using approximate charge relaxations and force evaluations. A canonical ensemble simulation scheme is also presented, where the additional numerical noise in the calculated forces due to the low precision is absorbed into a Langevin-like dynamics.

Quantum-based molecular dynamics simulations using tensor cores

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.