Efficient and Scalable Implementation of Hartree-Fock Exchange within the Ultrasoft Pseudopotential Branch of the CPMD code

Applicant

Prof. Dr. Bernd Meyer
Interdisciplinary Center for Molecular Materials (ICMM) and Computer-Chemistry-Center
Department of Chemistry and Pharmacy
Friedrich-Alexander-Universität Erlangen-Nürnberg

Project Overview

CPMD (Car-Parrinello Molecular Dynamics) is an ab initio molecular dynamics (AIMD) code. Atoms are propagated via Newton’s equation of motion and the forces acting on the atoms are calculated from first-principles by solving approximately the quantum-mechanical electronic structure problem using density-functional theory (DFT).

DFT has become an immensely popular method in chemistry, solid-state physics, and materials science for calculating parameter-free the properties of atomic assemblies, i.e. molecules, nanoparticles, liquids, and solids. Only in 2017, more than 30.000 scientific papers appeared which are based to a larger or smaller extent on DFT studies. It has been estimated that today about 1/6 of the worldwide high-performance computing (HPC) resources are devoted to DFT calculations, with a share of more than 30%, for example, at the UK facilities HECToR and ARCHER. It is very safe to assume that a very large fraction of these computational resources go to AIMD simulations, considering that for each time step a DFT evaluation has to be performed and that a typical AIMD run requires several 100.000 of these steps and not only a few 100 as in structural DFT geometry optimizations.