Density functional theory (DFT) has become a standard tool for ab-initio simulations for a wide range of applications. While the intrinsic cubic scaling of DFT was for a long time limiting the accessible system size to some hundred atoms, the recent progress with respect to linear scaling DFT methods has allowed to tackle problems that are larger by many orders of magnitudes. However, as these linear scaling methods were developed for insulators, they cannot, in general, be straightforwardly applied to metals, as a finite (electronic) temperature is needed to ensure locality of the density matrix. In this paper we show that, once finite electronic temperature is employed, the linear scaling version of the BigDFT code is able to exploit this locality to provide a computational treatment that scales linearly with respect to the number of atoms of a metallic system. We provide prototype examples based on bulk Tungsten, which plays a key role in finding safe and long-lasting materials for Fusion Reactors; however we do not expect any major obstacles in extending this work to cover other metals. We believe that such an approach might help in opening the path towards novel approaches for investigating the electronic structure of such materials, in particular when large supercells are required.

We acknowledge valuable discussions with María José Caturla and Chu-Chun Fu. S.M. acknowledges support from the MaX project, which has received funding from the European Union’s Horizon 2020 Research and Innovation Programme under Grant agreements 676598. M.A. acknowledges support from the Novartis Universität Basel Excellence Scholarship for Life Sciences and the Swiss National Science Foundation (P300P2-158407, P300P2-174475). We gratefully acknowledge the computing resources on Marconi-Fusion under the EUROfusion project BigDFT4F, from the Swiss National Supercomputing Center in Lugano (project s700), the Extreme Science and Engineering Discovery Environment (XSEDE) (which is supported by National Science Foundation grant number OCI-1053575), the Bridges system at the Pittsburgh Supercomputing Center (PSC) (which is supported by NSF award number ACI-1445606), the Quest high performance computing facility at Northwestern University, and the National Energy Research Scientific Computing Center (DOE: DE-AC02- 05CH11231).

Peer Reviewed