Optimizing density-functional simulations for two-dimensional metals

TL;DR

This study benchmarks various density-functional theory methods to identify efficient and accurate computational approaches for simulating two-dimensional metals, facilitating experimental comparisons and data-driven material discovery.

Contribution

It systematically compares DFT functionals and basis sets for 2D metals, revealing that local basis with PBE is sufficient for most purposes, reducing computational costs.

Findings

Local basis with PBE is adequate for 2D metal simulations.

Plane waves and hybrid functionals offer limited improvements.

Results support efficient DFT data generation for experimental and machine learning applications.

Abstract

Unlike covalent two-dimensional (2D) materials like graphene, 2D metals have non-layered structures due to their non-directional, metallic bonding. While experiments on 2D metals are still scarce and challenging, density-functional theory (DFT) provides an ideal approach to predict their basic properties and assist in their design. However, DFT methods have been rarely benchmarked against metallic bonding at low dimensions. Therefore, to identify optimal DFT attributes for a desired accuracy, we systematically benchmark exchange-correlation functionals from LDA to hybrids and basis sets from plane waves to local basis with different pseudopotentials. With 1D chain, 2D honeycomb, 2D square, 2D hexagonal, and 3D bulk metallic systems, we compare the DFT attributes using bond lengths, cohesive energies, elastic constants, densities of states, and computational costs. Although today most DFT studies on 2D metals use plane waves, our comparisons reveal that local basis with often-used PBE exchange-correlation is well sufficient for most purposes, while plane waves and hybrid functionals bring limited improvement compared to the greatly increased computational cost. These results ease the demands for generating DFT data for better interaction with experiments and for data-driven discoveries of 2D metals incorporating machine learning algorithms.