Benchmarking the performance of plane-wave vs. localized orbital basis set methods in DFT modeling of metal surface: a case study for Fe-(110)

Reproducing electronic structure of extended metallic systems is computationally demanding with the cost efficiency of this approach heavily dependent on both the density functional and the basis function used to approximate the electronic orbitals. It is well known that the generalized gradient approximation functional (GGA) is the most suitable and reliable approach for the description of metallic systems. As for the basis functions, two approaches dominate: the linear combination of localized basis functions (LB) such as Gaussian functions and the linear combination of plane waves (PW). Both have their own advantages and disadvantages, that may impact the efficiency and accuracy of the simulations. In this work, we use the VASP and the CRYSTAL14 suites of codes that employ plane waves and localized Gaussian basis sets, respectively, to establish a benchmark on their computational efficiency for the modeling of metal surfaces. The PW basis technique requires that the entire simulation box including the vacuum space be filled with plane waves which reduces the computational efficiency and limits the vacuum space. For its part, the LB method is based on atomic localized orbitals and does not require vacuum to model surfaces. Therefore, for calculations that require relatively large vacuum thickness such as modeling of adsorption, the LB method might be superior in terms of computational expense while providing the comparable accuracy.