2
methods can be computationally much more efficient [4, 5] by using the supercell technology with the periodic boundary
conditions to simulate any crystalline materials.
Due to the fundamental importance in various scientific areas including chemical physics, biophysics, environmental
science, interstellar medium, etc., research on small molecules is constantly developing both theoretically and experimentally.
Specifically, adsorption of small molecules on materials surfaces (including outer surfaces of materials or the surfaces of pores
in materials, e.g., molecular sieves [8, 9]) is often considered for both theoretical studies and applications. The supercell for
such a system by using a planewave DFT method often contains hundreds to thousands of atoms at least for obtaining reliable
adsorption properties. In addition, ab initio molecular dynamics (AIMD) simulations [10] for such systems are also selectively
implemented for, e.g., visualizing the diffusion paths of molecules on the surfaces, while AIMD simulations are even much
more demanding than normal structural optimizations. Thus, using AOBMs with higher accuracies but high computational
costs is impractical for such computations. Instead, the planewave DFT methods can be computationally practical due to the
high efficiency. However, when a planewave DFT method is applied to a specific system, the reliability of the method must be
first assessed because the accuracy of the DFT results can significantly depend on the electronic exchange-correlation energy
functionals. To this end, we mention an example. It is well-known that Perdew-Burke-Ernzerhof (PBE) functional [11] cannot
predict the interlayer spacing of graphite due to the absence of the dispersion corrections, e.g., the predicted lattice constant
= 8.870 Å along the [0001] direction of graphite from our previous PBE calculations [12] is notoriously much larger than the
experimental value of 6.6720 Å [13]. In contrast, the value of 6.701 Å from our optB88-vdW [14] calculations with dispersion
corrections can reproduce the above experimental value very well [12]. Thus, one should be particularly careful when
selecting a functional with or without dispersion corrections to calculate the weakly bonded systems like graphite.
As a semilocal functional, the generalized gradient approximation (GGA) generally has comparably good accuracies for
calculating ground state properties of neither weakly bonded nor strongly correlated systems. For example, recent extensive
tests on the lattice constants, bulk moduli, and cohesive energies (or atomization energies) of 44 strongly and 17 weakly
bonded solids from various local, semilocal, and nonlocal functionals (so-called “DFT Jacob’s ladder”) have been reported [15,
16]. For the weakly bonded systems, dispersion corrections need to be considered [17], while for strongly correlated systems,
DFT+U corrections are usually needed [18]. In addition, more nonlocal functionals (upper rungs of the DFT Jacob’s ladder) can
have higher accuracies but higher computational costs than more local functionals (lower rungs). Therefore, to appropriately
choose a functional before calculating a specific system, both accuracy and computational cost need to be balanced.
Recently, we have selected and applied the optB88-vdW functional [14], which typically considers dispersion correction
including van der Walls (vdW) interactions, to various vdW materials with guest atoms [12, 19–30] and silica polymorphs with
molecular groups consisting of C, H, and O [31]. These applications have already been proven very successful. The success is
not surprising, given that these systems include weakly bonded interlayer spaces, while for the weakly bonded systems, the