Electrostatic corrections

For charged cells or for calculations of molecules and surfaces with a large dipole moment, the energy converges very slowly with respect to the size $L$ of the supercell. Using methods discussed by Makov et al.^[1] and Neugebauer et al.^[2], VASP can correct for the leading errors (in many details, we have taken a more general approach, though).

The following flags control the behaviour of VASP:

NELECT, charged systems:

NELECT determines the total number of electrons in the system. The value may deviate from the default value, which is calculated assuming charge neutrality in the system. If NELECT differs from the default, an additional neutralizing background charge is applied by VASP. In this case, however, the energy converges very slowly with respect to the size L of the super cell. The required first order correction to the energy is given by

\frac{e^2q^2\alpha}{L\epsilon}

where q is the net charge of the system, α the Madelung constant of a point charge q placed in a homogeneous background charge -q, and ε the dielectric constant of the system. For atoms or molecules surrounded by vacuum, ε takes on the vacuum value ε=1. VASP can automatically correct for the leading error, by setting the IDIPOL and EPSILON tags in the INCAR file (see below).

It is important to emphasize that the total energy can not be corrected for charged slabs, since a charged slab results in an electrostatic potential that grows linearly with the distance from the slab (corresponding to a fixed electrostatic field). It is fairly simple to show that as a result of the interaction between the charged slab and the compensating background, the total energy depends linearly on the width of the vacuum. Presently, no simple a posteriori correction scheme is known or implemented in VASP. Total energies from charged slab calculations are hence useless, and can not be used to determine relative energies.

Note: If you are not convinced, simply vary the vacuum width and draw the energy versus the vacuum width.

EPSILON, dielectric constant:

IDIPOL, type of correction (monopole/dipole and quadrupole):

For systems with a net dipole moment, the energy converges slowly with respect to the size of the super cell as well. The dipole corrections (and quadrupole corrections for charged systems) fall off like 1/L³. Both corrections, dipole and quadrupole for charged systems, will be calculated and added to the total energy if IDIPOL is set.

There are four possible settings for IDIPOL (= 1 | 2 | 3 | 4).

For IDIPOL=1-3, the dipole moment will be calculated only parallel to the direction of the first, second or third lattice vector, respectively. The corrections for the total energy are calculated as the energy difference between a monopole/dipole and quadrupole in the current supercell and the same dipole placed in a super cell with the corresponding lattice vector approaching infinity. This flag should be used for slab calculations.

For IDIPOL=4 the full dipole moment in all directions will be calculated, and the corrections to the total energy are calculated as the energy difference between a monopole/dipole/quadrupole in the current supercell and the same monopole/dipole/quadrupole placed in a vacuum, use this flag for calculations for isolated molecules.

Note: strictly speaking quadrupole corrections is not the proper wording. The relevant quantity is

\int d^3{\mathbf r} \rho(\mathbf r) \Vert \mathbf r\Vert^2.

DIPOL, center of the net charge of the cell

This tag sets the center of the net charge distribution: DIPOL=R_center. The dipole is then defined as

\int ({\mathbf r}-{\mathbf R}_\mathrm{center})\rho_\mathrm{ions+valence} ({\mathbf r})d{\mathbf r}

If DIPOL is not set, VASP determines, where the charge density averaged over one plane drops to a minimum and calculates the center of the charge distribution by adding half of the lattice vector perpendicular to the plane where the charge density has a minimum (this is a rather reliable approach for orthorhombic cells).

LDIPOL and LMONO, enable dipole and/or monopole corrections to the potential:

These tags switch on the potential correction mode. Due to the periodic boundary conditions, not only the total energy converges slowly with respect to the size of the supercell, but also the potential and the forces are affected by finite size errors. This effect can be counterbalanced by setting LDIPOL=.TRUE. (dipole corrections) and/or LMONO=.TRUE. (monopole corrections) in the INCAR file. For LDIPOL=.TRUE.,a linear and for LMONO=.TRUE., a quadratic electrostatic potential is added to the local potential, correcting the errors introduced by the periodic boundary conditions. This is in the spirit of Neugebauer et al.^[2] (but more general and the total energy has been correctly implemented right away). The biggest advantage of this mode is that leading errors in the forces are corrected, and that the work-function can be evaluated for asymmetric slabs. The disadvantage is that the convergence to the electronic groundstate might slow down considerably (i.e., more electronic iterations might be required to obtain the required precision). It is recommended to use this mode only after pre-converging the orbitals without the LDIPOL flag, and the center of charge should be set in the INCAR file (DIPOL= center of mass). The user must also ensure that the cell is sufficiently large to determine the dipole moment with sufficient accuracy. If the cell is too small, charge might swap through the vacuum, causing very slow convergence (often convergence improves with the size of the supercell).

EFIELD, applied electrostatic field

For the current implementation, there are several restrictions; please read carefully:

Charged systems:
Quadrupole corrections are only correct for cubic supercells (this means that the calculated $1/L^3$ corrections are wrong for charged supercells if the supercell is non cubic). In addition, we have found empirically that for charged systems with excess electrons (NELECT $>$ NELECT $_{\rm neutral}$ ) more reliable results can be obtained if the energy after correction of the linear error ( $1/L$ ) is plotted against $1/L^3$ to extrapolate results manually for $L\to \infty$ . This is due to the uncertainties in extracting the quadrupole moment of systems with excess electrons.
Potential corrections are only possible for orthorhombic cells (at least the direction in which the potential is corrected must be orthogonal to the other two directions).