*DFT

Introduction

In this link the directives for modifying the DFT calculation are described (e.g. modifying grid, using ALDA and much more)

typical input (here correcting LDA):

**HAMILTONIAN
.DFT
LDA
*DFT
.SAOP
LBalpha

Single line following .SAOP : asymptotic potential (currently LB94 or LBalpha).

The specification of the functional is described in connection with the .DFT option in the .DFT section

The DFT module is described in references [Schipper2000], [Saue2002], [Fossgaard2003].

It uses standard non-relativistic functionals. This is justified in many cases since various studies, like refs. [Varga1999], [Varga2000], [Mayer1996], indicate that relativistic corrections to the exchange-correlation functionals have a negligible effect on spectroscopic constants; however, for example for core properties further studies are necessary.

Integration algorithm and thresholds

.SCREENING

Specify the screening threshold used in the BLAS-3 integration scheme.

Default:

.SCREENING
 1.0D-18

.POINTWISE

Switch to old (unscreened) BLAS-2 integration scheme (default before DIRAC11) instead of the new BLAS-3 scheme (default since DIRAC11).

.TINYDENS

Specify the tiny density threshold. Grid points with a density smaller than this threshold do not contribute to XC matrix elements.

Default:

.TINYDENS
 1.0D-14

Asymptotic correction

.GRAC

Activate the gradient regulated asymptotic correction scheme (GRAC), [Gruning2001], followed by two lines of input.

First line: Asymptotic potential (currently LB94 or LBalpha).

Second line (free format): Parameters α and β (see reference [Gruning2001]) the ionization potential, and the threshold for difference in HOMO eigenvalue below which asymptotic correction is switched on. A sufficiently converged density is needed before the correction is activated (before the bulk potential is shifted in order to reproduce the desired IP (see reference [Gruning2001])

Typical input (here correcting PBE0):

**HAMILTONIAN
.DFT
PBE0
*DFT
.GRAC
LB94
0.5 40.0 0.79248 1.0D-6

.SAOP!

Activate the statistical averaging of (model) orbital potentials (SAOP) as defined in reference [Schipper2000]. This implies (and activates) the ALDA kernel. The functional under .DFT is expected to be GLLBhole. The asymptotic potential is LBalpha.

Recommended input:

**HAMILTONIAN
.DFT
GLLBhole
*DFT
.SAOP!

.SAOP

The more general version of .SAOP!

Followed by a single line of input: asymptotic potential (currently LB94 or LBalpha).

Typical input (here correcting LDA):

**HAMILTONIAN
.DFT
LDA
*DFT
.SAOP
LBalpha

Spin magnetization TDDFT

.NOSAOP

Turn off spin density contribution to XC response.

.COLLINEAR

Use the collinear approximation as a definition of the spin density.

\begin{equation*} s={m_{z}}={\sum\limits_{i}}{ \phi_{i}^{\dagger}{{\Sigma}_{z}}{\phi_{i}} } \end{equation*}

instead of the default noncollinear definition

\begin{equation*} s = \vert \mathbf{m} \vert \end{equation*}

.BETASIGMA

use

\begin{equation*} \mathbf{m}=\sum_{i}{\phi_{i}^{\dagger}{\beta}{\mathbf{\Sigma}}{\phi_{i}}} \end{equation*}

instead of the default form

\begin{equation*} \mathbf{m} = \sum_{i} \phi_{i}^\dagger \mathbf{\Sigma} \phi_{i} \end{equation*}

Adiabatic local density approximation (ALDA)

.ALDA

Approximate all functional derivatives beyond the xc potential by SVWN derivatives. For hybrid functionals exact exchange is switched off in the solution of the response equation.

.XALDA

Same as .ALDA but keep the fraction of exact exchange of the xc functional under .DFT.

.ALDA+

Use .ALDA+ only for the Hermitian contribution (density contribution) and use the proper xc kernel for the anti-Hermitian part (spin density contribution).

.ALDA-

Use .ALDA- only for the anti-Hermitian contribution (spin density contribution) and use the proper xc kernel for the Hermitian part (density contribution).

.XALDA+

Use .XALDA+ only for the Hermitian contribution (density contribution) and use the proper xc kernel for the anti-Hermitian part (spin density contribution).

.XALDA-

Use .XALDA- only for the anti-Hermitian contribution (spin density contribution) and use the proper xc kernel for the Hermitian part (density contribution).

Other functionality

.GAUNTSCALE

Scale Gaunt integrals (if included) with the same factor as for Hartree-Fock exchange. This means that hybrid functionals include fractional HF Gaunt interaction, and pure functionals no HF Gaunt interaction at all. If this option is not given the Gaunt integrals will be included to 100%, meaning that even an LDA calculation will include full Hartree-Fock Gaunt interaction when the .GAUNT keyword is given in **HAMILTONIAN.

.OVLDIAG

Activate the overlap diagnostic for TD-DFT calculations of excitation energies according to [Peach2008]. This is only available in the development version.