**GRID

The numerical integration scheme uses the Becke partitioning [Becke1988a] of the molecular volume into atomic ones, for which the quadrature is performed in spherical coordinates. Radial integration is carried out using the scheme proposed by Lindh et al. [Lindh2001], while the angular integration is handled by a set of highly accurate Lebedev grids. Note that the Becke partitioning scheme generally uses Slater-Bragg radii for atomic size adjustments. For the heavier elements, with their more variable oxidation states, this can lead to errors. If the user invokes the .ATSIZE keyword, DIRAC tries to deduce relative atomic sizes from available densities, if it can find coefficients.

.RADINT

Specify the maximum error in the radial integration.

Default:

.RADINT
 1.0D-13

.ANGINT

Specify the precision of the Lebedev angular grid. The angular integration of spherical harmonics will be exact to the L-value given by the user. By default the grid will be pruned. The highest implemented value is 64.

Default:

.ANGINT
 41

.NOPRUN

Turn off the pruning of the angular grid.

.ANGMIN

Specify the minimum precision of the Lebedev angular grid after pruning.

.ANGMIN
 LEBMIN

Close to the nucleus, the precision of the angular grid will be less than L-value given by .ANGINT, but it will not be less than the integer LEBMIN. Note that giving a LEBMIN value greater than or equal to the L-value given by .ANGINT is equivalent to turning the pruning off.

Default:

.ANGMIN
 15

.ATSIZE

Generate new estimates for atomic size ratios for use in the Becke partitioning scheme. Relative sizes for a pair of atoms A and B are calculated from their contribution to the large component density along the line connecting A and B.

.IMPORT

*DFT
.IMPORT
grid_file

Import previously exported DFT grid.

For debugging you can also create your own grid file. The grid file is formatted - in free format. A grid file for three points could look like this:

3
0.1  0.1 0.1    1.0
0.01 0.2 0.4    0.9
9.9  9.9 9.0    0.8
-1

The first integer is the number of points, then come the x-, y-, and z-coordinate, and the weight for each point. The last line is a negative integer. Works also in parallel.

You can export the DIRAC grid simply by copying back the file “numerical_grid” from the scratch directory.

.NOZIP

DIRAC will try to remove redundant grid points when symmetry is present. Each reflection plane reduces the number of grid points by a factor of two. With this keyword you can turn this default symmetry grid compression off.

.4CGRID

Include the small component basis in the generation of the DFT grid even if you run a 1- or 2-component calculation.

This can be useful if you want to compare to a 4-component calculation using the identical integration grid.

By default the small component is ignored in the grid generation when running 1- or 2-component DFT calculations.

.DEBUG

Very poor grid - corresponds to

.RADINT
 1.0D-3
.ANGINT
 10

.COARSE

Coarse grid - corresponds to

.RADINT
 1.0D-11
.ANGINT
 35

.ULTRAFINE

A better grid than default - corresponds to

.RADINT
 2.0D-15
.ANGINT
 64

.INTCHK

Test the performance of the grid by computing the overlap matrix numerically and analyzing the errors. In addition, the error matrix can be printed.

Default (no test):

.INTCHK
 0

Error analysis:

.INTCHK
 1

Print error matrix:

.INTCHK
 2