| ADAPTIVE | An adaptive grid is used for the integration. This is the default. | ||
| FIXED | A fixed grid is used for the integration. | 
| MEDIUM | Medium grid accuracy is requested. This is the default. | ||
| COARSE | Coarse grid accuracy is requested. | ||
| FINE | Fine grid accuracy is requested. | ||
| REFERENCE | A (fixed) reference grid is requested. | ||
| TOL=  Real   | Adaptive grid tolerance. | ||
| RAD=  Integer   | Number of radial shells for adaptive grid. | 
| SCF | The converged SCF density is used for the adaptive grid generation. This is the default. | ||
| GUESS | The start density is used for the adaptive grid generation. | ||
| DIRECT | The grid coordinates and weights are rebuilt in each SCF cycle. This is the default for parallel runs. | 
 (MEDIUM) for the numerical integration of the exchange-correlation
energy and matrix elements of the associated potential. This tolerance
indicates the accuracy of the numerical integration of the diagonal elements
of the exchange-correlation potential matrix. With this setting, the converged
SCF energy accuracy is usually better than 100
 (MEDIUM) for the numerical integration of the exchange-correlation
energy and matrix elements of the associated potential. This tolerance
indicates the accuracy of the numerical integration of the diagonal elements
of the exchange-correlation potential matrix. With this setting, the converged
SCF energy accuracy is usually better than 100  Hartree [118]. The
adaptive grid COARSE and FINE options refer to grid tolerances of
Hartree [118]. The
adaptive grid COARSE and FINE options refer to grid tolerances of  and
and  , respectively. Thus, the stability of the numerical integration
can be easily checked by choosing different grid tolerances. The COARSE adaptive
grid should not be used for final energy or property calculations. With the TOL
option, a user-defined grid tolerance can be specified. Of course, this option is
only applicable for an adaptive grid. The same holds for the RAD option that can 
be used to define the number of radial shells for an adaptive grid. Its upper 
limit is 500 specified by the MAXRAD option in the parameter.h file. The
SCF and GUESS options specify the density used for the exchange-correlation 
potential which generates the adaptive grid. By default, the converged SCF density 
is used. This requires an extra SCF for the initial energy calculation. If the 
option GUESS is specified, the adaptive grid is built using the start density 
for the calculation of the grid generating exchange-correlation potential.
, respectively. Thus, the stability of the numerical integration
can be easily checked by choosing different grid tolerances. The COARSE adaptive
grid should not be used for final energy or property calculations. With the TOL
option, a user-defined grid tolerance can be specified. Of course, this option is
only applicable for an adaptive grid. The same holds for the RAD option that can 
be used to define the number of radial shells for an adaptive grid. Its upper 
limit is 500 specified by the MAXRAD option in the parameter.h file. The
SCF and GUESS options specify the density used for the exchange-correlation 
potential which generates the adaptive grid. By default, the converged SCF density 
is used. This requires an extra SCF for the initial energy calculation. If the 
option GUESS is specified, the adaptive grid is built using the start density 
for the calculation of the grid generating exchange-correlation potential.
Fixed grids are also available in deMon2k. They can be selected by the option FIXED in combination with the options COARSE, MEDIUM, and FINE. MEDIUM is the default for fixed grids. With the options COARSE, MEDIUM, and FINE, the pruned grids (50,194)p, (75,302)p and (99,590)p are selected. In this notation the first number refers to the radial shells and the second to the number of Lebedev grid points [119] on each of these shells. The "p" stands for pruned [120], which indicates that, for some radial shells, smaller Lebedev grids have been used. With the REFERENCE option, the fixed unpruned (200,1202) reference grid is specified. Due to its size, its application should be restricted to small systems.
The DIRECT option is used to avoid the I/O of the grid point coordinates and weights. With this option, the grid is built on the fly in each SCF step. This results in a small additional computational overhead. For serial runs, the I/O of the grid data is usually not significant. However, this behavior is architecture dependent. In parallel runs, the grid I/O becomes a bottleneck. Therefore, the DIRECT option is the default setting in such calculations.