Next: 39.2.2 Numerical derivatives of Up: 39.2 Numerical gradients Previous: 39.2 Numerical gradients


39.2.1 Choice of coordinates (COORD)

By default, the numerical gradients are computed relative to all variables on which the z-matrix depends. If the z-matrix depends on no variables or on $3N$ variables, the gradient is computed for all $3N$ coordinates and symmetrical displacement coordinates are used to evaluate the gradient. This yields the minimum computational effort.

These defaults can be modified using the COORD directive:

COORD,coord_type,[displacement_type]

where coord_type can be one of the following:

2cm

ZMAT
Compute the numerical gradients for all variables on which the geometry depends (default).
3N or CART
Compute the gradients for all $3N$ nuclear coordinates. This is the default if the z-matrix does not depend on variables or if the xyz input format is used. If this option is used and the original geometry is given in z-matrix form, the z-matrix is lost.

The specification of displacement_type is optional and only affects the numerical calculation of the gradient for $3N$ coordinates. It can also be given using

DISPLACE,displacement_type

displacement_type can be one of the following:

SYM
Use symmetrical displacements. This yields the minimum number of displacements and always preserves the symmetry of the wavefunction. This is the default and only recommended option.
CART
Displacements are generated for all $3N$ Cartesian coordinates. This is normally not recommended, since in cases in which molecular symmetry is present it generates far more displacements than needed. Also, the wavefunction symmetry is not preserved, and the calculation must be done in C1 symmetry.
UNIQUE
As CART, but symmetry-equivalent displacements are eliminated. Not recommended either.



Next: 39.2.2 Numerical derivatives of Up: 39.2 Numerical gradients Previous: 39.2 Numerical gradients

molpro@molpro.net
Sep 24, 2008