Size: 1336
Comment:
|
← Revision 7 as of 2021-06-01 12:25:07 ⇥
Size: 1496
Comment:
|
Deletions are marked like this. | Additions are marked like this. |
Line 11: | Line 11: |
NumGrad is used to calculate numerical gradients. It can be combined with BDFOPT module so as to perform geometry optimizations based on numeric gradients: simply write the $numgrad block after the $bdfopt block, followed by the modules needed in computing the energy. | NumGrad is used to calculate numerical gradients. It can be combined with BDFOPT module so as to perform geometry optimizations based on numerical gradients: simply write the $numgrad block after the $bdfopt block, followed by the modules needed in computing the energy. |
Line 38: | Line 38: |
Calculate XIANCI methods numerical gradient. | Calculate XIANCI methods' numerical gradient. |
Line 47: | Line 47: |
Do not exploit translational invariance. | Do not exploit translational invariance. This somewhat increases the cost of the calculation, but may sometimes help to reduce the deviation of the numerical gradient with respect to point group symmetry. |
Line 49: | Line 49: |
}}} |
NumGrad
Contents
NumGrad is used to calculate numerical gradients. It can be combined with BDFOPT module so as to perform geometry optimizations based on numerical gradients: simply write the $numgrad block after the $bdfopt block, followed by the modules needed in computing the energy. Note that translational invariance and the point group symmetry of the molecule are fully exploited to bring down the number of necessary displacements. Rotational invariance is not yet made use of, however.
General keywords
IPRT
- Print level.
RSTEP
- The numerical difference, default is 5.D-3.
SCF
- Calculate SCF numerical gradient.
MCSCF
- Calculate MCSCF numerical gradient.
XIANCI
- Calculate XIANCI methods' numerical gradient.
CCSD
- Calculate CCSD or CCSD(T) numerical gradient. If itriple 1 is specified in $ccsdso, the CCSD(T) numerical gradient is calculated, else the CCSD numerical gradient is calculated.
NOTRANSINVAR
- Do not exploit translational invariance. This somewhat increases the cost of the calculation, but may sometimes help to reduce the deviation of the numerical gradient with respect to point group symmetry.