Initial guess of the density matrix

flo fsch... at pci.uzh.ch
Fri Feb 12 12:11:06 CET 2010


Hi,

actually there is a second way, which could solve your problem, in
case the intial guess is not sufficient.
CP2K provides a section WFN_MIX in QS%PRINT. This keyword can be used
to reorder the MOs in the restart file.

It has to be used as follows:
You need to run a complete wavefunction optimization, and specify:

RESULT_MO_INDEX as the index of the MO to be modified (you can specify
the spin of this MO as well) and
ORIG_MO_INDEX to be the MO from which the WFN to be pasted originates
from.

If you want for example to swap the alpha and beta spins of two states
(1 and 2) you will have to specify

&WFN_MIX
           &UPDATE
           RESULT_MO_INDEX 1
           RESULT_SPIN_INDEX ALPHA
           RESULT_SCALE 0.000000000

           ORIG_MO_INDEX 2
           ORIG_SPIN_INDEX BETA
           ORIG_SCALE 1.0000000000

           ORIG_IS_VIRTUAL .FALSE.
       &END
       &UPDATE
           RESULT_MO_INDEX 2
           RESULT_SPIN_INDEX BETA
           RESULT_SCALE 0.0000000

           ORIG_MO_INDEX 1
           ORIG_SPIN_INDEX ALPHA
           ORIG_SCALE 1.0000000

           ORIG_IS_VIRTUAL .FALSE.
       &END
&END

like this the beta WFN of MO 2 is moved to alpha MO 1 and alpha MO1 is
moved to beta MO 2.
If you now do a restart using the RESTART_WFN (the modified WFN simply
replaces the original RESTART_WFN, therefore, the restart file won't
have a special name), the inital guess will have the state specified
above.

I don't know if this really fits your problem, but as far as I could
see, it should work.

regards
Flo

Don't know if this solution



More information about the CP2K-user mailing list