PC GAMESS-related discussion club

Alex Granovsky
gran@classic.chem.msu.su

you have at least three options to proceed.

First, you can try MCQDPT2 instead of MRCI. It is fast!

Second, PC GAMESS allows one to specify both SOCI and iexcit options in the $drt and $cidrt groups at the same time. In this case, iexcit refers to the maximum allowed electron excitation level.

Third, you can use $cidrt intact=.t. $end option.

Finally, you can further reduce memory needs by tuning parameters related to Davidson diagonalization in $gugdia group.

Provided you'll fit your run into available memory, I would recommend you to use pack2 option to reduce disk requirements for GUGA loops,
or even to switch to direct GUGA CI (dirci option) in the case you'll run out of free disk space.

Regards,
Alex Granovsky


On Thu May 17 '07, Kris wrote
-----------------------------
>Hi all,

>I am wanting to calculate a second order MRCI. The system has a reference CASSCF of 9 electrons in 12 orbitals resulting in more than 46000 references CSF's. Hence the total number of CSF's to use in the MRCI is over a 100 million. Is there any way of reducing this number of CSF without changing the size and occupation of the reference space?

>I was thinking that by excluding those CSF's with a total contribution say less than 0.000001%, would be one solution but i'm not sure if this is possible.

>Does anybody have any suggestions?

>Thanks

>Kris

>kwright06@qub.ac.uk