Posted by Per Ake Malmqvist on April 15, 2004 at 11:08:14:
In Reply to: CASPT2 analytic gradient was so bad??? posted by Evgeny K. Dolgov on April 15, 2004 at 09:32:20:
Unfortunately, you are quite right: We do *not* have analytic CASPT2 gradients. We have had
a guy working on this for a while, but the work was not finished. The situation now is that
we lack the man power to complete this programming, so nothing is planned for the
foreseeable future. Of course the programming project will be picked up again whenever we
have some person with necessary skills and interest to take it up, but right now everyone is
busy full-time with other tasks.
I fully agree that this would be a very valuable addition to MOLCAS.
You may consider buying MOLPRO. They have CASPT2 with gradients. Their CASPT2 formulation
is different, and for very large interacting spaces, there may be some inefficiency since
they do not use internally contracted semi-internal excitations, but in other respects,
their CASPT2 should give almost identical results to ours.