Posted by Jesper Wisborg Krogh on May 31, 2005 at 22:22:50:
In Reply to: CASPT2 error: DSLEV posted by Nicolas Ferré on May 30, 2005 at 11:03:38:
Hi,
Yes, you are absolutely correct that the iteration limit in dslev is hard coded to 30. I'm currently working on replacing dslev with a more robust routine, but until then I think the best advice is to use the 'JACObi' keyword in CASPT2. This will ensure that the Jacobi algorithm is used instead. Although this can be much slower, it's much more stable, and should hopefully be able to do the job.
/ Jesper Wisborg Krogh
: During a CASPT2 job, I got the following error:
: *** ERROR IN SUBROUTINE DSLEV ***
: QL METHOD DID NOT CONVERGE
: Calling history
: ---------------
: last entry: DSLEV
: DSLEV called by SBDIAG
: SBDIAG called by EQCTL
: EQCTL called by CASPT2
: Actually this error is related to some iteration limit (J=30) hard-coded into the essl dslev subroutine.
: I am not too familiar with the MaxIter, FockType of Jacobi options that may solve my problem. Which solution is the best ?