Re: SDCI with several roots

Posted by perake on November 21, 2001 at 08:37:29:

In Reply to: SDCI with several roots posted by Evgeniy on November 20, 2001 at 19:06:36:

Dear Evgeniy,
There are two problems here:
(1) When several roots are requested from the MRCI program, the start approximations
are taken from the reference CI. Thus, when you want several roots, the reference
space must be large enough so that it contains not just a single state. In fact, it
should be large enough so that the start approximations are reasonable ones, else
there unavoidable root flipping during iterations will make convergence problematic.
So the construction of the MRCI program makes it difficult to do a single-reference
SDCI if also excited states are needed. The calculated result would anyway be very
bad for the excited states, so is this really what you want to do? If you actually
need to do this calculation, please write me, and I will look into it.
(2) But you also want the roots for other symmetries. This is not possible with
the MRCI program, except of course that one may use a lower symmetry. But then the
entire calculation (SEWARD, SCF, etc) must be done in the lower symmetry. There is
presently no way of taking integrals and orbitals from a calculation using a certain
symmetry group and then compute wave functions that break this symmetry, nor can
RASSCF, MRCI or CASPT2 handle simultaneously wave functions of more than one irrep.

Best Regards,

Follow Ups:

Post a Followup

Name:
E-Mail:

Subject:

Comments:
: Dear Evgeniy, : There are two problems here: : (1) When several roots are requested from the MRCI program, the start approximations : are taken from the reference CI. Thus, when you want several roots, the reference : space must be large enough so that it contains not just a single state. In fact, it : should be large enough so that the start approximations are reasonable ones, else : there unavoidable root flipping during iterations will make convergence problematic. : So the construction of the MRCI program makes it difficult to do a single-reference : SDCI if also excited states are needed. The calculated result would anyway be very : bad for the excited states, so is this really what you want to do? If you actually : need to do this calculation, please write me, and I will look into it. : (2) But you also want the roots for other symmetries. This is not possible with : the MRCI program, except of course that one may use a lower symmetry. But then the : entire calculation (SEWARD, SCF, etc) must be done in the lower symmetry. There is : presently no way of taking integrals and orbitals from a calculation using a certain : symmetry group and then compute wave functions that break this symmetry, nor can : RASSCF, MRCI or CASPT2 handle simultaneously wave functions of more than one irrep. : Best Regards,