ACPF

Posted by Per Ake Malmqvist on April 08, 2003 at 16:19:07:

In Reply to: ACPF posted by B. Hajgato on April 08, 2003 at 15:52:51:

There will be a difference between two MRCI calculations if based on different
orbitals. The CASSCF orbitals will in general be better than the RHF ones. In the
present case (2 electrons in 2 orbitals, open-shell singlet) the SCF program is
unable to optimize the orbitals for this state -- perhaps you have set a fractional
occupation number to the SCF (occ. numbers 1.0 and 1.0 for the open orbitals). In that
case the SCF orbitals will be optimized for the ensemble average of the singlet and the
triplet. The CASSCF will be very close to a proper open-shell HF (main difference being that some small correlation effects may also become included, unless this is impossible
due to symmetry restrictions).

There will also be a difference between MR-ACPF and MRCI for the same calculation.
The MRCI calculation gives a variational bound, which may be preferable if a very
accurate calculation is intended, with an adequate reference space, and with not many
electrons, but MR-ACPF if often better since the energy is (almost) extensive, but the
energy is not variational. If many electrons are correlated, the MR-ACPF is preferred.

Follow Ups:

Post a Followup

Name:
E-Mail:

Subject:

Comments:
: There will be a difference between two MRCI calculations if based on different : orbitals. The CASSCF orbitals will in general be better than the RHF ones. In the : present case (2 electrons in 2 orbitals, open-shell singlet) the SCF program is : unable to optimize the orbitals for this state -- perhaps you have set a fractional : occupation number to the SCF (occ. numbers 1.0 and 1.0 for the open orbitals). In that : case the SCF orbitals will be optimized for the ensemble average of the singlet and the : triplet. The CASSCF will be very close to a proper open-shell HF (main difference being that some small correlation effects may also become included, unless this is impossible : due to symmetry restrictions). : There will also be a difference between MR-ACPF and MRCI for the same calculation. : The MRCI calculation gives a variational bound, which may be preferable if a very : accurate calculation is intended, with an adequate reference space, and with not many : electrons, but MR-ACPF if often better since the energy is (almost) extensive, but the : energy is not variational. If many electrons are correlated, the MR-ACPF is preferred.