Molcas Forum

Samer

Member

Registered: 2020-09-16

Posts: 4

Adding different symmetry orbitals in CASSCF

Hi everyone,

We recently ran the following test, two RASSCF calculations on the same system/geometry:

Calculation 1: 10 electrons and 10 orbitals in A” symmetry:

 &rasscf
 symmetry = 1
 lumorb
 nactel = 10 0 0
 inactive = 55 7
 ras2 = 0 10
 ciroot = 3 3 1
 rlxroot = 1

Calculation 2: Same system, 4 electrons and 2 orbitals in A’ symmetry, and 10 orbitals 10 electrons in A” symmetry:

 &rasscf
 symmetry = 1
 lumorb
 nactel = 14 0 0
 inactive = 53 7
 ras2 = 2 10
 ciroot = 3 3 1
 rlxroot = 1

My (naive) expectation was that the energies from the two jobs would be identical (and if not absolute energies then at least the excitation energies would be identical). The reason: The added A’ orbitals in the active space contribute no new configurations (symmetry=1 prevents A'->A”, only A"->A" allowed) and the A' orbitals do not mix with A” orbitals during orbital optimization. The A” active space orbital shapes and occupancies are identical in the two calculations, so we verified there is no change in A" active space.

The result: the energies are similar, but not identical. The absolute energies change (by ~8 kJ/mol for root 1 in this case), and also the relative energies of the three A” states (by up to ~4 kJ/mol).

The only explanation I could think of is that this energy change must be due to orbital relaxation (A’ active space orbital coefficients are optimized along with A”). I’m mainly surprised, however, that the A” relative state energies are affected. Am I missing something?

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,012

Re: Adding different symmetry orbitals in CASSCF

I guess one way to see how large the orbital relaxation effect is, is doing a 14,12 CASCI on the the 10,10 orbitals.

You should also check the CSF contributions for excitations involving the A’ orbitals. Note that single A'->A” excitations may be forbidden, but not double excitations, although those are presumably less important.

I wouldn't expect the absolute energies to be identical. At the very least you are including some "dynamical" correlation from the A’ active orbitals.

Samer

Member

Registered: 2020-09-16

Posts: 4

Re: Adding different symmetry orbitals in CASSCF

Hi Ignacio,

Of course, you are right! I was thinking just with a single excitation picture, I forgot about higher excitations! I was looking at occupations through Molden, which rounds to 2 decimal places and showed 2.00 occupation for A' orbitals, but it just occurred to me to look in the output, and A' orbital occupation is ~1.998, so there is some double-excitation character from A' to A".

Good idea about CASCI, will try that. Thanks!