How to use SIRIFC

[Addiw7]

Dear Dalton Users,

I would like to perform one-photon absorption energies and oscillator strength but with the DSCF and
properties calculations separately.

If I do DSCF and properties together in one input (see attached file), I get the exc. energy and osc. strength
of 3.875 eV and 0.774 for the first excited state.

If I do them separately and use SIRIFC file in the latter calculations (see attached files), I get the
exc. energy and osc. strength of 7.142 eV and 1.382.

From the general knowledge of my system, the latter answer is simply wrong.

I think that I start with properties calculations in a wrong way using SIRIFC file. What is wrong then?

Best regards,
Dawid Grabarek

[magnus]

Replace SIRIFC with NEWORB, or you can use the more intuitive restart option:

Code: Select all

**WAVE FUNCTIONS
.DFT
CAMB3LYP
.RESTART

and then provide the restart files through the dalton script using the -f option, e.g.:

Code: Select all

dalton -f scf-only.tar.gz -dal properties-only.dal -mol molecule.mol

[Addiw7]

Replace SIRIFC with NEWORB,

I tried that but then I need SIRIUS.RST instead of SIRIFC file and still the same issue appears unless I do .RUN PROPERTIES.
In the latter case, I need only one iteration for SCF to converge but SIRIUS still needs to be called. I will give a try to your last suggestion.

[magnus]

Ok, I misunderstood. In that case the restart will not help you either. I would assume that you would not need the **WAVE FUNCTIONS section but not sure if it makes a difference if it is not there.

[hjaaj]

Dawid, if you upload the three output files, I will find out what goes wrong.

[Addiw7]

Certainly, here are my logs.

If I use .RUN PROPERTIES together with .MOSTART SIRIFC I get this error message

SETORB FATAL ERROR Number of occupied orbitals bigger than total number of orbitals !

[hjaaj]

OK, I have now looked at it. The reason for the problem is that I originally implemented SIRIFC for MCSCF. Those who later implemented DFT did not add DFT information to the SIRIFC file (clearly not a priority for them then), therefore the DFT information is not transferred and one gets wrong results. Thus **PROPERTIES and **RESPONSE for DFT only works "together" with **WAVE FUNCTION, the DFT information is then transferred in memory. (By the way, I think your SIRIFC-wrong.out excitation energies therefore are Hartree-Fock excitation energies, based on the CAMB3LYP orbitals.)

So, until someone finds the time to program saving DFT information on SIRIFC and reading it back in, the only way is to not skip the SCF but restart the SCF from the converged orbitals, at the cost of one SCF iteration. By the way, you can also restart the SCF from SIRIFC, would give the same as the start from NEWORB.