Molcas Forum

meilani.wibowo

Member

Registered: 2015-11-25

Posts: 9

[SOLVED] Localisation

Hi all,
I am running calculations on a ethene dimer using CASSCF(4,4). The orbitals of this calculation are delocalized over the dimer. I wanted to localize the orbitals (HOMO and LUMO) into each ethene, thus I did localization calculations using both Pipek-Mezey and Boys methods but none of them gave me the localized orbitals that I am looking for. I checked the results by doing CASCI calculations using the output from localization (LocOrb) as the INPORB and the orbitals are still delocalized. Is there anyone who could help to explain why these orbitals are still delocalized?
Thank you.
Kind regards,
Meilani

Steven

Administrator

From: Lund

Registered: 2015-11-03

Posts: 95

Re: [SOLVED] Localisation

Are you using any point group symmetry?

---

Always check the orbitals.

meilani.wibowo

Member

Registered: 2015-11-25

Posts: 9

Re: [SOLVED] Localisation

I did not use any symmetry restriction.

Steven

Administrator

From: Lund

Registered: 2015-11-03

Posts: 95

Re: [SOLVED] Localisation

Are you localizing all orbitals or just the active orbitals?

If you localize only the active orbitals, you would need to have both the bonding+antibonding components to be able to localize them on 1 site. What are the 4 orbitals in your active space? Did you check if they are the pi orbitals? Did you plot them?

Unfortunately, I'm not that familiar with the localization module, did you try other types of localization such as Cholesky?

---

Always check the orbitals.

meilani.wibowo

Member

Registered: 2015-11-25

Posts: 9

Re: [SOLVED] Localisation

I only localised the active orbitals which are only the pi ornitals of each ethene (HOMO and LUMO). I always checked and plot them after each calculation has done.
Unfortunately, I only tried Pipek-Mezey and Boys, not the other methods.

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,012

Re: [SOLVED] Localisation

If I understand correctly, you localize the orbitals successfully (the LocOrb or local.molden files have correctly localized orbitals), but then you run a CASCI calculation (CASSCF with CIOnly?) and the orbitals you get after that are not localized. If this is correct, I think your problem is the RASSCF program is producing average delocalized orbitals anyway. Check the OutOrbitals keyword for other options, but I think all orbitals will be delocalized. In any case, the CAS energy is invariant to rotations within the active space, so once you've prevented rotations with the inactive and virtual spaces, the orbitals are in some way "equivalent", and the differences would only be in the CSF coefficients. I'm not sure how different input orbitals affect the CSF coefficients printed in the output, though...

meilani.wibowo

Member

Registered: 2015-11-25

Posts: 9

Re: [SOLVED] Localisation

Yes, you're right. My idea is first doing the CASSCF, checking the orbitals (delocalised orbitals) then localised the orbitals (RasOrb as the INPORB), finally doing CASCI (LocOrb as INPORB) to check if the orbitals are localized and how big the difference of the CSF coefficients using delocalised and localised orbitals. Because at some cases, it is easier to analyse different states using localised orbitals than delocalised ones. Of course the CAS energy is invariant under this transformation.

Steven

Administrator

From: Lund

Registered: 2015-11-03

Posts: 95

Re: [SOLVED] Localisation

I tested this with H2, and if I feed rasscf localized orbitals to do a CIONly calculation, the orbitals stay localized. At this point, I would recommend you to file a bug report so that you can attach input+output(+orbitals) together with the problem.

---

Always check the orbitals.

meilani.wibowo

Member

Registered: 2015-11-25

Posts: 9

Re: [SOLVED] Localisation

I know and understand now what was going on with my calculations. The localisation (4 orbitals) from the CASSCF(4,4) results using either Pipek Mezey or Boys gave me 4 localized pi orbitals in each monomer. But, my idea is to localized HOMO and LUMO of each monomer, thus what I did before is localisation of occupied and virtual orbitals from the active space. From these calculations, the occupied orbitals are localized but but not for the virtual ones. Therefore, the orbitals were not really localized as what I expected. Anyway, I solved the problem. I took the 4 localized pi orbitals from the localisation of 4 orbitals then write a simple code to localize them into HOMO and LUMO of each monomer. Thank you all for your response.

Steven

Administrator

From: Lund

Registered: 2015-11-03

Posts: 95

Re: [SOLVED] Localisation

Great. I marked the topic with [SOLVED].

---

Always check the orbitals.