CCSD(T) convergence issue

[Timothe Melin]

Dear Dirac users, 

I am trying to calculate the dissociation energy of LuF, to do so I do a CCSD(T) calculations with the dyall.ae2z, dyall.ae3z and dyall.ae4z and then converge to the basis set limit.

The results i obtained for ae2z and ae3z basis set are very far from each other D0=269 kcal/mol for dz and D0=343 kcal/mol for tz.

When I check my orbitals, they are the same and the electron correlated under NELEC are consistent between the molecule and the atoms alone. The only difference being to have the 5d electron in Lu but not in LuF.

My question are the following:

1) Is it possible for the 2 basis set to gives such different results

2) Is it possible to include a virtual orbital the Lu (5d) into the correlation. Like in the FSCC method but for CCSD(T) level?

I included both molecules output. If you need the atoms too let me know.

Best,

Timothe Melin

[Timothe Melin]

Oups... it looks like that I can not include my outputs

[Peterson, Kirk]

Dear Timothe,

 

not sure what exactly is wrong, but the dissociation energy of LuF is no more than 175 kcal/mol. This comes from (unpublished) CCSD(T)/CBS calculations in my group. Certainly you have to make sure you correlate the same electrons in the atoms as in the molecule but it sounds like if you correlated the 5d in Lu but not LuF this would make the dissociation energy too small, not too large. I'm guessing you've converged  to the wrong electronic state in Lu atom.  Either that or you have inconsistent virtual orbital cutoffs in Moltra.

 

regards,   -Kirk

--
You received this message because you are subscribed to the Google Groups "dirac-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/98151865-4ec7-45f1-b0f7-a9cd0d5ad560n%40googlegroups.com.

[Kenneth Dyall]

The 2z basis set is only semi-quantitative, so I would usually expect some difference. I think Kirk has a better explanation.

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/BYAPR01MB364057F7AB910BB965339829D6E30%40BYAPR01MB3640.prod.exchangelabs.com.

[Ilias Miroslav, doc. RNDr., PhD.]

Timothe,

put all your files into a cloud directory (like Google Drive), and share the link for reading.

Miro

---

From: dirac...@googlegroups.com <dirac...@googlegroups.com> on behalf of Timothe Melin <timothe...@gmail.com>
Sent: Monday, November 16, 2020 22:12
To: dirac-users <dirac...@googlegroups.com>
Subject: [dirac-users] Re: CCSD(T) convergence issue

 

Oups... it looks like that I can not include my outputs

--
You received this message because you are subscribed to the Google Groups "dirac-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/6b8b745c-6f2c-4d91-a63e-6c62b45ddb03n%40googlegroups.com.

[Timothe Melin]

Hey Miro, 

Here is the link with the outputs: https://drive.google.com/drive/folders/1g1gJy8-KoJNTJGeyx3AT1OxxepH6jeJN?usp=sharing

Thanks

[Timothe Melin]

Hi Kirk,

Thanks for the answer.

I agree with you that it should converge around 170-175 kcal/mol and that is what I have with CCSD(T) and DKH3.

I will look more into my atoms and Moltra see if I made a mistake somewhere.

I did check the electron correlation and in both atoms amd molecule the same orbitals are correlated with .NELEC for molecule and combination of .NELEC_gerade and .NELEC_ungerade

Best

Timothe

[Miroslav ILIAŠ]

Hi,

your LrF outputs are fine, converged, no errors.

But for the dissociation, LuF = Lu + F, you must have corresponding correlated shells in all three species - LrF, Lr and F. 

It should be possible to corelate Lu 5d orbitals in LuF, Lu. 

Note about open shells, Lu  [Xe].4f14.5d1.6s2  ( 2D3/2 ) and F (2P3/2) . Do you have proper, single determinant occupation of these in all your basis sets  ?  These affects the total energies.

 

Also, would be good to apply the basis set superposition error to De.

Best, Miro

[Timothe Melin]

Hi Miro, 

I added my atoms outputs for dz into the folder. 

I do have those 2 configuration for Lu and F. The 5d1 of Lu is spread among all 5d orbitals and same for the 2p5 of F.

The orbitals in both looks right and in the good order. 

I do not know if it is possible to correlate virtual orbital with RELCCSD with Dirac. The 5d electron of Lu is correlated but since in LuF this orbital is empty, it is part of the virtual. 

I know with FSCC it is possible to have virtual orbital into the correlated space but I do not think that you can get CCSD(T) energy with FSCC.

For BSSE. I will look into that.

Best,   

[Peterson, Kirk]

Dear Timothe,

 

everything looks fine to me. In fact, if I use the total energies from your output files (DZ basis), I get dissociation energies of 134.3 (DHF) and 152.8 [CCSD(T)] kcal/mol.

 

Ignore what I said previously about the 5d electron, I had 4d in my head for some reason. There is no problem with correlating the open 5d electron in Lu atom where of course it does not exist in LuF.

 

regards,  -Kirk

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/0c340986-f12c-4d2f-b815-d8b57fa05750n%40googlegroups.com.

[Timothe Melin]

Dear Kirk, 

It seems that I do not know how to copy paste... I did obtain 152.8 for the Dz basis set with the right energy. 

In the case of the Tz, I found the issue, I used a different basis set for Lu in the atomic calculations and I believed this is why the dissociation energy was so high.

Thank you everyone for looking at my files. It was a stupid mistake from my side.

Thanks again

Timothe