UKS/UCCSD(T)

More
1 year 8 months ago #1255 by vvp
UKS/UCCSD(T) was created by vvp
Hi!

The using of Kohn-Sham DFT orbitals for CCSD(T) calculations is quite a common approach today. I wonder whether it is possible in MRCC to perform UCC calculations by using UKS DFT orbitals, rather UHF ones. If so, could you please consult me on how to perform such calculation? 

Thank you in advance!

With best regards, 
Victor 

Please Log in or Create an account to join the conversation.

  • nagypeter
  • Offline
  • Premium Member
  • Premium Member
  • MRCC developer
More
1 year 3 months ago #1296 by nagypeter
Replied by nagypeter on topic UKS/UCCSD(T)
Hi Victor!

Yes, you can compute UCCSD(T) with UKS orbitals by adding the
dft=PBE
setting, where the dft keyword sets the employed functional, such as PBE in this example.

Note that the additional term needed for non-HF orbitals in the (T) theory is only implemented for the open-shell case in the UCCSD(T) code, but not in the closed shell CCSD(T) code. So currently you would need to set the open-shell treatment in the case of KS orbitals via scftype=uks if your molecule is closed shell.

I hope this helps and sorry for the late response,
Peter

Please Log in or Create an account to join the conversation.

More
1 year 2 months ago #1313 by vvp
Replied by vvp on topic UKS/UCCSD(T)
Hi again!

Thanks for your reply. The hint provided works. However, there an additional route available is to use QRO:

basis=cc-pVDZ
calc=CCSD(T)
scftype=uhf
basis_sm=def2-SVP
mult=2
charge=-1
symm=off
scflshift=0.2
scfiguess=small
scfmaxit=200
scfdiis=on
ccprog=ccsd
dft=b3lyp
qro=on
core=1

The QRO orbitals gives slightly different results compared to canonical UKS. However, the energy difference might be caused by freezing core-electrons in CCSD(T). My guess is that for KS-CCSD(T) one must correlate all electrons to get consistent results.

With best regards,
Victor

Please Log in or Create an account to join the conversation.

  • nagypeter
  • Offline
  • Premium Member
  • Premium Member
  • MRCC developer
More
6 months 2 weeks ago #1390 by nagypeter
Replied by nagypeter on topic UKS/UCCSD(T)
Dear Victor!

Sorry for the late response. Just letting you know, that there was a small inconsistency in the treatment of the (T) terms with UKS and QRO references, which we fixed in the 2023 release.

For clarity, the QRO option makes a projection of the UHF/UKS solution which becomes exact spin eigenfunction, and then the QRO-s are quasi-canonicalized for the UCC computations. Thus, except for exactly closed-shell molecules, the QRO and UHF/UKS are not the same and the difference depends on the level of spin contamination.

I hope this helps,
Peter

Please Log in or Create an account to join the conversation.

Time to create page: 0.040 seconds
Powered by Kunena Forum