Author Topic: RI-CCSD(T) with no F12  (Read 6707 times)

skkim

  • Newbie
  • *
  • Posts: 3
  • Karma: +0/-0
RI-CCSD(T) with no F12
« on: August 19, 2016, 06:51:05 AM »
With the older 6.x version, I used to use RICC2 to obtain CCSD(T)/basis energy. However, the CCSD(T) calculation is not supported in the RI-CC2 module in the 7.x version. In the new version, CCSDF12 module is recommended for the CCSD(T) calculation. I am wondering how RI-implemented CCSD(T) (without the F12 option) is calculated.

gunnar.schmitz

  • Developers
  • Jr. Member
  • *
  • Posts: 13
  • Karma: +0/-0
Re: RI-CCSD(T) with no F12
« Reply #1 on: August 19, 2016, 10:34:31 AM »
Dear skkim,

I guess you are wondering if the RI approximation is used for all 4 index integrals? This is not the case. If the RI approximation does not lead to more efficiency, conventional integrals are used. Turbomole calculates them in an integral direct fashion to save I/O. Some other codes switch for small systems to an algorithm where the MO integrals are precalculated and are therefore for this small system faster. The RI approximation  is used for the ring terms (C and D terms) which involve contractions of the doubles amplitudes with integrals with two occupied and two virtual indeces. Also the MO integrals in the triples correction are calculated with the RI approximation. Therefore CCSD(T) results with Turbomole might be slightly different from other codes where all the way down conventional integrals are used, but the errors are usually in the same range as the RI error in RI-MP2 calculations. 

I hope that was what your question was aiming for?

skkim

  • Newbie
  • *
  • Posts: 3
  • Karma: +0/-0
Re: RI-CCSD(T) with no F12
« Reply #2 on: August 21, 2016, 12:40:51 AM »
Thanks.

Then, how is the CCSD(T) calculation with no F12 done?

gunnar.schmitz

  • Developers
  • Jr. Member
  • *
  • Posts: 13
  • Karma: +0/-0
Re: RI-CCSD(T) with no F12
« Reply #3 on: August 22, 2016, 09:16:58 AM »
Okay, then I probably did your question wrong. You want to know how a CCSD(T) calculation is set up and executed? If you already did it with older Turbomole versions, what your questions implies, it remains nearly the same you only have to call the ccsdf12 program instead of ricc2.

  • In define you still go to the cc menu.
  • Select the RI basis with cbas
  • Choose there the ricc2 menu
  • and use "model ccsd(t)" to select CCSD(T).

or you edit your control file (after assigning a RI basis) and add the section

$ricc2
  ccsd(t)

or

$ccsdf12
  ccsd(t)

After that you run the ccsdf12 program. It should be rather simple. If it fails, I recommend that you contact the Turbomole Support.

I hope I did understand what you were asking for this time. If this is not the case please be a bit more precise. 

skkim

  • Newbie
  • *
  • Posts: 3
  • Karma: +0/-0
Re: RI-CCSD(T) with no F12
« Reply #4 on: August 23, 2016, 05:03:45 AM »
It worked O.K. Thank you very much for your help.

S. K. Kim