Choosing ccprog=ccsd via CFour interface

Dear Prof. Kallay,

For a large CCSD(T) calculation via CFour interface, is it possible to choose the ccsd module to run the calculation? The computational size is somewhat huge, and I would need efficiently parallelized execution using OpenMP parallelism. It seems that the default ccprog via CFour interface is the string-based mrcc module, and it requires ABCDTYPE=STANDARD in CFour to generate the fort.5x files correctly; using ABCDTYPE=AOBASIS caused CFour to die at the stage of xint.

Also I would like to learn that which one of ccsd and mrcc modules is faster for parallel CCSD(T) calculation with about 500~600 basis functions? Thank you for your kind advise.

Jen-Shiang

Dear Jen-Shiang,
For CCSD(T) calculations it is not worth using the the interface, you should use either cfour or mrcc in standalone mode.
For CCSD(T) calculations you should use the ccsd module, the mrcc module is designed for higher-order CC calculations. You can considerably accelerate the calculation using density fitting (calc=DF-CCSD(T)).

Dear Prof. Kallay,

Thank you for your response. I just figured out that ccprog=ccsd only supports closed-shell calculations, but my system is open-shell singlet and I might need to use ccprog=mrcc regardless of ROHF or UHF reference.

RETURNING FROM SCF ALGORITHM
======================================================================
************************ 2018-02-01 21:57:37 *************************
Executing ovirt...

ovirt, the routine of orbital optimization and integral transformation
Thu Feb 1 21:57:37 CST 2018
ccprog=ccsd is not supported for UHF calculations.
Please, use the ccprog=mrcc keyword!

Fatal error in ovirt.
Program will stop.

************************ 2018-02-01 21:57:37 *************************
Error at the termination of mrcc.
**********************************************************************

Dear Jen-Shiang,
For an open-shell singlet state you should not use CCSD(T). You should try some multi-reference method.

I was aware of this issue because the system is a singlet biradical. Just that our paper reviewer chose not to believe the energy profile from NEVPT2 theory and insisted that CCSD(T) "with a large basis set" should be performed. Thanks anyway for your comment.