Pseudopotential problems

[Yun_Pen_Wang]

Hi,

I am trying to use GAMESS to generate orbitals. In the GAMESS pseudopotentials, only ECP are defined. What basis set should be used together with those GAMESS pseudopotentials?
Before using any pseudo potential, it should be tested. The usual test case is the atomic ionization energy. So what is the accuracy we are expected to use Trail & Needs pseudopotentials for transition metals?

best wishes,
Yun-Peng Wang

[Mike Towler]

Hi Yun-Peng,

In the CASINO library of Trail-Needs pseudopotentials, corresponding Gaussian basis sets are given - for atoms H-Ne only - in formats appropriate for the CRYSTAL and GAUSSIAN programs. The only reason that the same basis sets are not given in the GAMESS format is that none of us have bothered to convert them (probably because we're not really GAMESS users). If someone who understands how GAMESS works wants to convert them, I am happy to incorporate them into the online table.

I might add a brief explanation of one possibly confusing point: the 'original pseudopotential' is the one given in the 'Tabulated representation' i.e. as numerical values on a set of grid points in the format that CASINO itself understands. Gaussian basis set quantum chemistry programs require that this pseudopotential is itself expanded in a basis set of Gaussians times powers of r. The GAUSSIAN and CRYSTAL representations are identical and differ only in the file format; the GAMESS representation is slightly different as this program artificially restricts the powers of r you can use so you end up with a 'slightly worse' representation of the original tabulated potentials compared to the GAUSSIAN and CRYSTAL files. However, they will be similar enough that you can use the same basis sets for all three codes - again differing only in file format.

I understand that Albert Defusco is fiddling with the GAMESS code base so that this artificial restriction may be removed in the near future.

If you want a basis set for an atom other than H to Ne, then we don't yet supply them simply because no-one has got around to generating any yet. I know we really should and this remains an important item on our 'Things to Do' list. Until then you can either generate your own - see the procedure in:

"Correlation Consistent Gaussian Basis Sets for H, B-Ne with Dirac-Fock AREP
Pseudopotentials: Applications in Quantum Monte Carlo Calculations"
J. Xu, M.J. Deible, K.A. Peterson and K.D. Jordan
Journal of Chemical Theory and Computation, 9, 2170-2178 (2013)

or you can ask someone nicely to do it for you..

Plane waves are so much easier aren't they?

Best wishes,
Mike

PS: see the original Trail-Needs papers (and some more recent ones) for your question about accuracy.

[Vladimir_Konjkov]

To make life easier for ORCA users, I create files with ECP and aug-cc-pVXZ-CDF bases in ORCA format for all available elements.

DF_AREP_Trail_Needs_gamess.bas - DF AREP Trail Needs ECP basis generated from gamess one ('slightly worse' representation)
DF_AREP_Trail_Needs_gaussian.bas - DF AREP Trail Needs ECP basis generated from gaussian one
aug-cc-pVDZ-CDF.bas - aug-cc-pVDZ-CDF basis for H, B, C, N, O, F, Ne
aug-cc-pVTZ-CDF.bas - aug-cc-pVTZ-CDF basis for H, B, C, N, O, F, Ne
aug-cc-pVQZ-CDF.bas - aug-cc-pVQZ-CDF basis for H, B, C, N, O, F, Ne
aug-cc-pV5Z-CDF.bas - aug-cc-pV5Z-CDF basis for H

All files autogenerated by script.

details of the use of bases discussed in https://orcaforum.cec.mpg.de/viewtopic.php?f=8&t=1609

Vladimir.