[Pw_forum] inconsistent quadrupolar coupling from gipaw calculation

[Davide Ceresoli]

...I wasn't able to obtain reasonable results for Co3O4 with elk.
Therefore I've run an all-electron calculation with an uncontracted
basis set with ORCA, for the Co(CO)$_4$H complex.

Here are the results (Vzz component in Hartree/bohr^2):

atom	 ORCA 	 NCPP	 NC-semi PAW+NC	 PAW+US
-----------------------------------------------
Co	 1.0747	 1.0368	 0.9011	 1.0684	 1.0526
H	 0.1372	 0.1343	 0.1353	 1.0821	 1.0797
C	-0.7097	-0.7821	-0.7891	-0.2583	-0.2306
C	-0.7100	-0.7825	-0.7893	-0.2585	-0.2307
C	-0.7100	-0.7825	-0.7893	-0.2585	-0.2307
C	-0.7176	-0.7880	-0.8000	-0.2700	-0.2430
O	-0.2675	-0.2194	-0.2307	 0.1459	-0.1818
O	-0.2681	-0.2201	-0.2320	 0.1446	-0.1833
O	-0.2681	-0.2201	-0.2320	 0.1446	-0.1833
O	-0.2651	-0.2121	-0.2286	-0.1086	-0.1802

NCPP: Co.pbe-tm-gipaw-dc.UPF  C.pbe-tm-gipaw-dc.UPF  H.pbe-tm-gipaw-dc.UPF 
O.pbe-tm-gipaw-dc.UPF
NC-semi: Co.pbe-tm-semi-gipaw.UPF  C.pbe-tm-gipaw-dc.UPF  H.pbe-tm-gipaw-dc.UPF 
O.pbe-tm-gipaw-dc.UPF
PAW+NC: Co.pbe-paw-gipaw-nh.UPF  C.pbe-tm-gipaw-dc.UPF  H.pbe-tm-gipaw-dc.UPF 
O.pbe-tm-gipaw-dc.UPF
PAW+US: Co.pbe-paw-gipaw-nh.UPF  C.pbe-rrkjus-gipaw-dc.UPF 
H.pbe-rrkjus-gipaw-dc.UPF  O.pbe-rrkjus-gipaw-dc.UPF

It seems now that the Co EFG does not depend too much on Co pseudopotential.
However, the choice of Co pseudo appears to influence dramatically the EFG of
the neighboring atoms. The NC pseudos are in better agreement with ORCA.

Uhm... the plot thickens... this is very strange. I need to investigate more.
I hope it's not a bug.

Davide



On 10/09/2017 02:13 PM, Davide Ceresoli wrote:
> Dear Jia,
>      I have to admit that this is the first time that NMR/EFG results
> depend so much on the choice of the pseudopotential. I've calculated
> a bunch of minerals, both with NC and US pseudos, and results are
> in good agreement each other.
> 
> I've tested several Co pseudos (NC, NC+semicore, PAW+semicore) on
> Co3O4 spinel and the EFG results span the entire real numbers range,
> both for Co and for O.
> 
> Which one is correct? I don't know. I'm in favor of pseudopotentials
> with semicore states. They should be closer to all-electron.
> 
> Do you have some reference with Co EFG calculated all-electron (wien2k,
> elk/exciting) on some simple system? If not, is there someone that
> could help us to setup a wien2k/elk calculation?
> 
> Best regards,
>      Davide
> 
>