The ABINIT Users Mailing List ( CLOSED )

["Eric J. Walter" <ewalter@particle.physics.wm.edu>]

Dear Abinit list, 

(I wasn't sure if this should go to the developer list or the regular
list, sorry if I picked the wrong one).

I have a question about the electric field gradient functionality the
developer releases 5.4.0 and 5.4.2.  I think it is fantastic that this
functionality is in Abinit now!  

I have run the Shistovite calculation v5/test51 and get the following
good results compared to a converged LAPW calculation:

LAPW EFG eigenvalues (au)  :  -0.170,  -0.032,  0.203
Abinit EFG eigenvalues (au):  -0.167,  -0.041,  0.208

This is using the converged values listed in the INPUT file for ECUT
and NGKPT.

Indeed if I keep increasing the ECUT values, the results do not
change, however, if I increase the PAWECUTDG values above 50, the
results DO change and get further away from the LAPW result:

PAWECUTDG(Ha)    Vzz     Vxx      Vyy
30             -0.156  -0.046    0.202
35             -0.158  -0.046    0.203
40             -0.159  -0.045    0.204
45             -0.160  -0.044    0.204

50             -0.167  -0.041    0.208 

60             -0.168  -0.040    0.208
70             -0.176  -0.035    0.212
80             -0.183  -0.031    0.214
90             -0.190  -0.027    0.217
100            -0.198  -0.021    0.220

Moreover, it appears that only the efg_ion contribution is changing.
The efg_paw and efg_el parts converged from 30 Ha on.  Looking a
little at the code, it seems that the efg_ion term depends on a
Gaussian width (just as in the usual Ewald sum for the ions) which is
set based on the value of the fine FFT mesh (I think).  This seems to
me to be the only place that the efg_ion term could know about the
value of PAWECUTDG.  Regardless, it seems that this term should be
robust to the choice of the Gaussian width and therefore shouldn't
depend on the value of PAWECUTDG?

Perhaps I am just doing something wrong?

I would be happy to post more information or supply my input files in
someone is interested in resolving this.

Thanks for your attention to this matter.

Sincerely, 

Eric J. Walter
Department of Physics
College of William & Mary