[Pw_forum] Questions related to Co and Fe Pseudopotentials in pslib 1and 0.31

[Andrea Dal Corso]

On Mon, 2015-04-27 at 15:20 +0000, Mostafa Youssef wrote:
> Dear all,
> 
> I have few questions related to the above mentioned PP's.
> 
> (1)  After generating the PAW and USPP versions of these PP (pslib1.0.0) using ld1 code provided with  q.e.5.1.2, I applied the test script.  The output seems to be fine in the case of USPP but for PAW extra few lines are printed. For example, in the case of LDA scalar-rel Fe, I get:
> 
> Making   Fe.pz-n-kjpaw_psl.0.2.4.test.in  ...  Done !
>      3 2     3D   1( 7.00)       -0.41920       -0.41398       -0.00522  !
> Making   Fe.pz-n-kjpaw_psl.1.0.0.test.in  ...  Done
> Making   Fe.pz-n-rrkjus_psl.0.2.4.test.in  ...  Done
> Making   Fe.pz-n-rrkjus_psl.1.0.0.test.in  ...  Done
> Making   Fe.pz-spn-kjpaw_psl.1.0.0.test.in  ...  Done
> Making   Fe.pz-spn-rrkjus_psl.1.0.0.test.in  ...  Done
> 
> And in the case of  Fe-rel I get:
> 
> Making   Fe.rel-pz-n-kjpaw_psl.0.2.4.test.in  ...  Done !
>      3 2 1.5 3D   1( 4.00)       -0.42153       -0.41623       -0.00530  !
>      3 2 2.5 3D   1( 3.00)       -0.41137       -0.40618       -0.00519  !
> Making   Fe.rel-pz-n-kjpaw_psl.1.0.0.test.in  ...  Done
> Making   Fe.rel-pz-n-rrkjus_psl.0.2.4.test.in  ...  Done
> Making   Fe.rel-pz-n-rrkjus_psl.1.0.0.test.in  ...  Done
> Making   Fe.rel-pz-spn-kjpaw_psl.1.0.0.test.in  ...  Done
> Making   Fe.rel-pz-spn-rrkjus_psl.1.0.0.test.in  ...  Done
> 
> Similar output obtained for Co PAW.  Are these extra printed lines indicative of some error in the generation?
> 

The atomic code puts a ! when the difference between all-electron and
pseudo eigenvalues is larger than 5 mRy and the test script searches and
prints those lines. This indicates that the transferability errors of
the Fe pseudopotential of version 0.2.4 was slightly larger than the
threshold. 


> 
> (2) The charge density cutoff recommended (in the UPF files) for these PP is about 12 *ecutwfc. This large cutoff for the charge density was also employed in a recent paper using Fe (pslib0.31) (http://arxiv.org/pdf/1502.01534.pdf) . However, when I did convergence test for the total energy on a unit cell of SrFeO3 , and SrCO3, and it looks like 4* ecutwfc is sufficient to achieve convergence. The convergence in the case of SrFeO3 (pslib0.3, USPP) can be found here (http://web.mit.edu/myoussef/www/Fe_16e_pslib.pdf).  The pink bound indicates that convergence was achieved within 25 mev/unit cell (5 meV/atom)  at ecutwfc=85 Ry and ecutrho =4 *ecutwfc. It seems to me that the recommend ecutrho it too strict.  Are there other tests that one needs to perform to check the convergence of charge density cutoff?
> 
The estimate of ecutrho written in the PP file is done using the largest
q vector that is used in the Bessel function expansion of the
pseudization function (ecutrho=2|q|^2). In the case of the charge it
refers to the largest q used to pseudize all the components of the
augmentation Q function. Usually it is a reasonable estimate, but it
remains an indication. The KE cut-off depends also on the quantity that
you are using to check the convergence. 


> (3) The recommended ecutwfc and ecutrho for this set of PP in Pslib do not change from PAW to USPP and do not depend on the functional. I did preliminary tests  and it seems so far that the values work for GGA also work for LDA consistent with the recommendation. I wonder if one could test convergence of ecutwfc and ecutrho for LDA only and apply the outcome to all other semilocal functionals, or are there cases where one encounters different convergence depending on the functional.
> 

It is usually like this (more or less independent on the functional),
but not always. In some cases the PP works for some functionals and not
for others. Maybe one can search PPs for which the KE cut-off is more or
less independent on the functional, but this has not been checked in
pslibrary and I would not be surprised if some PPs do not satisfy this
condition.

HTH,

Andrea

> 
> Thank you in advance for reading and help!
> Mostafa Youssef
> MIT
> 
> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum

-- 
Andrea Dal Corso                    Tel. 0039-040-3787428
SISSA, Via Bonomea 265              Fax. 0039-040-3787249
I-34136 Trieste (Italy)             e-mail: dalcorso at sissa.it