[Pw_forum] Regarding post processing code (epsilon.x)

[Andrea Ferretti]

Dear Mohaddeseh,

please note that few bugs have been recently fixed in epsilon.x with QE release 
5.4.0. If you have been using any other version of QE you may want to give a 
try to this latest release

reading from the logs

------------------------------------------------------------------------
r12218 | giannozz | 2016-03-15 18:17:56 +0100 (mar, 15 mar 2016) | 2 lines

Incorrect sum over pools in LSDA case (courtesy of Mariella Ippolito, CINECA)

------------------------------------------------------------------------
r12082 | ferretti | 2016-01-30 10:19:26 +0100 (sab, 30 gen 2016) | 11 lines

Contributions from Tae Yun Kim and Cheol-Hwan Park (Seoul National University)

Fixed in epsilon.f90:
* Intraband contribution of nspin=2 was twice as much as that of nspin=1.
* epsr(w --> infinity) went to 2 in nspin=2.
* It was assumed that electronic occupations are in the range of 0-2 for 
nspin=2 (0-1 for nspin=1). 
   But with 'mp' and 'mv' smearing this is not true, and can give ~8 % 
underestimation of epsi
   in the case of sodium metal.
* Instead of directly calculating the derivative of Fermi Dirac occupation,
   now we use w0gauss function which is already in the QE modules.

------------------------------------------------------------------------

Some more precise info about the code can be found here:
~/espresso-5.4.0/PP/Doc/eps_man.tex


As a general point, if you have any dubts that your calculations are not 
working properly you should be more specific (this input generates this output 
which eg shows these problems)... asking whether a code is expected to provide 
correct results is not very effective

concerning the comparison of epsilon.x with ph.x (or linear response in 
general), one needs to be careful about the physical content of the 
calculations:

* DFT linear response as implemented in QE can compute the (static)
   response of a system of interacting electrons treated within a given
   XC-functional (let's consider ions clamped for the moment)

* epsilon.x computes the independent particle response (both static and
   dynamic) of a Kohn-Sham system (neglecting both local fields and fxc 
contributions)

* epsilon.x does not take the non-locality of the pseudopot into account
   (it works as if p were the physical momentum of the electrons)

hope it helps
Andrea

> 
> Regarding epsilon.x code in QE, I was wondering to know whether it will lead 
> to logical results?
> Is it possible to compare the static dielectric constant obtained by 
> epsilon.x code with that of ph.x code?
> I was wondering if you would introduce me the paper which has used 
> postprocessing code (epsilon.x).
> Thanks in advance for your help!
> 
> Regards,
> Mohaddeseh
> --
> ---------------------------------------------------------
> 
> Mohaddeseh Abbasnejad,
> Room No. 323, Department of Physics,
> University of Tehran, North Karegar Ave.,
> Tehran, P.O. Box: 14395-547- IRAN
> Tel. No.: +98 21 6111 8634  & Fax No.: +98 21 8800 4781
> Cellphone: +98 917 731 7514
> E-Mail:     m.abbasnejad at gmail.com
> Website:  http://physics.ut.ac.ir
> 
> ---------------------------------------------------------
> 
>

-- 
Andrea Ferretti, PhD
S3 Center, Istituto Nanoscienze, CNR
via Campi 213/A, 41125, Modena, Italy
Tel: +39 059 2055322;  Skype: andrea_ferretti
URL: http://www.nano.cnr.it