<div dir="ltr">Dear Andrea,<div>It would be great if you could share the new epsilon.f90 code. If it takes time to merge it with the develop branch, can you put it in a separate one for now?</div><div>Best regards,</div><div>Michal Krompiec</div><div>Merck KGaA, Darmstadt, Germany</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Wed, 8 Jul 2020 at 20:31, Andrea Ferretti <<a href="mailto:andrea.ferretti@nano.cnr.it">andrea.ferretti@nano.cnr.it</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
Dear Anibal,<br>
<br>
you are right, it seems calculation="occ" is documented but no longer <br>
there. As far as I understand, the reason is exactly that it is preferred <br>
to directly compute the quantity of interest and perform convergence <br>
checks on it (epsilon.x should be fast enough, though).<br>
<br>
Regarding the anysotropy: as far as I remember epsilon.x does not <br>
implement symmetries, meaning that kpts need to span the whole BZ.<br>
If you run a scf and a nscf calculation using pw.x and does not pay <br>
attention to this (meaning you have not set nosym=.T. noinv=.T.), kpts <br>
will be symmetrized and only the IBZ wedge will be sampled. In turn this <br>
can lead to spurious anysotropy (besides non-correct results).<br>
<br>
hope it helps<br>
Andrea<br>
<br>
BTW: I have a newer version of epsilon.f90 contributed by Tae-Yun Kim <br>
(Seoul National University, South Korea) which fixes a number of these <br>
issues. Just haven't found the time to include it in the official distribution.<br>
<br>
> <br>
> The epsilon.x manual (in the PP/DOC folder) shows the possibility of calculating occupations using the key "occ" within the epsilon.x. It is<br>
> emphasized to be a good tool to analyze convergency against the broadening parameter and the k points sampling. However, the "occ"<br>
> calculation is not implemented (at least in the version I'm using - 6.4). Such a calculation was implemented with other packages?<br>
> <br>
> If not, is there a way of verifying the convergence other than explicitly changing the broadening and k points sampling?<br>
> <br>
> I'm working with an AuAl alloy, trying to evaluate the dielectric function. Using epsilon.x I've got anisotropic behavior that I was not<br>
> expecting for. Working with pure systems (Au and Al) I concluded that reducing conv_thr increases the epsilon.x output precision, returning<br>
> the isotropic behavior of the dielectric function. Therefore, to the alloy (with 12 atoms in the cell), I increased both the conv_thr (1e-13)<br>
> and k points (14 14 14). I still got the anisotropy. Should I go further (calculations with my actual computing power are becoming very time<br>
> and memory consuming)? <br>
> <br>
> Thanks in advance!!<br>
> <br>
> Anibal Bezerra<br>
> The Federal University of Alfenas<br>
> <br>
><br>
<br>
-- <br>
Andrea Ferretti, PhD<br>
S3 Center, Istituto Nanoscienze, CNR<br>
via Campi 213/A, 41125, Modena, Italy<br>
Tel: +39 059 2055322; Skype: andrea_ferretti<br>
URL: <a href="http://www.nano.cnr.it" rel="noreferrer" target="_blank">http://www.nano.cnr.it</a><br>
_______________________________________________<br>
Quantum ESPRESSO is supported by MaX (<a href="http://www.max-centre.eu/quantum-espresso" rel="noreferrer" target="_blank">www.max-centre.eu/quantum-espresso</a>)<br>
users mailing list <a href="mailto:users@lists.quantum-espresso.org" target="_blank">users@lists.quantum-espresso.org</a><br>
<a href="https://lists.quantum-espresso.org/mailman/listinfo/users" rel="noreferrer" target="_blank">https://lists.quantum-espresso.org/mailman/listinfo/users</a></blockquote></div>