[QE-users] ph.x from v6.3 reads wrong celldm

[Prof. Raphael Longuinhos Monteiro Lobato]

Hi,

Back on the celldm post (sorry for the six months delay to reply ... I 
will bring back some parts to remember):
http://lists.quantum-espresso.org/pipermail/users/2018-October/041418.html

and

http://lists.quantum-espresso.org/pipermail/users/2018-October/041419.html

Previously, I wrote:
+==
ph.x from pre-release q-e-qe-6.3Max reads from outdir unexpected values 
for celldm and writes these in the header of .dyn . It leads to 
"rescaled" distances in .axsf and polarizability in the dynmat.x out 
(omega, IR, Raman tensor and depolarization factor are as expected). 
Hand-correcting the .dyn header (correct values for celldm) solves it.

Tests in v6.2 and 6.2.2 leads to expected results (they read from outdir 
the expected values for celldm). ph.x prints wrong celldm while reading 
from outdir ... it is not related to io_dyn_mat write_dyn_mat_header ... 
also, the .xml files are the same.
+==

Prof. Giannozzi mentioned that
https://gitlab.com/QEF/q-e/commit/9a579539b65bd1502d1f28747e483a1d1059f5f8

could have solved the issue. Maybe I'm missing something, but it appears 
it did not.

The issue on the celldm can lead to other unexpected differences on 
post-processing the dyn.

In my case, the high-frequency (low frequency is okay) phonon dispersion 
near Gamma calculated with the original (wrong) celldm values and the 
ones using hand-correcting values are somewhat different. As example, 
the high-frequency LO mode obtained with the wrong cell_dm seems 
quadratic (or some d^2f/dk^2 >0 dependency) like using no LR 
electrostatics, while it becomes linear-like (I would expect some thing 
like that) when using the correct cell_dm. It is show in the images in 
appended. Also, there are some "discontinuities" in the branches I not 
used to see on results. Also, there are some extra inflection points in 
the region I show in the graphs that I would not expect (these appears 
using no LR electrostatics).

Note: The system is expected to be polar semiconductor. In the graphs nq 
07x07x01 and 14x14x01 uses nk is 14x14x1 (spacing of ~0.03*2pi/A). 
Later, I used both nk and nq of 16x16x01. Orange diamonds are direct 2D 
DFPT calculations (nk 14x14x01 or 16x16x01 leads to identical results in 
these direct 2D DFPT calculations). The highest frequency modes from 
direct 2D DFPT and from interpolation of dynmat are some what different 
(maybe the interpolation is not converged). Data with cell_dm corrected 
values are also indicated.

I will repeat all these using those versions I reported correctly 
reading cell_dm values.

Hope all it is not a "false alarm" (some unnoticed error of my part). 
Sorry bring problems, not solutions.

Best,

-- 
Associate Professor Raphael Longuinhos Monteiro Lobato
CV: http://lattes.cnpq.br/1475483971471395.
Research Group site: http://jenainassoares2.wixsite.com/nanomat (Portuguese only - English version on the way)
Departamento de Fisica, Universidade Federal de Lavras
Av. Dr. Silvio Menicucci 1001, bairro Kenedy
Post office box 3037
37200-000 - Lavras, MG, Brazil

-------------- next part --------------
A non-text attachment was scrubbed...
Name: f4.png
Type: image/png
Size: 32753 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20190313/a4c79a99/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: f3.png
Type: image/png
Size: 40974 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20190313/a4c79a99/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: f5.png
Type: image/png
Size: 43205 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20190313/a4c79a99/attachment-0002.png>