electron-phonon coupling as reported in School2018 tutorials

[SaraFiore]

Dear all,
I have to compute the electron - phonon coupling matrix element g and I was trying to follow the tutorial provided in School2018:
Electron-phonon coupling in QE - exercise 2.
It is reported :
if you are using your own compiled version of QE, copy the 4 Fortran files to the phonon folder and re-compile the code:
$ cp do_phonon.f90 elphon.f90 phq_readin.f90 symdyn_munu.f90 PATH/q-e/PHonon/PH
$ cd q-e/; make ph

My version is version 6.3 of Quantum-ESPRESSO (the one suggested in "Download & Install" section on epw.org.uk/Main/DownloadAndInstall).
Following this procedure I am not able to recompile. The error I get is

elphon.f90:492:25:

USE xml_io_base, ONLY : create_directory
1
Error: Symbol 'create_directory' referenced at (1) not found in module
'xml_io_base'

The module phq_readin.f90 give the same compile error against qe-6.3

How can I compute the electron-phonon coupling matrix elements as it is reported in the tutorial?

ibnd jbnd imode eig_i (eV) eig_j (eV) omega_nu (meV) |g| (meV)
...
4 4 6 9.359503 9.359503 95.957023 117.626599


thank you in advance for any help

Bests

[sponce]

Dear

As it is written on the school hands-on documents " Hands-on based on Quantum Espresso 6.2.2 ".

Therefore please use that version.

It is still possible to use the QE 6.3. You just have to change a few USE module statement.
For example the subroutine "create_directory" is no longer in the module xml_io_base but has been moved in a different module (try to grep in the code where it is).

Best wishes,
Samuel

[SaraFiore]

Dear sponce,
thank you very much for your fast answer, although...
Quantum Espresso 6.2.2 is not available on the download site, and its not tagged in the git repository either.
The following versions are tagged in the master Git repo:
git clone https://gitlab.com/QEF/q-e.git q-e

git tag
PW-1.2.0
PW-1.3.0
PW-1.3.1
qe-5.0
qe-5.0.1
qe-5.0.2
qe-5.1.0
qe-5.1.1
qe-5.1.2
qe-5.2.0
qe-5.2.1
qe-5.3
qe-5.4
qe-6.0.0
qe-6.1.0
qe-6.2.0
qe-6.2.1
qe-6.3
qe-6.3MaX
qe.6.3-rc2

6.2.1 is "close" but give this:

elphon.f90:21:28:

USE fft_interfaces, ONLY : fft_interpolate
1
Error: Symbol 'fft_interpolate' referenced at (1) not found in module 'fft_interfaces'
make: *** [elphon.o] Error 1



Can you please tell me where I can find 6.2.2?
Thank you very much in advance!

Best regards

[Nikita]

Dear Samuel and Sara,

It turns out that today I also have the same question. I tried to recompile ph module and get this error:
elphon.f90:21.28:

USE fft_interfaces, ONLY : fft_interpolate
1
Error: Symbol 'fft_interpolate' referenced at (1) not found in module 'fft_interfaces'
elphon.f90:337.28:

IF ( dffts%has_task_groups ) THEN
1
Error: 'has_task_groups' at (1) is not a member of the 'fft_type_descriptor' structure
elphon.f90:344.5:

ENDIF
1
Error: Expecting END SUBROUTINE statement at (1)
elphon.f90:389.34:

IF ( dffts%has_task_groups ) THEN
1
Error: 'has_task_groups' at (1) is not a member of the 'fft_type_descriptor' structure
elphon.f90:400.11:

ENDIF
1
Error: Expecting END DO statement at (1)
elphon.f90:403.37:

IF ( dffts%has_task_groups ) THEN
1
Error: 'has_task_groups' at (1) is not a member of the 'fft_type_descriptor' structure
elphon.f90:407.15:

ELSE
1
Error: Unexpected ELSE statement at (1)
elphon.f90:411.14:

ENDIF
1
Error: Expecting END DO statement at (1)
elphon.f90:460.28:

IF ( dffts%has_task_groups ) THEN
1
Error: 'has_task_groups' at (1) is not a member of the 'fft_type_descriptor' structure
elphon.f90:463.5:

ENDIF
1
Error: Expecting END SUBROUTINE statement at (1)
make[2]: *** [elphon.o] Error 1

Respectfully,
Nikita

[sponce]

Dear both,

I think the Q.E. 6.2.2 was a version but apparently indeed not tagged/saved.
You can just go to the commits around the time of the School.

For example:
commit acfcdc8c3411005e0ba89661fa1b8fd99bed6caa
Author: Pietro Bonfa <p.bonfa@nospam.com>
Date: Thu Mar 22 15:08:40 2018 +0100

Restored original CI.

which you can get by cloning the current gitlab version and then doing
git checkout acfcdc8c3411005e0ba89661fa1b8fd99bed6caa

Hope this works for you,
Best,
Samuel

[Nikita]

Dear Samuel,

Thank you for the link to the branch it worked for me.

Respectfully,
Nikita

[Vahid]

Dear EPW Users,

I have compiled the version of EPW (qe-6.2.2) used in the 2018 School including the 4 files for EPC calculation for semiconductors. In the Wed. 4 session, the g matrix is calculated for SiC using epw.x and ph.x (from an earlier session). The two results agree. In EPW calculation, polar scattering is included; however, in ph.x calculation, it seems that polar scattering is not included.

Is this because polar scattering in SiC is small or is there another reason for this agreement?

Thank you,
Vahid

Vahid Askarpour
Department of Physics and Atmospheric Science
Dalhousie University,
Halifax, NS, Canada

[carla.verdi]

Dear Vahid,
Direct ab initio calculations of the matrix elements using DFPT for selected q points, as done with ph.x in that example, include ALL kinds of e-ph interactions, including the polar one. When doing Wannier-Fourier interpolation starting from a relatively coarse BZ grid, the correct treatment to capture the polar coupling is to split the e-ph matrix element in a long and short-range part, interpolate the short-range part, and then add back the long-range one, as done by setting lpolar=.true. in EPW.
Hope this clarifies the issue,
Carla

[Vahid]

Dear Carla,

Thank you for this clarification. I need to hack the ph.x code to remove the polar term. Simply setting epsil=.false. in the phonon calculation is not apparently enough.

Best,
Vahid