LetzElPhC

[rishabhsrsw7]

Greetings forum,

I am trying to generate the elph_gkkp_* elements using LetzElPhC and yambopy, following the tutorial https://wiki.yambo-code.eu/wiki/index.p ... minescence. Dvscf calculations generate a ph_save and _ph0 folders. ph_save has one "dvscf1" file, which from my lelphc.log, I see that LetzElPhC uses to generate the gkkp elements.

My issue is: in ph_save only dvscf1 is generated, so the LetzElPhC calculation stops after the first q point, however, in _ph0 I see dvscf files at every q point, should I be copying those files into ph_save and renaming them dvscf1, 2, 3... etc?

Regards

[batman]

Dear Rishabh,

Can you try the following?

first delete the existing ph_save

Go to the directory where you performed your phonon calculations and run the following command

Code: Select all

lelphc -pp -F PH.X_input_file

PH.X_input_file is the input file of ph.x

You should see ph_save directory and check if all dvscf and dyn* are present.

Now perform your electron phonon calculation using the following command

Code: Select all

mpirun -n 2 lelphc -F lelphc.in

lelphc.in is the letzelphc input file

Code: Select all

nkpool          = 1
# Number of MPI pools for k-point parallelization
nqpool          = 1
# Number of MPI pools for q-point parallelization
start_bnd       = 1
# First band to include in the calculation
end_bnd         = 40
# Last band to include in the calculation
save_dir        = SAVE
# Path to the Yambo SAVE directory
ph_save_dir     = ph_save
# Path to the phonon save directory generated by the preprocessor
convention      = yambo # !!!!! ALways set this to yambo  when generate yambo gkkp databases !!!!!

Please go to the following link for more details (https://yambo-code.github.io/LetzElPhC/ ... age/usage/)

once you have the ndb.elph file, you can use yambopy to generate the databases

Code: Select all

from yambopy import LetzElphElectronPhononDB,ConvertElectronPhononDB

print(":: Load el-ph database ::")
lelph_obj = LetzElphElectronPhononDB('ndb.elph',div_by_energies=False)

print(":: Convert el-ph data to yambo format ::")
l2y       = ConvertElectronPhononDB(lelph_obj,'lelphc','SAVE','SAVE')

and check your save folder.

Let me know if it worked. if not please paste the warnings/errors and a sample input file to if we reproduce this behaviour.

Best Regards,
Murali

[rishabhsrsw7]

Dear Murali,

I generated the ph_save, it still generates only dvscf1 inside, all the dyn* files are however present.

I do get a warning when I generate the ph_save, which says

Warning : ldisp is set .false. in the ph.x calculation. Make sure dyn0 file is present with k-point compatible q-point grid.

Following this, the elph calculations stop after the first q point because no other dvscf files are present. Do I need to set ldisp to true for all the dvscf files?

Regards and Thank you

[batman]

Dear Rishabh,

It is always suggested to set ldisp=.true. and use the automatic uniform grid (that is commensurate with the k grid used in yambo) from Quantum ESPRESSO (this is much easier than manually specifying the points and less error prone).

In any case, since you already have the phonons, I suggest copying the following files manually to ph_save:

Code: Select all

data-file-schema.xml        from prefix.save
{pseudo_file}               all pseudopotentials in prefix.save
dyn0                        # you do not have this file when ldisp=.false. so you need to create this file (see below)
dyn1, dyn2, ...             dynamical matrix of each q-point in your list
dvscf1, dvscf2, ...         dvscf matrix of each q-point in your list
patterns.{i}.xml. # in _ph0/prefix.phsave/
tensors.xml # in _ph0/prefix.phsave/ (ignore this file if not present in the directory)

You need to create the dyn0 file manually in the following format and put it in ph_save:

Code: Select all

nq1  nq2  nq3
nq_iBZ
   qx   qy   qz
   qx   qy   qz
   ...

Where:
- nq1 x nq2 x nq3 is the uniform q-grid that is commensurate with k grid used in yambo
- nq_iBZ is the number of q-points in the irreducible Brillouin zone (iBZ) that you asked ph.x to compute
- The next nq_iBZ lines are the q-points in Quantum ESPRESSO units (2pi/alat). These are the same values you specify in your QE input file, be careful with the units (they must be in 2pi/a0).

[rishabhsrsw7]

Dear Murali,

Thanks a lot for your prompt help with this. I will do this and I believe this solves all the errors.

Regards

[rishabhsrsw7]

Hi Murali,

I did transfer all the dvscf files, and everything else. The run started fine, but when it starts to calculate the el-ph matrix elements at q=2, I got this error:

# [ Error !!!] : File : symmetries/rotate_eig_vecs.c, in function : rotate_eig_vecs at line : 71

Error msg : Unable to find the equalivalent atom for the symmetry operation. Wrong symmetry operation detected.

*************************************

All the atomic positions, celldm, and everything else is perfectly fine with scf and nscf runs. How do I solve this issue?


Regards

[batman]

This is happening because the Letz code is unable to find the equivalent atom under the symmetry operations when performing the rotation of phonon displacement vectors (as the errors says).

A very likely cause is that the atomic positions are not specified with sufficient precision. For example, if you set a coordinate to 0.3333330000 instead of the more precise 0.3333333333, QE is still able to identify the symmetry because it uses a tolerance of 1e-5. However, the Letz code uses a stricter tolerance of 1e-6, so it may fail to find the equivalent atom.

I have patched the code by relaxing the tolerance to 1e-5. Could you please clone the repository again from scratch (git clone https://github.com/yambo-code/LetzElPhC) and recompile it (make sure to simply reuse your make.inc to avoid setting it again)?

Let me know if this fixes the issue.

[rishabhsrsw7]

Dear Murali,

Thanks a lot, that worked perfectly. It created the ndb.elph database, and the gkkp files.

Regards