Computing the relaxed ion elastic tensor

Phonons, DFPT, electron-phonon, electric-field response, mechanical response…

Moderators: mverstra, joaocarloscabreu

Locked
Salah
Posts: 7
Joined: Mon Oct 12, 2020 5:33 am

Computing the relaxed ion elastic tensor

Post by Salah » Tue Dec 01, 2020 8:19 pm

Dear All,
I have been trying to compute the relaxed ion elastic tensor. I used DFPT for computing the shear and normal stress/strains components.

The relevant contents of the DFPT input file are:
rfstrs 3 # Response Function with respect to STRainS;

rfdir 1 1 1 # Response Function: DIRections; Gives the directions to be considered for response function calculations;

rfddk 1 # Response Function with respect to Derivative with respect to K; (d/dK);

rfatpol 1, 6 # The 1st element of the vector rfatpol, Response Function: ATomic POLarisation,
# & the 2nd element of the vector rfatpol, Here natom=6

Then, I used the "anaddb" utility for computing the elastic tensor. I am attaching hereto the input file and the output file on the "anaddb" utility.

The contents of the ".files" file for the "anaddb" utility are:
h-Primitive.in
h-Primitive.out
h-Primitiveo_DDB
dummy_moldyn
dummy_GKK
5b.3.hgh
7n.5.hgh
6c.4.hgh

As can be gathered from the output file of computing the elastic tensor using the "anaddb" utility, only the "clamped ion elastic tensor" is computed. The "relaxed ion elastic tensor" appears with all of its elements "Nan".

Thus, could you advise me on the reason behind that "Nan" suppression of the computation of the "relaxed ion elastic tensor" and help me in resolving this issue? I thankfully look forward to your consideration and reply at your earliest convenience.
With thanks and kind regards,
Salah

ebousquet
Posts: 469
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: Computing the relaxed ion elastic tensor

Post by ebousquet » Sun Dec 06, 2020 5:38 pm

Dear Salah,
How the SCF of each DFPT calculations are converged?
Best wishes,
Eric

Salah
Posts: 7
Joined: Mon Oct 12, 2020 5:33 am

Re: Computing the relaxed ion elastic tensor

Post by Salah » Mon Dec 07, 2020 11:22 am

Dear Eric,
Thank you for your reply and consideration. You are most welcome. I highly appreciate your valuable contributions to the Abinit Forum and your insightful feedback.

I have scrutinized the output file of the DFPT calculations and have found that the SCF in the DFPT has not converged. The following warning message appears in the output file:

scprqt: WARNING -
nstep= 3000 was not enough SCF cycles to converge;
maximum residual= 8.231E-02 exceeds tolwfr= 1.000E-10

Thus, although the current value of the “nstep” seems to be large enough for reaching convergence and the value of “tolwfr” seems to be reasonable enough for reaching convergence, the SCF convergence in the DFPT has not been reached. For your convenience, I am attaching hereto the DFPT input file and output file. It is noteworthy that I have removed some parts of the output file and have changed the extension of its name in order to enable this file to get through the online attachments constraints. I thankfully look forward to further hearing from you in this regard at your earliest convenience. I look forward to resolving this issue.

With thanks and kind regards,
Salah

mmazdz
Posts: 5
Joined: Fri Sep 13, 2019 9:13 am

Re: Computing the relaxed ion elastic tensor

Post by mmazdz » Mon Dec 07, 2020 9:35 pm

You need in ABINIT input also: rfphon 1 #do atomic displacement perturbation
Best wishes,
Marcin

Salah
Posts: 7
Joined: Mon Oct 12, 2020 5:33 am

Re: Computing the relaxed ion elastic tensor

Post by Salah » Tue Dec 08, 2020 6:55 pm

Dear Marcin,
Thank you for your reply and consideration. I highly appreciate your suggestion and your attempt to help in solving that issue.

I have scrutinized the keyword “rfphon” and have found that it is basically used in “Response Function with respect to PHONons” ( https://docs.abinit.org/variables/dfpt/#rfphon ). This is not the case here where I am seeking here to compute the relaxed ion elastic tensor. Actually, I am seeking here to compute the elastic constants, rather than phonons. In fact, I have even tried to add this keyword “rfphon” with the value “1” in the input file for the “ANADDB” computational utility, but the issue of the resulting “Nan” in the relaxed ion elastic tensor has remained the same.

Therefore, I hope you may kindly refine your suggestion for solving that issue in the direction of computing the relaxed ion elastic tensor, rather than in the direction of Phonons computations. I thankfully look forward to your consideration as well as to the consideration of the experienced contributors to the Abinit Forum for finding the root cause of that issue.
With thanks and kind regards,
Salah

Salah
Posts: 7
Joined: Mon Oct 12, 2020 5:33 am

Re: Computing the relaxed ion elastic tensor

Post by Salah » Wed Dec 09, 2020 12:39 am

Dear Marcin and Eric,
Many thanks for your insightful feedback and consideration. I highly appreciate the insightful suggestions of each of you. Your insightful suggestions have enlightened the way for me for resolving that issue.

I have modified the input file for the DFPT. The relaxed ion elastic tensor is now computed reasonably.

I have noticed that there is a small warning in the output file on the “ddb_piezo”:
(Unstable eigenvalue detected in force constant matrix at Gamma point).

I am attaching hereto this output file. Thus, I hope you may kindly advise me on the possible cause behind that warning statement. I thankfully look forward to your consideration as well as to the consideration of the experienced contributors to the Abinit Forum for finding the root cause of that warning statement.
With thanks and kind regards,
Salah

ebousquet
Posts: 469
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: Computing the relaxed ion elastic tensor

Post by ebousquet » Thu Dec 10, 2020 2:31 pm

Dear Salah,
Comment regarding the first problem: relaxed ions means you want the effect of atoms moving due to the change of interatomic force constants related to a strain perturbation such that you nee calculation of phonons to get it, see:
https://journals.aps.org/prb/abstract/1 ... .72.035105

Regarding the negative IFC in your calculation, it sounds like you have an unstable phonon mode in your system but you do not print them in the anaddb, could you add the print of phonon frequencies such that we can see if there is indeed an unstable mode?

Best wishes,
Eric

mmazdz
Posts: 5
Joined: Fri Sep 13, 2019 9:13 am

Re: Computing the relaxed ion elastic tensor

Post by mmazdz » Thu Dec 10, 2020 4:58 pm

Dear Salah,
I have noticed further problems with your results:
- Elastic Tensor must be symmetric or close to symmetry, yours is not
- If the crystal is mechanically stable, its Elastic Tensor is positive-definite (in Mandel notation all Kelvin moduli must be positive), yours is not: Kelvin moduli = {12.6588, 11.3204, -7.43272, 6.34902, 1.49906, -1.49094} (see attachment)
Best wishes,
Marcin
Attachments
3D.pdf
(85.73 KiB) Downloaded 368 times

Salah
Posts: 7
Joined: Mon Oct 12, 2020 5:33 am

Re: Computing the relaxed ion elastic tensor

Post by Salah » Thu Dec 31, 2020 10:47 pm

Dear Eric and Marcin,
Many thanks for your replies. I highly appreciate your insightful replies and helpful consideration.

I have been acting accordingly for several days continuously. In fact, I have improved the code accordingly and have run the computations again and have got better results. For your convenience, I am attaching hereto the input files and the output files in this regard.

Now, the tensor of the elastic constants is symmetric with all values of its elements positive, except for a couple of values which are almost zero. This can be due to a sort of vacuum in the unit cell, or due to not getting very tight convergence in the tolwfk. Thus, there can be 9% room for improvement.

The corresponding values of the elastic properties of the materials are reasonably comparable with the results of similar studies.

In addition, in the output of the DFT I have got improved results on phonons. Now, all the frequencies of phonons are positive, except for two frequencies whose values are small. This can be due to a sort of vacuum in the unit cell, or due to not getting very tight convergence in the tolwfk. Thus, there can be 9% room for improvement.

I am going later to further investigate in the near future the phonons results and analysis on this proposed system of materials. If I have further questions on that, I may make another post.

I wish you all best wishes of good health and a prosperous new year. I thankfully look forward to keeping in touch.

With thanks and kind regards,
Salah

Locked