Skip to Content.
Sympa Menu

forum - Problem calculating Born effective charges

forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

Problem calculating Born effective charges


Chronological Thread 
  • From: "Erik Ylvisaker" <eylvisaker@physics.ucdavis.edu>
  • To: <forum@abinit.org>
  • Subject: Problem calculating Born effective charges
  • Date: Wed, 1 Jun 2005 11:30:45 -0700

I looked around in the FAQs and the mailing list archives, but I didn't really find any answer to my question. I'm trying to calculate the effective charge tensors of the atoms in LiNbO2, which in fact has a double unit cell, so as far as Abinit is concerned, it's Li2Nb2O4.

My understanding of the theory behind Born effective charges is that if you sum up over any single element in all the effective charge tensors, they should add up to zero. This does not happen for (1,1) (1,2) (2,1) (2,2) elements, they come out to be around 2 or 4. Elements corresponding to the c direction do sum up to zero properly (or close enough). I have tried this calculation twice, once with nkpt = 40, generating using ngkpt = 8 8 8, and once with nkpt = 168 (as in the attached file) generated using ngkpt = 14 14 14, both using shiftk = 0, 0, 0.5. Both gave very similar results. I don't understand if there's some setting that is wrong in my input file, or if this is indicative of some sort of other problem. The first time I ran it I had problems converging, and it gave messages about using metallic occupation instead. So I have used occopt = 3 and tsmear = 0.001, and it seems to converge, but as mentioned above, the effective charge tensors appear to be wrong.

Attached are born.in, my input file, and born.out, parts of the output file that seem relevant. I can send the whole output file to anyone that's interested, it's 550kB though. These calculations were done using Abinit 4.2.4, although I didn't notice anything in the release notes for versions 4.3.4 or 4.4.4 that seemed to be relevant to this calculation, as I'm not doing strain perturbations.

Also, these calculations seem to take quite a lot of CPU time compared to other calculations in abinit, so if anyone has any tips on how to speed it up, I would appreaciate hearing them.

Attachment: born.out
Description: Binary data

Attachment: born.in
Description: Binary data



  • Problem calculating Born effective charges, Erik Ylvisaker, 06/01/2005

Archive powered by MHonArc 2.6.16.

Top of Page