Multibinit, Vampire or Tom's ASD

[Rick Anteur]

Hello

I would like to use multibinit. I was trying to install it in ubuntu 20.04 via virtualbox (install in win10), however, after each build I can't see abinit and multibinit anywhere. Can someone help me please?

I repeated the instructions to install abinit 9.x.x, which I found everywhere on the internet including the ABinit site, without success.

Can I install the abinit 9.x.x version (with multibinit) just so that sudo apt-get install abinit?

Otherwise someone knows how to fix the Vampire Execution Failure problem; or how to install the Tom's ASD package? Thank you very much!

Best regards!

[Xu He]

Hi,

- Is the build successful? Could you send how Abinit is installed? What are the error messages? Perhaps you can also post it on the Abinit forum (it has been recently moved to https://discourse.abinit.org ).  

- If the build is finished: Did you run "make install"? If not, it is in BUILDDIR/src/98_main where BUILDDDIR is the directory where Abinit is built.

Abinit9 can be installed via apt in ubuntu 21.04. (see https://launchpad.net/ubuntu/+source/abinit).

There are some differences in the input format of vampire 4 and 5. I am yet to learn what has been changed. I will transfer the format to version 5 and deprecate the output for version 4. If you want to use it now, perhaps you can read the tutorial of vampire and try to modify the files.

Tom's ASD is not publicly available yet.

Best wishes,

HeXu

--
You received this message because you are subscribed to the Google Groups "TB2J" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tb2j+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tb2j/CAMRadMX9pAYkUBipq7htr-N3ohb%3DWmiEipEgrBa3q7N%3DU07ntQ%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

[Rick Anteur]

Thanks for your advices.

I did the math with "site_symmetry = .true."

My question of great beginner, is for a calculation site_symmetry = .true. :

  if I use quantum-espresso; is it recommended to put in nscf nosym = .true.,?

[Xu He]

Hi,

Yes. I think nosym=.true. is always recommended in quantum-espresso when building Wannier functions, which will write the wavefunctions for all the k-points within the k-mesh.

Best regards,

HeXu

[Rick Anteur]

Hello, I performed calculations with wannier90 which either did not converge or that the wannier functions were not localized on the atoms. So I used "site_symmetry = .true." which unfortunately converges only if the projections are only chosen for the Mn atom, on the d orbital. My questions are: Knowing that around the fermi energy the d orbitals of Mn are very mainly represented, is it sufficient to consider only the d orbitals for Mn if I want to calculate the magnetic interactions? Must I also imperatively take those of germanium. 

Thank you!

[Xu He]

Hi Rick, 

Usually the ligand orbitals should be included in the projectors. TB2J takes the rigid spin rotation approximation, which assumes that "inside" one Wannier function, the spin orientation is coherent,  and the amplitude of the magnetic moment does not change with the perturbation of small spin rotation. This is usually a good approximation when the Wannier functions are well localized around the atom, but becomes worse when it is an d(Mn)-p(Ge) hybrid. 

To make the convergence better, there are some good guidelines here:

https://www.wanniertools.org/tutorials/high-quality-wfs/

Best wishes,

HeXu