CSC DFT+DMFT with VASP for Non-collinear Magnetism

[EthanMakaresz]

Hi,

I want to reproduce the results of figure 4 of this paper (https://journals.aps.org/prb/abstract/10.1103/PhysRevB.106.205103), where they use Haule's DFT+DMFT code to model a non-collinear magnetic (NCM) state with spin-orbit coupling (SOC). The arrangement of magnetic moments is shown in figure 2 (c). I'm new to DFT+DMFT and I already have some familiarity with TRIQS and VASP so I'd prefer to use one of the TRIQS implementations with VASP for the DFT part if possible.

I've seen that the VASP projectors don't currently support NCM, so is it possible to enforce particular magnetic moments in DMFT and perform non-spin polarized DFT? Can the same be done with SOC included in DMFT? The tutorials mention that SOC in DMFT is possible for a one-shot calculation, but I'm not sure about the charge self-consistent case.

This project is a side project for my PhD, so I'm mainly looking for some guidance about the difficulty of the work-arounds. If the solution is very complicated, or if support for NCM is coming soon, there are other projects I'd rather devote time to.

Thanks in advance for any advice!

Regards,
Ethan

[the-hampel]

Hi @EthanMakaresz ,

There are some new features that we currently merge and test in the unstable branch of dft_tools. First, non-collinear projectors should be supported since #218 has been merged. This is just for VASP -> TRIQS, i.e. one-shot DFT+DMFT calulcations without charge self-consistency (CSC). The documentation might be not up-to-date but this should work. So running a VASP calculation with non-collinear magnetism including SOC and projections via LOCPROJ should work.

Then full CSC calculations with SOC should work on the unstable branch now with VASP 6.5 I believe. Maybe @dariofiosca can confirm and say something about the status. The code is in an early testing stage but we are working on it and people who test this are welcome! Note, that solving the impurity problem with SOC can be quite tricky.

Best,
Alex

[EthanMakaresz]

Hi @the-hampel, @dariofiosca

Thanks for your help. Input files would be very useful. I'm not at the stage of testing with solid_dmft, I still need to go through the tutorials and I need to update to VASP 6.5, but I'd welcome the help when I get to that point.

I'll reach out to you when I get back. Would you prefer another post here or an email?

Regards,
Ethan

[the-hampel]

Hi Ethan,
from my side discussing here is preferable. That documents the status of the code for other users. If there is anything scientific project related that should be kept confidential we can switch to email. I do believe we also have to still update the documentation to reflect the changes we made to the VASP interface. Just so that you know, that this is not in the tutorials as of now.

Best,
Alex

[EthanMakaresz]

Hi,

I'm back from my holiday and working through the tutorials at the moment, so I should be ready to start testing next week. Can you let me know what changes need to be made to the VASP source code in version 6.5 for CSC calculations? I tried following the documentation when I installed VASP6.4.3 but it looks like LPRJ_WRITE is not called from main.F anymore, but instead from mlfw.F, from a subroutine which doesn't have EFERMI in scope.

Also, how can I install the version of dfttools I need for testing?

Regards,
Ethan

[the-hampel]

Hi Ethan,

I just worked on the official documentation: #269 and want to merge this the coming days. I will try to do it later today. With VASP 6.5. you do not need to do any code changes. That is the important summary ;-) . We implemented all necessary changes now in VASP.

I will let you know later if I merged the changes for the documentation

Best,
Alex

[the-hampel]

I updated the unstable documentation: https://triqs.github.io/dft_tools/unstable/ . I updated the tutorials:

• https://triqs.github.io/dft_tools/unstable/tutorials/svo_vasp/svo_notebook.html
• https://triqs.github.io/dft_tools/unstable/tutorials/nio_csc_vasp/nio_csc.html#

and the documentation on the interface with instruction on how to use it: https://triqs.github.io/dft_tools/unstable/guide/conv_vasp.html (removing all the code changes)

Best,
Alex