Post-processing (transport properties) using VASP?

[gwynchoi]

Hello,

I am trying to run a post-processing calculation of transport properties using TRIQS solid dmft but read that so far, the computation is based only on the Wien2k DFT code (https://tugraz.elsevierpure.com/ws/portalfiles/portal/5145978/CPC204_200.pdf).

Since I ran the TRIQS solid dmft calculations using VASP, I am wondering if there are any updates with VASP being used with TRIQS DFT+DMFT for calculating transport properties? If not, are there any other codes besides Wien2k such as quantum ESPRESSO that can be used?

Thank you.

Gwyn

[the-hampel]

Hi @gwynchoi ,

Although very poorly documented triqs/dft_tools has the ability to perform transport calculation for any set of Wannier90 input files. You only need to base your DMFT calculation on the Wannier90 converter in dft_tools (fully compatible with solid_dmft). Either coming from VASP or QE.

Once you have your wannier90_hr.dat file you can use the tranport functions here: https://triqs.github.io/dft_tools/latest/_python_api/triqs_dft_tools.sumk_dft_transport.html more specifically: https://triqs.github.io/dft_tools/latest/_python_api/triqs_dft_tools.sumk_dft_transport.recompute_w90_input_on_different_mesh.html#triqs_dft_tools.sumk_dft_transport.recompute_w90_input_on_different_mesh to prep the transport input for sumk. With or withoug DMFT self-energy. Then you can follow the normal steps to perform transport / optics calculations. An example workflow starting from a hr file can be found here: https://github.com/TRIQS/dft_tools/blob/3.3.x/test/python/w90_optics_Sr2RuO4.py . Note that this requires you to install https://wannier-berri.org/ , possibly you have to find some version that works with your current python env.

All of this is already part in the released 3.3.x branch. Note, that if you do perform CSC DFT+DMFT calculation you have to do some approximations since the optics post-processing needs to leverage Wannier interpolation, i.e. starting from the H(R) file, and cannot use the projector matrices. Easiest is to do this first for a single-shot DMFT calculation from wannier90 input. and later try for CSC calculations.

Another possibility is to use https://lxvm.github.io/AutoBZ.jl/stable/ . This package is written in Julia but employs advanced BZ integration techniques starting from H(R) and a self-energy to calculate optics with very high accuracy.

Best,
Alex

[the-hampel]

Hi @gwynchoi ,

I just wanted to have a quick look at your code but could not find the link anymore? Can you resend it and I can have a look at it!

Best,
Alex

[gwynchoi]

Hello @the-hampel ,

I was about to ask you a follow up question to the original thread! I have all of the wannier90 inputs ready as well as the one shot DMFT calculation. I was wondering how I could go about the CSC because it appears I need PLO for it, and cannot find much reference on how to connect wannier90 to CSC.

(For the codes, I am working on a new system with different parameters so when that is complete I can send them over.)

I appreciate your help!

Gwyn

[the-hampel]

Ah I see. Okay I assume you are talking about w90 through VASP using solid_dmft in CSC mode right? You have to change the VASP source code slightly for that: https://triqs.github.io/solid_dmft/md_notes/vasp_csc.html#enabling-csc-calculations-with-wannier90-projectors . I am not sure how up-to-date this instructions are. They are for sure valid for VASP 5.4.4 . You can check if this works for VASP 6.5.0 . Let me know if you need help. Another option is to use PLOs with VASP 6.5.0 . This works out of the box.

Best,
Alex

[gwynchoi]

Hello Alex,

By using PLOs with VASP 6.5.0, do you mean that I do PLO + CSC dmft first, and then W90 for transport? I am wondering how consistent this would be as well as the magnitude of approximation when it comes to publishing good results. I will make a new thread for any other questions that come up, thank you.

Gwyn

[the-hampel]

Ah no if this is about transport stick to W90 as the interface. The changes in the source code should be minimal. If this does not work let me know.

• Alex

[phibeck]

Hi @gwynchoi,

@the-hampel already gave you all the information. Sorry for the poor documentation, that's largely my fault. If you have any questions at all setting things up either with dft_tools or AutoBZ.jl, let me know!

Just a brief hint what to use when: if you're running calculations for a larger basis set and at higher temperatures, the dft_tools version works well. If you have a small basis set and will need good momentum resolution, I recommend looking into AutoBZ.jl.

Best,
Sophie