Is it possible that the Kwant-interface can deal with Polyatomic

[tron1989]

Hi Dominik,

I just found that your Tbmodels code group all the orbitals into one site when interfacing with kwant.

Is there a way to deal with Polyatomic case that simulate the real system which has different sites and different orbitals on each site?

Thanks
Xiangru

[Dominik Gresch]

Hi Xiangru,

If you have a model with multiple orbitals that share the same position, these should be grouped into a single site. So for example, if your .pos is [[0, 0, 0], [0, 0, 0], [0.5, 0.5, 0.5]], the first sub-lattice should contain the first two orbitals, while the second sub-lattice contains the last orbital.

Is this what you were looking for, or do you need something different? In general, the kwant interface is a bit limited as it is.

Cheers,
Dominik

[tron1989]

Hi Dominik,

Thanks for your reply.

I was thinking reproducing band structures firstly by Kwant from the wannier90 output files which there are more than one site (Wannier postion) and different number of orbitals on each site.

I checked TBmodels, I guessed it group all the orbitals into one site, and one site per unit cell. Because, it seems that there are no postion information provided in wannier90_hr.dat file.

Thanks
Xiangru

[tron1989]

Hi Dominik,

Sorry, I am idiot.

I just found that TBmodels can read Wannier centers, and it seems that it takes every wannier position as a sublattice.

Xiangru

[Dominik Gresch]

it takes every wannier position as a sublattice.

In general, yes - it will only group different orbitals into the same sublattice if they have the same position. If that's undesired, you can modify the pos attribute manually before creating the kwant model.

[tron1989]

Hi Dominik,

I have tested a case with 32 orbitals (considering SOC) on a laptop, it took much long time to run

model.add_hoppings_kwant(kwant_sys)

I had to kill it. Is there something wrong?

Because when only read wannier90_hr.dat file, it is easy to run.

Thanks
Xiangru

[Dominik Gresch]

Well, it's to be expected that it is slower than having them all at the same position. If they are all at the same site, it can set each 32x32 matrix of hoppings in a single step. When they are separate sites, each hopping needs to be set individually.

It will depend also on the range of your hoppings (how many non-zero hoppings there are), and the size of your kwant shape (i.e., if you are constructing a large finite structure that will be expensive).

Maybe a good middle ground is to group Wannier centers according to the atom that is nearest to them: This can be done automatically by using pos_kind='nearest_atom' when reading the Wannier90 output: https://tbmodels.greschd.ch/en/latest/reference.html#tbmodels.Model.from_wannier_files

[tron1989]

Hi Dominik,

Thank you very much. I have tested it successfully.
You code is really helpful. It will save me a lot of time as I planned to write by myself.

Best
Xiangru

[Dominik Gresch]

Good to hear, and you're very welcome!

I'll close this issue then, but feel free to open another one if you have more questions.

[Andrey E. Antipov]

Dominik, FYI Kwant now added vectorized setting of hoppings and onsites, so perhaps it can be used to speed up things.