search

zerothi / sisl

Electronic structure Python package for post analysis and large scale tight-binding DFT/NEGF calculations
https://zerothi.github.io/sisl
Mozilla Public License 2.0
182 stars 58 forks source link

Writing Hamiltonian with sisl from Wannier #760

Closed monx-ne closed 4 months ago

monx-ne commented 5 months ago

We are trying to write a .nc file for a tbtrans simulation using a Hamiltonian read with sisl (0.14.3) from a Wannier90 run. Since the Wannier90 Hamiltonian has complex terms, we run into the following message:

NotImplementedError: Currently we only allow writing a floating point Hamiltonian to the Siesta format

Is there a fundamental reason why it can not be complex? Is this issue related to siesta, sisl or tbtrans?

Also @nils-wittemeier might know/be interested in this issue. This might be relevant for studying other Hamiltonians that have complex terms, for instance topological insulators.

zerothi commented 5 months ago

There are two issues:

1) if your Wannier Hamiltonian is complex, likely the minimization process is not complete, at least that doesn't sound good! 2) Siesta and TBtrans does not allow one to do complex hamiltonians, only real-spin dependent components. If you want to use a complex Hamiltonian, you have to use the dH approach. There you can basically do whatever you want. ;)

zerothi commented 5 months ago

And since tbtrans isn't fully SO (then you'll have to play with Nils's branch, then you can't do a SO Hamiltonian with 8 components. So you basically have to emulate the 2 spin-components as individual orbitals.

zerothi commented 4 months ago

If there are no more information or feedback here, I'll close this soonish :) Thanks!