MLWFs - a few questions

[Jake Burner]

Hi Edward,

I'm sorry for so many questions, again I am not an expert and I want to make sure I am not making any mistakes when implementing these functionals. My questions again pertain to the Wannier functions I should use in the calculation. I am trying to do this as efficiently as possible, as the systems I am working with are quite large.

1. For periodic systems, do we need MLWFs for all bands, or can we consider only bands near the valence/conduction bands in the wannierization procedure? For example, if I have a system with 32 valence bands and 32 conductions bands (64 total), can I just do two separate wannierization procedures, one where I consider bands 20-32 and one where I consider 32-44? Then use these 24 MLWFs as the variational orbitals in the Koopmans calculation?  

2. Also, I assume there is no "optimal spread" that we should aim to get, as the minimum spread probably system-dependent?

Thank you so much for your continued assistance with these calculations!

Jake

[Jake Burner]

Whoops, correction about the example above, I meant considering bands 20-32 and 33-45, as an example (obviously not considering an occupied band in the unoccupied manifold)

[Edward Linscott]

Hi Jake,

No need to apologise, this is the place for questions!

1. yes, you can do Koopmans functional calculations on a restricted set of Wannier functions, but only if you are using the "DFPT" strategy for computing screening parameters. For the Delta SCF approach you need to Wannierise all of the electrons that aren't pseudised (as well as as many empty states as you are interested in). This comes with the caveat that I'm not 100% sure if this is properly supported by the workflow manager, as I can't remember testing this myself, so feel free to give this a try and if you get errors about there not being enough projections provided then get back to me!
2. the key quantities you want to monitor during a Wannierisation are...
1. the "pattern" of the spreads e.g. if I am Wannierising silicon with four sp3 orbitals, I expect to get four Wannier functions with almost identical spreads. If for some reason the spreads are different to one another then something unexpected has happened!
2. the quality of band interpolation that you get (i.e. does an interpolated band structure match what I get when using bands.x?)

You can always run all of these checks and play around with the Wannierisation by setting "task" to "wannierize".

Happy to help,

Edward

--
You received this message because you are subscribed to the Google Groups "koopmans-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to koopmans-user...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/koopmans-users/f9fd7e5c-c5dc-4e25-a3b7-f6f20a50c19cn%40googlegroups.com.