Running calculations on supercell

[Jake Burner]

Hi Edward,

I have to do 2x2x2 k-point sampling for the wannierization to converge with the primitive cell I am working with. Would it be sufficient to skip the "unfolding to supercell" step in the workflow if I use a 2x2x2 supercell in my calculations, and sample just the gamma point? I am mainly curious what the reason is behind extending the primitive cell to a supercell after obtaining the MLWFs (in the calculation of the screening parameters).

Thanks!

Jake

[Edward Linscott]

Hi Jake,

Thanks for reaching out!

First, let me answer your general question: the main reason behind this unfolding procedure is computational efficiency:

• computing the initial ground state and constructing MLWFs is more efficient using k-point sampling than doing it directly in a supercell
• meanwhile, the DSCF calculations for calculating the screening parameters (where we explicitly remove/add an electron to a localised state) need to be performed in a supercell to avoid the localized charged defect interacting with its periodic images

The computational expense of the unfolding procedure is negligible compared to that of the Wannierization. I can't think of any situtaion where it would be advantageous not to use the smallest possible cell when constructing the MLWFs.

Now, to answer your specific question: yes this is possible, but it is not recommended. If you start from a cell that is a 2x2x2 supercell of the primitive cell, then setting the k-point grid to [1, 1, 1] will mean that the screening parameters are performed on the same 2x2x2 supercell of the primitive cell (which I believe is the process that you are describing). However, if you do this, I think you will still see the "unfolding to supercell" step during the workflow because this step -- in addition to converting the k-point-indexed WFs to Gamma-only WFs -- converts the Wannier90 files to a format that the subsequent kcp.x calculations can read, so it is still necessary even when the k-point grid is [1, 1, 1]!

Hope that helps!

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/af2b6fad-a3c8-4f10-87cd-015f670ffc3bn%40googlegroups.com.