Add basic Wannier.jl integration #899
Conversation
|
Junfeng pushed a 0.3.2 Wannier update, unfortunately this requires Julia 1.8.4 at least. Is it acceptable to update Julia to 1.8, at least for testing? (I have 1.9 on my machine and did not catch this issue...) |
|
Yes #905. This was in the pipeline for other PRs as well. |
|
I added hydrogenic projections as an initial guess, because I wanted to try using them. The code was mostly adapted from #854 so I added a co-author note. On the basic example in DFTK, we go from 45 to 17 function calls for the optimization in the Wannierization procedure. |
|
@mfherbst CI is passing (except for nightly due to a |
|
Merge conflicts addressed. :) |
|
All comments addressed, and I moved the Wannier.jl and Wannier90 integrations to extensions as discussed earlier. We have to wait for another Wannier bump due to the Spglib bump (qiaojunfeng/Wannier.jl#35). |
src/external/wannier_shared.jl
Outdated
| """ | ||
| A p-like initial guess, using the difference of two Gaussians with different centers. | ||
| """ | ||
| function opposite_gaussians_projection(center1::AbstractVector, center2::AbstractVector) |
There was a problem hiding this comment.
You don't like the idea of directly calling it plike and slike ? I'm not sure myself, but wanted to argue.
There was a problem hiding this comment.
I'm not too sure myself. If you already know that you want a plike orbital you probably want to use HydrogenicWannierProjection anyway (or even the pswfc from the PSP in the future).
There was a problem hiding this comment.
yes, that's true. Good point. OTOH this makes it questionable why this function even exists in the main code. Might be good to include in the example to show how custom projectors could be made. Maybe it's better put there.
There was a problem hiding this comment.
Just removed it. I don't think that encouraging custom projectors is particularly a good idea, in principle it shouldn't be needed and the interface is still experimental. (I was using opposite_gaussians_projection before the hydrogenic guess was implemented, now it can be useful for comparison against the hydrogenic guess but that doesn't need to be in the mainline code).
There was a problem hiding this comment.
(I was using opposite_gaussians_projection before the hydrogenic guess was implemented, now it can be useful for comparison against the hydrogenic guess but that doesn't need to be in the mainline code).
Yes, but it would be good to keep the idea documented in the wannier example. That's what I meant. Sorry if I was not clear.
This PR adds basic
Wannier.jlintegration. The communication happens via the Wannier90 files at the moment; this is not the fastest, but it required the least amount of code. I split the Wannier code acrosswannier_shared.jl,wannier90.jlandwannier.jl, the last 2 requiring the relevant package.I also refactored the initial projection code, to make the choice of the initial projections more flexible. Hopefully that can be extended in the future to project onto atomic orbitals without changing the API again.
Following the recommendation in one of the #854 comments, I updated the example to mention
Wannier.jlfirst.