materialsproject
BUG: Smearing and number of frequency points sometimes leads to missing peaks in phonon DOS
The generation of the phonon dos relies on the tetrahedron method. If one does not sample enough frequency points, sharp peaks might be missed.
@kaueltzen spotted this issue in a larger scale phonon calculation of more unusual structures with very flat phonon bands.
More frequency points or Gaussian smearing could be the solution
@esoteric-ephemera also for your information - might be relevant for the task doc.
Yeah. Good question. One could integrate the dos and check if too many states got lost?
How is this handled in the electronic bandstructure workflows? It should also be an issue there (with isif -5 at least)
[I realized after writing that the situation might be slightly different for electronic structure as we typically do not calculate the DOS for the whole energy range]
There aren't any checks performed right now, even on the custodian side. The closest is a handler which warns the user if ISIF = -5 and EMIN / EMAX were set
@esoteric-ephemera I see! Yeah, maybe it's not too much of a problem in general but it is something to be aware of. We recently had a smaller dataset of a bit more unusual structures (probably more molecular) created, where this was highly relevant.