search

aiidateam / aiida-quantumespresso

The official AiiDA plugin for Quantum ESPRESSO
https://aiida-quantumespresso.readthedocs.io
Other
52 stars 77 forks source link

`PwRelaxWorkChain`: Revisit work chain logic #952

Open mbercx opened 1 year ago

mbercx commented 1 year ago

Fixes #733 Fixes #705

Will write full commit message when all changes are agreed upon.

mbercx commented 1 year ago

@sphuber some initial notes/questions/decisions:

  1. Regarding the meta-convergence:

    • In #705, I forgot to describe one of the most important reasons to more away from using the volume for the meta iterations: although it's rare that the volume condition rarely doesn't catch an issue with Pulay stresses, it will often run another geometry optimization unnecessarily.
    • Regarding the condition on the kpoints mesh not changing: should we be very strict, and expect the same mesh, or is it fine in case the previous run was done with a denser mesh than the output structure would require? Of course kpoints convergence will be achieved in the latter case, but there might be use cases where the user wants the kpoints mesh to approach the desired density as close as possible. For now I've decided to be strict.

  2. Should the initial relaxation be run by default? Currently it is, and if we don't make it the default I'm wondering how to allow the user to enable it while also being able to provide a protocol for this step.

  3. Re the documenation: I was planning to describe the work chain Purpose and Logic in the Topics section. What do you think?