atomate
atomate copied to clipboard
hackingmaterials
Workfunction workflow to be moved to atomate
Will do, but ultimately I want to combine it with the surface energy wf, reconstruction wf, and cohesive energy wf so they can run in conjunction.
I'm still working on integrating all the surface property related workflows into Atomate on my own branch. I'll push forward to get a pull request of a working version of the workflows within the next two weeks.
Thanks, Richard
On Fri, Feb 2, 2018 at 11:41 AM, Anubhav Jain [email protected] wrote:
@richardtran415 https://github.com/richardtran415 @shyuep https://github.com/shyuep any update on this issue?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/hackingmaterials/atomate/issues/147#issuecomment-362685919, or mute the thread https://github.com/notifications/unsubscribe-auth/AHqe1vX5DAvosnzJS69acV9ZEaGYGFRFks5tQ2TegaJpZM4O5TNT .
--
Richard Tran Phone: (415) 425-7660 Email: [email protected] https://acs-webmail.ucsd.edu/squirrelmail/src/compose.php?send_to=rit001%40ucsd.edu
@richardtran415 might want to elaborate, but we've worked together to modify the existing surface workflow such that it's more or less identical to the MVL approach. The main change was to ensure that reference structures for the slabs were oriented unit cells, to ensure that the k-point meshes were isomorphic for the a/b lattice directions. The remaining issues are to add and test that LOCPOT params are set in the inputs and that LOCPOT files are parsed in the outputs.
We also will need to address reconstructions, which will be slightly more complicated, but mostly a matter of specifying the appropriate supercell transformation to allow for relaxation with the correct periodicity.
Currently since reconstructions are rather difficult to guess a priori without experimental input, the ReconstructionGenerator on surface.py just takes in a conventional unit cell and determines the spacegroup of the cell (and wykoff positions if we want to deal with more complicated reconstructions, I'll have this implement soon) and applies a set of site translations, deletions, additions and supercell transformations to a slab from instructions provided in a json file. The instructions are based on already well-known reconstructions. The generate_all_slabs function has an option to include the reconstructions which is what the workflow in my previous fork of atomate uses. Currently, I'm trying to incorporate a similar framework for reconstructions into the workflow me and Joey have been working on. The incorporation of LOCPOT was so the workfunction can be obtained by default for all slab calculations since this is a useful surface properties for both clean and adsorbed surfaces. This is just a minor edit to MVLSlabSet, the workflow will handle the rest.
Thanks, Richard
On Fri, May 4, 2018 at 11:58 AM, Joseph Montoya [email protected] wrote:
I'll let @richardtran415 https://github.com/richardtran415 elaborate, but we've worked together to modify the existing surface workflow such that it's more or less identical to the MVL approach. The main change was to ensure that reference structures for the slabs were oriented unit cells, to ensure that the k-point meshes were isomorphic for the a/b lattice directions. The remaining issues are to add and test that LOCPOT params are set in the inputs and that LOCPOT files are parsed in the outputs.
We also will need to address reconstructions, which will be slightly more complicated, but mostly a matter of specifying the appropriate supercell transformation to allow for relaxation with the correct periodicity.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/hackingmaterials/atomate/issues/147#issuecomment-386700884, or mute the thread https://github.com/notifications/unsubscribe-auth/AHqe1qqzcTo267M5RddDW8v6vd6MWq02ks5tvKTbgaJpZM4O5TNT .
--
Richard Tran Phone: (415) 425-7660 Email: [email protected] https://acs-webmail.ucsd.edu/squirrelmail/src/compose.php?send_to=rit001%40ucsd.edu
I apologize for not having an update on this work sooner. I haven't spent too much time on the ReconstructionGenerator from pymatgen lately so I wasn't ready to incorporate it into @montoyjh 's workflow yet. I'll try to get this part into the workflow well before the work function stuff gets published.
Actually because the get_wfs_all_slabs() in adsorption.py already uses generate_all_slabs(), users should be able to run reconstructed slabs already. In the slab_gen_params, one can set the parameter "include_reconstructions" to True. This will prompt generate_all_slabs() to check a json file on pymatgen for any possible reconstructions compatible with the provided bulk_structure and return the list of slabs including any reconstructions.