Since the ab initio structures are periodic in all 3 directions, do the atoms are the atoms at the bottom of the slab also exposed to the vacuum on the top? To better illustrate my question, here’s an illustration of a catalyst + adsorbate system from the paper:
I drew the main simulation box, everything outside is a periodic image.
Consider the black atom at the bottom - I drew a cutoff radius around it. Since the boundary conditions are periodic, is my understanding correct that this black atom at the bottom of the slab is also exposed to the vacuum?
Correct. Despite periodicity in all 3 directions, the presence of the vacuum effectively prevents the atoms at the bottom of the surface from seeing the adsorbates in the periodic image (as intended).
The LMDBs do already account for PBC. You’ll see a cell_offset attribute in there as well, corresponding to what neighboring unit cell the neighbor is. If you are using the OCP models you will need to set use_pbc: True in your config (ocp/dpp.yml at 3ace0f225c54bae0413b26c6ed23ea09ff73e784 · Open-Catalyst-Project/ocp · GitHub) to properly calculate distances, angles, etc.
Thank you for the clarification. Initially I thought get_pbc_distances() updates the edge list with PBC in dimenet++. So .lmdb 's provided are all the final graphs structures with pbc. Thank you again !!
