Looking at example: https://materialsproject.org/materials/mp-390/
The bottom of this page shows a list with title “ICSD IDs”. I understand that these are cifs that were found to be similar hence grouped together here. Is it possible to determine which of these ICSDs were actually used to do the “mp-390” calculation? Is it the first entry in this list?
Currently, that’s not possible. It’s not necessarily the first in the list. We have to be careful about not reproducing the ICSD’s data, which is really their ID to structure representation, so we’re hesitant about being more specific.
That being said, you’ll notice that right above the ICSDs, we have a bunch of tasks listed and only a few are dark black. These are the calculations we’re using to build the materials properties. In this case, the structure comes from mp-1205164, probably because the final structure was slightly lower in energy. You can grab the input structure for that calculation, and then using pymatgen and your own set of ICSD CIFs, figure out which one it corresponds too.
It’s also important to note that we do structure matching on the relaxed structure, so although two CIFs from ICSD may not even actually match according to our structure matcher, the DFT relaxed structures do. That’s what we base our definition of a material on.