Are the structures in your database physical?


I’m doing a data mining project with the materials in your database, and I was wondering where the structures in the database comes from. I found that Materials Project is crowd-sourced with user-submitted structures, but are the user-submitted structures verified to be physical, i.e. exist in real life? Or are some of them just artificially made up, and submitted to see if the structure would have interesting properties? Thanks!


Some of the structures are hypothetical, but many are based on experimentally reported structures from the ICSD. Our structure matcher will group structures, and they will be associated with ICSD IDs if appropriate. Many ICSD structures are effectively duplicates of each other, which is why many of our materials are associated with multiple ICSD IDs. Furthermore, the materials shown on our site are energetically relaxed relative to original submitted structures, so the final structures are often not exactly what was reported in the ICSD, although we do provide the original structures as well. An example pymatgen.MPRester query to filter for materials with at least one associated ICSD ID:

from pymatgen import MPRester
mpr = MPRester()
mids_with_icsds = mpr.query(criteria={'icsd_ids.0': {'$exists': True}}, properties=['material_id'])

See our API documentation repository for more information on querying and on properties available to query.

I should mention also that we are currently (albeit slowly for now) working our way through processing an additional 18,952 ICSD structures and >20,000 structures from the Pauling File, which are also experimentally reported.