Behavior for retrieving average volume when not found in database

[abhijeetgangan]

The get_average_volume_from_database method returns ZeroDivisionError: division by zero when the composition is not found in the lookup table.

if (avg_vol := get_entry_from_dict(chem_env_key)) is not None:
     vols.append(avg_vol["avg_vol"] * avg_vol["count"])
     counts += avg_vol["count"]

This is because there is no fall back method when the composition isn't found and the counts and vol are both zero. A default would be using something like the Vegard's law where the volumes can be obtained with the single element structures.

Error when using the Composition("Ag50 Nb50"):

get_average_volume_from_database
    return sum(vols) / counts
           ~~~~~~~~~~^~~~~~~~
ZeroDivisionError: division by zero

[esoteric-ephemera]

If you take a look at get_average_volume_from_database, you can see that the default fallback is to use something like Vegard's law. We basically reweight the volume by composition / relative statistical count

For this method to fail, an element would have to be missing from the reference data entirely. What elements have you observed this behavior with?

[abhijeetgangan]

It happens when using the cached version of the lookup table for Ag and Nb. Both avg_vol["count"] and avg_vol["avg_vol"] return zero.

[esoteric-ephemera]

Are you maybe using an out of date version of the reference data? When I run the following:

from atomate2.common.jobs.mpmorph import _get_average_volumes_file

avg_vols = _get_average_volumes_file()
print(avg_vols[[ce in ("Ag","Nb") for ce in avg_vols["chem_env"]]])

I get non-empty entries:

      chem_env  with_oxi    avg_vol  count source
56326       Nb     False  18.022948     17   icsd
56830       Ag     False  16.986225     25   icsd
66755       Nb     False  18.592124     11     mp
92218       Ag     False  17.766298     14     mp