Just writing to flag that the docs indicate a Eu pseudopotential choice of Eu_3, while the InputSet in pymatgen seems to use Eu (which is consistent with the TaskDoc from the API).
In some preliminary testing, we’re finding that the Eu_3 is the best general choice for reproducing experimental heats of formation, although there’s still considerable error on EuO with Eu_3.
Hi @Joseph_Montoya, we did a pretty thorough benchmark in revising the pseudopotentials used by MP, and I didn’t turn up specific issues with the Eu pseudopotential. You can see the benchmark discussion here and the longer document here. These are with the PBE_64 release.
The tests indicate that Eu_h is probably the most transferable across chemical environments, but the improvement over Eu isn’t that marked. Would be curious how your testing methodology differs.
For the docs: The only mention of Eu_3 is in the MOF calculation section, which do not use the same parameters as our normal production calculations. The GGA/+U and r2SCAN sections both correctly list Eu.
Ah! This is so helpful, thank you! My confused resulted from when I clicked on the first returned hit from google on “Materials Project Pseudopotentials” and didn’t realize I was looking at the MOF-specific parameters.
Our testing methodology is pretty simple - just comparing the experimental and theoretical heats of formation, but we’re exploring modified correction schemes, so it may not be quite apples to apples with previous benchmarks. I’ll review these references and follow up if there’s anything notable for MP.
Thanks again!