Disagreement between methods of hull construction

I ran a (mostly) Materials Project compatible calculation of spinel (Fd-3m) Sc2CdS4, and using the Pymatgen/MPRester to build the hull diagram puts my calculation 0.2 meV/atom above the hull, with the MP Sc2CdS4 structure (mp-10953) as the stable decomposition product. However, in the Materials Project, that structure is itself shown as 292 meV/atom above the hull, and so the Pymatgen/MPRester decomposition ought to be CdS + Sc2S3.

It’s possible that this is an error in Pymatgen’s phase diagram, but my first guess is that something is wrong with the hull energy shown in Materials Project; 292 meV/atom above the hull doesn’t seem right for an experimentally synthesized sulfide. Just checking manually on the calculator:
Sc4Cd2S8 -> Sc4S6 + Cd2S2
(-146.5048/2 - 13.9748 + 87.3615)/14 = +0.010 meV/atom, indicating that Sc2CdS4 should be on the hull.

Hi @Steven_Hartman.

The data via the API and via pymatgen is correct, while the information on the materials details page is incorrect due to a bug in how our handling of sandboxed data (sandboxed data includes temporarily embargoed/non-public materials).

@shyamd is handling the fix and we will update our release notes with a description of the issue and a post-mortem as soon as possible.

Thanks also for reporting this and for your patience while we fix it! Reports about data issues are really valued and useful for us, since there are too many materials for us all to check manually, and our automated testing does not catch everything.

This should now be fixed, we’re still conducting a post-mortem on the issue however.