Values coming in hubbards attribute appear to be swapped for ‘O’ and transitional metals in ‘mp-17387’, ‘mp-17899’, ‘mp-22334’, ‘mp-26211’, ‘mp-26685’, ‘mp-26689’, ‘mp-26744’, ‘mp-26921’, ‘mp-530780’, ‘mp-531901’. For example, ‘mp-17899’: {‘O’: 3.9, ‘V’: 3.25, ‘Mn’: 0.0} (‘O’ <-> ‘Mn’), ‘mp-26744’: {‘O’: 4.38, ‘P’: 0.0, ‘Mo’: 0.0} (‘O’ <-> ‘Mo’).
Also, in ‘mp-18527’: {‘O’: 0.0, ‘P’: 4.38, ‘Mo’: 0.0, ‘Cs’: 0.0} (‘P’ <-> ‘Mo’).
Hi @Lsky, thanks for reporting this! We’ve had site permutation bugs previously. We have confidence the correct U values were used in the actual calculation, but we will need to investigate further to make sure the U values are reported accurately.
Hi @Lsky, thanks for your patience with this. We’re currently preparing our next data release, and it should be addressed then along with an announcement of changes made in our Materials Project Database Release Log thread. We’re currently aiming for happen in the next few weeks, but we’re not giving a specific date yet.
We’re trying not to change data in the database without these formal announcements because we think it’s important users are informed when data is changed (even in cases like this, when the change is a good thing like fixing a bug). It also gives us time to add checks to help prevent similar issues don’t happen in future data releases. And our U values are always standardized, so even if the reported values are swapped in these cases, we are using the same U values as specified in our docs.