Can't reconcile pourbaix decomposition energies between API and online diagram

My issue may be related to this previous topic but it has not received any response in the past month. Essentially, I’m trying to calculate delta Gpbx for a given materials id, however the calculated values don’t appear to match what’s on the website-generated pourbaix diagram. I’ve created a colab notebook which shows my calculation steps and also linked the online version of the diagram in the last cell.

Looking at the O2/H2O potential line, my calculated values are consistently lower than the mouse-over values in the online version. Am I missing something in the calculation? The steps I followed seem to work for other systems such as Fe2O3 where the energies are relatively close to the online version, but this particular example shows quite a large deviation (~1eV at pH 14).

Hi @guevarra,

I think other people more familiar with the Pourbaix infrastructure may be able to help better than I can, but have you tried filter_solids=False and/or have you tried an older version of pymatgen to see if the issue persists?

I suspect the current website version might use some non-default options enabled to make sure it runs in a timely manner, so might be subtly different to the pymatgen default.



Hi Matt,

Thanks for the input. Yes I’ve tried constructing the pourbaix diagram with filter_solids=False and while it adds more phase labels and boundaries to the plot, it doesn’t appear to affect the deltaGpbx. I’ve also tried setting the ionic concentration dict to the values used on the website, in addition to setting/omitting the composition dict but neither can make up for the 1 eV difference which I think is more than a bit subtle :wink:

Ah, yes. 1 eV is not so subtle! We’re still investigating, the website is running pymatgen 2019.8.14 currently if that helps.

Hi Guevarra,

Looks as if the website version is still being normalized per formula unit, rather than per atom, as per this post. I believe changes to either the units or the backend pymatgen version for MP are on their way.

I’ve modified the version of the colab notebook to demonstrate the differences between the versions, but I’m not sure if the changes are showing up for others or not. Thanks for catching this, and let me know if I can help further.

1 Like

Thanks Matt and Joseph! I’m glad this has been sorted out.

Thanks @Joseph_Montoya!

@guevarra, we’ve pushed a fix to the website code for this that should be deployed sometime next week.