I’ve been recently introduced to the work of Sendek et . Al, where they screened lithium-containing compounds from Materials Project using certain criteria. One of these is termed termed as the upper bound potential, which they defined as

Vox = G/xF

where x is the number of sodium atoms, and F the faraday constant.

I’ve ran into problems replicating these results however. The formation energy G I assume is the normalized formation energy per atom - hence for a compound, do I need to multiply this by the total number of atoms present in the formulaic structure?

Hi Exjustice,

Can you give a bit more detail? For example, can you link the manuscript and show an example of a calculation you’ve done which is inconsistent with their results?

(Also, I’d encourage you to contact the authors. We’re happy to help if we can with the MP part if we can, but it’ll be difficult to give you specific insight into what’s going wrong if their methods aren’t implemented in one of our codebases).

2 Likes

Hi Yijie,

MP presents *formation energy per atom* as the energy of a compound with respect to standard states (elements), normalized per atom. e.g., for Fe2O3 this is: [E(Fe2O3) - 2E(Fe) - 3/2E(O2)]/5.

If you use our pymatgen library, you can obtain our computed entries in the chemical system using `MPRester.get_entries_in_chemsys`

, add your compound as a `ComputedEntry`

, and finally generate a phase diagram, using `PhaseDiagram.get_form_energy_per_atom`

to obtain a formation energy per atom.

I do not know whether the particular paper uses formation energy per atom, but I hope the above clarifies MP’s approach.

Best,

Donny

1 Like

Hi Joseph

Thank you for the rapid and welcoming reply. I understand that this thread was for technical issues and hence this may be a bit of a strange question.

You may find the paper here

Looking at winstons reply (and example), im more and more convinced that multiplying by the number of atoms is the correct approach, given the description of the property on page 3.

Hi Donny

Thank you for the prompt and welcoming reply. I’ve compared your example to that of the paper (see my reply to Joseph), and have concluded that multiplying by the number of atoms in the compound is the right way forward.

I am trying to contact the authors of course to confirm