Finding the thermodynamic stability for the doped compounds through ML models.

[sachin-rangaswamy]

Hello everyone,
I am currently working on the doping of cathode coating materials and would appreciate some insights. To give a specific example, imagine I am working with a coating material like LiNiO₃. My objective is to screen possible multi-ion dopants (restricted to the first group of 3d-transition metals) for enhancing Li-ion conductivity and electronic conductivity. Additionally, the doped compound must be thermodynamically stable.

I’m currently using pymatgen to obtain decomposition data for doped compounds (e.g., Li16Ni14CuZnO48 → x * LiVO₃ + y * NiOₓ + ...) into stable phases. How would one proceed along further from here. However, I’d like to integrate the ML models in Matbench Discovery to evaluate thermodynamic stability efficiently. I have only the DFT calculations for few doped structures where I can't prepare the training set. Instead I would like use the Materials Projects data and proceed. If there are any other methods, please discuss it.

Could anyone suggest:

1. Which models/functions in Matbench Discovery are best suited for this?
2. How to build a workflow to test doped structures (CIF format) with these models?
   (Note: I am working with substitutional doping, and the example compound mentioned here is illustrative, not the actual material.)
   Your insights would be highly valuable! Thank you in advance