I’m writing my master thesis on a machine learning algorithm for the prediction of adsorption energies on metal oxides. In my approach I intend to use parameters of the electronic structure as potential descriptors for training the algorithm.
I was trying to reproduce the dft calculations (Single point) for the final relaxed state (adslab) in order to validate the VASP configuration that were described in the corresponding OC22 paper. (Later they will be used for the calculation of the clean surface in order to get the adsorption energy) It worked pretty well for certain materials (e.g. Al-O, Ti-O systems) that didn’t involve the Hubbard U (GGA+U) correction. I also managed to find the correct pseudopotentials that were provided in the .json file in the Open-Catalyst-Dataset GitHub page. However I’m struggling to reproduce materials that involve GGA+U calculations.
Would you give advice on how to set or possibly provide the correct INCAR files for the following bulk systems:
Thank you in advance and best regards,