BOWSR implementation in material discovery problems about maml HOT 3 CLOSED

The BOWSR algorithm is a structure relaxer that uses any energy model (in the paper, it is the MEGNet formation energy model, but it can be anything). To do the relaxation, you need an initial guess of a crystal structure. For materials discovery, a common way is to do substitutions of existing crystals with other elements. For example, say if you are looking for a new battery material, it is reasonable you might want to start from Na containing materials and change Na to Li to guess at potential new materials. E.g., to get LiFePO4 from say NaFePO4.

The initial structure guess can be done using any method you wish. BOWSR does not predict structures from composition alone. You need composition -> some list of plausible initial structures (unrelaxed) -> BOWSR -> relaxed structures with energies.

from maml.

Thank you for the very clear answer! Is there any method you would recommend for the initial structure guess? Any suggested paper to dig deeper into it?

from maml.

You can refer to https://matgenb.materialsvirtuallab.org/2018/09/25/Structure-Prediction-using-Pymatgen-and-the-Materials-API.html as one possible way.

from maml.