Question: How to make dataset?

[motonuko]

Hi

I have questions about datasets.

Materials Project database and API had changed a lot from when this repository started, and I can’t get the whole dataset written in the “cgcnn/data/material/mp-ids-○○.csv” file.

In CGCNN paper(https://arxiv.org/abs/1710.10324 ), it is written that

After removing ill-converged crystals, the full database has 46744 materials covering 87 elements, ...

My questions are as follows.

• How to collect “mp-ids-27430", “mp-ids-46744”, “mp-ids-3402" like dataset?
• What is the mean of ill-converged crystal ?

Thanks,

[txie-93]

Hi, thanks for your question!

• Each line in mp-ids-27430.csv is a materials id in the Materials Project database. To get the data, one option is to write some code using the pymatgen interface to download the structures and properties. An example code will be like:

from pymatgen.ext.matproj import MPRester

with MPRester("API_KEY") as m:
    results = m.query(criteria={'material_id': 'mp-1234'},
                      properties=['formation_energy_per_atom', 'structure'])
    formation_energy = results['formation_energy_per_atom']
    structure = results['structure']

You can get the "API_KEY" by registering an account at their website.

• We removed all crystals with a warning label in their calculations. And we only included crystals with fully-computed band structures for the band gap predictions. This may not be the best way to define ill-converged crystal. Our thought at the time was to only include the highest quality data.

Hope that is helpful!

[motonuko]

Thanks for your explanation! It is vary helpful for me.

I found a change in materials project (https://discuss.matsci.org/t/change-in-materials-project-ids/1268 ), and some 'materials_id' has changed to 'task_id'. Taking this change into account, I got similar datasets.

Thanks,