Obtain large numbers of band structures?

Hello,
I’m attempting to calculate the electron and hole effective masses of several thousand materials from the MP database, using sumo, which requires the bandstructure. I start by using a mongo-like query to find materials which have bandstructures and meet my other criteria, but putting “bandstructure” in the “properties” argument to MPRester.query() returns None, probably because “bandstructure” isn’t a supported property. For a small number of materials I could simply call get_bandstructure_by_material_id() for each, but I don’t want to hit the API thousands of times every time I run this query.
Is there a good way to do this, or is it beyond the current capabilities of the API?

Thanks

2 Likes

Hi Steven,

I think you can try this

from pymatgen.ext.matproj import MPRester
if __name__=="__main__":
    mpr = MPRester("your_MPkey")
    for id_ in range(1,1000):
        mp_id = "mp-" + str(id_)
        has_bs = False
        try:
            bs = mpr.get_bandstructure_by_material_id(material_id=mp_id)
            if bs is not None: has_bs = True
        except:
            pass
        if has_bs:
            do_something with bs (e.g., save it to a json file)

That method works well for a small number of materials, but each call of get_bandstructure_by_material_id takes about 5 seconds for me, so it isn’t practical at large scale. I think it would also degrade the performance of the MP database for other users, as well, due to the large number of API calls.

You’re right. Look at my post here

1 Like

I need the line mode band structure to calculate effective mass, not the uniform NSCF calculation.

You can set line_mode = True

1 Like

Pinging @shyamd @mkhorton @tschaume if they may be able to answer this

Does anyone know a solution for this? Alternatively, I could call get_bandstructure_by_material_id for each material, and let the script run overnight or something, but I don’t want to strain the Materials Project server.

Unfortunately, there is no good solution for this. The MP API, originally developed in 2012, was never designed for this. We are developing a new API that will be able to handle distributing all Bandstructures to a user.

For now, @Steven_Hartman’s solution is the best if you rate-limit yourself. We will try and build a google drive package of band-structures for everyone, but there are numerous backend restructuring tasks which are higher priority as our use and load continue to increase. The new API might be done before this.