Local queryable entire database for arbitrary convex hulls

Materials Project
1

Dear MP,

I’m interested in regularly checking the e_above_hull for arbitrary compositions. Right now, I’m doing this by gathering the relevant entries with an API call, then constructing the hull locally. The bottleneck is the API call.

There is one post about downloading the entire database, but I don’t just want all the entries. I want something which I can query in a similar way to produce these convex hulls “on the fly.” Any recommendations?

Thank you,
Ben

2

@bkmi could I ask how you are querying for entries? Also, depending on which hull you want, you might also find it useful to query our pre-computed phase diagrams for different chemical systems.

– Jason

3

I know it’s been a while but my project is coming to a close and I may need this functionality. Right now I query using mprester.get_entries_in_chemsys(chemsys). It works, but I need to query about 100,000 times and my script regularly gets stuck. To avoid the ratelimit, I stick to querying 10x / s maximum (although it’s probably a lot slower).

Any thoughts about this?

4

One other design choice, is that I’d like to query the hull at a certain date. For example in this paper UniMat: Scalable Diffusion for Materials Generation, they query the hull at the time of the creation of CDVAE (July 2021).

I think the best way to do this is to find the mpid for the first structure deposited in July 2021 and filter out all entries with an mp-id higher… But how do I figure out which mp-id was the first deposited in July 2021?

5

for those who try to do this in the future, the solution was to download a computed hull using matbench discovery.

1 Like