In our exploration of magnetic materials, we did some embedding via Orb v2 and some dimensionality reduction.
Here we have an HTML file generated from Python and Plotly that displays ~5,000 magnetic materials in 3 dimensions. To generate this visual, I took each material and "embedded" it by running it throug
In this post, I'll share a continuation of that work where we characterize materials by their distance to the known-to-be-good permanent magnet Nd2Fe14B (mp-616958).
The idea is that encoded in the structure would be the characteristics that create strong coercivity and anisotropy. This is just a hypothesis, and weak one at that, but maybe there is signal in there somewhere.
This dataset has a set of 34,000 ferro/ferrimagnetic materials from Materials Project, their formula, if they include rare earth elements, magnetic moment, volume, magnetic density, a predicted Curie temperature, and cosine distances to some known permanent magnets like NdFeB. Distances are based on a 256 dimension embedding from Orb v2 latent space.
With this dataset, we should be able to get a good sense of what might be a decent candidate worth further exploration.
Here are some that I've found so far:
Magnetic density: 0.142
Curie: 596 K
Distance to NdFeB: 0.151
CIF file for ZrFe12Si2B, a ferrimagnetic materials from Materials Project. https://next-gen.materialsproject.org/materials/mp-653838
Magnetic density: 0.155
Curie: 714 K
Distance to NdFeB: 0.166
Magnetic density: 0.092
Curie: 403 K
Distance to NdFeB: 0.204
Magnetic density: 0.112
Curie: 562 K
Distance to NdFeB: 0.196
, a ferrimagnetic material from Materials Project. https://next-gen.materialsproject.org/materials/mp-21666
Magnetic density: 0.175
Curie: 827 K
Distance to NdFeB: 0.227
The use of Orb v2 could be something to tweak in this process. Instead, maybe try to embed materials with CHGNet. The difference would be that we're getting similar materials from a physical and mechanical perspective and less of a magnetic perspective. This would be because of Orb missing magnetic understanding in it's interatomic potential representation.
Still running this experiment on 30,000 other materials. Takes a bit of time when there's that many materials.
Discover other posts like this one
Sharing some notes as I go through this paper:
Working on cleaning the data we have available and seeing what we've got for a MAE prediction model. This resource was nice and had all the raw files uploaded so that you can process them yourself and
Also known as the Magnetic Materials Database. I came to this database looking for magnetocrystalline anisotropy energy data for permanent magnet design. After scraping the data from the app, which is