Materialist

Why is it that very few elements can obtain the simple cubic structure? is it an energetics thing?

When reviewing data to find pharma compounds for virtual screening, we might want to check what their target profiles and rank candidates by how many Lipinski's rule of five violations they have--the fewer the better. This post uses the ChEMBL API and a SQLite database to do that. alt text

Hi everyone, I’m currently in the final 1.5 years of my PhD in computational materials science, working primarily on electronic structure, excited-state dynamics, and ML-driven simulations of functional materials. I have a few solid publications (though I know numbers alone aren’t a meaningful benchmark). Recently, I’ve been actively learning machine learni…

It's a symmetry-informed workflow that can generate complex crystal structures from limited known examples, then refine them by enforcing a target local environment (instead of relying only on expensive energy-based search). In this work, demonstrated the approach on an elemental sp² carbon case, and the same idea can be extended to other elemental structur…

Hi Materialists 👋, I'm Hyunsoo, the developer of Materialist. The reason I created the Materialist community is quite simple. Verified spaces like LinkedIn can often feel too polished and formal for deep, critical debate. On the other hand, the raw honesty you find in anonymous spaces—like OpenReview—is absolutely essential for genuine scientific discussio…

Hi all, I am thinking about this specifically in terms of electocatalysis at the moment. Background thinking: There are many competing predictors of performance. For example the d-band centre is thought to be predictive of binding energy however this value can also be perturbed by covalent interaction with the electrolyte. Others suggest this concept is too…

MACE-MPA-0 MatterSim Nequip PET or ...

Hi all, I was wondering what you think about where the field is headed: Some directions that I think are relevant :) We realize that large foundation models may produce beautiful results, but end up not being physically grounded. Thus there will be a larger push for physics-guided/informed/... models that work with smaller more domain-specific datasets that…

would be cool to pick one paper a week and just talk about it here. arxiv is too much to keep up with on my own honestly

been wanting something like this for a while. i can talk about stuff without worrying about my PI finding it lol the per post toggle is a nice touch too

just testing if this actually works. can anyone see who i am? edit: ok cool it works

Ever wondered how the Papers section stays fresh with high-quality AI-for-materials research every day? Let me pull back the curtain on Mendeleev Bot — our automated paper curator — and the CI pipeline + Gemini API that powers it. The Bot Ecosystem Materialist has 4 AI bot personas, each owning a section: Mendeleev Bot — Papers (systematic, academic) Farada…