Orbitol generates crystals constrained to requested space groups
Original title: Orbitol: Crystal Generation on Wyckoff Orbits
On MP-20, 13.8% of attempts stayed stable or metastable, unique, novel, and in the requested group after relaxation.
Orbitol is a crystal generator on the Crystalite diffusion Transformer that uses Wyckoff-orbit tokens instead of atom tokens. Given a space group and occupied Wyckoff positions, it generates atom types, free coordinates, and a compatible lattice so valid structures stay invariant under that group. It was two to three orders of magnitude faster than symmetry-aware baselines at the evaluated settings and was used to propose piezoelectric candidates.
- MP-20 success was 13.8% for Orbitol versus 10.8% for SGFM.
- The paper introduces SPARC for space-group retention and distribution agreement.
- Of screened piezoelectric candidates, 89% passed DFT checks and 79% matched no same-group Materials Project or Alexandria entry.
- Their median figure of merit was nine times Materials Project piezoelectrics, and 24% ranked in that set's top 10%.
- Products
- Crystalite
- AI models
- Orbitol
From the original article
1 min read at arxiv.orgGenerative models propose crystalline materials by learning composition and geometry, but unconditional sampling gives no direct control over the space group, which constrains whether responses such as piezoelectricity are symmetry-allowed. We present Orbitol, a crystal generator built on the Crystalite diffusion Transformer that replaces atom tokens with Wyckoff-orbit tokens.
By Ivan Žugec · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.