O-kRystal: dual-space orbital learning predicts HSE06 band structures at PBE cost
AnimaAnandkumar · x · 2026-10-18
Anima Anandkumar's team released an arXiv paper (2610.19919) introducing O-kRystal, a dual-space orbital-learning framework that uses a cheap PBE calculation as both energy baseline and structured orbital information source to predict beyond-DFT band corrections.
- Approach: combines real- and reciprocal-space orbital matrices to predict HSE06-level corrections per electronic state as a function of crystal momentum and spin.
- Accuracy: MAE of 0.171±0.009 eV for HSE06 band energies within 5 eV of the band-edge midpoint, across nearly 6,000 3D materials (insulating, metallic, magnetic, non-magnetic).
- Strongly correlated systems: on a benchmark of six materials, reduces PBE band-energy MAE by 68.7% and resolves orbital-dependent changes in transition-metal d vs ligand p state placement.
- Generalization: trained only at high-symmetry points yet maintains accuracy at unseen band-path points, reconstructs HSE06 densities of states, and transfers to G0W0 quasiparticle energies for 2D materials.
A notable AI-for-Science advance for faster quantum-material band predictions.
Related event: Caltech's O-kRystal Learns Band Structure Corrections Beyond DFT Accuracy(2 posts)→
More from Research
- PyNNDescent 0.7 benchmark: ~6x faster kNN graph builds on 64-core Xeon — leland_mcinnes · 2026-10-18
- Terence Tao: OpenAI's quasi-Riemann result is the biggest advance in 50 years — jracine22 · 2026-10-18
- What is wrong with AI writing - ICLR reviewing edition — bobrodsky · 2026-10-18
- Meta, MIT and Yale propose ReHarness: backward revision for agentic continual learning — HanGuo97 · 2026-10-18
- Fields Medalist Kontsevich: AI Agent Suggested a Better Math Definition — luislamb · 2026-10-18
- NYT Cites MIT–OpenAI Study Linking Extended Chatbot Use to Poorer Well-Being — patpat_mit · 2026-10-18