🔋 Automatic Generation of Inorganic Solid‐State Electrolytes via Unified Multi‐Modal Network

📝 شبکه عصبی چندوجهی USMNet با در نظر گرفتن ساختار، الکترولیت‌های جامد لیتیومی را پیش‌بینی و Li16ZnSiP2S16 را به‌صورت تجربی تأیید می‌کند.

📚 Journal: Small #Small
📅 Date: 2026-10-08
🔬 Research Article

💡 Novelty & Significance:
USMNet unifies crystal-structure, compositional, and physical descriptors to predict Li-ion conductivity, overcoming sparse data and structure dependence. It outperforms composition-only and structure-based baselines, enabling data-efficient screening of 168,675 candidates and identifying 21 electrolytes. Experimental validation confirms Li₁₆ZnSiP₂S₁₆ delivers 80% capacity retention after 600 cycles, demonstrating a generalizable, closed-loop strategy for solid-state electrolyte discovery.

🔗 10.1002/smll.75978

🏷 solid electrolyte · machine learning · ionic conductivity · materials discovery · all-solid-state battery