Superionic Ionic Conductor Discovery via Multiscale Topological Learning DOI Creative Commons
Dong Chen, Bingxu Wang, Shunning Li

и другие.

Journal of the American Chemical Society, Год журнала: 2025, Номер unknown

Опубликована: Июнь 5, 2025

Lithium superionic conductors (LSICs) are crucial for next-generation solid-state batteries, offering exceptional ionic conductivity and enhanced safety renewable energy electric vehicles. However, their discovery is extremely challenging due to the vast chemical space, limited labeled data, understanding of complex structure-function relationships required optimizing ion transport. This study introduces a multiscale topological learning (MTL) framework that integrates algebraic topology unsupervised efficiently tackle these challenges. By modeling lithium-only lithium-free substructures, extracts features two screening metrics, cycle density minimum connectivity distance, ensure structural diffusion compatibility. Promising candidates clustered via algorithms identify those resemble known conductors. For final refinement, pass undergo ab initio molecular dynamics simulations validation. approach led 14 novel LSICs, four which have been independently validated in recent experiments. success accelerates identification LSICs demonstrates broad adaptability, scalable tool addressing material

Язык: Английский

Superionic Ionic Conductor Discovery via Multiscale Topological Learning DOI Creative Commons
Dong Chen, Bingxu Wang, Shunning Li

и другие.

Journal of the American Chemical Society, Год журнала: 2025, Номер unknown

Опубликована: Июнь 5, 2025

Lithium superionic conductors (LSICs) are crucial for next-generation solid-state batteries, offering exceptional ionic conductivity and enhanced safety renewable energy electric vehicles. However, their discovery is extremely challenging due to the vast chemical space, limited labeled data, understanding of complex structure-function relationships required optimizing ion transport. This study introduces a multiscale topological learning (MTL) framework that integrates algebraic topology unsupervised efficiently tackle these challenges. By modeling lithium-only lithium-free substructures, extracts features two screening metrics, cycle density minimum connectivity distance, ensure structural diffusion compatibility. Promising candidates clustered via algorithms identify those resemble known conductors. For final refinement, pass undergo ab initio molecular dynamics simulations validation. approach led 14 novel LSICs, four which have been independently validated in recent experiments. success accelerates identification LSICs demonstrates broad adaptability, scalable tool addressing material

Язык: Английский

Процитировано

0