LIMG · Artificial Intelligence-Driven High-Fidelity Inverse Design of Solid-State Electrolytes

Линиите свързват координатора с партньорите.

Накратко на български

Цел на проекта

Solid-state electrolytes (SSEs) promise to revolutionize energy storage and carbon-neutral mobility industries, but their low ionic conductivity remains a barrier. Discovery of high-performance SSEs is hindered by a lack of high-quality lithium-diffusion data, unclear structure-property understanding, and the high cost of traditional computational screening. This project aims to build a Loop-Locked Intelligent Material Generation (LIMG) platform by coupling a pre-trained machine learning force field (MLFF) called SO3LR-SSE with advanced generative models. SO3LR-SSE explicitly incorporates non-local and many-body interactions to improve the fidelity of dynamic SSE databases and the transferability across diverse material systems. The project includes four main work packages (WPs 1-4, LIMG platform) plus one for data management, career development, and dissemination. WP1 will develop a high-fidelity, transferable MLFF for SSEs to speed up Li-diffusion dynamics simulations, forming the basis for generating a comprehensive SSE database of ionic conductivities in WP2. WP3 will use ML and the database to accurately predict ionic conductivity and elucidate structure-property relationships. WP4 will focus on inverse SSEs design with clear physical interpretability based on generative models. LIMG is expected to generate a large, reusable dataset and rapidly identify novel candidates across global chemical spaces for experimental follow-up. A planned secondment at TU Berlin will support MLFF development in a collaborative environment. For the fellow, this project will significantly strengthen skills in advanced AI algorithm development, materials design, and interdisciplinary collaboration, strengthening prospects for leadership in AI-driven materials research. LIMG aspires to be widely expanded to material systems, advancing innovative, interpretable, high-fidelity design of new energy materials and accelerating deployment of low-carbon energy storage technologies.

Оригинален текст от CORDIS (на английски).

Участници

• UNIVERSITE DU LUXEMBOURG · ESCH-SUR-ALZETTEКоординаторЛюксембург
• TECHNISCHE UNIVERSITAT BERLIN · BerlinГермания

Връзки

• Виж в CORDIS
• DOI: 10.3030/101265665