GlaDys – Glass dynamics at ab-initio accuracy via gen-AI

Research Project, 2026 – 2030

Glassy materials uniquely combine structural disorder with rigidity, leading to distinct mechanical, optical, and chemical properties. Organic glasses exhibit tunable optical characteristics, flexibility, and reduced weight, vital in optoelectronics and pharmaceuticals. However, accurately simulating their dynamics, especially the significant slowdown associated with the glass transition, remains challenging due to computational limitations.This GlaDys project aims to overcome these barriers by: (1) substantially extending simulation time scales to capture slow dynamic processes, and (2) investigating complex, multi-component molecular glass formers. Utilizing generative AI models and machine-learned interatomic potentials (MLIPs), we aim to achieve near-electronic-structure calculation accuracy at relevant time scales (10-100 µs). Structured into three work packages, we will develop AI-driven computational frameworks initially for inorganic glasses, then extend them to molecular systems, and finally apply them to perylene derivatives. The outcomes include a robust, generative AI-based simulation framework and novel insights into multi-component glass behaviors, supporting experimental interpretation and the informed design of advanced amorphous materials.