Embodied Intelligence Platform for Materials Synthesis

Artificial intelligence (AI) experimental systems demonstrate marked improvements in material research efficiency. However, challenges in generalization capabilities and experimental operations remain to be addressed. Here, a flexible Embodied Intelligence (EI) platform comprising a home-made EI robot and modular experimental components is designed and demonstrated to overcome these hurdles. The remotely trained EI robot generated action libraries enable consistent, human-like manipulations via encoded trajectories. By integrating large language models-assisted analysis, the synthesis of CsPbI3 quantum dots (QDs) is systematically studied using the EI robot executing key steps. Precise execution of manually uncontrollable manipulations via EI robot directly modulate growth kinetics, demonstrating effects of significance equivalent to conventional reaction parameters. The EI platform thus produces high-quality CsPbI3 QDs with an ultra-narrow full-width at half-maximum of 29.6 nm and a high photoluminescence quantum yield of 76.7%. Two distinct experimental series involving perovskite QDs diversification and extreme-condition synthesis are further conducted to confirm the adaptability and generalization capacity of the EI platform for diverse material systems. This framework facilitates the development of versatile AI systems and enhances the materials research efficiency.