System of Agentic AI for the Discovery of Metal-Organic Frameworks
🤖 AI Summary
Metal–organic frameworks (MOFs) face a fundamental challenge in CO₂ capture and atmospheric water harvesting due to the vast chemical space and the difficulty of simultaneously achieving structural novelty and synthetic feasibility. Method: To address this, we introduce MOFGen—a multi-agent AI system pioneering the “Agentic AI” paradigm for materials discovery. MOFGen integrates a large language model (for ligand proposal), a 3D crystal structure diffusion model (for framework generation), DFT-based quantum-mechanical geometry optimization (for structural refinement), and a hybrid rule- and machine learning–based synthesizability evaluator (for screening). Trained on all experimentally reported MOFs and comprehensive computational databases, it enables an end-to-end “AI conception → experimental validation” pipeline. Results: MOFGen generated hundreds of thousands of novel MOFs and organic linkers; high-throughput experimentation successfully synthesized five AI-designed MOFs, all confirmed by single-crystal X-ray diffraction and exhibiting target adsorption performance—marking the first fully validated, end-to-end AI-driven discovery in the MOF domain.
Application Category
📝 Abstract
Generative models and machine learning promise accelerated material discovery in MOFs for CO2 capture and water harvesting but face significant challenges navigating vast chemical spaces while ensuring synthetizability. Here, we present MOFGen, a system of Agentic AI comprising interconnected agents: a large language model that proposes novel MOF compositions, a diffusion model that generates crystal structures, quantum mechanical agents that optimize and filter candidates, and synthetic-feasibility agents guided by expert rules and machine learning. Trained on all experimentally reported MOFs and computational databases, MOFGen generated hundreds of thousands of novel MOF structures and synthesizable organic linkers. Our methodology was validated through high-throughput experiments and the successful synthesis of five"AI-dreamt"MOFs, representing a major step toward automated synthesizable material discovery.
Problem
Research questions and friction points this paper is trying to address.
Navigating vast chemical spaces for MOF discovery
Ensuring synthesizability of AI-generated MOF candidates
Accelerating material discovery for CO2 capture and water harvesting
Innovation
Methods, ideas, or system contributions that make the work stand out.
Agentic AI system with interconnected specialized agents
Generative models propose and optimize MOF candidates
Validated by synthesis of AI-generated MOF materials
🔎 Similar Papers
No similar papers found.
