MOFDiff: Coarse-grained Diffusion for Metal-Organic Framework Design

Explore a comprehensive lecture on MOFDiff, a coarse-grained diffusion model for designing metal-organic frameworks (MOFs). Delve into the innovative approach of using diffusion processes to generate MOF structures, with a focus on applications in gas storage and carbon capture. Learn about the challenges in traditional template-based methods and how MOFDiff addresses them through a denoising diffusion process. Discover the use of equivariant graph neural networks to respect symmetries in MOF generation. Examine the model's effectiveness in creating valid and novel MOF structures, particularly for carbon capture applications. Follow the journey from coarse-grained to all-atom MOF structures through a novel assembly algorithm. Gain insights into contrastive representation learning, sample MDF structures, and future directions in this field. Conclude with a Q&A session to deepen your understanding of this cutting-edge research in AI-driven material design.

- Intro + Background
- Results
- Coarse-Grained Diffusion
- Contrastive Representation Learning
- From CG to All-Atom MOFs
- Sample MDF Structures
- Future Directions
- Q+A