ml-generative-mattergen

What problem does it solve?

This Skill solves the bottleneck of manually creating candidate inorganic crystal structures by automating structure generation with a diffusion-based generative model.

Core Features & Use Cases

• Unconditional structure generation: Produce novel inorganic structures for exploration and downstream evaluation.
• Chemical system conditioning: Encourage generated structures to include a specified set of elements for targeted discovery.
• Property fine-tuning: Fine-tune MatterGen on custom structure/property datasets to generate structures conditioned on a property (e.g., formation energy).
• Workflow integration: Use generated CIFs as inputs to relaxation, stability analysis (e.g., E_hull), and property screening.

Quick Start

Ask the AI to generate 20 chemically conditioned structures for Li-Fe-P-O and save them to a local output directory by calling the MatterGen MCP generation function with chemical_system set to Li-Fe-P-O, guidance_scale to 1.0, num_structures to 20, and batch_size to 10.