Dandelion Docs
Welcome to the documentation site of Dandelion.

Introduction
Dandelion is a code designed to build comprehensive chemical databases by efficiently sampling reactive chemical space. Our approach combines fast tight-binding calculations with selective high-level refinement to generate high-quality datasets for Machine Learning Interatomic Potentials (MLIPs). The workflow consists of four phases:
- Reactant Preparation: Generate diverse 3D structures from GDB-13 database
- Product Search: Find reaction pathways using single-ended growing string method
- Landscape Search: Sample transition states using nudged elastic band method
- Refinement: Perform high-level DFT calculations using ORCA
This documentation guides the installation, setup, and usage of Dandelion.