Dandelion Docs

Welcome to the documentation site of Dandelion.

Dandelion

Introduction

overall_scheme

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:

  1. Reactant Preparation: Generate diverse 3D structures from GDB-13 database
  2. Product Search: Find reaction pathways using single-ended growing string method
  3. Landscape Search: Sample transition states using nudged elastic band method
  4. Refinement: Perform high-level DFT calculations using ORCA

This documentation guides the installation, setup, and usage of Dandelion.

Git repository

https://github.com/mhyeok1/dand/