AutoDFT: A Closed-Loop Multi-Agent Framework for Autonomous DFT Calculations
Researchers have developed AutoDFT, a novel closed-loop multi-agent framework designed to automate Density Functional Theory (DFT) calculations in materials science. Unlike previous LLM-based agents that only plan upfront, AutoDFT integrates LLM reasoning throughout the entire DFT process, from initial planning to real-time parameter generation and adaptive recovery from failures. This system demonstrates high success rates on a new benchmark and produces reliable property predictions, enabling users without deep computational expertise to perform first-principles calculations. AI
IMPACT Automates complex scientific calculations, potentially accelerating materials discovery and reducing reliance on expert computational chemists.