Harnessing Large Language Models to Collect and Analyze Metal-Organic Framework Property Data Set

Yeonghun Kang ¹, Wonseok Lee ¹, Taeun Bae ¹

⁰Department of Chemical and Biomolecular Engineering, Korea Advanced Institute of Science and Technology, Daejeon 34141, Republic of Korea.

Journal of the American Chemical Society +

|

January 21, 2025

View abstract on PubMed

Summary

Researchers created a large dataset of experimental metal-organic framework (MOF) data using AI. This enables better machine learning studies and a new system to recommend MOF synthesis conditions.

Area Of Science

Materials Science
Computational Chemistry
Data Science

Background

Accessing experimental metal-organic framework (MOF) data is challenging.
Existing data is often unstructured, hindering machine learning applications.
High-quality data is crucial for advancing materials science research.

Purpose Of The Study

To systematically collect and structure experimental MOF data from scientific literature.
To improve data availability and quality for machine learning in materials science.
To develop tools for analyzing MOF synthesis and properties.

Main Methods

Employed advanced large language models (LLMs) for data extraction.
Developed a systematic approach to compile MOF synthesis conditions and properties.
Created a structured database from over 40,000 research articles.

Main Results

Compiled a comprehensive, ready-to-use experimental MOF dataset.
Analyzed relationships between MOF synthesis, properties, and structure.
Identified a gap between simulation and experimental data, pinpointing contributing factors.
Developed a synthesis condition recommender system.

Conclusions

Experimental datasets are vital for advancing MOF research.
The developed system provides a practical tool for optimizing MOF synthesis.
AI-driven data curation enhances the utility of scientific literature for materials discovery.