Text-mined dataset of solid-state syntheses with impurity phases using Large Language Model.

Area of Science:

• Materials Science
• Chemistry
• Data Science

Background:

• Solid-state synthesis is crucial for inorganic materials like battery components and thermoelectrics.
• Current synthesis methods lack a general theory and clear reaction mechanisms, posing challenges.
• Existing literature datasets often overlook product phase purity and yield.

Purpose of the Study:

• To construct a comprehensive dataset of solid-state synthesis reactions.
• To analyze impurity phase formation during synthesis.
• To identify factors influencing product purity and yield.

Main Methods:

• Utilized a large language model (LLM) to extract data from scientific literature.
• Compiled a dataset of 80,806 solid-state syntheses.
• Included 18,869 reactions with identified impurity phases.

Main Results:

• The dataset validates thermodynamic predictions for impurity formation.
• Identified instances where impurities form despite the target phase's higher thermodynamic stability.
• Provides a valuable resource for understanding and optimizing solid-state synthesis.

Conclusions:

• The LLM-extracted dataset offers novel insights into solid-state reaction mechanisms.
• Highlights the complexity of impurity formation beyond simple thermodynamic stability.
• Facilitates the development of more reliable and efficient synthesis strategies for inorganic materials.