The Block Copolymer Phase Behavior Database

Area of Science:

• Polymer Science
• Materials Informatics
• Computational Chemistry

Background:

• Lack of a standardized data model for block copolymer self-assembly data.
• Need for accessible experimental and simulation data for polymer research.
• Challenges in curating and integrating diverse polymer datasets.

Purpose of the Study:

• To introduce the Block Copolymer Database (BCDB) as a comprehensive resource for block copolymer data.
• To establish a novel data schema for block copolymer self-assembly information.
• To facilitate data sharing, analysis, and machine learning applications in polymer science.

Main Methods:

• Development of a new data schema accommodating various block copolymer structures.
• Data curation from literature and integration of self-consistent field theory simulation data.
• Implementation of BigSMILES for chemical structure encoding and SMARTS for advanced searching.
• Utilizing SQL for querying characterization and phase information.
• Employing GPT-4 for automated literature screening and data identification.

Main Results:

• BCDB contains over 5400 experimental melt phase measurements and simulation data points.
• The database supports searching by repeat units, functional groups, and phase information.
• A protocol using GPT-4 achieved an F1 score of 0.74 in identifying relevant literature.
• Data can be downloaded for machine learning model training.

Conclusions:

• BCDB provides a crucial, accessible platform for block copolymer data, bridging experimental and simulation studies.
• The developed data schema and search functionalities enhance data discoverability and usability.
• AI-driven literature screening accelerates database expansion and data acquisition.