Exploration of Chemical Space Through Automated Reasoning

Area of Science:

• Materials Chemistry
• Computational Materials Science
• Artificial Intelligence in Materials Discovery

Background:

• The immense scale of possible material compositions presents a major hurdle in discovering new materials with desired properties.
• Traditional methods struggle to systematically evaluate novelty, stability, cost, and performance across vast chemical spaces.

Purpose of the Study:

• To introduce Comgen, a novel computational tool for efficient exploration of material composition space.
• To apply formal reasoning methods from computer science to materials discovery for the first time.
• To enable reasoning over expert-defined scientific domain knowledge for materials design.

Main Methods:

• Comgen utilizes logical methods and automated reasoning algorithms, adapted from theorem proving in computer science.
• User-defined criteria are converted into an abstract form for logical processing.
• The tool identifies compositions satisfying criteria or proves unsatisfiability, offering interpretable and provably correct outputs.

Main Results:

• Demonstrated Comgen's capability to identify suitable material compositions based on specified criteria.
• Successfully applied Comgen to case studies searching for new ionic conductors.
• Integrated Comgen into an automated workflow for proposing and quantitatively evaluating candidate materials.

Conclusions:

• Automated formal reasoning offers a powerful, interpretable, and provably correct approach to materials discovery.
• Comgen facilitates efficient exploration of composition space, accelerating the identification of novel materials.
• The tool has the potential to significantly advance materials chemistry by integrating domain knowledge and automated reasoning.