Generating information-rich high-throughput experimental materials genomes using functional clustering via multitree...

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• High-throughput screening accelerates materials discovery by rapidly synthesizing and characterizing large combinatorial libraries.
• Tiered screening strategies reduce the number of compositions tested at higher-complexity, lower-throughput stages.
• Effective down-selection algorithms are crucial for maximizing information gain from materials genomes.

Purpose of the Study:

• To develop advanced down-selection algorithms that prioritize information value over mere performance for enhanced materials discovery.
• To identify distinct 'property fields' within high-throughput data to guide intelligent sample selection.
• To introduce an informatics-based approach for clustering composition-property relationships.

Main Methods:

• Utilized information theory and multitree genetic programming for clustering composition-property functional relationships.
• Developed an algorithm to identify property fields within material composition libraries.
• Applied the methodology to a synthetic ternary library and a (Ni-Fe-Co-Ce)Ox catalyst library.

Main Results:

• Successfully identified four distinct property fields in a synthetic ternary composition-property map.
• Demonstrated the algorithm's capability to capture composition-catalytic activity relationships for the oxygen evolution reaction in a 5429-composition library.
• Validated the informatics-based clustering approach for property field identification.

Conclusions:

• The developed informatics-based clustering method effectively identifies property fields in high-throughput materials data.
• This approach enables advanced down-selection strategies, improving the efficiency of data-driven materials discovery.
• The methodology shows promise for accelerating the discovery of novel catalysts and materials with desired properties.