Pushing property limits in materials discovery via boundless objective-free exploration.

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Informatics

Background:

• Materials discovery often involves optimizing compounds for conflicting industrial demands.
• Traditional black-box optimization may fail when target properties are ill-defined.
• Automated methods are needed to navigate complex chemical spaces efficiently.

Purpose of the Study:

• To develop a novel algorithm for automated materials discovery.
• To address limitations of existing optimization methods in materials science.
• To enable open-ended exploration for novel chemical compounds.

Main Methods:

• Proposed BoundLess Objective-free eXploration (BLOX) algorithm.
• Utilized kernel-based Stein discrepancy in the property space.
• Applied BLOX to identify light-absorbing molecules from a drug database, minimizing density functional theory calculations.

Main Results:

• Successfully identified out-of-trend compounds in the intensity-wavelength property space.
• Experimentally verified optical properties of eight identified compounds using absorption spectroscopy.
• Demonstrated the effectiveness of BLOX in reducing computational cost for materials discovery.

Conclusions:

• BLOX is a powerful, objective-free exploration method suitable for open-ended scientific discovery.
• The algorithm effectively facilitates chemical repurposing and identifies novel materials.
• BLOX is expected to find broad applications across various scientific disciplines.