Beyond Optimization: Exploring Novelty Discovery in Autonomous Experiments.

Area of Science:

• Materials Science
• Physics
• Chemistry

Background:

• Autonomous experiments (AEs) integrate AI with automated platforms for scientific research.
• Current AEs often focus on optimizing predefined targets, potentially limiting the discovery of unexpected phenomena.

Purpose of the Study:

• To introduce a novel framework, INS²ANE (Integrated Novelty Score-Strategic Autonomous Non-Smooth Exploration), for enhancing the discovery of novel phenomena in autonomous microscopy.
• To move beyond conventional optimization-focused AEs and promote broader exploration of experimental spaces.

Main Methods:

• INS²ANE integrates a novelty scoring system to evaluate experimental result uniqueness.
• It employs a strategic sampling mechanism to explore under-sampled regions, even if initially less promising.
• The framework was validated on image-spectral data and implemented in autonomous scanning probe microscopy.

Main Results:

• INS²ANE significantly increases the diversity of explored phenomena compared to conventional optimization routines.
• The method enhances the likelihood of discovering previously unobserved phenomena.
• Demonstrated effectiveness in autonomous scanning probe microscopy experiments.

Conclusions:

• INS²ANE advances autonomous microscopy by enabling exploration of complex experimental spaces.
• The framework holds significant potential for accelerating scientific discovery through the uncovering of novel phenomena.
• This approach broadens the scope of autonomous experimentation beyond mere optimization.