Host: The Japan Society of Vacuum and Surface Science
In this study, we introduce an AI-robot system utilizing Bayesian optimization to expand the search space for solid-state inorganic compounds. This system fully automates sample transfer, thin film deposition, and growth condition optimization-all addressing the physical aspects of fabrication. The figure shows the design of our apparatus, which is equipped with a central robot arm that can access each satellite chamber for automatic thin-film growth and physical-property measurements. Using the values obtained from the physical-property measurements, the Bayesian optimization algorithm predicts the next growth condition to obtain better values. An example of Bayesian optimization is shown for the epitaxial growth of TiH2 thin films with two parameters. Through the automated thin-film synthesis and optimization, we aim to establish new materials research style to drastically accelerate solid-state inorganic materials research.
